NM-23 H1 immunohistochemistry is not useful as predictor of metastatic potential of colorectal cancer.

This study aimed to investigate whether immunohistochemical staining for nm23-H1 protein in the primary tumour is correlated with tumour stage, tumour differentiation, DNA ploidy, cell proliferative index, p53 status and patient survival time in colorectal cancer. Full-cross colorectal cancer biopsies were collected from 202 consecutive surgical specimens between 1987 and 1990. Immunohistochemical expression of nm23-H1 protein was investigated in cryosections, using a monoclonal anti-nm23-H1 antibody (clone NM 301). The staining pattern was classified as follows: strong homogeneous intensity, moderate homogeneous intensity, moderate focal intensity, or as negative. Immunohistochemical expression of p53 was investigated using a monoclonal anti-p53 antibody (DO-7). The DNA ploidy and cell proliferative index were determined by flow cytometry. Possible correlation between nm23-H1 staining patterns and the other studied tumour characteristics was explored at the end of 1994. Median survival time of living patients was 66 months, range 50-93 months. No correlation was found between various nm23-H1 staining patterns and tumour stage, cell proliferative index or p53 status. Nm23-H1-negative tumours and tumours with moderate focal staining intensity were less differentiated than tumours with strong homogeneous or moderate homogeneous staining intensity (P < 0.05). Of the nm23-H1-negative tumours, a significantly higher number was near-diploid rather than aneuploid, as compared with those expressing positive nm23-H1 (P < 0.05). The number of dead patients in Dukes' stages B and C did not correlate significantly with the nm23-H1 staining pattern. The nm23-H1 staining pattern alone, or combined with either of the other explored tumour characteristics, did not correlate with patient survival time. Immunohistochemical studies of the nm23-H1 protein expression are of minor value in the staging and prognostic prediction of colorectal cancer. ImagesFigure 1

The complex processes of tumour progression include both positive and negative regulatory elements, such as activation of oncogenes and inactivation of tumour-suppressor genes . The adenoma-carcinoma sequence in colorectal cancer is a well-known cascade of multistep genetic events including various mutations, some of which have been characterised (Cawkwell et al., 1994;Fearon, 1994). Accumulation of these and other unknown genetic changes is considered to be necessary for tumorigenesis, but none of the known changes can foretell metastatic potential (Fearon, 1994).
It was not long ago that the new human potential metastasis-suppressor gene family, 'non-metastatic' nm23, was identified (Steeg et al., 1988). nm23-H1 has been mapped to human chromosome locus 17q21.3-22, and the gene encodes a Mr 17 000 protein of unknown function (Backer et al., 1993;. Potential roles for the nm23 protein have been suggested, such as in the formation of a basement membrane, in tumour differentiation and cell proliferation (Caligo et al., 1995;Howlett et al., 1994;Lombardi et al., 1995).
A down-regulated expression of nm23-H 1, on either mRNA and/or protein level, has been reported in various human cancers (some with highly metastatic activities), such as malignant melanoma (Florenes et al., 1992;Xerri et al., 1994), squamous cell cancer of the lung (Huwer et al., 1994), hepatocellular cancer (lizuka et al., 1995), ovarian cancer (Mandai et al., 1994) and gastric cancer (Nakayama et al., 1993). Similar findings have been reported for breast cancer (Bevilacqua et al., 1989;Hennessy et al., 1991;Hirayama et al., 1991;Royds et al., 1993;Stahl et al., 1991;Tokunaga et al., 1993), although this was not the case in a study by Sastre-Garau et al. (1992). Mutations and deletions of the nm23 gene have been shown in a number of primary tumours, such as neuroblastomas, lung, breast and renal cancer (Leone et al., , 1993, but not in prostatic cancer (Brewster et al., 1994). Taken together, these results suggest that the nm23 gene may have an important role in the mechanism of metastasis in many solid tumour forms.
We have previously explored DNA ploidy, cell proliferative index (Lindmark et al., 1991) and p53 status (Kressner et al., 1996) in colorectal cancer, without being able to detect any substantial correlation to common clinicopathological characteristics and to patient survival time. Nevertheless, continued search for prognostic predictors in colorectal cancer is essential. Thus far, no factors have been identified capable of discriminating between patients truly cured by surgery and patients having subclinical micrometastases in Dukes' stages B and C (Dukes and Bussey, 1958). If such factors were to be available, additional therapy could be offered to selected patients running a high risk for tumour relapse. Furthermore, selected patients could be included in surveillance programmes.
The main goal in the present study was to evaluate whether the expression of the nm23-HI protein, possibly associated with events occurring later than p53-related events in tumour progression (Fearon, 1994), was associated with tumour stage and patient survival time in colorectal cancer. In addition, we also investigated the relation between the nm23-H 1 expression and the other studied tumour characteristics, as well as combinations thereof, in the prediction of tumour stage and patient survival.

