FISH mapping of Philadelphia negative BCR/ABL1 positive CML

Background Chronic myeloid leukaemia (CML) is a haematopoietic stem cell disorder, almost always characterized by the presence of the Philadelphia chromosome (Ph), usually due to t(9;22)(q34;q11) or its variants. The Ph results in the formation of the BCR/ABL1 fusion gene, which is a constitutively activated tyrosine kinase. Around 1% of CML patients appear to have a Ph negative karyotype but carry a cryptic BCR/ABL1 fusion that can be located by fluorescence in situ hybridisation (FISH) at chromosome 22q11, 9q34 or a third chromosome. Here we present FISH mapping data of BCR and ABL1 flanking regions and associated chromosomal rearrangements in 9 Ph negative BCR/ABL1 positive CML patients plus the cell line CML-T1. Results BCR/ABL1 was located at 9q34 in 3 patients, 22q11 in 5 patients and CML-T1 and 22p11 in 1 patient. In 3 of 6 cases with the fusion at 22q11 a distal breakpoint cluster was found within a 280 Kb region containing the RAPGEF1 gene, while in another patient and the CML-T1 the distal breakpoint fell within a single BAC clone containing the 3' RXRA gene. Two cases had a duplication of the masked Ph while genomic deletions of the flanking regions were identified in 3 cases. Even more complex rearrangements were found in 3 further cases. Conclusion BCR/ABL1 formation resulted from a direct insertion (one step mechanism) in 6 patients and CML-T1, while in 3 patients the fusion gene originated from a sequence of rearrangements (multiple steps). The presence of different rearrangements of both 9q34 and 22q11 regions highlights the genetic heterogeneity of this subgroup of CML. Future studies should be performed to confirm the presence of true breakpoint hot spots and assess their implications in Ph negative BCR/ABL1 positive CML.

variant translocation [1]. In ~1% of the CML patients the bone marrow cells appear to be Ph negative by G-banding, although the BCR/ABL1 fusion gene can be identified by molecular means and located by fluorescence in situ hybridisation (FISH) on chromosome 22q11, 9q34 or even a third chromosome. The biology and clinical significance of the genetic rearrangements in Ph negative BCR/ ABL1 positive disease have been widely discussed following the first descriptions [2][3][4][5]. Two mechanisms for the formation of the chimeric gene in masked Ph positive cells have been postulated: either by insertion of ABL1 into the BCR region (or vice versa) or by a multiple step model where a classical t(9;22) is followed by a translocation of both products and/or another autosome, thereby restoring the normal chromosome morphology. In both instances, more than the 2 breaks associated with classical t(9;22) are implicated. Although as early as 1990 Morris et al. [6] provided evidence that the insertion involves additional sequences distal to the 3' ABL1 site, the extent of the genomic rearrangements in this form of CML are unknown.
In view of the few studies published with a precise map of the insertions [7][8][9][10], we aimed to construct an accurate map of the insertions in the cell line CML-T1 [11] and 9 patients with Ph negative BCR/ABL1 positive CML using a range of FISH probes obtained from BAC clones. The fusion gene was identified at 9q34 (3 patients), 22q11 (5 patients and CML-T1) and 22p11 (3 patients), resulting in relocation of sequences well in excess of either 3' ABL or 5' BCR by means of a direct insertion (6 patients and CML-T1) or a sequence of events (3 patients). Recurrent distal breakpoints were found at the regions of RAPGEF1 and RXRA genes.

Methods
Nine archival bone marrow chromosome preparation samples of CML patients (7 females and 2 males) with Ph negative BCR/ABL1 positive disease collected from Ham-mersmith and Royal Free Hospitals (London, UK), together with the cell line CML-T1, were investigated ( Table 1). All samples tested positive for BCR/ABL1 fusion by PCR. Investigations were carried out on bone marrow samples obtained at presentation. G-banding and molecular cytogenetic analysis, including chromosome painting and FISH mapping with locus specific probes, were performed following protocols in routine use [12]. A minimum of 25 metaphase and over 100 interphase cells after short term in vitro culturing were analysed and results described following ISCN (2005). Five of the samples in this cohort (cases 1-3, 7 & 8) were part of another study [13].
