Oncogene (2008) 27, 3072–3080 & 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00 www.nature.com/onc ONCOGENOMICS High-resolution array CGH analysis of salivary gland tumors reveals fusion and amplification of the FGFR1 and PLAG1 in ring

F Persson1, M Winnes1, Y Andre´ n1, B Wedell1, R Dahlenfors2, J Asp1, J Mark2, F Enlund1 and G Stenman1

1Lundberg Laboratory for Cancer Research, Department of Pathology, Sahlgrenska University Hospital, Go¨teborg University, Go¨teborg, Sweden and 2Department of Pathology, Central Hospital, Sko¨vde, Sweden

We have previously identified a subgroup of pleomorphic adenomas most commonly occur in the parotid gland salivary gland adenomas with ring chromosomes of and less frequently in the submandibular and minor uncertain derivation. Here, we have used spectral salivary glands. Itis a benign tumorshowing a karyotyping (SKY), fluorescence in situ hybridization remarkable degree of morphological diversity, including (FISH) and high-resolution oligonucleotide array-CGH to epithelial and myoepithelial cells forming a variety of determine the origin and content of these rings and to patterns in an often mucoid/myxoid or condroid matrix. identify genes disrupted as a result of ring formation. Of Pleomorphic adenomas may cause problems in the 16 tumors with rings, 11 were derived from 8, clinical management due to its tendency to recur and the 3 from chromosome 5 and 1 each from chromosomes 1, 6 risk of malignant transformation. and 9. Array-CGH revealed that 10/11 r(8) consisted of Extensive cytogenetic studies have shown that pleo- amplification of a 19 Mbpericentromeric segment with morphic adenomas are frequently characterized by recurrent breakpoints in FGFR1 in 8p12 and in PLAG1 in chromosome translocations with consistent breakpoints 8q12.1. Molecular analyses revealed that ring formation atchromosome bands 8q12 or 12q14–15 (Stenman, consistently generated novel FGFR1–PLAG1 fusions 2005). We have previously identified the DNA-binding in which the 50-part of FGFR1 is linked to the coding transcription factor genes PLAG1 and HMGA2 as the sequence of PLAG1. An alternative mechanism of PLAG1 target genes of these translocations and shown that they activation was found in tumors with copy number gain of resultin fusion oncogenes (Schoenmakers et al., 1995; an intact PLAG1 gene. Rings derived from chromosomes Geurts et al., 1997; Kas et al., 1997). PLAG1 fusions are 1, 5, 6 or 9 did not result in gene fusions, but rather far more common than HMGA2 fusions and are found resulted in losses indicative of the involvement of putative in at least 50% of karyotypically abnormal tumors tumor suppressor genes on 8p, 5p, 5q and/or 6q. (Stenman, 2005). The breakpoints in PLAG1 occur in Our findings also reveal a novel mechanism by which the 50-noncoding part of the gene, leading to promoter FGFR1 contributes to oncogenesis and further illustrate swapping or substitution between PLAG1 and a fusion the versatility of the FGFR1 and PLAG1 genes in partner gene (Kas et al., 1997; Voz et al., 1998; A˚ stro¨ m tumorigenesis. et al., 1999; Asp et al., 2006) PLAG1 fusions usually Oncogene (2008) 27, 3072–3080; doi:10.1038/sj.onc.1210961; result in ectopic expression of a normal PLAG1 . published online 3 December 2007 The PLAG1 protein contains an N-terminal zinc-finger DNA-binding domain and a C-terminal Keywords: array-CGH; gene amplification; fusion transactivation domain (Kas et al., 1997, 1998). PLAG1 oncogene; FGFR1; PLAG1; salivary gland tumor activates transcription through binding to a bipartite DNA-binding consensus sequence consisting of a core sequence (GRGGC) and a G-cluster (RGGK) (Kas et al., 1998). It has been suggested that PLAG1 exerts its ONCOGENOMICS major oncogenic effects by induction of growth factors Introduction such as insulin-like growth factor 2 (IGF2) (Voz et al., 2004). The pleomorphic adenoma is the most common salivary We have previously identified a subgroup of pleo- gland tumor and accounts for about 60% of all salivary morphic adenomas with ring chromosomes of uncertain gland neoplasms (Eveson et al., 2005). Pleomorphic origin (Mark et al., 1983; Stenman, 2005). To determine the derivation and genetic content of these rings, we have now molecularly characterized 16 tumors with one Correspondence: Professor G Stenman, Lundberg Laboratory for or more ring chromosomes. Using spectral karyotyping Cancer Research, Department of Pathology, Sahlgrenska University (SKY), fluorescence in situ hybridization (FISH) and Hospital, Go¨ teborg University, Go¨ teborg SE-413 45, Sweden. E-mail: [email protected] high-resolution oligonucleotide array-CGH (aCGH) Received 23 August 2007; revised 29 October 2007; accepted 2 analyses, we could show that in 11 tumors, the rings November 2007; published online 3 December 2007 were derived from . Moreover, aCGH FGFR1–PLAG1 gene fusion and amplification F Persson et al 3073 analysis revealed that 10 of the r(8) chromosomes resulted in a fusion between the 30-partof FGFR1 consisted of amplification of a pericentromeric region (including the tyrosine kinase domain) and a putative 50- of chromosome 8 with breakpoints in the FGFR1 gene fusion partner on 6q or 10q. in 8p12 and in the PLAG1 gene in 8q12. Reverse aCGH analysis of the r(8) in case 11 revealed copy transcriptase PCR (RT–PCR) analysis