Molecular characterization of a novel amplicon at 1q21-q22 frequently observed in human sarcomas.

In a recent comparative genomic hybridization (CGH) study of a panel of sarcomas, we detected recurrent amplification of 1q21-q22 in soft tissue and bone tumours. Amplification of this region had not previously been associated with sarcoma development, but occasional amplification of CACY/S100A6 and MUC1 in 1q21 had been reported for melanoma and breast carcinoma respectively. Initial screening by Southern blot analysis showed amplification of S100A6, FLG and SPRR3 in several sarcomas and, in a first attempt to characterize the 1q21-q22 amplicon in more detail, we have now investigated the amplification status of these and 11 other markers in the region in 35 sarcoma samples. FLG was the most frequently amplified gene, and the markers located in the same 4.5-Mb region as FLG showed a higher incidence of amplification than the more distal ones. However, for most of the 14 markers, amplification levels were low, and only APOA2 and the anonymous marker D1S3620 showed high-level amplifications (> tenfold increases) in one sample each. We used fluorescence in situ hybridization (FISH) to determine the amplification patterns of two overlapping yeast artificial chromosomes (YACs) covering the region between D1S3620 and FLG (789f2 and 764a1), as well as two more distally located YACs in nine selected samples. Six samples had amplification of the YAC containing D1S3620 and, in three, 764a1 was also included. Five of these tumours showed normal copies of the more distal YACs; thus, it seems likely that an important gene may be located within 789f2, or very close. Two samples had high copy numbers of the most distal YACs. Taken together, FISH and molecular analyses indicate complex amplification patterns in 1q21-q22 with at least two amplicons: one located near D1S3620/789f2 and one more distal. ImagesFigure 1Figure 2Figure 3

Summary In a recent comparative genomic hybridization (CGH) study of a panel of sarcomas, we detected recurrent amplification of 1q21-q22 in soft tissue and bone tumours. Amplification of this region had not previously been associated with sarcoma development, but occasional amplification of CACY/S100A6 and MUC1 in 1q21 had been reported for melanoma and breast carcinoma respectively. Initial screening by Southern blot analysis showed amplification of S100A6, FLG and SPRR3 in several sarcomas and, in a first attempt to characterize the 1q21-q22 amplicon in more detail, we have now investigated the amplification status of these and 11 other markers in the region in 35 sarcoma samples. FLG was the most frequently amplified gene, and the markers located in the same 4.5-Mb region as FLG showed a higher incidence of amplification than the more distal ones. However, for most of the 14 markers, amplification levels were low, and only APOA2 and the anonymous marker D1S3620 showed high-level amplifications (> tenfold increases) in one sample each. We used fluorescence in situ hybridization (FISH) to determine the amplification patterns of two overlapping yeast artificial chromosomes (YACs) covering the region between Dl S3620 and FLG (789f2 and 764a1), as well as two more distally located YACs in nine selected samples. Six samples had amplification of the YAC containing DlS3620 and, in three, 764a1 was also included. Five of these tumours showed normal copies of the more distal YACs; thus, it seems likely that an important gene may be located within 789f2, or very close. Two samples had high copy numbers of the most distal YACs. Taken together, FISH and molecular analyses indicate complex amplification patterns in 1q21-q22 with at least two amplicons: one located near Dl S3620/789f2 and one more distal.
Keywords: amplification; chromosome 1; 1q21-q22; sarcomas Cytogenetic studies have demonstrated recurrent aberrations of chromosome 1, including deletions, translocations, trisomies and amplifications, in solid tumours as well as in haematological diseases (Dracopoli et al, 1994;Weith et al, 1996). Alterations of the long arm of chromosome 1 are found both in leukaemias and in solid tumours, and are among the most common chromosomal anomalies in human neoplasia. The aberrations can be seen as trisomy of the entire long arm (Oshimura et al, 1976), as an isochromosome lq (Kovacs, 1978) or as trisomy or duplication of a smaller region, especially lq23-lq32 (Rowley, 1977). It has been suggested that three or more copies of a gene (or genes) in this region provide a selective advantage to cancer cells. Furthermore, the finding of partial or complete lq trisomy being more frequent in recurrent than in primary tumours could suggest that this change may be associated with tumour progression (Weith et al, 1996).
