METHODS

Patient BUD

The male 18 year old patient BUD was the first born child of dizygotic twins. Delivery was by caesarean section at 37 weeks’ gestation. His birth weight was 2300 g, length 46 cm, and head circumference 32 cm. Neurofibromatosis was suspected shortly after birth, based on multiple café-au-lait spots. There was no family history of NF1.

All developmental milestones were delayed. He walked alone at five years. At age 13.8 years cognitive development was assessed at an eight year level. There were difficulties in gross and fine motor coordination.

At the age of 14, magnetic resonance imaging (MRI) of the brain showed several hyperintense T2 weighted lesions in the left and right lentiform nucleus and internal capsule on both sides. The spinal MRIs showed multiple neurofibromas affecting the nerve roots on both sides in nearly all segments of the vertebral column, often filling the neural foramina. At levels C1/2 to C4/5 and at the lumbosacral junctions, the lesions extended into the spinal canal, forming dumbbell tumours. Intraspinal tumours were present at C2/3 and C3/4. Multiple neurofibromas affecting the lumbosacral plexus were observed on both sides.

Physical examination at the age of 18 revealed 15 café-au-lait spots, skinfold freckling, and more than a thousand cutaneous neurofibromas. Height (162 cm) and head circumference (54 cm) were below the third centile. There was a thoracic scoliosis, genu valgum, deformity of the feet, and hyperextensible joints. A large fleshy nose gave his face a coarse appearance. Hands (19 cm) and feet (European shoe size 38½) were not enlarged.

FISH analysis

Chromosome spreads were prepared from lymphocytes and EBV transformed lymphoblastoid cells from the patient, and from blood lymphocytes from his father, using standard methods. BAC clones were purchased from the BACPAC Resource Center (www.chori.org/bacpac). Cloned polymerase chain reaction (PCR) products and BAC DNA used as fluorescent in situ hybridisation (FISH) probes were labelled with biotin-16-dUTP (Roche Diagnostics, Mannheim, Germany) and detected with FITC-avidin and biotinylated anti-avidin (Vector Inc, Burlingame, California, USA) or labelled with digoxygenin-11-dUTP (Roche Diagnostics) and detected by mouse anti-digoxygenin antibodies, and in a second step with anti-mouse antibody conjugated with Texas-Red (Dianova, Hamburg, Germany). Slides were counterstained with diamidinophenylindole (DAPI) and mounted with Vectashield antifade solution (Vector Inc).

PCR products used as FISH probes to narrow the deletion boundaries

PCR products were amplified from BAC DNA using the Expand Long-Template PCR system (Roche Diagnostics) and the primers listed in table 1. The PCR products were purified by GFX™ DNA purification columns (Amersham Pharmacia, Freiburg, Germany) and cloned using the TOPO-TA cloning kit (Invitrogen, Karlsruhe, Germany). The authenticity of recombinant clones was confirmed by sequencing or restriction enzyme digestion. DNA was prepared from the cloned PCR products with the MIDI DNA isolation kit (Qiagen, address) and labelled by nick translation.

BAC library screening to complete the map of the 17q11.2-12 region

Filters from the RPCI-11 human BAC library and the RPC 1, 3, and 5 human PAC libraries were obtained from the German Resource Center (www.rzpd.de) and the BACPAC Resource Center (RZPD) (www.chori.org/bacpac). Screening of the libraries was done by hybridisation with markers amplified by PCR and labelled by random priming with hexamer oligonucleotides and Klenow polymerase in the presence of ³²P-dCTP (Amersham Pharmacia) in buffer containing 7% SDS, 0.5 M Na phosphate (pH 7.2), and 1 mM EDTA at 65°C overnight. The positions of these markers are indicated in fig 3 and the primer sequences are available from the authors by request. Filters were washed four times at increasing stringency, with a final wash of 0.2×SSC/0.1% SDS for 45 minutes. Following autoradiography for one to two days at −70°C with intensifying screens, positive clones were identified by x/y coordinates and obtained from BACPAC resources or the RZPD.

