Tumor necrosis factor (TNF) is a cytokine that helps coordinate inflammation, antimicrobial defense, and the responses of tissues to injury. Produced especially by activated macrophages, but also by numerous other immune and tissue cells, it acts through two cell-surface receptors. Depending on the receptor, target cell, and surrounding signals, TNF can promote inflammatory gene expression, support cell survival, alter tissue organization, or initiate regulated cell death. Its name reflects its discovery through the destruction of experimental tumors, not its predominant function in normal physiology.
TNF occupies an important position between protective immunity and inflammatory disease. Local, controlled production helps recruit immune cells and contain infections; excessive or persistent activity contributes to tissue damage in disorders such as rheumatoid arthritis and inflammatory bowel disease. Drugs that inhibit TNF became a major class of biological therapies, while also demonstrating that the same molecule is required for important forms of host defense. Understanding TNF therefore requires attention to its molecular forms, receptor-specific signaling, and the cellular and anatomical setting in which it acts.
Discovery and changing terminology
The origins of TNF research lie in observations that bacterial products could cause hemorrhage and necrosis in experimental tumors. These findings developed against a longer history of attempts to provoke antitumor responses through infection, including William Coley’s use of bacterial preparations in the late nineteenth century. Those early treatments involved complex and variable immune stimulation rather than an identified molecule. The later experimental question was whether a host-derived factor could transmit part of the tumor-damaging effect produced by bacterial endotoxin.
In 1975, Elizabeth Carswell, Lloyd Old, and colleagues reported a factor in the serum of mice that had been primed with bacillus Calmette–Guérin and subsequently challenged with endotoxin. The serum caused necrosis of certain transplanted tumors, leading to the designation “tumor necrosis factor.” The activity was attributed to a host product rather than to direct poisoning of tumor cells by the bacterial material. This distinction helped establish the idea that activated immune cells release soluble mediators capable of producing powerful effects at distant sites.
A separate line of investigation concerned wasting during chronic infection. Anthony Cerami, Bruce Beutler, and colleagues studied a macrophage product called cachectin, which suppressed lipoprotein lipase activity and was implicated in altered lipid metabolism. During the 1980s, biochemical purification and molecular characterization established that cachectin and TNF were the same molecule. This connected experimental tumor necrosis to systemic changes in metabolism, fever, inflammation, and wasting. It also made clear that a name based on one experimental assay could obscure a much broader biological role.
Cloning of human TNF complementary DNA in 1984 enabled production of recombinant protein and detailed analysis of its structure and function. Researchers could now study a defined molecule rather than mixtures of inflammatory mediators. Subsequent work identified its receptors, mapped intracellular signaling components, and showed that TNF did not simply kill cells indiscriminately. Many cells survived TNF exposure and instead increased expression of inflammatory and protective genes.
Terminology changed as related molecules were characterized. The older name TNF-alpha distinguished this protein from “TNF-beta,” now generally called lymphotoxin-alpha. Contemporary molecular nomenclature usually uses TNF for the former and LTA for the latter. TNF also belongs to the larger TNF superfamily, which includes ligands involved in lymphocyte regulation, tissue organization, and cell death. Family membership indicates structural and evolutionary relationships, but it does not mean that all family members act through TNF receptors or reproduce TNF’s inflammatory effects.
Gene, protein structure, and molecular forms
The human TNF gene is located on chromosome 6p21.3, within the class III region of the major histocompatibility complex, near the genes encoding lymphotoxin-alpha and lymphotoxin-beta. The mature transcript encodes a precursor of 233 amino acids. This precursor is inserted into the cell membrane as a type II membrane protein: its amino-terminal region lies inside the cell, a single segment crosses the membrane, and its carboxy-terminal cytokine domain projects outside.
Membrane TNF is biologically active rather than merely an inactive storage form. It can engage receptors on adjacent cells, allowing close-contact communication between a TNF-producing cell and a responding cell. Proteolytic cleavage releases a soluble extracellular fragment of 157 amino acids, with a molecular mass of approximately 17 kilodaltons per subunit. The principal enzyme responsible is ADAM17, also historically called TNF-alpha-converting enzyme or TACE. Because ADAM17 also sheds other cell-surface proteins, changes in its activity can affect several signaling systems simultaneously.
Both membrane and soluble TNF assemble into homotrimers. Three identical subunits associate to produce the characteristic active ligand, with receptor-binding surfaces formed along the interfaces between neighboring subunits. Each cytokine domain adopts a beta-sandwich fold characteristic of many TNF-family ligands. X-ray crystallography helped establish this architecture and explain how a trimer can organize receptor molecules into signaling-competent assemblies. The biologically active soluble trimer has a mass of roughly 51 kilodaltons, distinct from the mass of an individual subunit.
