The role of PD‑1/PD‑L1 axis in liver diseases.

Introduction
The liver is characterized by a unique immune-tolerant microenvironment, a feature that is determined by its core physiological functions, including receipt of portal venous blood enriched in dietary antigens and microbial products and participation in systemic metabolism and detoxification. This immune-tolerant milieu is formed by the specialized architecture of hepatic sinusoids and by distinct resident cell populations, notably liver sinusoidal endothelial cells (LSECs), Kupffer cells (KCs) and tolerogenic dendritic cells (tolDCs). Immune-checkpoint molecules, exemplified by programmed cell death protein 1 (PD-1) and its ligand programmed death-ligand 1 (PD-L1), together with immunosuppressive cytokines such as interleukin-10 (IL-10), predominate within this environment.Under physiological conditions, the intrahepatic microenvironment is configured to finely regulate cellular immunity by inducing T cell dysfunction, deletion and exhaustion, and by promoting expansion of regulatory T cells (Tregs). Humoral immunity is modulated through induction of immunoglobulin A (IgA) production and clearance of immune complexes. High expression of PD-L1 by LSECs and expression of PD-1 on circulating CD4-positive T lymphocytes and CD8-positive T lymphocytes establish persistent, repetitive interactions that actively program CD8-positive T cells into a non-responsive state [1]; cytokine secretion is suppressed and CD8-positive T cells are rendered functionally inactive yet long-lived, thereby maintaining a stable tolerant state [2, 3], while differentiation toward short-lived effector cells (SLECs) or memory precursor effector cells (MPECs) is promoted [4].In the settings of liver transplantation and ischemia–reperfusion injury (IRI), high PD-L1 expression by KCs has been shown to exert protective effects [5, 6], and these mechanisms collectively contribute to the preservation of intrinsic hepatic immune tolerance. Nevertheless, PD-1/PD-L1 axis-mediated tolerance is disrupted in various chronic liver diseases. Persistent inflammation driven by hepatotropic viral infection, metabolic dysregulation, autoimmunity or bacterial infection culminates in hepatic fibrosis and the development of hepatic malignancies [7].
PD-1/PD-L1 signaling axis is a central component of immune-checkpoint pathways and is implicated in the maintenance of immune homeostasis and the prevention of autoimmunity [8]. PD-1, an inhibitory receptor, is predominantly expressed on the surface of activated T lymphocytes, B lymphocytes, natural killer (NK) cells and myeloid-lineage cells. Two ligands of PD-1 are recognized: PD-L1 and PD-L2. PD-L2 expression is principally induced by T helper 2 (Th2)-associated cytokines and is observed on antigen-presenting cells (APCs), including macrophages (Mφ), dendritic cells (DCs), mast cells (MCs) and subsets of B cells that respond to interleukin-4 (IL-4) and interferons (IFNs). PD-L1 exhibits a broader expression pattern and is present in both constitutive and inducible forms. Under physiological conditions, constitutive PD-L1 expression is detected on non-hematopoietic cells such as liver sinusoidal endothelial cells (LSECs) and hepatocytes (Hep), where local immune-privileged niches are established to protect tissues from excessive immune-mediated injury. Under pathological conditions, PD-L1 is induced, notably within the tumor microenvironment (TME) and chronic inflammatory milieus. Tumor cells frequently display high PD-L1 expression, and inflammatory cytokines such as interferon-gamma (IFN-γ) strongly induce PD-L1 on APCs and macrophages; hepatic stellate cells (HSCs) have been reported to exhibit inducible PD-L1 expression. Kupffer cell (KC) expression of PD-L1 is subject to both constitutive and inducible regulation.Engagement of PD-1 by its ligands initiates intracellular signaling cascades. Tyrosine residues within the cytoplasmic immunoreceptor tyrosine-based inhibitory motifs of PD-1 are phosphorylated, leading to recruitment of Src homology region 2 domain-containing phosphatase-1 and phosphatase-2 (SHP-1 and SHP-2). These phosphatases dephosphorylate key downstream components of the T cell receptor (TCR) signaling pathway, thereby inhibiting activation of pro-survival and proliferative pathways such as PI3K/AKT and Ras/mitogen-activated protein kinase (Ras/MAPK). The net effect is suppression of T cell proliferation, reduction in cytokine production, impairment of cytotoxic function and, in some instances, induction of apoptosis in T cells [9].Within the TME, PD-1/PD-L1 interactions are recognized as a principal mechanism underlying tumor-specific T cell dysfunction. Monoclonal antibodies targeting this pathway, termed immune checkpoint inhibitors (ICIs), have been developed to disrupt the inhibitory signal. PD-1 inhibitors, exemplified by pembrolizumab and nivolumab, bind with high affinity to PD-1 on T cells, whereas PD-L1 inhibitors, exemplified by atezolizumab and durvalumab, target PD-L1 on tumor cells or APCs. Blockade of the PD-1/PD-L1 interaction is intended to relieve T cell inhibition, restore proliferative capacity, cytotoxic activity and cytokine secretion of exhausted T cells, and thereby promote infiltration of activated T cells and clearance of tumor cells. With the expanding clinical use of ICIs in oncology, emerging issues such as ICI resistance and immune-related hepatic injury have attracted attention, and exploration of ICI applications in non-malignant diseases is ongoing.
The PD-1/PD-L1 signaling axis is recognized as a precisely regulated, multifunctional immune-regulatory system. Recent studies have demonstrated that PD-L1 is not confined to the plasma membrane but is distributed across distinct subcellular compartments, including a secreted, body-fluid-dispersed soluble form (sPD-L1); a nuclear fraction that interacts with transcriptional regulators (nPD-L1); and a mitochondrial pool that influences cellular energy metabolism (mtPD-L1). In addition, the PD-1/PD-L1 axis is subject to multiple layers of fine regulation, encompassing signals derived from the TME, intracellular signaling cascades, multilayered control of gene expression and modulatory inputs from the tumor stroma. Given the tight linkage between the PD-1/PD-L1 axis, immune regulation and chronic liver diseases, recent advances addressing this axis across diverse chronic hepatic disorders are reviewed herein, and the precise regulatory mechanisms operative in hepatic malignancies are synthesized with the objective of informing mechanistic understanding and the development of PD-1/PD-L1-based immunotherapeutic strategies.

Regulatory mechanisms of the PD‑1/PD‑L1 signaling axis in HCC
The most common primary hepatic malignancies are hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA). Among chronic liver diseases, the PD‑1/PD‑L1 signaling axis has been most extensively investigated in HCC, and PD‑1/PD‑L1‑targeted therapies have been implemented clinically in HCC [10].PD‑1/PD‑L1 expression in HCC and CCA will be examined. The complex, multilayered regulation of the PD‑1/PD‑L1 axis across diverse cell types will be delineated, encompassing epigenetic, transcriptional, post‑transcriptional, translational and post‑translational mechanisms. Modulation by the TME and by intracellular signaling pathways will be described, and contributions from stromal compartments will be summarized. In addition, challenges and recent insights regarding immune checkpoint inhibitor (ICI) therapies targeting the PD-1/PD-L1 axis in HCC will be summarized. Elucidation of the PD‑1/PD‑L1 axis in hepatic malignancies is expected to inform investigations of this axis in other chronic liver diseases.

HCC

Expression of the PD‑1/PD‑L1 axis in HCC
PD‑L1 and PD‑L2 positivity rates were reported to be low in HCC tissue. Concordant changes in PD‑1 and PD‑L1 levels were observed in HCC, suggesting a potential regulatory relationship between PD‑1 and PD‑L1; PD‑1 positivity was proposed to reflect substantial infiltration of antitumor immune cells within the TME [11, 12].TAMs were shown to upregulate PD‑L1 and PD‑L2 in response to IFN‑γ released by TILs. This induction was characterized as an adaptive immune‑resistance mechanism that inhibited the function of CD8(+) TILs, promoted CD8 + T cell exhaustion, and was associated with poorer prognosis [13].Expression of PD‑L1 was reported to correlate significantly with tumor size, tumor recurrence, and levels of protein induced by vitamin K absence or antagonist‑II (PIVKA‑II), whereas PD‑L2 expression was associated with histologic differentiation and reduced overall survival. Patients whose tumors exhibited CD8 + TIL infiltration but lacked PD‑L1/PD‑L2 expression demonstrated the most favorable survival, while patients lacking TIL infiltration and exhibiting PD‑L1/PD‑L2 expression had the worst prognosis [11].
The PD‑1/PD‑L1 axis should not be considered solely as a single, uniform entity; its cellular origins and functional heterogeneity within the complex TME must be acknowledged. In the settings of viral hepatitis and liver cirrhosis, the diversity of immune cell subsets within the TME is further increased. In patients with HBV‑HCC, dozens of distinct antigen‑specific CD8 + T cell populations were documented, targeting tumor antigens, HBV antigens, neoantigens, and non‑disease‑related antigens. In addition to heterogenous, severely exhausted CD8 + T cells with high PD‑1 expression, a population of functionally preserved, metabolically active, and exhaustion‑resistant Trm defined by CD103(+)CXCR6(+)CD69(+) phenotype was described; these Trm cells exhibited relatively low PD‑1 expression and were associated with improved responses to PD‑1‑directed therapy and favorable patient survival [14]. HBV‑specific Trm subsets with low PD‑1 and low thymocyte selection‑associated high mobility group box protein (TOX) expression were shown to produce IFN‑γ and TNF without CD57 expression, consistent with maintenance of high functional status for HBV control; such HBV‑specific Trm populations were correlated with superior prognosis. HBV‑specific CD8 + Trm cells were observed to secrete IFN‑γ and chemokines that induced intrahepatic recruitment of memory‑phenotype bystander CD8 + T cells defined by CCR5(+)CXCR3(+)IL‑7Rα(+) markers; these bystander CD8 + T cells were proposed to potentially regain antitumor activity within HCC [15].
In tumors exhibiting durable responses to combined atezolizumab and bevacizumab therapy, a subset of PD‑L1(high)CXCL10(+)TAMs was found to secrete CXCL9, CXCL10, and CXCL11, thereby recruiting peripheral T cells that supplemented intratumoral CXCR3(+) effector memory T cells (Tem). Upon intratumoral activation, CXCR3 + Tem cells were preferentially differentiated into cytotoxic PD‑1(-) CD8 + terminally differentiated effector memory T cells re‑expressing CD45RA (Temra). CD8 + Temra cells, which exhibited natural killer‑like properties, were reported to release lytic granules and to mediate direct cytolytic interactions with target cells; these cells did not express classical exhaustion markers nor TME signatures associated with cytokine and IFN‑γ signaling. Antigenic stimulation was documented to induce re‑expression of CD45RA on CD8 + Temra cells, and substantial T‑cell receptor (TCR) sharing was observed between intratumoral CD8 + Temra and peripheral blood T cells. Evolution of intratumoral CD8 + Temra toward more differentiated phenotypes and an increase in T cells bearing shared tumor–peripheral TCRs were associated with improved outcomes; activation of PD‑1(-)CD45RA(+) CD8 + Temra cells was correlated with superior prognosis following anti‑PD‑L1 antibody therapy [16].In HCC responsive to PD‑1 blockade, a triad of interacting cell populations was reported to be enriched and to be essential for differentiation of primary CD8 + T cells into antitumor effectors after PD‑1 inhibition: mature regulatory dendritic cells (mregDC), a CXCL13(+)CH25H(+)IL‑21(+)PD‑1(+)CD4(+) helper T cell subset (CXCL13 + Th), and PD‑1‑high progenitor CD8 + T cells. mregDC expressing high levels of CD86 and CD80 were shown to activate naïve CD8 + T cells via CD28‑dependent signals and to drive their differentiation into PD‑1‑high effector CD8 + T cells; mregDC‑derived IL‑15 supported survival and maintenance of PD‑1‑high effector CD8 + T cells. PD‑1(+)TCF‑1(+)CD8(+) progenitor T cells were found to share clonal relationships with effector‑like cells in responders, indicating ongoing local CD8 + T‑cell differentiation during ICI. In contrast, nonresponders were dominated by terminally exhausted CD39(hi) TOX(hi) PD‑1(hi) CD8 + T cells and by clonal expansion of Tregs that replaced CXCL13 + Th clones [17].
Resistance to PD‑1 blockade was further associated with increased infiltration of pathogenic T helper 17 cells (pTh17), which secrete IL‑17 A and induce PD‑L1 expression on HCC cells; combined inhibition of IL‑17 A and PD‑L1 was reported to markedly enhance CD8 + CTLs infiltration [18].MAIT cells, a liver‑enriched innate‑like T‑cell subset, were shown to become dysfunctional following interactions with colony stimulating factor 1 receptor positive(CSF1R+)PD‑L1(+)TAMs, acquiring high PD‑1 and CD69 expression [19, 20]. Tumor‑derived MAIT cells exhibited a canonical CCR7(-)CD45RA(-)CD45RO(+)CD95(+) effector‑memory phenotype, with upregulation of inhibitory receptors including PD‑1, cytotoxic T‑lymphocyte‑associated protein 4 (CTLA‑4), and T‑cell immunoglobulin and mucin‑domain containing‑3 (TIM‑3), together with reduced secretion of IFN‑γ, IL‑17, granzyme B, and perforin. Such tumor‑derived MAIT cells were reported to be reprogrammed toward pro‑tumor activity, and high intratumoral MAIT density was correlated with adverse clinical outcomes in HCC [21]. Similarly, CD3(+)CD56(+) natural killer T cells (NKT) in HCC were documented to exhibit impaired production of TNF‑α and IFN‑γ, downregulation of the activating receptor natural killer group 2 member D (NKG2D), and upregulation of PD‑1; restoration of antitumor function by PD‑1 blockade was demonstrated [22].Multiple strategies to enhance cellular immunity were reported. A bispecific Vα7.2×PD‑L1 antibody was developed to mediate physical bridging of MAIT cells to PD‑L1‑expressing tumor cells and was shown to selectively induce MAIT expansion, activation, cytokine production, degranulation, and cytotoxicity only in the presence of target cells. sPD‑L1 was identified as a suppressor of NK‑cell function; to overcome this inhibition, glypican‑3 (GPC3)‑directed chimeric antigen receptor NK cells (GPC3‑CAR‑NK) were combined with a high‑affinity sPD‑L1 variant (L3C7c‑Fc), which bound and neutralized sPD‑L1 and restored CAR‑NK antitumor activity [23, 24].