Among solid tumours, I q alterations had previously been reported for breast, lung and germ cell tumours. In breast cancer, comparative genomic hybridization (CGH), a method by which whole genomes may be surveyed for DNA sequence copy number changes (A Kallioniemi et al, 1992, has demonstrated frequent gains of the whole long arm of chromosome 1 (A Muleris et al, 1994). However, there have been few reports on amplification of specific genes located to chromosome 1. Amplification of the MYCLI gene at lp32 seems to be a common aberration in small-cell lung cancer (Nau et al, 1985;Makela et al, 1991) but has not been found in other tumour types (Dracopoli et al, 1994;Weith et al, 1996). Some melanoma cell lines show low-level amplification (or duplication) of CACY/SJOOA6', encoding calcyclin (Weterman et al, 1992). This gene is a member of the SIOOA family of calciumbinding proteins located within a cluster of SIOOA genes in lq21 (Schafer et al, 1995;Schafer and Heizmann, 1996;Maelandsmo et al, 1997). Another gene in lq21-q22, MUCI, coding for the epithelial tumour-associated antigen mucin 1 (Tsarfati et al, 1990), was found to be amplified in some breast cancers (Bieche et al, 1995;Bieche and Lidereau, 1997).
We recently studied DNA amplification in soft-tissue and bone sarcomas by CGH (Forus et al, 1995a,b) and found amplifications in about 50% of the tumours analysed. The amplicons most frequently observed were the well-characterized one at 12q I 3-q 15 and a novel one at lq21-q22. At the same time, gain of lq was reported in some osteosarcomas, but was not considered a common abnormality of such tumours (Tarkkanen et al, 1995). We found that amplicons at lq2l1-q22 were more frequent than those at 12ql3-q15, which have been detected in a substantial number of the sarcomas analysed (Forus et al, 1993(Forus et al, , 1994Khatib et al, 1993;Maelandsmo et al, 1995;Nilbert et al, 1995;Berner et al, 1996). These observations could indicate that increased copy numbers of one or more of the genes in the 1q21-q22 region could play a role in the development or progression of at least some sarcoma subtypes. Gains of lq material have also been reported by other investigators as an occasional event in rhabdomyosarcomas (Weber-Hall et al, 1996) and more frequently in liposarcomas (Szymanska et al, 1996 and Ewing's sarcomas (Armengol et al, 1997), often involving lq21-q22 as a minimal common region of gain. Recently, Larramendy et al (1997) reported infrequent gain of 1q material in malignant fibrous histiocytomas (MFHs), but with a more distal minimal common region, 1q24.
It might be significant that many of the tumours with amplifications of 1q are well-differentiated liposarcomas (WDLPS) and MFHs tumours that are often characterized by giant marker chromosomes containing amplified segments from various chromosomes (Heim et al, 1987;Orndal et al, 1992;Dal Cin et al, 1993;Pedeutour et al, 1993;Nilbert et al, 1994). In most WDLPS, such marker chromosomes carry segments from chromosome 12, correlating with MDM2 amplification (Pedeutour et al, 1994), but we have recently found that some have markers that also carry chromosome 1 material, and in most of these amplification of Iq21-q22 could be demonstrated by CGH (Pedeutour et al, 1998).
In an attempt to characterize the 1q21 -q22 amplicon in sarcomas, we have analysed the amplification pattern of SIOOA6 and ten other markers within the same 4.5 Mb region of lq2 I-q22, in addition to MUCI and two genes from other parts of the region, in 35 sarcomas previously studied by CGH. We have also used fluorescence in situ hybridization (FISH) to determine chromosome 1 and 2 copy numbers, as well as amplification patterns of four YAC clones from the 1q21-q22 region.

