• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjpatholAmerican Journal of Pathology For AuthorsAmerican Journal of Pathology SubscribeAmerican Journal of Pathology SearchAmerican Journal of Pathology Current IssueAmerican Journal of Pathology About the JournalAmerican Journal of Pathology
Am J Pathol. Dec 2003; 163(6): 2383–2395.
PMCID: PMC1892373

Gene Expression Patterns and Gene Copy Number Changes in Dermatofibrosarcoma Protuberans

Abstract

Dermatofibrosarcoma protuberans (DFSP) is an aggressive spindle cell neoplasm. It is associated with the chromosomal translocation, t(17:22), which fuses the COL1A1 and PDGFβ genes. We determined the characteristic gene expression profile of DFSP and characterized DNA copy number changes in DFSP by array-based comparative genomic hybridization (array CGH). Fresh frozen and formalin-fixed, paraffin-embedded samples of DFSP were analyzed by array CGH (four cases) and DNA microarray analysis of global gene expression (nine cases). The nine DFSPs were readily distinguished from 27 other diverse soft tissue tumors based on their gene expression patterns. Genes characteristically expressed in the DFSPs included PDGFβ and its receptor, PDGFRB, APOD, MEOX1, PLA2R, and PRKCA. Array CGH of DNA extracted either from frozen tumor samples or from paraffin blocks yielded equivalent results. Large areas of chromosomes 17q and 22q, bounded by COL1A1 and PDGFβ, respectively, were amplified in DFSP. Expression of genes in the amplified regions was significantly elevated. Our data shows that: 1) DFSP has a distinctive gene expression profile; 2) array CGH can be applied successfully to frozen or formalin-fixed, paraffin-embedded tumor samples; 3) a characteristic amplification of sequences from chromosomes 17q and 22q, demarcated by the COL1A1 and PDGFβ genes, respectively, was associated with elevated expression of the amplified genes.

Dermatofibrosarcoma protuberans (DFSP) is an uncommon soft tissue neoplasm of young adults with a propensity for locally aggressive growth. This tumor typically arises in the subcutaneous tissues of the trunk and proximal extremities and is composed of uniform spindle cells of uncertain histogenesis. Most DFSPs are cured by wide surgical excision, but local recurrence occurs in approximately 20% of cases. 1 In some instances, radiotherapy can be of benefit. 2 Death due to metastatic disease is very rare (<5%), 1 and there have been isolated reports of apparently successful treatments for metastatic disease with imatinib. 3,4 Virtually all cases of DFSP have a translocation that involves chromosomes 17 and 22, resulting in fusion of the collagen type I α I (COL1A1) and platelet-derived growth factor β (PDGFβ) genes. 5,6 The regulatory sequences of COL1A1 lead to increased expression of a fusion transcript that is processed to wild-type PDGFβ, which has transforming activity when expressed at high levels. 7,8 The translocation is most commonly located on a ring chromosome with varying amounts of DNA derived from chromosomes 17 and 22, and sometimes also from other chromosomes, 9 associated with amplification. To date, the areas of amplification have only been crudely mapped using conventional genomic hybridization. 10-12 Fusion of COL1A1 and PDGFβ is also encountered in giant cell fibroblastoma, a rare tumor typically seen in children. Currently these two tumors are thought to be variants of the same lesion. 5,6

Expression profiling using cDNA microarrays has resulted in molecular subclassification of a wide variety of tumors, including lymphomas, 13 lung carcinomas, 14 and breast carcinomas. 15 Gene expression profiling with cDNA microarrays has also proven useful as an aid in the differential diagnosis of mesenchymal tumors, where morphological features show significant overlap. 16-20 cDNA microarrays have been used not only for gene expression studies but also for high-resolution comparative genomic hybridization (array CGH). By using the same microarrays for both array CGH and cDNA expression studies on the same tumor samples, Pollack et al 21 found that changes in gene copy number were consistently associated with corresponding changes in expression of the affected genes. In the case of breast carcinoma, the majority of amplified genes showed increased expression, and alterations of gene copy number accounted for a significant component of the altered gene expression observed in an individual tumor.

In this study we used DNA microarrays to profile gene expression in DFSP and identified a large number of genes with consistently high expression in these tumors. Several potential differential diagnostic markers were identified, one of which (APOD) is described in detail in a separate report (West et al, in preparation). In addition, we characterized DNA copy number alterations in DFSP at a high resolution and correlated those changes with expression levels of affected genes. In the course of this study we found that array CGH could be used successfully to analyze formalin-fixed, paraffin-embedded tissue, circumventing the need for fresh frozen tissue, and making possible array CGH studies on samples from surgical pathology archives.