The distinction between soluble and membrane TNF has functional consequences. Soluble TNF efficiently activates TNF receptor 1, whereas membrane TNF is particularly effective at activating TNF receptor 2. These are preferences, not absolute binding exclusions: receptor activation depends on ligand presentation, receptor abundance, and the degree of receptor clustering. A soluble ligand in bulk fluid and a ligand constrained within a cell membrane impose different geometrical arrangements on the receptors they contact.
Membrane TNF can also participate in “reverse signaling.” In this situation, engagement of membrane TNF by a receptor or an anti-TNF agent produces responses within the TNF-bearing cell itself. This adds another layer to an interaction often represented simply as secretion followed by receptor binding. In tissues, TNF biology encompasses soluble communication, direct cell-to-cell contact, proteolytic release, and signaling in both participating cells. Consequently, interventions that neutralize soluble TNF alone need not have the same effects as agents that also bind membrane TNF.
Cellular sources and control of production
Activated monocytes and macrophages are major sources of TNF during acute inflammatory responses. Recognition of microbial products can trigger rapid transcription of TNF and release of the protein within hours. One widely studied stimulus is lipopolysaccharide, a component of the outer membrane of Gram-negative bacteria. Its recognition through Toll-like receptor 4 and associated proteins activates intracellular pathways that induce TNF together with other inflammatory mediators. TNF can therefore appear early in the response, before antigen-specific immunity has fully developed.
Production is not restricted to macrophages. Activated T cells, natural killer cells, dendritic cells, mast cells, and other leukocytes can produce TNF under appropriate conditions. Stromal cells and certain epithelial or other tissue cells also contribute in particular settings. The dominant source depends on the organ, stimulus, and stage of the response. A short burst from circulating or tissue myeloid cells has different consequences from persistent production by immune cells within an inflamed joint or intestinal lesion.
Control occurs at several stages of gene expression. Activation of transcription factors, including NF-κB and members of the AP-1 family, increases transcription of the TNF gene. The resulting messenger RNA contains AU-rich sequences in its 3′ untranslated region that regulate its stability and translation. RNA-binding proteins such as tristetraprolin promote transcript degradation, while inflammatory signaling can temporarily stabilize the transcript or increase its translation. This permits rapid production without requiring the cell to maintain continuously high TNF output.
Post-transcriptional regulation is physiologically important rather than merely a biochemical refinement. Experimental disruption of the AU-rich regulatory region of mouse Tnf produces excessive TNF expression and inflammatory disease, including arthritis and intestinal pathology. Such models show that the duration and magnitude of cytokine production can determine disease even when the cytokine’s amino-acid sequence is unchanged. They also explain why measurement of transcript abundance alone does not always predict the amount of active TNF reaching nearby cells.
Additional control occurs through protein processing, receptor availability, and negative feedback. ADAM17 regulates release of soluble TNF, while shedding of TNF receptors generates extracellular binding proteins that alter ligand availability. Anti-inflammatory mediators, including interleukin-10 and glucocorticoid-dependent pathways, can suppress TNF production. Repeated exposure to microbial stimuli can also change the responsiveness of myeloid cells. These interacting controls make TNF output a dynamic property of a cellular network, not a simple readout of whether a microbial stimulus is present.
Receptors and the organization of signaling
TNF acts primarily through two receptors: TNF receptor 1, encoded by TNFRSF1A, and TNF receptor 2, encoded by TNFRSF1B. They are often abbreviated TNFR1 and TNFR2 and are also known as p55/p60 and p75/p80, respectively, reflecting historically measured molecular masses. Both belong to the TNF receptor superfamily and contain extracellular cysteine-rich domains that participate in ligand recognition. Their intracellular regions, however, differ substantially, so binding of the same ligand can produce different signaling outcomes.
TNFR1 is expressed by most nucleated cell types and contains an intracellular death domain. Despite its name, this domain does not cause cell death automatically; it provides a protein-interaction surface used to assemble signaling complexes. Following receptor activation, adaptor proteins including TRADD and RIPK1 help recruit TRAF2 and cellular inhibitor-of-apoptosis proteins. These proteins organize a membrane-associated complex in which covalent ubiquitin modifications serve as docking and regulatory signals. Much of TNFR1’s normal activity begins with this inflammatory and cell-protective complex.