MATERIALS AND METHODS Specimens
Thirty-five human sarcomas of various subtypes were analysed. In a previous study, 16 of these had been found to have amplification of lq2 l-q22 by CGH, whereas the other 19 had not (Table   1 ) (Forus et al, 1995a,b). The samples were obtained directly from 21 patients with sarcomas of various subtypes, from 13 different sarcoma samples grown subcutaneously as xenografts in nude mice and from the liposarcoma cell line SW872 (ATCC). All tumours were classified according to the WHO International Histological Classification of Tumours (Schajowich, 1993;Weiss, 1994). The tissues were cut in small pieces and frozen in liquid nitrogen immediately after surgery and were stored at -70°C. Blood samples from healthy individuals were used as controls.
Southern blot analysis DNA extraction from tumour samples, preparation of filter blots and hybridization was performed as described previously (Forus et al, 1993). First, amplification patterns of ten different markers and genes were analysed: FLG, IVL, SPRR3, SPRR]B, SPRR2A, SOOA6 (CACY) and SIOOA2 (SIOOL), all of which are physically mapped within 2 Mb and are part of a 6-Mb YAC contig in 1q21 (Marenholz et al, 1996), as well as two markers that have been mapped distal to the 6-Mb contig in the order CRP-APOA2 and MUCI (Weterman et al, 1996; Chromosome 1 www page http://linkage.rockefeller.edu/chr I /). Based on the results obtained with these probes, we later included some additional anonymous clones that map proximal to FLG (DIS3620 and D1S3623) and distal to S10OA2 (D1S3625 and D1S3628) (Marenholz et al, 1996).
Southern blots were sequentially hybridized to probes from each locus and to a control probe from chromosome 2 (APOB) (Huang et al, 1985). Quantitation of signal intensity was achieved by two-dimensional densitometry on a Molecular Dynamics laser densitometer. The net signals from specific bands were corrected for unequal sample loading by calibration relative to the signal obtained with the APOB control probe. Mean signals from three (or more) different blots were used to measure the amplification levels in the tumour. As the percentage of tumour cells versus normal cells for each sample was not known, the actual amplification levels in the tumour cells may be higher than the presented values.
The signals were compared with signals from control samples with a normal karyotype (leukocytes) and interpreted as described. Borderline amplification -a signal two-to threefold more intense than signals from normal samples (i.e. average probe/APOB ratio for tumour divided by average probe/APOB ratio for normal sample between 2.0 and 2.9); low levelthreeto fivefold increase (tumour-normal value between 3.0 and 4.9); moderatefiveto tenfold increase (tumor-normal value between 5.0 and 9.9); and high -> tenfold increase (tumour-normal value 10 or higher).

Fluorescent in situ hybridization (FISH) to interphase nuclei Preparation of interphase nuclei
Frozen tumour tissue was pulverized in liquid nitrogen, transferred to a centrifuge tube and immediately fixed in 3:1 methanol-acetic acid. After centrifugation, the pellet was resuspended in 60% acetic acid. Two or three drops of the suspension were applied onto slides (Menzel Superfrost) prewarmed to 45-50°C, and left to dry at the same temperature. Slides were stored at -20°C before use.

In situ hybridization
Slides were thawed and immersed in 75% ethanol at 4°C for 1-2 h before use, air dried and denatured in 70% formamide/2 x SSC, pH 8.0 for 3 min at 74°C, washed three times in ice cold 2 x SSC, dehydrated in ethanol (70%, 90%, 96% and 100%) and air dried. Thereafter, slides were treated with proteinase K (0.1 ,ug ml-' in 20 mm Tris-HCl/2 mm calcium chloride, pH 7.0) for 10 min at room temperature, washed in 2 x SSC, dehydrated and air dried. Probes were denatured for 10 min at 80°C, prehybridized for 15-30 min at 37°C and applied to slides at room temperature. Hybridization was done overnight at 37°C. After hybridization, the slides were washed three times for 10 min in 50% formamide/ 2 x SSC at 45°C and then three times for 10 min in 2 x SSC at British Journal of Cancer (1998) 78(4), 495-503 . LS10 is a cell line. Some of the patients received chemoor radiotherapy before surgery. aThis patient received adjuvant chemotherapy after the primary tumour, which was excised 4 years before the removal of the sample analysed here. bThis was a therapy-induced osteosarcoma; the patient had Ewing's sarcoma 5 years before this tumour and received chemo-and radiotherapy.

