CADM1, a member of the immunoglobulin superfamily cell adhesion molecule, acts as a tumor suppressor in a variety of human cancers. CADM1 is also ectopically expressed in adult T-cell leukemia (ATL), conferring an invasive phenotype characteristic to ATL. Therefore, CADM1 plays dual roles in human oncogenesis. Here, we investigate the roles of CADM1 in small cell lung cancer (SCLC). Immunohistochemistry demonstrates that 10 of 35 (29%) primary SCLC tumors express CADM1 protein. Western blotting and RT-PCR analyses reveal that CADM1 is significantly expressed in 11 of 14 SCLC cells growing in suspension cultures but in neither of 2 SCLC cells showing attached growth to plastic dishes, suggesting that CADM1 is involved in anchorage-independent growth in SCLC. In the present study, we demonstrate that SCLC expresses a unique splicing variant of CADM1 (variant 8/9) containing additional extracellular fragments corresponding to exon 9 in addition to variant 8, a common isoform in epithelia. Variant 8/9 of CADM1 is almost exclusively observed in SCLC and testis, although this variant protein localizes along the membrane and shows similar cell aggregation activity to variant 8. Interestingly, both variant 8/9 and variant 8 of CADM1 show enhanced tumorigenicity in nude mice when transfected into SBC5, a SCLC cell lacking CADM1. Inversely, suppression of CADM1 expression by shRNA reduced spheroid-like cell aggregation of NCI-H69, an SCLC cell expressing a high amount of CADM1. These findings suggest that CADM1 enhances the malignant features of SCLC, as is observed in ATL, and could provide a molecular marker specific to SCLC. (Cancer Sci 2012; 103: 1051–1057)
Small cell lung cancer (SCLC) is one of the cancers that is refractory to therapeutic approaches, although patients with SCLC often respond favorably to combined modality chemotherapy at the initial treatment. To date, a molecular targeting therapy has not been developed for SCLC. Therefore, novel approaches to the diagnosis and treatment of SCLC on the basis of molecular alterations would be a prerequisite to control this refractory cancer. One of the most critical phenotypes to determine the prognosis of SCLC patients is its metastatic ability to distant organs, even in the very early stages of the disease. Corresponding to its metastatic phenotype, cultured cells from SCLC poorly adhere to plastic dishes and often grow in suspension. Thus, in spite of its epithelial origin, SCLC cells could also be categorized as cancer cells of less adhesive activity, such as lymphoma and leukemia. It is well known that the TP53 and the RB1 genes are inactivated in the majority of SCLC. In addition, promoter methylation of the RASSF1, point mutation of the KRAS2 and inactivation of the MYO18B genes are detected in a significant portion of SCLC.[3, 4] It is also reported that the genetic polymorphisms in the MTH1 gene are associated with the risk of SCLC. Disrupted functions of these gene products would be involved in the malignant features of SCLC. However, molecules directly involved in anchorage-independent growth of SCLC cells, especially those involved in cell adhesion, are not well investigated in SCLC.
The CADM1 on chromosomal fragment 11q23.2 was initially identified as a tumor suppressor in non-SCLC (NSCLC) by its tumor suppressor activity in nude mice.[6, 7] CADM1 encodes an immunoglobulin superfamily cell adhesion molecule (IgCAM) expressed in the brain, testis, lung and various epithelial tissues. The CADM1 protein plays important roles in epithelial cell adhesion through its homophilic trans-interaction between the adjacent cells. In contrast, CADM1 expression is frequently lost in various cancers in their invasive lesions. In fact, CADM1 inactivation by promoter methylation was observed in 44% of NSCLC and 20–50% of human tumors from prostate, breast, pancreas, stomach, esophagus, nasopharynx and uterine cervix. CADM1 associates with an actin-binding protein, 4.1B/DAL-1, and a member of the scaffold proteins, membrane-associated guanylate kinases, including MPP1, MPP2, MPP3, CASK and Pals2.[18-20] Loss of 4.1B or MPP3 in NSCLC, renal cancer and meningioma strongly suggests that 4.1B and MPP3 are downstream targets of CADM1 when it acts as a tumor suppressor in epithelial tissues.[17, 18, 21-23]
In contrast, CADM1 is reported to be ectopically expressed in adult T-cell leukemia (ATL) cells, providing a possible diagnostic marker for ATL. Moreover, CADM1-mediated adhesion between ATL cells and fibroblasts or endothelial cells appears to promote invasion into the skin or various organs, which are characteristic features of ATL. We recently demonstrated that, in ATL cells, CADM1 associates with Tiam1, a Rac-specific guanine nucleotide exchange factor, and promotes cell invasion by interacting with fibroblasts or endothelial cells in vitro. By associating with a distinct series of downstream molecules, CADM1 appears to play dual roles in human oncogenesis: as a tumor suppressor in epithelial cancers and as an oncoprotein that promotes invasion in ATL cells.
