MATERIAL AND METHODS

Patients

The 60 year old female investigated in this study (II.2, fig 1) has four café au lait spots on her thighs and forearms. Some minor freckling was found in the left axilla, but she did not have dermal neurofibromas or Lisch nodules. Her two sons, however, suffer from a severe form of NF1 (patients III.1 and III.2). In early childhood, they had multiple café au lait spots distributed all over the body and noticeably large hands and feet. In the younger brother, retarded psychomotor development was documented. Later in life, coarse facial features with hypertelorism were noticed, in addition to macrocephaly, axillary freckling, mild scoliosis, and multiple neurofibromas, which were too numerous to count. Multiple melanocytic naevi and low set nipples were additional features. The intelligence quotient of the sons was judged by their mother as normal, but was specifically tested only in her oldest son and was found to be in the normal range. Both attended regular school. Patient III.1 suffered from a malignant glandular schwannoma in the left gluteus which metastasised. At the age of 29 years, he died from intracranial bleeding before genetic investigations were started. His younger brother, patient III.2, who is still alive, had sphenoid wing dysplasia. At the age of 11 years, progressive dislocation of his hips and genu valgum were diagnosed, which had to be surgically corrected several times.

FISH analysis

Chromosome spreads were prepared from peripheral blood lymphocytes and skin fibroblasts of proband II.2 and blood lymphocytes of her son (the index patient III.2) according to standard methods. Skin fibroblasts of the mother were obtained by a punch biopsy of the left axillary region. BAC clones used as FISH probes were purchased from the BAC/PAC Resource Center (www.chori.org/bacpac) and have previously been characterised.^10,11,15 PCR products used as FISH probes, DJ1686/1863, DJHK10/11, and DJ1576/1578, were amplified with primers listed in table 1 using the Expand Long Template PCR System (Roche Molecular Biochemicals). The respective products were cloned with the TOPO TA Cloning System (Invitrogen) and labelled for FISH analysis. Two colour FISH was performed using BAC-DNA or the cloned PCR products labelled either with biotin-16-dUTP or digoxygenin-11-dUTP (Roche-Diagnostics, Mannheim, Germany). Hybridisation signals were visualised by FITC-avidin and biotinylated anti-avidin (Vector, Burlingame, USA) or with anti-digoxygenin antibodies and in a second step with anti-mouse antibodies conjugated to Texas-Red (Dianova, Hamburg, Germany). Slides were counterstained with diamidinophenylindole (DAPI) and mounted with Vectashield antifade solution (Vector, Burlingame, USA).

Genotyping and generation of hybrid cell lines

Markers analysed were D17S1873, D17S1841, D17S975, D17S1863, and D17S635 proximal to the NF1 gene and markers D17S1800, D17S1880, D17S907, D17S1833, D17S1788, and D17S1867 distal to the NF1 gene. Markers D17S1849 and D17S1166 within the NF1 gene were also investigated. Genotyping was performed with 6FAM labelled primers and capillary electrophoresis on an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems). For PCR experiments, we used DNA isolated from blood lymphocytes and hybrid cell lines carrying the deleted chromosome of patient III.2 (hybrid line Nos 4 and 15) and from hybrids with the normal chromosome 17 of III.2 (Nos 8 and 17) to identify the haplotypes. The hybrid cell lines were generated by PEG mediated fusion of a mouse cell line with peripheral blood lymphocytes of patient III.2 (GMP Genetics, Suffolk, UK). Hybrid cell lines were genotyped with the standard set of markers routinely used by GMP Genetics. Marker analysis in all other members of the family was performed using genomic DNA from peripheral blood.

Analysis of markers within the genomic interval between the KIAA0160 gene and its pseudogene

Dinucleotide repeat marker IVS27AC28.4 in intron 27b of the NF1 gene was amplified with Cy5 labelled PCR primers as described by Lazaro et al¹⁶ and analysed on an ALF-Express Sequencer (Amersham Pharmacia). The single nucleotide polymorphism SNP2 located in the KIAA1821 gene was amplified with primers described by Eisenbarth et al.¹⁷ Resulting PCR products were digested with the restriction enzyme MboI and analysed by agarose gel electrophoresis.

