PD-L1, or programmed death-ligand 1, is a cell-surface protein that helps regulate immune responses by binding principally to programmed cell death protein 1 (PD-1), an inhibitory receptor on activated immune cells. In humans, PD-L1 is encoded by the CD274 gene. Its interaction with PD-1 forms an immune checkpoint: a mechanism that limits immune activation, protects tissues from excessive inflammation, and contributes to tolerance of the body’s own components. Despite its name, PD-L1 is not primarily a molecule that directly causes programmed cell death.
The same regulatory pathway can be exploited by cancers to reduce immune attack. Antibodies that block PD-L1 or PD-1 have consequently become important treatments for several malignancies. PD-L1 expression is also measured in tumor samples to help select treatment in particular clinical settings, although it is an imperfect and context-dependent biomarker. Understanding PD-L1 requires distinguishing its normal physiological functions, its molecular regulation, its role in tumor–immune interactions, and the different ways its expression is assessed.
Discovery, nomenclature, and molecular structure
PD-L1 emerged from research on the B7 family of immune-regulatory molecules. In 1999, a group led by Lieping Chen described B7-H1 as a B7-family protein capable of influencing T-cell responses. In 2000, Gordon Freeman and colleagues identified this molecule as a ligand for PD-1 and demonstrated inhibition of lymphocyte activation through that interaction. The name PD-L1 reflects this receptor–ligand relationship. Early experiments produced apparently different functional effects under different stimulation conditions; the inhibitory PD-1 pathway subsequently became central to its biological interpretation.
Several names remain in use because they refer to different naming systems. CD274 is the official human gene symbol, CD274 is also a designation for the protein, and B7-H1 is its historical B7-family name. PD-L1 should not be confused with PD-L2, a distinct PD-1 ligand encoded by PDCD1LG2. Nor should the ligand be confused with PD-1 itself, which is encoded by PDCD1. This distinction matters clinically because blocking the receptor and blocking one of its ligands do not interrupt exactly the same interactions.
The human CD274 gene lies on chromosome 9 at 9p24.1, near PDCD1LG2 and JAK2. The canonical human PD-L1 protein contains 290 amino acids. It is a type I transmembrane glycoprotein: its amino terminus faces the extracellular space, a single membrane-spanning segment anchors it in the cell membrane, and a short carboxy-terminal tail extends into the cytoplasm. The extracellular portion contains an immunoglobulin variable-like domain and an immunoglobulin constant-like domain. The variable-like domain provides the principal surface involved in PD-1 binding.
The unglycosylated polypeptide has a predicted mass of approximately 33 kilodaltons, but mature PD-L1 commonly migrates at higher apparent molecular weights in laboratory assays because of glycosylation. Carbohydrate groups influence its folding, stability, trafficking, and recognition by some antibodies. Consequently, an apparent difference in PD-L1 abundance on a protein blot can reflect changes in processing as well as changes in protein production.
PD-L1’s intracellular tail is short and lacks the well-characterized inhibitory signaling motifs found in PD-1. The best-established checkpoint mechanism therefore consists of PD-L1 engaging PD-1 and initiating inhibitory signaling inside the receptor-bearing cell. Studies also describe PD-L1-associated functions within the cells that express it, but these depend on experimental context and are not equivalent to the canonical PD-1 signaling pathway. Structural work using X-ray crystallography has helped define the receptor-binding interface and explain how therapeutic antibodies obstruct it.
Expression and molecular regulation
PD-L1 expression varies substantially among cell types and physiological conditions. It occurs on antigen-presenting cells, including dendritic cells, macrophages, and B cells, and can also be induced on activated lymphocytes. Outside the hematopoietic system, expression occurs in epithelial, endothelial, and other tissue cells. Its distribution is therefore broader than that of a molecule restricted to professional antigen-presenting cells. Both baseline expression and inducibility are important: a tissue with little detectable PD-L1 at rest may express it strongly during an immune response.
