Rosettes may be seen in a variety of neoplasms, including neural tumors (neuroblastoma, medulloblastoma, retinoblastoma, ependymoma, schwannoma, and neurofibroma), neuroendocrine tumors (carcinoid, Wilms tumor), ovarian stromal tumor (granulosa cell tumor), epithelial tumor (thymoma), soft tissue sarcoma (hyalinizing spindle cell tumor with giant rosettes), and (rarely) non-Hodgkin's lymphoma and osteosarcoma (1,2). In neuroblastoma and peripheral neuroectodermal tumor, the rosettes are Homer-Wright rosettes. In these rosettes the nuclei of the neoplastic cells are peripherally situated and the cytoplasm locates centrally without a lumen. Flexner-Wintersteiner rosettes are commonly found in retinoblastoma and ependymoma. They resemble Homer Wright rosettes except that a lumen is present in the center of the rosettes.
Although pseudomeissnerian corpuscles have been cited in rare cases of congenital nevi, rosette formation has not been noted in melanocytic neoplasms (3–5). We describe an atypical melanocytic nevus with features of both blue nevus and congenital nevus. An associated proliferative nodule with many rosettes, neuroid cords, and balloon cells was present in the deep dermis.
MATERIALS AND METHODS
Four-micrometer-thick sections were cut from paraffin blocks and stained with hematoxylin-eosin and trichome. Additional paraffin sections of selected blocks were obtained for immunohistochemical studies, which were performed on an automated immunostainer (Ventana, Biotek System, Tucson, AZ) with use of the standard avidin-biotin peroxidase complex technique and heat-induced epitope retrieval buffer.
The following primary antibodies were used: S-100 protein (1:1,200; Dako, Carpinteria, CA), Mart-1 (M2–7C10, 1:100; Signet Laboratory, Dedham, MA), tyrosinase (T311, 1:50; Vector/Nova Castra, Burlingame, CA), HMB-45 (1:100; Dako), Ki-67 (MIB-1, 1:75; Zymed Laboratories, San Francisco, CA), CD34 (QBend, 1:40; Dako), CD31 (JC/70A, 1:80; Dako), CD57 (Leu-7, 1:10; Becton Dickinson, Franklin Lakes, NJ), type IV collagen (CIV 22, 1:50; Dako), neurofilament protein (2F11, 1:1,600; Dako), glial fibrillary acidic protein (1:5,600; Dako), tyrosine hydroxylase (1:200; Daisorin), neuronspecific enolase (BBS/NC/VI-H14, 1:4,000; Dako), synaptophysin (1:100; Dako), protein gene product 9.5 (3IA3, 1:200; Biogenesis, Brentwood, NH), epithelial membrane antigen (1:200; Dako), and vimentin (1:200; Dako).
For comparison, 78 congenital nevi from 65 pediatric patients were examined for rosette formation, neuroid cords, pseudomeissnerian corpuscles, and balloon cell change. The clinical information and follow-up data were obtained from the patients' medical records. The age of the patients ranged from 1 month to 18 years (median: 4 years; mean: 4.7 years). The ratio of males to females was 2:3. The sizes of the nevi were available in 45 cases and ranged from 0.2 to 40.8 cm (median: 4.3 cm). Six of these nevi were giant congenital nevi of a size greater than 20.0 cm. None of these nevi has recurred or undergone subsequent malignant transformation from 1 to 14 years after surgery.
A 59-year-old man presented with an asymptomatic 3.0 × 2.0 cm pigmented lesion in the right conchal bowl of his external ear, extending onto the antitragus area. The lesion was soft and hairy and had been present for many years. A reexcision was performed, and the patient is well, without evidence of disease 1 year after surgery.
Histologic sections showed scattered solitary melanocytes at the dermal epidermal junction. A proliferation of nonatypical and pigmented melanocytes was seen in the dermis surrounding adnexal structures, characteristic of a congenital nevus (Fig. 1). In areas within the superficial dermis, there were epithelioid melanocytes devoid of nuclear atypia. They exhibited maturation with progressive depth. In others, pigmented spindled and dendritic cells infiltrated thickened collagen bundles in a pattern of a blue nevus. In other areas, spindle and wavy nevus cells were seen within a background of fibroblasts and collagen, resembling a neurofibroma. Focally, there were neuroid cords characterized by cylindrical structures with nuclei arranged in a palisade within the neurofibromatous background. Within the deep dermis and subcutaneous tissue, there was an 8.0-mm nodule of epithelioid melanocytes.
The periphery of the nodule merged with the surrounding nevus cells (Fig. 2). The neoplastic cells within the proliferative nodule had nuclear atypia, melanin pigment, focal pseudonuclear inclusions, and balloon cell change. Within this nodule there were many rosettes formed by cells with coarse cell processes radiating toward a center, resembling Homer-Wright rosettes (Fig. 3). The neoplastic cells forming the rosettes also had melanin pigment and balloon cell change (Fig. 4). Melanin pigment was noted in the central cell processes. No prominent mitotic figures, necrosis, lymphovascular invasion, or significant inflammatory infiltrate was noted.