Tumor microenvironmental regulation of PD‑1/ PD‑L1
The TME was characterized as a highly immunosuppressive milieu that facilitates tumor immune evasion. The TME was described to encompass canonical features such as hypoxia, increased extracellular matrix stiffness, and nutrient/energy deprivation, together with a complex milieu of cytokines and soluble mediators derived from neoplastic cells and infiltrating immune‑inflammatory cells. Accumulating evidence indicated that TME characteristics and extracellular factors, including cytokines, chemokines, metabolic by‑products and extracellular vesicle cargoes, regulate the PD‑1/PD‑L1 signaling axis through multiple molecular and cellular mechanisms.
Hypoxia was identified as a principal feature of the TME. Under hypoxic conditions, HIF‑1α expression was reported to be upregulated, and HIF‑1α was shown to transcriptionally increase expression of PD‑L1 [25]. CircPRDM4 was described to act as a scaffold that recruited HIF‑1α to the CD274 promoter, thereby stabilizing the interaction; concurrently, HIF‑1α was reported to transcriptionally upregulate the LncRNA MIR155HG. Hypoxic stress was indicated to promote nuclear export of interleukin enhancer binding factor 3 (ILF3), and MIR155HG binding to ILF3 was shown to enhance HIF‑1α mRNA stability, creating a positive regulatory loop [26, 27].In peri‑tumoral, non‑hypoxic regions, BCL2‑associated transcription factor 1 (BCLAF1) was reported to interact with cullin 3 (CUL3) to promote ubiquitination and degradation of prolyl hydroxylase domain‑containing protein 2 (PHD2), thereby inhibiting the PHD2–von Hippel–Lindau (VHL) pathway responsible for HIF‑1α degradation and enabling HIF‑1α accumulation [28].Multiple molecular pathways were described to converge on HIF‑1α to drive PD‑L1 upregulation. Under glucose deprivation in HCC cells, solute carrier family 7 member 11 (SLC7A11) was reported to be induced via an IL‑1β/ERK/Sp1 axis; SLC7A11‑mediated glutamate export reduced intracellular alpha‑ketoglutarate (α‑KG) levels, resulting in HIF‑1α stabilization and PD‑L1 induction [25]. Gap junction beta‑2 protein (GJB2) was shown to promote ubiquitination and degradation of inhibitor of kappa B alpha (IκBα), activating NF‑κB signaling and, via a HIF‑1α/GLUT‑1/PD‑L1 axis, increasing PD‑L1 expression [29]. The splicing factor serine/arginine‑rich splicing factor 10 (SRSF10) was reported to suppress IFNγ/α signaling and to upregulate PD‑L1 through an IFNα–FosB–HIF‑1α axis [30].Hypoxia‑driven intercellular communication within the TME was also implicated in PD‑L1 regulation. Hypoxia was reported to stimulate HCC cells to release exosomes containing miR‑1290; uptake of these exosomes by TAMs resulted in Akt2 inhibition and consequent upregulation of PD‑L1 [31]. Conversely, myeloid cells were shown to deliver exosomal miR‑223 to HCC cells, targeting HIF‑1α downstream effectors CD39/CD73–adenosine signaling and thereby decreasing PD‑L1 expression [32].Unexpectedly, hypoxia was reported to activate nuclear PD‑L1 (nPD‑L1) transcriptional activity and thus to participate in regulation of pyroptosis. Under hypoxia, PD‑L1 was shown to interact with p‑STAT3 to promote nuclear translocation of nPD‑L1 and to enhance transcription of gasdermin C (GSDMC). Following tumor necrosis factor alpha (TNFα) stimulation, GSDMC was reported to be specifically cleaved by caspase‑8 to generate an N‑terminal fragment (GSDMC‑N) that formed membrane pores and induced pyroptotic cell death. In this context, PD‑L1 was proposed to convert TNFα‑induced apoptosis into pyroptosis, thereby contributing to tumor necrosis [33].
Interferons were identified as key soluble mediators within the TME. IFN stimulation of tumor cells was documented to markedly increase expression of PD‑L1. IFN‑α, which is widely used in treatment of HBV infection, was shown to upregulate PD‑L1 and, via activation of AKT and c‑Jun N‑terminal kinase (JNK) signaling, to induce transcription of CXCL8 [34].Interferon‑gamma–mediated regulation of PD‑L1 was reported to occur through canonical and noncanonical routes. IFN‑γ was indicated to induce PD‑L1 transcription via downstream interferon regulatory factor 1 (IRF1) activity [35], and additional indirect regulatory mechanisms were described, including modulation via acetylated MEF2D [36, 37] and upregulation by the RNA helicase DDX1 (DDX1) [38].Notably, IFN‑γ was reported to regulate intracellular, mitochondrial functions of PD‑L1 (mtPD‑L1). IFN‑γ induced PD‑L1 upregulation and promoted PD‑L1 translocation to mitochondria, which elicited dynamin‑related protein 1 (Drp1)‑dependent mitochondrial fission and glycolytic metabolic reprogramming. Concomitant increases in glutathione peroxidase 4 (GPX4), ROS and lipid peroxidation (LPO) production were reported, leading to enhanced resistance to ferroptosis [39].Aberrant activation of the EGFR was described as common in HCC and as being driven by multiple ligands including epidermal growth factor (EGF) and TGF‑β, with consequent upregulation of PD‑L1. Golgi membrane protein 1 (GOLM1) was reported to phosphorylate EGFR, enhance downstream phosphorylation of STAT3‑Tyr705, and thereby increase PD‑L1 transcription [40]. EGFR activation was further shown to engage p38 mitogen‑activated protein kinase (p38 MAPK), which downregulated miR‑675‑5p; miR‑675‑5p was reported to bind the 3’‑UTR of PD‑L1 mRNA and to enhance its stability [41].
Recent findings indicated that alterations in extracellular matrix stiffness and accumulation of lactate also regulated the PD‑1/PD‑L1 axis. Extracellular matrix stiffening was reported to induce upregulation of PD‑L2 via activation of SET and MYND domain containing 3 (SMYD3) and increased trimethylation of histone H3 at lysine 4 (H3K4me3). PD‑L2 was described to bind ferritin light chain (FTL) mRNA, enhancing FTL mRNA stability and protein abundance; elevated FTL was reported to increase resistance to cystine/glutamate antiporter (xCT; also known as SLC7A11)‑dependent ferroptosis [42].Protein arginine methyltransferase 3 (PRMT3) was shown to mediate asymmetric dimethylation of pyruvate dehydrogenase kinase 1 (PDHK1), thereby enhancing PDHK1 kinase activity and promoting lactate production. PRMT3 was further reported to facilitate histone H3 lysine 18 lactylation (H3K18la), and H3K18la at the PD‑L1 promoter was indicated to strengthen promoter binding and to promote lactate‑induced PD‑L1 expression [43]. Lactylation was here defined as a post‑translational modification involving covalent addition of a lactyl moiety to lysine residues and was implicated as a link between metabolic reprogramming and epigenetic regulation of immune checkpoints.

Intracellular signaling pathways regulating the PD‑1/PD‑L1 axis
HCC cells were characterized by a high ROS burden and distinctive metabolic phenotypes, and multiple molecular regulators were reported to participate in the fine‑tuning of intracellular signaling pathways, resulting in aberrant pathway activation that modulated PD‑L1 expression.

