Increased levels of soluble CD226 in sera accompanied by decreased membrane CD226 expression on peripheral blood mononuclear cells from cancer patients

Using the Sandwich ELISA, we found that the concentration of sCD226 was significantly higher in serum from cancer patients than in serum from normal subjects: the ranges of serum sCD226 being 0.1–23 (median 6.3) ng/mL in normal subjects and 0.2–60 (median 11.5) ng/mL in cancer patients (P < 0.001, Fig. 1A). This increase in serum sCD226 was evident in most of the tumor types we tested except liver cancer (Fig. 1B), a finding consistent with the fact that all of the tumor types shown here have been identified to express CD155 and/or CD112 [8, 17, 19‐22]. It is known that many membrane molecules can be shed from the cell surface and released into the circulation in soluble form in some pathological conditions such as tumor [23‐26]. To investigate whether the increased sCD226 in cancer patients might be shed from the cell surface, we examined the mCD226 on PBMC from both patients and normal subjects. Accordingly, it was found that the expression of mCD226 by PBMC from cancer patients was lower than that from normal subjects, with a statistically significant difference (P < 0.001, Fig. 1C). The reduction of mCD226 expression on PBMC from cancer patients was not surprising, since it had been reported that CD226 was reduced on CD56^dim NK cells from myeloma patients with active disease compared with patients in remission and healthy controls [11].

It has been well accepted that the cytotoxicity mediated by NK cells and effector T cells is one of the major mechanisms of immune surveillance against tumor cells. However, malignant cells often have developed strategies that counteract immune surveillance of the hosts, such as secreting soluble activating killing receptors and/or their ligands to avoid or reduce this kind of killing [27‐30]. To investigate the possible effect of sCD226 on this cytolytic function against malignant cells, and whether sCD226 could be one of the tumor escape strategies, a CD226-Fc fusion protein was used to mimic the function of sCD226 in vitro. NK cytolytic activity was then measured by co-incubation of normal PBMC with the NK-sensitive K562 cell line, which expresses both CD226 ligands (Fig. 2A), in a standard ⁵¹Cr-release assay, with or without CD226-Fc fusion protein. As expected, CD226-Fc fusion protein could inhibit the cytotoxicity of NK cells against K562 cells in a dose-dependent manner and the inhibition was of statistic significance when the concentration was 250, 500, and 1000 ng/mL (P < 0.05, Fig. 2B and Fig. 2C). Although the effective concentration of CD226-Fc required for inhibition in vitro was 10–20 fold greater than that observed in patients' sera, the situation in vivo is likely to be quite different. First, it is known that many different inhibitory molecules are shed by tumor cells, therefore sCD226 is likely to constitute only one of a number of molecules that each contribute to inhibition; second, natural sCD226 in vivo may contain structural differences from the CD226-Fc fusion protein, which was expressed and purified artificially and in a form of homodimer linked by human Ig Fc fragment, and these may yield a higher affinity of binding to CD155 and/or CD112. It can be speculated, therefore, that the increased sCD226 in sera from tumor patients might play a role in blocking or inhibiting the binding of mCD226 to its ligands on the tumor cell, thereby inhibiting tumor cell lysis.

Having observed the increased serum sCD226 in cancer patients and presenting a possible role for the increased sCD226, we then sought to determine the origin of the sCD226. Based on the fact that many transmembrane molecules are cleaved by certain proteases and shed into the circulation, and that several kinds of proteases can be induced in vitro by treatment with PMA [31‐33], we investigated the effect on the regulation of sCD226 and mCD226 of PMA-stimulated normal PBMC and Jurkat cells by the treatment with 15 different protease inhibitors (data not shown). As shown in Fig. 3A and 3B, three of the 15 protease inhibitors (1–10-Phenanathroline: a metalloprotease inhibitor; AEBSF: a serine protease inhibitor; N-Ethylmaleimide: a cysteine protease inhibitor) significantly decreased sCD226 level and increased mCD226 expression of the cultured cells treated with PMA. These data indicate that CD226 can be shed from the cell surface by proteolysis and released into the circulation in soluble form, and this shedding might involve more than one protease. Similarly, many other transmembrane receptors and cell adhesion molecules, can be shed from the cell surface by proteolysis and released into the circulation in soluble form, and such shedding can involve more than one protease [34]. Our finding that this process likely occurs in a number of cancers is perhaps not surprising since it is known that in these conditions many proteases are up-regulated [35].