An important output is activation of NF-κB, a family of transcription factors. In the canonical pathway, activation of the IκB kinase complex promotes degradation of inhibitory IκB proteins, permitting NF-κB to enter the nucleus and regulate target genes. TNFR1 also activates mitogen-activated protein kinase pathways, including those involving p38 and JNK. These pathways influence cytokine production, adhesion molecules, stress responses, and cell survival. Their combined activity explains how TNF can reprogram a cell without killing it.
TNFR2 lacks a death domain and is expressed more selectively, including on subsets of lymphocytes, endothelial cells, and other cells whose expression changes with activation. It recruits TRAF proteins and can activate canonical and noncanonical NF-κB signaling. TNFR2 is particularly relevant to the expansion and function of some regulatory T cells, although its effects are not exclusively suppressive. In other settings, it supports activated immune cells or tissue responses. Its restricted and inducible distribution makes anatomical and cellular context especially important.
Receptor signaling also depends on higher-order organization, internalization, and competition for shared intracellular components. TNFR2 can alter the availability of TRAF2 and inhibitor-of-apoptosis proteins, thereby changing how TNFR1 signals in the same cell. Soluble receptor fragments may bind TNF outside cells and modify its persistence or effective concentration. Thus, the two receptors are not simply independent switches labeled “damage” and “repair.” Their outputs emerge from receptor abundance, ligand form, signaling-complex composition, and the state of the responding cell.
Cell survival, apoptosis, and necroptosis
One of the central findings of TNF research is that the same receptor can support either survival or death. In many cells, the immediate response to TNFR1 activation is survival combined with inflammatory gene expression. NF-κB induces proteins that restrain death pathways, including c-FLIP and members of the inhibitor-of-apoptosis system. Ubiquitin scaffolds within the receptor-associated complex help maintain this state. Consequently, purified TNF often produces little direct killing in cells whose protective signaling machinery remains intact.
When these safeguards fail, TNFR1 signaling can shift toward apoptosis. Cytoplasmic signaling assemblies can recruit FADD and caspase-8, activating a protease cascade that dismantles the cell in an organized manner. In some cell types, mitochondrial amplification involving BCL-2-family proteins is required for efficient execution. The final result includes characteristic changes such as chromatin condensation, cellular fragmentation, and packaging of material for removal. Apoptosis is therefore a regulated cellular program, not a direct corrosive effect of the cytokine.
A different outcome is necroptosis, a regulated form of membrane-disrupting cell death. In appropriate cellular settings, particularly when caspase-8 activity is absent or inhibited, RIPK1 and RIPK3 can assemble a signaling platform that activates MLKL. Activated MLKL associates with cellular membranes and disrupts membrane integrity. Unlike the usual containment associated with apoptosis, this process can release intracellular material that stimulates surrounding cells. Whether necroptosis occurs depends on the presence and regulation of the required molecular machinery.
The choice among survival, apoptosis, and necroptosis is affected by ubiquitination, deubiquitination, phosphorylation, protein abundance, and the metabolic condition of the cell. RIPK1 illustrates this complexity: it can function as a scaffold supporting survival, while its kinase activity contributes to certain death pathways. It is therefore misleading to describe all RIPK1-dependent biology as equivalent. Experimental inhibitors that block kinase activity may leave scaffold functions intact, producing effects different from deletion of the protein.
These mechanisms matter beyond laboratory cell lines. Regulated cell death can help remove infected or damaged cells, but inappropriate activation can compromise epithelial barriers and sustain inflammation. Conversely, excessive protection from death may permit persistence of abnormal cells. TNF’s contribution to cancer is similarly context-dependent: it can participate in tumor-cell killing and vascular injury, while chronic TNF-driven inflammation can also support processes associated with tumor development. Its historical name captures one possible outcome within a much wider signaling system.
Host defense, inflammation, and tissue physiology
TNF is an important mediator of innate immunity. At sites of infection or injury, it activates vascular endothelial cells to express adhesion molecules and produce chemokines. These changes help circulating leukocytes slow down, adhere to the vessel wall, and move into affected tissue. TNF also increases the responsiveness of phagocytes and interacts with other cytokines to coordinate antimicrobial activity. The effectiveness of this response depends on concentrating the signal in the appropriate place and limiting its duration.
A particularly well-established role is the control of mycobacterial infection. TNF works with interferon-gamma and other signals to support macrophage activation and the organization of granulomas, structured cellular aggregates that help contain persistent pathogens. Granulomas are dynamic tissues rather than inert walls; their organization depends on continuing exchanges between macrophages, lymphocytes, and stromal cells. Experimental loss of TNF signaling and clinical experience with TNF blockade both demonstrate its importance in containment of tuberculosis.