STAT
The STAT signaling pathway was reported to be aberrantly and persistently activated in HCC cells. The STAT protein family comprises seven members (STAT1-4, STAT5A, STAT5B, STAT6); STAT1 and STAT3 were shown to bind directly to the CD274 promoter and to upregulate PD‑L1, and crosstalk between the STAT pathway and the interferon (interferon; IFN)‑regulated transcription factor interferon regulatory factor 1 (IRF‑1) was described [44, 45].Activation of STAT proteins was reported to occur downstream of membrane receptors for diverse growth factors (e.g. EGF, HGF) and inflammatory cytokines (e.g.IL‑6), which induced JAK–mediated phosphorylation of STATs. Phosphorylated STAT monomers were indicated to dimerize, translocate to the nucleus, and regulate target gene transcription. STAT proteins were described as not being readily directly druggable, and the STAT pathway was proposed to exert context‑dependent pro‑tumorigenic and tumor‑suppressive effects.Complex, multilayered regulation of STAT1 that resulted in PD‑L1 upregulation was reported. RNA binding motif protein 12 (RBM12) was shown to bind JAK1 mRNA via its RNA recognition motif (RRM) domain and to recruit eukaryotic translation initiation factor 4A2 (EIF4A2), thereby enhancing JAK1 protein translation and activating the JAK1/STAT1 axis [46]. After IFN‑γ stimulation, hexokinase domain containing 1 (HKDC1) was reported to interact with alpha‑smooth muscle actin (ACTA2) and to present STAT1 to the plasma membrane IFN‑γ receptor 1 (IFNγR1), promoting STAT1‑Tyr701 phosphorylation and nuclear translocation [47]; conversely, phospholipase A2 group VII (PLA2G7) was indicated to suppress levels of p‑STAT1 [48].High mobility group box 1 (HMGB1) signaling via TLR4/ERK/p65 was reported to upregulate full‑length gasdermin D (GSDMD‑FL) and to activate caspase‑1, producing the N‑terminal GSDMD fragment (GSDMD‑N). GSDMD‑N–induced calcium influx was described to promote calmodulin–histone deacetylase (HDAC) interactions and to facilitate p‑STAT1 nuclear entry and binding to the CD274 promoter [49]. Nicotinamide phosphoribosyltransferase (NAMPT) was shown to support nicotinamide adenine dinucleotide (NAD+)–dependent maintenance of α‑KG levels, preserving ten‑eleven translocation methylcytosine dioxygenase 1 (TET1) expression and activity; STAT1 was reported to promote TET1 binding to the IRF1 promoter and to sustain IRF1 demethylation [50]. The long noncoding RNA negative regulator of the interferon response (NRIR) was indicated to promote cytidine/uridine monophosphate kinase 2 (CMPK2)‑mediated ATP generation, thereby activating STAT1, increasing IRF1 and ultimately upregulating PD‑L1 via a STAT1–IRF1 axis [51].
STAT3 activation was described as another central mechanism promoting cancer stemness and immune evasion and as a key STAT family member involved in PD‑L1 upregulation [52, 53]. Regulation of STAT3 was reported at multiple levels: TLR9 signaling reduced poly‑ADP ribosylation (PARylation) and thereby decreased STAT3 degradation while increasing STAT3‑Tyr705 phosphorylation [54]; brain‑specific angiogenesis inhibitor 1‑associated protein 2‑like protein 2 (BAIAP2L2) was shown to interact with JAK1 and to potentiate JAK1/STAT3 signaling [55]. MicroRNA‑195 expression was reported to be induced by IRF1; miR‑195 bound checkpoint kinase 1 (CHEK1/CHK1) mRNA, suppressing CHK1 expression and resulting in increased p‑STAT3 levels [56, 57]. Lnc RNA TUG1, heat shock factor 1 (HSF1), and its downstream apolipoprotein J (APOJ) were indicated to regulate PD‑L1 expression via a JAK2–STAT3 axis [58, 59].Additional regulatory nodes affecting JAK2/STAT3 phosphorylation were described. HECT and RLD domain containing E3 ubiquitin protein ligase 2 (HERC2) was reported to couple endoplasmic reticulum‑resident protein tyrosine phosphatase non‑receptor type 1 (PTP1B), thereby inhibiting PTP1B‑mediated dephosphorylation of JAK2 and STAT3 [60]. Tankyrase‑binding protein 1 (TNKS1BP1) was shown to interact with TRIM21 to promote ubiquitination of CCR4‑NOT transcription complex subunit 4 at lysine 239 (CNOT4‑K239), leading to CNOT4 degradation, suppression of JAK2/STAT3 signaling and consequent downregulation of PD‑L1 expression [61].

β-Catenin
β‑catenin was frequently reported to be overexpressed in HCC cells and to participate in tumor initiation, growth, and metastasis; cytoplasmic stabilization and nuclear accumulation of β‑catenin were associated with poor prognosis [62]. Expression and activity of β‑catenin were reported to be tightly regulated by Wnt signaling and the PI3K/Akt pathway. Under conditions of Wnt pathway activation and Akt activation, glycogen synthase kinase 3 beta (GSK3β) was inhibited, resulting in impaired destruction‑complex‑mediated phosphorylation‑dependent degradation, increased β‑catenin accumulation and nuclear translocation. In addition, Akt was shown to phosphorylate β‑catenin at serine 552 (Ser552), which stabilized the protein and promoted nuclear import.
Multiple mechanisms that converged on β‑catenin to regulate PD‑L1 expression were described. Leukocyte‑associated immunoglobulin‑like receptor 1 (LAIR1) and NUAK family SNF1‑like kinase 1 (NUAK1) were reported to promote GSK3β Ser9 phosphorylation, thereby stabilizing β‑catenin and facilitating its nuclear entry; subsequent transcriptional activation of the cellular Myc oncogene (c‑MYC) and direct β‑catenin–mediated effects were indicated to upregulate PD‑L1 [63, 64]. Forkhead box O1 (FoxO1) was shown to interact with the CD274 promoter and to act as a transcriptional repressor; synoviolin 1 (SYVN1) was reported to promote ubiquitination‑dependent degradation of FoxO1 while concurrently enhancing β‑catenin nuclear translocation, resulting in PD‑L1 upregulation [65].Fibroblast growth factor 19 (FGF19) was described to increase PD‑L1 via dual mechanisms. FGF19 and hepatocyte growth factor (HGF) signaling through fibroblast growth factor receptor 4 (FGFR4) and c‑MET, respectively, were reported to converge on ERK1/2 to induce ETS variant transcription factor 4 (ETV4), with ETV4 activation leading to stimulation of Wnt/β‑catenin signaling [66]. Separately, selective engagement of FGFR4 by FGF19 was shown to drive insulin‑like growth factor 2 mRNA‑binding protein 1 (IGF2BP1) and downstream PI3K/Akt signaling, with joint promotion of PD‑L1 expression [67]. Dickkopf‑related protein 1 (DKK1) was reported to interact with its receptor cytoskeleton‑associated protein 4 (CKAP4) to activate Akt and thereby augment β‑catenin signaling [68]. Transmembrane 4 L six family member 1 (TM4SF1) was shown to enhance interaction between AKT1 (AKT1) and 3‑phosphoinositide‑dependent protein kinase‑1 (PDPK1), promoting AKT phosphorylation, downregulating the cyclin‑dependent kinase inhibitors p16 and p21, and thereby inhibiting non‑secretory senescence [69].
Multiple molecules were reported to downregulate PD‑L1 via modulation of β‑catenin. GSK3β‑Tyr56 was identified as a novel phosphorylation site; hepatocyte growth factor (HGF) receptor MET was reported to phosphorylate and activate GSK3β at Tyr56. Activated GSK3β was shown to phosphorylate PD‑L1 at threonine 180 and serine 184 (T180, S184), thereby promoting ubiquitin‑dependent degradation of PD‑L1 [70].The innate immune effector interferon‑stimulated gene 12a (ISG12a) was reported to stabilize Axin, facilitating β‑catenin ubiquitination and suppression of Wnt /β‑catenin signaling, with consequent downregulation of PD‑L1 [71]. The IL‑27 receptor WSX1 (also known as IL‑27Rα) was shown to transcriptionally repress the delta isoform of phosphoinositide 3‑kinase (PI3Kδ), resulting in inactivation of AKT, relief of inhibitory phosphorylation on GSK3β, activation of GSK3β and subsequent reduction of PD‑L1 expression [72].Targeted inhibition of the glycolytic enzyme phosphoglycerate mutase 1 (PGAM1) was reported to induce cellular energy stress and ROS‑dependent suppression of AKT, which similarly resulted in decreased PD‑L1 levels [73].
Reciprocal regulation of GSK3β by PD‑L1 was also described. In CD133(+) liver cancer stem cell (CSC)‑like populations, PD‑L1 was reported to interact with serum/glucocorticoid regulated kinase 2 (SGK2) and to activate SGK2. SGK2 was shown to increase inhibitory phosphorylation of GSK3β, thereby attenuating β‑catenin degradation and promoting CSC expansion and epithelial‑mesenchymal transition (EMT) [74].

mTOR
Tumor protein p53 (TP53) was reported to be one of the most extensively studied tumor suppressor genes and to be mutated in approximately 29.1–58.0% of HCC cases; TP53 mutation status was described to significantly influence PD‑L1 expression. Loss of TP53 was reported to activate the mechanistic target of rapamycin (mTOR) signaling pathway, resulting in PD‑L1 upregulation. In TP53 wild‑type HCC cells, inhibition of mTOR complex 1 (mTORC1) was shown to disrupt the interaction between p53 and E2F transcription factor 1 (E2F1), thereby promoting E2F1 nuclear translocation and consequent PD‑L1 induction. By contrast, in TP53‑mutant HCC, mTORC1 inhibition was reported to activate autophagy, leading to enhanced PD‑L1 degradation; thus, mTORC1 regulation of PD‑L1 was indicated to be highly dependent on TP53 status [75].A novel serine–threonine kinase, vaccinia related kinase 2 (VRK2), was described to promote phosphorylation of transketolase (TKT) at threonine 287 (Thr287) and to recruit the E3 ubiquitin ligase F‑box and leucine rich repeat protein 6 (FBXL6), thereby facilitating TKT ubiquitination and activation. Activated TKT was reported to inhibit ROS‑activated mTOR, with resultant increases in PD‑L1 and VRK2 expression [76]. In addition, the ubiquitin‑like modifier FAT10 (FAT10) was shown to upregulate PD‑L1 via activation of the PI3K/AKT/mTOR [77].
Figure 2༎PD‑L1 expression is modulated by microenvironmental cues and by multiple intracellular signaling pathways. Hypoxia has been shown to transcriptionally induce PD‑L1 via stabilization and activation of HIF‑1α. Convergent regulation at the post‑transcriptional and post‑translational levels—including alterations in mRNA stability, subcellular localization and protein phosphorylation—has been reported to influence PD‑L1 abundance and activity; upstream stimuli acting through these mechanisms culminate in activation of JAK/STAT, PI3K/Akt and Wnt/β‑catenin signaling, which further promote PD‑L1 upregulation. Epigenetic modifications within the PD‑L1 promoter region, such as DNA methylation and histone phosphorylation/acetylation, have been identified as important determinants of transcriptional output. At the transcriptional level, PD‑L1 expression is induced by a spectrum of transcription factors, including c‑MYC, NRF2, IRF, MEF2D, LMNB, p62, p65, SPI1, SOX2 and SOX18, in response to specific signaling inputs.Figure created with BioRender.

