(2009) 23, 332–339 & 2009 Macmillan Publishers Limited All rights reserved 0887-6924/09 $32.00 www.nature.com/leu ORIGINAL ARTICLE

Detection and molecular monitoring of FIP1L1-PDGFRA-positive disease by analysis of patient-specific genomic DNA fusion junctions

J Score1, C Walz2, JV Jovanovic3, AV Jones1, K Waghorn1, C Hidalgo-Curtis1, F Lin1, D Grimwade3, F Grand1, A Reiter2 and NCP Cross1

1Wessex Regional Genetics Laboratory, University of Southampton, Salisbury, UK; 2III. Medizinische Universita¨tsklinik, Medizinische Fakulta¨t Mannheim der Universita¨t Heidelberg, Mannheim, Germany and 3Department of Medical and Molecular Genetics, King’s College London School of Medicine, London, UK

To evaluate current detection methods for FIP1L1-PDGFRA in the presence of the FIP1L1-PDGFRA fusion predicts a hypereosinophilic syndrome (HES), we developed a means to favourable response to the small molecule inhibitor rapidly amplify genomic break points. We screened 202 cases and detected genomic junctions in all samples previously and most positive patients show dramatic responses to therapy identified as RT-PCR positive (n ¼ 43). Genomic fusions were with rapid normalization of peripheral counts and amplified by single step PCR in all cases whereas only 22 (51%) achievement of nested or real-time quantitative reverse tran- were single step RT-PCR positive. Importantly, FIP1L1-PDGFRA scription polymerase chain reaction (RQ-PCR) negativity.6–10 was detected in two cases that initially tested negative by RT- Identification of patients who are positive for FIP1L1-PDGFRA is PCR or fluorescence in situ hybridization. Absolute quantita- therefore critical to appropriate clinical management.11,12 tion of the fusion by real-time PCR from genomic DNA (gDNA) Exclusion of FIP1L1-PDGFRA is also important as some patients using patient-specific primer/probe combinations at presenta- 13 tion (n ¼ 13) revealed a 40-fold variation between patients may benefit from alternative therapies such as mepolizumab. (range, 0.027–1.1 FIP1L1-PDGFRA copies/haploid genome). In Unambiguous detection of FIP1L1-PDGFRA, however, is follow up samples, quantitative analysis of gDNA gave 1–2 log complicated by several factors. Fluorescence in situ hybridiza- greater sensitivity than RQ-PCR of cDNA. Minimal residual tion (FISH) to detect heterozygous of CHIC2,a disease assessment using gDNA showed that 11 of 13 patients surrogate marker for FIP1L1-PDGFRA, can be difficult because achieved complete molecular response to imatinib within a 9 median of 9 months (range, 3–17) of starting treatment, with a of the inherent background fluorescence of and the sensitivity of detection of up to 1 in 105. One case relapsed with variability in the proportion of cells involved in the malignant 14,15 an acquired D842V . We conclude that detection of clone, which is often low and may overlap with the FIP1L1-PDGFRA from gDNA is a useful adjunct to standard intrinsic background false-positive rate of this technique. diagnostic procedures and enables more sensitive follow up of Reverse transcriptase polymerase chain reaction (RT-PCR) is positive cases after treatment. complicated by the considerable diversity of break points Leukemia (2009) 23, 332–339; doi:10.1038/leu.2008.309; within FIP1L1, the complex of FIP1L1 and published online 6 November 2008 4,9,16 Keywords: imatinib; FIP1L1-PDGFRA; HES; MRD the variable use of cryptic splice sites in the . Indeed, FIP1L1-PDGFRA transcripts can often only be detected by sensitive nested RT-PCR assays, even in previously untreated patients.4,11,17 Nested RT-PCR has a greatly enhanced risk of contamination resulting in false-positive results, potentially Introduction leading to misdiagnosis and inappropriate treatment. Although real time quantitative RT-PCR (RQ-PCR) can detect FIP1L1- Persistent can arise from clonal disorders of the PDGFRA in many cases,7 the diversity of mRNA fusions makes bone marrow, or from a number of reactive conditions. The the design of workable primer/probe sets that are capable of diagnosis of idiopathic hypereosinophilic syndrome is one of detecting all variants very difficult. Currently it is unclear to exclusion and the disease is characterized as a persistent state of what extent these concerns actually impact on current unexplained eosinophilia (total eosinophil count 41.5 Â 109/l) diagnostic methods to assess this, we have developed a for longer than 6 months, with evidence of organ dysfunction multiplex Long-Range PCR (LR-PCR) technique to detect due to eosinophilic tissue infiltration and the consequent release genomic DNA (gDNA) break points which are unique to each of their granules and toxic contents.1 If clonal proliferation of patient, enabling a definitive diagnosis. We demonstrate that the eosinophils can be shown, then the disease should be amplification of DNA break points enables more robust reclassified as chronic eosinophilic leukaemia and treated detection of FIP1L1-PDGFRA in patients at diagnosis and a accordingly.2,3 more sensitive means of detecting minimal residual disease The finding of a cryptic interstitial deletion on (MRD) in patients undergoing targeted therapy. 4q12 that fuses the FIP1-Like-1 gene (FIP1L1) to the platelet- derived growth factor a gene (PDGFRA) revolutionized 4,5 the diagnosis of chronic eosinophilic leukaemia. Importantly, Materials and methods