Evaluation of results
Hybridized slides were examined visually using a Zeiss Axioplan microscope equipped with appropriate single-bypass filters for exitation of DAPI, FITC, Texas Red and rhodamine/CY3, and double bypass filters for excitation of DAPI/rhodamine and DAPI/FITC. The slides were manually scanned at 63x or lOOx magnification with DAPI excitation to localize the interphases.
Nuclei that were either partially or totally overlapping or not intact were not analysed. For each probe, the number of spots was counted in at least 150 nuclei.
'Probes cDNA and genomic probes The following probes were used: pHX5 FLG (Presland et al, 1992), containing a part of the coding region from the 3' end of the human filaggrin gene, kindly provided by Drs Fleckman and Presland; IVL (pI-2) (Eckert and Green, 1986) containing the 3' end cDNA of human involucrin, kindly provided by Drs Easley and Green; cDNA probes for the genes SPRRIB, SPRR2A and SPRR3 (Gibbs et al, 1993;Hohl et al, 1995), kindly provided by Dr Backendorf; a cDNA clone for human calcyclin (CACYI SJOOA6)(pMW1) (Weterman et al, 1992), kindly provided by Dr Bloemers, and CAN19; a near full-length cDNA probe for the human SJOOLISJOOA2 (Lee et al, 1992), kindly provided by Dr Sager. All these probes are located within the same 2-Mb region in 1q21 (Mahrenholz et al, 1996). Three other genes from Iq21-q22 were also checked: pCRP-5, a cDNA clone for human C-reactive protein (CRP) (Tucci et al, 1983); cDNA for apolipoprotein All (APOA2) (Rogne et al, 1989); and pMUClO, a genomic clone for British Journal of Cancer (1998) 78(4), 495-503 Figure 1 Representative Southern blot hybridizations demonstrating the amplification pattern of the FLG, SPRR1B, SPRR2A, SPRR3and S100A21oci (A), Dl S3620 and Dl S3623 (B) and APOA2 (C). DNA from each sample was digested with Hindlll and sequentially hybridized to probes as indicated to the right. Leucocyte DNA was included as a control for normal copy number, and a probe for APOB was used to calibrate for unequal sample loading. An 'A' below a panel indicates that the corresponding gene is amplified more than threefold, an (A) indicates borderline amplification (signal two to three times more intense than the signal from normal DNA). LS, liposarcoma; OS, osteosarcoma; MFH, malignant fibrous histiocytoma; MS, malignant peripheral nerve sheath tumour (malignant schwannoma) the epithelial tumour-associated antigen MUC1 (Swallow et al, 1987) (UK DNA probe bank). Anonymous probes from the 1q21 region that are located approximately 4 Mb apart were: D1S3620 and D1S3623, which are centromeric to FLG, and D1S3625 and D1S3628, which are telomeric to CACYIS]O0A6 (Marenholz et al, 1996). A cDNA probe for the APOB gene on human chromosome 2, kindly provided by Dr Breslow (Huang et al, 1985), was used to calibrate for unequal sample loading.

Yeast artificial chromosome (YAC) clones
We used two YACs from the Centre D'Etude du Polymorphisme Humain (CEPH) mega-YAC library, 789f2 and 764al (Marenholz et al, 1996). These YACs are part of a 6-Mb contig in 1q21 and cover the region from FLG to DIS3620 (Marenholz et al, 1996).
The two other YACs, 935b12 and 883h6, were from the CEPH library described by Albertsen et al (1990) and have been mapped to the 1q21-q22/q23 region (CEPH-Genethon map).