In all samples, FISH with the commercially available LSI BCR/ABL1 Dual Color, Dual Fusion Translocation Probe ("D-FISH", Vysis, Downers Grove, IL, USA) was initially performed using manufacturer's protocol to identify the chromosome location of the BCR/ABL1 fusion gene. FISH mapping was carried out with Bacterial Artificial Chromosomes (BAC) clones obtained from the BACPAC Resources Center (Children's Hospital Oakland Research Institute, Oakland, CA, USA), the Sanger Centre (Cambridge, UK) and Invitrogen (Paisley, UK) ( Table 2). Clones were grown in Luria-Bertani medium with Chloramphenicol (20 μg/ml), BAC DNA extracted with a QIA-GEN Large-Construct Kit (Qiagen, West Sussex, UK) and directly labelled with either Spectrum Orange or Spectrum Green dUTPs with a Nick Translation Kit (Vysis). BAC DNA from chromosomes 9 and 22 belonging to the Human 32 K Clone Set from the BACPAC Resources Center was amplified with a GenomePlex Single Cell Whole Genome Amplification Kit (Sigma-Aldrich, Dorset, UK) and labelled as described before.
BAC RP11-164N13 was used to target BCR gene. Since this BAC covers both major and minor BCR breakpoints, it is found split when BCR gene is rearranged. RP11-61N10 covers BCR non-coding sequences centromeric to the breakpoint and it was used to identify 5' BCR region. RP11-83J21, which covers the 3' end of ABL1 incorporating the whole of the coding region, was used to identify the sequences telomeric of the ABL1 breakpoint, whereas RP11-57C19 was used to identify the ABL1 sequences centromeric of the breakpoint.
The BAC clones and genes were located according to the UCSC database, hg17 (University of California Santa Cruz, CA, USA) [14]. Mapping data for the 32 K human clone set was obtained from the BACPAC Resources Center website [15] and used to assess the size of the sequences found to be rearranged. In addition, sub-telomeric probes directly labelled from the regions of 9q, 22q and 16q were used (Stretton Scientific Ltd, Stretton, UK).
All tests were carried out as dual colour, dual probe FISH. Digital imaging and karyotyping were carried out using a SmartCapture and SmartType FISH workstation (Digital Scientific Ltd, Cambridge, UK).
Array CGH analysis (aCGH) was carried out on the cell line CML-T1. The aCGH was performed using two platforms -1 Mb BAC clone chip (SGI2600) [16] and oligonucleotide (105 K Agilent) [17] following manifacturer's protocol, while data processing and presentations were carried out using 'in house' software as reported [18,19].

Results
A summary of the molecular cytogenetic investigations carried out on BCR/ABL1 positive samples from 9 CML patients with normal bone marrow (BM) karyotype as well as the cell line CML-T1 with masked Ph chromosome is presented Figure 1.

FISH analysis using commercial BCR/ABL1 D-FISH probe (Vysis)
The BCR/ABL1 D-FISH probe (Vysis) contains sequences covering both genes labelled in different colours, so that rearrangements affecting them can be visualised at chromosome level. The ABL1 probe (red) targets a 650 Kb region of 9q34.1 including the whole of the ABL1 gene (173.8 Kb), thus spanning the common breakpoint, and extends 5' of ABL1 to incorporate the ASS gene. The BCR probe (green) is represented by two 600 Kb regions of 22q11.2 separated by a 300 Kb gap, one of the regions covering the entire BCR gene and extending 5' of it in order to include the IGLV gene, and the other starting 300 Kb telomeric of BCR and ending at 900 Kb 3' of the gene. The application of this probe revealed the BCR/ABL1 fusion at three different chromosome sites: at 22q11.2 (samples no [1][2][3][4][5][6], at 22p11 (sample no 7) and at 9q34 (samples no 8-10). There was evidence for the formation of the reciprocal ABL1/BCR fusion in just one patient (no 4), which showed 2 fusion signals at der (22) and der (9). In the remaining cases only 1 fusion signal was observed, irrespective of the fusion gene location.