Materials and methods
Patients Two-hundred and two potentially curable colorectal cancer patients, with no preoperative indications of tumour spread (120 colon, 82 rectum), were operated on between January 1987 and November 1990. None of the patients received adjuvant chemotherapy; however, 28 patients with rectal cancer obtained preoperative radiotherapy to 25 Gy in 5 days (Glimelius et al., 1995). There were 117 women and 85 men; mean age was 71 years (range 40 -92 years). A total of 169 patients were potentially cured with a radically excised tumour in Dukes' stages A -C. Thirty-three patients had either non-radical surgery on distant metastases and were designated Dukes' stage D. Survival was measured from the time of resection until follow-up at the end of 1994. Median survival time of 104 living patients was 66 months (range 50 -93 months).
Tumour biopsies Full-cross tumour sections, collected from 202 surgical specimens, were frozen in dry-ice isopentane and stored at -70°C. Serial cryosections were used for immunohistochemistry, and adjacent tumour tissue was used for flow cytometry analysis. Biopsies for routine histopathology were taken from all tumours.

Antibodies
Mouse monoclonal anti-nm23-H1 antibody, cloned NM301, from Becton and Dickinson (San Jose, CA, USA), and mouse monoclonal anti-p53 antibody, DO-7 (Dakopatts, Glostrup, Denmark) were used in concentrations of 1: 10 and 1: 500 respectively. Biotinylated horse anti-mouse IgG from Vector Laboratories (Burlingame, CA, USA) was used in dilution 1: 200. Antibodies were omitted and replaced by mouse IgG or dilution buffer, to test for specificity. Immunohistochemical staining Cryosections (6 ,um) were fixed in ice-cold acetone for 15 min. Incubation was performed for 60 min with the anti-nm23-H1 antibody, which was diluted in phosphate-buffered saline (PBS), supplemented by 0.1% bovine serum albumin (BSA) and 5% normal horse serum for blocking of nonspecific staining. After repeated rinsing, endogenous peroxidase was extinguished with 3% hydrogen peroxide in 100% methanol for 15 min. Following incubation with biotinylated horse anti-mouse secondary antibody for 30 min, staining was performed with the avidin -biotin complex technique, using Vectastatin ABC Elite kit from Vector Laboratories with aminoethylcarbazole as peroxidase substrate. Finally, the sections were counterstained with Mayer's haematoxylin.
Studies on DNA ploidy, cell proliferative index and p53 status are described in papers by Lindmark et al. (1991) and Kressner et al. (1996).
Histopathological evaluation Tumour stage and tumour differentiation Tumour differentiation was assessed according to the WHO recommendations (Morson and Sobin, 1976), and tumour staging according to Dukes' classification system (Dukes and Bussey, 1958).
Staining patterns obtained using the anti-nm23-HJ antibody The staining patterns were classified into four types: strong homogeneous intensity, moderate homogeneous intensity,  (Figure 1). The intraobserver variability was estimated in a second blind evaluation.
Statistics The relation between nm23-H1 staining patterns and the other studied tumour characteristics was determined by x2 analysis, where P < 0.05 was considered significant. Survival curves were constructed using the life-table (actuarial) method (Peto et al., 1977;Lawless, 1982).    (Figure 1). The intensity of the cytoplasmic staining was either strong or moderate in virtually all tumour cells in 24 and 54 sections respectively (Table I).
Sixty-four sections showed focally distributed moderate staining intensity. Sixty sections stained entirely nm23-H1 negative. The intraobserver variability was low, with unequal classification in 4 of 210 (2%) tumours. Tumour adjacent normal bowel epithelium, present in 37 of 202 sections, generally stained with strong homogeneous intensity.
nm23-HJ staining patternstumour stage and tumour differentiation No correlation was observed between the staining patterns and tumour stage ( nm23-HJ staining patterns -p53 status The distribution of p53-positive and -negative tumours did not show any significant difference in relation to the nm23-HI staining patterns in 178 of the 202 tumours in which both stainings were available ( nm23-HJ staining patternspatient survival time Patient survival curves showed no significant difference according to the various nm23-H1 staining patterns ( Figure  2). A similar observation was made when each of the studied tumour characteristics was anlaysed in relation to nm23-H1 (data not shown).