Correspondence: Professor NCP Cross, Wessex Regional Genetics Patient samples Laboratory, University of Southampton, Salisbury NHS Foundation Peripheral blood or bone marrow samples were referred for Trust, Salisbury, Wilts SP2 8BJ, UK. E-mail: [email protected] molecular diagnostic analysis following informed consent Received 2 September 2008; revised 8 October 2008; accepted 8 according to the Declaration of Helsinki. The use of residual October 2008; published online 6 November 2008 material for research was approved by the South Wiltshire Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 333 Research Ethics Committee. Of the cases analysed by RQ-PCR AAGGAAAGAG-30) and PDA.Ex.22.2R (50-CCTGAAGAGTGA from cDNA, results are included from three (cases E359, E370 CCATCCTGCTGT-30) in the second step. PCR products were and E630) that were published in a previous study.7 directly sequenced using the second step primers.

Nested reverse transcriptase-PCR Expand long-range PCR Leukocytes were lysed with guanidium thiocyanate and total Genomic DNA was extracted from peripheral blood leuko- RNA extracted using Qiagen RNeasy system (Qiagen, West cytes by standard procedures. For detection of genomic Sussex) following the manufacturers guidelines. The RNA was DNA break points, the Expand Long Template PCR System reverse transcribed into cDNA with random hexamers and (Roche, Burgess Hill, UK) was used with System 2 and an tested for the adequate cDNA quality by amplification of the annealing temperature of 64 1C according to the manufac- normal BCR gene as described earlier.18 This test also excluded turer’s instructions. All primers were designed using Primer3 the presence of BCR-ABL. FIP1L1-PDGFRA was tested for (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) to by both single step and nested RT-PCR using primers FIP1E6-F1 be 23 bases long with a melting temperature of 66 1C. Fifteen (50-CACCTGGAAGCATTAATGGAG-30) and PDA-R1 (50-TGAG FIP1L1 forward primers were designed approximately every 5 kb AGCTTGTTTTTCACTGGA-30) for first step and FIP1E6-F2 (50-A from FIP1L1 intron 5 to intron 16, with the exception of a region GTTCCACTCTTAGAGGTAG-30) and PDA-R2 (50-GGGACCGG of approximately 12 kb in intron 10 for which primers could not CTTAATCCATAG-30) for the nested step, a procedure modified be designed due to the highly repetitive nature of the sequence. slightly from that described earlier.4 For cases that were For the purposes of the multiplex technique, the 15 FIP1L1 relapsed, the region of the fusion encoding the kinase domain forward primers were split up into three mixes of five primers was specifically amplified from cDNA using primers FIP.Ex. ensuring that the primers in each group were more than 15 kb 10.1F (50-CAGCAGGGAAGAACTGGAAACTCA-30) and PDA. apart to avoid amplification of more than one break point Ex.22.1R (50-GTGGTCTGGATGAGCAGAGACTGAG-30) in the product (see Figure 1 and Table 1). A generic reverse primer in first step followed by PDA.Ex16.1F (50-TGTCCCCATGCTAGA PDGFRA intron 12 located downstream of the break point

Int 10 Int 11 Int 13 Int 15 Mix 1: (22850) (40225) (56273) (70370)

Int 9 Int 10 Int 11 Int 13 16 Mix 2: (15351) (34060) (44469) (60580) (75275)

Int 9 Int 11 Int 12 Int 14 Int 16 Mix 3: (20848) (38554) (48718) (65181) (80314)

9 10 11 12 13 14 15 16 17 18

Position of breakpoints for nested RT- Position of FIP1L1 forward primer Position of breakpoint PCR negative cases Figure 1 Positions of FIP1L1-screening primers and patient gDNA break points. The thick horizontal line represents the FIP1L1 gene from exon 9 to the end of exon 18 and the black vertical lines represent each exon. The FIP1L1 forward screening primers are represented by the horizontal arrows with their names and which of the screening mixes they are in, above, except the FIP1L1 int 5 (11403) forward primer which was included in Mix 1. The position of the respective genomic break point for each patient is represented by a vertical arrow, with the two smaller arrows showing break point locations of the two patients who were nested RT-PCR negative but gDNA multiplex positive.

