Deregulated Expression of c-mos in Non-Small Cell Lung Carcinomas: Relationship with p53 Status, Genomic Instability, and Tumor Kinetics

Genomic instability is a hallmark of malignant cells and occurs either in the form of microsatellite instability or in the form of CIN(1).² Microsatellite instability is reflected by alterations in polymorphic, short, tandem repeat sequences and is associated in a small fraction of colorectal carcinomas with germ-line or somatic mutations of the DNA mismatch repair genes (2). On the other hand, CIN, which represents gains or losses of segments or whole chromosomes and leads to abnormal numbers of chromosomes (aneuploidy),is likely to occur in most human malignancies (1). The molecular basis of the latter is just beginning to be investigated. It is reasonable to suggest that alterations of the genes involved in replication, chromosome condensation, and segregation, as well as inactivation of checkpoint genes that monitor DNA damage and proper assembly of the spindle, represent the main events leading to CIN(3).

Of the rapidly increasing list of molecules that participate in the cell-cycle checkpoint pathways, p53 is the most widely studied. p53 is a transcription factor that mediates G₁ arrest in response to genotoxic stress, allowing time for DNA repair or apoptosis in case that repair is not feasible (4). It is also activated by oncogenic stimuli that trigger an antitumorigenic response(5) and by antimicrotubule drugs that elicit a G₂ arrest (6). Thus, p53“senses” alterations at both checkpoints,G₁-S and G₂-M. These diverse “safeguard” properties of p53 make it the most common genetic target in human malignancies (7). Although the downstream effects of p53 have been investigated in detail(4), our knowledge about the upstream biochemical signals that activate p53 are still poorly understood. Increasing evidence suggests that phosphorylation is one type of upstream signal that triggers the p53 regulatory functions. p53 phosphorylation is mediated by several cellular kinases including ataxia-telangiectasia mutated kinase (8), DNA-dependent protein kinase(9), Jun-NH₂ kinase(10), casein kinases I and II, cyclin-activating kinase complex, and others (9). In a recent report, Fukasawa and Vande Woude (11) demonstrated in vitro that the levels of wt p53 increased in response to high mos/MAPK expression,which induces growth arrest or apoptosis depending on the phase of the cell cycle. They also observed that abrogation of p53 was followed by a reduction of growth arrest, a 2–3-fold higher transformation efficiency of mos in mouse embryo fibroblasts, and chromosome instability.

The c-mos proto-oncogene is the cellular homologue of the v-mos oncogene, a product of Moloney murine sarcoma virus,identified in the early 1980s. It is located at chromosome region 8q11–12 and encodes a M_r39,000 protein with serine-threonine kinase activity (12, 13). c-mos is an upstream activator of the MAPK pathway,phosphorylating MAPK kinase (14). It plays an important role in oocyte maturation, in which, as a component of the cytostatic factor, it arrests oocytes at metaphase II by stabilizing the maturation-promoting factor (14). Moreover, c-mos activity is associated with the appropriate formation and orientation of the meiotic spindle that leads to asymmetric division of the oocyte and the production of the first polar body (15). Although the role of c-mos in oocyte maturation is well established,very little is known about its expression and functions in human somatic cells (16). Constitutive expression of c-mos in somatic cells, such as mouse fibroblasts, induces oncogenic transformation (17). There is strong evidence that the transformation activity of c-mos, like many other oncogene products, is exerted when it is expressed in the G₁ phase. c-mos has been proposed to act as a mitogenic stimulus by modulating,via the MAPK pathway, the activities of many downstream G₁ targets including c-fos,c-jun, c-myc, S6 kinase II, and TCF/Elk-1(14, 18 and references therein). It has been suggested that c-fos is the main effector of the c-mos/MAPK pathway leading to cellular transformation (14, 19). Furthermore, two recent studies (20, 21) showed that serum-starved v-mos-transformed cells had elevated levels of certain cyclins (D, E, and A), cyclin dependent kinases(p33^cdk2 and p43^cdc2), and S-phase specific E2F complexes, suggesting that the inability to down-regulate these critical cell cycle regulatory molecules may contribute to neoplastic transformation. Interestingly, it seems that the role of c-mos in somatic cell transformation is not limited only to the G₁ phase but involves mitosis as well. The effects of c-mos in the M phase are characterized by meiotic-like modifications that lead to the production of binucleated cells and may indicate a novel mechanism of CIN (22).