Genotyping

To determine both the parental origin of the deletion and the recombination mechanism underlying this rearrangement, marker haplotypes were reconstructed and the segregation of marker genotypes was investigated using DNA isolated from peripheral blood lymphocytes of patient BUD and his family members. Ten polymorphic microsatellite markers were analysed, five proximal (D17S841, D17S1873, D17S1841, D17S975, D17S1294) and four distal to the deletion boundaries of patient BUD (D17S907, D17S1833, D17S1788, D17S1867). Marker GGAA7D11 is located within the deleted region. Oligonucleotide primer sequences were obtained from the Genome Database and the respective forward primer was 5′ end-labelled with 6-FAM (Applied Biosystems, Foster City, California, USA). PCR products were analysed by electrophoresis on 5% denaturing polyacrylamide gels on a 396 ABI DNA sequencer using the GENESCAN (version 2.1) software (Applied Biosystems). We phased the parental haplotypes on the basis of the most parsimonious explanation of the observed genotypes.

RESULTS

Characterisation of the proximal deletion breakpoint

In a previous study, we showed by FISH analysis that BAC R-271K11 hybridised to both chromosome 17 homologues in the majority of NF1 deletion patients, but did not produce a second signal on metaphase chromosomes of the patient BUD. The probe R-271K11s, which spans the region of BAC R-271K11 from 141950 to 148526 (according to AC005562), was clearly deleted in patient BUD, but present in the other deletion patients.¹⁰ To locate the proximal end of the deletion in patient BUD precisely, we narrowed down the deletion boundary with additional BACs (fig 1) from the 17q11.2 region. In contrast to the more proximal BAC R-281M11, the BACs R-778K9 and 2349P21 were completely missing on one chromosome 17. We now found a highly reduced hybridisation signal with BAC R-271K11 on the second chromosome, indicating that it was not completely deleted and that some sequences from its insert still hybridised to the affected chromosome 17. Three subfragments from the proximal portion of the BAC insert were used in FISH experiments (table 1). For the PCR probes DJ2384/2385, DJ2386/2387, and DJ2447/2448, signals on both chromosomes were noted (fig 2A), whereas probe DJ2516/2517 was clearly hybridising to only one chromosome 17 homologue (fig 2B). According to the position of the PCR primer pairs on BAC R-271K11, the proximal break occurred between 49221 and 81044 (numbering according to GenBank entry AC005562 of BAC R-271K11; table 1). Thus the proximal breakpoint of BUD lies far outside of the LCR interval (112000 and 147000 on R-271K11) which recurs on BAC R-640N20, on the distal side of the NF1 gene.^9,11,13

Identification of the distal breakpoint

To confine the distal end of the deletion, we assembled a BAC contig for the 3.3 mbp distal to the WI-12393 related LCR on BACs R-640N20 and R-474K4. This contig covers the previously identified clones¹⁸ and BAC R-337L1,¹⁰ and ends up with BAC R-362K1, as shown in fig 3.

Overlapping clones were identified by BLAST searches using completely sequenced BAC clones, BAC end sequence, and clone pools which were screened with known STS markers from the existing radiation hybrid (Standford, Genebridge 4) and genetic linkage maps (Marshfield genetic map). This assembly was confirmed by FISH analyses with several BAC clones from this region on metaphase chromosomes from different deletion patients. The deletion of patient BUD extends up to BAC R-799D4, which is the most distal clone that is completely deleted on the affected chromosome 17. Interphase nuclei of peripheral blood leucocytes were also investigated by FISH with BAC R-799D4 and a differentially labelled BAC R-55A13 (AC015651), which maps outside of the deleted region at 17q23. The interphase nuclei of patient BUD always lacked the second signal for R-799D4 (n = 200), but displayed two signals for R-55A13. By contrast, FISH analysis with BAC R-686D22 resulted in two signals of unequal intensity on chromosome 17 homologues, indicating that this BAC is hitting the distal breakpoint region. BAC R-1094M14, which starts in the middle of R-686D22 and extends beyond its distal half, hybridised equally well to both chromosomes 17. In order to restrict the distal deletion boundary more precisely, we undertook FISH with cloned PCR products DJ2438/2439 and DJ2522/2523 (table 1; fig 3). Both probes are deleted on one chromosome 17 of patient BUD (fig 2C). Subsequently, we generated probes DJ2788/2789 (fig 2D), DJ2725/2726, DJ2781/2780, and DJ2730/2729, which hybridised to both chromosomes of patient BUD and thus delineate the distal border of the deletion. According to these observations, the distal deletion breakpoint is located in the region between the SLFN1 and the SLFN3 genes (fig 3).