In the present study, we demonstrate the high incidence of CADM1 overexpression in SCLC and its strong correlation with anchorage-independent cell growth. Enhancement of tumorigenicity in nude mice by CADM1 suggests that CADM1 acts as an oncoprotein that promotes malignant growth in SCLC cells, as was reported in ATL cells. Furthermore, a splicing variant of CADM1 that we found in SCLC cells is highly specific to SCLC, except for testis. CADM1 could provide a novel molecular target for the diagnosis and growth suppression of SCLC cells.
Materials and Methods
Tumor samples, cell lines and animals
Five primary NSCLC tumors and a primary SCLC tumor, as well as corresponding non-cancerous tissues from the same patients, were surgically resected and histologically diagnosed at the Department of Diagnostic Pathology, National Cancer Research Center Hospital, Japan. Human samples were analyzed in accordance with the institutional guidelines (IMSUT 20–36). A total of 16 SCLC and 10 NSCLC cells were obtained, as described in Supplementary Table S1. The cells were cultured according to the suppliers' recommendation. BALB/cA-nu (nu/nu) mice were from Charles River Laboratories Japan (Yokohama, Japan). All animal experiments were performed according to the Guidelines for Animal Experiments at the University of Tokyo.
Human adult lung and brain poly-A+ RNA was obtained from Clontech (Palo Alto, CA, USA). Poly-A+ RNA and total cellular RNA were extracted using the FastTrack 2.0 Kit (Invitrogen, Carlsbad, CA, USA) and reverse-transcribed by Superscript II (Invitrogen), as described previously. Primers used for amplification of a 150-bp fragment of CADM1 are 5′-TTTTCTAGCAGTGAACTCAAAGTATCAT-3′ (sense) and 5′-GATATCGATCATCAGATTACGTGGTG-3′ (anti-sense), while those of a 646-bp fragment of beta-actin as an internal control in the same reaction are 5′-AAATCTGGCACCACACCTT-3′ (sense) and 5′-AGCACTGTGTTGGCGTACAG-3′ (anti-sense). For quantitative RT-PCR analysis, real-time monitoring of a PCR reaction of 55 cycles was performed using the Light Cycler system (Roche Molecular Systems, Indianapolis, IN, USA) and the Light Cycler Fast Start DNA SYBR Green I kit according to the manufacturer's instructions. Relative expression of CADM1 was measured in three independent experiments.
Analysis of splicing variants
cDNA fragments of CADM1 corresponding to exons 7-11 were amplified by PCR using a pair of primers, 5′-GTGATGGTAACTTGGGTGAGAGTC-3′ (sense) and 5′-CCAGAATGATGAGCAAGCACAG-3′ (antisense). The 5′ end of the sense primer was labeled with Texas Red, and PCR was carried out as described previously.(9) After denaturing at 95°C for 3 min, the PCR product was subjected to PAG containing 6 M urea for 180 min at 45°C using SF5200 (Hitachi Electronics Engineering, Tokyo, Japan), with cooling systems as described previously. The results were analyzed using a DNA Fragment Analyzer (Hitachi Electronics Engineering).