Table 1 Primers used in the multiplex gDNA PCR

Mix Primers Sequence 50–30

1 FIP Intron 5 (11403) F TCTTGGTCATTAACCACCCACCA FIP Intron 10 (22850) F CTCGGCCAGATCACTTACCTTTTG FIP Intron 11 (40225) F AAATGTGGAGTTTGAGGCATTGCT FIP Intron 13 (56273) GTGGCGAGTTGCTCATATCTAGGG FIP Intron 15 (70370) F TTGGAACATGCCTTTCTCCACCT 2 FIP Intron 9 (15351) F AAATAGGGCCATGATGAGCAGGA FIP Intron 10 (34060) F ATGCTGGGTTTCCAACGTTCTGT FIP Intron 11 (44469) F GGGTTGTGAAACAATTTGGTGGTC FIP Intron 13 (60580) F AGGGGAAAGAAGAGGGACGTTTG FIP Exon 16 (75275) F GACACCAGCAAGCAGTGGGACTA 3 FIP Intron 9 (20848) F CTTTGCCTCATGGGAAAGGAGAG FIP Intron 11 (38554) F TGAACCTAAAAGGGCCAGACAACA FIP Intron 12 (48718) F GTGGGCCTGAGTTTTGATTGTGA FIP Intron 14 (65181) F CCCGATATCAAAACAGGGCAAAG FIP Intron 16 (80314) F CCCCTGCCCACACAGATGTAAT All PDA intron 12 R1 TGTGCAAGGGAAAAGGGAGTCTT

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 334 cluster in exon 12 was used in all mixes. The multiplex PCRs performed using PrimerDesign (Southampton, UK) master mix were optimized initially with 16 known FIP1L1-PDGFRA under standard conditions, with 5 ml of gDNA (approximately positive samples and 20 healthy controls to ensure no non- 100 ng). PCRs were set up on a RotorGene CAS1200 robot and specific amplification. For all subsequent screens we included run on the RotorGene 6000 platform (Cambridge, UK). positive, negative and no template controls on each run. In addition, a Long-Range PCR using ABL primers (ABL F 50-TTCT GGGGAAATTGCCTGTCATT-30 and ABL R 50-AAACATTCTGC Relative quantification of gDNA fusions for MRD CGCATCTGGATT-30 giving a control band of 5.4 kb) was detection performed on each sample to ensure good quality gDNA. The PDGFRA and albumin (ALB) gDNA probes and primers were tested on serial dilutions of normal control gDNA, to ensure that they gave good standard curves, as defined 20 RQ-PCR of gDNA break points elsewhere. Patient samples were then run with the appropriate Quantitative gDNA analysis was performed for 13 cases that forward primers and relative amounts of FIP1L1-PDGFRA 24 were treated with imatinib and for whom presentation and at calculated using the DCt method with normalization against least three follow up samples were available. PDGFRA primers the ALB results for each sample. To assess the analytical and probes were designed on Molecular Beacon software by sensitivity with which residual disease could be excluded for Sigma (Gillingham, UK). For the PDGFRA probe to be suitable negative results, we estimated the number of haploid genomes for detection of all possible break points it had to be designed in as described above. The results for relative quantification were a very small region 30 of the exon 12 break point cluster region expressed as the ratio of gDNA FIP1L1-PDGFRA/ALB. All and therefore locked nucleic acids (LNA) were incorporated to samples, negative controls and no template controls were run increase specificity of the probe19 LNA probe 50-cctCccAa in triplicate and samples only classed as positive if amplification gActCcct-30 (capital letters indicate LNA oligos). All FIP1L1 was seen at least in two of the three wells with a Ct value of 40 patient-specific forward primers were designed using Primer3 or less. with the same annealing temperature as the generic PDGFRA reverse primer (55 1C). The gDNA RQ-PCR was tested initially on FIP1L1-PDGFRA positive presentation samples, normal FISH controls and no template controls to ensure patient-specific FISH was performed on six patients for whom cytogenetic amplification of break points. The probe was also tested on cultures were available, using the Poseidon FIP1L1-CHIC2- PDGFRA dual-labelled deletion break probe (Kreatech, serial dilutions of normal control DNA with the PDGFRA 25 forward and reverse primers (PDGFRA-F: 50-ACTAGTGCTTGG- Amsterdam, Holland) and hybridized as described earlier. TAAGTTCC-3; PDGFRA-R: 50-CCTATAAATTGTAAAGTTGTGT For all patients 100 interphase cells were scored. GC-30) and on three patients with the appropriate FIP1L1 forward primers to ensure the assays conformed to the EAC 20 RQ-PCR from cDNA D 7 criteria, that is, a Rn of at least one in diagnostic material and RQ-PCR from cDNA was performed as described except the 4 an efficiency 95%. Absolute quantification was used for RotorGene CAS1200 robot and RotorGene 6000 were used, the determining the levels of gDNA fusion at presentation and probe was labelled with FAM and BHQ1 and 2 ml of cDNA was relative quantification was used to determine patients response used for each reaction using PrimerDesign master mix. New to imatinib therapy. forward primers were designed for two break points. As a control for cDNA quality and quantity, normal ABL transcripts were quantified by RQ-PCR as described earlier 20 and results Absolute quantification of gDNA fusions in pre-imatinib were expressed as the ratio of FIP1L1-PDGFRA/ABL. samples Each patient-specific gDNA fusion was amplified using Hotstart PCR amplification conditions (AmpliTaq Gold, Roche) with Results 64 1C annealing temperature and 30 cycles. Each fusion was cloned using the TOPO pCR4 kit (Invitrogen, Paisley, UK), Amplification of FIP1L1-PDGFRA from gDNA extracted, linearized by restriction digest, cleaned and quanti- We screened gDNA from 202 cases referred for IHES or fied on a NanoDrop ND-1000 spectrophotometer. For each persistent unexplained eosinophilia, none of whom had been patient/primer/probe combination we estimated the routine treated with imatinib. We found that all cases previously analytical sensitivity of the gDNA assay to be o10 copies in a determined to be positive by RT-PCR were also positive by 4 5 background of 10 –10 normal cells using 500 ng input gDNA gDNA multiplex LR-PCR (n ¼ 43; Figures 1 and 2). Encoura- using serial dilutions of plasmid DNA in normal control DNA gingly, genomic fusions were amplified by single step PCR in all (data not shown). The number of copies of the fusion for each cases whereas only 22 (51%) of these samples were single step patient sample was then calculated and serial dilutions of each RT-PCR positive. Amplicons ranged in size from 404 bp–8.6 kb patient-specific plasmid made into normal control gDNA and in all cases were confirmed by sequence analysis, showing (Roche). Each pre-imatinib sample was then compared against that the FIP1L1-PDGFRA junctions were unique to each patient. its plasmid standard curve. As a control for the amount of amplifiable gDNA for each sample we estimated the number of haploid genomes by RQ-PCR for the albumin gene. To do Improved detection of FIP1L1-PDGFRA from gDNA this, all patient samples were compared against a standard curve compared with nested RT-PCR for the albumin gene using previously described conditions21 Unexpectedly, FIP1L1-PDGFRA was detected by single step and serial dilutions of human genomic DNA (Roche). Each result PCR of gDNA in two cases that had been previously scored as was then converted into estimates of the actual amounts of nested RT-PCR negative. Repeat samples from these patients amplifiable gDNA in mg and multiplied by 3.3 Â 105 to estimate were tested by RT-PCR and both were found to be positive (one the number of amplifiable haploid genomes.22,23 PCRs were case was positive in only two of three replicate nested