In view of the diverse activities of c-mos in various cell lines and its apparent link with p53, we examined in a series of 56 NSCLCs, which have been used in previous analyses of a G₁ phase protein network (22), the following: a) the status of c-mos at the protein, mRNA, and DNA levels; b) its relationship to genomic instability (aneuploidy) and the kinetic parameters of the tumors (PI and AI); and c) its association with p53 alterations and their concomitant relationship with the above parameters. To the best of our knowledge this information has not been addressed thus far.

A total of 56 NSCLCs and adjacent normal lung tissue were analyzed. These tumors were part of a panel of 68 NSCLCs, classified according to the WHO criteria and Tumor-Node-Metastasis system(23), previously investigated for a G₁ phase protein network including p53 (Ref.24; Table 1).

For immunohistochemical analysis, the following antibodies were used: P-19 (class, IgG goat polyclonal; epitope, COOH terminus of human c-mos; Santa Cruz, Bioanalytica, Greece); DO7 (class, IgG2b, mouse monoclonal; epitope, residues 1–45 of p53; Dako, Kalifronas, Greece);MIB-1 (class, IgG1, mouse monoclonal; epitope, Ki-67 nuclear antigen;Oncogene Science, Biodynamics, Greece); and p-ERK/E-4 (class, IgG 2a mouse monoclonal; epitope corresponding to the amino acid sequence containing the phosphorylated Tyr 204 of ERK1/2).

IHC was performed by an indirect streptavidin-biotin-peroxidase method, as described previously (23).

The human cervical cell line ME180 was used as positive control for c-mos expression (16). In addition, the specificity of P-19 anti-mos antibody was tested by preincubating the latter with the appropriate control peptide (Santa Cruz, Bioanalytica). Elimination of immunostaining verified c-mos specificity. Lung carcinoma specimens from our previous study (23), with well-characterized p53 status, were used as controls for p53 protein reactivity. The DB lymphoma c-mos negative cell line (16) was used as a negative control. The specificity of E-4 anti-p-ERK antibody was tested by preincubating the latter with the appropriate control peptide (Santa Cruz, Bioanalytica). Elimination of immunostaining verified p-ERK specificity. Furthermore, in each set of immunoreactions antibody of the corresponding IgG fraction, but of unrelated specificity was used as a negative control.

Cytoplasmic and membranous immunoreactivity was considered to be evidence of c-mos expression. Nuclear staining in the presence of cytoplasmic and/or membranous reactivity was also regarded as specific,as described previously (25). Sole nuclear staining was disregarded. IHC was evaluated by examining all of the discreet areas of each tumor specimen. Tumors were considered c-mos(P) when >30% of the tumor cells were stained; otherwise, they were scored negative (N). In a recent immunohistochemical study, we divided the positively stained tumors in two groups: (+), <30%; and (2+), >30%(26). However, for the present study, the criteria were reassessed because only the carcinomas, in which >30% of the cells showed immunoreactivity, were associated with increased c-mos mRNA levels.

Tumors were considered positive when >20% of tumor cells showed nuclear staining; otherwise, they were scored as negative.

Tumor cells were evaluated as positive when staining was mainly nuclear because activated ERK translocates to the nucleus to gain access to its transcription factor substrates (18).

Tumor cells were evaluated as positive when nuclear staining, without cytoplasmic background, was observed. The PI was calculated as the percentage of MIB-1 positive cells in five to seven HPFs (at least 1000 cells were assessed). Slide examination was made by three independent observers (V. G. G., P. F., and P. Z.). Interobserver variability was minimal (P < 0.001).

For DNA extraction, contiguous 5-μm sections were microdissected as described previously (23).

DNA was extracted from 50μg of neoplastic material using the phenol/chloroform/isoamylalcohol method (27).

Cancerous material with >90% tumor cells was used for RNA extraction, because microdissection is not suitable for RNA handling methods. RNA was extracted by Trizol reagent (Life Technologies, Inc.,AntiSel, Greece) according to the manufacturer’s instructions.

To examine the mRNA levels of c-mos in tumor and adjacent normal tissues, we performed a semiquantitative multiplex RT-PCR method, as described previously (28). Briefly,target mRNA fragment is coamplified with a larger reference mRNA fragment of a ubiquitously expressed molecule. The relative ratios of the amplification products in the tumor samples reflect the relative proportion of input mRNAs and are compared with the relative ratios in the corresponding normal tissue.

The primers used in the reaction were the following: target cDNA,the amplimers for the c-mos fragment were 5′-GCC TGC TCT TCC TCC ACT CG-3′ (pos. 803) and 5′-AGT ATG TGC TGC CGC TCC CC-3′ (pos. 1100); reference cDNA, a 548-bp β-actin fragment was used as reference.