Mechanisms of transcriptional and Post‑transcriptional control of PD‑1/ PD‑L1
PD-1(PDCD1) was localized on human chromosome 2 (2q37.3), whereas CD274 was localized on human chromosome 9 (9p24.1) adjacent to CD273 and Janus kinase 2 (JAK2); amplification or mutation of genes in this chromosomal region was reported to result in overexpression of PD‑L1 in multiple malignancies. The PDCD1 promoter region was described to contain binding sites for nuclear factor of activated T cells (NFAT), activator protein 1 (AP‑1), and IRF9, among others, and to be responsive to TCR signaling, with transient expression under physiological conditions. The CD274 promoter was reported to include binding sites for HIF‑1α, specificity protein 1 (SP1), STAT3, NF‑κB, c‑MYC, and c‑JUN [78].Several upstream signaling molecules were reported to regulate CD274 transcription. Hepatocyte growth factor (HGF) was shown to transcriptionally upregulate PD‑L1 via its receptor MET proto‑oncogene (c‑Met) and downstream transcription factor p65 (RELA) [79]. IFN‑γ was indicated to regulate PD‑L1 through the antagonistic actions of IRF‑1/IRF‑2, with IRF‑1 promoting PD‑L1 transcription [35]. TGF‑β1 was reported to activate SRY‑box transcription factor 18 (SOX18), which directly bound the CD274 promoter to increase PD‑L1 transcription and concomitantly induced transcriptional activation of CXCL12 to foster an immunosuppressive microenvironment [80]. DNA binding inhibitor 1 (ID1) was described to facilitate c‑MYC binding to the CD274 promoter region [81].Transcriptional regulation of PD‑L1 by post‑translational modulators of transcription factors was also reported. PR domain zinc finger protein 1 (PRDM1) was shown to enhance transcription of ubiquitin specific peptidase 22 (USP22); USP22‑mediated deubiquitination of transcription factor SPI1 (SPI1) was reported to reduce SPI1 degradation and thereby promote PD‑L1 transcription [82]. Lamin B2 (LMNB2) was indicated to transcriptionally upregulate PD‑L1 and to be subject to speckle‑type POZ protein (SPOP)‑mediated ubiquitination and degradation, a mechanism proposed to maintain physiological PD‑L1 levels [83]. Interaction between the mitochondrial translocator protein (TSPO) and p62 was reported to impair autophagy, resulting in p62 accumulation; p62 was described to compete with Kelch‑like ECH‑associated protein 1 (KEAP1) and thereby stabilize nuclear factor erythroid 2–related factor 2 (Nrf2), leading to PD‑L1 upregulation and inhibition of ferroptosis [84].Regulation of PD‑L1 by RNA‑binding proteins and noncoding RNA pathways was reported. Lin‑28 homolog B (Lin28B) was shown to suppress miR‑let‑7, thereby increasing SRY‑box transcription factor 2 (SOX2) and octamer‑binding transcription factor 4 (OCT4); SOX2/OCT4 transcriptional motifs were reported to bind the promoters of CD274 and indoleamine 2,3‑dioxygenase 1 (IDO1), leading to transcriptional upregulation of PD‑L1 and IDO1 [85].
Noncanonical transcriptional regulators and chromatin modifications were also implicated. Epidermal growth factor (EGF) signaling via epidermal growth factor receptor (EGFR) was described to promote nuclear translocation of PKM2, enhance PKM2 phosphorylation and phosphorylation of histone H3 at threonine 11 (H3‑Thr11), and thereby increase CD274 transcription [86]. Myocyte enhancer factor 2D (MEF2D) was reported to be demethylated by lysine demethylase 1 A (KDM1A), and demethylated MEF2D was indicated to bind the CD274 promoter and activate transcription [87]. MEF2D acetylation status was described to be regulated reciprocally by p300 acetyltransferase and sirtuin 7 (SIRT7), with consequences for PD‑L1 transcriptional control [36]. Finally, TIP60 (TIP60)‑mediated acetylation of KIAA1429 was reported to increase KIAA1429 stability; acetylated KIAA1429 was shown to upregulate lysine demethylase 5B (KDM5B) via an m6A–YTH N6‑methyladenosine RNA binding protein 1 (YTHDF1)‑dependent mechanism, leading to suppression of forkhead box O1 (FoxO1) expression and consequent PD‑L1 upregulation [88].
Post‑transcriptional regulation of PD‑L1 expression was reported to be mediated extensively by non‑coding RNAs (ncRNAs), which were shown to upregulate PD‑L1 by multiple mechanisms. Non‑coding RNAs comprise RNA species that do not encode proteins and include long non‑coding RNAs (lncRNAs) and circular RNAs (circRNAs), among others.Stabilization of PD‑L1 messenger RNA (mRNA) by competitive ncRNA interactions was described. The lncRNA MIAT was reported to negatively regulate miR‑411‑5p, resulting in increased PD‑L1 mRNA levels [89]. The lncRNA HOXA‑AS3 was shown to act as a molecular sponge for miR‑455‑5p, thereby upregulating PD‑L1 mRNA [90]. A human circular RNA designated hsa_Circ_0005239 was reported to sequester miR‑34a‑5p, with consequent PD‑L1 mRNA upregulation [91]. Lnc‑CCNH‑8 was described to sponge miR‑217 to elevate PD‑L1 mRNA and additionally to stabilize PD‑L1 protein via a miR‑3173/plakophilin 3 (PKP3) axis [92]. By contrast, β‑glucuronidase (GUSB) was reported to promote expression of miR‑513a‑5p, which led to PD‑L1 downregulation [93].Non‑coding RNAs were also reported to function as protein scaffolds to modulate PD‑L1 transcription. The circPRDM4 was shown to facilitate binding of HIF‑1α to the CD274 promoter, enhancing HIF‑1α‑mediated transcriptional activation of PD‑L1 [26]. Conversely, circCCNY was reported to bind heat shock protein 60 (HSP60) and to promote SMAD‑specific E3 ubiquitin protein ligase 1 (SMURF1)‑mediated ubiquitination and degradation of HSP60; HSP60 degradation was described to release Raf kinase inhibitor protein (RKIP), thereby inhibiting the mitogen‑activated protein kinase (MAPK) pathway and, via a MAPK/c‑MYC/PD‑L1 axis, downregulating PD‑L1 expression [94].
Extracellular vesicle‑mediated intercellular transfer of ncRNAs within the TME was implicated in PD‑L1 regulation. HSCs were reported to secrete exosomes containing circWDR25 that entered HCC cells; circWDR25 was shown to sponge miR‑4474‑3p and thereby increase PD‑L1 expression in recipient tumor cells [95]. Hepatocellular carcinoma cells were described to release the lncRNA PCED1B‑AS1 within exosomes; PCED1B‑AS1 sequestered miR‑194‑5p in other HCC cells and in TAMs. Because miR‑194‑5p was demonstrated to downregulate both PD‑L1 and PD‑L2, its depletion by PCED1B‑AS1 resulted in net upregulation of PD‑L1 and PD‑L2 within the TME [96].A complex multimodal regulatory role for a tumor‑derived circRNA, circCCAR1, was reported in modulation of the PD‑1/PD‑L1 axis. Synthesis of circCCAR1 in HCC cells was promoted by p300 and eukaryotic translation initiation factor 4A3 (EIF4A3). N6‑methyladenosine modification of circCCAR1 was mediated by Wilms tumor 1‑associating protein (WTAP), and m6A marks enhanced circCCAR1 stability via binding to insulin‑like growth factor 2 mRNA‑binding protein 3 (IGF2BP3). CircCCAR1 was reported to sponge miR‑127‑5p, and a circCCAR1/miR‑127‑5p/WTAP feedback loop was described to further increase WTAP levels and thereby reinforce circCCAR1 stability. Secretion of circCCAR1 was shown to be dependent on heterogeneous nuclear ribonucleoprotein A2/B1 (hnRNPA2B1); exosomes containing circCCAR1 were taken up by CD8(+) T cells, wherein circCCAR1 stabilized PD‑1 expression. In addition, EP300‑induced expression of cell division cycle and apoptosis regulator protein 1 (CCAR1) was reported to enhance CCAR1 interaction with β‑catenin and thereby to promote PD‑L1 transcription [97].
N6‑methyladenosine modification was identified as a prominent post‑transcriptional regulatory mechanism of the PD‑L1 axis, acting both to stabilize PD‑L1 mRNA and to promote PD‑L1 transcription. Leucine‑rich pentatricopeptide repeat‑containing protein (LRPPRC) was reported to directly increase PD‑L1 mRNA stability [98]. Lipopolysaccharide (LPS) exposure was shown to promote m6A modification of the LncRNA MIR155 host gene (MIR155HG) in HCC cells via METTL14; MIR155HG stabilization was further mediated through an ELAV like RNA binding protein 1 (ELAVL1; also known as HuR)‑dependent pathway, and MIR155HG was reported to sponge miR‑223 thereby stabilizing PD‑L1 mRNA [99]. Taurine upregulated gene 1 (TUG1) was indicated to be upregulated by METTL3 ‑dependent m6A modification; TUG1 functioned as a sponge for miR‑141 and miR‑340 to increase PD‑L1 and CD47 expression, and TUG1 was reported to interact with Y‑box binding protein 1 (YBX1) to promote PD‑L1 and CD47 transcription [100]. In addition, YTH N6‑methyladenosine RNA binding protein 2 (YTHDF2) was shown to recognize an m6A mark within the 5’‑untranslated region (5’‑UTR) of ETS variant transcription factor 5 (ETV5) mRNA, recruit eukaryotic translation initiation factor 3 subunit B (eIF3b) to enhance ETV5 translation, and thereby augment PD‑L1 transcription [101].

Mechanisms of translation and Post‑translational control of PD‑1/PD‑L1
Post‑translational modifications of PD‑L1 in HCC have been principally investigated for phosphorylation, ubiquitination, and lactylation, each modification being implicated in regulation of PD‑L1 activation, stability, subcellular localization, and molecular interactions(Fig. 1). Various molecular mechanisms that stabilize PD‑L1 via glycosylation and ubiquitin‑related processes have been reported.Guanosine monophosphate synthetase (GMPS) was described as an adaptor linking the Sect. 61 translocon complex subunit alpha (Sec61α) to the catalytic subunit STT3A of the oligosaccharyltransferase (OST) complex; GMPS enhanced interaction between PD‑L1 and STT3A, promoted PD‑L1 glycosylation, and simultaneously reduced ubiquitin tagging, thereby increasing PD‑L1 stability [102]. CKLF‑like MARVEL transmembrane domain containing 4 (CMTM4) was reported to facilitate PD‑L1 plasma‑membrane recycling and to prevent ubiquitin‑dependent proteasomal degradation, stabilizing membranous PD‑L1 independently of IFN‑γ signaling [103]. B‑cell lymphoma‑associated transcription factor 1 (BCLAF1) was shown to interact with the E3 ubiquitin ligase speckle‑type POZ protein (SPOP), competitively inhibiting SPOP‑mediated PD‑L1 ubiquitination and degradation [104]. The deubiquitinase OTU deubiquitinase, ubiquitin aldehyde binding 2 (OTUB2) was demonstrated to bind PD‑L1 and to impede endoplasmic reticulum‑associated degradation (ERAD) pathway‑mediated ubiquitination of PD‑L1 in the endoplasmic reticulum [105]. Protein O‑fucosyltransferase 1 (POFUT1) was reported to prevent TRIM21‑mediated ubiquitination and degradation of PD‑L1 in a manner independent of POFUT1 enzymatic activity [106]. Ubiquitin‑specific peptidase 47 (USP47) was identified as a PD‑L1 deubiquitinase that stabilized PD‑L1 through removal of ubiquitin chains [107].
Golgi membrane protein 1 (GOLM1) was reported to promote COP9 signalosome subunit 5 (CSN5)‑mediated PD‑L1 deubiquitination and to inhibit Rab27b expression in the trans‑Golgi network (TGN), thereby facilitating trafficking of PD‑L1‑containing exosomes [108]. GOLM1 was further described to interact selectively with epidermal growth factor receptor (EGFR), assisting EGFR and other receptor tyrosine kinases (RTK) to anchor in the TGN and recycle to the plasma membrane, sustaining phosphorylation of STAT3(Tyr705), nuclear translocation, and transcriptional upregulation of PD‑L1 [40, 109]. By contrast, the E3 ubiquitin ligase Parkin was shown to induce ubiquitination and degradation of PD‑1; loss of Parkin was reported to attenuate antitumor immune responses, reduce hepatic mitophagy, and contribute to formation of an immunosuppressive microenvironment [110].
Chaperone‑mediated autophagy (CMA) was described as an alternative degradative route. Cyclin‑dependent kinase 5 (CDK5) was reported to phosphorylate PD‑L1 at threonine 290 (T290), promoting association with the chaperone heat shock cognate 70 kDa protein (HSC70); dysregulation of CDK5 in HCC was associated with reduced PD‑L1 degradation [111].Recent work has also implicated PD‑L1 lactylation as a functional modification. The acetyltransferase p300 (EP300) and histone deacetylase 2 (HDAC2) were reported to mediate PD‑L1-K189 lactylation and delactylation, respectively. HDAC2‑mediated delactylation of PD‑L1 facilitated vimentin‑dependent nuclear translocation of PD‑L1; nuclear PD‑L1 was shown to interact with the transcription factor Ying Yang 1 (YY1) to transcriptionally upregulate squalene epoxidase (SQLE), the rate‑limiting enzyme in cholesterol biosynthesis, thereby accelerating cholesterol synthesis and promoting tumor growth [112].

