Main

Gallager and Helwig1 initially described neurothekeomas in 1980 as benign cutaneous tumors of nerve sheath differentiation. Subsequent authors suggested neurothekeomas represented variants of dermal nerve sheath myxomas,2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12 which were initially described in 1969 by Harkin and Reed.13 The presumed relationship between neurothekeomas and nerve sheath myxomas was inferred from histopathologic similarities between these two entities and the resemblance of neurothekeomas to non-neoplastic neural structures. Additionally, some authors demonstrated that the immunohistochemical profile of neurothekeomas was compatible with cultured cells of nerve sheath origin from rats, which suggested that neurothekeomas represented cells of immature nerve sheath differentiation.14, 15, 16, 17

However, subtle morphologic differences, divergent immunohistochemical staining in human tissue and ultrastructural studies led many authors to question the relationship between neurothekeomas and nerve sheath myxomas.1, 7, 9, 11, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29 Recent work by Fetsch et al,28 using a combination of morphological, immunohistochemical and ultrastructural data, strongly argues that nerve sheath myxomas represent a myxoid variant of schwannoma. Using similar studies, Hornick and Fetsch both underscore previous suggestions that neurothekeomas represent a variant of cutaneous fibrohistiocytic tumors, which are benign fibroblastic and histiocytic tumors with a wide range of histomorphologic appearances.27, 29

To date, there have been no molecular studies examining these entities. Recent biotechnological advances have optimized conditions to improve the quality and quantity of RNA extracted from formalin-fixed paraffin-embedded tissues.30, 31, 32 These now permit molecular analysis, including microarray-based whole-genome-wide testing, of clinically small lesions from archival tissues.

We report the first microarray-based gene expression profile study of neurothekeomas, nerve sheath myxomas, schwannomas and cellular fibrous histiocytomas. We successfully used formalin-fixed paraffin-embedded material from archived diagnostic tissue. Our data further clarifies the relationship between these entities and demonstrates that clinically significant gene expression profiles can be obtained from routine biopsy material.

Materials and methods

A search for dermatopathology cases coded as ‘neurothekeoma’, ‘nerve sheath myxoma’, ‘cellular fibrohistiocytoma’ and ‘schwannoma’ was performed on the University of California, Los Angeles (UCLA), Department of Pathology database. The search was limited to cases from the previous 3 years, as early literature indicated that this period of storage did not adversely affect RNA extraction for microarray analysis in formalin-fixed paraffin-embedded blocks.30, 33, 34

All slides were re-reviewed by three authors (SS, SB and SD) and appropriately categorized based on histological and immunohistochemical criteria set forth by the Armed Forces Institute of Pathology, World Health Organization and Hornick and Fletcher's reviews of neurothekeomas and cellular fibrohistiocytic tumors (Table 1).22, 27, 28, 29, 35, 36 Exclusion criteria included cases with scant tissue remaining in the block and cases without unanimous agreement on the definitive diagnosis. Analysis was also limited to blocks of tissue with at least 40% tumor cells. In all, 28 cases were selected from Department files or retrievable from originating hospitals for cases seen in consult. These cases included patients diagnosed with myxoid, mixed or cellular neurothekeomas (nine cases), dermal nerve sheath myxomas (five cases), cellular fibrous histiocytomas (seven cases) and dermal schwannomas (seven cases; Figures 1, 2, 3, 4). Areas of the slide (∼5 mm2) with maximal tumor tissue were circled and corresponding areas were macrodissected from the tumor blocks for RNA isolation.

Table 1 Reported clinical, histological and immunological findings of neurothekeomas, cellular fibrous histiocytomas and dermal nerve sheath myxomas.27, 28, 29, 36
Full size table
Figure 1
figure 1

(a) Dermal schwannoma (SCH no. 4) with antoni A and B areas and prominent hyalinized vessels (H and E; × 20). (b) Low power showing typical circumscription and diffusely positive S100 protein staining (S100; × 4).

Full size image
Figure 2
figure 2

(a) Cutaneous cellular fibrous histiocytoma (BFH no. 2) centered within the dermis with overlying epidermal hyperplasia (H and E stain; × 20). (b) Cellular fibrous histiocytoma (BFH no. 5) with typical dermal collagen entrapment and haphazard cellular pattern (H and E stain; × 28). (c) Positive staining for factor 13a in BFH no. 5 (factor 13A; × 200).

Full size image
Figure 3
figure 3

(a) Myxoid neurothekeoma (MyxNK no. 3) centered within the dermis with broad fibrous septae (H and E; × 20). (b) Higher power showing epithelioid cytomorphology within a prominent myxoid matrix (H and E; × 200). (c) Cellular neurothekeoma (CNK no. 1) with a plexiform, whorled pattern of cells within the dermis (H and E; × 60). (d) Cellular neurothekeoma (CNK no. 6) composed of plump epithelioid cells in a vaguely whorled pattern and faintly myxoid matrix (H and E; × 100).