Because the c-mos gene is intronless, a positive PCR product could theoretically be obtained from DNA contaminating the RNA solutions. To exclude the latter possibility, we performed the reactions in the presence or the absence of reverse transcriptase before cDNA amplification. Contaminated RNA solutions were treated with DNase and/or 4 m LiCl₂ and tested again as described above. To exclude the possibility of having false-positive results attributable to a carry-over effect in our assay, RNA was isolated from several murine cell lines and included in the experiments as negative controls. Because the primers used do not share sequence homology with murine c-mos (12, 29), no specific PCR product was expected in the murine cells tested in the absence of human c-mos sequence contamination. To avoid possible RT-PCR intra-assay alterations that could affect the original transcript ratio (target versus reference), 2-fold serial dilutions of RNA solutions were used in control amplifications. From these reactions a low number of PCR cycles was established for assessing RNA transcript ratios.

The relative ratio of c-mos:β-actinwas quantified using an image analysis system (Media Cybernetics,Silver Spring, Maryland). Because β-actinamplification is not necessarily similar to that of c-mos,comparison of the relative ratio of c-mos andβ -actin between corresponding tumor and normal specimens was performed [c-mos/β-actin(tumor):c-mos/β-actin(normal)]. Performing the above procedure in 100 pairs of normal-normal specimens to determine the normal variability of this ratio, we found it to be 1.00 ± 0.18. When the tumor:normal relative ratio was within this range, c-mos mRNA levels were considered to be normal. Values higher than 1.18 were interpreted as c-mos mRNA overexpression. All of the reactions were performed twice.

Thirty μg of RNA were electrophoretically separated on 1%4-morpholinepropanesulfonic acid-formaldehyde gel and transferred to nylon membrane by capillary blotting. Equal loading was confirmed by staining with ethidium bromide. Hybridization was performed with a 40-bp antisense c-mos 3′end-labeled FITC-conjugated dUTP probe spanning nucleotides 1061–1100 of the published cDNA sequence (12). The probe was labeled using the enhanced chemiluminescence 3′-oligolabeling and detection system(Amersham Life Science, Buckinghamshire, England). The filter was prehybridized in the buffer supplied by the manufacturer at 42°C for at least 2 h and hybridized at 42°C for 12 h in the same buffer that contained the c-mos 3′end FITC-dUTP-labeled probe. The membrane was washed, twice with 5× SSC-0.1% SDS at 42°C for 5 min, once with 1× SSC-0.1% SDS at room temperature for 5 min,and once with 0.1× SSC-0.1% SDS at room temperature for 5 min. For detection, the appropriate anti-FITC antibody was applied and c-mos mRNA expression was analyzed using an enhanced chemiluminescence system (Amersham Corp., Arlington Heights, IL). RNA from the ME180 and the DB cell lines were used as c-mospositive and negative controls, respectively.

To examine amplification of the c-mos gene in tumor tissues, we performed the d-PCR method as described previously (30). Briefly, this technique is based on the coamplification of a target DNA fragment with a shorter reference DNA fragment of a single-copy gene. The relative ratios of the amplification products in the tumor samples reflect the relative proportion of DNA fragments and were compared with the relative ratios in the adjacent normal tissue. It should be noted that theoretically the tumor:normal relative ratio in normal tissue equals 1, because normally no gene amplification occurs.

We used a 245-bp region of the IFN-γ gene as a reference fragment. The sequences of the primers used to amplify the fragment were as follows: 5′ ACA AGG CTT TAT CTC AGG GGC CAA C 3′ (pos. 4972) and 5′ AAG CAC CAG GCA TGA AAT CTC C 3′ (pos. 5195). The target DNA sequence was the c-mos fragment identical to that amplified in the Multiplex RT-PCR assay described above.

The relative ratio of c-mos:IFNγwas quantified using an image analysis system (Media Cybernetics). Values >2 were regarded as indicative of c-mos gene amplification.

The method was performed on matched normal and tumor DNA as described previously (23). On the basis of the nucleotide sequence of the c-mos gene reported by Watson et al. (12), amplimers were designed using the Oligo software (version 4.01; National Biosciences, Inc., Plymouth,MN) to produce four partially overlapping fragments (Fig. 1): fragment 1, 5′ GTC TCT TCA TTC ACT CCA GCG 3′ (pos: 194) and 5′ CTT GTT CAC TTG CTT TAT GGC 3′ (pos: 515); fragment 2, 5′ TGG GAG CTG GAG GGT TTG GC 3′ (pos: 438) and 5′ TCC AGT GCG GCA GTG AGG CT 3′ (pos:750); fragment 3, 5′ GCC TGC TCT TCC TCC ACT CG 3′ (pos: 803) and 5′AGT ATG TGC TGC CGC TCC CC 3′ (pos: 1100); and fragment 4, 5′ AAA TGA CTA CCA AGC AGG CG 3′ (pos: 1052) and 5′ ACC AAG TTT TCA GTC AGC CG 3′(pos: 1284). The amplimers for p53 have been reported previously(23).