In order to exclude the possibility that the sCD226 was, in fact, intact mCD226 derived from dead cells, we characterized the molecular weight of sCD226 both in sera samples from healthy donors and cancer patients and in the culture supernatants from normal PBMC and Jurkat cells treated with PMA for 24 h. All the four samples were aliquoted into 3 tubes: one was precipitated with normal mouse IgG-Sepharose 4B and probed with an anti-CD226 mAb FMU4 [18], one was precipitated with LeoA1-Sepharose 4B and probed with anti-SED (staphylococcal enterotoxin D) mAb, and the other was precipitated with LeoA1-Sepharose 4B and probed with FMU4. As shown in Fig. 4, protein bands were only observed when the samples were precipitated with LeoA1-Sepharose 4B and probed with FMU4. As expected, a band with molecular weight of ~50 kDa, coincident with the extracellular domain of CD226 molecule, was found in all the four kinds of sera and supernatant samples. Interestingly, another weaker band, with molecular massed of ~45 kDa, was observed in sera samples, but not in supernatant samples. Based on these findings, we speculate a reasonable explanation for the smaller molecule might be the variety of proteases in vivo, leading to the cleavage of sCD226 from different sites of the intact molecule. However, the possibility of heterogeneous glycosylation of sCD226 in vivo cannot be excluded.

Samples of patients' sera were obtained from 259 patients with various kinds of tumors, before treatment, within 1 week after entrance to Xijing Hospital, the Fourth Military Medical University. The series included a total of 13 esophageal carcinoma (7 male and 6 female), 45 gastric carcinoma (25 male and 20 female), 24 colorectal carcinoma (11 male and 13 female), 23 liver cancer (13 male and 10 female), 16 breast cancer (female), 15 cervix/ovarian/endometrial cancer (female), 46 lung cancer (28 male and 18 female), 28 leukemia/lymphoma (14 male and 14 female), and 49 others (including tumor in central nervous system, in kidney, in bladder, in parotid gland, in thyroid gland, in submaxilary gland, in suprarenal gland, in prostate gland, cholangiocarcinoma, melanoma, nasopharyngeal carcinoma, laryngocarcinoma, soft tissue sarcoma, and osteogenic sarcoma). Normal sera were isolated from 129 healthy volunteers (69 male and 60 female). Both the patients and the volunteers were 25–65 years old. They had no chemotherapy and had no significant signs or symptoms of microbial infection at the time of entrance. This study was approved by the institutional review board of the University and all individuals provided written and/or oral informed consent. The sera were stored at -20°C until use.

100 μL of the anti-CD226 mAb, LeoA1 [1] (2.5 mg/L in 0.05 M sodium carbonate buffer, pH9.5) was added to each well of an ELISA plate (Nunc, Roskilde, Denmark) and incubated overnight at 4°C. After three washes, serum samples or standard CD226 serially diluted with PBS containing 1% BSA and 0.1% (v/v) Tween-20 were added to the wells and incubated for 1 h at 37°C. After extensive washing with PBS containing 0.1% (v/v) Tween-20 (PBS/Tween), the wells were reacted with another anti-CD226 mAb [18] which was conjugated with HRP and diluted in PBS containing 3% PEG for 1 h at 37°C. Color development was performed by adding 100 μL TMB (eBioscience, CA, USA) for 10 min at 37°C and stopped by 2 M H₂SO₄. The absorbance at 450 nm was determined with a microplate reader (BioRad, CA, USA).

PBMC were freshly isolated from peripheral blood of healthy adults or cancer patients by Isopaque-Ficoll (Hao Yang, Tianjin, China) gradient centrifugation and incubated with LeoA1 or control IgG1 (BD BioSciences, CA, USA) at 4°C for 30 min, followed by washing and incubation with FITC-labeled goat anti-mouse Ig (Dako, Glostrup, Denmark) at 4°C for 30 min and then analyzed on a FACScan (BD BioSciences). Expression of CD155 and CD112 on K562 cells were analyzed similarly.

