Nakajima et al1 first described the results of S100protein staining in formalin-fixed paraffin-embedded
material for a large number of tumors and tissues in 1982and concluded that S100 was a useful diagnostic tool intumor diagnosis. This work also showed that S100 wasalso variably expressed in tissues of adipocytic, chon-droid, and myoepithelial derivation, as well as staininginterdigitating dendritic cells and Langerhans cells.1 Laterstudies showed that S100 was expressed in many non-neural crest-derived tumors including Ewing sarcoma,rhabdomyosarcoma, and synovial sarcoma.2–5 Despite itslimited specificity, S100 is commonly used as the majortool to determine neural crest differentiation and is fre-quently deployed by pathologists when faced with aspindle cell neoplasm of uncertain differentiation. S100 isa family of proteins containing 2 EF-hand calcium-binding motifs. S100B is more abundantly expressed inglial cells, melanocytes, adipocytes, and chondrocytes.S100 protein antibodies are often polyclonal but aretypically strongly reactive against S100B.
Sox10, a member of the sex-determining regionY-related HMG-box family, is a recently characterizedmarker of neural crest differentiation that has also beenimplicated as a neural crest stem cell marker.6,7 Severalhuman pathologies are associated with Sox10 mutationsincluding Hirschprungs disease and Waardenburg syn-drome.8–10 Intact Sox10 signaling is required for normalcentral and peripheral nervous system myelination.11–13
Previous studies have demonstrated that Sox10 is a more
Received for publication July 27, 2011; accepted December 4, 2011.From the *Department of Pathology, Stanford University Medical
Center, Palo Alto, CA; and wGenetic Pathology Evaluation Centre,University of British Columbia, Vancouver, BC, Canada.
The authors declare no conflict of interest.Reprints: Jason Raj Karamchandani, MD, 300 Pasteur Dr, R241A, Palo
Alto, CA 94025 (e-mail: [email protected]).Copyright r 2012 by Lippincott Williams & Wilkins
sensitive and specific marker than S100 for schwannianand melanocytic tumors.14 In this study we sought toconfirm the findings reported by Nonaka and colleaguesin a larger set of neural crest tumors as well as in a widearray of benign and malignant soft tissue neoplasms ofknown and unknown histogenesis to determine the sen-sitivity and specificity of Sox10 as compared with S100for clinical diagnostic use. We also compared Sox10 andS100 mRNA expression in schwannomas, neurofibromas,malignant peripheral nerve sheath tumors (MPNSTs),and synovial sarcomas.
MATERIALS AND METHODS
Gene Expression ProfilingThe data set consisted of 122 cases (40 synovial
sarcomas, 26 schwannomas, 26 neurofibromas, and 28MPNSTs). Prior analyses on these cases have been pub-lished15 and the data set is publicly available in the Stan-ford Microarray Database (http://smd.stanford.edu).16
Tissue Microarray ConstructionTissue microarrays (TMAs) were assembled ac-
cording to established protocols. The following tissuemicroarrays were used in this study: TA19, TA138,TA166, TA171. These microarrays have been described inpublished studies.17,18 TA138 was also previously eval-uated for Sox10 and S100B reactivity in the 2008 study byNonaka et al.14
ImmunohistochemistryFour-micrometer TMA sections and conventional
sections were stained for both Sox10 and S100. Reagentsand conditions including clone, vendor, and dilution aresummarized in Table 1. Standard 4-mm paraffin block tissuesections from 14 synovial sarcomas, 6 cellular schwanno-mas, 1 Ewing sarcoma/primitive neuroectodermal tumorand 1 rhabdomyosarcoma were also examined for bothSox10 and S100 expression. The polyclonal S100 antibodyused shows strong labeling of S100B, weak labeling ofS100A1, and very weak labeling of S100A6.
Nuclear staining in tumor cells was required for apositive interpretation for both S100 and Sox10 stains.The pattern of immunostaining was scored on a 3-tieredscale (0 to 2), with a score of 0 indicating no staining, ascore of 1 indicating focal nuclear staining, and a score of2 indicating staining in the majority of lesional cells. Forthose cases where multiple TMA scores were evaluatedfrom the same specimen and where scores differed, thehighest score was assigned. For the purpose of calculatingsensitivity and specificity, a score >0 was counted aspositive.