The DNA fragments that showed mobility shifts were purified using the QIA Gel Extraction kit (Qiagen, Bioanalytica, Greece) and analyzed further with automated sequencing. Cycle sequencing for c-mos was performed on Gene Amp PCR system 9700 (PE-Applied Biosystems, Warrington, United Kingdom) using the ABI PRISM BigDye Terminator Cycle Sequencing Kit (PE-Applied Biosystems) according to the manufacturer’s directions. In brief, 100 ng of purified DNA were mixed with 1.6 pmol of each primer, 0.75 μl of BigDye mixture, and 3.25 μl of the Diluent solution in a 10-μl reaction. The thermal profile of the reaction comprised: 1 min at 95°C, 25 cycles of 20 s at 95°C, 15 s at 58°C, 30 s at 68°C, and 3 min 30 s at 60°C, followed by a final extension at 60°C for 5 min. The products were electrophoresed in a denaturing gel on an ABI PRISM 377 DNA Sequencer (PE-Applied Biosystems). Results were analyzed using the Genescan and Genotyper software (PE-Applied Biosystems). The p53 exons that harbored putative mutations were amplified with primers tailed with M13 forward and reverse primers and sequenced as formerly described (23). All of the PCR and sequencing reactions were performed and confirmed twice by two independent laboratories.

To examine allelic alterations of c-mos and p53, we chose two markers: D8S285, which lies within the vicinity (telomeric boundary) of c-mos, and D17S179E, a pentanucleotide marker located within the first intron of p53. Primers were designed using the Oligo software (version 4.01; National Biosciences, Inc.).

The primers were: D17S179E, 5′-AGTAAGCGGAGATAGTGCCA-3′(ABI-HEX labeled) and 5′-GCACTGACAAAACATCCCCT-3′; and D8S285, 5′-CAGAATCTTTGCTACCTACA-3′ and 5′-CAGTTTTTATGGGTTTATGG-3′.

The reaction mixture consisted of: 1× GeneAmp buffer II, 250μ m deoxynucleotide triphosphates, 2 mmMgCl_2, 0.25 μm of each primer, and 2 units of AmpliTaq Gold polymerase. The PCR amplification parameters were: initial denaturation for 12 min at 95°C, 30 cycles of 30 s at 94°C, 30 s at 55°C, and 30 s at 72°C, followed by a final extension of 20 min at 72°C. The D17S179E products were analyzed on an ABI-PRISM 377 Automatic Sequencer (PE-Applied Biosystems). The products of D8S285 were analyzed on 10%native gels and stained with silver, and results were evaluated by densitometry. All of the LOH results were confirmed with a second PCR.

The variability of the reaction and the establishment of the cutoff level for scoring AIm have been described previously(31). Briefly, when the allele ratio values were ≤0.65 or≥1.54, samples were scored as LOH; whereas when their values covered the range 0.77–1.23, they were scored negative.

The samples were stained according to the Thionin-Feulgen procedure (32). The measuring and evaluation procedures have been reported formerly (28, 33). Cases with >5% of cells with DNA content above the 5c limit were considered as aneuploid (33).

Double-strand DNA breaks were detected by TUNEL according to the method of Gavrieli et al. (34).

We used tissue sections incubated with DNase I before treatment with terminal deoxynucleotidyltransferase as positive controls and sections incubated in terminal deoxynucleotidyltransferase buffer without the presence of the enzyme as negative controls.

Cells were considered to undergo apoptosis when nuclear staining,without cytoplasmic background, was observed. AI was estimated as the percentage of apoptotic cells in 10 HPFs (counted cells, 900–10,000). Slide examination was performed by three independent observers (P. F.,V. G. G., P. Z.). Interobserver variability was minimal(P < 0.001).

The possible associations between c-mos and p53 status independently, c-mos/p53 patterns with PI, logAI, ploidy status, and clinicopathological parameters were assessed with the nonparametric Kruskal-Wallis and Pearson χ² tests (Table 2). Furthermore, ANOVA was used to evaluate more specifically the possible association between p53 status, c-mos/p53 patterns, and PI and logAI (Table 3). Logistic regression model formulation was applied for estimating possible associations between p53 status, c-mos/p53 patterns, and ploidy status (Table 3). Finally, Kaplan-Meier survival curves were plotted for the parameters examined in the present study. Differences between survival curves were examined by log-rank testing. All of the analysis was performed with the SAS statistical package. The statistical difference was considered significant when the Pwas <0.05.