A major inducer is interferon-γ, a cytokine produced by activated T cells and natural killer cells. Binding of interferon-γ to its receptor activates JAK1 and JAK2, followed by STAT1 and downstream transcriptional regulators such as IRF1. These events increase CD274 transcription. Type I interferons can also promote expression through related signaling mechanisms. This establishes a local negative-feedback circuit: immune activation produces inflammatory signals, those signals induce PD-L1, and PD-L1 limits further activation of PD-1-expressing cells.
Other inflammatory and tissue signals contribute in a context-dependent manner. Toll-like receptor stimulation can induce PD-L1 in myeloid cells, while hypoxia can increase expression through pathways involving hypoxia-inducible factors. Oncogenic signaling involving pathways such as PI3K–AKT and MAPK can also support expression in some tumors. These mechanisms are not interchangeable. A tumor expressing PD-L1 because of an active local immune response may have a different immune environment from one whose expression is driven predominantly by a tumor-intrinsic genetic alteration.
Regulation also occurs after transcription. The stability of CD274 messenger RNA is influenced by its untranslated regions and RNA-binding proteins. Structural alterations that disrupt the gene’s 3′ untranslated region can increase expression by removing regulatory constraints. MicroRNAs can affect transcript abundance or translation, while protein modification and trafficking determine how much PD-L1 reaches and remains on the surface. The ubiquitin–proteasome system participates in PD-L1 turnover, and proteins such as CMTM6 help maintain surface PD-L1 by protecting it from degradation.
These regulatory layers explain why measurements of the gene, transcript, and surface protein are related but not equivalent. Increased CD274 copy number does not specify the exact amount of accessible membrane protein; abundant RNA does not guarantee sustained surface expression; and a single tissue sample may not capture induction that occurs later during treatment. In experimental studies, the timing of cytokine exposure, the cellular compartment examined, and the method used to detect PD-L1 can therefore materially change the result. PD-L1 is best understood as a dynamically regulated component of an immune response rather than a fixed attribute of a cell.
Receptor interactions and inhibitory signaling
The principal inhibitory action of PD-L1 begins when it binds PD-1 on another cell. PD-1 is induced on activated T cells and is also found on other immune populations. Productive signaling depends not merely on the presence of receptor and ligand but on their spatial relationship to activating receptors at the cell–cell interface. During antigen recognition, PD-1 can accumulate near signaling complexes associated with the T-cell receptor, allowing an extracellular binding event to modify intracellular activation.
The cytoplasmic portion of PD-1 contains an immunoreceptor tyrosine-based inhibitory motif and an immunoreceptor tyrosine-based switch motif. Following receptor engagement in an activating environment, tyrosines in these motifs can become phosphorylated and recruit phosphatases, particularly SHP2. These enzymes remove phosphate groups from components of activation pathways. Biochemical studies have identified CD28, a major costimulatory receptor, as an important target of PD-1-associated inhibition, although the balance of affected substrates depends on the experimental system and cellular context.
The consequences include reduced signaling through pathways that support proliferation, cytokine production, nutrient use, and survival. PD-L1–PD-1 engagement can decrease production of interleukin-2, restrain cell-cycle progression, and reduce effector activity. It does not invariably switch a T cell completely off, nor does it necessarily kill the cell. Its effect is better described as raising the threshold for activation and reducing the strength or persistence of an ongoing response. Antigen strength, costimulation, receptor abundance, and the cell’s differentiation state all influence the outcome.
PD-L1 also interacts with CD80, another B7-family molecule that participates in T-cell costimulation. An important configuration is a cis interaction, in which PD-L1 and CD80 associate on the same cell membrane. Such association can reduce the availability of PD-L1 for PD-1 binding and modify CD80 interactions with other immune receptors. Experiments have shown that this arrangement can preserve CD80-mediated CD28 costimulation while limiting some inhibitory interactions, although its overall effect depends on the molecular and cellular setting. Thus, PD-L1 abundance alone does not fully specify its functional availability.
