© 1999 Cancer Research Campaign Article no. bjoc.1998.0087 Loss of heterozygosity (LOH), malignancy grade and clonality in microdissected prostate cancer

The aim of the present study was to find out whether increasing malignancy of prostate carcinoma correlates with an overall increase of loss of heterozygosity (LOH), and whether LOH typing of microdissected tumour areas can help to distinguish between multifocal or clonal tumour development. In 47 carcinomas analysed at 25 chromosomal loci, the overall LOH rate was found to be significantly lower in grade 1 areas (2.2%) compared with grade 2 (9.4%) and grade 3 areas (8.3%, P = 0.007). A similar tendency was found for the mean fractional allele loss (FAL, 0.043 for grade 1, 0.2 for grade 2 and 0.23 for grade 3, P = 0.0004). Of 20 tumours (65%) with LOH in several microdissected areas, 13 had identical losses at 1–4 loci within two or three areas, suggesting clonal development of these areas. Markers near RB, DCC, BBC1, TP53 and at D13S325 (13q21–22) showed higher loss rates in grades 2 and 3 (between 25% and 44.4%) compared with grade 1 (0–6.6%). Tumour-suppressor genes (TSGs) near these loci might, thus, be important for tumour progression. TP53 mutations were detected in 27%, but BBC1 mutations in only 7%, of samples with LOH. Evaluation of all 25 loci in every tumour made evident that each prostate cancer has its own pattern of allelic losses. © 1999 Cancer Research Campaign

The clinical outcome of prostate cancer is strongly related to its differentiation and malignancy grade. Although well-differentiated tumours do not significantly affect patientsÕ survival, lessdifferentiated neoplasms have a major impact on prognosis (Hanash et al, 1972;Bšcking et al, 1982;Dhom, 1991). A significant proportion of prostate carcinomas are heterogeneous and pluriform neoplasms, which consist of several histological patterns with different biological properties (Dhom, 1991). Whether these patterns result from mulifocal or clonal tumour development is still poorly understood.
Few LOH studies have systematically analysed the relationship between malignancy grade or clinical stage and the frequency of allelic losses (Isaacs and Carter, 1991;Trapman et al, 1994;Macoska et al, 1995;Cunningham et al, 1996;Latil et al, 1996;Vocke et al, 1996), or taken into account intratumoral heterogeneity of prostate carcinoma (Konishi et al, 1995;Mirchandani et al, 1995). There is only one report of a significant increase of LOH at a single chromosomal locus (8p12Ð21) in advanced tumour stages (Trapman et al, 1994).
The aim of the present work was to evaluate whether increasing malignancy of prostate carcinoma goes along with a general increase in the frequency of LOH and the number of chromosomal loci concerned, and whether LOH typing can help to distinguish between multifocal or clonal tumour development. We analysed 47 carcinomas for LOH at 25 chromosomal loci near known or putative TSGs, evaluating 1Ð7 areas per tumour by microdissection. Two TSGs, TP53 and BBC1 (the breast basic conserved gene) (Adams et al, 1992;Cleton-Jansen et al, 1995), have been screened for both LOH and mutations.

Tissue samples
Tissues from 47 prostate cancers (16 grade 1, 14 grade 2 and 17 grade 3 carcinomas; grading according to Bšcking and Sommerkamp (1980); Helpap et al (1985); age of patients 55Ð87 years) were obtained from transurethral resections or radical prostatectomy specimens. Thirty-seven tumours were uniform and ten pluriform carcinomas with two or three different malignancy grades. Tissues were formalin-fixed and routinely embedded into paraffin. Representative samples of all malignancy grades of every neoplasm (1Ð7 areas per tumour) were prepared for LOH analysis by microdissection under microscopic control. In total, 19 grade 1, 45 grade 2 and 33 grade 3 areas were examined.

DNA extraction
DNA was extracted from selected tumour areas and normal prostate control tissues after routine deparaffination and proteinase K digestion for 12 h using the QIAamp tissue kit (Qiagen, Hilden, Germany).
PCR was performed in a final volume of 10 µl containing 10 ng of template DNA, 50 mM potassium chloride, 10 mM tris-HCl, pH 8.3, 200 mM of each dNTP, 0.1% gelatin and 10 pmol of each primer. 0.25 units of Taq-DNA polymerase (Gibco BRL) were used. Magnesium chloride concentrations ranged from 1.5 to 2.5 mM, depending upon primer pairs. PCR reactions were carried out on a Biometra UNO-thermocycler. PCR mix in 0.5 ml tubes was overlaid with paraffin oil. For PCR, initial denaturation at 94°C for 3 min was followed by 30 cycles (94°C, 30 s; 52Ð61°C, 40 s; 72°C, 60 s) and a final elongation step of 10 min at 72°C.

Gel electrophoresis
PCR products were diluted 1.5:1 in loading buffer (formamide, bromophenol blue and xyleneÐcyanol) and denatured at 95°C for 5 min. Twelve microlitres of this mixture were run on an 8% polyacrylamide urea sequencing gel at 70 W for 2.5 h in tris-borate buffer. Amplification products were detected by a silverstaining method developed for sequencing gels (von Deimling et al, 1993;Bender et al, 1994).