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 335

E370, E1025, E1131 E1336 E43 E591 E513 E513 E370 E1279 E176 E1336 E1131 Ladder Ladder E1279 E176 E614

E370 (FIP1L1-PDGFRA positive)

12 kb Negative controls 2 kb Mix 1 Normalized fluorescence Normalized E176 E614 E591 E513 E759 E359 E1336 E905 E1401 E1131 E606 E513 E1279 E939 Ladder Cycle number

12 kb 100% 2% 0.01% 2 kb 20% 1% 0.001% Mix 2 10% 0.1% neg Normalized fluorescence Normalized Ladder E243 E243 E344 E1025 E606 E606 E344 E1025

12 kb 2 kb Cycle number Mix 3 Ct

Figure 2 Agarose gels showing examples of FIP1L1-PDGFRA LR-PCR multiplex results. The smallest band amplified was E176 mix 2 at 404 bp, and the largest band amplified was E370 mix 1 at 8680 bp. Ladder ¼ 1 kb Plus (Invitrogen, Paisley, Scotland). Concentration

Figure 3 Examples of FIP1L1-PDGFRA RQ-PCR analysis. (a) RQ-PCR graph showing normalized fluorescence against number of cycles. The three lines showing increasing fluorescence correspond to gDNA from reactions). Sequencing of the amplicons revealed typical mRNA three FIP1L1-PDGFRA-positive cases: E1025 (green line), E1131 (red fusions between FIP1L1 exon 13 for both patients and part of line) and E370 (purple line). No amplification was seen from either the PDGFRA exon 12. Despite the fact that both initial cDNAs had normal control gDNAs or the no template controls. (b) RQ-PCR graph passed our routine quality control procedure (qualitative showing a dilution series of E606 gDNA in normal control gDNA as amplification of normal BCR mRNA18), we speculated that indicated with the calculated cycle threshold line marked (lines showing no amplification are the normal control gDNAs and the no the discrepant results were due to relatively poor quality template control). (c) E606 standard curve showing the cycle threshold cDNA, possibly exacerbated by the known variability in data against concentration of gDNA, r2 value of 0.998 and FIP1L1-PDGFRA expression between cases. Quantitation of amplification efficiency of 95%. A similar analysis was performed normal ABL transcripts by RQ-PCR revealed 1–2 Â 103 tran- for each case. scripts per 2 ml of cDNA, which is perfectly adequate to detect other fusions such as BCR-ABL in pretreatment specimens, but which is relatively low compared with the level of X104 median of 0.13 copies of FIP1L1-PDGFRA per haploid genome (sensitivity X0.01%) that is generally considered as suitable (equivalent to a median 13% fusion-positive cells) with a 40-fold quality for MRD analysis.26 variation between patients (range, 2.7–100% fusion-positive cells).