Full size image
Figure 4
figure 4

(a) Multilobulated and circumscribed dermal nerve sheath myxoma (NSM no. 3) with prominent fibrous bands (H and E; × 10). (b) Higher power demonstrates prominent stellate to spindled cells within abundant myxoid matrix (H and E; × 28). (c) Acral skin with a nerve sheath myxoma (NSM no. 1) demonstrating a plexiform pattern of paucicellular, spindled cells within a myxoid matrix (H and E; × 40). (d) Strong-positive staining of stellate to spindled cells with S100 protein (S100; × 40).

Full size image

A variety of histopathologic data were collected, including the presence of epidermis/adnexal structures, degree of inflammation and percent of tumor/fibrous/dermal/adipose tissue in the areas of macrodissection (Table 2). Immunoperoxidase studies using the avidin–biotin–peroxidase technique performed included S100 protein (Ventana, Tucson, AZ, USA; predilute), α-smooth muscle actin (Ventana; predilute), factor XIIIa (Calbiochem, Merck & Co., Whitehouse Station, NJ, USA; 1:30), neuron-specific enolase (Ventana; predilute), and NKI/C3 (BioGenex, Fremont, CA, USA; 1:1600).

Table 2 Summary of histologic features of macrodissected areas selected for gene microarray analysis
Full size table

All RNA extraction and microarray experiments were performed at the UCLA Department of Pathology Clinical Microarray Core Laboratory under the direction of one of the authors (XL). Total RNA was isolated using the Ambion RecoverAll (Applied Biosystems/Ambion, Austin, TX, USA) kit according to the manufacturer's instructions. RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA) and purity/concentration was determined using a NanoDrop 8000 (NanoDrop Products, Wilmington, DE, USA). Microarray targets were prepared using NuGEN WT-Ovation Formalin-Fixed Paraffin-Embedded RNA Amplification System and FL-Ovation cDNA Biotin Module V2 (NuGEN Technologies, San Carlos, CA, USA) and then hybridized to the Affymetrix GeneChip U133 Plus 2.0 Array (Affymetrix), all according to manufacturers' instructions. The arrays were washed and stained with streptavidin phycoerythrin in Affymetrix Fluidics Station 450 using the Affymetrix GeneChip protocol, and then scanned using an Affymetrix GeneChip Scanner 3000.

The acquisition and initial quantification of array images were conducted using the AGCC software (Affymetrix). The subsequent data analyses were performed using Partek Genomics Suite Version 6.4 (Partek, St Louis, MO, USA). Differentially expressed genes were selected at ≥threefold and P<0.005. Cluster analysis and principal component analysis were conducted with Partek default settings. Biofunctional analysis was performed using Ingenuity Pathways Analysis Software Version 7.6 (Ingenuity Systems, Redwood City, CA, USA).

Results

The clinical and immunohistochemical features of our cases are similar to those recently described in the literature (Tables 3 and 4).27, 28, 29

Table 3 Summary of clinical findings of cases selected for microarray analysis
Full size table
Table 4 Summary of positive immunohistochemical findings in cases selected for microarray analysis
Full size table

The gene expression profiles of these four histological entities are shown in Figures 5, 6, 7. Greater numbers of differentially expressed genes between nerve sheath myxomas and neurothekeomas suggest that they are not related (Figure 5). This data, in conjunction with cluster analysis (Figure 6) and principal component analysis (Figure 7), demonstrate two molecular groups: a nerve sheath myxoma/schwannoma group and a neurothekeoma/cellular fibrous histiocytoma group.

Figure 5
figure 5

Number of genes differentially expressed between selected cutaneous spindle cell lesions (>threefold change and P-value <0.005).

Full size image
Figure 6
figure 6

Cluster map constructed using 452 differentially expressed genes between neurothekeomas and nerve sheath myxomas. The nerve sheath myxomas and schwannomas are clustered into one group, whereas the neurothekeomas and cellular fibrous histiocytomas fall into another group (>threefold change; P-value <0.005).

Full size image
Figure 7
figure 7

Principal component analysis using 452 differentially expressed genes between neurothekeomas and nerve sheath myxomas demonstrates one major group composed of schwannomas and nerve sheath myxomas and another major group composed of neurothekeomas and cellular fibrous histiocytomas (>threefold change; P-value <0.005).