confirmed that number gain of the 8q10Àq21.11 segmentincluding the r(8) in all cases resulted in a novel FGFR1–PLAG1 PLAG1 (Table 1; Figure 2b). The proximal breakpoint gene fusion. Our results reveal a new mechanism of was in the centromere and the distal breakpoint within formation of gene fusions and subsequent amplification or immediately distal to the CRISPLD1 gene. Similarly, in ring chromosomes in epithelial tumors. aCGH analysis of case 12 revealed gain of one copy of chromosome 8, including an intact PLAG1 gene. aCGH analysis of two of the three tumors with r(5) revealed similar, butnotidentical,breakpointsin 5p14.1 Results and 5q21.1 with concomitant losses of the segments distal to these breakpoints (Figure 2c). In case 13, the Cytogenetic and SKY analyses 5p14.1 breakpoint was between the CDH9 and CDH6 The karyotypic findings based on G-banding, SKY and genes, and in case 14, itwas between CDH9 and aCGH are summarized in Table 1 and Figure 1a. All CDH10. The corresponding breakpoints in 5q21.1 were tumors had pseudodiploid or hyperdiploid karyotypes between SLCO4C1 and SLCO6A1, and SLCO6A1 with clonal structural aberrations. Sixteen tumors had and PAM, respectively. The third case with r(5) had one or more ring chromosomes and one tumor had slightly different breakpoints located in 5p11 and 5q22 trisomy 8 as the sole anomaly. SKY analysis revealed (Table 1). the derivation of all ring chromosomes and markers and Of the remaining two cases with ring chromosomes, showed that each ring only contained material from a that is, r(6) in case 16 and r(9) in case 17, only the 6p14.1 single chromosome (Figure 1a). Eleven tumors had rings breakpoint in the r(6) was associated with a copy derived from chromosome 8 and three had rings derived number alteration, allowing a more detailed mapping by from chromosome 5, one of which also had an r(1). The aCGH. This breakpoint was located between the two remaining tumors had rings derived from chromo- IMPG1 and HTR1B genes. somes 6 and 9, respectively. Two tumors had super- The variable size of the r(8) both within individual numerary r(8) and four tumors contained 2–4 copies of tumors and between tumors prompted us to study the the r(8). copy number of PLAG1 in r(8) from seven tumors. FISH analyses using PLAG1-specific probes and a aCGH and FISH analyses of tumors with r(8) reveal CEP8 probe revealed multiple dual signals on all r(8) consistent copy number gains of 8p12–q12.1 as well as in micronuclei from several cases (Figures 1b To fine map the breakpoints and determine the genetic and c). The number of PLAG1 and CEP8 signals in each content of the rings, we performed aCGH analysis of 14 r(8) varied from two to eight (Table 1). The sizes of the tumors. A detailed description of all gains and losses rings correlated well with the number of PLAG1 and recorded are given in Table 1. In general, losses were CEP8 signals. However, FISH analysis revealed a higher four times more common than gains. However, the most copy number of PLAG1 than aCGH did. This difference prominentfinding was a consistentcopy number gain of may be explained by in vitro selection of cells with high the pericentromeric segment 8p12Àq12.1 in eight copy number gain of PLAG1, intratumor heterogeneity tumors with r(8). Other recurrent copy number alter- and/or by the fact that most aCGH-platforms quantita- ations were losses of one or more regions within 3p21.3 tively underestimate copy number changes in log2 ratios and 7q11.21–q11.23 in six and four tumors, respectively, (Pollack et al., 1999; Brennan et al., 2004). 8p12-pter in three tumors and 5p14.1-pter and FISH analysis of case 12, which had trisomy 8 as the 5q21.1-qter in two tumors. There were also four cases sole anomaly, revealed three copies of PLAG1. Rings with loss of partly overlapping segments of 6q. derived from chromosomes other than 8, that is r(1), Analyses of the 44K arrays revealed consistent r(5), r(6) and r(9), were consistently negative for PLAG1 breakpoints within or close to the FGFR1 gene in 8p12 sequences. Case 15, which had a t(3;8) as well as r(1) and and close to the PLAG1 gene in 8q12 in tumors with r(5), showed a splitsignal for YAC 166F4, consistent r(8). To refine the mapping of these breakpoints, we with a CTNNB1–PLAG1 gene fusion. analyzed four tumors with high-resolution 244K arrays. Detailed analysis of these arrays unequivocally demonstrated that the breakpoints had occurred within Identification of a novel FGFR1–PLAG1 gene fusion in the 50-parts of the FGFR1 and PLAG1 genes in all four tumors with r(8) tumors (Figure 2a), suggesting a possible fusion between To test the hypothesis that the r(8) might result in a these two genes as a result of the formation of r(8). In fusion between the FGFR1 and PLAG1 genes as three of the 11 tumors with r(8), there was a suggested by aCGH, we performed RT–PCR analysis concomitant loss of the segment distal to 8p12. In cases of 14 of the 16 tumors with ring chromosomes, including 3, 6 and 8, the 8p12-pter segment was translocated onto all but one of the cases with r(8). The transcriptional 6q14.1, 10q22.1 and 6q22.1, respectively. In none of orientation of both genes (Figure 2a) is compatible with these cases, aCGH indicated that the translocations had a simple ring closure resulting in a fusion of the 50-part