Patients with BCR/ABL1 fusion residing on chromosome 22q11.2
The BAC clone RP11-83J21 covering 3' ABL1 moved to 22q11 in 6 cases (no 1-6), whereas RP11-57C19 remained at der(9) in all but one (no 1, Figure 1). In patient no 1, we identified a cryptic genome loss at der (9)  clones RP11-57C19 to RP11-17O4. However, it was not possible to assess the full extend of the deletion due to lack of material.
In all 6 cases the inserted 9q34 sequences exceeded the 3' boundaries of the ABL1 gene and estimated to be 720 Kb long in 1 case (no 1), 1 Mb in 2 cases (no 2-3), between 1.6 Mb and 2.2 Mb in 1 case (no 6) and 3.9 Mb in 2 cases (no [4][5]. The estimated sizes of the insertions were calculated based on the information of the BAC clones chromosome location available at the UCSC database (genome build 35). In cases no 1-5 the insertions were found to stretch distally falling into two sub-groups: a "small" insertion (720 Kb-1 Mb) and a "large" insertion (3.9 Mb) ( Figure 2). The distal boundary of the "small" insertion seen in 3 patients (no 1-3) fell within a region covered by 3 overlapping BAC clones (CTD-2107G12, RP11-40A7 and RP11-323H21), thus forming a breakpoint cluster ( Figure 2) which was estimated to be 280 Kb long and found to house several genes: POMT1, UCK1 and RAPGEF1. In further two samples (patient no 4 and CML-T1) the inserted material was found to be larger (estimated size of 3.9 Mb) and the distal breakpoint fell within a single BAC clone (RP11-92B21) containing the 3' end of the RXRA gene (Figure 2).
High-resolution aCGH analysis of the cell line CML-T1 identified a gain at 9q34.1 starting at ABL1 breakpoint at 130.6 and covering 3.9Mb in 3' direction until the distal part of the RXRA gene at 134.5, confirming the FISH mapping data ( Figure 3). This gain resulted from the duplication of the der (22) , showing the duplication of the masked Ph and that the genomic loss affects not the der(9) but the "normal" homologue (the 2 green and 2 red signals from the two normal 9 are missing). The top box on the right shows 4 green, 2 red signals as seen in interphase tetraploid cells with deletion, while the bottom box on the right shows 6 green, 4 red signals as seen in the interphase tetraploid cells without deletion.
had previously revealed a deletion of 5' ABL1 in a tetraploid sub-clone. We mapped the length of the deletion, which was found to be 8.7 Mb long (from RP11-138E2 to 9q telomere) and affect not the der(9)ins(22;9) but the "normal" 9. This was demonstrated by FISH when cohybridizing a BAC clone found within the insert (in green) and a BAC from chromosome 9 centromeric or telomeric to the inserted sequences (in red). Tetraploid cells with a deletion displayed 2 red signals from two chromosome 9 and 4 green signals from the 4 masked Ph chromosomes. Therefore not only two red signals were deleted, but also the 2 green signals from the "normal" 9 were missing, demonstrating that the loss had occurred at the "normal" homologues.
Regarding BCR flanking regions, all probes tested from chromosome 22 remained at their original locations in cases no 1, 2, 3 and 5. However, in case no 4, sequences distal to the BCR breakpoint and housed by the BAC clones RP11-164N13, RP11-529P21, RP11-223O10 and RP11-698L6 were found embedded within the der(9) chromosome at the 9q34 region ( Figure 4). Both RP11-164N13 and RP11-529P21 gave a split signal pattern being found at the masked Ph and derivative 9, apart from the normal 22. The distal breakpoint of the 22 sequences found at der(9) fell between RP11-698L6 and RP11-594L10.
Even more complex rearrangements involving a third chromosome were revealed in another patient (no 6) (Figure 5). FISH painting and FISH mapping identified a three BCR/ABL1 at 22q11.2 in patient no 4 results from a multiple step mechanism  (22) and 22q11.2 sequences inserted on der(9) (black arrows) can be explained by two consecutive translocations: an initial t(9;22)(q34;q11.2) followed by a second reciprocal translocation between the two products, with breakpoints distal to both BCR/ABL1 and ABL1/BCR fusion genes.