Variability of FIP1L1-PDGFRA expression Previous RQ-PCR studies revealed considerable variation in Comparison of gDNA RQ-PCR and FISH FIP1L1-PDGFRA expression levels in patients prior to imatinib The gDNA RQ-PCR results suggested that the clone size may be treatment.7 To investigate this further, we estimated the too small to be easily detected by FISH in some cases. Of the 13 proportion of fusion-positive cells in pretreatment patient cases analysed above, six had contemporaneous peripheral samples as assessed by RQ-PCR of gDNA fusion junctions blood-derived fixed cells suitable for interphase FISH. As shown compared with the ALB gene as a measure of total DNA. For on Table 2, the results between the two techniques were broadly each patient we designed specific gDNA primers that were used concordant. The two patients with only 3% FIP1L1-PDGFRA in combination with a generic probe. As expected, primer/probe cells estimated by gDNA RQ-PCR (cases E513 and E2002) had 6 combinations were specific to each patient (Figure 3a) and and 4% deleted positive cells detectable by FISH compared with demonstrated linear results on analysis of a dilution series of a background false-positive rate of 5%.14,15 This suggests that presentation DNA (Figures 3b and c). Absolute quantification of FISH screening alone might miss some FIP1L1-PDGFRA-positive the fusion at presentation (n ¼ 13) by gDNA RQ-PCR revealed a cases.

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 336 Comparison of cDNA and gDNA methods to assess the PDGFRa domain in the most recent samples sensitivity of detecting MRD revealed only normal sequence for E370, but E2245 had an A to Having established that amplification from gDNA was more T sequence change at position 2473, which is predicted to result reliable than cDNA for detection of FIP1L1-PDGFRA in pre- in a D842V substitution. This sequence change was not imatinib samples, we next went on to examine the utility of detectable in the pre-imatinib sample (Figure 5). gDNA analysis for assessing levels of MRD. Initially, we designed patient-specific nested gDNA primers for seven cases and found that 11/31 post-imatinib data points were discordant Comparison of gDNA and cDNA RQ-PCR when compared with nested RT-PCR, with 10/11 being gDNA Of the 13 patients monitored using gDNA after imatinib positive, cDNA negative (not shown). This suggested that gDNA treatment, nine were amenable to RQ-PCR analysis from cDNA 7 analysis was a more sensitive indicator of MRD. To examine this because of limitations imposed by the position of the probe and in more detail, we developed patient-specific RQ-PCR gDNA seven of these had sequential samples available for analysis. Of assays. We analysed 66 samples from 13 patients who under- the 31 post-treatment data points available for comparison, five went treatment with imatinib (median dose ¼ 100 mg/day, were discordant (gDNA positive, cDNA negative; see Figure 6). median follow-up ¼ 25 months; range, 9–58) and initially In contrast, there were no instances where cDNA-based compared the results to those obtained by qualitative single RQ-PCR was positive and gDNA-based RQ-PCR was negative. step or nested RT-PCR. The gDNA RQ-PCR also showed 1–2 logs higher sensitivity for Comparison of gDNA RQ-PCR and nested RT-PCR revealed exclusion of residual disease that tested negative for FIP1L1- large discrepancies with 49% (31/63) samples positive by gDNA PDGFRA compared with cDNA (Figures 4 and 6). RQ-PCR but only 31% (19/61) were positive by nested RT-PCR, again highlighting the inaccuracies of the standard RT-PCR technique. Nevertheless, MRD assessment using gDNA showed Discussion that 11 of 13 patients achieved a complete molecular response to imatinib within a median of 9 months (range, 3–17) of starting In contrast to other leukaemia-specific gene fusions, nested treatment (Figure 4). The median sensitivity with which FIP1L1- RT-PCR is routinely required to detect FIP1L1-PDGFRA in many 4,9 PDGFRA could be excluded was 1 in 30 000. pretreatment cases even when good quality cDNA is used. Two cases (E370 and E2245) failed to achieve a molecular response. Sequencing of the region encoding the FIP1L1-

Presentation Table 2 Comparison of the percentage of FIP1L1-PDGFRA- positive cells estimated by gDNA RQ-PCR and FISH

Sample ID gDNA RQ-PCR Interphase FISH (% cells positive) (% cells positive)

E359 20 41 Relapse E513 3 6 E566 12 14 E606 11 63 E759 22 26 E2002 3 4 Figure 5 Sequence electropheregrams showing the emergence of the imatinib resistance mutation in the PDGFRa kinase domain changing Abbreviations: FISH, Fluorescence in situ hybridization; RQ-PCR, real- codon GAC (Asp) in the presentation sample to GTC (Val) in the time quantitative reverse transcription polymerase chain reaction. relapse sample.