Materials and Methods

Case Selection

Nine cases of DFSP were used for this study, obtained from the Departments of Pathology of Stanford University Medical Center, Vancouver Hospital and Health Sciences Centre, University of Washington Medical Center, and Cleveland Clinic Foundation. Fresh frozen tissue, stored at −80°C was available in each case. The Institutional Review Board at Stanford University Medical Center approved the study. The diagnosis of DFSP was based on the light microscopic appearance of the tumors, with confirmatory immunohistochemical staining (CD34) performed in each case. All but one of the DFSPs reacted for CD34, consistent with known staining patterns. For gene expression studies, a total of 36 soft tissue tumors (STT) were studied, including the nine cases of DFSP. Non-DFSPs studied were leiomyosarcoma (eight cases), malignant fibrous histiocytoma (six cases), of which two cases could be subclassified as myxofibrosarcoma, gastrointestinal stromal tumor (GIST, five cases), synovial sarcoma (six cases), nodular fasciitis (one case), and epithelioid fibrous histiocytoma (one case). One of the GIST cases (STT1823) was a recurrence from a prior lesion (STT094). The nine DFSP cases, the two myxofibrosarcomas, the nodular fasciitis, the fibrous histiocytoma, and the recurrence of the GIST lesion are new to this report. The other lesions functioned as comparison for gene expression level determination in DFSP and were previously published as part of a prior study. 16

Array CGH studies were performed using fresh frozen tissue on four of the nine DFSP cases, and also using paraffin embedded tissue for three of the four cases. For comparison, we performed array CGH on four cases of solitary fibrous tumor (SFT), a soft tissue tumor known to have a relatively simple karyotype 22 that rarely harbors chromosome abnormalities in the 17q or 22q regions. 23 DNA was also isolated from paraffin- embedded material of six fibrous histiocytoma specimens for aCGH studies. In each case the diagnosis was made based on conventional light microscopic examination, with appropriate adjuvant immunohistochemical studies.

Histology and Immunohistochemistry

All cases were reviewed by two pathologists (B.R. and M.v.d.R.). For immunohistochemistry, sections were stained with CD34 (Becton Dickinson, Mountain View, CA) and APOD (Novocastra, Newcastle-on-Tyne, United Kingdom) using the Envision detection system (DAKO, Carpinteria, CA).

cDNA Expression Microarray Analysis

The cDNA microarrays used in this study included about 28,000 unique characterized genes or ESTs represented by a total of 41,859 unique cDNAs printed on glass slides by the Stanford Functional Genomics Facility (http://www.microarray.org/sfgf/jsp/home.jsp). The details of the construction of these arrays were described previously. 15 Preparation of tumor mRNA, labeling, and hybridization were performed as described in an earlier publication. 15 Briefly, after confirmation of the presence of viable tumor by frozen section, tissue was homogenized in Trizol reagent (Invitrogen, Carlsbad, CA) and total RNA was prepared; mRNA was then isolated using the FastTrack 2.0 method following the manufacturer’s protocol. Preparation of Cy-3-dUTP (green fluorescent) labeled cDNA from 2 μg of reference mRNA and Cy-5-dUTP (red fluorescent) labeled cDNA from 2 μg of each tumor specimen mRNA, microarray hybridization and subsequent analysis was performed as described. 15 The reference mRNA was isolated from a pool of 11 cell lines. 15 After washing, the microarrays were scanned on a GenePix 4000 microarray scanner (Axon Instruments, Foster City, CA) and, after normalization of fluorescence intensities to control for experimental variation, fluorescence ratios (tumor/reference) were calculated using GenePix software. The primary data tables and the image files are freely available from the Stanford Microarray Database 24 (http://genome-www4.stanford.edu/MicroArray/SMD/). Data were selected using the following criteria: 1) only cDNA spots with a ratio of signal over background of at least 1.5 in either the Cy3 or Cy5 channel were included; 2) genes were included for further analysis only if the corresponding cDNA spots provided data that passed criterion 1 on at least 29 of 36 arrays (80% good data); 3) to focus on genes with high variations in expression in these tumors, we selected genes whose expression level differed by at least fourfold in at least two specimens from that gene’s geometric mean expression level across all 36 specimens. Hierarchical clustering analysis 25 and significance analysis of microarrays (SAM) 26 were then performed as described previously. 16