Oncogene 3074 Oncogene

Table 1 Clinical–pathological, cytogenetic and molecular data on 17 pleomorphic adenomas with ring chromosomes or trisomy 8. (r(8) chromosomes as well as partial and complete gains of chromosome 8 are shown in bold) Case Sex/age Size (cm) Karyotype based on G-banding, PLAG1 status aCGH Copy number by aCGH (years) SKY, FISH and aCGH (copies/ring) Gainsa Lossesa

1 M/69 1.2 47,XY,del(7)(q11.21q11.23), der(8)(pter- FGFR1–PLAG1 44K 8p12–q12.1 [19] 7q11.21–q11.23 [13.1] p12::q12.1 -q12.1-neo-q12.1-qter, (4) +r(8)(p12q12.1) 2 M/53 1.7 46,XY, der(1)t(1;8)(p33;q12.1), r(8)(p12q12.1) FGFR1–PLAG1 44K 8p12–q12.1 [19] 1pter–p33 [50], 8pter–p12 [38.4]

(6) FGFR1–PLAG1 3 M/69 1.9 46,XY,der(6)t(6;8)(q14.1;p12), de- FGFR1–PLAG1 44K 8p12–q12.1 [19] 6q14.1-q16.3 [22.7], 8q23.3-qter [30], 11pter– r(7)ins(7;8)(p22;q12q23.3), r(8)(p12q12.1), (8) p14.3 [26.4] der(11)t(6;11)(q16.3; p14.3) 4 F/40 3.0 46,XX,der(3)t(3;8)(p26;q12.1), FGFR1–PLAG1 244K 8p12–q12.1 [19] 3p21.3 multiple deletions [2], 6q24.1–q25.2

der(6)t(6;21)(q24.2;q21),del(7)(q31), (4) [10], 6q25.3-qter [14], 8pter–p12 [38.4], amplification and fusion gene r(8)(p12q12.1),der(21)t(7;21)(q31;q21)/46, idem, 16q22.1–q23.1 [9], 21q21.3 [0.5] del(16)(q22.1q23.1) 5 M/67 2.2 47–48,XY, dup(2)(q22.3q24.3), FGFR1–PLAG1 44K 2q22.3–q24.3 [23], 8p12- 3p21.3 [2.5] +r(8)(p12q12.1)x1–2 (6) q12.1 [19]

6 F/81 3.0 46–47, XX, der(3)t(3;7)(p14.2;p13)del(3)(q12.3), FGFR1–PLAG1 244K 8p12-q12.1 [19] 3pter-p14.2 [61.4], 3q12.3-qter [94], 8q12.1 Persson F der(7)t(7;8)(p13;q12.1) t(8;10)(q24;q22.1), [0.75] r(8)(p12q12.1)x1–2, der(10)t(8;10)(p12;q22.1) 7 M/70 1.5 46,XY,der(4)t(4;8)(q31;q12.1), r(8)(p12q12.1) FGFR1–PLAG1 n.d. 8 M/49 2.0 46,XY, der(6)t(6;8)(q22.1;p12), r(8)(p12q12.1), FGFR1–PLAG1 244K 8p12-q12.1 [19] 2q34 [0.65], 2q37.2 [1.0], 6q22.1-qter [55], al et der(17)t(8;17)(q12;q25) 8p23.2–p23.3 [5] 9 F/22 0.6 46,XX, der(1)t(1;2)(p35.3;p21), FGFR1–PLAG1 244K 8p12–q12.1 [19] 1pter-p35.3 [27.7], 8pter–p12 [38.4], 10q11.21 der(2)t(2;3)(p21;q21), der(3)t(3;8)(q21;q12.1), [1.5] r(8)(p12q12.1) 10 K/57 2.0 46–48,XX,t(3;6)(q21;p21.3), r(8)(p12q12.1)x1–3, FGFR1-PLAG1 n.d. der(13)t(8;13)(q12;q12.1) (3) 11 M/31 0.5 46–49,XY,del(8)(q10q21.11), r(8)(q10q21.11) Copy number 44K 8q10–q21.11 [30] — x1–4 gain (2) 12 M/60 2.0 47,XY, +8 3 copies 44K 8 — 13 M/69 3.0 46,XY,r(5)(p14.1q21.1) Normal 44K — 3p21.3 [1.0; 1.8], 5pter–p14.1 [28.4], 5q21.1- qter [78.4], 7q11.21 [4.4], 12q14.1–q14.3 [3.1], 15q15.1 [1.4] 14 F/57 1.5 46,XX,t(2;7)(q31;p13), r(5)(p14.1q21.1) Normal 44K — 3p21.3[1.6], 5pter–p14.1[25], 5q21.1–qter[78], 11p14.1[0.2], 22q12.2[1.7] 15 F/13 1.5 46, XX,t(3;8)(p21;q12.1)/46,XX, idem, n.d. n.d. r(1)(p11q43B44), r(5)(p11q22) 16 F/64 1.5 46,XX, r(6)(p21q14.1), der(11)t(6;11)(p21;q21) Normal 44K — 3p21.3 [1.0; 1.7], 5q13.1 [0.5], 6q14.1-qter [92.4], 7q11.23 [4.6], 9p13.3 [1.0], 9q33.3- q34.11 [2.8], 11q21-qter [40], 12q24.23-q24.31 [3.2], 17p13.2 [1.0] 17 M/77 2.5 46,XY,t(2;8)(p21B22;q23), del(7)(q11q22)/ Normal 44K — 7q11.22–q21.11 [13] 46,XY,der(1)t(1;9)(q44;q13), t(2;8) (p21-22;q23), r(9)(p24q13)

aApproximate sizes in megabases of gained and lost segments are shown in square brackets. FGFR1–PLAG1 gene fusion and amplification F Persson et al 3075

Figure 1 SKY and FISH analyses of tumors with ring chromosomes. (a) Partial SKY karyotypes of cases 1–6, 10, 11 and 13–17. Ring chromosomes and translocated chromosome segments are indicated. (b) FISH analysis of selected r(8) chromosomes of cases 3–5, demonstrating amplification of PLAG1 (green signals) and a satellite sequences (red signals) within the rings. Note the differences in the size of the rings and of the number of signals within each ring. (c) Micronucleus from case 3 with amplification of PLAG1 (green signals) and a satellite sequences (red signals). The chromosomes and nuclei are counterstained in blue with 40,60-diamidino-20- phenylindole dihydrochloride. SKY, spectral karyotyping; FISH, fluorescence in situ hybridisation. of FGFR1 to the 30-partof PLAG1. Using primers transcripts containing exon 1 of FGFR1 fused to either located in exon 1 of FGFR1 and in exon 4 of PLAG1,we exon 2 or 3 of PLAG1 and two tumors expressed could show that 9 of the 10 cases with r(8) expressed chimeric transcripts consisting of exon 2 of FGFR1 FGFR1–PLAG1 fusion transcripts (Figure 3a). aCGH fused to either exons 2 or 3 of PLAG1 (GenBank analysis of the 10th case revealed identical breakpoints accession numbers EF525168, EF525169, EF525170 in FGFR1 and PLAG1, strongly indicating that this case and EF525171). The breakpointin FGFR1 in the also has an FGFR1–PLAG1 gene fusion. latter two cases was in intron 2, resulting in a fusion Four differenttypesof FGFR1–PLAG1 fusion trans- of 50-noncoding sequences as well as 91 nucleotides of cripts were detected (Figure 3a). Nucleotide sequence the coding region to the acceptor splice site of PLAG1 analysis showed that seven tumors expressed chimeric exon 2 or 3.