Other cellular components within the tumor stroma

TAM
TAMs were reported to arise predominantly from monocyte‑derived macrophages (MoMFs) and to be skewed toward an M2‑like phenotype, constituting a major component of the HCC‑infiltrating stroma(Fig. 2). Except for certain subpopulations, HCC tumor cells were described to express low levels of PD‑L1, whereas the majority of intra‑tumoral inflammatory cells exhibited higher PD‑L1 expression. High PD‑L1 expression by TAMs was indicated to represent a principal source of PD‑L1 within HCC and to contribute importantly to recurrence from minimal residual disease (MRD) after therapy [113].Compared with IFN‑γ‑dependent induction of PD‑L1, TAM‑expressed PD‑L1 was reported to be relatively less IFN‑γ‑dependent and more stably maintained, implying regulation by multiple additional factors. CD14(+) monocytes, which were recruited into the TME as TAM precursors, were described to undergo a glycolytic metabolic switch; ERK signaling was reported to induce expression of complement component CD93, and CD93 was shown to upregulate PD‑L1 via an AKT–GSK3β axis. CD93(+)monocytes were further reported to secrete the multifunctional proteoglycan versican, which inhibited migration of CD8(+) T cells [114].S100A9(+)CD14(+)monocytes and IgA(+) monocytes were described to be enriched in the TME and to express high levels of PD‑L1;IgA(+) monocytes were reported to secrete IgA immune complexes that stimulated M2 TAMs and, via YAP signaling, upregulated PD‑L1 in TAMs [115, 116].
Following monocyte differentiation into TAMs, high expression of OIT3 was indicated to activate NF‑κB signaling and to increase PD‑L1 expression [117]. Suppression of apolipoprotein C‑I (APOC1), which was described to be highly expressed in TAMs, was reported to increase ferroptosis and to reprogram M2‑type TAMs toward an M1 phenotype, resulting in PD‑L1 downregulation and restoration of anti‑tumor activity [118]. Deficiency of GSK3β within TAMs was shown to reduce ubiquitin‑dependent PD‑L1 degradation, thereby increasing PD‑L1 abundance in TAMs [119].
HCC tumor cells were reported to modulate PD‑L1 expression in TAMs. Tumor‑derived Sonic hedgehog (Shh) and lysyl oxidase‑like 4 (LOXL4) were indicated to drive STAT3 and STAT1, respectively, leading to PD‑L1 upregulation in TAMs [120, 121]. Angiotensin‑converting enzyme 2 (ACE2) expression in HCC cells was shown to act through an ACE2/Ang1‑7/Mas receptor (Mas) axis to inhibit NF‑κB signaling and to reduce secretion of CCL5; decreased CCL5 levels were reported to attenuate JAK–STAT3 activation in TAMs and thereby to lower TAM PD‑L1 expression [122]. Schlafen family member 11 (SLFN11) was shown to compete with RNA binding motif protein 10 (RBM10) for binding, blocking TRIM21‑mediated RBM10 ubiquitination and degradation; stabilized RBM10 was reported to promote NUMB (NUMB) expression, to suppress Notch signaling and tumor‑derived CCL2, and thereby to inhibit TAM M2 polarization and reduce PD‑L1 expression [123].
Trace elements and cellular energy metabolism were also reported to modulate PD‑L1 expression in TAMs. Zinc deficiency in HCC was shown to impair TLR4 function and to reduce PD‑L1 endocytosis in TAMs [124]. SLC7A11, which participates in glutathione biosynthesis, was reported to promote TAM M2 polarization and PD‑L1 upregulation via a SOCS3–STAT6–PPARγ axis [125]. Tumor‑derived fibronectin 1 (FN1) was described to activate HIF‑1α in TAMs, inducing the glycolytic enzyme PKM2 and driving a glycolytic polarization characterized by HLA‑DR(high) CD86(high). PKM2 was reported to upregulate PD‑L1 and to promote production of anti‑tumor IL‑12p70. Targeting glycolysis was indicated to attenuate PD‑L1‑mediated immune escape, but it was cautioned that such interventions could also impair innate anti‑tumor immunity [126].
Tumor‑derived lactate was shown to increase nuclear protein 1 (NUPR1) expression in TAMs via histone lactylation; NUPR1 was reported to inhibit ERK and JNK signaling, to enhance M2 polarization, and to elevate expression of PD‑L1 and signal regulatory protein alpha (SIRPα) [127]. Interestingly, dietary interventions such as short‑term starvation (STS) were indicated to regulate PD‑L1. Exosomal PD‑L1 (exoPD‑L1) derived from TAMs was reported to attenuate the efficacy of ICI. STS was shown to inhibit TAM exoPD‑L1 secretion via modulation of a fructose‑1,6‑bisphosphatase 1 (FBP1)/AKT/Ras‑related protein Rab‑27 A (Rab27a) signaling axis [128].

TAN
Neutrophils (NEUT) were reported to contribute to anti‑tumor responses through release of factors such as thrombospondin‑1 (TSP‑1) and hydrogen peroxide (H2O2), and were identified as an important cellular component of the TME. However, tumor‑infiltrating neutrophils, termed tumor‑associated neutrophils (TANs), were indicated to be polarized by the TME toward an immunosuppressive N2 phenotype, which was associated with upregulation of PD‑L1.HCC cells were reported to promote TAN infiltration and to induce PD‑L1 upregulation through multiple mechanisms. The CRK‑like proto‑oncogene, adaptor protein (CRKL) was shown to enhance secretion of VEGFα and CXCL1, thereby increasing TAN recruitment; tumor‑derived lactate was reported to drive TAN polarization toward the N2 phenotype and to elevate PD‑L1 expression in TANs [129, 130]. Hepatocytes within the TME were described to secrete serum amyloid A (SAA), which recruited TANs and, via activation of lactate dehydrogenase A (LDHA)/STAT3 signaling to stimulate glycolysis, induced PD‑L1 upregulation and release of the tumor‑suppressive factor macrophage migration inhibitory factor (MIF) by TANs [131].

CAF
Cancer‑associated fibroblasts (CAFs) and HSCs within the tumor stroma were reported to express PD‑L1 and to contribute to establishment of an immunosuppressive TME. CAFs were described to express the IgA receptor CD71, and stimulation with IgA was reported to markedly increase PD‑L1 levels in CAFs [132]. Under hypoxic stress, CAFs were shown to secrete exosomes containing circHIF1A; uptake of these exosomes by HCC cells resulted in binding of circHIF1A to the RNA‑binding protein human antigen R (HuR), which in turn stabilized PD‑L1 mRNA and increased PD‑L1 expression [133].HSCs were reported to be regulated by HCC cells and to upregulate PD‑L1 in response. Tumor‑derived exosomes enriched in miR‑500a‑3p were indicated to be taken up by HSCs and by peripheral blood mononuclear cells (PBMCs). miR‑500a‑3p was shown to act via suppressor of cytokine signaling 2 (SOCS2) to modulate the JAK3–STAT5A/STAT5B axis, resulting in PD‑L1 upregulation in HSCs and stabilization of PD‑1 expression in PBMCs, thereby promoting an immunosuppressive milieu [134].

ICI therapy targeting the PD‑1/PD‑L1 axis in HCC
The landscape of advanced HCC management has been profoundly altered by ICI therapy, with clinical use having been rapidly advanced from later-line settings to first-line treatment and with combination regimens established as foundational strategies. Although phase III trials of PD-1-directed monoclonal antibodies (nivolumab and pembrolizumab) as first-line monotherapy failed to meet their primary endpoints [135, 136], regimens exemplified by atezolizumab plus bevacizumab (an “immunotherapy + anti-angiogenesis” strategy) and durvalumab plus tremelimumab (a “dual-immunotherapy” strategy) have demonstrated significant overall survival benefit and have been adopted as new first-line standards [137, 138]. Monotherapy with PD-1 inhibitors is nonetheless retained as a valuable option for defined patient subgroups.Future investigative efforts are expected to prioritize elucidation of mechanisms of ICI resistance and evaluation of more precisely tailored combination approaches. Examples include integration of ICIs with locoregional modalities such as transarterial chemoembolization (TACE) to increase tumor immunogenicity and potentiate systemic antitumor immunity [139]. Ultimately, the identification of responsive patient cohorts by means of multi-omics profiling and the design of individualized treatment algorithms incorporating ICIs, targeted agents and local therapies are proposed as strategies to extend survival benefits to a greater proportion of patients with HCC.
Clinical application of ICIs remains confronted with substantial challenges, principally limited objective response rates that reflect both primary and acquired resistance, and the lack of validated biomarkers capable of accurately predicting therapeutic efficacy. Investigation of the impact of ICIs on the PD‑1/PD‑L1 signaling axis and of the mechanisms underlying immune‑related adverse events (irAEs) induced by ICIs is expected to facilitate prolongation of response durability and reduction of treatment‑related toxicity.Modulation of the PD‑1 receptor by ICI was reported. An increased frequency of a PD‑1 isoform, Δ42PD‑1 was observed in T cells from patients who exhibited continued disease progression after treatment with anti‑PD‑1 antibody; Δ42PD‑1(+) T cells were described as more profoundly exhausted and tumor‑infiltrating, and tumor‑infiltrating Δ42PD‑1 + T cells were reported to promote HCC progression via TLR4 signaling [140].Effects of PD‑L1‑targeting antibodies on tumor cells were also documented. PD‑L1 antibody treatment was reported to stabilize glycosylated PD‑L1 receptors on the HCC cell surface, enhancing direct interaction with the AXL receptor tyrosine kinase (AXL) and thereby activating downstream AKT and ERK1/2 signaling, which induced tumor cell proliferation [141].PD‑1 blockade was further associated with transcriptional and signaling adaptations within tumor cells. Upregulation of Yes‑associated protein 1 (YAP1) was induced in HCC cells following anti‑PD‑1 antibody treatment; YAP1 was reported to engage a positive‑feedback circuit involving JAK1/STAT1/p‑STAT3(Y705) to increase PD‑L1 expression, potentially attenuating antitumor immunity [44].
Concomitant therapies administered with ICI in hepatocellular carcinoma were noted to influence ICI efficacy, and drug‑drug and drug‑tumor interactions were proposed to warrant systematic evaluation to optimize clinical benefit and limit immune‑related hepatic toxicity.Lenvatinib was reported to inhibit fibroblast growth factor receptor 2 (FGFR2) tyrosine kinase activity, resulting in dephosphorylation of phosphatidylinositol glycan anchor biosynthesis class L(PIGL)-Y81. Loss of PIGL‑Y81 phosphorylation abolished interaction with importin α/β1, leading to cytoplasmic retention of PIGL and prevention of its nuclear translocation. Consequently, disruption of PIGL‑mediated interference with the c‑Myc/BRD4 promoter interaction was relieved, and expression of CCL2 and CCL20 was increased, thereby promoting an immunosuppressive TME [142].
Multiple mechanisms that re‑sensitize tumors to ICI have been identified. Schlafen family member 11 (SLFN11) expression was markedly upregulated in ICI‑responding patients; SLFN11 was reported to competitively bind and stabilize RNA binding motif protein 10 (RBM10), suppress Notch signaling and CCL2 transcription, reduce TAMs infiltration and M2 polarization, and decrease PD‑L1 expression [123]. Metabolic and epigenetic enzymes and receptors altered after ICI were proposed as synergistic targets. Peroxisome proliferator‑activated receptor‑γ (PPARγ) was shown to transcriptionally activate vascular endothelial growth factor alpha(VEGF‑α)‑driven expansion of myeloid‑derived suppressor cells (MDSCs) and to induce CD8(+) T‑cell dysfunction; selective PPARγ antagonism was reported to reprogram the TME and to restore sensitivity to anti‑PD‑L1 therapy [143].ATP citrate lyase (ACLY) catalysis of citrate to acetyl‑coenzyme A (acetyl‑CoA) for fatty acid and cholesterol biosynthesis was targeted to overcome resistance: ACLY inhibition induced peroxidation of polyunsaturated fatty acids (PUFAs) and mitochondrial damage, thereby activating the cyclic GMP‑AMP synthase–stimulator of interferon genes (cGAS‑STING) innate immune pathway and overcoming anti‑PD‑L1 resistance [144]. The active metabolite SN‑38 or metformin was reported to activate forkhead box O3 (FOXO3), suppress c‑Myc/STAT3 signaling, and downregulate PD‑L1 expression [145]. MER proto‑oncogene tyrosine kinase (MerTK) was shown to upregulate solute carrier family 7 member 11 (SLC7A11) via the ERK/SP1 pathway, thereby limiting ferroptosis, recruiting MDSCs and promoting an immunosuppressive TME; inhibition of MerTK by sitravatinib was reported to promote tumor ferroptosis and to reduce MDSC infiltration [146].
ICI therapy was also associated with a risk of drug‑induced liver injury (DILI) [147]. In a Japanese retrospective cohort of 135 patients treated with anti‑PD‑1, approximately 5% of patients developed PD‑1 inhibitor‑related DILI, and MAFLD was identified as a risk factor for PD‑1 inhibitor‑related DILI [148]. Mechanistic studies indicated that PD‑L1 was also highly expressed in normal hepatocytes; treatment with atezolizumab enhanced phosphorylation of RIP3/RIPK3 and mixed lineage kinase domain‑like pseudokinase (MLKL), increased LDH release, reduced cell viability and proliferation, and induced PD‑L1–mediated necroptotic cell death in hepatocytes [149].

