Perivascular Cells Harboring Multiple Endocrine Neoplasia Type 1 Alterations Are Neoplastic Cells in Angiofibromas

Neoplasia of tissue is caused by clonal proliferation of cells, resulting in benign or malignant tumors. In many tumors, particularly malignant ones, the proliferating cell population is readily identifiable as an invasive mass that is composed primarily of a single type of tumor cell. The tumor is identified and classified according to the histological and architectural differentiation of the proliferating cell. For example, most colon cancers exhibit characteristic features, such as glandular architecture, epithelial lining, and intracellular and intraglandular mucin, consistent with the differentiation of normal colorectal epithelium (1).

Slower-growing benign tumor cells, however, tend to preserve the architectural background. The background stromal tissue may proliferate as well, in response to either physical irritation or factors secreted by the neoplastic cells, and the resulting tumors are frequently composed of different types of cells. In this situation, it may be difficult to elucidate the truly neoplastic component on morphological grounds alone.

Angiofibromas are small tumors of the skin that are composed of a variety of cellular elements, including fibroblasts, neurosustentacular cells, dermal dendrocytes, and dilated vessels (2), intermixed with deposits of dermal collagen. Due to the complex histological architecture of angiofibromas, their etiology is currently considered to be hamartomatous rather than neoplastic (2).

Angiofibromas are frequently found in patients with MEN1² (3). Mutation and deletion analyses of MEN1-associated neuroendocrine tumors revealed two genetic hits, germ-line mutation of the MEN1 tumor suppressor gene combined with allelic deletion of the opposite wild-type allele, as an essential pathogenetic mechanism (4, 5, 6, 7, 8, 9). In contrast to the rather monotonous histological pattern of neuroendocrine tumors, however, angiofibromas display a variety of cellular elements, many of which may be reactive rather than neoplastic. In this study, we applied a combined approach of morphological, FISH, and genetic deletion analysis, which succeeded not only in detecting specific genetic changes in angiofibromas but also in localizing the neoplastic component.

Cutaneous angiofibromas from five patients with MEN1 were retrieved from the files of the Laboratory of Pathology, National Cancer Institute. Multiple tumors were removed from one patient, and touch preparations were prepared from two of them for FISH analysis.

Unstained 6-μm sections on glass slides were deparaffinized with xylene, rinsed in ethanol from 100 to 80%, briefly stained with H&E, and rinsed in 10% glycerol in 10 mm Tris (pH 8)-1 mm EDTA buffer. A slightly modified microdissection procedure (10) was performed under direct light microscopic visualization using a 30-gauge needle, as described previously. Tumor cells were procured (Fig. 2) from the following areas: (a) control tissue for analysis of constitutional DNA from epidermis; (b) putative tumor cell complexes located around vascular cells; and (c) mesenchymal cells from angiofibroma without putative tumor cell complexes.

Procured cells were immediately resuspended in 5 μl of buffer containing Tris-HCl (pH 8.0), 1.0 mm EDTA (pH 8.0), 1% Tween 20, and 0.1 mg/ml proteinase K and were incubated at 37°C overnight. The mixture was boiled for 5 min to inactivate the proteinase K, and 2 μl of this solution were used for PCR amplification of the DNA.

All cases were examined for LOH with microsatellite markers for the MEN1 locus at chromosome 11q13 (D11S449 and PYGM). The markers were chosen on the basis of ease of amplification and informativeness. Each PCR sample contained 2 μl of template DNA, as described above; 10 pmol of each primer; 20 nmol each of dATP, dCTP, DGTP, and DTTP; 15 mm MgCl₂; 0.1 unit of Taq DNA polymerase; 0.05 μl [³²P]dCTP (6000 Ci/mmol); and 1 μl of 10× buffer in a total volume of 10 μl. PCR was performed with 35 cycles, each consisting of denaturing at 94°C for 1 min, annealing at 55°C for 1 min with D11S449 and PYGM, and extending at 72°C for 90 s. The final extension was continued for 10 min.