Fine mapping of the deletion breakpoints

The region of the deleted interval and the adjacent segments were analysed by PCR using polymorphic and non-polymorphic STS markers and DNA from a hybrid cell line carrying only the deleted chromosome 17 of patient III.2. PCR products flanking the deletion boundaries, DJ1911/1910, DJ1561/1562, and DJ1948/1936, were amplified with primers DJ1911 (5′ CTGCACATGGCATTGTTTGATT 3′) and DJ1910 (5′ AAACCCTCCTTGTGCCCTACTC 3′), DJ1561 (5′ GTCCAATAAG CAAGAAAAGAGCAA 3′) and DJ1562 (5′ TGGGGAAGAAAACATCAGTCAA 3′), DJ1948 (5′ GAGAAAATGAAAGGAGAGCAAGAA 3′) and DJ1936 (5′ TCTTAGTGCCTCTGGGAGCAA 3′). PCR products were directly sequenced by cycle sequencing on an ABI 377 sequencer (Applied Biosystems) and analysed using the GCG software (Genetics Computer Group) to determine their origin and to evaluate whether sequences are derived from one or two distinct loci.

RESULTS

Deletion detection by haplotype analysis and FISH

Analysis of polymorphic markers showed a large deletion of the NF1 gene region in the index patient III.2, as hemizygosity for markers D17S635, D17S1849, D17S1166, D17S1800 was observed in the DNA of peripheral blood lymphocytes (fig 1). To confirm these findings, FISH was performed with BAC 142O6 (AC079915), which covers the proximal region of the NF1 gene.¹⁵ A deletion of the corresponding region on one chromosome 17 was observed in all blood lymphocytes of patient III.2 (n=50). Upon this finding, the mother (patient II.2) of the index patient was also investigated by FISH. In her blood lymphocytes, the deletion was detected in 70% of the metaphases analysed (n=50), whereas in fibroblast cultures, the deletion of BAC 142O6 was noted only in 15% of all metaphases (n=20) (data not shown).

Delineation of the deletion boundaries by FISH

To determine the extent of the deletion, we performed FISH with further BACs that have previously been mapped to the NF1 gene region (figs 2 and 3).^10–15 On metaphase chromosome spreads of the index patient III.2 and his mother, BACs 785C15, 3047G17, 499I20, 805L22, and 542B22 are deleted on one chromosome 17. Using BACs 2349P21 and 307A16 as FISH probe (fig 3A, B), a reduced signal was observed on one chromosome 17 homologue suggesting that these BACs span the deletion breakpoints on the affected chromosome. Convincing reduction of signal intensities was not observed using BACs 271K11 and 474K4 as FISH probes, which span the WI-12393 gene derived LCRs (fig 2A). To confine the deletion boundaries more precisely, FISH was performed with the cloned PCR products DJ1576/1578 amplified from BAC 2349P21 and DJHK10/11 amplified from BAC 307A16 (fig 2, fig 3C, D). Both probes are deleted on one chromosome 17 homologue. FISH probe DJ1686/1863, spanning exon 1 of the functional WI-12393 gene, hybridised to both chromosomes 17q11.2 and is therefore not deleted. These findings strongly suggest that the proximal deletion breakpoint maps to the region of the KIAA0160 pseudogene and the distal breakpoint to the functional KIAA0160 gene (fig 2B).