Full size image

Genes upregulated in nerve sheath myxomas and schwannomas compared with both neurothekeomas and cellular fibrous histiocytomas primarily included those involved in peripheral nerve sheath maintenance or development (Table 5), with the most statistically significant differentially gene expressed being S100B.37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 The principal differentially expressed genes upregulated in neurothekeomas and cellular fibrous histiocytomas over nerve sheath myxomas and schwannomas consisted of those encoding various metalloproteinases and glycoproteins. These genes are involved in extracellular matrix growth and remodeling, and cell adhesion in fibroblasts and macrophages (Table 6).49, 50, 51, 52, 53, 54, 55, 56

Table 5 Upregulated genes in dermal nerve sheath myxomas (NSMs) and dermal schwannomas (SCHs) vs neurothekeomas (NKs) and cellular fibrous histiocytomas (BFHs).37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48
Full size table
Table 6 Upregulated genes in neurothekeomas (NKs) and cellular fibrous histiocytomas (BFHs) vs dermal nerve sheath myxomas (NSMs) and dermal schwannomas (SCHs).49, 50, 51, 52, 53, 54, 55, 56
Full size table

There were relatively few statistically significant genes differentially expressed between neurothekeomas and cellular fibrous histiocyomas, and between nerve sheath myxomas and schwannomas. For neurothekeomas and cellular fibrous histiocytomas, these genes primarily included various collagen proteins. Between nerve sheath myxomas and schwannomas, differentially expressed genes included those encoding neuronal cell intercellular signaling, collagen and cell adhesion molecules.

One outlier (nerve sheath myxoma no. 1) did not cluster with the schwannomas/nerve sheath myxomas group, despite typical histological and immunohistochemical features. Upon analysis of the cluster map, this sample demonstrated broadly decreased expression of almost all genes compared with the other samples (Figure 6). Outside processing and storage of the block, technical limitations/artifact inherent to microarray analysis of formalin-fixed paraffin-embedded tissue or a higher proportion of fibrous tissue in this case (60% vs an overall average of 35% in the other cases) may account for this outcome (Table 2).

Discussion

Whole-genome microarray and transcription profiling analysis simultaneously interrogates tens of thousands of genes, and thus provides sophisticated molecular images of tissues. However, until recently, this technology could not be applied to formalin-fixed paraffin-embedded tissue blocks. Formalin fixation causes chemical cross-linking of nucleic acids to each other and to proteins, as well as fragmentation of nucleic acids. Paraffin embedding results in further mechanical stresses to RNA. Cumulatively, these lead to suboptimal nucleic acid quality and hampered reverse transcription and amplification reactions.57, 58, 59, 60 Thus, in the past, gene expression profile studies were restricted to tissues with higher intrinsic nucleic acid quality, such as cell lines, fresh blood or fresh frozen tissue.61, 62, 63, 64, 65, 66

Recent biotechnological advances have improved RNA isolation and amplification techniques, resulting in successful use of formalin-fixed paraffin-embedded tissues to identify gene signatures in human tumors. These advances include newer formalin-fixed paraffin-embedded tissue extraction protocols, cDNA-mediated annealing, selection, extension and ligation, the addition of random hexamer priming to oligo(dT) priming and newly developed array platforms with redesigned probes.30, 31, 32, 33, 34, 57, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82 Although sensitivity is low, high specificity and positive predictive value suggest that transcript detection is reliable from formalin-fixed paraffin-embedded tissue.33

Although RNA isolation techniques from formalin-fixed paraffin-embedded tissues have improved, problems remain, largely because of the greater fragmented nature of RNA in formalin-fixed paraffin-embedded tissue.30, 32, 75, 78 Magnitudes of differential expression are often higher in formalin-fixed paraffin-embedded tissue than in fresh frozen samples, because of the increased background noise; this increases the odds of false-positive results.30, 83 Despite these concerns, several studies have shown comparable biological information between genome-wide microarray analysis from matched fresh frozen and formalin-fixed paraffin-embedded tissue.30, 32, 72, 74, 75, 77, 78

The use of formalin-fixed paraffin-embedded rather than fresh frozen tissue clearly allows more practical applications of gene microarray analysis. The National Cancer Institute estimates that 85% of all cancer surgeries are performed in community hospitals, which historically have not had the resources for procuring snap frozen tumor samples.84 Small tumors often must be completely paraffin embedded for diagnostic studies. In some cases, the diagnosis is not apparent until microscopic examination, at which time fresh tissue is unavailable. Formalin-fixed paraffin-embedded tissue allows for improved correlation between histopathologic findings and molecular signatures, allowing for better understanding of gene expression profiles. Finally, use of archival formalin-fixed paraffin-embedded tissue allows for large-scale retrospective studies; as clinical and outcome data can be associated with these tissues, this permits more clinically relevant research.