-2 E243

E43 -1 E359 0 E591 E759 1 E1025 2 E1279 FIP1L1-PDGFRA E370 3 E630 4 E614 E1131 5

Log reduction E606 E2245 6 0 102030405060 Time from presentation (months) Figure 4 Graph showing log reduction from patient-specific baselines in the levels of gDNA FIP1L1-PDGFRA fusion following time of start in imatinib therapy for 13 patients. The solid data points represent positive and empty data points represent negative results, respectively.

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 337 RT RT-PCR single positive 0 E359 RT-PCR nested positive -1 RT-PCR nested negative -2 RT-PCR failed -3 cDNA RQ-PCR positive -4 cDNA RQ-PCR negative -5 gDNA RQ-PCR positive 0102030 gDNA RQ-PCR negative

RT RT

0 E614 0 E606 -1 -1 -2 -2 -3 -3 -4 -4 -5 -5

01020300102030 FIP1L1-PDGFRA

RT RT 0 0 E370 -1 -1 E1131 -2 -2 -3 -3 Log reduction -4 -4 -5 -5

01020300102030

RT RT 0 0 -1 E2245 -1 E130 -2 -2 -3 -3 -4 -4 -5 -5

01020300102030 Months on imatinib

Figure 6 Graphs showing log reduction in FIP1L1-PDGFRA fusion over time (in months) following start of imatinib therapy. Three different techniques for minimal residual disease detection were compared; RT-PCR (RT; triangles), cDNA RQ-PCR (pink lines and squares) and gDNA RQ-PCR (blue lines and circles).

We tested several different FIP1L1 and PDGFRA primers but negative by FISH, however it is important to stress that our were unable to identify a combination that was able to routinely analysis was performed on peripheral blood rather than bone amplify fusions by single step amplification (not shown). To try marrow. In our six cases, the median proportion of FISH-positive and improve the detection of FIP1L1-PDGFRA, we developed a cells was only 20% compared with a median of 80% in 10 cases single step, multiplex long-range PCR technique based on for whom bone marrow was analysed.14 Although further detection of gDNA break points. We designed and optimized comparative FISH data of blood and bone marrow are three multiplex PCRs each consisting of 5 forward primers warranted, these results suggest that the absence of CHIC2 spanning the break point region within FIP1L1 (intron 5 to intron deleted cells by interphase FISH should be interpreted with 16) and a single reverse primer in PDGFRA intron 12. The caution when peripheral blood is used for analysis. multiplexes amplified FIP1L1-PDGFRA genomic junction se- RQ-PCR from gDNA confirmed a considerable degree of quences in all 43 cases previously shown to be fusion positive variation in the proportion of FIP1L1-PDGFRA-positive cells by single step or nested RT-PCR. FIP1L1-PDGFRA was also between patients prior to imatinib therapy. However, this detected in two patients previously thought to be nested RT-PCR variation was much smaller (40-fold versus 1000-fold) than that negative, despite the initial cDNAs having passed standard previously described for FIP1L1-PDGFRA mRNA levels,7 quality control criteria which have been proved to be robust for suggesting that the difficulties in RT-PCR detection is only partly diagnostic detection of BCR-ABL.18 On account of the high because of variable clone size. This is consistent with the degree of variability in FIP1L1-PDGFRA expression between observation that gDNA PCR was able to amplify FIP1L1- patients, we suggest that the minimum cDNA quality criteria PDGFRA junction sequences in all pre-imatinib cases by single similar to those employed for residual disease analysis26 should step amplification, whereas only half were single step RT-PCR also apply to diagnostic RT-PCRs for FIP1L1-PDGFRA. If these positive. For assessment of MRD following imatinib treatment, criteria are applied then it appears that nested RT-PCR is a we demonstrated that gDNA RQ-PCR was 1–2 logs more robust technique to detect FIP1L1-PDGFRA in pretreatment sensitive than RNA-based analysis. Nevertheless, we found that cases. 11 of 13 cases achieved a complete gDNA molecular response We also found that the gDNA multiplex PCR easily detected to imatinib in a median of 9 months, confirming the excellent FIP1L1-PDGFRA in two cases that were negative or borderline response of this disease to targeted therapy.4,6,8,15 The superior