Expression of c-mos was observed in 14 of 52 informative cases (27%). Staining was mainly cytoplasmic and membranous, but in certain cases nuclei reacted with the anti-mos antibody as well. Expression was confined to the cancerous areas (Fig. 2,a). Considerable intratumor staining heterogeneity was noticed. The relationship between c-mos immunohistochemical status,smoking habits, histology, lymph node status, and stage is presented in Table 2. Statistical analysis revealed an association between c-mos expression and stages grouped as I and II/III [4 of 23 (17.4%) versus 10 of 29 (34%); P = 0.018 by Pearson χ²; Table 2].

The c-mos mRNA levels were examined in normal and tumor paired samples by a semiquantitative multiplex RT-PCR assay. Interestingly, c-mos mRNA was detected in all of the normal and tumor specimens including the cell lines used as negative controls in IHC analysis. The possibility of false positive results was excluded by using the appropriate controls (see “Materials and Methods”). However, multiplex RT-PCR revealed in 15 of 53 informative cases (28%)significantly higher levels of c-mos mRNA in the tumors compared with the corresponding normal tissues (Table 1; Fig. 3). In these samples, the relative ratio ranged between 3 and 7.2 compared with the value 1.00 (±0.18) characteristic of the normal status (see “Materials and Methods”). It is of the most interest that complete concordance was found between c-mos immunohistochemical positivity and c-mos mRNA overexpression(Table 1). Furthermore, RNA from six matched normal-tumor cases (17, 26, 48, 56, 58, and 68) with c-mos protein expression was available for Northern blot analysis. The levels of c-mos mRNA were higher in the tumor samples than in the corresponding normal ones, which confirms the multiplex RT-PCR results. Two c-mos transcripts of 3.5 and 1.7 kb were clearly detected in two tumor samples, whereas in the other four cancerous ones only the 3.5 kb band was definite. All of the normal counterparts expressed the 3.5 kb transcript (Fig. 4).

Three of the 56 samples (5%) showed tumor-specific mobility shifts(Fig. 5). Sequence analysis revealed one missense mutation and two silent ones. The missense mutation resulted in a single-base substitution of arginine (R) to leucine (L; CGG→ CTT) at codon 22 (case 31), whereas the silent mutations resulted in an alanine (A) substitution(GCC→GCA) at codon 260 (case 43) and in serine (S) substitution(AGC→AGT) at codon 324 (case 48; Table 1). d-PCR in our series did not reveal any c-mos gene amplification. Given the latter, the observed AIm in 4 of 47 informative cases possibly reflects LOH at the chromosomal locus D8S285 (Fig. 6,a), which lies at the telomeric boundary of c-mos(Table 1).

To examine the downstream biochemical effect of c-mos expression, we performed an immunohistochemical analysis of phosphorylated ERK1/2 in a group of c-mos positive and c-mos negative cases (Table 1). Staining was mainly nuclear (Fig. 2 b). Expression was observed in the cancerous areas and in the surrounding normal tissues as well. We did not detect any differences in ERK1/2 staining among the c-mos(P) and c-mos(N) group of patients.

Expression of p53 was observed in 33 of 55 informative cases (60%). The association of p53 with the clinicopathological parameters of the patients is summarized in Table 2. A significant association was noticed between p53 positive staining and smoking [30 of 44 (68%) versus 3 of 11 (27%); P = 0.013 by Pearson χ²; Table 2; and P = 0.020; odds ratio, 5.714 (1.313–24.871)by logistic regression analysis].

Sequence analysis revealed p53 mutations in 24 of 56 cases(43%). Twenty mutations were missense, two were silent (cases 43 and 62), and two were frameshift (cases 27 and 44; Table 1). One more sample (case 20) yielded a tumor-specific mobility shift on SSCP analysis, but further sequencing was not performed because material was not available. A highly significant association was observed between p53 gene mutations and p53 immunostaining [20 of 33 (61%) versus 4 of 22 (18%); P = 0.005 by Pearson χ²]. AIm analysis with D17S179E (Fig. 6,b) showed LOH in 14 of 21 informative samples (67%). Seven of these cases (50%) were accompanied by p53 point mutations in the remaining allele(Table 1). Because the IHC status of p53 is highly correlated with sequence analysis and the number of informative specimens at the D17S179E locus was rather low, we decided to consider p53 immunohistochemical positivity as an indicator of p53 gene alterations for subsequent analyses.