The distinction between PD-L1 and PD-L2 adds another layer. Both can engage PD-1, but their tissue distribution and regulation differ. An antibody directed against PD-1 blocks receptor engagement by both ligands, whereas a PD-L1-directed antibody generally leaves PD-1–PD-L2 binding intact. Some PD-L1 antibodies also interfere with PD-L1–CD80 binding. These differences establish that anti-PD-1 and anti-PD-L1 drugs are mechanistically related but not molecularly identical. They do not, by themselves, establish a universal clinical advantage for one class over the other.
Physiological tolerance and responses to infection
PD-L1 contributes to peripheral immune tolerance, the mechanisms that restrain potentially damaging immune responses after lymphocytes have developed. Unlike central tolerance, which eliminates or modifies many self-reactive lymphocytes during development, peripheral mechanisms operate in mature tissues and secondary lymphoid organs. PD-L1 expressed by antigen-presenting cells or tissue cells can limit responses by PD-1-expressing lymphocytes encountering antigen outside a strongly activating setting. This reduces the likelihood that ordinary tissue antigen recognition will escalate into sustained injury.
Its physiological role is supported by complementary forms of evidence. Mouse studies disrupting components of the PD-1 pathway have demonstrated susceptibility to autoimmune disease, with the phenotype depending on genetic background and experimental conditions. In humans, inflammatory adverse effects following therapeutic blockade provide evidence that this pathway normally restrains immune activity in multiple organs. These observations establish a protective role for the pathway, but they do not mean that every PD-L1-expressing tissue is equally dependent on it or that PD-L1 alone determines tolerance.
PD-L1 is also part of the regulation of ongoing inflammation. Activated immune cells can damage infected or abnormal cells, but the same effector molecules can injure neighboring tissue. Interferon-driven PD-L1 induction creates a feedback mechanism that can reduce this collateral injury. In organs exposed repeatedly to environmental antigens or microbial products, such regulation helps balance defense with tissue integrity. Placental expression and animal studies also implicate the PD-1 pathway in maternal–fetal immune regulation, although reproductive tolerance depends on multiple interacting mechanisms.
During persistent infection, PD-L1 may have a dual effect. Restraining T-cell activity can limit immunopathology, yet it can also reduce the ability to clear an infectious agent. Experiments in chronic lymphocytic choriomeningitis virus infection in mice helped establish the connection between PD-1 signaling and T-cell exhaustion, a state associated with prolonged antigen exposure and altered effector function. Blocking the pathway in that model could restore aspects of antiviral T-cell activity. Human chronic infections also show associations between sustained antigen exposure, PD-1 expression, and impaired lymphocyte function.
Exhaustion is not synonymous with PD-1 expression. Recently activated, functional T cells can express PD-1, and exhausted cells possess broader transcriptional and epigenetic changes. PD-L1 blockade therefore does not simply reverse every feature of chronic stimulation. Moreover, improving pathogen-directed activity and reducing tissue damage are separate biological outcomes that may conflict. The role of PD-L1 in infection depends on the pathogen, disease stage, affected organ, and host response; evidence from a particular infection model cannot be transferred directly to all infectious diseases.
Tumor immune evasion and the tumor microenvironment
In cancer, PD-L1 can inhibit immune cells that recognize tumor-associated or mutation-derived antigens. Cytotoxic T cells may enter a tumor, recognize peptide–major histocompatibility complex molecules, and produce interferon-γ. Tumor cells and neighboring stromal or immune cells can respond by increasing PD-L1. Engagement of PD-1 then reduces the activity of nearby T cells. This arrangement is often called adaptive immune resistance because an antitumor response itself induces a mechanism that limits that response.
PD-L1 can also arise through tumor-intrinsic mechanisms. Amplification of chromosome region 9p24.1 can increase the dosage of CD274 and PDCD1LG2. This region may also include JAK2, potentially reinforcing signaling that promotes ligand expression. Such alterations are particularly relevant in classical Hodgkin lymphoma and primary mediastinal large B-cell lymphoma. In other cancers, oncogenic signaling, viral proteins, or changes in transcript regulation can contribute. These examples show why identical-looking PD-L1 staining patterns need not have identical molecular causes.