Of the 78 congenital nevi reviewed, rosette formation or balloon cell change was not identified in any, including the six giant congenital nevi. Structures resembling Meissner corpuscles were identified in three cases, and two of these were giant congenital nevi. Neuroid cords were seen in 11 cases, 1 of which was a giant congenital nevus.
The majority of the neoplastic cells in the dermis and within the rosettes and the balloon cells expressed S-100 protein, Mart-1 (Fig. 5), tyrosinase, neuron-specific enolase, and vimentin. HMB-45 and Ki-67 (MIB-1) labeled only rare neoplastic cells within the proliferative nodule. The tumor cells were negative for synaptophysin, protein gene product 9.5, CD57, neurofilament protein, glial fibrillary acidic protein, tyrosine hydroxylase, epithelial membrane antigen, CD31, and CD34. The fibrillary material in the center of the rosettes was weakly positive for S-100 protein, Mart-1, and tyrosinase while negative for trichome, type IV collagen, neurofilament protein, glial fibrillary acidic protein, and tyrosine hydroxylase.
The distinctive rosette structures of our tumor are similar to the Homer-Wright rosettes. One might argue that these rosettes are actually multinucleated giant cells; however, the findings of distinct pigment within the individual cells as well as the presence of balloon cells within the rosettes suggest a structure composed of multiple cells. The tumor cells are melanocytic in origin because they not only contain melanin pigment but also are reactive for multiple melanocytic markers. The central fibrillary tangle most likely represents cell cytoplasm because it stains weakly for S-100 protein, Mart-1, and tyrosinase. Melanin pigment is even noted within these coarse cell processes.
In effect, similar to the Homer-Wright rosettes, the nuclei locate peripherally and the cytoplasm forms the center of the rosette. The negative trichome and type IV collagen stains exclude the possibility of a collagen core. Their morphologic appearance also suggests a neural origin; however, numerous neural markers, including PGP9.5, CD57, neurofilament protein, and glial fibrillary acidic protein, are negative.
We retrospectively examined 78 congenital nevi from 65 pediatric patients at our institution and found no evidence of either rosette formation or balloon-cell change in any of these congenital nevi. Six of these cases were giant congenital nevi. Neurotization in the form of pseudomeissnerian corpuscles and neuroid cords was noted in 3 and 11 cases, respectively. In one of seven giant congenital nevi described by Hendrickson and Ross and in another case report by Jerdan et al., pseudomeissnerian corpuscles within a neurofibromatous stroma were seen in the deep dermis and subcutaneous tissue (3,4).
In a study of 55 giant congenital nevi by Reed et al., Schwann cells, Wagner-Meissner corpuscles, or Verocay bodies were seen in 27% of cases; however, rosette formation was not described (5). In addition, in a large series of 80 giant congenital nevi by Ruiz-Maldonado, rosette formation was not mentioned in the histologic description (6).
The main differential diagnosis of our tumor includes neural lesions that form rosette-like structures, such as neuroblastoma-like epithelioid schwannoma, neuroblastoma-like neurilemmoma, benign epithelioid schwannoma, and dendritic cell neurofibroma with pseudorosettes (7–10). Histologic features of schwannoma such as spindled neoplastic cells, palisading of nuclei, and cellular whorl formation are not seen in our tumor. CD57-positive stellate cells, characteristic of dendritic cell neurofibroma with pseudorosettes, were not seen in the center of the rosettes in our tumor (10). In addition, the superficial component of our tumor was clearly a melanocytic proliferation.
Studies have shown Mart-1 and tyrosinase to be highly specific markers for both benign and malignant melanocytic lesions (11,12). Mart-1 and tyrosinase immunolabeling of the tumor cells within the rosettes and the lack of immunolabeling for neural markers (PGP9.5, CD57, GFAP, and neurofilament protein) argue against a tumor of neural origin.
The differential diagnosis also includes melanotic schwannomas, cutaneous neurocristic hamartoma, cellular blue nevus, and neuroectodermal tumor. Forty to fifty percent of melanotic schwannomas have psammoma bodies, especially those that are associated with Carney's complex (13). In cutaneous neurocristic hamartomas, the absence of a junctional component, a decrease in the number of hair follicles, diffuse HMB-45 immunoreactivity, and CD34 staining of the surrounding stroma would be noted (14). A cellular blue nevus would show cellular nodules of intersecting fascicles of spindle melanocytes (15), whereas the neoplastic cells within the proliferative nodule of our case were predominantly epithelioid.
Our tumor might be considered by some authors to be a neuroectodermal neoplasm arising in a congenital nevus (4); however, this case was negative for synaptophysin, protein gene product 9.5, neurofilament protein, tyrosine hydroxylase, and glial fibrillary acidic protein.
The biologic behavior of this tumor is uncertain because of the presence of nuclear atypia. The lack of mitotic figures, the absence of lymphocytes within or beneath the proliferation, the low Ki-67 labeling of the neoplastic cells within the proliferative nodule, and the focal HMB-45 immunoreactivity are features that make implausible the diagnosis of a malignant melanoma arising in a nevus.