Oncogene FGFR1–PLAG1 gene fusion and amplification F Persson et al 3076

Figure 2 aCGH profiles of chromosomes 8 and 5 in tumors with r(8) and r(5). (a) Profiles of overlayed chromosomes 8 from four tumors (cases 4, 6, 8 and 9) with r(8) demonstrating copy number gain of 8p12–8q12.1 and loss of 8pter-p12. To the right is shown zoomed gene views of the 8p12 and 8q12.1 breakpoint regions. Note the distinct breakpoints within the FGFR1 and PLAG1 genes. The transcriptional orientation of the two genes is indicated by arrows. (b) aCGH profile of chromosome 8 from the fusion-negative tumor with r(8) and copy number gain of 8q10–q21.11 (case 11). The location of the PLAG1 gene within the gained region is indicated. (c) aCGH profiles of overlayed chromosomes 5 from two tumors with r(5) (cases 13 and 14). Note that both tumors have almost identical breakpoints with concomitant deletions of the distal parts of 5p and 5q. aCGH, array-CGH.

As expected, case 11 with an r(8)(q10q21.11) as well We also evaluated the expression of the PLAG1 as the cases with trisomy 8, r(5), r(6) and r(9) were all protein by immunohistochemistry in five tumors (cases negative for the FGFR1–PLAG1 fusion (Figure 3a). 1–5) with r(8), resulting in fusion and copy number gain Cases 11 and 12 were also analyzed for expression of of FGFR1 and PLAG1. All tumors overexpressed the two recently described cryptic PLAG1 gene fusions, PLAG1 protein in the majority of cells and cell types. CHCHD7–PLAG1 and TCEA1–PLAG1 (Asp et al., The staining was mainly nuclear. In line with the real- 2006). In addition, case 4, which had a der(3)t(3;8), was time PCR results, the highest expression level was noted analyzed for expression of the CTNNB1–PLAG1 fusion. in case 3 (Figure 3d). RT–PCR analyses revealed that none of these tumors To study the impact of copy number gain of 8p12– expressed any of these fusion transcripts (data not q12.1 on other genes located within this segment, we shown). also analyzed the expression of FGFR1 as well as of C8orf4 (TC1;8p11.21), MYST3 (8p11.21), DKK4 (8p11.21–p11.1), MCM4 (8q11.21) and SNAI2 Overexpression of PLAG1 in tumors with fusion and (8q11.21) using real-time quantitative PCR. As ex- amplification of the FGFR1 and PLAG1 genes pected, case 12, which was trisomic for chromosome 8, To study the consequences of fusion and amplification showed the highest expression level of FGFR1 of the FGFR1 and PLAG1 genes on PLAG1 expression, (Figure 3c). In the remaining tumors, there were no we performed real-time quantitative PCR on 10 of the substantial differences in FGFR1 expression between 12 tumors with r(8) or trisomy 8. All tumors over- tumors with r(8)/copy number gain and tumors with expressed PLAG1 relative to normal salivary gland CTNNB1–PLAG1 gene fusions. Similar results were tissue (Figure 3b). In general, the expression levels were also obtained for C8orf4 (TC1), MYST3, DKK4, equal or higher in tumors with r(8) and FGFR1–PLAG1 MCM4 and SNAI2 (data not shown). fusions compared to tumors with CTNNB1–PLAG1 fusions. The tumor with the highest expression level of PLAG1 had also the highest number of PLAG1 copies Discussion per r(8) (case 3). Case 12, which was trisomic for chromosome 8, expressed similar levels of PLAG1 as the Here, we have used a combination of SKY, FISH and cases with CTNNB1–PLAG1 fusions. oligonucleotide aCGH to determine the origin and

Oncogene FGFR1–PLAG1 gene fusion and amplification F Persson et al 3077 genomic content of ring chromosomes in pleomorphic from chromosome 5 and one each from chromosomes 1, adenomas and to identify possible genes disrupted as a 6 and 9. High-resolution aCGH analysis using 44K result of ring formation. Of the 16 tumors with one or arrays as well as the most recently released 244K more rings, 11 were derived from chromosome 8, three oligonucleotide arrays revealed that the chromosome 8 rings in all but one case consisted of amplification of a pericentromeric segment with consistent breakpoints in the FGFR1 gene in 8p12 and in the PLAG1 gene in 8q12.1. RT–PCR and nucleotide sequence analyses confirmed that all 10 cases with r(8)(p12q12.1) had a novel FGFR1–PLAG1 gene fusion. All fusion-positive tumors expressed two alternatively spliced fusion transcripts. The majority of tumors (78%) expressed transcripts consisting of exon 1 of FGFR1 fused either exon 2 or 3 of PLAG1, and a few tumors (22%) expressed transcripts consisting of exon 2 of FGFR1 fused to either exon 2 or 3 of PLAG1. These transcript variants are consistent with breakpoints located in intron 1 of PLAG1 and in intron 1 or more rarely in intron 2 of FGFR1. Chimeric transcripts including exon 2 of FGFR1 and exons 2 or 3 of PLAG1 are predicted to encode truncated FGFR1–PLAG1 fusion of 52 and 55 aminoacids, respectively, as well as a full-length PLAG1 protein. Whether these truncated fusion proteins, which do not contain functional FGFR1 tyrosine kinase domains, are biologically active, however, remains to be shown. Since the breakpoints in FGFR1 and PLAG1 occurred in the 50-noncoding regions of both genes, the major molecular consequence of this fusion is likely to be activation of PLAG1 expression due to promoter substitution. Real-time quantitative PCR analysis con- firmed that all FGFR1–PLAG1 positive tumors over- expressed PLAG1 relative to normal salivary gland tissue and that the expression levels were equal or higher in tumors with FGFR1–PLAG1 fusions and amplifica- tion compared to tumors with CTNNB1–PLAG1 fusions. The tumor with the highest number of PLAG1 copies per r(8) also expressed the highest level of PLAG1. The fact that there were no significant differences in expression levels of six other cancer- associated genes in 8p12–q12.1, including FGFR1,in tumors with r(8) or copy number gain of PLAG1 compared to tumors with CTNNB1–PLAG1 fusions

