Correlation between genetic alterations and growth of human malignant glioma xenografted in nude mice

In order to develop preclinical models of malignant astrocytomas and oligodendrogliomas, a series of 54 resected gliomas (37 from oligodendroglial lineage and 17 from astrocytic lineage) were xenografted subcutaneously into nude mice. Molecular alterations commonly observed in gliomas subtypes, including LOH 1p and 1q, LOH 19q, LOH 10p and 10q, LOH 9p, TP53 and PTEN mutations, EGFR amplification, CDKN2A homozygous deletion and telomerase reactivation were systematically screened in the original and xenografted tumours. In all, 23 gliomas grew in nude mice. The most anaplastic tumours were selected as shown by pathological and molecular studies of the original tumour as well as shorter survival in patients whose tumours were successfully grafted. Comparison between the two growth profiles showed that 10q LOH and EGFR amplification gave a tumorigenic advantage. With a few exceptions, the genetic pattern was remarkably stable before and after growth in nude mice. These results suggest that subcutaneous xenografts are useful and reproducible models to analyse the molecular profile of malignant astrocytoma and oligodendroglioma. This represents the first step to improve our understanding of the correlations between molecular alterations and response to standard or experimental therapies.

Over the last decade, important advances have been made in the understanding of the molecular pathways involved in the progression of astrocytomas and oligodendrogliomas . Recent data also suggest that acquired genetic alterations, such as LOH 1p -19q, could be important predictors of prognosis and response to therapy (Cairncross et al, 1998;Smith et al, 2000;Ino et al, 2001). Unfortunately, the lack of clinically relevant models of gliomas seriously affects fine analysis of the correlation between genotype and treatment response. This is particularly clear for oligodendrogliomas, which are very difficult to maintain in in vitro cultures (Merrill and Matsushima, 1988;Post and Dawson, 1992;McLaurin et al, 1995). In this setting, wellcharacterised xenografts of gliomas could be of help, as specified by the recommendations of the European Medicines Evaluation Agency (EAEMP, 1999) and the EORTC Laboratory Research Division (EORTC LRD, 2001). However, it remains to be demonstrated that astrocytic and oligodendroglial tumours can grow reproducibly in nude mice while maintaining the same profile of genetic alterations as in the original tumour.

Tumour classification
A set of 54 gliomas was surgically resected at the Pitié-Salpêtrière hospital, Paris and the Centre Hospitalier Régional, Nancy. An informed consent was obtained for all patients. After pathological review according to the current WHO guidelines (MK) (Kleihues and Cavenee, 2000), except for the addition of a recently individualised subgroup of glioblastoma with an oligodendroglial component (GBMO) (He et al, 2001), the tumours were classified as oligodendroglioma (one O), mixed gliomas (two OA), anaplastic oligodendroglioma (29 AO), anaplastic oligoastrocytoma (one AOA), four GBMO and finally a subset of 17 glioblastoma (GBM) that was used as a control group.

Xenograft protocol
Each tumour was xenografted subcutaneously into the scapular area. Following a maximal delay of 2 h after surgical resection, small fragments (B50 mm 3 ) of tumour tissue were xenografted onto at least three anaesthetised nude mice (Swiss nu/nu) aged 6 -8 weeks. Before xenograft, tumours were kept in foetal calf serumfree DMEM (Sigma-Aldrich, St Louis, MO, USA). The mice were maintained under clean room conditions and received sterile rodent food and water ad libitum. Their care and housing were in accordance with institutional guidelines as put forth by the Ministère de l'Agriculture et de la Forêt, Direction de la Santé et de la Protection Animale, Paris, France and the standards required by the UKCCCR guidelines (Workman et al, 1998).
The mice were kept alive for a maximum of 12 months until the growth of the xenografts. The tumours were measured once a week in two perpendicular dimensions and their volumes were estimated using the formula (width) 2 Â (length)/2. Latency was defined as the time between first transplantation and the appearance of a palpable tumour. The doubling time was calculated once it became stable, that is, usually after the fifth passage. The mice were killed when the tumour volume reached 2000 mm 3 and the xenografts were maintained by serial transplantation.