RESULTS
Gene Expression AnalysisThe relative Sox10 and S100B gene expression levels
were assessed in 122 publically available cases of MPNST,scwhannoma, neurofibroma, and synovial sarcoma. Theraw expression data were mean centered by individualarray. None of the 40 cases of synovial sarcoma showedrelatively an increased expression of SOX10, though 18/40 cases had relatively an increased S100B mRNA ex-pression (Fig. 1). A paired Student t test confirmed thatsynovial sarcomas express significantly lower levels ofSOX10 than S100B (P=7.9�10�9). The majority ofboth schwannomas and neurofibromas showed relativelyan increased expression of SOX10 and S100. Sox10 andS100 expression was highly correlated in schwannomas(Pearson correlation coefficient r=0.89), and neuro-fibromas (r=0.79). MPNSTs had variable but signi-ficantly correlated levels of S100 and SOX10 expression(r=0.79).
Immunohistochemical StudiesWe examined Sox10 and S100 protein expression of
these markers in 1012 specimens representing 66 diag-nostic entities and 33 normal tissues. These studies in-cluded 174 cases previously studied by Nonaka andcolleagues, which included 9 cases of melanoma, 15 casesof dermatofibrosarcoma protuberans, 43 cases of neuro-fibroma, 15 cases of synovial sarcoma, 5 cases of clear-cellsarcoma, 61 cases of MPNST, 4 cases of perineurioma,
and 22 cases of schwannoma. Immunostaining was per-formed independently in our lab, and scored by a blindedobserver. Consistent with previously published results, weidentified 5 non-neural sarcomas that showed immun-histochemical S100 protein expression: synovial sarcoma(12/79, 15%), Ewing sarcoma (3/14, 21%), rhabdomyo-sarcoma (4/17, 24%), chondrosarcoma (3/4, 75%), andextraskeletal myxoid chondrosarcoma (5/11, 45%).Strong and diffuse S100 staining was observed in a subsetof all of these cases (Table 2). Only in one of these cases arhabdomyosarcoma showed focal Sox10 positivity, andnone showed strong and diffuse Sox10 staining (Fig. 2).
Sox10 and S100 were both reliably expressed at theprotein level in schwannomas (45/50, 90%), and neuro-fibromas (52/55, 95%). Of the 5 Sox10-negative schwan-nomas, 2 of these cases also showed no evidence of S100expression (Fig. 3). Sox10 and S100 dual positivity wasalso identified in granular cell tumor (n=26), and gan-glioneuroma (n=2). Excluding MPNSTs, in tumors ofneural crest origin, Sox10 was positive in 137 of 148(sensitivity 93%) cases and S100 was positive in 140 of148 cases (sensitivity 95%).
Both Sox10 and S100 showed highly variable pro-tein expression in MPNSTs. Of 78 cases, 44 cases (56%)showed no evidence of either Sox10 or S100 expression.Eighteen cases were dual positive. Thirteen cases (17%)showed S100 expression without Sox10 expression, and 3cases (4%) were positive for Sox10 and negative for S100(Fig. 3). The sensitivity of Sox10 and S100 for the de-tection of MPNST was 27% and 40%, respectively.
We examined a total of 668 cases of mesenchymalneoplasms of nonschwannian, nonmelanocytic soft tissueorigin with Sox10 and S100. The results are summarizedin Table 3. These cases include the aforementioned sar-comas. Sox10 staining was seen in 5 cases (99% speci-ficity). S100 was positive in 53 of 668 cases (91%specificity). Dual Sox10 and S100 staining was seen incases diagnosed as malignant fibrous histiocytoma (focal)and sarcoma not otherwise specified (strong and diffuse).These cases may represent unrecognized MPNSTs. Asingle case of nodular faciitis showed focal Sox10 stainingand strong and diffuse staining for S100. A single case ofdigital fibroma showed focal Sox10 staining (no tissuewas available for S100 staining). Five glomus tumors werestained. Three of these cases showed S100 positivity, 2strong and diffuse, and 1 with focal staining. Sox10 wasalso focally positive in the latter case.