Figure 3 Expression of FGFR1–PLAG1 fusion transcripts and protein. (a) RT–PCR analysis revealed expression of FGFR1– PLAG1 fusion transcripts in all (cases 1–8 and 10) but one (case 11) of the cases with r(8). Tumors with trisomy 8, r(5), r(6) and r(9) are negative for the FGFR1–PLAG1 fusion. Four types of FGFR1– PLAG1 fusion transcripts were detected, corresponding to chimeric transcripts containing exon 1 of FGFR1 fused to either exons 2 or 3 of PLAG1 and to transcripts containing exon 2 of FGFR1 fused to either exons 2 or 3 of PLAG1.(b and c) Real-time quantitative PCR analysis of PLAG1 (b) and FGFR1 (c) expression in FGFR1– PLAG1-positive tumors with r(8) chromosomes (cases 1–4, 6–8 and 10; blue bars) and in fusion-negative tumors with copy number gain of PLAG1 (cases 11 and 12; light-blue bars) compared to normal salivary gland tissue (NSG.) and four tumors with CTNNB1–PLAG1 gene fusions (grey bars). (d) Immunostaining of the PLAG1 protein in an FGFR1–-PLAG1-positive tumor with r(8) and amplification of the fusion gene (case 3). Note the moderate to very strong (arrows) staining intensity of PLAG1 in the majority of nuclei.RT–PCR, reverse transcriptase PCR.

Oncogene FGFR1–PLAG1 gene fusion and amplification F Persson et al 3078 further emphasizes the critical role of PLAG1 in these constitutive tyrosine kinase activation of FGFR1 by tumors. However, we cannot exclude the possibility that oligomerization. The FGFR1 fusions in EMS are clearly other genes within 8p12–q12.1 could have oncogenic differentfrom the FGFR1–PLAG1 fusions described effects. here, in which the 50-partof FGFR1, excluding the The r(8) chromosomes demonstrated considerable tyrosine kinase domain, are fused to the entire coding inter- and intratumor heterogeneity in both number sequence of PLAG1. Moreover, previous studies have and size. The number of rings per cell varied between shown that amplifications of 8p11–12, including the one and four, and the number of PLAG1 copies per ring FGFR1 gene, are found in several tumor types, including varied between two and eight. This variation may be for instance breast (Ugolini et al., 1999; Reis-Filho et al., caused by continuous rearrangement of the rings 2006) and colorectal carcinomas (Nakao et al., 2004). through breakage–fusion–bridge cycles as previously Amplifications of the 8p11–12 region are often accom- demonstrated for other types of low- and high-grade panied by loss of distal 8p, suggesting that one or more malignanttumors (Gisselsson et al., 1999, 2000). tumor suppressor genes are located on 8p (Pole et al., Of the 11 tumors with r(8), one was negative for the 2006). In three of our tumors with fusion and FGFR1–PLAG1 fusion. aCGH and FISH analyses amplification, there was also a concomitant loss of revealed that case 11 instead had a copy number gain 8p12-pter. Whether these losses affect the same gene(s) of 8q10–q21.11 with at least two copies of PLAG1 per as in breastand colorectalcarcinomas remains tobe ring. This case is therefore equivalent to case 12, which shown. was trisomic for chromosome 8 and PLAG1. The fact Spectral karyotyping and aCGH analyses revealed that that both tumors overexpressed PLAG1 atsimilar levels two tumors with r(5) had similar but not identical as tumors with PLAG1-fusions demonstrates that copy breakpoints in 5p14.1 and 5q21.1. The ring formation in number gain of an apparently normal PLAG1 gene is an these cases, however, did not result in any gain of alternative mechanism for activation of this gene. chromosome 5 material. Nor did the breakpoints indicate Copy number gains or amplifications of gene fusions that the formation of the rings could have resulted in are very rare and have previously been observed only in gene fusions. Similarly, aCGH analysis of the break- a few types of neoplasms, including chronic myeloid points in the r(6) and r(9) did not indicate that gene leukemia with BCR–ABL fusions (Gorre et al., 2001; fusions mighthave been generatedas a resultof the Johansson et al., 2002). T-cell acute lymphoblastic formation of these rings. Rather, our findings suggest leukemia with NUP214–ABL1 fusions (Graux et al., that the rings could result in loss of one or more putative 2004) and dermatofibrosarcoma protuberans with tumor suppressor genes located distal to the breakpoints COL1A1–PDGFB fusions (Abbott et al., 2006). Inter- on the short and long arms of chromosomes 5, 6 and 9. estingly, these studies have suggested that amplification Interestingly, six of the present tumors with ring of fusion genes may indicate progression towards a chromosomes also showed submicroscopic deletions of more malignantphenotypeand an aggressive clinical one or more regions within 3p21.3. Malignancy-related course. Whether this holds true also for pleomorphic deletions involving this region have previously been adenomas with genomic instability and FGFR1–PLAG1 found not only in the majority of epithelial cancers but gene fusions and/or copy number gain of PLAG1 also in many lymphoid malignancies (Kok et al., 1997; remains to be shown. Kost-Alimova et al., 2003; Kost-Alimova and Imreh, Our findings of FGFR1–PLAG1 gene fusions and 2007). Extensive experimental and clinical data suggest subsequentamplificationin ring chromosomes reveal a that this region harbors one or more tumor suppressor novel mechanism by which FGFR1 contributes to genes that when lost contribute to a malignant tumorigenesis. FGFR1, which belongs to a family phenotype (Kost-Alimova et al., 2003; Kost-Alimova of four receptor tyrosine kinases, is a 92 kDa trans- and Imreh, 2007). It is therefore tempting to speculate membrane protein with an extracellular domain con- that deletions within 3p21.3 could be an important taining three immunoglobulin-like C2-type domains, a genetic event that contributes to malignant transforma- heparin-binding site, a transmembrane domain and an tion of pleomorphic adenomas. intracellular thyrosine kinase domain (Eswarakumar et al., 2005). FGFR1 is expressed in mosttissues including normal fetal and adult salivary gland tissue Materials and methods (Hughes, 1997; Hoffman et al., 2002). Interestingly, FGFR1 is also involved in the development of salivary Tumor material glands by regulation of branching morphogenesis Fresh samples from 17 primary pleomorphic adenomas of the (Hoffman et al., 2002). parotid gland were obtained from the Central Hospital, Sko¨ vde, Previous studies have clearly implicated FGFR1 in the Sweden. Histopathological re-examination of the tumors con- pathogenesis of a variety of neoplasms. For example, in firmed the diagnosis of pleomorphic adenoma in all cases. The clinical–pathological characteristics of the tumors are shown in the 8p11 myeloproliferative syndrome (EMS)/stem cell Table 1. The tumors from eight patients (cases 3–6, 10, 11, 14 leukemia-lymphoma syndrome, gene fusions are found and 15) have not been reported previously. In the remaining nine 0 in which various 5 -partners are fused to the tyrosine tumors, the cytogenetic findings based on G-banding alone have kinase encoded domain of the FGFR1 gene as a result been reported (Mark et al., 1988, 1997). of chromosomal translocations (Xiao et al., 1998; For real-time PCR analysis, we used fresh frozen tumor Roumiantsev et al., 2004). The fusion proteins induce tissue from four additional pleomorphic adenomas with