DNA EXTRACTION
Paired blood and tumours could be obtained for 31 initial tumours, blood could not be obtained for 12 tumours, and unfortunately tumour samples were not fresh frozen in 11 cases. Blood DNA was extracted using the Nucleon BACC3 DNA Extraction kit (Amersham Bioscience, Piscataway, NJ, USA) and tumoral DNA was extracted using the QIAamp DNA minikit, as described by the manufacturer (Qiagen, Venlo, NL, USA).
Screening of the PTEN/MMAC1 and TP53 gene mutations PTEN/MMAC1 mutations were screened by the denaturing gradient gel electrophoresis (DGGE) technique in the entire coding sequence of the nine exons and their corresponding splice junctions using previously described primers (Zhou et al, 1999). TP53 mutations were screened by the DGGE technique for exons 5 -8 and their using previously described primers (Hamelin et al, 1994). DNA showing altered DGGE profiles were sequenced bidirectionally using the Perkin Kit and sequencer. When a DNA variant was found, the corresponding blood DNA was sequenced in order to differentiate somatic events from constitutional variants (polymorphism or germline mutation).
Screening of EGFR gene amplification and P16/CDKN2A gene homozygous deletion EGFR amplification was screened by semiquantitative PCR using primers and protocol described previously (Hunter et al, 1995). P16/CDKN2A homozygous deletions were screened by semiquantitative PCR using primers and protocol previously described (Walker et al, 1995).

Screening of telomerase activity
Telomerase activity was screened by Telomeric Repeat Amplification Protocol (TRAP) technique using the TRAPeze kit (Intergen Co, Purchase, NY, USA) as described by the manufacturer.

Statistical. analyses
Descriptive statistics for continuous variates are provided as mean, standard deviation, median, minimum and maximum. For categorical variates, frequency distribution, median, minimum and maximum are provided.
The relation between categorical variates is described using the w 2 method, or Fisher's exact test when the w 2 method is not appropriate.
Prognostic factors for tumour growth delay are identified using univariate analysis (log rank test) and multivariate analysis fitting Cox's proportional hazard regression models. All tests are considered significant at the 0.05 significance level. Odds ratios are presented with their 95% confidence interval.

Patient characteristics and relationship with tumour growth
The 54 surgical samples were obtained from 33 men and 21 women, with a median age of 57. 4 years (range 20 -77). The postop median survival was 11.03 months after a median follow-up of 31.9 months (range 1 -37.2). Out of 54 tumours, 23 were established in nude mice (42.2%). In total, 14 were derived from oligodendroglial tumours (one AOA, nine AO, four GBMO) and nine from GBM. When a tumour was tumorigenic, it grew on at least 60% of mice at the first transplantation and on 80 -100% mice on later transplantations.
None of the three low-grade gliomas were established in nude mice, while 23 out of 51 high-grade gliomas did (45%).

Histology after xenograft
After xenograft in nude mice, tumours presented a dedifferentiated phenotype devoid of any particular pattern with small round or fusiform cells. Grafted oligodendrogliomas displayed a complete loss of 'honeycomb' appearance ( Figure 1B, D). The cellular density was very high. Neither endothelial hyperplasia nor endothelial proliferation could be observed. Many mitotic and apoptotic figures were seen in the same samples, while central areas of necrosis were almost constant. Tumours lost their glial differentiation whatever their initial type (oligodendroglial or astrocytic), except one tumour with an initial morphology of mixed anaplastic glioma, which exhibited an oligo-like differentiation after xenograft (ABE).

Genetic alterations of primary tumours and relationship with tumour growth
As summarised in Table 1   chromosomes 1p, 1q, 19q, 10p, 10q, 9p was performed for 31 tumours and for telomerase activity in 22 tumours. Tumours were separated into two groups. The first one consisted of tumours that grew in nude mice for at least two transplantations (n ¼ 19) and the second group of those that did not (n ¼ 24). Comparison of the two groups showed that growth in nude mice was correlated to 10q LOH and EGFR amplification (P ¼ 0.01 and 0.0035, respectively). In addition, when genetic alterations linked to anaplasia were pooled (chromosome 10 loss, chromosome 9p loss, EGFR amplification, p16 deletion, PTEN mutation and telomerase reactivation), nongrowing tumours showed significantly fewer alterations than growing tumours (P ¼ 0.029). Among the whole series, no alteration could be correlated to a nontumorigenic feature.

Genetic alterations after establishment in nude mice
Comparison of genetic alterations was made between primary tumours (at the time of surgery) and first transplantations in nude mice, and then between the first and subsequent transplantations. Tumours showed a striking genetic stability before and after establishment in mice (Table 2). No alteration disappeared after xenograft and only two new alterations appeared in xenografts (a 1p loss for ODA-4-GEN and a PTEN homozygous deletion for ODA-20-THO). After the first transplantation, the molecular profile of alterations remained constant.

Correlations between spontaneous growth rate and genetic alterations
The latency (time between the first transplantation and the appearance of a palpable tumour) and the doubling time for 12 growing tumours are summarised in Table 3 showing that even if latency differed between tumours, the doubling time remained relatively stable. For the different genetic alterations tested, multivariate analysis showed that only EGFR amplification was associated with a higher growth rate (P ¼ 0.0082).