Twenty nine of 51 tumors (60.6%) were scored as aneuploid (Fig. 7). No statistically significant association between the ploidy status and the clinicopathological features of the tumors was found. The AI ranged from 0.13% to 10.67% with a mean value 2.037 ± 2.029% (Fig. 8). No significant association was found between AI, smoking, and the pathological subtype of the carcinomas. In the present study,proliferative activity was reflected by the PI using Ki-67, the most reliable antibody for estimating growth fraction by IHC(35). The percentages of Ki-67 positive cells in the cancerous areas ranged from 6.2 to 70.4%, and the mean PI was 35.06 ± 12.16% (Fig. 2 c). The PI was significantly higher in squamous cells than adenocarcinoma cells [P = 0.006; MD = 9.57 (2.94, 16.21) by ANOVA] and in smokers than nonsmokers[P = 0.019 by Kruskal-Wallis test]. The aneuploid tumors demonstrated a significantly higher PI[P = 0.006; MD = 9.57 (2.94,16.21) by ANOVA] and lower AI [P = 0.031 by Kruskal-Wallis test and P = 0.017; MD of logAI = −0.628 (−1.139, −0.117) by ANOVA] than the diploid ones.

Correlation of c-mos and p53 status independently with ploidy and tumor kinetics showed that the cases with c-mos overexpression were more frequently associated with aneuploidy and appeared to have a lower AI than the c-mos(N) group, but these differences did not reach significance (Tables 2 and 3). Only the p53(P) carcinomas were significantly associated with aneuploidy, increased PI, and decreased AI in comparison with p53(N) cases (Tables 2 and 3).

When c-mos expression was examined in association with p53 status, four c-mos/p53 patterns were observed (Table 2). The most frequent ones were the c-mos(N)/p53(P) and c-mos(N)/p53(N) patterns, with 19 patients(37%) each, followed by the c-mos(P)/p53(P) phenotype that accounted for 12 cases (23%). Pearson χ² analysis showed a correlation between c-mos/p53 expression patterns, smoking(P = 0.011; Table 2) and disease stage(P = 0.018; Table 2). However, these associations were found to be independent associations when the logistic regression model was applied. The c-mos(P)/p53(P) and c-mos(N)/p53(P) phenotypes had significantly lower AI scores and were more frequently associated with aneuploidy than were the c-mos(N)/p53(N) tumors (Table 3; Fig. 9), whereas regarding proliferation, only the c-mos(N)/p53(P) profile showed significantly higher PI values than the c-mos(N)/p53(N) pattern(Table 3; Fig. 9).

Survival analysis (follow-up duration up to 37 months) using the Kaplan-Meier method did not reveal any statistically significant association between patient outcome, c-mos status, p53 overexpression,and c-mos/p53 patterns.

There is little published information on alterations of the constituents of the MAPK cascade in lung cancer(36). We are aware of only one report that has shown low expression of c-mos in mouse lung tissue (37). Our study,which deals with the status of c-mos and its relationship with its putative downstream sensor p53, tumor growth parameters, and genomic instability in human NSCLCs, has yielded several novel findings.

Overexpression of c-mos was observed in 27% of the tumors. The observation that staining was restricted mainly to the cytoplasm and membrane of the malignant cells is in accordance with the fact that c-mos is a cytoplasmic serine-threonine kinase (14). However, immunostaining was also noticed in the nuclei of some cancerous cells. The phenomenon of nuclear translocation has been observed with components of the MAPK module and is possibly related to the end result of transcriptional activation (38). Furthermore, Min Wang et al. (39) showed that microinjected c-mos in mammalian somatic cells can enter the nucleus and bind to the kinetochores disrupting the normal mitotic progression. Thus, nuclear staining of c-mos may be associated with its ability,demonstrated in vitro, to superimpose a meiotic process on the mitotic program of somatic cells, negatively influencing chromosome stability (22). This may be of relevance to our finding that 77% of c-mos positive cases in our series were associated with aneuploidy. Heterogeneous immunoreactivity was also noticed within tumors, possibly reflecting either c-mos cell-cycle fluctuations(16, 40) or clonal expansion of c-mos positive cells,during cancer progression. The higher expression of c-mos protein in stage II/III (34%) than in stage I carcinomas (17.4%; P = 0.018) supports the latter view.