Western blotting and deglycosylation
Western blotting was performed as described previously using a rabbit anti-CADM1 polyclonal antibody (pAb) (CC2) and a goat anti-GAPDH pAb (V-18; Santa Cruz Biotechnology, Santa Cruz, CA, USA) as a control. N-linked glycans were digested by peptide N-glycosidase F (New England Biolabs, Beverly, MA, USA) as described previously. Briefly, 20 μg of protein from cell lysates in a 90 μL lysis buffer was denatured with 10 μL of a glycoprotein denaturing buffer (5% SDS, 10% beta-mercaptoethanol) at 100°C for 10 min and then incubated with 10 μL of a G7 buffer (0.5 M sodium phosphate, pH 7.5) and 10 μL of 10% NP-40 containing 1500 U of peptide N-glycosidase F at 37°C for 5 h.
Paraffin sections of primary SCLC were purchased from Cybrdi (Gaithersburg, MD, USA). The sections were incubated with a rabbit anti-CADM1 pAb (CC2) and visualized by Envision kit/HRP (DAB) (Dako) as described previously. All sections were counterstained with hematoxylin. For analyses of the subcellular localization of CADM1, cells were seeded on culture slides for 2–3 days, permeabilized and stained with anti-CADM1 pAb(CC2). CADM1 protein and actin filaments were visualized with Alexa Flour 488-conjugated anti-rabbit IgG and Alexa Flour 568 phalloidine, respectively, and detected by a Biozero BZ-8000 (Keyence, Tokyo, Japan).
Plasmids and transfection
Splicing variants of CADM1 containing the sequences corresponding to exons 7, 8 and 11 (variant 8) (3, 25) and exons 7, 8, 9 and 11 (variant 8/9) were cloned into the expression vector pcDNA3.1 (Invitrogen). Transfection was carried out using FuGENE 6 (Roche Diagnostics, Indianapolis, IN, USA) and stable cell clones were selected against 200 μg/mL of hygromycin.
Cell aggregation assay
A cell suspension (1 × 105 cell/well) in normal HBSS with Ca2+ and Mg2+ or HBSS without Ca2+ and Mg2+ was reseeded in Ultra Low Cluster six-well plates (Corning Incorporated, Corning, NY, USA) and rotated at 37°C for 20–60 min, as described previously. The extent of aggregation of cells was represented by the ratio of the total particle number at time t of incubation (Nt) to the initial particle number (N0).
The siRNA sequences targeting CADM1 were from position +194 to +212 and +204 to +222 of the CADM1 mRNA relative to the first adenine of the initiation codon at position 1. For constructing an expression vector of CADM1 siRNA, ribonucleotides of the above sequence were synthesized and cloned into pSilencer 3.0-H1 (Ambion, Austin, TX, USA) to generate the shRNA, followed by the insertion of the neomycin resistance gene into the vector at the SspI site. Then, the CADM1 shRNA expression vector or pSilencer-negative control-neo siRNA (Ambion) was transfected into NCI-H69 cells using DMRIE-C (Invitrogen) and selected against 500 μg/mL of G418 in the culture medium for 14 days to obtain the pooled clones.
Analysis of tumorigenicity
A suspension of 1 × 106 cells in PBS (0.2 mL) was subcutaneously injected into the flanks of 5–6-week-old female BALB/cA-nu (nu/nu) mice. Tumor size was measured twice weekly using a Vernier caliper. Tumor volume (V) was calculated using the values of the largest (A) and smallest (B) diameters according to the formula V = 0.5 × AB2.
Experimental differences were tested for statistical significance using the independent t-test. The software Stat View 5.0 (SAS Institute, Cary, NC, USA) was used for the analysis. Differences with P-values of <0.05 were considered significant.