TNF also connects local inflammation with systemic physiology. It contributes to fever through cytokine networks and prostaglandin-dependent changes in temperature regulation, and it promotes production of other inflammatory mediators. Together with interleukin-1 and interleukin-6, it helps coordinate responses involving the liver, vascular system, endocrine pathways, and brain. These effects are often indirect and mutually reinforcing. A systemic response should therefore not be attributed to TNF alone when several interacting cytokines are active.
Effects on metabolism helped establish the identity of TNF with cachectin. TNF can suppress adipocyte lipoprotein lipase activity, alter lipid handling, and interfere with insulin signaling in experimental systems. Persistent inflammation also promotes muscle protein breakdown and changes appetite through multiple pathways. TNF contributes to these networks, but cachexia in cancer or chronic disease is not explained by a single mediator. Other cytokines, disease-specific factors, reduced intake, and altered energy expenditure may all participate.
TNF can influence repair as well as injury. Transient signaling helps coordinate debris clearance, immune-cell recruitment, and regenerative responses, while prolonged signaling may impede tissue recovery or promote destructive remodeling. In bone, TNF can enhance osteoclast formation and activity in conjunction with other signals, helping explain inflammatory bone erosion. In neural tissues, TNF can affect glial and neuronal functions, but the outcome depends strongly on receptor usage and local conditions. Across organs, duration, concentration, and cellular source are more informative than a simple division into beneficial and harmful TNF.
Contributions to human disease
Persistent TNF signaling is a major component of rheumatoid arthritis. In inflamed synovial tissue, macrophages, fibroblasts, lymphocytes, and other cells form mutually reinforcing cytokine networks. TNF increases endothelial activation, leukocyte recruitment, and production of mediators that promote cartilage degradation and bone erosion. Work by Marc Feldmann, Ravinder Maini, and colleagues helped establish TNF as a therapeutically actionable component of this network. Clinical improvement after TNF blockade provided evidence that it was not merely an incidental marker of inflammation.
TNF also contributes to several forms of inflammatory bowel disease, particularly Crohn disease and ulcerative colitis. Its effects include activation of intestinal immune cells, modulation of epithelial barrier function, and regulation of cell survival and death in the mucosa. These processes occur within a larger interaction among host genetics, intestinal microbes, immune regulation, and environmental exposures. TNF inhibition can control disease in many patients, but variable responses show that intestinal inflammation is not uniformly dependent on a single cytokine.
In psoriasis, psoriatic arthritis, and axial spondyloarthritis, TNF participates in inflammatory circuits involving immune cells and tissue-resident cells. Its relationship to the interleukin-23/interleukin-17 axis differs by tissue and disease. The effectiveness of TNF blockade across several conditions illustrates a shared inflammatory mechanism, whereas differences between drugs and diseases reveal the limits of treating all TNF-associated inflammation as equivalent. TNF can also contribute to uveitis and other inflammatory disorders for which specific inhibitors have established indications.
Excessive systemic inflammatory signaling can contribute to circulatory dysfunction and organ injury during sepsis. TNF was identified as an important mediator in experimental endotoxin shock, prompting trials of TNF-directed interventions. However, results from many broad clinical studies did not reproduce the dramatic benefits seen in simplified animal models. Human sepsis varies in pathogen, timing, immune state, and organ involvement; the same patient may experience inflammatory injury together with impaired antimicrobial immunity. These complexities limit direct translation from a single experimental trigger to a heterogeneous clinical syndrome.
Inherited disease provides another perspective. TNF receptor-associated periodic syndrome, or TRAPS, is caused by pathogenic variants in TNFRSF1A. It involves recurrent inflammatory episodes and is not simply a state of excessive circulating TNF. Altered receptor folding, trafficking, cellular stress, and downstream inflammatory responses contribute to its biology. More generally, elevated TNF in a blood or tissue sample does not establish that it is the initiating cause of a disease. Its causal importance is strongest where genetic, experimental, pathological, and treatment evidence converge.
TNF-targeted medicines and their limitations
Therapeutic TNF inhibition developed from experimental immunology into clinical practice during the 1990s. In the United States, infliximab and etanercept first received approval in 1998, for Crohn disease and rheumatoid arthritis, respectively. Subsequent agents included adalimumab, certolizumab pegol, and golimumab. These medicines established cytokine blockade as a major approach to chronic inflammatory disease. Their approved indications differ by molecule and jurisdiction, so the clinical properties of one TNF inhibitor cannot automatically be assigned to another.