Identification of the deletion boundaries by PCR

PCR analysis of several STS using DNA from hybrid cell lines that carry only one chromosome 17 homologue of the index patient III.2 allowed us to confirm the presence of a large deletion (fig 2A). The markers that are framed in fig 2A were absent in the hybrid lines 4 and 15, but could be amplified from DNA of hybrids 8 and 17. To narrow down the boundaries more precisely, PCR products located within duplicated KIAA0160 and WI-12393 regions were amplified from the hybrid cell lines carrying the deleted chromosome 17 homologue and were sequenced. The location of these three primer pairs, DJ1948/1936, DJ1561/1562, and DJ1910/1911, is indicated in fig 2B. Sequence analysis of amplified products and comparison with the known sequences for 2349P21 (AC127024), 778K9 (AC023266), 307A16 (AC003041), and 640N20 (AC090616) enabled us to determine the presence or absence of the functional WI-12393 gene and the distally located WI-12393 pseudogene fragment, as well as the functional KIAA0160 gene and its pseudogene sequences. The PCR product DJ1910/1911 amplified from total DNA of the hybrid cell line 4 carrying only the deleted chromosome 17 contained sequences from both the functional WI-12393 gene and its distally located pseudogene. This result clearly indicates that the deletion does not include these regions and that the breakpoint lies centromeric to intron 2 of the WI-12393 gene. The DJ1910/1911 segment occurs only within the LCR sequence on the distal side of the NF1 gene and is not present in the proximal LCR.¹⁵ Examination of the PCR fragment DJ1948/1936, however, showed that this product was exclusively derived from the KIAA0160 pseudogene fragment located on BAC 2349P21. The sequence of the DJ1948/1936 product was unique and did not contain the functional KIAA0160 gene which on the other hand could be coamplified from the hybrids with the normal chromosome 17. By contrast, analyses of the PCR product DJ1561/1562 showed sequences that were solely amplified from the functional KIAA0160 gene, which is located on BAC 640N20 and 307A16, whereas the PCR product from the normal chromosome was mixed up with sequences from the pseudogene. Therefore, we conclude that the distal and proximal breaks are located within a duplicated segment between exons 5 and 9 of the KIAA0160 gene and its pseudogene, respectively.

Mechanism underlying the deletion

To determine the recombination mechanism resulting in the deletion, polymorphic markers were also analysed in the grandmother, I.2, of the index patient. Haplotypes were determined in patient III.2 using the hybrid cell lines 4,15 and 8,17 carrying only the deleted or normal chromosome of the index patient III.2. The phases of the haplotypes of the grandmother I.2 were inferred on the most parsimonious assumption that the haplotype of II.2 is not the result of a recombination event between grandmaternal haplotypes within the interval analysed (fig 1). According to our findings, intrachromosomal recombination between highly homologous KIAA0160 sequences was associated with the deletion.

To show that the deletion occurred somatically in proband II.2, we investigated further markers in the genomic interval between the KIAA0160 pseudogene and the functional KIAA0160 gene and observed heterozygosity of proband II.2 in fibroblast cultures for the dinucleotide repeat marker IVS27AC28.4 in intron 27b of the NF1 gene and for the single nucleotide polymorphism SNP2 located in the KIAA1821 gene, flanking the NF1 gene in the 3′ direction (fig 4). The heterozygosity of these markers in normal cells of proband II.2 excludes the possibility that the germline deletion occurred during meiosis in the grandmother, and was partially corrected by mitotic recombination or segmental conversion of the deleted chromosome during early embryogenesis in somatic cells of proband II.2.

DISCUSSION

The NF1 gene has one of the highest mutation rates in humans (~1 × 10⁴/gamete/generation) and about 50% of all NF1 patients have no family history of the disease. This high mutation rate cannot be simply explained by the enormous size of the NF1 gene which spans 350 kb. It has been suggested that the high proportion of sporadic NF1 patients mirrors not only new mutations in the germline of a parent but also postzygotic NF1 gene mutations associated with gonadosomatic mosaicism in the transmitting unaffected parent (fig 5).¹⁸ Somatic mosaicism is clearly implicated in the pathogenesis of segmental neurofibromatosis,¹⁹ but only documented at the cytogenetic level for one patient.²⁰ With regard to point mutations, somatic mosaicism of the NF1 gene is difficult to detect and to analyse in tissues, but may be one of the factors contributing to the relatively high number of sporadic cases and to the considerable variability of clinical signs and symptoms in NF1 patients.