Labeled amplified DNA was mixed with an equal volume of formamide loading dye (95% formamide, 20 mm EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol). Samples were then denatured for 5 min at 95%, loaded onto a gel consisting of 6% acrylamide (acrylamide-bisacrylamide, 49:1), and electrophoresed at 1800 V for 90 min. After electrophoresis, the gels were transferred to 3-mm Whatman paper and dried. Autoradiography was performed with Kodak X-OMAT film (Eastman Kodak, Rochester, NY). A case was considered informative for a polymorphic marker if normal tissue DNA showed two different alleles (heterozygosity). The criterion for LOH was complete or near complete absence of one allele in the tumor DNA as defined by direct visualization.

Touch preps were made from fresh tumor specimens from four patients. After fixation in methanol-acetic acid (3:1) for 20 min, slides were air- dried, equilibrated in 2× SSC solution, and dehydrated in ethanol series of 70, 80, 90, and 100%. In situ hybridization was performed using cosmid clone c10B11 containing the MEN1 as a probe. The DNA was labeled with digoxigenin-11-dUTP by nick translation (Boehringer Mannheim) and ethanol-precipitated in the presence of 50× herring sperm DNA and 50× Cot-1 fraction of human DNA. The DNA pellet was resuspended in Hybrisol solution (50% deionized formamide, 10% dextran sulfate, and 2× SSC) to a final concentration of 25 ng/ml. Slides were denatured in 70% formamide-2× SSC at 72°C for 2 min with the following incubation in cold (−20°C) ethanol series of 70, 80, 90, and 100% for 2 min each and air-dried. Probes were denatured at 78°C for 10 min and then incubated for 30 min at 37°C for preannealing. A total amount of 250 μg of DNA probe was applied on the slide. α-Satellite repetitive DNA, specific for chromosome 11 (Oncor), was denatured separately and mixed with the cosmid probe just prior to hybridization. Overnight hybridization was done in a humidified chamber at 37°C.

Posthybridization washes were at 45°C in 50% formamide-2× SSC (5 min, three times each), 1× SSC (5min, two times each), and 0.1× SSC (5min, two times each). Detection was performed using avidin-FITC and anti-digoxigenin-rhodamine (40 min at 37°C), followed by washing in 4× SSC-0.1% Tween 20 solution at 45°C (2 min, three times each) and counterstaining with 4′,6-diamidino-2-phenylindole-antifade (0.25 mg/ml).

Hybridization signals were scored using a Zeiss Axiophot epifluorescence microscope and two-color images were captured on a Photometrics charged coupled device camera (Photometrics, Ltd., Tucson, AZ) using IP Lab Image software (Signal Analytics Corporation, Vienna, VA). At least 100 interphases with strong hybridization signals were scored. Presence of >20% cells with one MEN1 signal was interpreted as an allelic loss. Normal control (normal tissue) showed 3% of cells with one MEN1 signal.

We used the surgical pathology material of five patients with MEN1 and cutaneous angiofibromas (3). In four of the patients, germ-line MEN1 mutations were identified. The tumors measured between 0.3 and 1.0 cm in size and were located at the nose, upper lip, earlobe, chin, shoulder, arm, or groin. From one patient (patient 4), multiple tumors were removed. For initial genetic analysis, angiofibroma tissue was procured by dissecting the dermal angiofibromatous area from H&E-stained slides. Tissue from epidermis and/or adnexal structures adjacent to the angiofibroma nodule was used as normal control tissue. PCR-based LOH analysis of the angiofibromatous tissue, however, consistently failed to reveal LOH. Failure to detect LOH, however, can be interpreted in two different ways:

(a) Angiofibromas consist of cells with constitutive genotype; accordingly, tumor cells and normal tissue from the same patient will yield identical results upon genetic testing. This interpretation would be consistent with the current concept, according to which angiofibromas are not composed of truly neoplastic tissue but rather represent masses of mature, disorganized cells, so-called “hamartomas” (2, 10, 11).