The relevant ligand is not confined to malignant cells. Tumor-associated macrophages, dendritic cells, and other nonmalignant cells can express substantial PD-L1 within the tumor microenvironment. These cells may influence T cells inside the tumor, at its invasive margin, or in lymphoid tissues where antitumor responses are initiated and maintained. Experimental models have demonstrated contributions from both tumor-derived and host-derived PD-L1, with their relative importance varying by model. Clinical scoring systems that include immune-cell staining reflect part of this biological complexity.
Spatial organization matters. A tumor with PD-L1 concentrated where lymphocytes contact malignant cells differs from a tumor with broad expression but few infiltrating T cells. Likewise, abundant PD-L1 cannot compensate for the absence of recognizable antigens, defective antigen presentation, or barriers that prevent lymphocytes from entering tumor tissue. The checkpoint operates within a larger sequence involving antigen release, antigen presentation, lymphocyte priming, trafficking, recognition, and killing. Failure at any of these stages can limit the effect of blocking PD-L1.
PD-L1 is therefore neither a universal marker of aggressive cancer nor a simple measure of immune suppression. In some settings, its expression is associated with an ongoing immune response; in others, it accompanies tumor-intrinsic mechanisms of escape. Reported relationships with prognosis differ by cancer type, disease stage, cellular compartment, and assay. A prognostic association—an association with outcome regardless of treatment—must also be distinguished from a predictive association, which concerns the likelihood of benefit from a particular therapy. Confusing these concepts can lead to misleading interpretations of tumor-expression studies.
Clinical measurement and interpretation
The principal clinical method for assessing PD-L1 is immunohistochemistry on formalin-fixed, paraffin-embedded tumor tissue. An antibody detects PD-L1, and a pathologist evaluates staining according to a defined assay and scoring system. Commonly used antibody clones include 22C3, 28-8, SP142, and SP263. An assay comprises more than the antibody clone: staining platform, tissue handling, detection chemistry, interpretation rules, and clinical validation all contribute. A result is clinically meaningful only in relation to the relevant tumor type and treatment setting.
One widely used metric is the tumor proportion score, or TPS. It is the percentage of viable tumor cells showing qualifying partial or complete membrane staining:
For particular indications in non-small-cell lung cancer, thresholds such as 1% and 50% have been used to define clinically relevant expression groups. These thresholds are not universal biological boundaries. Their significance comes from the patient populations, assays, drugs, and outcomes evaluated in the supporting clinical studies.
Another metric is the combined positive score, or CPS:
CPS incorporates selected immune cells as well as tumor cells and is conventionally reported with a maximum value of 100. It has been used in several tumor-specific treatment settings, including certain head and neck and upper gastrointestinal cancers. Other assays use immune-cell scoring based on the area occupied by stained immune cells rather than the same cell-counting approach. Scores produced by different systems are therefore not directly interchangeable.
Preanalytical and sampling factors introduce uncertainty. Fixation, specimen age, decalcification, and limited tumor content can affect assessment. PD-L1 may vary between different areas of one tumor, between a primary tumor and metastases, or between samples obtained before and after treatment. Small biopsies may miss a positive region. Interpretation also requires distinguishing malignant cells from macrophages and other immune cells, a task that can be difficult in heavily inflamed or poorly differentiated tumors. Assay-comparison studies have found substantial similarities among some tumor-cell assays, but this does not make all assays interchangeable for every purpose.
PD-L1 is an imperfect predictive biomarker. Some PD-L1-positive tumors do not respond to checkpoint blockade, while some tumors classified as negative do respond. Treatment decisions in particular settings may also incorporate tumor histology, stage, molecular alterations, microsatellite instability, mismatch-repair status, or tumor mutational burden. These measurements address different biological properties and should not be treated as synonyms for PD-L1 expression. Blood-based soluble PD-L1 measurements and imaging approaches remain distinct from established tissue-based scoring and do not generally substitute for a validated tissue assay.