Southern blot analysis
Initial screening of selected tumour samples using probes for SJOOA6, FLG and SPRR3, which are located in the same 2-Mb region of 1q21-q22, detected amplifications at various levels in all the tumours tested and, therefore, these and four other markers in the region were analysed in more samples. We also included MUCI, as this gene has been reported to be amplified in breast cancer, and two other genes in the region, CRP and APOA2, for which probes were available. All the probes from 1q21-q22 detected amplification in some of the samples (Figures 1 and 2). Nine samples had more than a threefold increase in gene dosage of one or more of the genes British Journal of Cancer (1998) 78(4) (Marenholz et al, 1996;Schafer and Heizmann, 1996) and genetic (Dracopoli et al, 1994;Weterman et al, 1996) mapping data. Loci are listed in their order from the centromeric (left) to the telomeric (right) side. An empty column indicates > 200 kb distance between two loci. Between S100A2 and S100A6 there are at least three additional Si OOA genes that have not been tested in this study (Schafer et al, 1995;Schafer and Heizmann, 1996). The physical distances between S100A6, CRP and APOA2 are not known, but the genetic distance between S100A6 and CRP is around 9 cM (Murray et al, 1994), as discussed by Mahrenholz et al (1996). MUC1 has not been located relative to the other genes, but linkage analysis gives the order (centromere to telomere) MUC1-CRP-APOA2 (Dracopoli et al, 1994;Weith et al, 1996). The densitometrically determined levels of amplification are divided into four categories as indicated, based on average signals from at least three different blots. Tumour types and numbers are given to the left [LMS, leiomyosarcoma; LS, liposarcoma; MFH, malignant fibrous histiocytoma; MS, malignant peripheral nerve sheath tumour (malignant schwannoma); OS, osteosarcoma]. The number of samples with amplification (two to three fold increase or more) of each locus is listed below. The CGH column lists the previously detected 1 q21-q22 amplifications (Forus et al, 1 995a,b). A '-' sign = no 1q21-q22 amplification detected by CGH. ND, not done ( Figure 2) and ten other samples showed borderline amplification (twoto threefold increase). Among these were three samples in which no gain of lq21-q22 was detected by CGH (LS22, MFH3x and OS21). The remaining 16 tumours had a normal copy number of all the genes tested.
FLG, encoding human epidermal profilaggrin (Presland et al, 1992), was the most frequently amplified gene, but in most samples only borderline amplification of this gene was found. The genes localized close to FLG (IVL, SPRR3, SPRRIB, SPRR2A, SIOOA6 and SIOOA2), were amplified in fewer samples and were also at borderline or low levels in most cases. However, SPRR3, SPRR2A, SIOOA6 and SIOOA2 showed amplification levels above fivefold in some samples (LMS2x, LS 13 and MS8x). Three samples had amplification of all seven genes at variable levels (Figure 2, LMS2x, LS2 and LS 13). Among the loci that has not been mapped to the above-mentioned 6-Mb contig, CRP was the only one that was included in the amplicon in LS6, and APOA2 was the only gene amplified above tenfold in one of the liposarcomas (LS21). MUCI was amplified in five samples, i.e. in fewer cases than any of the other genes tested.
The region between FLG and SIOOA2 was more frequently amplified than loci located telomeric to this interval (MUCI, APOA2 and CRP). We therefore determined the amplification pattems also of some additional anonymous clones that map proximal to FLG (D1S3620 and D1S3623) and distal to SJOOA2 (DIS3625 and D1S3628), delineating a 4.5-Mb region in 1q21-q22 (Marenholz et al, 1996). Copy numbers of these probes were determined in those 19 samples that had previously revealed amplification of one or more of the genes. As shown in Figure 2, D1S3623, located proximal but close to FLG (Marenholz et al, 1996), was as frequently amplified as FLG, whereas the more proximal marker, D1S3620, was amplified in fewer cases. D1S3620 was highly amplified in MS8x, whereas D1S3623 and FLG showed normal copy numbers, but the sample had another amplified cluster covering IVL through SPRR2A. D1S3625 and D1S3628, localized distal to S1]OA2, showed a similar amplification pattern to S1OOA2.