Four cases of perineurioma were evaluated. Noneshowed Sox10 expression.
We identified a subset of sarcomas in which Sox10stained a greater number of the tumor cell nuclei. Thesetumors included clear-cell sarcoma (4/7 cases) and des-moplastic melanoma (7/9 cases).
We stained 33 normal tissue types for Sox10. Sox10-negative normal tissues included adrenal, appendix,bladder, cervix, colon, duodenum, endometrium, epi-didymis, esophagus, gallbladder, heart, kidney, liver,lung, ovary, pancreas, parathyroid, placenta, prostate,seminal vesicle, skeletal muscle, skin, small bowel, spleen,stomach, testis, thyroid, tonsil, umbilical cord, and ute-rus. A number of normal tissues were Sox10 positive.These included the Schwann cells of peripheral nerve, aswell as myoepithelial cell population of the breast andacinar and myoepithelial cells of salivary gland tissue(salivary gland duct epithelium was Sox10 negative). Ashas been reported in prior studies, mast cells stainedpositively for Sox10. The staining pattern was predom-inantly cytoplasmic, but nuclear staining was also seen.14
DISCUSSIONNonaka et al14 have characterized Sox10 as a sen-
sitive and specific marker for schwannian and melano-cytic tumors. We confirmed this conclusion on a largerexperimental data set, both in terms of numbers ofspecimens and pathologies evaluated. They reported noevidence of Sox10 expression in synovial sarcoma,rhabdomyosarcoma, and chondrosarcoma. Our resultssupport this assessment: synovial sarcoma (0/79), rhab-domyosarcoma (0/17), and chondrosarcoma (0/4). Wealso observed a strong S100 protein expression in caseswhere no Sox10 protein expression was identified otherthan non-neural, nonmelanocytic sarcoma types: Ewingsarcoma (3/14) and extraskeletal myxoid chondrosar-coma (5/11). The use of Sox10 in the place of S100 whenfaced with a spindle cell tumor of uncertain differentiationmay prevent misdiagnosis as MPNSTs. The MPNSTsevaluated in this series were diagnosed using strictcriteria and either arose in patients with NF1, or wereclearly derived from a major nerve with compatiblehistology.17
We think that the absence of Sox10 expression is ofparticular clinical use in cases of cellular schwannoma, asthis tumor can demonstrate morphologic similarities tofibrosarcoma, leiyomyosarcoma, and synovial sarcoma,particularly in cases with monophasic histology, or inneedle biopsy specimens. Synovial sarcoma in particularhas reported to express S100 in up to 30% of cases.19
Distinguishing between these entities on the basis ofmorphology can be particularly challenging on smallneedle biopsy specimens. If presented with a cellular-spindled S100-positive neoplasm, fluorescence in situhybridization to evaluate for t(X;18) or immunohisto-chemical staining with TLE1 may be necessary to excludesynovial sarcoma.17,20 In our study, we examined 79 casesof synovial sarcoma (65 cases on TMA). Of the 12
S100-positive cases (15%), 4 of these cases featuredstrong, diffuse staining. None of these cases showedSox10 positivity. In addition, no significant SOX10 gene
expression was identified in 40 cases of synovial sarcoma.Given the gene and protein expression results describedin this study, if a tumor shows strong Sox10 protein
expression, a diagnosis of synovial sarcoma is unlikely.Negative Sox10 staining in combination with TLE1expression would conversely provide strong support for adiagnosis of synovial sarcoma.17,20
Sox10 was reliably positive in both neurofibromasand schwannomas, in both cases staining the schwanniancell population. Unique to Sox10 is its exclusive nuclearlocalization in Schwann cells, which typically also dem-onstrate cytoplasmic S100 staining. Both nuclear andcytoplasmic S100 positivity is present in Schwann cellsand melanocytes, and consequently the lack of nuclearpositivity is a potential diagnostic pitfall.21