Array Comparative Genomic Hybridization

Of the over 41,000 cDNA sequences represented on the microarrays used for this study, the chromosomal localization is known for 35,151 distinct mapped cDNAs, which represent 24,540 different Unigene clusters and 3225 cDNAs not yet represented in Unigene clusters. Tumor DNA from frozen or formalin-fixed, paraffin-embedded tissue and reference DNA (normal gender-matched human leukocytes) were extracted; see protocols on accompanying website (http://genome-www.stanford.edu/DFSP/). Frozen tumor DNA and reference DNA were digested with DpnII before further processing. Gel electrophoresis of digested and non-digested DNA isolated from formalin-fixed, paraffin-embedded tissue was run to determine DNA fragment size. Labeling of DNA isolated from tumor samples, after light-microscopic confirmation of the presence of non-necrotic tumor, was performed as described previously 21 (http://cmgm.stanford.edu/pbrown/protocols/index.html). Briefly, 2 to 4 μg of tumor DNA was fluorescently labeled (Cy5) in a volume of 50 microliters, mixed with reference DNA labeled with Cy3, and hybridized overnight to the array. After washing, the array slides were scanned on a GenePix Scanner (Axon Instruments) and fluorescence ratios (test/control) calculated using GenePix software. Only cDNA spots with a ratio of signal over background of at least 1.5 in the Cy3 channel were included in further analysis. Chromosomal localization of the mapped genes was assigned as described previously 21 and is based on Goldenpath data from June 28, 2002. For CGH data the copy number for each locus was based on a moving average of the five nearest cDNA clones centered on that locus. 21 For chromosome 17 and 22, 1224 and 527 genes were selected respectively, and the same genes were used to display centered expression levels for four DFSPs and three SFTs in Figure 4 [triangle] .

Figure 4.
Comparison of aCGH data of chromosomes 17 and 22 in four DFSP with mRNA levels of same genes relative to aCGH data and expression levels in SFT.

Statistical Analysis

For statistical tests the SPSS version 10.0 statistical software package (SPSS, Inc., Chicago, IL) was used.

Results

Histology and Immunohistochemistry

The nine cases of DFSP studied here were all typical DFSP based on their histological features and/or immunohistochemical staining profiles (Figure 1) [triangle] . Clinical and immunohistochemical features of the DFSPs are shown in Table 1 [triangle] . The 27 cases of non-DFSP tumors chosen for comparison by gene expression profiling included four tumors with histological features similar to those of DFSP. One, STT169, (Figure 1) [triangle] was a cellular fibrous histiocytoma, the lesion with which DFSP is most often confused. The distinction between these two neoplasms is critical, as histiocytomas are almost invariably benign and require less aggressive treatment. Three other lesions, two myxofibrosarcomas (STT625, STT640) and one case of nodular fasciitis (STT604) can be confused histologically with the myxoid variant of DFSP. The clinical and immunohistochemical features of the four DFSP mimics are also shown in Table 1 [triangle] . The histology of the nine DFSPs, and the four lesions in its differential diagnosis can be found on the supplemental Figure 1 [triangle] (http://genome-www.stanford.edu/DFSP/). The histology of the remaining 23 STT can be found on the supplemental website of an earlier publication (http://genome-www.stanford.edu/sarcoma/).

Figure 1.
Hematoxylin and eosin staining of a representative case of DFSP (STT491) (A), and a cellular fibrous histiocytoma (STT169) (D), with confirmatory immunohistochemical staining for CD34 (B and E), and APOD (C and F). The DFSP is positive for CD34 (B), while ...
Table 1.
Clinical Data of All DFSP Cases and of the Four DFSP Mimics

Gene Expression Data

This report focuses on variation among these tumors in expression of 4687 of the 28,597 genes or ESTs represented on our microarrays. These were genes for which we consistently were able to obtain technically adequate measurements and where expression levels varied substantially among the 36 samples analyzed (Figure 2A) [triangle] . We first used an unsupervised hierarchical clustering method to highlight groups of tumors with similarities in global gene expression patterns. Hierarchical clustering analysis organizes genes into groups with similar expression patterns, facilitating recognition of functional themes in the expression patterns, and similarly organizes the soft tissue neoplasms into distinct groups. The results are displayed in Figure 2A [triangle] in the form of a table in which the expression pattern of each gene in each tumor is represented using a color key and dendrograms are used to represent relationship among the tumors. The nine cases of DFSP clustered closely together based on their global gene expression patterns, indicating that they were closely related to each other and significantly different from the other tumors profiled (Figure 2B) [triangle] . The 4 cases representing alternative diagnoses in the differential diagnosis of DFSP (STT640, STT625, STT169, STT604) clustered separately from the cases of DFSP.