Oncogene FGFR1–PLAG1 gene fusion and amplification F Persson et al 3079 t(3;8)(p21;q12) translocations and known CTNNB1–PLAG1 RT–PCR and nucleotide sequence analyses gene fusions. Total RNA was extracted from frozen tumor tissue using the TRIzol reagent(Invitrogen,Carlsbad, CA, USA). DNase- Cytogenetic, spectral karyotype and fluoresence in situ treated (DNA-free; Ambion, Austin, TX, USA) total RNA hybridization analyses was subsequently converted to cDNA using the SuperScript Primary cultures were established from fresh, unfixed tumor First-Strand Synthesis System (Invitrogen). As a control for specimens (Nordkvist et al., 1994). Chromosome preparations intact RNA and cDNA, RT–PCR reactions for expression were made from primary cultures or early passage cells, and of GAPDH were performed on all cDNAs (Asp et al., 2006). these were subsequently G-banded and analysed using The FGFR1–PLAG1 fusion transcript was amplified by PCR standard procedures. using the primers FGFR1-602F-50-TCGCACAAGCCACG Spectral karyotyping analysis was performed on 13 tumors GCGG30 (exon 1 of FGFR1) and PLAG1-445AS-50-GGAA (cases 1–6, 10, 11 and 13–17) using the SkyPaint probe kit CTGCCCAACTCCACTA-30 (exon 4 of PLAG1). PCR (ASI-Applied Spectral Imaging Ltd, Migdal Ha’Emek, Israel) amplification of the CHCHD7-PLAG1 and TCEA1-PLAG1 as described previously (Sjo¨ gren et al., 2000). fusion transcripts were performed as described previously (Asp Fluoresence in situ hybridization analysis was performed et al., 2006). PCR products were gel-purified and sequenced on metaphase chromosomes from eight cases using the using an ABI PRISM 310 Genetic Analyzer (Applied following PLAG1-specific probes: CEPH YAC 166F4, Biosystems, Foster City, CA, USA). The resulting sequences PAC233, PAC234 and PAC235 (Kas et al., 1997; Ro¨ ijer were analyzed using the BLAST-tool provided by the NCBI et al., 2002). Fluorescence signals were digitalized, processed (http://www.ncbi.nlm.nih.gov). and analyzed using the CytoVision image analysis system (Applied Imaging International Ltd, Newcastle-Upon-Tyne, UK). The number of PLAG1 and CEP8 signals were evaluated Quantitative real-time PCR analysis in at least 25 metaphases from each of the eight tumors. Quantitative real-time PCR analysis was performed using the AB 7500 FastReal-Time PCR system (Applied Biosystems). The following genes were analysed using TaqMan Gene Expression Array-CGH analysis assays; PLAG1 (Hs00965048_m1),FGFR1(Hs00241111_m1), Genomic DNA was isolated from frozen tumor tissue using R DKK4 (Hs00205290_m1),MCM4(Hs00381533_m1),MYST3 the QIAamp DNA Mini Kit(Qiagen GmbH, Hilden, (Hs00198899_m1), SNAI2 (Hs00161904_m1) and C8orf4 Germany). aCGH analysis was performed using high-resolu- (Hs00535539_s1) (Applied Biosystems). All samples tion 44K and 244K 60-mer oligonucleotide CGH-arrays were assayed in triplicates. To enable detection of possible conta- containing approximately 43 000 and 236 000 probes, respec- minating genomic DNA, we analysed non-reversed transcribed tively (arrays G4410B and G4411B sourced from the National total RNA in parallel with the cDNAs. The relative expression Center for Biotechnology Information (NCBI) genome Build levels of all genes were calculated using the comparative C 35; AgilentTechnologies Inc., Palo Alto,CA, USA). These t method (DDCt) and the SDS Software v1.3.1 (Applied arrays have an average spatial resolution of about 35 and Biosystems) using GAPDH (Hs99999905_m1) as endogenous 6.4 kb, respectively. The aCGH experiments performed control and cDNA from normal salivary gland tissue as according to the recommendations of the manufacturer calibrator (Livak and Schmittgen, 2001). (Barrett et al., 2004). Briefly, 3 mg each of tumor DNA and sex-matched reference DNA (Promega p/n G1471) were digested with AluI and Rsa1 for 2 h at37 1C. Digested tumor Immunohistochemistry and reference DNAs (1.5 mg of each) were labeled with Cy3- For immunohistochemistry, tissue sections were treated dUTP and Cy5-dUTP (Perkin-Elmer Life and Analytical as previously described and incubated overnight at 4 1C with Sciences Inc., Wellesley, MA, USA), respectively, and were a PLAG1 polyclonal antibody (diluted 1:200) raised against thereafter pooled and mixed with human Cot-1 DNA in a peptide in the N-terminal part of the protein (Asp et al., hybridization buffer (Agilent Oligo aCGH Hybridization Kit; 2006). Bound antibodies were visualized using the indirect Agilent Technologies), denatured and hybridized to the immunoperoxidase technique DAKO EnVision þ System arrays. The hybridizations were carried out in hybridization (DakoCytomation, Glostrup, Denmark). Control sections chambers, placed in a rotating oven, at 65 1C for 40 h. The were incubated identically, except for the primary antibody, arrays were subsequently washed, dried and scanned using a which was replaced by normal rabbitserum/mouse IgG. G2505B AgilentDNA microarray scanner (AgilentTechnol- ogies). Microarray images were analyzed using Feature Extraction v.9.1 (Agilent Technologies) with linear normal- Acknowledgements ization (protocol CGH-v4_91), and the data were subsequently imported into the CGH Analytics software v.3.4.27 (Agilent We thank Ulric Pedersen for expert help with the illustrations. Technologies). Detection of gains and losses were based on This work was supported by grants from the Swedish Cancer (i) the z-score algorithm (threshold 2.5) with a moving average Society, the IngaBritt and Arne Lundberg Research Founda- of 500 kb and (ii) visual inspection of the log2 ratios. In tion, the Assar Gabrielsson Research Foundation for Clinical general, log2 ratios X0.4 in atleastfive consecutiveprobes Cancer Research and the Sahlgrenska University Hospital were considered as a reliable copy number alteration. Foundations.

