The immunoreceptor NKG2D provides costimulatory signals to CD8 T cells and potently activates NK cells, even overcoming inhibitory signals by MHC class I molecules.1, 2 Among the NKG2D ligands (NKG2DL) is the surface glycoprotein MICA, which is broadly expressed by epithelial and hematopoietic tumors but not in healthy tissue.3, 4 In mice, NKG2DL expression by tumor cells strongly stimulates antitumor immunity, leading to tumor rejection by NK cells and/or CD8 T cells.5, 6 Because the strength of the antitumor response is critically dependent on NKG2DL surface levels,5 downregulation of NKG2DL expression likely constitutes a novel tumor immune escape mechanism. We previously demonstrated that human tumor cells reduce NKG2DL surface levels by proteolytic shedding of MICA from the cell surface.7 sMICA was also reported to cause systemic downregulation of NKG2D surface expression, thereby impairing lysis of tumor cells by human CD8 T cells.8 As a consequence of MICA shedding by tumors, sMICA can be detected in sera of patients with various malignancies.4, 7, 8 These results indicate an important role of MICA release in tumor immune escape but also raise the question of whether sMICA is of diagnostic value in patients with malignancies.9 Besides a small study that revealed a correlation between sMICA serum levels and disease progression in 23 patients with prostate cancer,10 there are yet no data available regarding the correlation of sMICA levels with specific tumor entities, aggressiveness of disease and hence the potential implementation of sMICA as a diagnostic and prognostic marker in cancer.
Here, in a large study including 512 individuals, we analyzed sMICA levels in sera of cancer patients and compared these to levels in patients with matched benign diseases and healthy individuals. In addition, we compared sMICA levels between different tumor entities and addressed the correlation of sMICA with tumor stage and differentiation to evaluate sMICA as a marker in the differential diagnosis and staging of cancer.
AUC, area under the curve; HRP, horseradish peroxidase; MAb, monoclonal antibody; MICA, major histocompatibility complex class I-related-chain A; NK, natural killer; NSE, neuron-specific enolase; ProGRP, progastrin-releasing peptide; ROC, receiver operating characteristic; sMICA, soluble MICA; ULBP, UL16-binding protein.
Material and methods
Serum samples were from 512 individuals, including 62 healthy donors, 154 patients with benign diseases and 296 patients with malignant tumors. The group with benign diseases included 72 patients with benign gastrointestinal disorders (adenoma, polyposis, colitis, Crohn's disease, gastritis, peptic ulcer disease, pancreatitis, cholecystolithiasis and others), 37 with lung disorders (tuberculosis, sarcoidosis, allergic, autoimmune and infectious lung diseases; and others), 37 with gynecologic (ovarian cysts, endometriosis, uterus myomatosus and others) and 8 with other benign diseases (nodular goiter, lipoma and others). Serum samples were obtained at the time of diagnosis or at acute stage of disease before start of the respective therapy, i.e., prior to surgery, anti-inflammatory or antibiotic treatment, etc.
The group with malignant diseases comprised 88 patients with colorectal cancers and 40 with other gastrointestinal cancers (esophageal, gastric, liver and pancreatic cancers), 19 with lung cancers, 52 with breast cancers, 43 with ovarian cancers, 19 with other gynecologic cancers (cervical, endometrial and vulva cancers), 19 with renal cancers and 16 with prostate cancers. Serum samples from all cancer patients were obtained prior to surgery, which was the most frequent treatment modality, or before start of chemotherapy or radiotherapy. The study was approved by the local ethics committee.
Serum levels of sMICA were determined by our previously described sandwich ELISA.4, 7 In brief, plates were coated with the capture anti-MICA MAb AMO1 at 5 μg/ml in PBS, then blocked by addition of 15% BSA, washed and incubated with the standard (recombinant sMICA*04 produced in Escherichia coli) or patient sera. Patient sera were diluted 1:3 in 7.5% BSA. Next, plates were washed, incubated with the detection MAb BAMO3 at 1 μg/ml, washed again and then incubated with anti-mouse IgG2a-HRP (1:8,000 dilution; Southern Biotechnologies, Birmingham, AL). Finally, plates were washed and developed using the TMB Peroxidase Substrate System (KPL, Gaithersburg, MD). Absorbance was measured at 450 nm.
MICA ELISA was evaluated by assessing intra- and interassay variations (n = 10 and n = 8, respectively), as well as the variation between various sample formats, such as serum, EDTA plasma and citrate plasma (n = 5, respectively).