Patient with the BCR/ABL1 fusion residing at 22p11
The BCR/ABL1 fusion was unexpectedly found at 22p11 in 1 patient (no 7) ( Figure 6). FISH with RP11-164N13 showed one signal at the normal chromosome 22 and a split signal at the 22p11 and 22q11 regions from the other homologue. RP11-61N10 was found at the normal 22 and only at the 22p11 region of der (22), thus confirming the location of BCR/ABL1 fusion at the p arm of the derivative 22. RP11-529P21, RP11-223O10 and 22qter remained on 22q11, while all probes tested from RP11-83J21 to 9qter were found at 22p11.

Patients with the BCR/ABL1 fusion residing at 9q34.1
The BCR/ABL1 fusion was located by FISH at band 9q34.1 in 3 patients (no 8-10) and thought to result from a direct insertion of 22q11.2 sequences. As expected, the telomeric breakpoint was found in all cases within the BAC clone RP11-164N13, which was always seen at both der(9) and der (22) (Figure 7). RP11-61N10 was always found at 9q34, however, due to lack of material it was not BCR/ABL1 resides at 22q11.2 as a result of a cryptic three-way rearrangement between chromosomes 9, 22 and 16 in patient no 6 Sequences distal to BCR breakpoint are found at their original location while 5' BCR and 9q34 sequences distal to ABL1 breakpoint (including the telomere) are relocated at 22p11. This could be explained by an initial t(9;22)(q34;q11) followed by a three-way translocation between 9q34, 22q11 and 22p11, which would require 5 breaks (red arrowheads). (f) The unusual location of BCR/ABL1 at 22p11 could also be a result of an initial ins(9;22)(q34;q11q11) followed by a translocation between der(9) and the p arm of der (22). This sequence of events would need also the same amount of breaks (red arrowheads) and therefore cannot be rule out.
possible to assess the proximal breakpoint of the 22q11 insertion. All BACs tested from 9q34 remained at their respective location on chromosome 9 in all 3 patients.
In patient no 8, the insertion was accompanied in all cells by loss of a region at least 1.77 Mb long from 22q11.2 and immediately distal to the BCR breakpoint, covered from BCR/ABL1 fusion resides at 9q34.1 in 3 patients  (9), plus 1 green signal at der (22), as seen in patients no 9 and 10. RP11-164N13 is therefore split, with the 5' BCR sequences relocated at der (9). (c) BCR/ABL1 D-FISH probe (Vysis) in patient no 8 showing 1 fusion signal at der(9), 1 red signal at 9 and 1 green signal at 22. The green signal from der(22) is deleted. (d) A representative metaphase cell with co-hybridization of FISH probes RP11-61N10 and RP11-594L10 giving a fusion signal at normal 22 and one red signal from RP11-61N10 at der(9), confirming that the BCR sequences moved to der(9) are centromeric of the breakpoint. There is again one green signal missing, because the deletion at der (22) includes not only 3' BCR but also at least 1.77 Mb distal to the breakpoint. the 3' end of RP11-164N13 to at least RP11-765G14 (Figure 1).

Discussion
Since the first description of a Ph negative BCR/ABL1 positive CML patient [2], several studies have been published reporting similar cases. Most of them are focused on the presence and location of the BCR/ABL1 fusion in CML patients with masked Ph chromosome and commonly achieved by application of commercial FISH probes, which have been proved to be very useful to identify the presence and location of the BCR/ABL1 fusion gene in CML patients with no distinguishable Ph chromosome. These studies have established the importance of FISH tests for the diagnosis and therapy monitoring of Ph negative BCR/ABL1 positive CML. However, the commercial probes don't provide enough information to understand the mechanisms involved in the formation of the masked Ph chromosome.
We used FISH mapping with BAC probes in order to study the formation of the BCR/ABL1 fusion and the underlying genomic rearrangements in 9 CML patients with Ph negative BCR/ABL1 positive CML and the cell line CML-T1. The formation of the fusion gene resulted from the relocation of not only the 3' ABL1 sequences within the BCR region at chromosome 22q11.2 or 5' BCR sequences within ABL1 region at 9q34.1, but also a considerable amount of flanking material, leaving the chromosome morphology apparently intact.
