ADAM and MMP are zinc-dependent metalloproteinases belonging to the metzincin superfamily. These are the main substances involved in the shedding of NKG2DL (Fig. 1). The catalytic domains of these two enzymes have features conserved within the zincin enzyme family. The first is a catalytic sequence, HEXXHXXGXXH/D, that binds zinc ions, where three histidine residues coordinate with the zinc ion, and the substrate forms a fourth coordination bond. This is followed by a 1,4-β turn containing glutamic acid (Met-turn) that helps stabilize the catalytic process (33). Humans express 21 ADAMs, of which 13 are catalytically active. ADAM 9, 10, 12, 15, 17 and 19 are widely expressed in somatic cells and hydrolyze various extracellular regions of membrane proteins (34). MMPs are mainly secreted from cells, degrade various proteins in the extracellular matrix, and participate in tissue remodeling. Membrane-bound MMPs (MT-MMP) are anchored to the plasma membrane via transmembrane or GPI-anchoring domains (35).

ADAM10 and ADAM17 [also known as tumor necrosis factor alpha (TNF-α) converting enzyme] are involved in the shedding process of MICA from the plasma membrane. Under stress conditions, such as malignancy, the expression of MICA increases and endoplasmic reticulum protein 5 (ERp5) translocates from the endoplasmic reticulum to the plasma membrane surface, thereby activating ADAM. In the α3 domain of MICA, there is a conserved 6-peptide sequence between two cysteine residues, allowing ERp5 to form a complex with MICA. Through this interaction, a transient disulfide bond is formed, leading to conformational changes (36,37). This structural change allows ADAM10 and ADAM17 to approach MICA, and catalyze the hydrolysis between the α3 domain and transmembrane region of MICA, leading to its shedding from the plasma membrane, resulting in soluble MICA (sMICA). Studies on the carboxyl-terminus of sMICA have revealed that the ADAM cleavage site is not fixed, and the efficiency of ADAM-induced shedding is positively correlated with the length of the stalk region (38).

Hydrolysis of NKG2DL by the metalloproteinases ADAM and MMP has been widely studied. Previous research indicates that ADAM10 and ADAM17 have shedding effects on various NKG2DLs, including MICA, MICB and ULBP2. By contrast, the hydrolytic activity of ULBP3 is low and there is no definitive evidence that metalloproteinases directly participate in ULBP1 hydrolysis (38–41). Numerous studies have shown that ADAM and MMP can cause NKG2DL shedding in various tumors; however, most experiments have used broad-spectrum metalloproteinase inhibitors without careful differentiation (34,38–41).

Tissue inhibitors of metalloproteinases (TIMPs) are endogenously synthesized metalloproteinase inhibitors. Of note, four TIMPs (TIMP1, TIMP2, TIMP3, and TIMP4) inhibit the activities of MMP and ADAM. Increased TIMP3 expression reduces the activity of ADAM17, thereby reducing the release of sMICA, sMICB and sULBP2 (42).

The transfer of NKG2DL to exosomes and its release into the serum is another way it is shed from the plasma membrane (Fig. 1). Detergent-resistant membrane domains (DRMs) are membrane microdomains on the plasma membrane that cannot be fully dissolved using nonionic detergents, such as Triton X-100, at low temperatures. They are also rich in cholesterol and sphingolipids. GPI-anchored proteins (GPI-APs) are highly enriched in these areas. When cells are stimulated, these regions form large, stable lipid rafts that facilitate protein interactions and signal transduction on the membrane (43). the composition of exosomes is related to lipid rafts as they have similar lipid components and are enriched in GPI-APs (44). Additionally, due to mutations in the transmembrane region of MICA*008, its tail is truncated and acquires a GPI-anchoring sequence, allowing its release through exosomes (45). MICA, MICB and ULBPs have also been observed in exosomes (41,46). However, only ULBPs are GPI-anchored, whereas MICA and MICB have a transmembrane region and a cytoplasmic tail.

Transmembrane proteins, such as MICA, MICB and ADAM17, also accumulate in lipid rafts. In these areas, ADAM17 has high hydrolytic activity on various membrane proteins. Palmitoylation of cysteine residues in the cytoplasmic tails of MICA and MICB signals their location in DRMs, and inhibiting palmitoylation effectively inhibits shedding (40,47). However, this inhibitory effect was incomplete, indicating that shedding still occurred outside the DRMs.