Mutational analysis of TP53 and BBC1 genes by SSCP and DNA sequencing
For single-strand conformational polymorphism (SSCP) analysis, exons 5Ð8 of the TP53 gene and the two exons of the BBC1 gene were amplified by PCR (magnesium chloride concentration 1.5 mM; 35 cycles: 94°C, 30 s; 60Ð61°C, 60 s; 72°C, 60 s). Gel electrophoresis was carried out on non-denaturing polyacrylamide gels (6% or 14%, acrylamide:bis-acrylamide 1:30 with glycerol or 1:99 without glycerol, running time 16 h at 8 W at room temperature). Single strands were detected by silverstaining (see above).
Shifted SSCP bands were excised and reamplified by PCR using conditions described above. PCR products were purified with QIAquick PCR Purification Kit (Quiagen). For cycle sequencing, 1 pmol µl Ð1 sense or antisense-primer (1.6 µl), 2 µl DNA sequencing kit (Dye Terminator Cycle Sequencing Ready Reaction Mix; Perkin Elmer) and 10Ð30 ng template DNA (2Ð3 µl) were used in a final volume of 10 µl (PCR conditions: 96°C, 10 s; 50°C, 5 s; and 60°C, 4 min; 25 cycles). Reaction products were ethanol precipitated, mixed with 4 µl of loading buffer (formamide/EDTA) and denatured for 10 min at 95°C. Products were then electrophoresed through 6% denaturing acrylamide gels using an automatic sequencer (ABI Prism genetic analyser 373, Perkin Elmer). All 97 tumour areas and normal prostate control tissues were examined by SSCP and, in case of shifts, by direct sequencing.

Evaluation of LOH and statistical analysis
Allelic losses were evaluated by visually comparing alleles of normal DNA with those of tumour DNA. Calculation of fractional allele loss (FAL) was carried out by dividing the number of chromosomal arms with LOH by the total number of informative arms. The H-test of KruskalÐWallis was used to test for statistical differences.

LOH rates at the 25 chromosomal loci within all tumour samples
The frequency of allelic losses at all 25 chromosomal loci within the 97 tumour areas is given in column 3 of Table 1. No losses at all were found at loci D8S87 (8p21Ðp12), D12S270 (12q), D17S855 (17q21, within BRCA1) and at D18S543 (18q, DCC-region). Figure 1 AÐD shows examples of allelic losses. No microsatellite instabilities were observed in the present series of tumours.

The overall LOH frequency and the fractional allele loss (FAL) are related to malignancy grade
When calculating the mean frequency of allelic losses for all chromosomal markers for the three malignancy grades, losses were found in only 2.2% of grade 1 areas (s.d., 3.4%; minimum, 0%; maximum, 9.1%; median, 0%), but in 9.4% of grade 2 (s.d., 8.3%; minimum, 0%, maximum, 29.4%; median, 5.9%) and in 8.3% of grade 3 areas (s.d., 6.3%; minimum, 0%; maximum, 23.5%; median, 7.8%). The difference between grade 1 on the one hand and grades 2 and 3 on the other was statistically significant at P = 0.007. A similar tendency was found for the mean fractional allele loss (FAL): 0.043 for grade 1 areas (s.d., 0.06; minimum, 0; maximum, 0.12; median, 0), 0.2 for grade 2 areas (s.d., 0.18; minimum, 0; maximum, 0.57; median, 0.12) and 0.23 for grade 3 areas (s.d., 0.18; minimum, 0; maximum, 0.57; median, 0.12). Again the difference between grade 1 and grades 2 and 3 was statistically significant at P = 0.0004. There also was a remarkable difference in the number of chromosomal loci affected by LOH between grade 1 on the one hand and grades 2 and 3 on the other. Only six loci were  affected in grade 1 compared with 19 in grade 2 and 16 in grade 3 areas (see also columns 4Ð6 of Table 1).

LOH rates at several loci differ in the three malignancy grades
Considerable differences in LOH frequency between the three histological grades were found at several loci (columns 4Ð6 of Table 1 and Figure 2). In grade 2 and 3 areas, loss rates between 25% and 44.4% were seen for markers near RB, DCC, BBC1, TP53 and at D13S325 (13q21Ð22). These markers showed losses in only 0Ð6.6% (D13S318 near RB) in grade 1 areas. Figure 3 is an overview of the LOH typing at the 25 loci within all 97 tumour areas. In 24 of the 47 tumours (tumours 20Ð43), 2Ð7 different areas have been separately evaluated for LOH at all loci. Four of these 24 tumours (tumors 20, 21, 26 and 39) showed no losses at all. Fourteen of the 20 resting cancers (tumours 23, 24, 27Ð33, 35Ð38 and 42) presented losses of the same allele at 1Ð4 loci within two or three different areas. In 5 of these 14 tumours (tumours 29, 31, 33, 35 and 36), these areas were of different grades. It is also evident from Figure 3 that each individual cancer has its own pattern of LOH. No identical patterns were seen among the 47 tumours.