CCA
Cholangiocarcinoma (CCA) was characterized as a highly desmoplastic tumor in which malignant epithelial cell populations substantially outnumber stromal cells; two major anatomical subtypes were recognized: extrahepatic cholangiocarcinoma (eCCA) and intrahepatic cholangiocarcinoma (iCCA). Only approximately 25% of eCCA cases were reported to harbor actionable driver alterations, and roughly 11% were classified as having immune‑molecular alterations with higher expression of PD‑1 on lymphocytes (> 5% lymphocytes) and PD‑L1 on tumor or stromal cells (> 1% tumor or stromal cells) [150].A greater density of TILs was associated with earlier disease stage, lower invasive potential, and improved overall survival (OS), whereas a high proportion of PD‑1(hi)CD8(+) TILs correlated with advanced Tumor‑Node‑Metastasis (TNM) stage [151]. In iCCA, PD‑L1 positivity was observed in approximately one‑third of tumor cells and tumor‑infiltrating immune cells, whereas PD‑L1 expression by peritumoral epithelial cells and peritumoral immune cells was comparatively low [152]. High PD‑L1 expression on iCCA tumor cell surfaces was presently interpreted as primarily induced by pre‑existing activated CD8 + T cells and was therefore described as a “battlefield” marker; in this context, tumor PD‑L1 overexpression was associated with favorable prognosis [153].TAMs were identified as a principal source of PD‑L1 in CCA. An increase in PD‑L1(+)CD68(+)TAMs was associated with shortened overall survival in cholangiocarcinoma, and TAMs were proposed to cooperate with PD‑1‑high CD8 + T cells to suppress antitumor immunity [154]. Attempts to deplete TAMs alone failed to retard tumor progression because compensatory expansion of granulocytic myeloid‑derived suppressor cells (G‑MDSCs) further diminished T cell responses and promoted immune escape; combined inhibition of TAMs and G‑MDSCs was reported to enhance the efficacy of ICI [155].Additional immune perturbations were noted in iCCA, including a marked reduction in the proportion of NK cells and enrichment of immunosuppressive CD39(+)FoxP3(+)CD4(+)Treg together with exhausted‑phenotype PD‑1 + CD39 + CD8 + T cells; these populations were associated with poor prognosis in iCCA [156]. The TME in CCA was thus considered highly complex, and comprehensive characterization of the CCA TME was reported to be incomplete.
CCA was characterized as an immune “desert,” and several mechanisms regulating the PD‑1/PD‑L1 signaling axis have been recently identified with potential to enhance antitumor immunity. Kirsten rat sarcoma viral oncogene homolog (KRAS) mutation, one of the most frequent molecular alterations in iCCA, was reported to cause sustained activation of downstream extracellular signal‑regulated kinase (ERK) signaling; inhibition of ERK signaling was shown to activate tumor autophagy and to reduce PD‑L1 expression [157].Reduced expression of liver kinase B1 (LKB1) in iCCA was associated with poor prognosis; genetic silencing of LKB1 was reported to increase PD‑L1 expression in iCCA cells and in tumor‑derived exosomes, concomitant with decreased LKB1 content in exosomes [158, 159]. Beyond canonical signaling regulation, PD‑L1 in CCA was found to be subject to post‑transcriptional and post‑translational control. Y‑box binding protein 1 (YBX1) was shown to stabilize STAT1 mRNA translation in a 5‑methylcytosine (m5C)–dependent manner, thereby activating the STAT1/PD‑L1 axis [160]. The N6‑methyladenosine (m6A) demethylase AlkB homolog 5 (ALKBH5) was reported to inhibit m6A modification within the 3’‑UTR of PD‑L1 mRNA, preventing YTH N6‑methyladenosine RNA binding protein 2 (YTHDF2)–dependent degradation of PD‑L1 transcript [161]. Conversely, METTL14 was shown to install m6A marks on the 3’‑UTR of SIAH E3 ubiquitin protein ligase 2 (SIAH2) mRNA, promoting YTHDF2‑dependent degradation of SIAH2 transcript and thereby decreasing SIAH2‑mediated PD‑L1 proteasomal degradation [162]. Methyltransferase‑like 3 (METTL3)–mediated m6A modification was reported to stabilize a circular RNA derived from circSLCO1B3 was shown to promote iCCA proliferation and metastasis via the miR‑502‑5p/HOXC8/SMAD3 axis and to inhibit the E3 ubiquitin ligase adaptor speckle‑type POZ protein (SPOP), resulting in PD‑L1 stabilization [163].
Tumor cell–derived extracellular vesicles were implicated in modulation of TAMs PD‑L1. Exosomes enriched in miR‑183‑5p from iCCA cells were reported to upregulate PD‑L1 on TAMs via the miR‑183‑5p/PTEN/AKT/PD‑L1 pathway [164]. Notably, PD‑L1 expressed by activated HSCs was reported to promote tumor growth through an immune‑independent mechanism: HSC PD‑L1 expression stabilized TGF‑β receptor I and II (TβRI; TβRII), thereby enhancing TGF‑β–induced HSC activation; inhibition of PD‑L1 on HSCs was shown to suppress HSC activation and to impede iCCA growth [165].

The PD‑1/PD‑L1 axis in chronic liver diseases

HBV/HCV
During acute hepatitis B virus (HBV) infection, PD-1 was transiently upregulated in vivo, and PD-L1 expression on KCs was increased to initiate differentiation of CD8 + T cells into effector cells; liver-resident natural killer (LrNK) cells were likewise induced to upregulate PD-L1. Blockade or conditional deletion of PD-L1 was shown to enhance intrahepatic accumulation of NK cells and CD8 + T cells and to facilitate HBV clearance and memory formation [166–168]. Following establishment of chronic HBV infection, PD-1 and PD-L1 were observed to be significantly increased in hepatic cells, immune cells, and plasma. PD-1 expression was primarily detected on lymphocytes infiltrating the portal tracts, PD-L1 expression was predominantly detected on lymphocytes, hepatocytes, and LSECs, whereas PD-L2 expression was restricted to KCs and DCs [169].Innate immune cells characterized phenotypically as CD3(+)CD161(++)TCR Vα7.2(+) mucosal-associated invariant T (MAIT) cells were reported to contribute to HBV control. In chronic HBV infection, peripheral blood MAIT cell frequency was reduced and PD-1 expression on these cells was increased, a phenomenon that was proposed to result from persistent immune activation induced by HBV [170]. Maternal-to-child transmission was identified as a major route leading to chronic HBV infection: maternal hepatitis B e antigen (HBeAg) was reported to upregulate PD-L1 expression on offspring KCs and, upon HBeAg re-stimulation, to drive M2-like polarization of KCs that supported HBV persistence; PD-L1 antibody-mediated blockade on Kupffer cells was shown to promote activation of offspring CD8 + T cells and facilitate HBV clearance [171].In addition to cell-surface PD-L1, soluble programmed death-ligand 1 (sPD-L1) concentrations were found to be markedly elevated in the plasma of HBV-infected patients. Lower sPD-1 levels were associated with spontaneous seroclearance of hepatitis B surface antigen (HBsAg), whereas plasma sPD-L1 levels were higher in patients with HBV-related hepatocellular carcinoma (HBV-HCC) than in those with chronic hepatitis B (CHB) or cirrhosis [172–174]. Genetic polymorphisms within the PD-1/PD-L1 axis have been correlated with CHB prognosis: the PD-1.9 TT genotype was identified as a risk factor for development of HCC in CHB patients, whereas the PD-1.5 CT genotype was associated with a reduced HCC risk [172]. In the PD-L1 3’-untranslated region (3′-UTR) single nucleotide polymorphism rs2297136, the AA genotype was associated with CHB progression, while the GG genotype was reported to exert a protective effect against progression to HCC [173].
HBV components were reported to modulate the PD-1/PD-L1 axis through multiple mechanisms, thereby further suppressing adaptive immune interactions, inducing T cell exhaustion, and promoting an immunosuppressive microenvironment. During chronic HBV infection, upregulation of the T cell–specific transcription factor eomesodermin (Eomes) was observed; consequent enhancement of transcription factors such as nuclear factor of activated T cells cytoplasmic 1 (NFATc1), B‑lymphocyte‑induced maturation protein‑1 (Blimp‑1), and forkhead box O1 (FoxO1) was reported to drive PD-1 expression and to induce CD8 + T cell exhaustion [175].HBsAg was shown to act via the myeloid differentiation primary response 88 (MyD88)/nuclear factor kappa‑B (NF‑κB) signaling pathway to upregulate human leukocyte antigen‑E (HLA‑E) and PD-L1 on immunosuppressive monocytes. These monocytes were reported to induce NK cells to produce IL‑10 and to express high levels of PD‑1, thereby suppressing autologous T cell activation and impairing viral clearance; similar effects were observed for LrNK cells [176, 177].
In HBV‑HCC, tumor cells were characterized by aberrant expression of human leukocyte antigen‑DR (HLA‑DR) together with high PD‑L1 expression. Such tumor cell phenotypes were proposed to concurrently attract CD8 + T cells via antigen presentation while inhibiting their effector function through PD‑L1 overexpression, thereby facilitating tumor immune evasion [178].(Fig. 3)Mechanistic investigations demonstrated that HBV‑induced activation of STAT3 led to activation of the spalt‑like transcription factor 4 (SALL4), which repressed transcription of microRNA‑200c (miR‑200c) in HBV‑HCC cells, resulting in PD‑L1 upregulation [179].Additional HBV proteins were shown to modulate oncogenic signaling and DNA‑repair pathways to increase PD‑L1 expression. The hepatitis B virus X protein (HBx) and other viral components were reported to activate the phosphatase and tensin homolog (PTEN)/β‑catenin/c‑Myc signaling cascade; furthermore, interactions between viral proteins and the DNA repair enzyme poly(ADP‑ribose) polymerase 1 (PARP1) were described to inhibit PARP1 nuclear translocation, contributing to PD‑L1 upregulation in HBV‑HCC cells [180, 181]. Conversely, persistent HBV presence was reported to induce expression of interferon‑stimulated gene 12a (ISG12a) in hepatocellular carcinoma cells; ISG12a interaction with tripartite motif containing 21 (TRIM21) was shown to inhibit phosphorylation of protein kinase B(AKT) and β‑catenin, thereby suppressing PD‑L1 expression [182].
The effects of chronic hepatitis B virus (HBV) infection and treatment of HBV‑HCC on the PD‑1/PD‑L1 signaling axis have been identified as areas of ongoing controversy. An inverse correlation between HBV DNA viral load and PD‑1 expression was reported; reduction of HBV DNA was shown to facilitate upregulation of the tumor‑suppressive miR‑138, which binds the 3’‑UTR of PD‑1 mRNA and suppresses PD‑1 expression [183].The nucleoside reverse transcriptase inhibitor (NRTI) entecavir (ETV), a commonly used agent for CHB, was reported to upregulate CKLF‑like MARVEL transmembrane domain‑containing 6 (CMTM6) in hepatocytes irrespective of HBV infection status, leading to increased hepatocyte surface expression of PD‑L1; this effect was proposed as a potential explanation for retrospective observations of a higher incidence of HCC among patients treated with ETV compared with tenofovir disoproxil fumarate (TDF) [184].In HBV‑HCC, the intratumoral core immune microenvironment was documented to be markedly suppressed. Tumor‑margin tissues were found to be significantly enriched for HBV‑specific PD‑1‑expressing CD8 + tissue‑resident memory T cells (Trm) that exhibited features of exhaustion and demonstrated more pronounced responses following PD‑L1 blockade, indicating potential benefits of ICI in this compartment; however, ICI was also associated with a risk of immune‑related hepatic injury [185]. In a cohort of cancer patients with chronic HBV infection treated concomitantly with antiviral therapy and PD‑1 inhibitors, decreases in HBsAg levels were observed in approximately 71.6% of patients, increases in HBsAg levels were observed in 28.3% of patients, hepatitis B virus reactivation (HBVr) occurred in 6% of patients, and immune‑related adverse events (irAEs) occurred in 20% of patients [186]. Conversely, immune checkpoint inhibitor(ICI) therapy was reported to accelerate HBsAg seroclearance in cancer patients with baseline HBsAg < 100 IU/mL in a separate study [187]. Collectively, clinical evidence indicated that the net benefit of ICI therapy for HBV‑HCC patients may exceed associated risks, although therapeutic regimens and patient selection criteria were identified as requiring further optimization.HBV‑HCC patients were noted to receive polypharmacy in clinical practice. The small molecule 2,5‑dimethyl‑celecoxib (DMC) was reported to activate the AMPK pathway and to promote HBx‑induced PD‑L1 ubiquitination and degradation via ring‑box 1 (RBX1)‑mediated ubiquitin ligase activity [188]. Studies investigating the effects of commonly co‑administered agents on the PD‑1/PD‑L1 axis in HBV‑HCC were identified as scarce; further investigation of such drug interactions was recommended to facilitate optimization of systemic therapy for patients with HBV‑HCC.
Investigations of the PD‑1/PD‑L1 signaling axis in hepatitis C virus (HCV) infection have been limited. Single‑cell RNA sequencing of liver specimens from patients with chronic hepatitis C demonstrated that interferon‑stimulated genes (ISGs) were downregulated following antiviral therapy; viral load was inversely correlated with ISG expression in neutrophils (Neut), monocytes, and DCs, and neutrophils exhibiting high ISG expression were found to upregulate PD‑L1/PD‑L2 expression [189].The HCV core protein (HCVc) was reported to induce upregulation of PD‑L1 on KCs via Toll‑like receptor 2 (TLR2) signaling, a process that was proposed to facilitate establishment or maintenance of chronic HCV infection [190]. In HCV‑related hepatocellular carcinoma (HCV‑HCC), hepatoma cells harboring the HCV JFH‑1 replicon were shown to activate NF‑κB signaling in response to HCVc, leading to upregulation of intestine‑specific homeobox (ISX) and PD‑L1; this axis was associated with promotion of hepatic fibrosis, metabolic dysregulation, and an immunosuppressive TME [191].