Characterisation of the deleted interval in patient BUD and comparison with the mouse syntenic region on chromosome 11

The deletion of patient BUD encompasses 4.7 mbp including the 14 known genes between the proximal and distal NF1 LCR at 17q11.2—CYTOR4, FLJ12735, FLJ22729, CENT2A, MGC13061, NF1, OMG, EVI2B, EVI2A, KIAA1821, MGC11316, HCA66, KIAA0160, and WI-12393—and 27 additional genes located distally to this region, ranging from FLJ11040 to the SCHLAFEN (SLFN) gene cluster (fig 3; table 2).

Comparative analysis of the NF1 gene region in human and mouse undertaken by us previously¹⁴ showed that 12 of the 14 functional genes in the NF1 gene region flanked by LCRs were also physically linked in the mouse genome on chromosome 11, but their order differed significantly.¹⁴ The subsequent region—beginning with the FLJ11040 gene and ending with the SLFN gene cluster at the distal breakpoint region of patient BUD—is, however, strictly conserved with regard to the number and arrangement of genes in both species (table 2). In the region around the distal deletion boundary of patient BUD, the human and the mouse genomes again displayed significant discrepancies. Comparing the SLFN genes that are bordered by the MGC7875 and PEX12 single copy orthologues, we noticed only six human members (SLFN1-6) of the SLFN gene family, but 10 murine genes in the otherwise conserved human–mouse interval (fig 4).

Comparison with other long range deletions of 17q

Through the last decade, three NF1 patients with large deletions at 17q11-12^5,9,15,16 and one patient with a 7 mbp deletion at 17q11.2-21¹⁷ have been reported. The markers and FISH probes used in these studies have been identified by their sequence and placed on our physical map accordingly (fig 5). The proximal deletion boundary of patient UWA106-3 described by Kayes et al^15,16 was localised between D17S1294 and the SLC6A4 gene,¹⁹ the distal deletion breakpoint to the interval between the markers D17S73 and FB6F10 (figs 3 and 4). The proximal breakpoint of patient UWA155-1 described by Dorschner et al⁹ was located within the LCR bordered NF1 interval between markers SHGC35088 and FB12A2, the distal deletion breakpoint between the markers D17S1656 and stSG50857 telomeric to the LCR flanking the 1.4 mbp region.⁹ The proximal breakpoint of the deletion of patient ID806¹⁷ was located in the interval between marker D17S58 and the CRYBA1 (Crystallin β A1) gene. Thus the proximal breaks of the five patients clearly differ in their location. The distal breakpoints of these patients, however, have not been determined precisely. In three patients—UWA106-3, UWA155-1, and patient 3724A⁵—the deletion ends in the very large ACCN1 gene which spans more than 1 mb, including the four markers D17S57, D17S73, D17S115, and FB6F10 (figs 3 and 5). The deletion reported by Upadhyaya et al¹⁷ is cytogenetically visible and is thus the largest. It ends somewhere between D17S73 and the RARA gene. The distal deletion breakpoint of our patient BUD is telomeric to the ACCN1 gene in the region between the SLFN1 and SLFN3 genes. Blast searches with the entire WI-12393 gene sequence did not retrieve additional LCR copies in the region between D17S58 and the SLFN genes.