Figure 2.
Hierarchical cluster analysis of gene expression profiles of 36 tumors, including nine DFSPs. Each row represents the relative level of expression for a single gene, centered at the geometric mean of its expression level across the 36 samples. Each column ...

DFSP specimens were distinguished from the other neoplasms by a large cluster of 465 genes whose patterns of variation in expression among these tumors were highly correlated (correlation coefficient with groupwise mean pattern of >0.63, see black bar, Figure 2A [triangle] ). Within this cluster is a subcluster of 98 genes including the PDGFβ and PDGFRB genes, shown in Figure 2C [triangle] . The complete dataset is available in a searchable format on the accompanying website (http://genome-www.stanford.edu/DFSP/), where the 465 genes are also listed as Web Table 2 [triangle] . The DFSPs expressed high levels of PDGFβ transcripts, consistent with the known significance for this tumor of the t(17;22) involving PDGFβ. Only one of the nine DFSPs, STT3053, did not show an elevated level of PDGFβ expression. The lack of PDGFβ mRNA expression as seen by gene array analysis in this case (using two separate cDNAs for PDGFβ mRNA) was confirmed by in situ hybridization studies (data not shown). Despite the lack of elevated PDGFβ expression in case STT3053 this is still considered a classic example of DFSP, based on histology and immunohistochemistry. The histology is available on the accompanying website and the tumor reacted for CD34 (see Table 1 [triangle] ). The expression pattern of several thousands of genes, shared by case STT3053 and the other DFSP cases further supports this. Importantly, the receptor for PDGFβ, PDGFRB, was also highly expressed in the group of DFSPs. High levels of CD34 transcripts and low expression of CD68 in DFSPs compared to non-DFSPs, are consistent with the known immunohistochemical profile of these tumors (Figure 2, D and E) [triangle] . 27-29 A cellular fibrous histiocytoma (case STT169) unexpectedly showed moderately high levels of CD34 mRNA. However, immunohistochemical staining of this tumor shows that CD34 protein is expressed by endothelial cells, rather than the tumor cells themselves (Figure 1E) [triangle] . Although APOD mRNA was highly expressed in both DFSPs and the cellular fibrous histiocytoma (STT169), immunohistochemistry for APOD protein showed expression only in the DFSPs but not in the histiocytoma (Figure 1F) [triangle] . The full explanation for this discrepancy is unknown. However posttranslational modifications such as glycosylation may account for this effect. In addition in several instances, discrepancies between mRNA levels and protein levels have been described, for example in bcl2 expression. 30

Table 2.
First 50 Genes with Highest Differential Expression in DFSP According to SAM Analysis

We used a supervised analytical method, SAM (Significance Analysis of Microarrays), as an alternative way to search for differentially expressed genes that distinguish DFSP from the other sarcomas. 26 A list of the 259 genes (288 cDNAs) most consistently expressed at higher levels in DFSP, as determined by SAM, is available on the accompanying website (supplementary Table 2 [triangle] , http://genome-www.stanford.edu/DFSP/) with their chromosomal location. The top 50 genes, most responsible for the distinction between DFSP and the other soft tissue lesions, is shown in Table 2 [triangle] . These genes representing candidate diagnostic markers for DFSP, include APOD, which stained all nine DFSPs in this study and failed to react with the single histiocytoma (despite high levels of mRNA in that lesion). The utility of APOD as an immunohistochemical marker is described in detail in a separate report (West et al, in preparation).