Granular cell tumors are typically benign tumors ofthe peripheral nerve characterized by monotonous ovoidto polyhedral cells with distinct cell borders and abundanteosinophillic granular cytoplasm. Sox10 showed strongand diffuse nuclear positivity in all 26 cases of granularcell tumor (Fig. 4). The granular cell tumors in this studyarose in a number of sites including skin, soft tissue, andthe aerodigestive tract (larynx and esophagus). Evidenceof the Schwann cell derivation of granular cell tumor wasin part based on S100 immunoreactivity,22 and our resultsprovide further support that granular cell tumors are ofneural crest derivation.23
All 4 cases of perineurioma in this series showed noevidence of Sox10 expression. This is consistent with thecurrent understanding that perineurium is derived fromtransformed mesenchymal tissue as opposed to neuralcrest.24 These results are also in-keeping with the tradi-tionally S100-negative immunophenotype of perineurialtumors.19
Variable staining for S100 in MPNSTs is well de-scribed in the literature.23,25,26 Our evaluation of S100gene expression in MPNST correlates with inconsistentstaining for S100 by immunohistochemistry. Sox10 geneexpression was similarly variable, though correlated withS100 expression. In our series Sox10 was a less sensitive
FIGURE 3. Immunohistochemical expression of Sox10 (x axis)and S100 (y axis) in synovial sarcoma and peripheral nervesheath tumors (discrete values randomly “jittered” to allow forvisualization of multiple cases). MPNST indicates malignantperipheral nerve sheath tumor.
TABLE 3. Sox10 and S100 Expression in Tumors ofNon-neural Crest Origin
marker than S100 in cases of MPNST. Though the ma-jority of cases of MPNST were negative for both markers(44/78, 56%), S100 was positive in 13 cases that wereSox10 negative, whereas Sox10 was positive in 3 casesthat were negative for S100. Given the 16 cases ofMPNST that were positive for only one of these markers,we suggest that a combination of these markers should beused when considering a diagnosis of MPNST in thesetting of an undifferentiated tumor.
Sox10 has previously been shown to be superior toS100 in the identification of desmoplastic melanoma.27 Ourstudy confirmed these findings, and Sox10 showed strongerexpression than S100 in 7 of 9 unique cases. In addition,Sox10 showed stronger staining than S100 in 4 of 7 cases ofclear-cell sarcoma. As is the case with S100, Sox10 pos-itivity does not distinguish MPNST from melanoma.
We did observe that Sox10 is not exclusively ex-pressed in neural crest tissue, and that it stained the my-oepthelial cells of breast tissue and myoepithelial andacinar cells of salivary gland tissue and as such is not apurely specific marker of neural crest derivation. None-theless we think that in the realm of soft-tissue neoplasmsthat Sox10 is significantly more specific than S100 forperipheral nerve sheath tumors and should be used in theplace of or along side S100 for diagnostic purposes.
REFERENCES1. Nakajima T, Watanabe S, Sato Y, et al. S-100 protein in Langerhans
cells, interdigitating reticulum cells and histiocytosis X cells. Gann.1982;73:429–432.
2. Shimada H, Newton WA Jr, Soule EH, et al. Pathologic features ofextraosseous Ewing’s sarcoma: a report from the IntergroupRhabdomyosarcoma Study. Hum Pathol. 1988;19:442–453.
3. Coindre JM, de MA, Trojani M, et al. Immunohistochemical studyof rhabdomyosarcoma. Unexpected staining with S100 protein andcytokeratin. J Pathol. 1988;155:127–132.
4. Fisher C. Synovial sarcoma. Ann Diagn Pathol. 1998;2:401–421.5. Fisher C, Schofield JB. S-100 protein positive synovial sarcoma.
Histopathology. 1991;19:375–377.
6. Paratore C, Goerich DE, Suter U, et al. Survival and glial fateacquisition of neural crest cells are regulated by an interplay betweenthe transcription factor Sox10 and extrinsic combinatorial signaling.Development. 2001;128:3949–3961.
7. Kuhlbrodt K, Herbarth B, Sock E, et al. Sox10, a novel transcrip-tional modulator in glial cells. J Neurosci. 1998;18:237–250.