(b) Angiofibromas are composed of truly neoplastic cells, admixed with abundant coproliferating, mature, reactive cells. Consequently, detection of LOH assays would be obscured by the presence of cells with constitutive genotype.

Two additional studies supported the latter hypothesis. (a) We performed touch preparations from fresh angiofibroma tissue of four MEN1 patients and analyzed the cells using FISH (12). The results showed 48–64% cells with allelic deletion and complementary numbers of cells without allelic deletion (Fig. 1, a and b) and, therefore, provided evidence for a subset of cells to be affected by a “second genetic hit” of MEN1. (b) We performed mutation analysis of 20 angiofibromas from patients without evidence of hereditary disease (13). Although we failed to detect allelic deletion of the MEN1 locus in these sporadic tumors, we identified missense MEN1 mutations in two cases. Again, the results suggested that the analyzed samples consisted of both neoplastic and reactive cells because both a mutation nucleotide and the wild-type nucleotide were detected (Fig. 1, c and d). We concluded from FISH results of MEN1-associated angiofibromas and mutation analysis of sporadic angiofibromas that these tumors contain neoplastic cells that are intermixed with a substantial proportion of nonneoplastic reactive cells. Therefore, detection of allelic deletion by PCR based LOH analysis would require selective procurement and analysis of the neoplastic cell compartment.

Searching for specific cell populations that could potentially represent genetically altered neoplastic cells, we performed immunohistochemical studies of the five cases using consecutive serial sections. In view of the neuroendocrine phenotype of classic MEN1-associated tumors, we applied markers for synaptophysin, chromogranin, and neuron-specific enolase, which, however, failed to detect any immunoreactive cells. Immunostaining with antivimentin, a relatively nonspecific marker for a variety of mesenchymal cells, revealed conspicuous clustering of mesenchymal cells around small vessels in all five cases. Cytologically, these cells resembled fibrous or histiocytic dermal cells (Fig. 2). The vimentin-positive cell clusters appeared identical to those described previously as “partial” or complete “perithelial” or “histiocytic” coats or “hamartial germs” (11). The number of the perivascular cell clusters varied markedly between the cases.

Guided by the hypothesis that the clusters of perivascular cells represented the “true” neoplastic component of angiofibroma, we selectively microdissected these cell clusters for subsequent genetic analysis. Manual dissection of the cell clusters was facilitated by an increased adhesiveness of the perivascular cells to each other, compared to the surrounding spindle cells (Fig. 3). In most cases, however, we were unable to separate the endothelial cells from the dissected cell cluster. By cell counting, however, we assured that the ratio of the number of potentially neoplastic perivascular cells to the number of probably reactive endothelial cells was not smaller than 5:1.

LOH analysis of the dissected samples confirmed the hypothesis that the perivascular cells represent neoplastic cells. The majority of samples showed LOH of the MEN1 locus (Fig. 4). Different clusters taken from different biopsies from the same patient always showed loss of the same allele. Furthermore, by comparing the allelic loss with that of multiple other neuroendocrine tumors in patient 1, we were able to corroborate that it was the same allele that was lost in all tumors (data not shown).

Genetic analysis of tumors has been proven to be a powerful tool in identifying changes of a variety of genes that are believed to be causative or associated with certain types of neoplasia. In conjunction with tissue microdissection, genetic analysis can be applied to small, histologically definable groups of cells (14). Here, we introduce genetic analysis of tumor tissue as a flexible and powerful tool to more closely identify tumors of unknown cellular origin.

In contrast to analysis of genetic mutation, allelic deletion (LOH) analysis of tumors requires relatively pure tumor cell samples. Most tumors, however, are composed of proliferating neoplastic cells, admixed with coproliferating, “reactive” nonneoplastic cells. If the number of reactive cells exceeds that of neoplastic cells in a tumor sample, the LOH result will be obscured. Cutaneous angiofibromas are characterized by dermal fibrosis and associated vascular proliferation with unknown histogenesis (10). Because of the variety of cellular constituents which bear close cytological resemblance to normal dermal cellular elements, angiofibromas are not recognized as true neoplasms (2, 10). In this study, we performed morphological, immunohistochemical and genetic analysis of a series of cutaneous angiofibromas. The goal of the study was not only to prove a true genetic association of angiofibroma development with MEN1 but to identify these tumors as true neoplastic processes and more closely identify the neoplastic cell compartment.