Array Comparative Genomic Hybridization

Array CGH analysis of four DFSP cases revealed abnormalities of chromosome arm 17q and 22q in all cases (Figure 3) [triangle] , corresponding to the most commonly identified cytogenetic abnormality in DFSP, a ring chromosome containing portions of chromosomes 17q and 22q, r(17:22). 23 This ring chromosome results in fusion of the COL1A1 and PDGFβ genes. In contrast, 4 cases of solitary fibrous tumors, showed no significant areas of amplification or deletion. The areas of amplification of chromosomes 17 and 22 in DFSP start at the PDGFβ and collagen 1A1 genes, such that the PDGFβ and collagen 1A1 constitute one of two borders in a contiguous amplified segment of 22q, and 17q respectively, (Figure 3 [triangle] , Panel A, B). One or more additional genetic abnormalities were identified in every case and gains and losses at some loci were seen in more than one case (see supplemental Figure 3 [triangle] ). For instance, we observed recurrent gains of chromosome 5 (STT1971, STT1984), chromosome 7p (STT154, STT1971), chromosome 8 (STT154, STT1971, STT1984), chromosome 18 (STT1971, STT1984), and chromosome 21 (STT1971, 1984). In addition, loss of large parts of chromosome 19 was observed in STT1971 and STT1984. Many of the apparent losses and gains affected small segments of chromosomes and would not be expected to be detectable cytogenetically or by conventional comparative genomic hybridization.

Figure 3.
Pairwise comparison of array CGH on three cases of DFSP with DNA isolated from fresh frozen tumor and from formalin-fixed, paraffin-embedded tumor showing nearly identical regions of amplification on chromosomes 17 and 22. In contrast, five cases of solitary ...

Array CGH Analysis of Frozen versus Formalin-Fixed, Paraffin-Embedded Tissue

To test the feasibility of array CGH analysis of formalin-fixed, paraffin-embedded tumor samples we analyzed both fresh frozen and formalin-fixed, paraffin-embedded material from the same tumor, in three cases of DFSP. In each case the essential results were very similar (Figure 3) [triangle] . In an attempt to demonstrate the potential clinical usefulness of aCGH in the differential diagnosis of other lesions, we performed aCGH on five cases of fibrous histiocytoma, for which only paraffin material was available. All cases showed a lack of significant areas of gene amplification by this technique, including in the areas of chromosome 17 and 22 affected in DFSP (Figure 5 [triangle] and accompanying website).

Figure 5.
Array-based DNA copy number profiles of DFSP STT154 (A) and fibrous histiocytoma 089 (B). Individual cDNAs are arranged according to their known location along human chromosomes, to produce a representation of tumor DNA copy number changes mapped onto ...

Correlation of Array CGH and Gene Expression Data

Comparison of array CGH and gene expression data could be done in four cases of DFSP. For these cases the copy number and gene expression were compared for the 27,765 genes for which the chromosomal location has been mapped. A correlation was observed between increased gene copy number and increased gene expression. This influence of DNA copy number on gene expression is evident in the colored map representations of DNA copy number and mRNA levels for 1224 genes on chromosome 17 and 527 genes on chromosome 22 (Figure 4) [triangle] . From visually inspecting Figure 4 [triangle] it appears that there is a considerable difference in levels of expression between the regions of chromosomes 17 and 22, bounded by the COL1A1 and PDGFβ genes, respectively. To test this observation more formally, using the difference in average mRNA levels between the DFSP and SFT specimens, we considered a change-point model 31 with a single change in average expression level occurring on chromosome 17 and another change on chromosome 22. The maximum likelihood estimate for the change-point on chromosome 17, based solely on gene expression measurements, was the gene PHB (prohibitin Hs.75323, AA055656, chromosome 17, start 47192492), 12 mapped genes toward the p terminus from COL1A1 (Hs.172928, R48843, chromosome 17, start 47864928). On chromosome 22, the maximum likelihood estimate of the change-point was MGAT3 (mannosyl (β-1,4-)-glycoprotein β-1,4-N-acetylglucosaminyltransferase Hs.348978, AA421473, chromosome 22, start 36501033),five mapped genes toward the q terminus from PDGFβ (Hs.1976, W72000, chromosome 22, start 36234216). A 95% confidence region for the location of the change-point on chromosome 17 barely excluded the COL1A1 gene; a 95% confidence region for the location of the change-point on chromosome 22 included the PDGFβ gene. Despite the significant correlation between amplification levels and expression levels it is clear that there remain many examples of amplified genes that are not overexpressed.

Of the 259 genes whose elevated expression was most consistently associated with a diagnosis of DFSP, 12% (31 genes) were localized to the regions of recurrent amplification on chromosomes 17 (23 genes) and 22 (8 genes), as determined by array CGH (Table 3) [triangle] , although the amplified segments of chromosomes 17 and 22 contain only 3% of all mapped genes (P < 0.001, Binomial test-exact). Although genes in regions of amplification tended to be more highly expressed, many of the genes that distinguish DFSP from other tumors are in regions without apparent changes in DNA copy number.