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 338 sensitivity of gDNA analysis may be particularly useful for novel 9 Cools J, Stover EH, Gilliland DG. Detection of the FIP1L1-PDGFRA approaches to disease management such as determination of fusion in idiopathic hypereosinophilic syndrome and chronic optimal dosing for individual cases.27 Analysis of gDNA also has eosinophilic leukemia. Methods Mol Med 2006; 125: 177–187. an advantage in that results can be related readily to numbers of 10 Klion AD, Noel P, Akin C, Law MA, Gilliland DG, Cools J et al. Elevated serum tryptase levels identify a subset of patients with a malignant cells, something that is problematic for RT-PCR-based myeloproliferative variant of idiopathic hypereosinophilic syn- approaches. drome associated with tissue fibrosis, poor prognosis, and imatinib Of the two cases that did not achieve a complete molecular responsiveness. Blood 2003; 101: 4660–4666. response, one was found to have a D842V mutation. This 11 Chung KF, Hew M, Score J, Jones AV, Reiter A, Cross NC sequence variant has not been reported earlier in FIP1L1-PDGFRA, et al. Cough and hypereosinophilia due to FIP1L1-PDGFRA but is the most common mutation seen in imatinib-resistant, fusion gene with tyrosine kinase activity. Eur Respir J 2006; 27: 28 230–232. PDGFRA-mutated gastrointestinal stromal tumours. D842V 12 Vigna E, Lucia E, Gentile M, Mazzone C, Bisconte MG, Gentile C corresponds to the D816V mutation seen in systemic , et al. PDGFRalpha/FIP1L1-positive chronic eosinophilic leukemia an activating mutation that is also refractory to inhibition with presenting with retro-orbital localization: efficacy of imatinib imatinib.29,30 The suboptimal response in the second case was treatment. Cancer Chemother Pharmacol 2008; 61: 713–716. believed to be a consequence of poor treatment compliance. 13 Rothenberg ME, Klion AD, Roufosse FE, Kahn JE, Weller PF, In summary, the exquisite sensitivity of FIP1L1-PDGFRA- Simon HU et al. Treatment of patients with the hypereosinophilic positive cases to imatinib means that accurate detection of the syndrome with mepolizumab. N Engl J Med 2008; 358: 1215–1228. 14 La SR, Specchia G, Cuneo A, Beacci D, Nozzoli C, Luciano L et al. fusion is very important for optimal clinical management. We The hypereosinophilic syndrome: fluorescence in situ hybridization have shown here that analysis of gDNA break points assists in detects the del(4)(q12)-FIP1L1/PDGFRA but not genomic rearrange- the initial diagnosis of FIP1L1-PDGFRA-positive disease and ments of other tyrosine kinases. Haematologica 2005; 90: 596–601. consequently we have adopted this into our routine laboratory 15 Pardanani A, Ketterling RP, Brockman SR, Flynn HC, Paternoster practice. We have also shown that analysis of gDNA break SF, Shearer BM et al. CHIC2 deletion, a surrogate for FIP1L1- points provide a means to monitor MRD with high levels of PDGFRA fusion, occurs in systemic mastocytosis associated with eosinophilia and predicts response to imatinib mesylate therapy. sensitivity. Although this may be of benefit for specific research Blood 2003; 102: 3093–3096. applications, this approach involves a considerable amount of 16 Gotlib J, Cools J, Malone III JM, Schrier SL, Gilliland DG, Coutre work and it remains to be established if the extra log in SE. The FIP1L1-PDGFRalpha fusion tyrosine kinase in hyper- sensitivity is of routine clinical benefit. eosinophilic syndrome and chronic eosinophilic leukemia: implications for diagnosis, classification, and management. Blood 2004; 103: 2879–2891. Acknowledgements 17 Score J, Curtis C, Waghorn K, Stalder M, Jotterand M, Grand FH et al. Identification of a novel imatinib responsive KIF5B-PDGFRA This work was supported by Leukaemia Research (UK), Deutsche fusion gene following screening for PDGFRA overexpression in patients with hypereosinophilia. Leukemia 2006; 20: 827–832. Jose´ Carreras Leuka¨mie-Stiftung e.V. (Grant no. DJCLS R06/02, 18 Cross NC, Melo JV, Feng L, Goldman JM. An optimized multiplex AH) and the European LeukemiaNet within the 6th European polymerase chain reaction (PCR) for detection of BCR-ABL Framework Program for Research and Technology Development. fusion mRNAs in haematological disorders. Leukemia 1994; 8: 186–189. 