Typically, the lipid bilayer in lipid raft regions is thicker than that in other areas of the plasma membrane. Hence, transmembrane proteins located in the lipid rafts may encounter hydrophobic mismatches. This phenomenon was observed experimentally. Although both MICA and ULBP1-3 co-localized with lipid raft regions, the association of MICA with DRMs was much lower than that of ULBP3 (48).

Uniquely, despite ULBP2 being a GPI-AP, it predominantly becomes soluble through metalloprotease-mediated proteolysis. ULBP2 is more susceptible to metalloprotease attack compared with ULBP3, and the presence of metalloprotease inhibitors significantly increases ULBP2 in exosomes while reducing it on the cell surface (41).

ULBP4 (RAET1E) and ULBP5 (RAET1G) are members of the RAET1 family that contain transmembrane regions. In 2004, Bacon et al (49), while studying RAET1G, discovered alternative splicing that led to truncation of the transmembrane region, predicting that this would turn the molecule into a soluble form, and named it RAET1G2. Subsequently, in 2009, research confirmed that both RAET1G and RAET1G2 were expressed in various tumor cells. Truncated soluble RAET1G2 can reduce the expression NKG2D in NK cells, leading to immunosuppression (30). In 2007, Cao et al (50) discovered an alternative splice form of RAET1E and named it RAET1E2, which similarly results in the loss of the transmembrane region, becoming a soluble molecule (Fig. 1) and reducing the cytotoxicity of NK cells.

Being the most polymorphic member of the non-classical MHC class I gene family, the polymorphism of the MICA gene not only affects the quantity of MICA on the cell membrane and its affinity for the NKG2D receptor, but also influences the shedding of MICA (51). Exon 5, which encodes the MICA transmembrane region, contains short tandem repeat sequences (GCT) that encode for alanine (52). The MICA-A5.1 gene (also known as MICA*008) has a purine insertion after the second GCT, leading to a frameshift mutation and a premature stop codon, giving MICA-A5.1 a GPI anchor point and the shortest transmembrane domain compared with that of other MICA proteins (52). Although the shorter transmembrane domain reduces the sensitivity of MICA-A5.1 to MMP (52), the GPI anchor prioritizes embedding into lipid rafts, thus recruiting MICA-A5.1 to exosomes and releasing more sMICA through the exosomes (53). MMP inhibitors, while inhibiting hydrolytic shedding of MICA, promote the exosome pathway, leading to the release of more sMICA (54). This indicates that drugs that suppress MICA shedding (such as MMP inhibitors) may not be as effective in patients with MICA-A5.1.

The association between MICA single nucleotide polymorphisms (SNPs) and sMICA levels has also garnered attention. SNPs in MICA can regulate its expression, and the degree of MICA expression is often correlated with sMICA levels. For instance, the SNP rs1051792 causes the 129th amino acid site in the MICA α2 domain to change from valine (Val) to methionine (55). In patients with multiple myeloma (51), those with the MICA-129Val/Val genotype release more sMICA, which is linked to the increased expression and quantity of MICA. In rs2596542 (A/G), the G allele promotes MICA expression, leading to an increase in sMICA generation (56).

NKG2DL is rarely expressed in normal tissues. When cells are subjected to various stress stimuli, such as oncogenic factors, infections, physical injuries, or inflammatory responses, they enhance NKG2DL expression through various mechanisms (1). For instance, when cells undergo heat stress, heat shock factor 1 binds to the heat shock element in the MIC promoter sequence, thereby promoting the expression of MICA and MICB through the heat shock response. This leads to increased expression of MICA/B membrane proteins in tumor cells (15). However, in leukemia and lymphoma, heat and oxidative stress promote the secretion of MICA, ULBP1 and ULBP2 in exosomes, thus reducing NK cell cytotoxicity (46).