FISH analysis of chromosome 1 and 2 copy numbers
Although the variable gene dosages observed along the chromosome were consistent with regional low-level amplification, we wanted to ascertain that the tumours did not have extra copies of  Centromere 1 and 2 as well as four YAC clones from 1q21-q22/q23 have been analysed by FISH in selected tumours with amplifications in 1 q21-q22 detected by CGH and/or molecular analysis and in two samples with normal copy numbers of the genes (MS2x and MFH25). For each sample, at least 150-200 nuclei were counted. YACs 789f2 and 764a1 are partly overlapping. YACs 935b12 and 883h6 are located very close together, but more distal (not shown). Centromere analyses: for each sample, the average centromere counts are shown. All tumours had a balanced chromosome 1 :2 ratio, but LS22, MFH3x and OS4x had abnormal (increased) chromosome numbers and were not analysed with the YACs (NA). YAC analyses: for each sample, the percentage of nuclei with abnormal copy numbers is shown. Bold text: high copy number gains of a YAC (at least ten signals). Underlined text: three to four YAC signals in more than 30% of the nuclei. In two samples with amplification of 789f2 (LS13) and 789f2 and 883h6 (LS21), a fraction of the nuclei showed four to six signals, as indicated by split boxes. Italic: 10-20% of the nuclei showed three to four signals. In addition, some samples gave three to four signals with one or two of the YACs tested in less than 10% of the nuclei. Notably, in LS21, two different parts of the tumour (a, b) showed different amplification patterns.
chromosome 2 (used for normalization). We therefore determined the copy numbers of centromeres 1 and 2 by interphase FISH on nuclei from ten of the samples with amplicons detected by molecular analysis and/or CGH, as well as two tumours with apparently normal copy numbers of all the markers tested (MS2x and MFH252). All the samples tested had similar mean numbers of centromere 1 and 2 per nucleus ( Table 2). Seven of the tumours showed normal centromere counts in most of the nuclei, whereas three (LS22, MFH3x and OS4x) had four to six copies of each.

FISH analysis of YAC copy numbers
The molecular results gave no clear indications as to whether the region studied was likely to contain the core of the amplicon, although the high amplification levels of D4S3620 found in MS8x could indicate that this marker was close. The region between DlS3620 and FLG has not been mapped in detail, but two YACs that cover this region, 789f2 and 764al, were available [the positions of these YACs was mapped by Mahrenholz et al (1996) and are indicated in Figure 2]. We studied the amplification pattern of these YACs only in samples with normal average cen 1 and cen 2 copy numbers, including seven tumours with amplifications in this region and two control samples (MS2x and MFH25). The samples with increased centromere 1 and 2 copies were excluded from the study because it would be very difficult to obtain reliable results.
Two additional YACs were selected from the CEPH-Genethon map of region lq21-q23 and mapped by FISH analysis distal to 764al (not shown). However, FISH on normal metaphases could not determine whether 935b12 or 883h6 is the most distal YAC as they gave overlapping signals. Results from FISH analyses with these four YACs are shown in Table 2 and Figure 3.
As shown in Table 2, 789f2, covering D1S3620, detected high copy numbers (more than ten, but in most samples uncountable) in nuclei from five of the samples and three to four signals in one (LS3x). However, there was considerable heterogeneity among the nuclei. In MFH36 and MS8x, a major fraction of the nuclei (75% or more) exhibited amplification; in the other samples, LS2, LS3x, LS6 and LS 13, only about 50% or less. 764a 1, covering D I S3623, detected high copy numbers in three samples and three or four signals in one, but, except for MFH36, only about 50% of the nuclei or less had these aberrations. The more distal YACs were amplified in fewer samples. 883h6 and 935b12 detected high copy numbers only in LS3x, involving more than 70% of the nuclei. We analysed two different pieces from the last sample, LS2 1. As shown in Table 2, considerable heterogeneity was detected: one part of the tumour had no amplifications, whereas in the other part 789f2 and 883h6 showed high copy numbers. In addition, a fraction of the nuclei also showed four to six signals with these two YACs. 789f2 and 883h6 were always co-amplified in this part of the tumour, but only in 46% of the nuclei analysed. The other YACs, 764al and 935bl2, detected three or four signals in 16% and 38% of the nuclei respectively.