Table 3.
31 Genes Amplified on chr17 or 22 and Differentially and Highly Expressed in DFSP

The full data sets for the array CGH and gene expression microarray studies can be accessed through the Stanford Microarray Database website and through http://genome-www.stanford.edu/DFSP/.

Discussion

Dermatofibrosarcoma protuberans was first described in 1924, but the cell of origin for this lesion remains unknown, and cases of DFSP continue to pose difficulties in differential diagnosis. 32 The uncertainties result from the tumor’s histology, consisting of bland spindle cells that lack distinctive growth or cytological features, and the lack of specific immunohistochemical markers for this tumor. The correct diagnosis is critical, as DFSPs are locally aggressive tumors that need a wide surgical excision, whereas other lesions in the differential diagnosis of DFSP, such as fibrous histiocytoma, can be treated less aggressively. The local relapse rate of DFSP is around 20%. Relapses are often due to incomplete surgical resection. 1

Gene expression profiling has been applied to a wide range of tumors and has repeatedly been successful in recapitulating established histological classifications, 13-16 and in identifying clinically relevant new subtypes of tumors, when routine diagnostic approaches have failed to do so. 13-15,33-38

In the case of mesenchymal tumors, we have previously shown that some types of sarcomas, such as synovial sarcomas, neural tumors, and gastrointestinal stromal tumors, are very homogenous, based on their gene expression profile. In contrast, others, such as liposarcomas, leiomyosarcomas, or malignant fibrous histiocytomas, show considerable heterogeneity and often overlapping molecular features. 16

The nine cases of DFSP in this study clustered closely together relative to a diverse set of non-DFSP sarcomas based on their gene expression patterns. These tumors were clearly separable from histologically similar STT by the expression of a distinct set of genes, including PDGFβ, one of the fusion partners of the COL1A1:PDGFβ translocation. This finding reemphasizes the important role that the fusion product is thought to play in the tumorigenesis of DFSP. In addition, PDGFBR, the receptor for PDGFβ, was highly expressed in DFSP, suggesting that an autocrine stimulatory loop may contribute to the tumorigenesis. In a previous study we had made a similar observation for synovial sarcoma, where SSX, one of the fusion partners of the t(X;18) SYT:SSX translocation, was present in the cluster of relatively highly expressed genes that distinguished synovial sarcoma from the other STT. 16

To investigate the potential utility of gene expression profiles in the differential diagnosis of DFSP, we also analyzed in this study four tumors that can be misdiagnosed as DFSP. The DNA microarray analysis clearly showed that these tumors were molecularly distinct from DFSP.

cDNA microarray analysis provides an efficient way to search for new diagnostic markers for routine immunohistochemistry. CD34 is a commonly used marker in the diagnosis of DFSP but, like many single markers, it is neither completely sensitive nor specific. 39 The characteristic gene expression pattern of DFSP included several new candidate diagnostic markers, including APOD, EGR2, PLA2R1, and MEOX1. Antibodies against APOD stained tumor cells of all DFSPs in this study, but did not stain the fibrous histiocytoma, despite the high levels of APOD mRNA detected in this tumor. A more detailed description of the reactivity of the APOD antibodies will be described elsewhere (West et al, manuscript in preparation).

Several genes that are normally expressed during early embryogenesis and involved in somitogenesis, including MEOX1, HEY1, and HEYL, 40-44 were characteristically expressed in DFSP. During murine embryogenesis, MEOX1 is expressed in the dermomyotome 40-42 and in sclerotomally derived cells and muscle cells that form the body wall, 41 perhaps paralleling the predominant localization of DFSPs in the dermis of the trunk, and suggesting the possibility that DFSPs originate from these early embryonic mesenchymal cells. In a genome-wide gene expression study of normal human vasculature, HEY1 and HEYL were found to be relatively highly expressed in the microvascular endothelial cells of skin, dermis, lung, and intestine (Jen-Tsan Chi, manuscript in preparation). Preliminary studies have shown that antibodies against HEY1 react with DFSP tumor cells (data not shown).