Contrary to most stress stimuli that elevate NKG2DL membrane protein levels and boost tumor immunity, hypoxia induced by the tumor microenvironment promotes the production of sNKG2DL, exacerbating immune suppression. Under hypoxic conditions, hydroxylation of hypoxia-inducible factor (HIF)-α in tumor cells is inhibited. The non-hydroxylated HIF-α accumulates and binds with HIF-β, collectively acting on the hypoxia-responsive element of the ADAM10 gene, elevating ADAM10 expression and promoting MICA shedding (57). Upregulation of circ_0000977 modulates this process during hypoxia (58). Circ_0000977, acting as a miRNA sponge, binds to miR-153, reducing the inhibitory effect of miR-153 on HIF-α and ADAM10 (58). Targeting this pathway, nitric oxide (NO) mimetics, such as glyceryl trinitrate and DETA-NO, can promote the degradation of HIF-α, reducing MICA shedding induced by hypoxia. They activate the NO-cyclic guanosine monophosphate-protein kinase G pathway, enhancing prolyl hydroxylase-mediated degradation of HIF1-α, interfering with the accumulation of HIF-α under hypoxic conditions, thereby decreasing ADAM expression (58).

Genotoxic stress, such as that caused by chemotherapeutic drugs and ionizing radiation, can produce reactive oxygen species, leading to DNA damage, protein misfolding, and inactivation. This overactivates the extracellular signal-regulated kinase pathway, triggering cellular senescence (59). During cell cycle arrest, senescent cells activate and secrete a senescence-associated secretory phenotype, including metalloproteinases (59), which can stimulate tumor progression. In a neuroblastoma model, low doses of chemotherapeutic drugs induced cellular senescence and stimulated lncRNA MALAT1 to bind to miR-92a-3p. This reduces the inhibitory effect of miR-92a-3p on ADAM10 and promotes the shedding of MICA and MICB (60). However, drugs, such as sorafenib (61), have the opposite effect in hepatocellular carcinoma, inhibiting ADAM9 and ADAM10 and reducing MICA shedding. Therefore, it is crucial to assess the effects of drugs on NKG2DL shedding carefully.

The tumor microenvironment plays an important role in the development and metastasis of tumors, and the generation of sNKG2DL is regulated by cells and molecules in the tumor microenvironment. Cancer-associated fibroblasts promote tumor cell migration by reshaping the microenvironment. Research has shown that cancer-associated fibroblasts secrete high levels of active MMPs, causing MICA/B shedding from the surfaces of melanoma cells (62). Platelets release exosomes containing soluble ADAM10 and ADAM17 into tumor cells, facilitating the shedding of NKG2DL from the tumor cell surface (63). In lymph node mesenchymal stromal cells, both transcription and expression levels of ERp5 and ADAM10 were revealed to be elevated, promoting the shedding of MICA and ULBP3 from the surface of lymph node mesenchymal stromal cells and Reed-Sternberg cells (64).

Metalloproteinases, which are essential enzymes for the generation of sNKG2DL, influence sNKG2DL production based on changes in their expression and content. Cytokines in the TME affect the expression of metalloproteinases through various signaling pathways and transcription factors. IFN-γ affects MMP expression via STAT, TGF-β through Smad, and IL-1β and TNF-α through the mitogen-activated protein kinase pathway (65). Thus, cytokines may influence the shedding of NKG2DL through these pathways. IL-1β facilitates ADAM9-mediated NKG2DL cleavage (66). TGF-β induces ADAM17 mRNA expression, promoting sMICA and sULBP2 production (67,68). Furthermore, indoleamine 2,3-dioxygenase 1 regulates ADAM10 expression via the indoleamine 2,3-dioxygenase 1-Kyn-AhR pathway, thereby inducing sMICA shedding and immune evasion (69).

Morimoto et al (25) found that MUC1-C suppresses MICA/B expression and promotes sMICA/B production through various pathways. MUC1-C can activate NF-κB, promoting H3K27 trimethylation mediated by EZH2 and methylation of the MICA/B promoter region by DNA methyltransferase, thus suppressing MICA/B expression. Another possible mechanism is that MUC1-C facilitates MICA/B shedding mediated by ERp5. Notably, RAB27A plays an important role in exosome secretion. MUC1-C also forms a complex with RAB27A, which effectively promotes the release of exosomes containing MICA/B.

