Methods

SAMPLE COLLECTION

Haemangioma tissue was collected from three centres. Tissue from one centre was only available as archived paraffin wax blocks. For these samples, the only clinical details available were diagnosis and histology of haematoxylin and eosin stained sections. Further fresh tissue was obtained from two other surgical centres. For these lesions, the four that amplified successfully were all in the proliferative phase and located on the face (US1: boy with 1.5 cm lesion beneath nose resected at 9 months of age; US2: girl with 2–3 cm lesion on forehead resected at 3 months of age; US3: girl with 2 cm lesion on lower lip resected at 5 months; US4: girl with 1 cm lesion on right upper eyelid resected at 6 months of age). Individual US4 was reported to have several haemangiomas.

MICRODISSECTION

Formalin fixed, paraffin wax embedded, methyl red stained 10 μm thick sections of infantile haemangiomas in the proliferative phase were obtained from three surgical centres. After dewaxing with xylene, sections were overlaid with 200 μl of microdissection buffer (20 mM Tris/HCl, 1 mM EDTA, 0.05% Tween 20, pH 8.6). Under a dissecting microscope, the region of interest was scraped from the slide using a sterile disposable pipette tip. For each lesion analysed, this procedure was performed twice. To identify the genotype expected in normal tissue from an individual, matched DNA extracted from blood was available for some of the haemangioma samples. Where no peripheral blood DNA was available for analysis, the complete tissue sample from an independent section of the same lesion, comprising skin, fatty tissue and lesion, was removed into buffer as a control reaction to give the expected genotype.

Buffer and tissue were transferred to a microcentrifuge tube and incubated with 20 μl of proteinase K (20 mg/ml) for 16 hours at 50°C, followed by heating to 100°C for 10 minutes to inactivate the proteinase K. Aliquots (2 μl) of this extract were subsequently used for PCR amplification of markers.

MARKER ANALYSIS

Primers used were from Marshfield set 8, obtained from Research Genetics™. Primer sequences and marker maps are available via the Marshfield Medical Research Foundation web site (www.marshmed.org/genetics/). Primers were FAM, HEX, or TET labelled. PCR amplification from 2 μl of DNA solution was performed according to the manufacturer's recommended conditions in 96 well plates. After amplification, products were run on an ABI377^TM sequencer and analysed using GeneScan v2.02^TM software. Markers clearly showing the presence of two alleles in the control tissue, and with one or two alleles of the correct size in the microdissection tissue sample, were considered “informative” and were analysed further. The ratio of the peak areas of the smaller to larger fragment was calculated for the control sample from the individual, and for each of the two samples from the lesion. These were designated ql (lesion) or qc (control). Loss of heterozygosity was taken as having occurred if ql/qc deviated significantly (< 0.5 or > 2.0) from the expected value of 1.^{6, 7}

STATISTICAL ANALYSIS

The number of markers showing LOH on chromosome 5q was compared with the number of markers showing LOH on chromosomes 5p and 9. Statistical analysis involved the construction of a two by two table and the use of the χ² test with one degree of freedom.

Results

Six proliferative phase haemangiomas were successfully microdissected and amplified. In total, 86 sample/marker combinations were successfully amplified and were informative. These data are shown in tables 1 and 2 and are summarised in table 3. The total number of LOH events identified on chromosome 5q was compared with the number of LOH events identified in control markers from chromosome 5p. In a subset of lesions, markers from chromosome 9 were analysed. This was initially performed because the endoglin gene, responsible for vascular dysplasia hereditary haemorrhagic telangiectasia type 1 (HHT), maps to 9q34, although we recognise that a different pathological process is involved in lesion formation in HHT.

Using the χ² test, significantly more markers showed LOH on chromosome 5q than on chromosome 5p (p < 0.05) or chromosome 9 (p < 0.05).

Examining data from individual lesions, three haemangiomas showed LOH for three or more markers tested on 5q. In these lesions (UK2, US3, and US4), the LOH was distal to D5S1453, although no distal limit to the region of LOH on 5q was identified.