MAFLD
In metabolic dysfunction‑associated fatty liver disease (MAFLD) and Metabolic Dysfunction-Associated Steatohepatitis (MASH) liver tissue, innate and adaptive immune cells were reported to be skewed toward proinflammatory phenotypes. Hepatocyte‑expressed innate immune TLR2, TLR4, TLR5, TLR9 were activated, resulting in production of proinflammatory cytokines IL‑1β and IL‑18. Activation of hepatocellular inflammasomes and increased cell death were observed to exacerbate hepatic inflammation, and the ensuing immune activation and proinflammatory cytokine milieu were reported to promote hepatocyte injury and activation of HSCs [192].Within the adaptive immune compartment, CD4 + effector T cells were shown to be activated early and to polarize into distinct helper subsets (Th1, Th2, Th17), while increases in cytotoxic CD8 + T cells and natural killer T cells (NKT cells) were associated with induction of liver injury and progression of MAFLD; experimental depletion of CD8 + T cells was reported to markedly attenuate hepatic inflammation [192]. To limit CD8 + T cell‑mediated damage, KCs were found to upregulate PD‑L1, arginase 1(ARG1), and IL‑10 and to downregulate major histocompatibility complex class II (MHC‑II) molecule expression, thereby acquiring a tolerogenic phenotype that suppressed deleterious immune activation. Notwithstanding, in contrast to normal KC‑mediated intrahepatic tolerance, KCs within the MASH context were reported to be insufficient to fully prevent immune‑mediated hepatic injury; the precise mechanisms underlying this insufficiency remain to be elucidated [193].Changes in NK cell populations were documented in MAFLD livers: increased frequencies of natural killer group 2 member D(NKG2D)(+) and CD69 + NK cells were observed, whereas a subset characterized as Siglec7(-)CD57(+)PD‑1(+)CD56(dim) NK cells exhibited increased PD‑1 expression concomitant with a reduction in peripheral CD56dim NK cell frequency. These findings were interpreted as evidence of NK cell functional impairment, a condition that was associated with increased risks of hepatic fibrosis and tumorigenesis [194]. Such an immunosuppressive milieu was proposed to represent an important factor permitting development of MAFLD‑associated hepatocellular carcinoma (MAFLD‑HCC) without antecedent cirrhosis, although mechanistic details remain incompletely defined.Mechanistically, hepatocellular free fatty acids (FFAs) in MAFLD were reported to upregulate PD‑L1. High FFA levels were shown to trigger reactive oxygen species (ROS) production via NADPH oxidase 4 (NOX4) and mitochondrial dysfunction, thereby inducing zinc finger protein 24 (ZNF24) expression and inhibiting ZNF24 conjugation with small ubiquitin‑like modifier proteins (SUMOylation), which culminated in PD‑L1 upregulation [195]. Notably, basic leucine zipper ATF‑like transcription factor (BATF) expression was increased in hepatocytes from high‑fat diet (HFD)‑induced MAFLD mouse models; BATF was reported to reduce triglyceride (TG) accumulation through suppression of hepatocellular PD‑1 expression and administration of PD‑1 antibody was found to ameliorate high‑fat‑induced hepatocellular steatosis. These observations indicate that PD‑1 and downstream signaling pathways participate in regulation of lipid metabolism in the MAFLD liver [196].
A progressive intrahepatic accumulation of nonconventionally activated PD‑1(+)CD8 + T cells was reported in MASH liver tissue; these CD8 + T cells exhibited aberrant activation, metabolic dysregulation, markedly reduced motility, and a proinflammatory phenotype [197].In Metabolic Dysfunction-Associated Steatohepatitis‑associated hepatocellular carcinoma(MASH‑HCC), immune cells were observed to be enriched in tumor‑adjacent normal tissue but reduced within tumor cores, with pronounced heterogeneity in cellular abundance, phenotype, and function. PD‑L1 expression on tumor cell surfaces was reported to be low and to show limited direct linkage to immune cells. High PD‑1‑expressing CD4 + T cells and CD8 + T cells were described to form local interactions with PD‑L1‑high Treg, tumor‑associated macrophages (TAMs), and monocytic myeloid‑derived suppressor cells (M‑MDSCs) in tumor and peritumoral regions; such interactions were reportedly absent in virus‑associated HCC [198].Subpopulation analysis indicated that a CD44(+)CXCR6(+)PD‑1(+)CD8(+) T cell subset was recruited into the liver and tumors by CXCL16 produced by DCs and TAMs; administration of an anti‑CD122 antibody reduced the number of CXCR6 + PD‑1 + T cells and inhibited HCC growth [199].In a MASH‑HCC rat model, treatment with PD‑1 antibody (programmed cell death protein 1 antibody) increased intratumoral numbers of PD‑1 + CD8 + T cells without inducing tumor regression; prophylactic PD‑1 antibody administration was reported to significantly increase MASH‑HCC incidence, as well as tumor nodule number and size. Concomitant increases in intrahepatic PD‑1 + CXCR6 + CD8 + T cells, TOX(+) T cells, and TNF(+) T cells were observed. These data supported a model in which CD8 + T cells contributed to MASH‑HCC induction via tissue‑damaging, rather than tumor‑suppressive, activities; comparable phenotypic and functional features were reported in patients with MAFLD [200]. A clinical meta‑analysis was reported to indicate shortened overall survival in MASH‑HCC patients receiving anti‑PD‑1/PD‑L1 therapy, suggesting reduced responsiveness to immune checkpoint blockade that may be attributable to MASH‑associated aberrant T cell activation, tissue injury, and impaired immune surveillance [200]. The cellular heterogeneity and complexity of the immune compartment in MAFLD‑HCC were therefore proposed to limit benefit from generalized PD‑1/PD‑L1 blockade, and the development of strategies targeting specific immune cell subsets was recommended to improve therapeutic outcomes.Single nucleotide polymorphisms (SNPs) in the PDCD1 (PD-1 gene) were associated with MAFLD and MAFLD‑HCC. The PDCD1 rs13023138 C > G variant was reported to correlate closely with severe steatosis, advanced fibrosis, and hepatocellular carcinoma, constituting an independent risk factor for comprehensive liver injury [201]. The rs10204525 C > T allele was identified as a predominant allele in 60–70% of Asian populations; presence of this variant was associated with higher PD‑1 expression, and the variant was carried by more than 35% of female patients with MAFLD‑HCC. The PDCD1 rs7421861 A > G polymorphism was reported to yield a G allele that inhibited PD‑1‑mediated immune exhaustion and thereby conferred protection against HCC, whereas the wild‑type A allele was observed more frequently in MAFLD‑HCC patients [202].

Liver failure
Acute liver failure (ALF) and acute‑on‑chronic liver failure (ACLF) were identified as principal forms of liver failure, both of which were associated with profound immune dysregulation and a high incidence of sepsis‑related mortality. During ALF, extensive hepatocyte death and excessive activation of local and systemic inflammatory responses were observed. Defects in both innate and adaptive systemic immunity were reported during ALF, including reduced secretion of proinflammatory cytokines by monocytes, impaired phagocytosis of Escherichia coli, and a phenotypic shift of monocytes/macrophages toward an anti‑inflammatory M2 phenotype, resulting in hyporesponsiveness to microbial stimuli. Phagocytic capacity and cytokine‑secreting ability of PD‑1(-)KCs were reported to be restored, and administration of PD‑1 antibody was shown to facilitate recovery of monocyte function and to increase bacterial clearance by KCs [203, 204].ALF‑induced sepsis was associated with an increase in F4/80(+)PD‑1(+)KCs; KC‑derived PD‑1 signaling was reported to engage PD‑L1 that was highly expressed on LSECs, thereby triggering apoptosis of LSECs, loss of vascular repair capacity, and disruption of sinusoidal barrier function. In this context, LSEC PD‑L1 was observed to shift from a role in maintaining immune tolerance to a driver of endothelial injury during sepsis [205].
ACLF was characterized as a major cause of rapid mortality in patients with chronic liver disease and was defined by dysregulated balance between proinflammatory and anti‑inflammatory responses during the course of chronic disease. Multiple immune cell compartments, including NK cells, CD4(+) T cells, CD8(+) T cells, neutrophils, and myeloid monocytes, were reported to exhibit functional perturbations in ACLF [206]. A subset of plasma cells defined by CD27(+)CD38(+) phenotype was shown to preferentially accumulate in ACLF livers in proportion to disease severity; this subset expressed increased levels of PD‑L2, granzyme B, and IL‑10 secretion [207]. Markers of M2‑polarized macrophages were found to be increased in ACLF liver tissue [208].In hepatitis B virus–related ACLF (HBV‑ACLF) patients, PD‑1 expression on CD8 + T cells was reported to be elevated. Activation of the PD‑1/PD‑L1 axis was associated with reduced glycogen uptake by CD8 + T cells and decreased expression of glucose transporter 1 (GLUT1), hexokinase 2 (HK2), and pyruvate kinase M2 (PKM2), indicating suppression of glycolysis that was proposed to contribute to PD‑1‑mediated CD8 + T cell dysfunction in ACLF [209]. Cirrhosis was identified as a common substrate for ACLF; approximately 40% of hospitalized cirrhotic patients were reported to be at risk for ACLF, and sepsis was likewise frequent and severe in this population. In cirrhotic livers, enrichment of HLA‑DR(+)CD8 + T cells with high PD‑1 expression was observed, and high PD‑1 expression was correlated with poor prognosis [210]. KCs in cirrhotic livers were reported to overexpress PD‑L1 and to exhibit impaired phagocytic function; blockade with anti‑PD‑L1 antibody was shown to aid restoration of KC phagocytic activity and to reduce bacterial dissemination [211].Because of a paucity of suitable animal models, detailed characterization of cellular dysregulation in ACLF remained incomplete; additional preclinical studies were therefore deemed necessary to determine whether blockade of the PD‑1/PD‑L1 axis could confer clinical benefit to patients with ACLF.

