Despite a well-characterized lack of specificity, pathologists routinely employ S100 in the diagnosis of
neural crest-derived tumors. Recent studies have shown that Sox10 is a reliable marker of neural crest differentiation that is consistently expressed in schwannian and melanocytic
tumors. We sought to validate these results in a larger series of
soft tissue neoplasms of both neural crest and non-neural crest origin, and to further characterize the sensitivity and specificity of Sox10 for use in clinical diagnosis. We evaluated Sox10 and S100
mRNA levels in 122 cases of
peripheral nerve sheath tumors and
synovial sarcoma and used immunohistochemistry for Sox10 and
S100 protein expression in 1012 tissue specimens. This study includes 174 tissue microarray cases previously reported by Nonaka and colleagues, which include cases of
melanoma,
dermatofibrosarcoma protuberans,
neurofibroma,
synovial sarcoma,
clear-cell sarcoma,
malignant peripheral nerve sheath tumor (
MPNST),
perineurioma, and
schwannoma.
Synovial sarcomas expressed significantly higher levels of S100B than Sox10 (P=7.9×10), and no significant Sox10
mRNA expression was identified in
synovial sarcoma (n=40), whereas 18/40 cases showed comparatively increased levels of S100
mRNA. The majority of
schwannomas (n=26) and
neurofibromas (n=28) showed relatively an increased expression of both Sox10 and S100
mRNA. MPNSTs (n=28) showed variable levels of Sox10 and S100
mRNA expression, and these expression levels were highly correlated (Pearson correlation coefficient r=0.79). In contrast, immunohistochemistry performed on a larger and more varied number of cases highlighted significant differences between the 2
proteins. We identified 5 non-neural, nonmelanocytic
sarcoma types in which a subset of cases showed
S100 protein expression:
synovial sarcoma (12/79, 15%),
Ewing sarcoma (3/14, 21%),
rhabdomyosarcoma (4/17, 24%),
chondrosarcoma (3/4, 75%), and
extraskeletal myxoid chondrosarcoma (5/11, 45%). For each of these entities, we identified cases with strong and diffuse S100 staining. Of these cases, only 1 case of
rhabdomyosarcoma showed focal Sox10 positivity. In 78 cases of
MPNST, S100 increased the sensitivity (31/78, 40%) as compared with Sox10 (21/78, 27%), but the majority of these cases were negative for both Sox10 and S100 (44/78, 56%). Sox10 proved superior to S100 in the detection of desmoplastic
melanoma (7/9, 78%) and
clear-cell sarcoma (4/7, 57%). We also report for the first time Sox10 expression in 26 cases of
granular cell tumor, further supporting the neural crest derivation of this
tumor. Excluding
MPNST, S100 and Sox10 showed similar sensitivity in
tumors of neural crest origin (140/148, 95% and 137/148, 93%, respectively). In summary, Sox10 shows an increased specificity for
tumors of neural crest origin compared with S100: Sox10 was positive in only 5 of 668 cases (99% specificity) in nonschwannian, nonmelanocytic
tumors, whereas S100 was positive in 53 of 668 cases (91% specificity). Sox10 should be used in the place of or along with S100 in soft tissue
tumor diagnosis.