Discussion

We have demonstrated LOH for DNA markers on chromosome 5q in a proportion of proliferative haemangiomas. This suggests that somatic mutational events on 5q have a role in haemangioma formation.

Analysis for LOH is invariably complicated by the contamination of tumour tissue with normal surrounding tissue. The cut off point used for the identification of LOH assumes that the mutational event is seen in > 50% of the cellular DNA sampled. The preferential amplification of one (usually the smaller) allele is a recognised complication of PCR amplification from small amounts of DNA. However, in the 19 LOH events seen on 5q, six of 19 events involved loss of the smaller allele. Statistical comparison of the number of LOH events seen on 5q with those on 5p and chromosome 9 also reduces this as a source of error.

The chromosomal regions showing LOH on 5q overlap with the candidate interval demonstrated for hereditary haemangioma by Walter et al,⁵ even though we recognise that the candidate interval published is very large, and that different loci might be involved.

Our data are the first to suggest that sporadic haemangioma is associated with somatic mutation events. These data showing LOH involving a subset of markers on 5q, in association with the identification of a locus for hereditary haemangioma on 5q,⁵ are suggestive of a tumour suppressor gene for haemangiomas in this region.

There are many loci within the overlapping candidate regions, and several of these might have a role in the control of endothelial cell proliferation and differentiation. The vascular endothelial growth factor receptor 3 gene (FLT4) is mutated in hereditary lymphoedema,⁸ although haemangiomas are not reported as a feature in these affected families. The platelet derived growth factor receptor β is expressed in vascular smooth muscle cells, and has a role in the recruitment of pericytes by developing blood vessels.⁹

The interferon regulatory factor 1 (IRF1) is a known tumour suppressor gene, which has been shown to be involved in gastric cancer and some forms of leukaemia.¹⁰ However, no specific role for IRF1 in the endothelial cell has been established. The SPARC (secreted protein acidic and rich in cysteine) gene encodes a protein that is secreted by endothelial cells, particularly after certain types of injury.¹¹ SPARC also has a role in bone matrix formation. The fibroblast growth factor receptor 4 and the fibroblast growth factor 18 also map to the candidate region. All these genes have been sequenced in individuals from autosomal dominant haemangioma families, but no causative mutation has been identified in these genes (J Walter, personal communication, 2001).

The endothelial cell proliferation that gives rise to a haemangioma might be caused by the complete loss of a gene that acts as an endothelial growth suppressor. This situation would be similar to that seen in the “two hit” model for retinoblastoma.¹² Under this model, the high frequency of sporadic haemangioma might be explained in two ways. Mutations in the haemangioma predisposing gene could be relatively common in the population, with the “second hit” occurring relatively infrequently. Alternatively, the region of chromosome 5q containing the haemangioma predisposing gene could be particularly prone to mutation in endothelial cells.

This tumour suppressor model for the formation of sporadic haemangiomas would be consistent with the clinical presentation. Most lesions arise in a single random location, although occasionally multiple lesions arise within a larger confluent area of skin. Disseminated haemangiomatosis can also occur. A single lesion might arise from a mutation in a single endothelial cell; multiple lesions would arise from a mutation in an endothelial precursor that multiplies to give multiple affected endothelial cells within one region, or occasionally throughout the body.

An alternative genetic model would be that the function of the gene mapping to the chromosome 5q locus might be particularly sensitive to gene dosage effects. In this case, either germline or somatic mutation leading to haploinsufficiency for the gene would be sufficient to predispose to haemangioma. Subsequent haemangioma formation would have to be initiated by a localised environmental insult, presumably arising in utero.

If loss of a tumour suppressor gene causes proliferation, this does not explain why haemangiomas usually undergo involution after proliferation. This could be explained if the proliferating cells subsequently differentiate, or are subject to a limitation in the number of cell divisions that they can undergo before apoptosis occurs. Even in the presence of a somatic mutation, haemangioma proliferation and involution will be influenced by growth factors in the microenvironment and changes in circulating hormones at the time of birth and during growth.

Our study provides further evidence that there is a genetic component to the aetiology of sporadic haemangiomas of childhood. Further studies are required to refine the candidate interval and identify the genes involved.