AILD
Autoimmune liver disease (AILD) was defined to encompass autoimmune hepatitis (AIH), primary biliary cholangitis (PBC), and the less common primary sclerosing cholangitis (PSC). Although specific pathogenic mechanisms remained incompletely elucidated, PD‑1 deficiency was reported to induce AILD in murine models, and PD‑L1 deficient mice exhibited constitutive accumulation of activated CD8(+) T cells in the liver; PD‑L1 was therefore proposed to participate in clearance of CD8 + T cells. Collectively, these observations supported a role for the PD‑1/PD‑L1 signaling axis in maintenance of immune tolerance in AILD [212].In liver tissue from AIH and PBC patients, PD‑1 and its ligands PD‑L1 and PD‑L2 were reported to be markedly upregulated; the increases were more prominent in AIH, and PD‑L1 elevation exceeded that of PD‑L2. sPD‑1 and sPD‑L1 concentrations were likewise elevated in PBC patients and correlated positively with disease progression [213]. Upregulation of PD‑L1 and PD‑L2 in AIH and PBC was indicated to be minimally dependent on IFN‑γ signaling, implying regulation by additional factors [214].Naturally occurring CD25(+)CD4(+)Treg and PD‑1 signaling were implicated in negative regulation of AIH development. In AIH patients, reductions in Treg frequency and impairment of Treg function were observed; simultaneous depletion of Treg and PD‑1 signaling was reported to precipitate fatal AIH in experimental systems, and PD‑1 deficiency was required for autoreactive CD4 + T cells to induce AIH. Whether primary alterations in PD‑1 or its ligands occur in patients with fulminant AIH remained unclear [212]. A population of autoreactive soluble liver antigen (SLA)‑specific PD‑1(+)CXCR5(−)CCR6(−) CD27(+)CD4(+) follicular helper T cells (Tfh) was described to exhibit a memory phenotype and to secrete IL‑21, thereby supporting B cell differentiation and production of autoantibodies.These SLA‑specific Tfh cells were reported to be sensitive to immunosuppressive therapy yet to retain autoreactive potential under treatment, functioning effectively as peripheral reservoirs for AIH and potentially contributing to disease relapse after therapy [215].
In PBC patients, PD‑L1 expression on intrahepatic cholangiocytes was reported to be reduced, and cytotoxic T lymphocytes (CTLs) within hepatic tissue were observed to upregulate T‑box expressed in T cells (T‑bet), a transcription factor associated with Th1 polarization, resulting in suppression of PD‑1 expression; experimental silencing of the PD‑1/PD‑L1 signaling axis was demonstrated to enhance CTL‑mediated injury to cholangiocytes [216]. Intrahepatic subsets of T cells in PBC were reported to secrete IL‑17, which induced PD‑L1 upregulation on cholangiocytes and thereby suppressed CD8(+) T cell‑mediated inflammatory bile duct damage [217].A relatively high proportion of CD69(+)CD62L(-)PD‑1(+)CXCR6(+)CD8 + Trm was identified among hepatic immune populations; these cells displayed elevated expression of activation and cytotoxicity‑related markers and were capable of inducing cholangiocyte apoptosis. Targeting of PD‑1 by chimeric antigen receptor T cells (CAR‑T cells) directed against PD‑1 was reported to selectively deplete hepatic CD8 + Trm and to ameliorate PBC in experimental models [218].Subsets of Tfh that preferentially promote antibody responses, specifically CXCR5(+)PD‑1(+)CD4(+)Tfh and CCR7(low)CXCR5(+)PD‑1(+) CD4 + Tfh, were shown to be significantly increased in PBC liver tissue. These Tfh populations exhibited enhanced expression of the co‑stimulatory molecule OX40 and other inducible T cell co‑stimulatory factors, correlated with titers of anti‑mitochondrial M2 antibodies and IgM, and were enriched in patients with cirrhosis [219]. Regulation of the PD‑1/PD‑L1 axis in PBC was reported to be incompletely understood. Expression of the chloride/bicarbonate anion exchanger 2 (AE2) was found to be reduced in the livers and lymphocytes of PBC patients. In Ae2a, b(−/−) mice progressive intrahepatic accumulation of CD8(+) tumor‑infiltrating lymphocytes (TILs) was observed with aging; epigenetic downregulation of PD‑1 and reduced apoptosis were implicated in clonal expansion of CD8 + T cells and development of PBC‑like pathology [220]. These findings indicated that additional molecular pathways contribute to PD‑1/PD‑L1 regulation in PBC.The use of ICIs in patients with malignancy and concurrent AILD remained a subject of debate. A clinical study reported that, during the first year of ICI therapy in patients with cancer and comorbid AILD, no marked deterioration of liver function was observed and only approximately 13% of patients developed hepatic irAEs, suggesting that AILD should not be considered an absolute contraindication to ICI therapy in all cases [221].
Therapeutic modulation of the PD‑1/PD‑L1 signaling axis in AILD was proposed to hold substantial potential. Agonistic PD‑1 antibodies were reported to represent a novel therapeutic avenue for autoimmune diseases by reinforcing inhibitory signaling [222]. Persistent issues that contributed to high relapse rates in AILD included autoreactive Tfh and Trm populations that sustain antibody production and tissue autoreactivity; strategies combining PD‑1‑targeted CAR‑T cells to selectively eliminate pathogenic effector subsets and immunomodulatory mesenchymal stromal cells(PI‑MSCs) overexpressing PD‑L1 and intercellular adhesion molecule‑1(ICAM‑1) to enhance Treg‑mediated immunoregulation were reported to enable specific clearance of pathogenic cells and augmentation of Treg function, thereby inducing sustained drug‑free remission in experimental settings [223].

Conclusions and future perspectives
The PD-1/PD-L1 signaling axis in hepatic malignancies and chronic liver diseases is systematically summarized herein. As an important immunosuppressive pathway, the PD-1/PD-L1 axis has been shown to play a critical role in the maintenance of physiological hepatic immune tolerance. Regulatory mechanisms of the PD-1/PD-L1 axis have been investigated most extensively in HCC, and a multilayered synthesis is provided from the perspectives of multiple intratumoral cell types and of signals derived from the TME, intracellular signaling cascades, gene-expression control and stromal modulation. Investigation of PD-1/PD-L1 regulatory mechanisms has been demonstrated to facilitate identification of novel therapeutic targets, and, concomitant with advances in genetic engineering, genetically modified oncolytic viruses and other engineered vectors have been proposed as potential delivery platforms targeting the PD-1/PD-L1 axis [224].Nevertheless, substantial limitations in current knowledge remain. The immune microenvironments of chronic liver diseases driven by different etiologies—including HBV, HCV, alcohol-related injury and metabolic factors—are highly heterogeneous, and the etiological-specific effects on PD-1/PD-L1 regulation and the underlying mechanisms remain largely undefined. The etiology-specific roles of the PD-1/PD-L1 axis across distinct chronic liver diseases have therefore been insufficiently characterized. In addition, the depth and specificity of mechanistic studies are presently inadequate: the PD-1/PD-L1 axis has frequently been treated as a uniform entity, with insufficient attention given to the cellular origins and functional heterogeneity of PD-1 and PD-L1 within the complex hepatic microenvironment. The precise mechanisms by which PD-1 expressed on different cell lineages contributes to immune suppression, viral persistence or fibrogenic activation, and the expression patterns, regulatory controls and pathological significance of distinct PD-L1 isoforms and subcellular pools in liver disease, remain to be elucidated through targeted, high-resolution investigations.
Although combination regimens incorporating ICIs have produced significant survival gains in advanced HCC, substantial limitations to clinical application persist. Existing regimens are associated with overall objective response rates of approximately 30% and with both primary and acquired resistance, phenomena that are attributed in part to the pronounced heterogeneity of the TME. Concurrently, as ICIs are moved earlier in therapeutic algorithms and as diverse combination strategies emerge, optimization of treatment sequencing—for example, the optimal integration and timing of ICIs with TACE and with targeted systemic agents—remains an important area for investigation.The absence of robust, predictive biomarkers constitutes a central barrier to individualized therapy, such that a subset of patients is exposed to the risk of irAEs without deriving clinical benefit. In contrast to the established role of ICIs in HCC, the potential clinical benefit of ICIs in various chronic liver diseases is highly conditional and remains exploratory; therapeutic utility and associated risks are dependent upon disease etiology and the prevailing immune state.
In chronic infection with HBV or HCV, virus-specific T lymphocytes frequently exhibit an exhausted phenotype with high PD-1 expression. Blockade of the PD-1/PD-L1 pathway by ICIs can reinvigorate these cells and restore their capacity to clear infected hepatocytes, representing a potential pathway toward a functional cure. The principal hazard of this approach is the induction of uncontrolled, widespread immune activation, which may precipitate fulminant, potentially fatal hepatitis.
MAFLD is characterized by complex dysregulation of innate and adaptive immune responses. In MASH, infiltration by exhausted CD8-positive T lymphocytes is frequently observed, yet the PD-1/PD-L1 axis may exert a complex, potentially protective role in limiting MASH-associated inflammation and fibrogenesis. Current paradigms do not support routine use of ICIs in MASH; instead, mechanistic elucidation of PD-1/PD-L1 function in this context is advocated to improve understanding of immunometabolic regulation and to identify specific immune subsets that might serve as precise targets for future immune-modulatory interventions. For autoimmune liver diseases, including AIH and PBC, use of ICIs is presently regarded as relatively contraindicated because release of immune checkpoints is likely to precipitate or markedly exacerbate autoimmune liver injury, with potentially catastrophic consequences. Conversely, the development of PD-1 agonistic antibodies has raised the prospect that targeted activation of the PD-1/PD-L1 axis could represent a therapeutic strategy for autoimmune liver disorders, meriting further investigation. Patients with liver failure are often rendered immunoparalyzed and thereby highly susceptible to secondary infections, which represent a leading cause of mortality. It has been hypothesized that appropriately timed, targeted modulation of immune checkpoints might reverse immunoparalysis and improve anti-infective defenses. However, this concept remains largely theoretical; the intervention would risk disrupting the fragile balance between immune tolerance and hyperinflammation and therefore requires validation through rigorous preclinical and clinical research.
Overall, the functional role of the PD-1/PD-L1 axis is observed throughout both physiological and pathological processes of the liver. Investigation of the PD-1/PD-L1 axis is expected to elucidate the immunological heterogeneity present across different chronic liver diseases and to inform the development of therapeutic strategies targeting this axis.