Expression of CADM1 in small cell lung cancer
Expression of the CADM1 in 16 SCLC and 10 NSCLC cells was examined by quantitative RT-PCR (Fig. 1a and Fig. S1). A significant amount of CADM1 mRNA was detected in 14 of 16 (88%) SCLC cells, whereas CADM1 mRNA was reduced or absent in all 10 NSCLC cells, as reported previously. Interestingly, 2 SCLC cells, SBC3 and SBC5, lacking CADM1 expression, grew attached to the plastic dishes, similarly to all 10 NSCLC cells examined. In contrast, all 14 SCLC cells expressing CADM1 mRNA showed anchorage-independent growth, providing a strong association between CADM1 expression and the growth ability in suspension (P < 0.0001). Western blotting revealed that 11 of 14 SCLC cells showed a considerable amount of CADM1 protein. However, CADM1 protein is absent in SBC3 and SBC5, as well as in 9 NSCLC cells (Fig. 1b). Immunohistochemical studies revealed that CADM1 was highly expressed along the cell–cell attachment sites in 10 of 35 (29%) primary SCLC (Fig. S2).
Identification of the splicing variant of CADM1 in small cell lung cancer
Broad signals for CADM1 from SCLC cells in western blotting suggested possible glycosylation (Fig. 1b). Thus, 4 SCLC cells, NCI-H69, NCI-H510, Lu134 and Lu139, expressing a high amount of CADM1, were treated with PNGase F and examined by Western blotting. The molecular weight of CADM1 protein was greatly reduced after treatment with PNG-F, indicating that CADM1 is modified by N-glycosylation, as reported previously, although the degree of N-glycosylation does not correlate with the malignant feature of these 4 cell lines (Fig. 1c). A couple of distinct signals, however, were still recognized in all SCLC cells after the treatment of N-glycosydase. Because four splicing variants of CADM1 corresponding to exons 8–10 were reported previously (Fig. 2a),[29, 30] we determined the sequences of CADM1 cDNA corresponding to exons 7–11 by RT-PCR and sequencing. As shown in Fig. 2b, normal human lung cDNA expressed a single major isoform, variant 8, containing the sequences corresponding to exons 7, 8 and 11. In contrast, for normal human brain cDNA, there were three major isoforms, including variant 8(–) containing exons 7 and 11, variant 9/10 containing exons 7, 9, 10 and 11, and a novel variant 9 containing exons 7, 9 and 11. For human brain cDNA, there were three additional isoforms with trace amounts, including variant 8, variant 8/9 containing exons 7, 8, 9 and 11, and another novel variant 8/9/10 containing exons 7, 8, 9, 10 and 11. Notably, all 14 SCLC cells expressing CADM1 contain a significant amount of variant 8/9, as does variant 8 (Fig. 2b,c). In contrast, only a traceable amount of variant 8/9 is detected in 5 NSCLC cells, as is the case in normal human or mouse tissues, except for the testis (Figs. 2b,c and 3). A primary SCLC tumor expresses a significant amount of both variant 8/9 and variant 8 of CADM1, in contrast to 5 other primary NSCLC tumors (Fig. 2d). These findings indicate that the expression of variant 8/9 of CADM1 is specific to SCLC, except for the testis.
Cell aggregation activity and tumorigenicity of small cell lung cancer cells by CADM1
To understand the biological significance of variant 8/9 and variant 8 of CADM1, we obtained SBC5 derivatives expressing significant amounts of variant 8/9 (SBC5-v8/9) and variant 8 (SBC5-v8) and with a control vector (SBC5-c) (Fig. 3a). Immunohistochemical analyses demonstrated that variant 8 and variant 8/9 of CADM1 are expressed at the cell–cell attachment sites and co-localized with F-actin in confluent cells (Fig. 3b,c). In a cell aggregation assay, both SBC5-v8/9 and SBC5-v8 cells formed large aggregates after 60 min of incubation, whereas SBC5-c cells formed fewer aggregates at the same time point (Fig. 4a). The number of particles was significantly reduced in SBC5-v8/9 and SBC5-v8 cells when compared with SBC5-c cells (Fig. 4b). Cell aggregation activity, however, was not essentially changed in the presence or absence of Ca2+ and Mg2+ in HBSS, suggesting that the aggregation is mediated by CADM1 as one of the Ca2+-independent IgCAM molecules.