8. Kelsh RN. Sorting out Sox10 functions in neural crest development.Bioessays. 2006;28:788–798.
9. Herbarth B, Pingault V, Bondurand N, et al. Mutation of the Sry-related Sox10 gene in Dominant megacolon, a mouse model forhuman Hirschsprung disease. Proc Natl Acad Sci USA. 1998;95:5161–5165.
10. Kuhlbrodt K, Schmidt C, Sock E, et al. Functional analysis ofSox10 mutations found in human Waardenburg-Hirschsprungpatients. J Biol Chem. 1998;273:23033–23038.
11. Osaka H, Hamanoue H, Yamamoto R, et al. Disrupted SOX10regulation of GJC2 transcription causes Pelizaeus-Merzbacher-likedisease. Ann Neurol. 2010;68:250–254.
12. Bondurand N, Girard M, Pingault V, et al. Human Connexin 32, agap junction protein altered in the X-linked form of Charcot-Marie-Tooth disease, is directly regulated by the transcription factorSOX10. Hum Mol Genet. 2001;10:2783–2795.
13. Pingault V, Bondurand N, Le CC, et al. The SOX10 transcriptionfactor: evaluation as a candidate gene for central and peripheralhereditary myelin disorders. J Neurol. 2001;248:496–499.
14. Nonaka D, Chiriboga L, Rubin BP. Sox10: a pan-schwannian andmelanocytic marker. Am J Surg Pathol. 2008;32:1291–1298.
15. Subramanian S, Thayanithy V, West RB, et al. Genome-widetranscriptome analyses reveal p53 inactivation mediated loss ofmiR-34a expression in malignant peripheral nerve sheath tumours.J Pathol. 2010;220:58–70.
16. Hubble J, Demeter J, Jin H, et al. Implementation of GenePatternwithin the Stanford microarray database. Nucleic Acids Res. 2009;37:D898–D901. (Database issue).
17. Terry J, Saito T, Subramanian S, et al. TLE1 as a diagnosticimmunohistochemical marker for synovial sarcoma emerging fromgene expression profiling studies. Am J Surg Pathol. 2007;31:240–246.
18. Nielsen TO, Hsu FD, O’Connell JX, et al. Tissue microarrayvalidation of epidermal growth factor receptor and SALL2 insynovial sarcoma with comparison to tumors of similar histology.Am J Pathol. 2003;163:1449–1456.
20. Jagdis A, Rubin BP, Tubbs RR, et al. Prospective evaluation ofTLE1 as a diagnostic immunohistochemical marker in synovialsarcoma. Am J Surg Pathol. 2009;33:1743–1751.
22. Armin A, Connelly EM, Rowden G. An immunoperoxidaseinvestigation of S-100 protein in granular cell myoblastomas:evidence for Schwann cell derivation. Am J Clin Pathol. 1983;79:37–44.
23. Scheithauer BW, Woodruff JM, Erlandson RA. Armed ForcesInstitute of Pathology (U.S.), Universities Associated for Researchand Education in Pathology. Tumors of the Peripheral NervousSystem. Washington, D.C: Armed Forces Institute of Pathology; 1999.
24. Parmantier E, Lynn B, Lawson D, et al. Schwann cell-derivedDesert hedgehog controls the development of peripheral nervesheaths. Neuron. 1999;23:713–724.
25. Daimaru Y, Hashimoto H, Enjoji M. Malignant peripheral nerve-sheath tumors (malignant schwannomas). An immunohistochemicalstudy of 29 cases. Am J Surg Pathol. 1985;9:434–444.
26. Weiss SW, Langloss JM, Enzinger FM. Value of S-100 protein inthe diagnosis of soft tissue tumors with particular reference tobenign and malignant Schwann cell tumors. Lab Invest. 1983;49:299–308.
27. Ramos-Herberth FI, Karamchandani J, Kim J, et al. SOX10immunostaining distinguishes desmoplastic melanoma from excisionscar. J Cutan Pathol. 2010;37:944–952.
FIGURE 4. Immunohistochemcial expression of Sox10 ingranular cell tumor (�200).
Karamchandani et al Appl Immunohistochem Mol Morphol � Volume 00, Number 00, ’’ 2012
6 | www.appliedimmunohist.com r 2012 Lippincott Williams & Wilkins