In summary, ADAM and MMP are the core enzymes responsible for producing sNKG2DL and are critical targets for regulating sNKG2DL. Their roles in tumor development, invasion, and related regulatory factors have been extensively studied. Although the precise mechanisms by which certain regulatory factors affect sNKG2DL are yet to be elucidated, their potential impact on sNKG2DL production cannot be overlooked. Considering the multifaceted roles of ADAM and MMP in regulating tumor growth, inducing apoptosis, and promoting invasion, targeted interventions against these proteinases may yield more pronounced antitumor effects.

Immunological molecular markers play a vital role in tumor diagnosis and are increasingly used in clinical practice. Numerous studies have shown significant elevations in sNKG2DL levels in the sera of various patients with cancer, suggesting its potential as an auxiliary diagnostic biomarker for tumors. It has been proven to be effective in some cancers (81). Particularly in pancreatic ductal adenocarcinoma, CA19-9 (a commonly used clinical biomarker for pancreatic cancer diagnosis) levels can increase under several benign conditions, affecting its specificity (82). However, using sMICA and sMICB as biomarkers has the potential to overcome the diagnostic limitations of CA19-9 with higher specificity and sensitivity (83,84). Nevertheless, distinguishing between malignant and benign kidney tumors using sMICA alone has limited applicability (85). Moreover, a study with 512 participants with various cancers concluded that when potential confounding factors were considered, the sMICA level was a valuable tool for differential diagnosis and treatment efficacy evaluation (86). However, the same conclusion has not been drawn for sMICB (87).

Whether the level of sNKG2DL has diagnostic value depends on the tumor type. However, despite the limited number of studies, sNKG2DL has demonstrated great potential for auxiliary or early diagnosis of some cancers and can be used to differentiate between benign and malignant tumors (81–87). Additionally, multiple experiments have suggested combining detection with other biomarkers to improve results.

Serum sNKGDL levels have been observed to be closely associated with higher tumor grading and staging (86,87). It has now been observed in various cancers and serves as a biomarker to assist in determining tumor malignancy or in early screening (88,89). Notably, a report on liver cancer observed that the concentration of sMICA/B decreased in the T4 stage or poorly differentiated tumors (G3). This outcome might be due to severe impairment of liver cell function in patients with end-stage cancer, such as protein synthesis disorders, or it might be due to post-transcriptional regulation, such as proteasomal degradation (90).

Although no studies have explicitly indicated that serum sNKG2DL levels can be used clinically for tumor grading and staging, the significant correlation between them suggests great potential in this regard. However, a study has noted that serum-sNKG2DL is a superior indicator of systemic tumor manifestations than local tumor differentiation (91). Their roles in grading and staging, however, require further evaluation.

sNKG2DL reflects biological behaviors centered on cancer cells and the state of tumor immune surveillance, potentially serving as a predictive factor for rapid tumor development, lymph node metastasis, or distant metastasis. sNKGDLs have been found to correlate with tumor size, progression, and/or metastasis (84,88). For instance, Qiu et al (85) found that sMICA concentration was significantly associated with lymph node metastasis, distant metastasis, and vascular invasion in renal cell carcinoma. Given the current absence of reliable noninvasive serum biomarkers for the diagnosis and monitoring of patients with renal cell carcinoma, this noninvasive detection method appears to have an advantage (85). Additionally, sNKG2DL may also be related to the transformation of precancerous lesions into tumors, as it has been observed to increase under various precancerous conditions (92,93). For example, MICA shedding is a crucial determinant of host immunity in the progression from monoclonal gammopathy of undetermined significance (a precancerous condition) to mature multiple myeloma (93).

Furthermore, sNKGDLs can be used to predict tumor prognosis by correlating them with overall survival, recurrence rates, treatment outcomes, and drug resistance, among other indicators. Strong correlations have already been observed in numerous cancers (81,94), with some being considered independent indicators of poor prognosis. Notably, Paschen et al (94) observed that sULBP could serve as a prognostic biomarker, outperforming the already established melanoma serum marker B100P. However, comparisons are still lacking in a number of cancers, and the broader implications of sNKG2DL require further evaluation. In a meta-analysis, Zhao et al concluded that sMICA/B is a marker of tumor prognosis and immunotherapy efficacy, corroborating the views expressed in the present review (95). However, some studies have indicated that sNKGDLs are not related to tumor size, metastasis, or poor prognosis (86,96). This may be due to factors, such as detection methods or sample sources, suggesting the need for a standardized detection criterion to make results comparable across different laboratories (8).