For clinical evaluation, sera of cancer patients were compared with sera of healthy persons and patients suffering from nonmalignant diseases. For general comparison of 2 groups, statistical calculations were performed by Wilcoxon's test. Distribution of concentrations across the categories of an ordered variable, e.g., stage and grade, were tested for an upward trend using ANOVA on ranks of values.
To assess the diagnostic potential of the assay, cut-offs were chosen at the 95th percentile to patients with relevant nonmalignant diseases and at the 95th percentile to healthy persons, and sensitivities for cancer patients were calculated. Additionally, ROC curves were established to cover the entire spectrum of sensitivity and specificity. This analysis was done for the whole sample of patients as well as for the most representative subgroups with lung, gynecologic and gastrointestinal cancers.
p < 0.05 was considered statistically significant. For all statistics, SAS software (version 8.2; SAS Institute, Cary, NC) was used.
First, we subjected the MICA ELISA to a systematic and thorough evaluation regarding its methodical quality. We determined a median intra-assay variation (n = 10) of between 4.0% and 7.4% and a median inter-assay variation (n = 8) of between 17.0% and 24.7%. When testing various formats, serum and EDTA plasma showed only minor differences (n = 5, median deviation 6.9%, range 2.8–17.2%), whereas the differences were somewhat more pronounced when comparing serum and citrate plasma (n = 5, median deviation 16.4%, range 0.2–25.0%).
Subsequently, we analyzed levels of sMICA in serum of 296 patients with various cancers including colorectal and other gastrointestinal cancers, lung cancer, breast cancer, ovarian and other gynecologic cancers, renal and prostate cancers (Table I). In parallel, we analyzed sera from 62 healthy donors and 154 patients with benign diseases who were chosen for investigation according to the most frequent cancer groups and their relevance for differential diagnosis in clinical practice.
Table I. sMICA (pg/ml) in Healthy Individuals, Patients with Benign Disorders and Patients with Malignant Disorders, Respectively
Other gastrointestinal cancers
Other gynecologic cancers
Healthy individuals revealed significantly lower sMICA values (median <30 pg/ml) than patients with benign diseases (median 84 pg/ml, p = 0.005) and cancer patients (median 161 pg/ml, p < 0.0001) (Table I, Fig. 1a,b).
Levels of sMICA differed significantly (p < 0.0001) between cancer patients and patients with benign disorders representing the most relevant control group for differential diagnosis. Patients with malignant pulmonary and gynecologic cancers exhibited significantly higher values than healthy individuals (p < 0.0001) and the respective benign control groups (p < 0.0001 and p = 0.042, respectively) as illustrated by ROC curves in Figure 2a,b. Interestingly, the ROC curves for lung cancer were similar when healthy persons or patients with benign lung diseases were used as the control group. The resulting AUC values of 0.789 and 0.805, respectively, demonstrate the diagnostic potential of MICA in this subgroup. Also, in malignant gastrointestinal diseases, significantly higher sMICA levels were detected than in healthy individuals (p < 0.0001); but sMICA levels did not differ significantly between malignant and benign gastrointestinal diseases (p = 0.238, Fig. 2c). The distribution pattern of sMICA levels in the investigated malignancies was similar in most cancers, with the median being the highest for lung cancer (Table I).
When the cut-off was chosen at the 95th percentile to healthy persons (330 pg/ml), sensitivity for the detection of malignant disease in the whole sample of cancer patients was 23.0% and ranged in the various tumor types between 12.5% and 28.9% (Table II). When the cut-off was chosen at the 95th percentile to the relevant nonmalignant diseases, the sensitivity for all cancers was 9.5%, for gastrointestinal cancer 7.0%, for gynecologic cancer 11.3% and for lung cancer 26.7% (Table II).
Table II. Sensitivities of sMICA for Cancer Detection when Cut-Offs are Chosen According to 95% Specificity against Healthy Persons and against the Relevant Benign Control Group
Cut-off at 95% specificity to healthy persons
Cut-off at 95% specificity to the relevant benign control group
sMICA levels in cancer patients correlated with the extent of disease according to the UICC classification (p = 0.015). While there was no association between sMICA levels and tumor size (p = 0.456), cell differentiation (p = 0.271) or lymph node involvement (p = 0.674), sMICA levels correlated significantly with metastasis (p = 0.007, Table III). In a detailed analysis according to the various tumor types, a significant correlation with metastasis was found for gastrointestinal tumors (p = 0.032), with similar, but not significant, tendencies for renal and gynecologic cancers (p = 0.077 and p = 0.097, respectively). Since in other tumor types, such as lung cancer, high sMICA levels were already observed in early stages, no association with tumor stage was found (Table III).