For some samples, one or two of the YACs tested gave three or four signals in a smaller fraction of the cells (less than 20%). As all these samples showed higher amplification levels with at least one of the other YACs, these aberrations were considered to be of less importance.

DISCUSSION
Gain of chromosomal sequences, or amplification, is frequently found in human cancers. Functional analysis and clinical observations have supported the hypothesis that amplification and 'MFH25 was called MFH43 in a previous study (Forus et al, 1995a). 1q21-q22 amplification in human sarcomas 501 A Figure 3 FISH analysis using YACs from 1q21-q22/q23 on interphase nuclei from MFH36, MS8x and LS2. (A-C) interphase nuclei from MFH hybridized with (A) digoxygenin (dig)-labelled 789f2 detected by FITC (in green) and biotin-labelled 764a1 detected by avidin-CY3 (in red), (B) biotin-labelled 883h6 and (C) 935b12, detected by avidin-Cy3 (in red). YAC 789f2 and 764a1 detect amplification and were always co-amplified, but 789f2 is amplified at higher levels (A). 883h6 and 935b12 show normal copy numbers in most of the nuclei (883h6 gives three signals in one). (D and E) interphase nuclei from MS8x hybridized with biotin-labelled 789f2 (D) and 935b12 (E). 789f2 detects high-level amplification whereas 935b12 gives normal signals. (F) interphase nucleus from LS2 hybridized with digoxygenin (dig)-labelled 789f2 (in green) and biotin-labelled 764a1 (in red). The YACs were always co-amplified, but also, here, 789f2 is amplified at higher levels overexpression of cellular oncogenes are important for tumorigenesis or tumour progression (Alitalo and Schwab, 1986). In its simplest form, one might expect that amplification and overexpression of a single dominant (proto)oncogene could provide a selective growth advantage to the tumour cells. If so, all amplicons would contain this 'driver' gene and its closely flanking markers and other amplified sequences would only be random passengers of the amplicon. However, studies of 11 q 13 amplification in breast cancer and 12q 13-15 amplification in sarcomas have revealed more complex situations with multiple amplification units and several candidate genes in the same region (Gaudray et al, 1992;Berneret al, 1996;Wolf et al, 1997).
We have previously demonstrated frequent amplification of the 12q 13-15 segment, including the CDK4, MDM2 and HMGIC genes, in human sarcomas (Forus et al, 1993;Maelandsmo et al, 1995;Berner et al, 1996Berner et al, , 1997. In addition, CGH analysis detected amplification of the 1q2l -q22 region in even more samples from this tumour panel (Forus et al, 1995a, b), indicating that lq21-q22-located genes may also play an important role in the development and/or progression of such tumours. Similar observations have also been reported by other investigators (Szymanska et al, 1996;Armengol et al, 1997).
In this first molecular analysis of the lq21-q22 amplifications in sarcomas, we determined the amplification status of 11 markers located in a region covering 4.5 Mb, including S10OA6, which had been reported to be amplified in melanomas (Weterman et al, 1992). We also analysed three more distal markers, including MUCI, amplified in some breast carcinomas (Bieche and Lidereau, 1997;Bieche et al, 1995).