Mechanism underlying the deletion

We analysed several polymorphic markers on 17q in the parents and two brothers of the index patient. Lack of inheritance of the paternal allele for the GGAA7D11 marker located within the deleted interval clearly indicates the paternal origin of the deletion (fig 6). Studies with markers flanking the deletion interval closely showed that the deletion arose on the paternal chromosome during spermatogenesis by an intrachromosomal mechanism.

To investigate the possibility of somatic mosaicism in BUD’s father, we undertook FISH analysis with a BAC from the deletion interval, R-41C23, which flanks the NF1 gene on the 3′ side.^10,14 On metaphase chromosomes (n = 50) and interphase nuclei (n = 50) of the father, two fluorescent signals were observed, indicating that the chromosomal deletion most probably arose during germ cell maturation.

DISCUSSION

Previous definition of the 1.4 mbp deletion interval shared by most patients with NF1 microdeletion led to the identification and localisation of new LCRs flanking the breakpoint regions. The patient described here belongs to a small subgroup of four patients whose deletion intervals are not bordered by these LCRs. Their telomeric or centromeric breakpoints fall far beyond these boundaries and have not so far been located at the molecular sequence level.

The deletion of the patient BUD described here is more far reaching in telomeric direction and thus we extended our clone and sequence map up to the distal breakpoint. Major portions of this region have been sequenced and assembled as HTGS data by the Human Genome Consortium towards the end our studies and were partially reported recently.²¹ The existence of low copy repeats on BAC 640N20 has prevented an automatic assembly of the adjoining telomeric region. Furthermore, some of the genes between the LCR on BAC R-640N20 and the distal breakpoint of patient BUD were not fully annotated and reviewed. We confirmed and fully established the map by FISH experiments with additional BAC and PAC clones that were initially screened to cover the deleted region fully.

Comparative analyses of the human and murine genome sequences enabled us to identify all functional genes that are co-deleted in our patient BUD and may contribute to the clinical phenotype (table 2; fig 3). Thus our patient is not only hemizygous for the 14 genes between the proximal and distal LCRs at 17q11.2, inclusive of the NF1 gene, but also for the 27 genes following distally.

The most conspicuous clinical finding in patient BUD is the large number of bilateral spinal neurofibromas. Although spinal neurofibromas develop as isolated asymptomatic tumours in up to 36% of NF1 patients,^22–24 multiple symmetrically distributed spinal tumours with clinical implications are very rare in classical NF1 patients. Familial spinal neurofibromatosis (FSNF, MIM 162210) is a rare distinct form of NF1 which is characterised by multiple symmetrical spinal nerve root neurofibromas and café au lait spots in all affected family members.²⁵ FSNF is associated with NF1 mutations, but is not caused by specific mutations in the NF1 gene.^26–28 It has been suggested that a gene closely linked to the NF1 locus is modifying the NF1 phenotype in FSNF.²⁸ Most probably, the analyses of large deletion patients with or without multiple spinal neurofibromas will help to identify this modifying locus. Unfortunately, a search for spinal neurofibromas by MRI has not been done in the patients ID806 and 3724A. Patient UWA155-1 was reported to have one spinal neurofibroma, but it is unknown whether he carries multiple asymptomatic spinal tumours.⁹ However, patient BUD, described here, and patient UWA106-3,⁹ who both have large deletions, developed numerous spinal tumours.

As the number of spinal neurofibromas has not been investigated and quantitated systematically by MRI in patients with LCR mediated NF1 microdeletion, we cannot really exclude the genes of the 1.4 mbp interval, but in view of the clinical data reported so far, we suggest that one of the genes distal to the 1.4 mbp interval is implicated in the development of spinal neurofibromas. The most suspicious region for such a locus is the segment between the distal NF1 LCR and the marker FB6F10, which is deleted in both patients and encompasses nine functional genes. Hemizygosity with reduced activity of this modifier gene might predispose patients BUD and UWA106-3 to the development of multiple spinal neurofibromas. Similarly, a mutation in the modifier locus or a hypomorphic allele of the latter could contribute to the phenotype in patients with FSNF.