The concordance between c-mos immunohistochemical positivity and high c-mos mRNA levels along with the absence of gene amplification suggests that its deregulated expression is mainly a consequence of increased transcription. This is an intriguing finding,because normally c-mos is specifically transcribed in germ cells and plays a critical role in meiosis, whereas in somatic cells,it is transcriptionally silent or expressed at low levels(16). It has been shown that c-mos in somatic cells is suppressed by a NRE, located between 392 and 502 bp upstream from its initiation site (41). A candidate c-mos repressor was identified in the nuclear extracts of several somatic cell lines but not in male germ cells in which c-mos was transcribed (42). Disruption in the repressor-NRE interaction (e.g., deletion or mutations of the NRE, impaired function, or expression of the repressor) may possibly account for c-mos transcriptional activation in our cases. To our knowledge, there is only one previously reported study(43) in primary human tumors, in which detection of c-mos mRNA in thyroid carcinomas is described. The authors also observed abnormal transcripts in a medullary thyroid carcinoma(43). However, the lack of internal reference material(adjacent normal thyroid tissue) in that study makes it difficult to judge the role of c-mos in thyroid carcinogenesis. On the other hand,we have clearly demonstrated increased expression of c-mos in lung cancerous tissue compared with adjacent normal tissue. It is noteworthy that the c-mos gene shares sequence homology with the HPV E2 gene, and cross-reactivity of these elements may lead to false-positive results (44). However, we have recently investigated our NSCLC series for HPVs but found no indication of HPV infection (45). Finally, Northern analysis of six cases with c-mos protein overexpression revealed two c-mostranscripts of 3.5 and 1.7 kb. The 3.5-kb band was clearly detected in all of the matched normal-tumor samples, whereas the 1.7-kb transcript was detected in two tumor samples. This finding is consistent with the report of Li et al. (16), who detected similar c-mos messages in various cell lines. The significance of these transcripts in c-mos regulation remains to be clarified.

At the DNA level, we did not observe amplification of the c-mos gene, a finding which is in line with the results of previously published studies in primary NSCLCs (46), NSCLC cell lines (47), head and neck squamous carcinomas(48), and primary gastric carcinomas (49). Structural abnormalities within c-mos and its flanking region were detected in pleomorphic adenomas by Stenman et al. (50). Restriction fragment analysis indicated that these genetic lesions were the result of multiple subtle mutations rather than rearrangements (50). Sequencing of c-mos in our cases showed one missense and two silent point mutations. We did not find the previously reported polymorphism (G to T transversion) at codon 105 (51, 52). The missense mutation observed in case 31 resulted in a R to L substitution at codon 22. Its significance is unclear because c-mos protein in this patient was undetectable and c-mos mRNA levels were normal. Codon 22 lies near Ser-3 and Ser-16, the phosphorylation of which is important for c-mos stability (53, 54) and interaction with MKK(55). In addition, Ser-3 is phosphorylated in vitro by ERKs, which forms a positive feedback loop(56). Thus, one possibility is that this mutation in the NH₂-terminal region destabilizes c-mos and disrupts the mos-ERK feedback loop, which affects signal transduction through the ERK pathway.

The mos/MKK/ERK pathway is involved in the cellular processes of growth (reviewed in Refs. 14, 38, 40, and 57)and differentiation (58). In vitro studies have shown that the exact outcome of this signal transduction pathway[proliferation (reviewed in Refs. 14, 38, 40, and56), cell cycle arrest (59), apoptosis(59), and differentiation (58)] depends on various parameters, such as the cell type (14, 58), the cell-cycle phase (59), and the levels of c-mos protein(40). This diversity of function is not surprising for MKKKs like c-mos, because unlike MAPKs, which are specifically recognized by their corresponding MKKs, MKKKs interact with a number of MKKs, enabling diversity in the cellular response (38). By taking into account the above apparent opposite actions of c-mos, we were led to investigate its relationship with the tumor kinetic parameters of proliferation and apoptosis. No significant differences were found between the c-mos(P) and the c-mos(N) group of patients(Table 2), which contrasts with the results of Fukasawa et al. (59), who reported that elevated c-mos levels lead to apoptosis or cell cycle arrest. One possible explanation for this apparent discrepancy is that in many of our c-mos(P) cases, a downstream inducer of apoptosis may have been defective. Prompted by the recent study of Fukasawa and Vande Woude (11),who showed that c-mos/MKK/ERK-induced apoptosis or growth arrest is p53 dependent, we sought to investigate the relationship between c-mos expression and p53 status and found that c-mos overexpression was indeed frequently accompanied by p53 alterations (Table 1). Moreover, the fact that the c-mos(P)/p53(P) immunophenotype had significantly lower AI values and was more frequently associated with aneuploidy only from the c-mos(N)/p53(N) profile, but not from the c-mos(N)/p53(P) pattern (Table 3), suggests that p53 status is the main determinant of ploidy status and apoptosis in our series. This finding also strengthens the concept that wt p53 plays a “safeguard” role in preventing oncogene-mediated activation of the mos/MKK/ERK pathway(11). A similar wt p53 “safeguard” function seems to apply to other active oncogenes such as ras (5, 60) and β-catenin (61). The finding that the group of tumors with p53 alterations showed increased proliferative activity (P = 0.004),decreased apoptotic rate (P < 0.001), and genomic instability (aneuploidy; P = 0.005)compared with the group of patients with normal p53 status (Table 2)underscores the central role of wt p53 as the cellular gatekeeper of growth and division (62).