Neurotization can be seen in melanocytic nevi and has been attributed to senescence (16). As a result, various neural structures such as neuroid cords and pseudomeissnerian bodies can be seen. It is possible that the rosettes within our tumor represent modified pseudomeissnerian bodies.
This concept of a single histogenesis of melanocytic nevi with neurotization increasing with age has been challenged by the descriptions of neural features in giant congenital nevi (3–5). The precursor of melanocytes, normal as well as neoplastic, most likely is not a melanoblast—a primitive cell committed to melanocytic differentiation. There is evidence that primitive cells of neural crest origin may not become committed in vivo until they have migrated to the local environment (17).
The neural crest is a temporary embryonic structure whose cells migrate as soon as they are formed, colonizing the entire embryo (18,19). Because of their broad differentiating capability, these stem cells can give rise to diverse cellular phenotypes, including neurons, neural support cells such as Schwann cells, melanocytes of skin and internal organs, melanophores of ocular iris, and endocrine and paraendocrine cells, depending on the microenvironment (20,21). It is possible that neoplastic cells that display different morphologic patterns such as ours are derived from common pluripotential stem or neural crest cells whose differentiation is modulated by the microenvironment.
The authors thank Dr. Arthur G. Weinberg, of Children's Medical Center, Dallas, Texas, for allowing us to review the congenital nevi cases and for reviewing the manuscript, and Drs. George P. Lupton and Maria-Magdalena Tomaszewski, of the Armed Forces Institute of Pathology, Washington, D.C., for reviewing the case.
1. Koo CH, Shin SS, Bracho F, et al. Rosette-forming non-Hodgkin's lymphomas. Histopathology
2. Okada K, Hasegawa T, Yokoyama R. Rosette-forming epithelioid osteosarcoma: a histologic subtype with highly aggressive clinical behavior. Hum Pathol
3. Hendrickson MR, Ross JC. Neoplasms arising in congenital giant nevi: morphologic study of seven cases and a review of the literature. Am J Surg Pathol
4. Jerdan MS, Cohen BA, Smith RR, et al. Neuroectodermal neoplasms arising in congenital nevi. Am J Dermatopathol
5. Reed WB, Becker SW, Becker SW Jr, et al. Giant pigmented nevi, melanoma, and leptomeningeal melanocytosis: a clinical and histopathological study. Arch Dermatol
6. Ruiz-Maldonado R, Tamayo L, Laterza AM, et al. Giant pig mented nevi: clinical, histopathologic, and therapeutic considerations. J Pediatr
7. Fischer C, Chappell ME, Weiss SW. Neuroblastoma-like epithelioid schwannoma. Histopathology
8. Kindblom LG, Meis-Kendblom JM, Havel G, et al. Benign epithelioid schwannoma. Am J Surg Pathol
9. Goldblum JR, Beals TF, Weiss SW. Neuroblastoma-like neurilemmoma. Am J Surg Pathol
10. Michal M, Fanburg-Smith JC, Mentzel T, et al. Dendritic cell neurofibroma with pseudorosettes: a report of 18 cases of a distinct and hitherto unrecognized neurofibroma variant. Am J Surg Pathol
11. Busam KJ, Chen Y-T, Old LJ, et al. Expression of melan-A (MART-1) in benign melanocytic nevi and primary cutaneous melanomas. Am J Surg Pathol
12. Hofbauer GF, Kamarashev J, Geertsen R, et al. Tyrosinase immunoreactivity in formalin-fixed, paraffin-embedded primary and metastatic melanoma: frequency and distribution. J Cutan Pathol
13. Scheithauer BW, Woodruff JM, Erlandson RA. Tumors of the Peripheral Nervous System
. Fascicles 24. Third Series. Washington, DC: Armed Forces Institute of Pathology, 1999:105-76.
14. Mezebish D, Smith K, Williams J, et al. Neurocristic cutaneous hamartoma: a distinctive dermal melanocytosis with an unknown malignant potential. Mod Pathol
15. Rodriguez HA, Ackerman LV. Cellular blue nevus: clinicopathologic study of forty-five cases. Cancer
16. Maize J, Foster G. Age-related changes in melanocytic naevi. Clin Exp Dermatol
17. Perris R, von Boxberg Y, Lofberg J. Local embryonic matrices determine region-specific phenotypes in neural crest cells. Science
18. Kissel P, Andre JM, Jacquier A. Formation and fate of the neural crest and the pigment cells. In: Kissel P, Andre JM, Jacquier A, eds. The neurocristopathies
. New York: Masson Publishing, 1981:1-25.
19. Weston JA. The migration and differentiation of neural crest cells. Adv Morphogen
20. Clarke DL, Johansson CB, Wilbertz J, et al. Generalized potential of adult neural stem cells. Science
21. Le Dovarin N. Migration and differentiation of neural crest cells. Curr Top Devel Biol