Therapeutic blockade and clinical development
Therapeutic PD-L1 blockade uses an antibody to prevent inhibitory ligand–receptor engagement. The major approved agents in this class include atezolizumab, durvalumab, and avelumab. They are monoclonal antibodies directed against PD-L1, rather than PD-1. Their structures differ, including features of the Fc region that influence interactions with other immune components. Avelumab retains an Fc region capable of mediating antibody-dependent cellular cytotoxicity in laboratory systems, whereas atezolizumab and durvalumab have Fc modifications intended to reduce certain effector functions. The contribution of these differences to clinical outcomes is context-dependent.
Atezolizumab received its first United States approval in 2016, and avelumab and durvalumab received initial United States approvals in 2017. Subsequent development established roles for PD-L1 blockade in several malignancies and treatment settings. The range of indications has changed over time, with some approvals added and others withdrawn as confirmatory evidence developed. A drug’s established use therefore depends on cancer type, disease stage, prior treatment, combination regimen, and jurisdiction, rather than simply on whether a tumor expresses PD-L1.
Several trials illustrate distinct applications. The PACIFIC trial established durvalumab as consolidation therapy after concurrent chemoradiotherapy for selected patients with unresectable stage III non-small-cell lung cancer without disease progression. IMpower133 demonstrated a survival benefit from adding atezolizumab to carboplatin and etoposide in previously untreated extensive-stage small-cell lung cancer. JAVELIN Bladder 100 demonstrated improved survival with avelumab maintenance in patients with advanced urothelial cancer whose disease had not progressed after initial platinum-based chemotherapy. These are different therapeutic strategies: consolidation, first-line combination treatment, and maintenance treatment.
PD-L1 blockade does not directly kill every PD-L1-expressing tumor cell. Its central purpose is to alter immune regulation so that antitumor responses can develop or function more effectively. Responses may be durable, but benefit is not universal. Clinical outcomes depend on the tumor’s biological features, the available immune response, and the treatment context. Expression thresholds used for one drug–cancer combination cannot automatically be applied to another, and some effective regimens do not require a positive PD-L1 test.
Because the pathway contributes to normal tolerance, blocking it can produce immune-mediated adverse effects. Recognized manifestations include pneumonitis, colitis, hepatitis, thyroid dysfunction, other endocrine disorders, nephritis, and skin inflammation; rarer events can affect the heart, nervous system, or other organs. These effects differ mechanistically from many conventional chemotherapy toxicities, although combination regimens can produce both. Clinical trials and prescribing information characterize their occurrence and severity. Tumor PD-L1 expression is not a reliable general measure of an individual patient’s risk of immune-related toxicity.
Primary and acquired resistance
Resistance to PD-L1-directed therapy is commonly divided into primary resistance, in which meaningful benefit is absent from the outset, and acquired resistance, in which disease progresses after an initial response or period of control. These categories describe clinical patterns rather than single molecular mechanisms. Tumors can resist treatment despite abundant PD-L1 because the ligand–receptor interaction is only one part of the antitumor immune response. Conversely, treatment can fail even when an antibody successfully occupies its molecular target.
One major limitation is inadequate tumor recognition. T cells require appropriate antigens presented by major histocompatibility complex molecules. Loss of beta-2-microglobulin, defects in antigen processing, or reduced expression of presentation machinery can make malignant cells less visible to cytotoxic T cells. Changes in the antigenic composition of a tumor during treatment can also favor escape. A tumor with few effective immune targets cannot necessarily become immunologically vulnerable simply by removing PD-L1-mediated inhibition.
Defective inflammatory signaling provides another route. Alterations affecting interferon-response pathways, including JAK1 or JAK2 in some documented settings, can reduce tumor responsiveness to interferon-γ. The consequences extend beyond PD-L1 expression: interferon signaling also affects antigen presentation and other components of immune recognition. Thus, loss of an interferon response may simultaneously decrease PD-L1 and impair processes needed for successful immunotherapy. A low PD-L1 result in this context does not imply absence of immune escape.