One would expect that markers closest to the 'core' of the amplicon would be amplified most frequently and at the highest levels. The most frequently amplified markers were moderately increased in most of the samples, and only APOA2 and the anonymous marker Dl S3620 were amplified more than tenfold in one sample each (in LS21 and MS8x respectively). Thus, these criteria were not fulfilled for any of the markers tested here. The relevance of the moderate copy number increases is unclear at present, but it seems likely that they are due to amplification processes within the segment, probably selected for by one or more unknown oncogene(s) in the region. As DIS3620 detected high-level amplification in one sample, we analysed the region around this marker in more detail by FISH in nine selected samples, using two partly overlapping YACs covering sequences between DlS3620 and FLG. YAC 789f2, which includes DlS3620 and extends towards D1S3623 (Figure 2), detected high copy numbers in six of these nine tumours (LS2, LS6, LS 13, LS2 1, MFH36 and MS8x, Table 2), but in only one part of LS2 1. The more distal 764al was amplified in only three of these tumours (LS2, LS 13, and MFH36). No YACs near APOA2, the other region with high-level amplification, were available, but we analysed the amplification pattern of 935b12 and 883h6, which we have mapped distal to 764al (Table 2). Like APOA2, both YACs detected amplification in LS3x. In LS2 1, showing the highest copy numbers of APOA2, only one part of the tumour showed amplification of 883h6, again indicating tumour heterogeneity.
In five of the samples with amplification of 789f2, the more distal YACs (935b12 and 883h6) showed normal copy numbers in most of the nuclei. These results could suggest that in a subset of British Journal of Cancer (1998) 78(4), 495-503 0 Cancer Research Campaign 1998 tumours a target gene is located within 789f2, or even more proximal, whereas more distal sequences seem to be of less importance. In LS3x, on the other hand, the focus of the amplicon seems to be more distal, i.e. in the region of 883h6, 935b12 and APOA2, whereas LS21 may have at least two amplicons, one near 789f2 and another one in the region of APOA2 and 883h6. The presence of multiple amplified regions on the lq arm in sarcomas has been shown previously (Forus et al, 1995a).
It is somewhat puzzling that for most samples the high-level amplifications detected by FISH were often found only in about 50% of the nuclei. One possibility is that the cells with normal copy numbers in tumours with amplification are normal cells, in which case the amplification levels determined from the Southern analyses would be underestimated. However, such a large fraction of normal cells would be surprising, e.g. in the homogeneous WDLPS samples. Conversely, the variation in copy numbers between the nuclei could be an indication of clonal variations within the tumours and, then, Southern analysis would only reveal the average copy numbers. This interpretation is supported by the heterogeneous amplification status detected in some of the tumours (e.g. LS21). It seems likely that the results shown here (Table 2) reflect tumour heterogeneity as well as, to some extent, the presence of normal cells. Tumour cell percentage in the different pieces was not evaluated before DNA isolation, therefore we do not know to what extent the presence of normal cells may have affected the measured amplification levels.
Most of the samples analysed were from primary tumours of histological grade 3 or 4 (Table 1), and there is no correlation between malignancy grade or any other known clinical parameters and the presence of Iq21-q22 amplifications. Many of the patients with osteosarcoma have received chemotherapy before surgery, but the presence of the 1q21 -q22 amplicon and preoperative treatment is not correlated. Therefore, it seems unlikely that such amplifications could be therapy induced. As the lq21-q22 amplicon is present in primary as well as in recurrent and metastatic samples, it is possible that amplification of oncogenes in this region plays a role in the development of these tumours but may be less important for progression and metastasis.
The previous CGH analyses of this tumour panel (Forus et al, 1995a,b), as well as other studies of osteosarcomas and liposarcomas (Tarkkanen et al, 1995;Szymanska et al, 1996), indicate that lq21-q22 amplifications are more frequent than those at 1 2q I 3-q 15 and thus may be of great importance. In this first molecular characterization of the I q2 1-q22 amplifications in sarcomas, we observed frequent but relatively low copy number increases for most of the markers tested, but also some high-level amplifications. Taken together, our results indicate more than one core of the amplicon and further analyses are required to determine important sequences more precisely and to find the relevant genes.