The agents have distinct molecular designs. Infliximab, adalimumab, and golimumab are monoclonal antibodies with different developmental origins and binding properties. Certolizumab pegol is a polyethylene-glycol-conjugated antibody fragment that lacks an Fc region. Etanercept is a fusion protein containing the extracellular portion of TNFR2 joined to an immunoglobulin G Fc region. Unlike the TNF-specific monoclonal antibodies, etanercept also binds lymphotoxin-alpha. These structural differences influence pharmacology and the ways the agents interact with soluble and membrane-associated targets.
Blocking TNF reduces activation of downstream inflammatory networks rather than simply removing one isolated laboratory measurement. In responsive disease, this can diminish inflammatory-cell recruitment, lower production of other mediators, and reduce tissue damage. Nevertheless, substantial differences exist among conditions: etanercept is not an effective treatment for Crohn disease, whereas several anti-TNF antibodies are effective. Proposed explanations include differences in binding behavior, membrane TNF interactions, reverse signaling, and effects on target cells; no single mechanism accounts for every clinical distinction.
The principal limitations follow partly from TNF’s protective functions. TNF inhibitors are associated with serious infections, including tuberculosis reactivation and certain opportunistic infections. Hepatitis B reactivation is another recognized risk in susceptible individuals. Product labeling also addresses issues such as worsening heart failure in certain settings, demyelinating disorders, and malignancy-related warnings. Rare paradoxical inflammatory reactions, including psoriasis arising during treatment, further illustrate that suppressing one pathway can change the balance of a wider immune network.
Treatment response may be absent from the outset or diminish over time. Contributing factors include disease mechanisms that are not predominantly TNF-dependent, differences in drug exposure, increased clearance, and formation of antibodies against the therapeutic protein. Biosimilar versions of several TNF inhibitors have expanded their availability; these are evaluated for high similarity to their reference products and the absence of clinically meaningful differences. Outside systemic inflammatory treatment, recombinant TNF also has a specialized regional oncology application: tasonermin is used with melphalan in isolated limb perfusion for selected soft-tissue sarcomas, limiting systemic exposure to a cytokine whose widespread administration is highly toxic.
Experimental methods and unresolved questions
TNF has been investigated through biochemical assays, structural biology, cell culture, genetic models, and clinical studies. Early functional assays measured cytotoxicity in sensitive cell lines, whereas modern immunoassays detect TNF protein using specific antibodies. These approaches measure different properties. An immunoassay may detect protein that is bound to another molecule or not fully biologically active, while a functional assay depends on the sensitivity and signaling machinery of the chosen cells. Results therefore require interpretation in relation to what the assay actually measures.
Circulating TNF is an imperfect biomarker of tissue-level activity. Production can be highly localized, concentrations can change rapidly, and soluble receptors alter ligand distribution and persistence. Sample handling and assay design also affect measurements. Tissue immunohistochemistry, flow cytometry, and transcript-based methods help identify producing or responding cells, but none alone captures the complete pathway. Transcript detection does not establish secretion, and receptor expression does not establish that a receptor is currently signaling.
Genetically modified mice have been particularly important in separating TNF production, receptor function, and downstream cell-death mechanisms. Deletion of Tnf, Tnfrsf1a, or Tnfrsf1b produces different phenotypes, while cell-specific alterations reveal roles that whole-animal experiments can obscure. Such studies must be interpreted alongside species differences and experimental context. Mouse responses to a defined infectious challenge or induced arthritis model do not reproduce every feature of human chronic disease, especially its duration, treatment history, and environmental diversity.
Single-cell RNA sequencing, spatial methods, proteomics, and advanced imaging increasingly reveal which cells produce TNF and which cells respond within diseased tissues. These techniques address a central limitation of bulk measurements: two tissues with similar average TNF expression may contain very different cellular arrangements. A TNF-producing macrophage adjacent to an epithelial cell, a blood vessel, or a regulatory lymphocyte may generate different consequences even when the total amount of cytokine is comparable.
An important research direction is selective modulation rather than complete blockade. Experimental approaches include preferential inhibition of soluble TNF, antagonism of TNFR1, and stimulation of TNFR2 in defined settings. Their rationale is to separate damaging inflammatory or death signals from protective immune and reparative functions. That separation remains difficult because receptor activities overlap and change with tissue context. Another unresolved issue is how to identify, before treatment, which patients have disease maintained predominantly by TNF-dependent circuits; no single widely applicable molecular test currently resolves that heterogeneity.