References

Abbott JJ, Erickson-Johnson M, Wang X, Nascimento AG, Oliveira Asp J, Persson F, Kost-Alimova M, Stenman G. (2006). CHCHD7- AM. (2006). Gains of COL1A1-PDGFB genomic copies occur in PLAG1 and TCEA1-PLAG1 gene fusions resulting from cryptic, fibrosarcomatous transformation of dermatofibrosarcoma protuberans. intrachromosomal 8q rearrangements in salivary gland pleomorphic Mod Pathol 19: 1512–1518. adenomas. Genes Chromosomes Cancer 45: 820–828.

Oncogene FGFR1–PLAG1 gene fusion and amplification F Persson et al 3080 A˚ stro¨ m AK, Voz ML, Kas K, Ro¨ ijer E, Wedell B, Mandahl N et al. Livak KJ, Schmittgen TD. (2001). Analysis of relative gene expression (1999). Conserved mechanism of PLAG1 activation in salivary data using real-time quantitative PCR and the 2(-Delta Delta C(T)) gland tumors with and without chromosome 8q12 abnormalities— method. Methods 25: 402–408. identification of SII as a new fusion partner gene. Cancer Res 59: Mark J, Dahlenfors R, Ekedahl C. (1983). Specificity and implications 918–923. of ring chromosomes and dicentrics in benign mixed salivary gland Barrett MT, Scheffer A, Ben-Dor A, Sampas N, Lipson D, Kincaid R tumours. Acta Pathol Microbiol Immunol Scand [A] 91: 397–402. et al. (2004). Comparative genomic hybridization using oligonucleo- Mark J, Dahlenfors R, Wedell B. (1997). Impactof the in vitro tide microarrays and total genomic DNA. Proc Natl Acad Sci USA technique used on the cytogenetic patterns in pleomorphic 101: 17765–17770. adenomas. Cancer Genet Cytogenet 95: 9–15. Brennan C, Zhang Y, Leo C, Feng B, Cauwels C, Aguirre AJ et al. Mark J, Sandros J, Wedell B, Stenman G, Ekedahl C. (1988). (2004). High-resolution global profiling of genomic alterations with Significance of the choice of tissue culture technique on the long oligonucleotide microarray. Cancer Res 64: 4744–4748. chromosomal patterns in human mixed salivary gland tumors. Eswarakumar VP, Lax I, Schlessinger J. (2005). Cellular signaling by Cancer Genet Cytogenet 33: 229–244. fibroblast growth factor receptors. Cytokine Growth Factor Rev 16: Nakao K, Mehta KR, Fridlyand J, Moore DH, Jain AN, Lafuente A 139–149. et al. (2004). High-resolution analysis of DNA copy number Eveson JW, Kusafuka K, Stenman G, Nagao T. (2005). Pleomorphic alterations in colorectal cancer by array-based comparative genomic adenoma. In: Barnes L, Eveson JW, ReichartP, Sidransky D (eds). hybridization. Carcinogenesis 25: 1345–1357. World Health Organization Classification of Tumours. Pathology NordkvistA, Mark J, Gustafsson H, Bang G, Stenman G. (1994). Non- and Genetics of Head and Neck Tumours. IARC Press: France, random chromosome rearrangements in adenoid cystic carcinoma of pp 254–258. the salivary glands. Genes Chromosomes Cancer 10: 115–121. Geurts JM, Schoenmakers EF, Ro¨ ijer E, Stenman G, Van de Ven WJ. Pole JC, Courtay-Cahen C, Garcia MJ, Blood KA, Cooke SL, (1997). Expression of reciprocal hybrid transcripts of HMGIC and Alsop AE et al. (2006). High-resolution analysis of chromosome FHIT in a pleomorphic adenoma of the parotid gland. Cancer Res rearrangements on 8p in breast, colon and pancreatic cancer reveals 57: 13–17. a complex pattern of loss, gain and translocation. Oncogene 25: Gisselsson D, Ho¨ glund M, Mertens F, Johansson B, Dal Cin P, Van 5693–5706. den Berghe H et al. (1999). The structure and dynamics of ring Pollack JR, Perou CM, Alizadeh AA, Eisen MB, Pergamenschikov A, chromosomes in human neoplastic and non-neoplastic cells. Hum Williams CF et al. (1999). Genome-wide analysis of DNA copy- Genet 104: 315–325. number changes using cDNA microarrays. Nat Genet 23: 41–46. Gisselsson D, Pettersson L, Ho¨ glund M, Heidenblad M, Gorunova L, Reis-Filho JS, Simpson PT, Turner NC, Lambros MB, Jones C, WiegantJ et al. (2000). Chromosomal breakage-fusion-bridge events Mackay A et al. (2006). FGFR1 emerges as a potential therapeutic cause genetic intratumor heterogeneity. Proc Natl Acad Sci USA 97: target for lobular breast carcinomas. Clin Cancer Res 12: 6652–6662. 