These data lead to speculation as to the role of c-mos overexpression in a cancerous cell, because its relationship with tumor kinetics and genomic instability seems mainly to be dependent on the integrity of p53. One possible explanation is that overexpressed c-mos may bypass various G₀ steps, normally initiated by growth factors, that render the tumor cells insensitive to and independent of growth regulatory signals. Support for this hypothesis derives from a recent study (63) that exhibited that v-mos transformed murine fibroblasts were characterized by rapid growth and decreased serum requirements while simultaneously inhibiting platelet-derived growth factor-mediated signal transduction. Additional evidence comes from the reports of Rhodes et al.(20) and Afshari et al. (21), who showed that serum-starved v-mos-transformed NIH3T3 cells were unable to down-regulate the activity of critical cell cycle regulatory molecules, which included cyclins D, E, and A, cyclin dependent kinases, and S-phase specific E2F complexes. In an attempt to examine the downstream biochemical effects of c-mos overexpression, we studied the status of phosphorylated ERK1/2, but we did not find any differences between the c-mos(P) and c-mos(N) groups. This finding is expected because ERKs are activated by numerous upstream signals(38). In keeping with the latter, we observed that the c-mos(N) cases were accompanied with K-ras mutations, whereas the c-mos(P) cases possessed none. This observation implies alternative modes of activation of the Ras/Raf/MEK/ERK pathway.³Interestingly, the two c-mos(P) cases without p53 alterations were aneuploid. In this particular subset of tumors, genomic instability may be the outcome of p53 independent mechanisms (reviewed in Ref.1) in which c-mos could be implicated (22, 39). In such a process, constitutive maintenance of maturation-promoting factor during late S phase could result in premature chromosome condensation events generating free DNA ends,which are highly recombinogenic (64). Nevertheless, the small number of c-mos(P)/p53(N) cases renders the above suggestion speculative. It is noteworthy that these cases were adenocarcinoma cells and, compared with squamous cells and small cell lung carcinomas,adenocarcinoma cells exhibit the lowest p53 mutation rates(36). Finally, because sensitivity to certain microtubule-active drugs depends on p53 repression of microtubule-associated protein 4 (65) and c-mos activates the microtubule-associated protein kinase cascade (66, 67), the pattern of c-mos and p53 status may help in certain cases to determine the appropriate chemotherapeutic strategy.

Another interesting finding was that four c-mos(N) cases (8.5%) showed AIm at the D8S285 locus, which lies at the telomeric boundary of c-mos (Table 1). This relatively low incidence is in agreement to the observed lack of c-mos gene amplification. Given the latter, AIm in these four cases reflects allele loss rather than amplification. This might seem inconsistent with its role as an oncogene (68) in these cases. One explanation is that LOH at D8S285 causes loss of activity to a novel tumor suppressor gene. Yet, in case that allelic loss applies to c-mos, then its role in these tumors is even more complicated. Two studies (69, 70) using c-mos^−/− knockout mice showed that female mutants had an elevated risk of developing ovarian teratomas, but male mutants were phenotypically normal. A plausible interpretation for our result is that because c-mos, as an MKKK, is located at the“intersection” of the MAPK pathway, its deletion, in this group of carcinomas, may play a complementary role to the actions of another defected molecule thus permitting the cancerous cell to survive. The finding of Teng et al. (71), who detected deletions of MKK4 in various cell lines, is in keeping with our results.

To summarize, we have obtained evidence that c-mos protein is relatively frequently overexpressed, attributable to increased transcription, in NSCLCs. Its effects on tumor kinetics and genomic instability seem to depend mainly on the integrity of p53.

We would like to thank Prof. Seth Love, Dr. Thanos D. Halazonetis, and Dr. Tariq Enver for their valuable guidance in completing our manuscript.