Formation of tumors may be based on reactive, hamartomatous, or neoplastic etiology. Regardless of the specific etiology, however, most tumors have a complex architecture, being composed of a variety of cellular elements including vascular, mesenchymal, and other components. If the etiology is neoplastic, i.e., due to clonal proliferation of one of its cellular constituents, the tumor can be assumed to be composed of both neoplastic and reactive elements. However, the etiology of many tumors will remain obscure unless the reactive and neoplastic components can be separated, procured, and genetically analyzed.

Inherited tumor syndromes serve as valuable models for detailed genetic study. Tumors occur in different organs, and frequently, they occur in multiplicity within the same organ. Each individual tumor appears to arise independently, and knockout of the wild-type tumor suppressor allele is hypothesized to represent an early or even initiating event (15). Furthermore, extensive studies of tumors in various hereditary tumor syndromes, including MEN1, von Hippel-Lindau disease, and others, have characterized the second hit as allelic deletion. Therefore, the tumor cells in these hereditary syndromes are genetically “marked” by a deletion event that can be sensitively detected whenever the tumor cells are selectively procured. In contrast, procurement of nonneoplastic cells (or of tissue probes consisting of predominantly nonneoplastic cells) will demonstrate presence of both tumor suppressor gene alleles.

Here, we applied selective tumor cell analysis to cutaneous angiofibromas, which frequently occur in association with MEN1. Angiofibromas are benign tumors that are predominantly composed of mesenchymal spindle cells and vessels. On the basis of the bland morphological features, these tumors have been considered to represent reactive/hamartomatous masses rather than true neoplasia.

Two lines of evidence, however, suggested that angiofibromas are not exclusively composed of reactive cells. The first evidence was provided by the observation of MEN1 deletion with cosmid clone c10B11 containing MEN1 in tumor nuclei that had been prepared by touching angiofibroma tissue onto glass slides (12). The second evidence was derived from SSCP and sequencing analysis of sporadic angiofibromas, which revealed bands representing both constitutive and aberrant genotypes (13). The applications of both FISH and/or SSCP/sequencing analysis, however, are limited because they do not allow to study cells in their original architectural context; therefore, morphology and location of the genetically altered cells remained unknown.

In patient 1, whose cutaneous lesions were analyzed in this study, we were consistently able to demonstrate MEN1 deletions in other MEN1-associated tumors. Consistently, it was the same allele that was lost in each individual neoplasm. However, initial analysis of this and other MEN1 patients’ angiofibromas failed to reveal MEN1 deletion. In retrospect, the absence of MEN1 deletion was caused by an abundance of tissue cells with normal constitutive genotype. In contrast, after selective procurement of perivascular cell complexes, previously called hamartial germs (11), we were consistently able to demonstrate loss of one MEN1 allele.

Although our study provides evidence for the presence and the location of neoplastic cells in angiofibromas, the histogenesis of these cells remains unclear. Recent experiments support the concept that MEN1-associated tumors may originate from pluripotent cells that may differentiate along different pathways, including mesenchymal and epithelial lineage (16). The rich vascularization of angiofibromas and noncutaneous, MEN1-associated neuroendocrine tumors may indicate angiogenetic properties of neoplastic cells with MEN1 deletion.

In conclusion, we support previous evidence that MEN1-associated angiofibromas arise as part of the spectrum of MEN1 disease that has so far been predominantly characterized by development neuroendocrine tumors. MEN1-associated angiofibromas represent true neoplastic processes. The neoplastic cells are concentrated in perivascular location and reveal mesenchymal immunophenotype. Further studies will have to closer characterize the neoplastic cell in other MEN1-associated tumors.

We thank members of the NIH Interinstitute Endocrine Program for clinical contributions.