19 Costa JM, Ernault P, Olivi M, Gaillon T, Arar K. Chimeric LNA/ DNA probes as a detection system for real-time PCR. Clin Biochem References 2004; 37: 930–932. 20 Gabert J, Beillard E, van der Velden VH, Bi W, Grimwade D, 1 Vardiman JW, Harris NL, Brunning RD. The World Health Pallisgaard N et al. Standardization and quality control studies of Organization (WHO) classification of the myeloid neoplasms. ‘real-time’ quantitative reverse transcriptase polymerase chain Blood 2002; 100: 2292–2302. reaction of fusion gene transcripts for residual disease detection in F 2 Bain BJ. Eosinophilic leukaemias and the idiopathic hypereosino- leukemia a Europe Against Cancer program. Leukemia 2003; 17: philic syndrome. Br J Haematol 1996; 95: 2–9. 2318–2357. 3 Brito-Babapulle F. The , including the idiopathic 21 Pongers-Willemse MJ, Verhagen OJ, Tibbe GJ, Wijkhuijs AJ, hypereosinophilic syndrome. Br J Haematol 2003; 121: 203–223. der Hass V, Roovers E et al. Real-time quantitative PCR for the 4 Cools J, DeAngelo DJ, Gotlib J, Stover EH, Legare RD, Cortes J detection of minimal residual disease in acute lymphoblastic et al. A tyrosine kinase created by fusion of the PDGFRA and leukemia using junctional region specific TaqMan probes. FIP1L1 as a therapeutic target of imatinib in idiopathic Leukemia 1998; 12: 2006–2014. hypereosinophilic syndrome. N Engl J Med 2003; 348: 1201–1214. 22 Brisco MJ, Condon J, Sykes PJ, Neoh SH, Morley AA. Detection and 5 Griffin JH, Leung J, Bruner RJ, Caligiuri MA, Briesewitz R. quantitation of neoplastic cells in acute lymphoblastic leukaemia, Discovery of a fusion kinase in EOL-1 cells and idiopathic by use of the polymerase chain reaction. Br J Haematol 1991; 79: hypereosinophilic syndrome. Proc Natl Acad Sci USA 2003; 100: 211–217. 7830–7835. 23 Jeffreys AJ, Wilson V, Neumann R, Keyte J. Amplification of 6 Baccarani M, Cilloni D, Rondoni M, Ottaviani E, Messa F, human minisatellites by the polymerase chain reaction: towards Merante S et al. The efficacy of imatinib mesylate in patients with DNA fingerprinting of single cells. Nucleic Acids Res 1988; 16: FIP1L1-PDGFRalpha-positive hypereosinophilic syndrome. Results 10953–10971. of a multicenter prospective study. Haematologica 2007; 92: 24 Livak KJ, Schmittgen TD. Analysis of relative data 1173–1179. using real-time quantitative PCR and the 2(-Delta Delta C(T)) 7 Jovanovic JV, Score J, Waghorn K, Cilloni D, Gottardi E, Method. Methods 2001; 25: 402–408. Metzgeroth G et al. Low-dose imatinib mesylate leads to rapid 25 Baxter EJ, Hochhaus A, Bolufer P, Reiter A, Fernandez JM, Senent L induction of major molecular responses and achievement of et al. The t(4;22)(q12;q11) in atypical chronic myeloid leukaemia complete molecular remission in FIP1L1-PDGFRA-positive fuses BCR to PDGFRA. Hum Mol Genet 2002; 11: 1391–1397. chronic eosinophilic leukemia. Blood 2007; 109: 4635–4640. 26 Hughes T, Deininger M, Hochhaus A, Branford S, Radich J, Kaeda J 8 Metzgeroth G, Walz C, Erben P, Schmitt-Graeff A, Haferlach C, et al. Monitoring CML patients responding to treatment with Grimwade D et al. Safety and efficacy of imatinib in chronic tyrosine kinase inhibitors: review and recommendations for eosinophilic leukemia and hypereosinophilic syndromeFa phase- harmonizing current methodology for detecting BCR-ABL trans- II study. Br J Haematol 2008, published online 20 October 2008, cripts and kinase domain and for expressing results. DOI: 10.1111/j.1365-2141.2008.07294.x. Blood 2006; 108: 28–37.

Leukemia Improved detection and monitoring of FIP1L1-PDGFRA J Score et al 339 27 Klion AD, Robyn J, Maric I, Fu W, Schmid L, Lemery S et al. 29 Frost MJ, Ferrao PT, Hughes TP, Ashman LK. Juxtamembrane Relapse following discontinuation of imatinib mesylate therapy for mutant V560GKit is more sensitive to Imatinib (STI571) compared FIP1L1/PDGFRA-positive chronic eosinophilic leukemia: implica- with wild-type c-kit whereas the kinase domain mutant D816VKit tions for optimal dosing. Blood 2007; 110: 3552–3556. is resistant. Mol Cancer Ther 2002; 1: 1115–1124. 28 Heinrich MC, Corless CL, Blanke CD, Demetri GD, Joensuu H, 30 Pardanani A, Elliott M, Reeder T, Li CY, Baxter EJ, Cross NCP Roberts PJ et al. Molecular correlates of imatinib resistance in et al. Imatinib for systemic mast-cell disease. Lancet 2003; 362: gastrointestinal stromal tumors. J Clin Oncol 2006; 24: 4764–4774. 535–536.

Leukemia