DISCUSSION
These data indicate that human xenografts are useful models to study the molecular pathways involved in malignant oligodendrogliomas and astrocytomas. However, successful grafting occurred only in anaplastic tumours in terms of pathological grading of the original tumours, genotype analysis and prognostic significance for donor patients.
The results of the molecular analysis of the primary tumour (before grafting) are in agreement with the pathological data. Overall, growing tumours had significantly more molecular alterations than the nongrowing ones. This finding is particularly striking for EGFR amplification and LOH on chromosome 10, two alterations highly associated with increased malignancy (Bigner and Vogelstein, 1990;von Deimling et al, 1992;Reifenberger et al, 1996). It has previously been shown that EGFR amplification gives a tumorigenic advantage in GBM (Humphrey et al, 1988;Muleris et al, 1994), a feature that we often found associated with stable double-minute extra-chromosomal elements on caryotypic analy-  Tumour  P  Q  Q  P  P  Q  TP53  PTEN  EGFR  CDKN2A  Telomerase activity   RAI  LOH  LOH  LOH  -LOH  -mut  ---Absent  ODA-5-RAI p1  LOH  LOH  LOH  -LOH  -mut  - RAI, RAV, GEN, THO ¼ initial tumours. ODA-4-GEN, ODA-14-RAV, ODA-4-GEN, ODA-20-THO ¼ xenografted tumours; p x ¼ passage X; LOH ¼ loss of heterozygosity; mut ¼ mutation; amp ¼ genomic amplification; hd ¼ homozygous deletion; ND ¼ not done. sis (data not shown) (Bigner et al, 1989). In addition, EGFR amplification was present in 4/11 successfully grafted oligodendrogliomas in our series, indicating that the tumorigenic role of this alteration affects various subtypes of gliomas. Interestingly, overexpression of EGFR in transgenic mice has been shown to be involved in tumour progression of oligodendroglioma (Ding et al, 2003). Similarly, we found that LOH on chromosome 10 was also extremely frequent not only in GBM (100%, 7 out of 7) but also in growing oligodendrogliomas (73%, eight out of 11). Thus, the loss of the putative tumour suppressor gene located on chromosome 10 is important for tumorigenicity in both tumour subtypes. However, we did not find evidence of involvement of the PTEN gene, which is located on chromosome10, in this series.
The fact that pathological and molecular factors predictive of successful grafting were associated with higher malignancy explaining that survival was shorter in donor patients whose tumours grew in nude mice (9.1 months) as compared to survival of patients whose tumours were rejected (15.4 months) (P ¼ 0.0003).
An important characteristic of this model is the stability of the genetic alterations in the xenografts as compared to the primary tumours, a finding also observed after successive passages in nude mice. Such stability, which was somewhat controversial in previous studies (Bigner et al, 1989;Muleris et al, 1994;Goike et al, 1999;Jeuken et al, 2000), appears to hold both for GBM and anaplastic oligodendrogliomas. Nevertheless, rare differences before and after growth on nude mice were shown as illustrated in Table 2. Rather than selection of a minor subclone, it is likely, as shown in Figure 2, that these alterations were already present in the original tumour, but were hardly visible or even undetectable because of the presence of 'contaminating' normal tissue.
An advantage of this model is therefore to obtain tumours devoid of normal cells, improving the quality of molecular analyses (Figure 2).
Despite a stable genetic profile, the phenotype of the grafted tumour cells underwent striking changes whatever the histological features of the original tumour. After xenograft, all but one glioma lost their glial differentiation and presented a dedifferentiated phenotype in contrast with xenografts derived from other human cancers that maintained several of their histological characteristics (Rofstad et al, 1990a, b;Poupon et al, 1993;Kolfschoten et al, 2000;Bras-Goncalves et al, 2001;de Pinieux et al, 2001;Krasagakis et al, 2001).
Whether the morphological homogeneity of grafted tumours represents a nonspecific change or supports the view that many gliomas have a common cell of origin (Daumas-Duport et al, 1997) leading to various morphological appearance in patients remains unsettled.
In summary, xenografting malignant astrocytomas and oligodendrogliomas is a useful method for studying and probably refining knowledge of the spectrum of genetic alterations in these tumours. This method can be of help in analysing the correlations between genotype and response to chemotherapy or various experimental agents. In addition, it may further benefit from new approaches such as gene expression by microarray studies in order to identify genes associated with chemosensitivity.

Allele 1
Allele 2  Figure 2 Loss of heterozygosity patterns before and after establishment on nude mice. Genescan patterns showed a hardly detectable LOH in the initial tumour, which in contrast appeared very clearly after the first passage onto nude mice.