We then tested the tumor-forming or suppressing activity of CADM1 variants in SCLC. When 1 × 106 cells of SBC5-v8/9, SBC5-v8 and SBC5-c cells were injected subcutaneously into the back of Balb/c nu/nu mice, SBC5-v8/9 and SBC5-v8 cells developed palpable tumors by 31 days after injection and continued to grow large tumors. By contrast, SBC5-c cells did not form palpable tumors by 70 days after injection (Fig. 4c). These results demonstrate that variant 8/9 and variant 8 of CADM1 promote a tumorigenic phenotype of SCLC.
Suppression of spheroid-like cell growth in vitro by RNAi-mediated downregulation of CADM1 in small cell lung cancer
Finally, we suppressed CADM1 expression of NCI-H69 by RNAi and obtained H69/shCADM1 expressing shRNA against CADM1 and H69/shCON expressing control shRNA. Both variant 8/9 and variant 8 of CADM1 were significantly reduced in amount in H69/shCADM1 cells in comparison with H69/shCON cells (Fig. 5a). H69/shCON cells formed large brown spheroids in culture, as did the parental NCI-H69 cells. By contrast, H69/shCADM1 cells formed flatter sheet-like gray aggregation rather than spheroids (shown in Fig. 5b). When we counted the aggregates of more than 10 cells as one particle, large spheroids were observed in 33% of H69/shCON particles, but in only 11% of H69/shCADM1 particles. In contrast, flat sheet-like cell aggregates was observed in only 22% of H69/shCON, but in 34% of H69/shCADM (Table S2). These results suggest that CADM1 might play a role in the formation of large 3-D spheroids of SCLC cells in vitro.
The present study demonstrates that CADM1 could provide a promising target for both the diagnosis and treatment of SCLC because CADM1 is a membrane protein. For possible diagnosis, splicing variant 8/9 of CADM1 could be a promising marker of SCLC. As shown in Fig. 1, the CADM1 protein is expressed in 11 of 16 (69%) SCLC cells and 10 of 35 (29%) primary SCLC tumors. Furthermore, we found that all SCLC cells expressed a unique variant 8/9 of CADM1 in addition to a variant 8, commonly expressed in the epithelia. Considering its specific expression in SCLC and testis, variant 8/9 would be a very promising candidate as a diagnostic marker of SCLC. It is noteworthy, however, that a very small peak corresponding to variant 8/9 was also detected in mouse brain, esophagus, jejunum and kidney. A small peak might be derived from a small population of SCLC-precursor cells of neuroendocrine origin. Alternatively, variant 8/9 might be expressed as a minor isoform of CADM1 in cells from these tissues. In this study, we identified six splicing variants of CADM1, including two novel variants, variant 9 and variant 8/9/10. Variant 9 is expressed in the testis, colon and brain, whereas variant 8/9/10 is expressed as a small peak in the brain (Fig. S1). These variable fragments are located just outside the cell membrane of the CADM1 protein, connecting the domain of three immunoglobulin loops to the transmembrane domain as a possible hinge. It is noteworthy that cancer-specific splicing variants of CD44 also differ in the fragments just outside the cell membrane. Exons 8, 9 and 10 of CADM1 contain 28, 11 and 18 amino acids, respectively, where exons 8 and 9 contain threonine-rich repeats that could be modified by O-glycosylation. The numbers of possible O-glycosylation sites in the fragments corresponding to exons 8, 9 and 10 are 16, 2 and 0, respectively.[29, 30] Therefore, variants 8 and 8/9 contain 16 and 18 possible O-glycosylation sites, respectively, whereas variant 8(−), expressed in the brain, does not contain possible O-glycosylation sites. Therefore, variant 8(−) and variant 8 or 8/9 proteins would show a marked difference in a higher structure, whereas the difference between variants 8 and 8/9 might be less remarkable. CADM1 in SCLC is also highly glycosylated through N-glycosylation, as shown in Fig. 1c. However, the degree of glycosylation does not correlate with their malignant features in at least 4 SCLC cells examined. Thus, the pathological significance of N-glycosylation and O-glycosylation of CADM1 in SCLC remains unclear. These variants are also intriguing as a serum marker because the CADM1 protein shows possible shedding by several proteases, including ADAM10 and γ-secretase. However, we were unable to investigate the pathological significance of CADM1 expression in primary SCLC because the paraffin sections of primary SCLC that we examined were provided commercially without detailed clinicopathological information.