We are the first to report gene expression profiles for nerve sheath myxomas, schwannomas, neurothekeomas and cellular fibrous histiocytomas. Similar grouping of schwannomas and nerve sheath myxomas by principal component analysis and hierarchical clustering strongly suggests that these lesions are closely related. Functional gene analysis of the microarray profiles revealed that many of the highest differentially expressed genes in schwannomas/nerve sheath myxomas compared with neurothekeomas/cellular fibrous histiocytomas involve peripheral nerve sheath development and maintenance. This data strongly supports previous histomorphologic studies, indicating that nerve sheath myxomas and schwannomas are related neoplasms of peripheral nerve sheath origin and, despite some similar histological features, are unrelated to neurothekeomas.27, 28, 29

Neurothekeomas expressed genes that primarily encoded glycoproteins and metalloproteinases, often from macrophages and fibroblasts involved in matrix growth and remodeling, angiogenesis and mesenchymal cell differentiation.49, 50, 51, 52, 53, 54, 55, 56 The S100B gene was the most significantly differentially expressed gene between nerve sheath myxomas and neurothekeomas. This further provides molecular support for recent work reporting negative S100 protein staining by immunohistochemistry in neurothekeomas of all subtypes.27, 29 Interestingly, our microarray data demonstrated increased podoplanin gene expression in neurothekeomas over nerve sheath myxomas. This molecular finding correlates with recent immunohistochemical observations that podoplanin (D2-40) protein expression is seen in neurothekeomas.85

Our gene expression profile data for neurothekeomas and cellular fibrous histiocytomas support previous arguments that these are related neoplasms.1, 7, 9, 11, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 29, 86 Both groups showed similar expression of genes involved in matrix remodeling, and growth involving fibroblasts and histiocytes. Analysis of the differential gene expression data between neurothekeomas and cellular fibrous histiocytomas was less instructive. The vast majority of the top differentially expressed genes between these tumors encoded various uncharacterized loci or nonspecific immunoglobulin loci, zinc fingers, chemokine receptors or proteins involved in intracellular transport. There is no definite known relationship between these gene products and cellular fibrous histiocytomas in the literature. Thus, the significance of these findings is unclear. A small number of differences in expression of genes involved in connective tissue matrix and tissue remodeling were noted, and it is possible that this is secondary to the inclusion of both myxoid and mixed types of neurothekeomas.

Our data does not resolve the question of whether neurothekeomas are simply dermal variants of the rarer plexiform variant of fibrohistiocytic tumors. Neurothekeomas and plexiform fibrous histiocytomas share nearly identical demographic and anatomic distributions, as well as many overlapping histological features.7, 27, 29, 87, 88, 89 There are no reliable immunohistochemical markers that distinguish these lesions.29, 89 The majority of plexiform fibrous histiocytomas are clinically benign, similar to neurothekeomas. Taken together, all these features have led some authors to suggest that neurothekeomas and plexiform fibrohistiocytomas are the same lesion.89, 90 It has been suggested that the deeper subcutaneous location of plexiform fibrous histiocytomas is responsible for the higher rates of local recurrence and rare instances of metastases.24, 87, 88, 89, 90 With one exception, the cases of plexiform fibrous histiocytomas seen at our institution during the study were outside consultations for which we were not able to obtain material for testing. Molecular profiling by gene microarray analysis would be a critically helpful adjunct in determining the relationship between these entities.

Our molecular data provides critical insights into the relationship between these entities. However, practically, diagnosis of these entities will rely on light microscopy and immunohistochemistry for the present time. Gene microarray studies are expensive compared with standard pathology techniques. Recent large histological and immunohistochemical studies of these entities have substantially clarified the microscopic features of each.27, 28, 29 Furthermore, the clinical implications of misdiagnosis are minimal, as the vast majority of these are clinically benign. Collectively, these features do not support use of very expensive microarray technology for routine diagnosis.

In summary, our gene expression array data strongly supports that nerve sheath myxomas are of peripheral nerve sheath origin and are distinct neoplasms from neurothekeomas. Furthermore, our data shows many similarities in gene expression between neurothekeomas and cellular fibrous histiocytomas, supporting recent arguments that these entities may in fact be related. Further studies to examine the molecular profile of fibrous histiocytomas and neurothekeomas are indicated, in particular the rare plexiform variant of fibrous histiocytomas, which shows many histological and clinical similarities to neurothekeomas. Additionally, our data shows that advances in RNA isolation, amplification, and labeling protocols allow successful gene microarray analysis on routine clinical biopsy samples that are processed as formalin-fixed paraffin-embedded tissues.