mAbs targeting the MICA α3 domain, such as 7C6 and 6E1, can inhibit the shedding of MICA/B and increase its surface expression, suppressing tumor metastasis (Fig. 3) (119–121). When used in conjunction with HDAC inhibitors, HDAC inhibitors enhance the expression of the MICA/B gene, whereas MICA/B antibodies stabilize proteins synthesized on the cell surface, thereby achieving improved antitumor effects (121,122). Additionally, the Fcγ segment of the mAb can mediate the antibody-dependent cellular cytotoxicity effect, triggering immune cells to kill target cells (121). Since the α3 domain is relatively conserved, it can overcome the effects of MICA/B polymorphism (119). Recently, Badrinath et al (123) designed a tumor vaccine targeting the MICA α3 domain, which can induce the body to produce antibodies against the MICA α3 domain, inhibiting the proteolytic shedding of MICA/B from tumor cells, enhancing the cytotoxic function of NK cells, and increasing the cDC1-mediated cross-presentation of tumor antigens to CD8⁺T cells, inducing responses against MICA/B in both CD4⁺ and CD8⁺ T cells (123).

The mAb targeting sMICA, B10G5, is considered to neutralize free sMIC in the plasma, thus alleviating the immunosuppressive effect of sMIC (Fig. 3). Although B10G5 can also recognize membrane-bound MIC, it binds at a site different from that of NKG2D, thus it does not interfere with the function of NKG2D. On the contrary, owing to its antibody-dependent cellular cytotoxicity action, it may enhance the reactivity of NK cells toward tumor cells (124). Basher et al (125) observed that the monotherapy with B10G5 to clear sMIC effectively controlled tumor growth and eliminated lung metastasis. Additionally, since the membrane-bound MIC is already low in patients in the late stages of cancer, the effect of inhibiting its shedding is minimal; thus, using the neutralizing antibody B10G5 can have a superior effect at this stage (124).

MULT1 is a ligand for NKG2D in mice, and compared with MICA/B, it is a high-affinity, soluble ligand. It can upregulate the expression of membrane NKG2D and prevent NK cells from being desensitized by sNKG2DL, thus emerging as a potential mechanism to enhance antitumor immunity. In an experimental design developed by Deng et al (76), cells were generated to induce the production of secMULT1 (an extracellular fragment of HA-MULT1 secreted by induced fibroblasts). It was found to be resistant in various mouse models, in which secMULT1 mobilized or activated antitumor effector cells. As the adjustment of NK cell reactivity by sNKG2DLs depends on their affinity to NKG2D, the preclinical development of this new class of candidate drugs may reveal new pharmacokinetics and pharmacodynamic guidelines (126).

Given the critical role of ADAM and MMP in the release of sNKG2DL, specific inhibition of ADAM and MMP may enhance the recognition of tumor cells by the immune system. Multiple studies have observed a reduction in sNKG2DL content in supernatants after treatment with metalloproteinase inhibitors (72,119,122), with some studies concurrently observing elevated levels of membrane-bound NKG2DL (Fig. 3) (60,131). However, most experiments remain at the preclinical cell-level stage, and studies on the in vivo efficacy and potential side effects are either insufficient or absent. Furthermore, TIMPs, which are endogenous proteinase inhibitors, inhibit MMPs when their expression is upregulated. Fatty acid amide hydrolase inhibitors (132) and demethylating agents (azacytidine and decitabine) (42) inhibited sNKG2DL shedding by upregulating expression of TIMPs. Targeted therapies that inhibit multiple enzymes and thus reduce sNKG2DL production are feasible approaches. However, some inhibitors have been reported to have side effects or toxicity, which limit the application of this method (133).

Experiments have observed that phenylarsine oxide, an inhibitor of the protein disulfide isomerase ERp5, can inhibit the release of sNKG2DL (Fig. 3). However, its effective concentrations are toxic. The combination of phenylarsine oxide with HDAC inhibitors and MMP inhibitors may have a synergistic effect (134,135).