This study also provides a possible therapeutic target of SCLC. Of 14 SCLC cell lines expressing CADM1, 11 showed spheroid growth in vitro. By contrast, CADM1 is not detected in 2 SCLC cells showing attached growth, strongly suggesting that CADM1 overexpression could promote anchorage-independent cell growth, one of the indicators of the malignant phenotype of cancer cells. CADM1, therefore, appears to act as an oncoprotein in SCLC, as it does in ATL, rather than a tumor suppressor. This hypothesis was supported by the following three findings in the present study: (i) CADM1 expression in SCLC cells correlated well with spheroid formation in vitro; (ii) introduction of CADM1 into SCLC cells enhanced aggregation in vitro and tumor formation in nude mice; (iii) suppression of CADM1 expression by shRNA reduced a population of spheroid growth in vitro. Spheroid growth of SCLC shown in this study is abrogated by serum depletion, indicating that this is not a feature of currently recognized cancer stem cells. However, the enhancement of cell aggregation in vitro and tumorigenicity in vivo by CADM1 indicate that CADM1 promotes the malignant features of SCLC, in contrast to NSCLC. The distinct roles of CADM1 in oncogenesis in SCLC and NSCLC would be due to the different downstream cascade in these two subtypes of lung cancer. One of the candidate downstream targets in SCLC is Tiam1, as observed in ATL, although preliminary results show that Tiam1 protein is detected in both NCI-H69 cells expressing CADM1 with anchorage-independent proliferation and SBC5 cells lacking CADM1 with anchorage-dependent proliferation. Phosphoinositide 3-kinase (PI3K) is another possible downstream molecule. A previous study demonstrates that PI3K is constitutively active in SCLC and could mediate anchorage-independent proliferation of SCLC cells. Furthermore, a subset of SCLC showed oncogenic mutation of the PIK3CA gene, although possible crosstalk of CADM1 and PI3K is not yet demonstrated. Additional studies are required to obtain a better picture of the CADM1 cascades in SCLC. It is also noteworthy that instead of spheroids, sheet-like aggregation was predominantly observed in NCI-H69 when CADM1 expression was reduced. This sheet-like aggregation was further broken to generate tiny particles or single cells when treated with EDTA (data not shown). These findings suggest that CADM1 might play a role in 3-D spheroid aggregation, whereas 2-D sheet-like aggregation is mainly caused by Ca2+-dependent cell–cell adhesion activities, including those by cadherins. In summary, we demonstrate that variant 8/9 of CADM1 is specifically expressed in SCLC and might promote spheroid-like cell growth and tumorigenicity, providing a possible molecular target for the diagnosis and treatment of SCLC.
This work was supported in part by a Grant-in-Aid for Young Scientists (B) (20790276, 22700914 for MI and 21790309 for MSY) and a Grant-in-Aid for Scientific Research (B) [22300336 for YM] from the Ministry of Education, Culture, Sports, Science, and Technology of Japan; a Grant-in-Aid for the Third-Term Comprehensive Control Research for Cancer from the Ministry of Health, Labor, and Welfare of Japan; and a Grant from the Program for the Promotion of Fundamental Studies in Health Sciences from the National Institute of Biomedical Innovation (ID05-10 for YM).