5357–5362. Ro¨ ijer E, NordkvistA, Stro ¨ m AK, Ryd W, BehrendtM, Bullerdiek J Gorre ME, Mohammed M, Ellwood K, Hsu N, Paquette R, Rao PN et al. (2002). Translocation, deletion/amplification, and expression et al. (2001). Clinical resistance to STI-571 cancer therapy caused by of HMGIC and MDM2 in a carcinoma ex pleomorphic adenoma. BCR-ABL gene mutation or amplification. Science 293: 876–880. Am J Pathol 160: 433–440. Graux C, Cools J, Melotte C, Quentmeier H, Ferrando A, Levine R Roumiantsev S, Krause DS, Neumann CA, Dimitri CA, Asiedu F, et al. (2004). Fusion of NUP214 to ABL1 on amplified episomes in Cross NC et al. (2004). Distinct stem cell myeloproliferative/T T-cell acute lymphoblastic leukemia. Nat Genet 36: 1084–1089. lymphoma syndromes induced by ZNF198–FGFR1 and BCR– Hoffman MP, Kidder BL, Steinberg ZL, Lakhani S, Ho S, Kleinman FGFR1 fusion genes from 8p11 translocations. Cancer Cell 5: HK et al. (2002). Gene expression profiles of mouse submandibular 287–298. gland development: FGFR1 regulates branching morphogenesis Schoenmakers EF, Wanschura S, Mols R, Bullerdiek J, Van den in vitro through BMP- and FGF-dependent mechanisms. Develop- Berghe H, Van de Ven WJ. (1995). Recurrent rearrangements in the ment 129: 5767–5778. high mobility group protein gene, HMGI-C, in benign mesenchymal Hughes SE. (1997). Differential expression of the fibroblast growth tumours. Nat Genet 10: 436–444. factor receptor (FGFR) multigene family in normal human adult Sjo¨ gren H, Wedell B, Meis-Kindblom JM, Kindblom LG, Stenman G. tissues. J Histochem Cytochem 45: 1005–1019. (2000). Fusion of the NH2-terminal domain of the basic helix-loop- Johansson B, Fioretos T, Mitelman F. (2002). Cytogenetic and helix protein TCF12 to TEC in extraskeletal myxoid chondrosarco- molecular genetic evolution of chronic myeloid leukemia. Acta ma with translocation t(9;15)(q22;q21). Cancer Res 60: 6832–6835. Haematol 107: 76–94. Stenman G. (2005). Fusion oncogenes and tumor type specificity— Kas K, Voz ML, Hensen K, Meyen E, Van de Ven WJ. (1998). insights from salivary gland tumors. Semin Cancer Biol 15: 224–235. Transcriptional activation capacity of the novel PLAG family of Ugolini F, Adelaide J, Charafe-JauffretE, Nguyen C, Jacquemier J, zinc finger proteins. J Biol Chem 273: 23026–23032. Jordan B et al. (1999). Differential expression assay of chromosome Kas K, Voz ML, Ro¨ ijer E, A˚ stro¨ m AK, Meyen E, Stenman G et al. arm 8p genes identifies Frizzled-related (FRP1/FRZB) and Fibro- (1997). Promoter swapping between the genes for a novel zinc finger blast Growth Factor Receptor 1 (FGFR1) as candidate breast protein and b-catenin in pleomorphic adenomas with t(3;8)(p21;q12) cancer genes. Oncogene 18: 1903–1910. translocations. Nat Genet 15: 170–174. Voz ML, A˚ stro¨ m AK, Kas K, Mark J, Stenman G, Van de Ven WJ. Kok K, Naylor SL, Buys CH. (1997). Deletions of the short arm of (1998). The recurrent translocation t(5;8)(p13;q12) in pleomorphic in solid tumors and the search for suppressor genes. adenomas results in upregulation of PLAG1 gene expression under Adv Cancer Res 71: 27–92. control of the LIFR promoter. Oncogene 16: 1409–1416. Kost-Alimova M, Imreh S. (2007). Modeling non-random deletions in Voz ML, Mathys J, Hensen K, Pendeville H, Van Valckenborgh I, cancer. Semin Cancer Biol 17: 19–30. Van Huffel C et al. (2004). Microarray screening for target genes of Kost-Alimova M, Kiss H, Fedorova L, Yang Y, Dumanski JP, Klein the proto-oncogene PLAG1. Oncogene 23: 179–191. G et al. (2003). Coincidence of synteny breakpoints with malig- Xiao S, Nalabolu SR, Aster JC, Ma J, Abruzzo L, Jaffe ES et al. nancy-related deletions on human chromosome 3. Proc Natl Acad (1998). FGFR1 is fused with a novel zinc-finger gene, ZNF198, in Sci USA 100: 6622–6627. the t(8;13) leukaemia/lymphoma syndrome. Nat Genet 18: 84–87.

Oncogene