Discussion
It was not possible to clarify in this report the possible clinical significance of the differences observed in the nm23-H1 staining patterns, as no correlation was found between the nm23-H1 staining patterns and the other clinicopathological tumour characteristics and patient survival time. Nor was there any indication of a relation between nm23-H 1 staining patterns and the formation of a basement membrane, indirectly observed as tumour differentiation, or cell proliferation, suggested by Howlett et al. ), Lombardi et al. (1995) and Caligo et al. (1995. These findings support observations made in some studies showing that the nm23-H1 protein expression is independent of the tumour stage in colorectal cancer (Haut et al., 1991;Lacombe et al., 1991;Yamaguchi et al., 1993;Zeng et al., 1994). Reduced expression was shown to be associated with progressive tumour stage and distant metastasis in studies by Ayhan et al. (1993) andTannapfel et al. (1995). Moreover, Yamaguchi et al. (1993) found that the expression of both mRNA and protein was significantly lower in tumours associated with liver metastasis than in those without such metastasis. In a study by Royds et al. (1994), there was only a marginal significance between the association of death from colorectal cancer and the nm23 status.
Conflicting conclusions have been drawn in various reports, based on the nm23-H1 expression on the mRNA level. No correlation to the tumour stage was described by Haut et al. (1991), Myeroff andMarkowitz (1993), andZeng et al. (1994);Yamaguchi et al. (1993) presented a different view in a study.
Allelic deletion and/or mutation of the nm23-H1 gene has, in some papers, been shown to correlate with metastatic progression in colorectal cancer, even if no correlation between nm23-H1 protein and the initial tumour stage could be observed (Campo et al., 1994;Cohn et al., 1991;Wang et al., 1993). Steeg et al. (1991) claimed that colorectal cancer provides an example where nm23 mRNA levels remain constant, but allelic deletion is correlated with the development of distant metastasis, and furthermore, that this retained nm23 mRNA level may be explained by the fact that the remaining allele compensates for the lost one. No changes on the nm23-HI DNA level in primary colorectal cancer (Bafico et al., 1993;Cawkwell et al., 1994;Whitelaw and Northover, 1994), or in liver metatases (Heide et al., 1994), have been shown by others. Okada et al. (1994) detected allelic loss in only 3 of 29 (10%) informative colorectal cancers and Iacopetta et al. (1994) in only 3 of 19 (16%).
Nm23-H1 allelic losses are considered as secondary events, and specific nm23-H 1 mutations have not been observed frequently in colorectal cancer (Bafico et al., 1993. A lateacting suppressor gene at or near the nm23-H1 locus has been suggested (Cohn et al., 1991). It has been questioned whether nm23 acts via the traditional recessive suppressor gene model, and alterations, other than reduced nm23 expression, have been proposed as relevant to tumour metastasis .
Based on findings in this colorectal cancer study and on the available literature, it may be concluded that there are tissue-specific differences in the relative importance of the nm23-H1 gene, and that the immunohistochemical expression of nm23-H1 protein has been proven to be unrelated to tumour progression and patient survival time. Thus, this supports the opinon that the role of the nm23-HI genetic alteration must be analysed in the context of association with other genetic changes, rather than with the corresponding protein expression.