NK cytolytic activity was measured after co-incubation of normal PBMC with the NK-sensitive K562 cell line by a standard ⁵¹Cr-release assay. Briefly, K562 cells were labelled with ⁵¹Cr (Amersham biosciences, NJ, USA, 100 μCi/10⁶ cells) for 2 h at 37°C. After three washes, ⁵¹Cr-labelled target cells were incubated with various concentrations of CD226-Fc fusion proteins, control protein (human Ig), or anti-CD155/CD112 mAb for 30 min at 37°C. Meanwhile, freshly isolated PBMCs were incubated with 10% of normal human serum or normal murine serum for 30 min at 37°C. Then the target cells and effector cells were pooled together (1×10⁴/well K562 and 1×10⁵/well PBMC) and incubated for 4 h at 37°C. The percentage of cytotoxicity was calculated as percentage specific lysis, as follows:

Freshly isolated normal PBMC or Jurkat cells were cultured in RPMI 1640 (Hyclone, UT, USA) supplemented with 10% FCS, 100 U/mL penicillin/streptomycin, 10 mM HEPES, and 50 μM β-mercaptoethanol at 37°C in 5% CO₂. The cells (1×10⁶ cells/mL) were stimulated with PMA (50 ng/mL, Sigma-Aldrich, MO, USA) for 6 h, washed twice, and co-cultured with or without different protease inhibitors (Sigma-Aldrich, INHIB-1 and 131377) for 6 h. Although the concentration of the 15 protease inhibitors was not all the same, the optimized concentration for 3 effective protease inhibitors (1–10-Phenanathroline, AEBSF, and N-Ethylmaleimide) was all 1 mM. Then sCD226 level in the supernatants was detected by the ELISA mentioned above, and mCD226 expression on the cells was analyzed by flow cytometry.

Twelve ml of serum from cancer patients and healthy adults, respectively, and 30 ml of culture supernatant from normal PBMC and Jurkat cells treated with PMA (50 ng/mL) for 24 h were used to characterize the molecular weight of sCD226 by immunoprecipitaion and Western blot analysis. Briefly, LeoA1 or normal mouse IgG was covalently coupled to CNBr-activated Sepharose 4B (GE Healthcare, London, the United Kingdom) according to the manufacturer's instructions. After preclearing, the sera or supernatant samples were aliquoted and incubated with 100 μL of LeoA1-Sepharose 4B or normal mouse IgG-Sepharose 4B (negative control for immunoprecipitation) at 4°C overnight. After 4 times washing (20 min incubation for each washing) with Washing Buffer (10 mMTris-HCl, 140 mM NaCl, 0.5 mM MgCL₂, 0.5 mM CaCL₂, 0.02% NaN₃, pH7.4) and following the last centrifugation, the supernatant was aspirated and 40 μl of 2 × loading buffer was added to the bead pellet, vortexed, heated (90–100°C, 5 min), and centrifuged. The supernatant was gently collected, without disturbing the pellet, and then loaded onto 12% SDS-PAGE as the order in Figure 4, and transferred to one complete nitrocellulose membrane (Millipore, Massachussets, USA). The membrane was blocked with 5% (w/v) bovine skim milk in PBS at room temperature for 1 h, then was cut into individual lanes using scissors, and incubated with anti-SED mAb (negative control for Western blot) or anti-CD226 mAb FMU4 [18] at 4°C overnight. Then the membranes were washes six times with PBS-T and incubated with HRP labeled goat anti-mouse IgG secondary antibody (DAKO) at room temperature for 1 h. After four times washes with PBS-T, the individual membrane lanes were put together one by one and enhanced chemiluminescence (ECL) reagent (GE Healthcare) was applied to the membranes according to manufacturer's introductions, which were then exposed to an x-ray film (Kodak, Rochester, NY, USA).

The Mann-Whitney U test or Kruskal-Wallis test was used to determine the significance of the differences of sCD226 and mCD226 expression between the groups. And the paired-samples T test was used to compare percentage of cytotoxicity between the groups. A P value < 0.05 was regarded as significant. Analysis was performed with SPSS (SPSS Inc., Chicago, IL, USA).