The fusion gene was located at 9q34.1 in 3 patients, at 22q11.2 in 5 patients and CML-T1, and at 22p11 in another patient. 5 out of the 6 cases with a 9q34 insertion at 22q11 displayed recurrent distal breakpoints that fell within two gene bearer regions. Thus, in 3 patients a common breakpoint cluster of 280 Kb was found. According to UCSC database, this region houses 3 genes: POMT1 (protein-O-mannosyltransferase1), UCK1 (uridine-cytidine kinase 1) and RAPGEF1 (guanine nucleotide releasing factor for RAP1; also known as C3G). In another patient and CML-T1 the breakpoint fell within a single BAC clone encompassing the 3' end of RXRA gene (retinoid × receptor alpha). Although further investigations were not carried out in this study, both RAPGEF1 and RXRA belong to pathways the disruption of which may be relevant to the evolution of the patients. RAPGEF1 has been shown to have transformation suppressor activity towards several oncogenes [20,21] and also interact with BCR/ABL1 [22]. RXRA belongs to a family of nuclear receptors that target multiple signalling pathways [23] and its downregulation has been showed to be critical for neutrophil granulocytes differentiation [24]. Other studies found RXRA to contribute in acute promyelocytic leukaemia transformation [25,26]. Although we have no direct evidence for the immediate involvement of either RAPGEF1 or RXRA genes, their potential role merits further investigation.
Two other mapping studies that found similar insertions have been published [7,8]. In the first study, the authors used FISH mapping to identify the rearrangements involved in 2 Ph negative CML patients with variant translocations. A 3 Mb insertion from 22q11 into ABL1 was identified in 1 patient, while the other had a 9q34 insertion at the BCR region with a distal breakpoint falling within the clone RP11-353C22 (genome address: 31,278,588,248). This result matches with our findings since the latter BAC is located within the same 280 Kb common breakpoint region identified in 3 patients of our study. On the other hand, Valle et al. [7] found a 5.6 Mb insertion of 9q34 sequences into BCR. This insertion is larger than the ones identified in our study and was not accompanied by any deletions or other rearrangements.
Deletions of 5' ABL1 and/or 3' BCR sequences at the der(9) chromosome in patients with classical and variant Ph translocations [27] have been shown to have adverse prognostic value in CML patients treated with interferon [28], although their impact in patients being treated with tyrosine kinase inhibitors is controversial [29][30][31]. Dual colour, dual fusion translocation FISH probes spanning the BCR and ABL1 genes are very useful for revealing these events but they have a limited value in interphase nuclei in patients with masked Ph, since often the merging of the 5' BCR and 3' ABL1 signals by simple insertion leads to an apparent loss of one fusion signal that can be falsely assessed as deletion [13]. Thus, D-FISH (Vysis) in a patient with a direct ins(22;9) gives a 2 red, 1 green, 1 fusion signal pattern, which is the same pattern obtained in case of a typical t(9;22) with deletion of 5' ABL1 at der (9). If the patient has a direct ins(9;22) the D-FISH signal pattern is 1 red, 2 green, 1 fusion, which could be mistaken for a typical t(9;22) with deletion of 3' BCR at der (9). Furthermore, Ph negative BCR/ABL1 positive patients with either a cryptic deletion of 5' ABL1 or 3' BCR show a 1 red, 1 green, 1 fusion D-FISH pattern, also typical for a t(9;22) with deletion of the reciprocal ABL1/BCR fusion. Therefore, when a deletion signal pattern is detected by interphase FISH with a dual colour, dual fusion probe, it is essential to look also at the metaphases in order to be able to differentiate a classical t(9;22) with deletion from a simple insertion.