TP53
Using SSCP analysis and direct sequencing, 18 mutations of exons 5Ð8 of TP53 were found within the 97 tumour areas (Table 2), which concerned 14 tumours. Twelve codons were affected by missense mutations, three by silent mutations and one by a Tinsertion leading to a stop codon at position 144 (Table 2). Only two carcinomas (tumour 39 and 42) had TP53 mutations within two different areas (grade 3). These mutations were identical in tumour 42 (codon 134) and different in tumour 39 (codon 165 and 184). Mutations were present in only 4 of the 11 samples with LOH at TP53 (36%) ( Table 2). No correlation was found between tumour grade and mutation frequency.

BBC1
Mutations within the BBC1 gene were found in 7 of the 97 samples (six tumours; Table 3). They affected six codons (Table 3). Only carcinoma 21 had two mutations in two areas (grade 1 and 2) which were different (codon 129 and codon 279). In only 1 of the 14 probes with LOH close to BBC1 was a mutation at codon 290 found (tumour 40, area 2, Table 3). Again, no correlation was evident between tumour grade and mutation frequency.

DISCUSSION
It is well known that life expectancy of prostate cancer patients is strongly related to tumour grade. Although survival rates are not affected by grade 1 tumours, grades 2 and 3 significantly worsen prognosis (Hanash et al, 1972;Bšcking et al, 1982;Dhom, 1991). We wanted to know whether this augmentation of malignant potential goes along with an increase in LOH frequency. We actually found a significant tendency towards both an increase of the overall LOH frequency and the number of affected chromosomal loci with malignancy, when comparing grade 1 areas with grades 2 and 3. This tendency probably reflects an augmenting genetic instability which leads to LOH by mechanisms such as chromosomal losses or interstitial deletions. It is not clear why we found no significant differences between grades 2 and 3 in this series of tumours because microdissection was carried out under microscopic visual control. It can, however, not be excluded that chromosomal loci not examined in the present study differ in their LOH rates between these two grades. A significant proportion of prostate carcinomas consists of several histological patterns differing in morphology and malignant potential. Whether these patterns are of clonal or multifocal origin is still unclear. LOH typing is a suitable method to study clonality, and special stress was laid upon this point. In 24 of the 47 tumours, 2Ð7 areas have been systematically analysed for LOH at all 25 chromosomal loci. Four of these tumours had no allelic losses at all. Thirteen of the 20 remaining cancers showed losses at the same allele at 1Ð4 loci within two or three different areas. In five tumours, these areas were even of different grades. These findings show that there is at least some degree of clonality in prostate carcinoma.
Mutational analyses of TP53 and BBC1 did not contribute essentially to the question of clonality in the present series of tumours because only two of them had TP53 mutations and only one had BBC1 mutations in two different areas. Mutations were identical in carcinoma 42 (TP53), but different in tumours 39 (TP53) and 21 (BBC1). The findings of two other studies of TP53 mutations are rather in favour of a multifocal origin of prostate carcinoma (Konishi et al, 1995;Mirchandani et al, 1995).
We found TP53 mutations in 27% of samples with LOH, which may suggest a certain importance of this TSG in prostate cancer. All mutations concerned the DNA-binding domain encoding region. Those at codons 165, 184, 193, 200, 218, 239, 243 and 296 have not yet been reported (Bookstein et al, 1993;Navone et al, 1993;Chi et al, 1994). Mutations at codons 200 and 243 are known not to affect the DNA-binding properties of the p53 protein (Lin et al, 1994).
The finding of LOH at 16q24Ðqter (D16S539) near BBC1 or D16S444E and the demonstration of mutations within this gene in prostate cancers are novel. BBC1 is a recent candidate tumoursuppressor gene of breast cancers which express it less strongly than benign fibroadenomas (Cleton-Jansen et al, 1995). Homologues have been identified in a wide range of species (Adams et al, 1992;Helps et al, 1995), but the function of the protein is still unknown. The fact that only 7% of prostate cancer samples of this series with LOH near BBC1 also had mutations is nevertheless not in favour of an important role of this gene in prostate cancer.
Some of the investigated chromosomal loci near RB, DCC, TP53 and at D13S325 (13q21Ð22) were found to be more often affected by LOH in grades 2 and 3 compared with grade 1. TSGs close to these sites might, therefore, be important for tumour progression. Fitting in with this view, a suppression of tumorigenicity of prostate cancer cell lines DU-145, TSU and PC-3 which contain mutated Rb or TP53 genes has been achieved upon introduction of the normal genes (Bookstein et al, 1990;Isaacs and Carter, 1991).
A comparison of the 47 carcinomas finally makes evident that each tumour actually has its own pattern of allelic losses when evaluating all 25 loci. Although the LOH typing carried out in this study is far from complete, it is nevertheless tempting to speculate that different combinations of genetic events could result in similar malignant phenotypes of prostate carcinoma, as has also been suggested for other tumours (for review see Macdonald and Ford, 1997).