Interstitial microdeletions, however, can be approached by FISH more easily and thus somatic mosaicism has been detected in a few NF1 patients at the cytogenetic level.^8,20–26 Large deletions of the NF1 region have been reported in six patients who all showed generalised NF1 symptoms not restricted to a particular body region. Four patients had deletions of the whole NF1 gene as shown by FISH analysis^23–26 and two were ascertained by marker analysis.^21,22 Moreover, Tinschert et al²⁰ described a patient with segmental NF1 who is mosaic for an interstitial NF1 deletion. Breakpoint boundaries in all these seven cases have not been analysed so far and thus the underlying mechanism of mitotic rearrangements remains unclear.

Here, we describe a female with minor signs of the disease, who, however, is mosaic for a large deletion of the NF1 gene region. She transmitted this deletion to her son, who developed a severe microdeletion syndrome. The deletion encompasses a segment of ~1.3 Mb and is the result of non-allelic recombination between the KIAA0160 pseudogene on the proximal side and the functional KIAA0160 gene located distally to the NF1 gene (fig 2). This is the first published case of an NF1 microdeletion mediated by recombination within the KIAA0160 gene. In 29 of 60 NF1 patients with constitutional microdeletions characterised so far with respect to the deletion boundaries, the breakpoints were identified in highly homologous segments of the WI-12393 gene derived duplicons which flank the NF1 gene region.^12–14 The KIAA0160 gene and its pseudogene are also duplicated segments, which are located in close proximity to the WI-12393 gene derived low copy repeats (fig 2A). The KIAA0160 gene, recently termed JJAZ1 (Joined to Juxtaposed with Another Zinc Finger gene) has been shown to be disrupted by somatically acquired translocations t(7;17)(p15;q21) in endometrial stromal sarcomas. This type of translocation not only fuses the promoter and 5′ end of the JAZF1 gene on chromosome 7 to the coding portion of the KIAA0160 (JJAZ1) gene on chromosome 17,²⁷ but may also reduce the physiological activity of the KIAA0160 product as a result of hemizygosity at the KIAA0160 locus. While it seems premature to construct a link between hemizygosity of the KIAA0160 locus and tumour progression, it might be suspected that the triple event, loss of one JAZF1 allele, creation of a new fusion transcript between JAZF1 and KIAA0160, as well as the loss of one KIAA0160 copy triggers the development of endometrial sarcomas as a whole. Similarly, haploinsufficiency of KIAA0160 (JJAZ) may contribute to the outgrowth and progression of neurofibromas in microdeletion patients. Our speculation that KIAA0160 (JJAZF1) serves some tumour suppressive function, which may also be disturbed in endometrial stromal tumours, is consistent with the high risk and early development of tumours observed in patients carrying constitutional microdeletions.

The present study shows that the KIAA0160 gene in conjunction with its pseudogene represents a second recombinational pitfall in the NF1 gene region, besides the WI-12393 gene derived duplicons. Genotype analyses clearly indicate that the deletion occurred somatically in proband II.2 by an intrachromosomal (sister chromatid based) mechanism (figs 1 and 3). According to previous studies, low copy repeat mediated de novo deletions are believed to be triggered by meiotic recombination during germ cell development.^14,28 This notion was based on PCR experiments that failed to identify the patient specific junction product in the blood of their healthy parents.¹⁴ López Correa et al²⁸ observed a strong maternal bias for 17q11.2 microdeletions and found that the 1.5 Mb spanning deletions occurred predominantly during maternal germ cell development. In five of the six informative families, interchromosomal recombination (non-sister chromatid exchange) was the underlying mechanism that caused the rearrangements.

Our study implies that somatic recombination between duplicated sequences at 17q11.2 during early embryonic development can lead to gonadosomal and tissue mosaicism with only mild signs of NF1, which do not meet the diagnostic criteria. Subsequent germline transmission of such a microdeletion, however, can generate a very severe form of NF1 in the offspring with developmental retardation and high tumour load. The risk of somatic mosaicism in patients with very mild or ambiguous signs of NF1 who do not fulfil the usual diagnostic criteria is widely ignored or underestimated, but can have dramatic consequences, as in our family.