Batista et al [32] reported a Ph negative BCR/ABL1 positive patient with an insertion of ABL1 into BCR associated with a deletion of the ASS -5' ABL1 region. Zagaria et al [8] also reported two cases of CML patients with masked Ph and associated deletions: one patient with a cryptic ins(9;22) and a deletion of 5' ABL1 and 3' BCR regions; the other patient with a multi-step variant translocation and a deletion of around 400 Kb telomeric of the ABL1 gene. In addition to them, De Melo et al [13] identified with commercial triple-colour FISH probes 5' ABL1 deletions in 2 patients and 3' BCR deletion in 1 patient out of 14 CML patients with masked Ph. Our study confirmed and sized such deletions in 2 patients which where also part of De Melo's cohort. CML-T1 also had a 8.7 Mb deletion of 9q34 material in one of the sub-clones, but in this case the loss was found to affect the homologue not involved in the BCR/ABL1 formation.
We identified a duplication of the chromosome 22 harboring the BCR/ABL1 fusion accompanied by loss of the normal homologue in 1 out of 9 patients in this study plus the cell line CML-T1. Such duplications of the BCR/ ABL1 bearing chromosome (either chromosome 22 or 9) seem to be a relatively common event in Ph negative BCR/ ABL1 positive CML patients, being accompanied by loss of the normal homologue in most of the cases and seen both in chronic phase and blast crisis [33][34][35].
Regarding the formation of the fusion gene in Ph negative BCR/ABL1 positive CML patients, Morris et al [6] suggested two possible mechanisms: a one step model, where BCR/ABL1 results from a simple insertion of either 3' ABL1 into BCR or 5' BCR into ABL after three genomic breaks; and a multiple step model, with an initial classical t(9;22)(q34;q11) followed by a second translocation of both products and/or a third chromosome, requiring a minimum of 4 genomic breaks.
Our study provided evidence for the existence of both mechanisms implicated in the formation of the fusion gene in Ph negative patients. We found a simple insertion (one step event) in 6 out of 9 patients (no 1-3, 8-10) and CML-T1 (no 5), with no evidence of secondary rearrangements within the regions flanking BCR and ABL1 breakpoints apart from the accompanying deletions seen in 2 patients and CML-T1. On the other hand, traces of sequential rearrangements indicating a multiple step mechanism were found in 3 patients. Patient no 4 had a 9q34 insertion at chromosome 22 with bits from 22q11 distal to the breakpoint embedded within the der(9), suggesting an initial t(9;22)(q34;q11) followed by further translocation of both products. Patient no 7 carried the BCR/ABL1 fusion at 22p11. Sequences downstream of the breakpoint remained at their original location, while 5' BCR and all 9q34 sequences distal to ABL1 breakpoint (including the telomere) were relocated at 22p11, which could be explained by an initial t(9;22)(q34;q11) followed by a three-way translocation between 9q34, 22q11 and 22p11. This sequence of events would have required 5 breaks (2 for the translocation and 3 for the second one). However, an initial ins(9;22) followed by a reciprocal translocation between 9q34 and 22p11 would also require 5 breaks and therefore cannot be ruled out. Finally, patient no 6 had a three way cryptic t(9;22;16)(q34;q11;q?13) with 9q34 sequences embedded within the der (22), suggesting an initial ins(22;9) followed by a translocation between chromosomes 16 and der (22)ins (22;9). These data show not only that the two mechanisms do happen, but also that they are not excluding options. An example of the latter is patient no 6, where an initial direct ins(22;9) would be part of the spectrum of rearrangements that had restored the normal morphology of the der (22).

Conclusion
In summary, we found that the BCR/ABL1 fusion resulted from relocation of not only the 3' ABL1 sequences within BCR at 22q11.2 or 5' BCR sequences within ABL but also a considerable amount of flanking material, with distal recurrent breakpoints of the excised 3' ABL1 sequences at RAPGEF1 and RXRA regions. BCR/ABL1 resulted from a direct insertion (one step mechanism) in 6 patients and CML-T1, while in 3 patients the fusion gene was a result of a sequence of events (multiple steps). Finally, the presence of different rearrangements of both 9q34 and 22q11 regions demonstrates the genetic heterogeneity of this subgroup of CML. Future studies should be performed to confirm the presence of true breakpoint hot spots and assess their implications in Ph negative BCR/ABL1 positive CML.