Table III. Correlation of sMICA with UICC Stages and Metastasis (M-stages); Medians of sMICA (pg/ml) for the Various Stages and Grade of Correlation by p-Values are given for all Cancers and for the Various Tumor Entities
Medians of sMICA in various stages
Correlation withstage, p
Medians of sMICA in various M-stages
Correlation with M-stage, p
Previous findings indicate that diminished expression of NKG2DL on the tumor cell surface by enhanced proteolytic shedding constitutes a novel tumor immune escape mechanism.7 Accordingly, elevated levels of sMICA were found in sera of patients with various malignancies.4, 7, 8 A small study revealed a correlation between sMICA serum levels and disease progression in 23 patients with prostate cancer.10 However, so far, no thorough evaluation has been performed on the diagnostic potential of sMICA levels in various malignant diseases and their correlation with stage and aggressiveness of disease.
Here, in a large study including 512 individuals, we analyzed sMICA levels in sera of cancer patients and compared them to levels in patients with the respective organ-specific benign diseases and in healthy individuals. In addition, we compared sMICA levels between different tumor entities and addressed a correlation of sMICA with tumor stage and differentiation to evaluate sMICA as a marker in differential diagnosis and staging of cancer. We performed experiments to determine the methodical quality of our recently developed MICA ELISA and found it to be sufficiently accurate for clinical testing. It is important to note that both sample formats, serum and EDTA plasma, are suited for sMICA quantitation.
Analysis of serum samples revealed that healthy individuals overall had significantly lower sMICA values than patients with benign diseases and cancer patients, and sMICA levels differed significantly between cancer patients and patients with benign disorders. The latter finding is noteworthy since benign diseases such as adenomas, polyps, cysts, infections, etc. are particularly relevant in differential diagnosis. Nevertheless, sMICA levels of cancer patients showed a considerable overlap with both healthy and benign groups. This observation corresponds well with findings regarding established tumor markers, such as cytokeratin-19 fragments (CYFRA 21-111) and CA 15-3,12 and more recently described markers, such as S-100 protein13 and ProGRP.14 This might be due to a remarkable variability among the tumors as far as expressing and releasing these tumor-associated molecules. In addition, all these markers are elevated in benign diseases such as renal insufficiency or infections, which limits their diagnostic specificity. Though this limits the use of tumor markers for screening purposes, they are valuable tools in differential diagnosis and therapy monitoring when potentially disturbing factors are taken into account.
A considerable number among the patients with malignancies were seronegative for MICA, which is in line with the results of immunohistochemical analyses that revealed a significant number of tumors lacking MICA expression.3 In benign disorders, higher sMICA concentrations were observed more frequently in patients with gastrointestinal diseases, which may, at least in part, be due to the propensity of the gastrointestinal tract to express MICA.15 The best discrimination between cancer patients and both control groups using sMICA was obtained in lung cancer. Thus, sMICA might be well suited for differential diagnosis of this particular malignancy, and further studies will have to compare its value with other relevant tumor markers which are currently in use.16 Interestingly, even early stages of lung cancer disease display markedly elevated sMICA levels, while when analyzing the whole group of cancer patients and some other tumor entities, we found a clear correlation of sMICA levels with tumor stage. This might reflect variations in different tumor types with regard to expression and shedding of MICA. sMICA levels were particularly elevated in patients with distant metastases but showed no correlation with tumor size, cell differentiation or lymph node involvement. Thus, the presence of sMICA in serum appears rather to be an indicator for systemic manifestation of malignancy than for local tumor extent or differentiation grade. Elevated sMICA levels in metastatic vs. localized disease may reflect differences in tumor vascularization. However, they may also be a consequence of enhanced MICA shedding, which might have impaired the confining immunosurveillance by cytotoxic lymphocytes.
Release of other NKG2D ligands, such as MICB or ULBP, may also play a role in the escape of tumor cells from immune surveillance.17 While sULBP molecules have not yet been described, expression and release of MICB have been demonstrated in patients with leukemia.4 The clinical relevance of sMICB and the correlation of sMICB levels to sMICA levels are currently being investigated in patients with various malignant and benign disorders.
In conclusion, our study provides comprehensive information regarding the clinical value of sMICA analysis. We demonstrate that sMICA levels are significantly higher in sera of patients with various malignancies than in patients with benign disorders, who in turn reveal significantly higher levels than healthy donors. The power of discrimination between cancer sera and controls varied between different tumor entities and was highest in pulmonary diseases. In cancer patients, sMICA levels correlated with cancer stage and metastasis. Further studies are ongoing to elucidate the value of sMICA for monitoring response to therapy and its prognostic relevance in malignant diseases.
We thank Drs. H.-G. Rammensee and G. Pawelec for critical reading of the manuscript.