Comparing the breakpoint regions in patient BUD at the sequence levels, we did not encounter blocks of low copy repeats around the proximal and distal breakpoints, which could have triggered chromosomal mispairing and unequal homologous recombination during meiosis, as has been observed in patients with 1.4 mbp deletions.^11,13 The proximal breakpoint of BUD is located between position 49221 and 81044 on BAC R-271K11 (AC005562), about 65 kbp centromeric to the proximal LCR of the NF1 gene region, which triggers the common 1.4 mbp deletion by recombination with the distal LCR.

A third LCR, which is similar to the regional LCRs at 17q11.2, has been mapped to 17q24.⁹ This distal LCR also contains fragments of the WI-12393 gene. Our mapping of the deletion breakpoint in patient BUD rules out the involvement of another WI-12393 derived LCR, including the one at 17q24.

The comparisons of the long range deletions at 17q in five different patients including our patient BUD^{5,9,10,15–17} strongly suggest that all deletion breakpoints in these patients are unique and do not coincide with pairs of LCRs situated at the proximal and distal boundaries.

The fact that the distal breakpoints in three patients (UWA106-3, 155-1, and patient 3724A^5,9,15) occur within the 1 mbp spanning the ACCN1 gene suggests that this region is accessible and transcribed during germ cell development. The distal deletion breakpoint of patient BUD, however, maps far away from the ACCN1 gene to the region of the SLFN genes. In the region around the distal deletion boundary of patient BUD, the human and the mouse genomes again displayed significant discrepancies. Comparing the SLFN genes, we noticed only six human members (SLFN1-6) of the SLFN gene family in the human genome, but 10 murine genes in the otherwise conserved human–mouse interval (fig 4). These findings indicate that the SLFN gene region has been reshaped by multiple breaks and subsequent rearrangements during the evolution of rodents and humans, and that the deletion of patient BUD occurred in this region of recent evolutionary rearrangements.

The proximal deletion breakpoint of patient BUD falls within a 200 kbp large gene-free segment represented by BAC R-271K11. This BAC spans various blocks of low copy repeats—a partial SMS-REP,²⁹ a SMURF2 pseudogene fragment, the known WI-12393 related LCR, sequences from chromosome 19—and is followed by the KIAA0160 pseudogene fragment covered by the BAC R-778K9. These blocks originate from different genomic locations and have been inserted successively into this region during primate evolution.¹⁴ We therefore suggest that the proximal breakpoint region again carries intrinsic features which facilitate double stranded breaks and rearrangements.

Approximately 80% of small intragenic mutations are paternally inherited,^30–32 whereas meiotically generated microdeletions of the NF1 gene region are predominantly of maternal origin.^6,8,9,32,33 In 17 of 19 informative cases (89%) analysed by López-Correa et al,¹³ the deletions occurred on the maternal allele. By contrast, the deletions of patient UWA106-3,^15,16 patient ID806,¹⁷ patient 96-2,^12,33 and patient PF¹² that do not end at the proximal and distal NF1 LCRs were of paternal origin, with one exception (patient 3724A⁵).

To gain insight into the mechanism of recombination, six families containing patients with de novo deletions were previously analysed by haplotyping.³³ All five deletions on the maternal chromosome occurred by crossovers between non-sister chromatids, whereas the paternal chromosome acquired a de novo deletion by intrachromosomal (sister chromatid) breaks. Interestingly, this latter deletion (patient 96-2) is not bordered by LCRs and is probably restricted to the NF1 gene. Likewise, the deletion of patient BUD arose by an intrachromosomal mechanism during male germ cell development. Length variation of deletions across the NF1 locus were also observed in neurofibromas of NF1 patients which lost the normal NF1 allele by a mitotically acquired interstitial deletion of the normal 17-homologue.³⁴ As shown previously, even interstitial deletions between LCRs can occur intrachromosomally during mitosis in somatic cells, probably driven by the recombination between misaligned LCRs in sister chromatids.^35,36 In view of these observations it seems quite reasonable to assume that non-LCR-driven interstitial deletions are primarily generated during mitotic cell division in particular in the course of spermatogenesis.