The surrounding tissue can create additional barriers. Abnormal vasculature, dense stromal architecture, suppressive myeloid cells, and regulatory T cells can restrict T-cell entry or function. Local nutrient competition and suppressive metabolites can impair effector activity even after checkpoint blockade. Other inhibitory pathways may remain active, including those involving LAG-3, TIM-3, and TIGIT. Their presence does not establish that each is the dominant cause of resistance, but it illustrates why a single checkpoint is rarely the only regulatory influence in a tumor.
The state of the T-cell population is also important. Studies have identified progenitor-like exhausted T cells with greater capacity to proliferate after PD-1-pathway blockade than more terminally differentiated exhausted populations. Effective responses can involve expansion or recruitment of T-cell clones rather than simple reactivation of every lymphocyte already present. These findings connect treatment outcome to the availability of responsive immune-cell populations and to lymphoid as well as intratumoral environments.
Resistance research has motivated combinations with chemotherapy, radiotherapy, antiangiogenic agents, and other immunotherapies. Their rationale is to address complementary barriers, such as antigen release, vascular access, or separate inhibitory pathways. A plausible biological rationale alone does not establish clinical benefit: efficacy and toxicity must be evaluated for each regimen in appropriately designed randomized controlled trials.
Experimental models and unresolved questions
PD-L1 research uses complementary methods because no single measurement captures its abundance, location, binding state, and biological effect. Flow cytometry measures surface expression on defined cell populations and can distinguish tumor cells from immune subsets in dissociated samples. Tissue imaging preserves spatial relationships but samples fewer molecular features. RNA sequencing measures transcriptional programs, while protein assays assess products that may be regulated independently of RNA abundance. Agreement among these methods strengthens an interpretation; disagreement can identify an additional regulatory layer rather than a technical failure.
Genetic experiments help distinguish correlation from function. Targeted disruption of CD274 in tumor cells can be compared with disruption in host cells to identify the source of functionally important ligand. CRISPR screens have identified regulators of PD-L1 expression and stability, including pathways affecting interferon responses and protein trafficking. Such experiments require attention to model design: a cell-culture assay may reveal regulation of surface PD-L1 but cannot reproduce lymphocyte trafficking, tissue architecture, or a complete host immune response.
Mouse tumor models, organoids, and human tumor–immune cocultures address different aspects of this problem. Transplanted mouse tumors permit controlled comparisons but may not reproduce the slow evolution and heterogeneity of human cancer. Humanized mouse models introduce human immune components yet do not fully reconstruct human immunity. Organoid and coculture systems can preserve selected tumor properties while lacking normal vasculature or systemic immune compartments. Results therefore gain significance through replication across models and connection to human tissue or clinical observations.
Forms of PD-L1 outside the conventional cell surface remain active research subjects. Soluble forms can arise through processes including alternative transcripts or proteolytic release. PD-L1 can also be carried by extracellular vesicles, including small vesicles often described as exosomes. Experimental studies show that vesicle-associated PD-L1 can interact with immune cells, but its concentration, origin, and functional importance vary. Assays for circulating PD-L1 species differ in what they detect, which complicates comparisons and limits the interpretation of a single blood measurement.
Another unresolved area is the significance of PD-L1-associated signaling within tumor cells. Reported effects include changes in survival pathways, stress responses, and treatment sensitivity, but findings depend on cell type, protein localization, and experimental manipulation. These proposed cell-intrinsic functions must be separated from the established effect of PD-L1 on PD-1-bearing immune cells.
Spatial profiling, single-cell analysis, and repeated sampling during therapy are being used to relate PD-L1 to particular cellular neighborhoods and changing immune states. A central question is whether these approaches can produce reproducible biomarkers that improve on a single pretreatment staining score. Any such biomarker must demonstrate not only biological association but also analytical reliability and treatment-specific clinical validity.