Allele loss on chromosomes 10 and 17p and epidermal growth factor receptor gene amplification in human malignant astrocytoma related to prognosis.

Patients with high-grade astrocytomas have a poor prognosis. However, considerable variation exists within this group of patients with respect to post-operative survival. In order to determine whether genetic alterations might be of help in subdividing this group, we used allele loss on chromosomes 10 and 17p and epidermal growth factor receptor (EGFR) gene amplification in the tumours as genetic parameters and determined their prognostic value. A series of 47 malignant (grade III and grade IV) tumours were genetically characterised, and four types of tumours were found. Type 1 tumours had loss of heterozygosity on chromosome arm 17p (LOH 17p) as the sole genetic alteration. Patients with this type of tumour were relatively young (mean age 39 years) and had a median survival period of 17 months. Type 2 tumours displayed only allele loss on chromosome 10 (LOH 10), type 3 tumours had LOH 10 + LOH 17p and type 4 tumours contained LOH 10 + EGFR gene amplification. Patients with types 2, 3 and 4 tumours were older (mean ages 59, 65 and 54 years respectively) and had a shorter survival (median duration 6, 3 and 2 months respectively) than type 1 patients. Multivariate analysis indicated that the genetic subdivision was a significant prognostic variable. In this study, age proved to be of minor importance with regard to survival. Our study revealed a predominance of frontally located tumours in patients with type 1 tumours, i.e. with LOH 17p only.

Finally. amplification of several proto-oncogenes. including MYCN. GLI and MDMV2. and the anonymous marker D17S67 has been noted in a low percentage of high-grade astrocytomas (Collins & James. 1993: Bijlsma et al.. 1994. Although. in general. prognosis is poor for patients with high-grade astrocytoma. considerable variation exists in this group of patients with respect to post-operative survival. The genetic alterations found in high-grade astrocytoma might be of help in defining subsets of tumours that correlate with clinical outcome. Recently. using LOH 10. LOH 17p and EGFR gene amplification as genetic parameters. Von Deimling et al. (1993) proposed a genetic subdivision of glioblastoma multiforme that correlates with mean age of the patient. One subset was characterised by the presence of LOH 17p and the absence of EGFR gene amplification. This type of tumour was primarily found in younger patients. The other subset had EGFR gene amplification, but no LOH 17p. and was primarily found in older patients. LOH 10 was not useful in subdividing the tumours in their study. It is well documented that for high-grade astrocytoma a negative relationship exists between advancing age and period of postoperative survival (Burger & Green. 1987: Salford et al.. 1988. Taken together. these data suggest that patients with LOH 17p but no EGFR gene amplification have a better prognosis than those with EGFR gene amplification and no LOH 17p.
The objective of the present study was 2-fold. Firstly. we wanted to determine whether the genetic subdivision cited above also applies to the series of high-grade astrocytomas that we studied. Secondly. considering the strong influence of age on survival, we wanted to determine whether such a genetic subdivision has a prognostic significance that is independent of age.

Materials and methods
Tumour samples Tumour samples were obtained from patients undergoing brain surgery at the Academic Medical Centre in Amsterdam. Two specimens (cases 534 and 543) were obtained from the Department of Neurosurgery of the University Hospital of Utrecht. The Netherlands. All tumour samples were primary lesions that were removed prior to radiotherapy. Only homogeneous high-grade tumours (grades III and IV) were used for RFLP (restriction fragment length polymorphism) analysis. except for tumour samples 316. 449 and 673. These samples represent high-grade parts of heterogeneous high-grade tumours that were previously molecularly analysed (Leenstra et al., 1992). The molecular data for tumours 1197. 1672 and 1683 have been published before (Bijlsma et al.. 1994). For DNA extraction. tumour tissue samples were quickly frozen in liquid nitrogen and stored at -80°C until use. Peripheral venous blood samples were collected prior to surgery. Routine histopathological examination was performed on H&Eand Gomori-stained sections of paraffin-embedded blocks of formalin-fixed material and compared with frozen specimens. Tumours were graded according to Kernohan and Sayre (1952).

Clinical data
The clinical data were obtained by reviewing clinical and out-patient records. Survival data were obtained either from clinical records or by written information from general practitioners. Following surgerv. all patients were treated according to a standardised fractionated radiotherapy protocol (60 Gy). None of the patients received chemotherapy.

Genetic alterations
The study included 47 high-grade tumours. These were, according to Kernohan and Sayre (1952). ten grade III and 37 grade IV astrocytomas. For the purpose of this investiga-tion thev wvere considered as one group because. in the Kernohan system. patients with grade III and grade IV tumours do not differ in median post-operative survival (Scanlon & Taylor. 1979: Daumas-Duport et al.. 1988: Revesz et al.. 1993. The genetic alterations that were found and the relevant clinical data of the individual patients are shown in Table I. LOH for chromosome 10 markers was detected in 29 (62%) tumours. Most tumours displaved LOH for all informative markers. indicating loss of a whole copy of chromosome 10. Deletion of part of chromosome 10 was apparent in seven tumours. The LOH profile for each of these tumours is depicted in Figure 1. Recent reports (Ransom et al.. 1992: Rasheed et al.. 1992: Fults & Pedone. 1993: Karlbom et al.. 1993 suggest the presence of at least three regions of common deletions on chromosome 10. i.e. in the telomeric part of 10p and in the centromeric and telomeric parts of 10q. However. excluding tumour 1208. our series of tumours with partial deletions indicates that the central region of chromosome 10. between loci D10S17 and D10S85. is critical for astrocytoma tumorigenesis. In tumour 1208. allele loss was restricted to the telomeric 10q marker DlOS285. which is most probablv outside the critical region (cf. Ransom et al.. 1992: Rasheed et al.. 1992: Fults & Pedone. 1993. For this reason. case 1208 was considered as having no LOH for the critical region on chromosome 10 in the statistical analysis that follows. LOH for chromosome arm 17p A-as found in 15 (32%) tumours. The LOH data for nine tumours suggested loss of the whole p arm. In the remaining six tumours deletions of part of chromosome 17p were seen. Except for tumour 314.
the deletion in all these tumours included the tumoursuppressor gene TP53 in chromosomal band 17pl3.1. Both copies of the TP53 gene were retained in tumour 314. This is demonstrated in Figure 2. Figure 2a shows hybridisation of the TP53 cDNA probe pProSP53 to BglII-digested tumour and coresponding leucocyte DNA. The 12 and 9kb allelic fragments remain in the tumour. Figure 2b shows the same Southern blot. but in this case hybridised with probe LEW502. identifying locus D17S66. which is distal to TP53 in chromosomal band 17plI3.1. The 2.5 kb allele at locus D17S66 is clearly absent in the tumour. LOH was also found at D17S68 in chromosomal band 17pl3.2 but not at ten other informative loci on 17p. including D17S5 and D17S28 at the tip of 17p (data not shown). We could not detect any sequence abnormality in exons 5-8 of the TP53 gene. in which most of the known TP53 gene mutations have been found (Hollstein et al.. 1991). Therefore. as the interstitial deletion in tumour 314 most probably did not affect the TP53 gene. the tumour was considered as having no LOH 17p for the TP53 gene region in the statistical analysis that follows. The deletion of both copies of the TP53 gene in tumour 1197 and LOH for the TP53 gene region in tumour 1672 have been reported previously (Bijlsma et al.. 1994). Significant EGFR gene amplification (more than five times the level in normal tissue) wsa seen in nine (19%) tumours.
Amplification levels ranged from eight to 33 times (Table I). Although EGFR gene amplification is usually found in a higher proportion of high-grade astrocytomas. a similar low percentage has been reported in another study (Venter & Thomas. 1991). In accordance with earlier observations (Von Deimling et al.. 1992b). EGFR gene amplification almost exclusively (in eight of nine cases) occurred in tumours that also displayed LOH 10.
Genetic alterations in relation to patients'age at operation Based on the genetic alterations presented in Table I. we defined four groups of patients. These were patients with only LOH 17p (n = 7). with only LOH 10 (n = 14). with LOH 10+ LOH 17p (n = 6) and with LOH 10+ EGFR gene amplification (n = 7). Two patients with genetic alterations (cases 1672 and 1683) did not fit into any of these groups and were excluded from further analysis. Eleven patients had none of the studied genetic alterations in their tumours.
Genetic alterations in relation to post-operative survival Kaplan -Meier survival curves were constructed for the four groups of patients (Figure 3). The median survivals are given in Table II. The survival curves were compared by log-rank test. The resulting P-values were adjusted for multiple comparisons by the procedure of Hommel (see Wright, 1992). The difference in median survival between the group of patients with only LOH 17p and the group of patients with LOH 10 or with LOH 10 + LOH 17p was marginally significant (P-values of 0.055 and 0.072 respectively). The difference in survival between the LOH 17p only patients and the patients with LOH 10 + EGFR gene amplification was significant (P = 0.039). The groups of patients with LOH 10 only, LOH 10 + LOH 17p and LOH 10 + EGFR gene amplification did not differ significantly in median survival between each other (P-value for each comparison>0.17).
Genetic alterations in relation to other clinical data All tumours (n = 7) with LOH 17p and no other genetic alterations occupied a frontal (frontal or frontoparietal) position (Table I). This suggests a significant (P<0.0001, x2 test) predominance of frontally located tumours in the group of patients with LOH 17p as the sole genetic alteration in the tumour.

Multivariate analysis
The variables age, grade, duration of symptoms, tumour location, extent of tumour resection and genetic subdivision were analysed for prognostic significance by univariate analysis. The variables grade, duration of symptoms, tumour location and extent of tumour resection did not reach significant levels in this analysis. To explore the effect of age and genetic subdivision on survival, Cox's regression model was used. Elimination of the genetic subdivision was a significant step (P = 0.026); elimination of age was not significant (P = 0. 17).
By using LOH 10, LOH 17p and EGFR gene amplification as genetic parameters, we have subdivided our senres of highgrade astrocytomas into four types. Type 1 has LOH 17p without concurrent genetic changes, type 2 has LOH 10 as the sole genetic alteration, type 3 is characterised by both LOH 10 and LOH 17p and type 4 is characterised by LOH 10 and EGFR gene amplification. The mean age at operation was significantly lower for the patients with the type 1 tumours than for the patients with types 2, 3 and 4 tumours (  Figure 2 RFLP analysis at loci TP53 and D17S66 in tumour 314. Tumour (T) and corresponding blood leucocvte (L) DNAs were digested with BgilII. The resulting fragments were separated in an 0.8% agarose gel, transferred to a nylon filter and hybridised with pProSP53 (TP53, a). After removal of the signal. the filter was rehybridised with LEW502 (D17S66. b). Sizes of restriction fragments are given in kbp.

Survival in months
Fgre 3 Kaplan-Meier survival curves for patients whose tumours contained only LOH 17p (U. type 1). only LOH 10 (--. type 2). LOH 10 + LOH 17p ( . type 3) or LOH 10 + EGFR gene amplification ( -. tWpe 4). tumours do not match completely because of the different grading systems that were used (glioblastomas multiforme and grade III + IV astrocytomas). Our type 1 patients differed from the type 2. 3 and 4 patients with regard to their median surv,ival time. although the difference was only significant (P = 0.039) between the type 1 and type 4 patients (Table II and Figure 3). Type 1 patients differ from the type 2. 3 and 4 patients in mean age at operation. and age is an important variable with regard to survival (Burger & Green. 1987: Salford et al.. 1988. Therefore. the longer survival in the type 1 patients could be explained by their younger age. However. multivariate analysis revealed that. after allowing for age. the genetic subdivision remained a variable with prognostic significance.
In this analysis, age was assumed to be linearly related to the logarithm of the hazard rate. If this assumption is not correct. then. owing to the lack of overlap in age between the type 1 patients and the other groups of patients. no adjustment for age would be possible.
Comparable studies with regard to survival in patients with high-grade astrocytomas and with the same set of genetic alterations have, to our knowledge. not been published. Bigner et al. (1988) studied EGFR gene amplification in malignant gliomas but could not detect a significant effect on median survival. In contrast to this, Hurtt et al. (1992) reported a significant and age-independent decrease in median survival in patients with EGFR gene amplification. If we consider EGFR gene amplification as an independent variant in Cox's regression model. then it has age-corrected prognostic significance (P = 0.004). However, as noted by Von Deimling et al. (1992b) and as shown in Table I. EGFR gene amplification is strongly associated with LOH 10. Our type 4 patients with LOH 10 + EGFR gene amplification had a similar mean age to the type 2 patients with LOH 10 onlv (59 and 54 years respectively). The difference in survival between both groups of patients is not significant (P = 0.17). thereby questioning the prognostic importance of EGFR gene amplification. Thus, the contradictory reports on the prognostic importance of EGFR gene amplification could be caused by differences in the chromosome 10 status of the tumours that were examined. In all our type 1 patients. the tumour occupied a frontal position. As noted before, these patients were relatively young at presentation and had a favourable prognosis. This observation genetically supports other clinical studies reporting a longer sur'Vival time for patients with frontally located lesions (North et al.. 1990;Curran et al., 1992;Ayoubi et al.. 1993).
In summary. our data suggest that histopathologicallv high-grade astrocytomas constitute a genetically heterogeneous group of tumours. They can be subdivided by using allele loss on chromosomes 10 and 17p and EGFR gene amplification as genetic parameters. The genetic subdiv%ision into four types correlates with mean age and median survival of the different groups of patients. Under certain assumptions, the genetic subdivision has major prognostic significance. The discrepancies between the results presented here and those of others may in part be explained bv the small sample sizes in the studies.