DFSPs have already been characterized by several cytogenetic studies. 23 The presence of a ring chromosome has been the most consistent genetic abnormality identified in DFSP, being present in more than 75% of cases, sometimes as the only detectable cytogenetic abnormality. 23,45,46 By conventional cytogenetics and comparative genomic hybridization, the ring chromosome was shown to be composed of sequences from chromosomes 17 and 22, 12,47 leading to the recognition that the fusion of COL1A1 and PDGFβ in DFSP is related to formation of the ring chromosome, or in occasional cases leads to a t(17:22) translocation. 5 Deregulated expression of wild-type PDGFβ protein has transforming activity in NIH3T3 cells 7,8 and is thought to play a critical role in the genesis of DFSP, but it is not clear whether other genetic abnormalities are also required for tumor development. Nishio et al 10 have performed conventional comparative genomic hybridization on 12 cases of DFSP, finding 0 to 3 genetic abnormalities per tumor (mean 1.9). Besides changes in chromosomes 17 and 22, they found gains of copy number of chromosome 8q24.1-qter in three tumors. They did not identify any instances of loss of DNA sequences. By array CGH analysis, we found that abnormalities of chromosomes 17 and 22, consisting of an amplification of large regions of chromosome 17 and 22 bounded by the genes involved in t(17;22), COL1A1, and PDGFβ, were present in all four cases we studied. This suggests that these abnormalities may have a role in tumorigenesis, in concert with deregulated expression of PDGFβ.

Previous studies have shown that paraffin-embedded tissue can be used for array CGH analysis using 2400 element BAC arrays. 48 We found that formalin-fixed, paraffin-embedded material is also suitable for array CGH using higher resolution genome-wide cDNA arrays, expanding the potential application of array CGH to large-scale retrospective studies of cancer. The average resolution of aCGH can be estimated by dividing the haploid genome size (1.5 billion bp) by the number of mapped genes on the array (~26,000), which equals ~60 kb. This is a conservative estimate of resolution, because genes are generally more closely spaced in the gene-rich portions of chromosomes. By using this conservative estimate, array CGH has a resolution >150 times that of conventional CGH, which is estimated at 10–20 MB for deletions. 49 The aCGH technique can thus be used to distinguish DFSP from lesions in its differential diagnosis, in cases where only formalin-fixed material is available, as is shown here for five cases of fibrous histiocytoma.

The use of the same cDNA microarrays for array CGH and gene expression profiling allows correlation of changes of gene copy number with gene expression on a genome wide scale. Pollack et al 21 were able to show a strong correlation between copy number changes and gene expression, and concluded that at least 12% of variation in gene expression among the breast carcinomas examined was directly attributable to changes in gene copy number. Unlike breast carcinomas, DFSPs show one consistent genetic abnormality (COL1A1-PDGFβ fusion) with relatively few other genetic alterations. Examination of gene expression in DFSP identified elevated expression of PDGFβ as a consistent feature. Array CGH showed amplification of the PDGFβ gene. In addition, 31 of the 259 genes that we found by SAM analysis to be consistently more highly expressed in DFSP than in other STT were located in the amplified regions of chromosomes 17 and 22. This highlights the potential role of other genes in the amplified regions of chromosomes 17 and 22 in DFSP tumorigenesis, as suggested previously by others. 11 Several of these 31 genes appear particularly worthy of further attention. TBX2 (Table 3) [triangle] , a gene in the amplified region of chromosome 17, encodes a transcription factor that can immortalize cell by down-regulating CDKN2a (p14ARF) and that has been found to be expressed at an elevated level in a subset of human breast carcinomas. 50 TBX2 is highly expressed in DFSP. Protein kinase C α is encoded by another of the 31 genes that are both amplified and highly expressed in DFSP. PRKA has roles in regulation of cell growth, differentiation, and transformation. 51,52 Although a generic protein kinase C inhibitor (calphostin C) does not inhibit DFSP cell growth in vitro, 53 an anti-sense oligonucleotide that specifically inhibits expression of PRKA has been shown to be active in several phase I studies in patients with a variety of malignancies. 54,55 GRB2, amplified on chromosome 17 (Table 2) [triangle] and highly expressed in DFSP, belongs to the group of proteins with SH2 and SH3 domains that link tyrosine kinases to intracellular pathways. 56 For instance, after activation of a PDGFRB receptor, GRB2 binds as an adaptor molecule to the receptor and then complexes with SOS1, which leads to RAS activation (see web figure 4 [triangle] ). 57 In a mouse model that used the polyomavirus middle T antigen to induce mammary carcinomas, the onset of mammary tumorigenesis was significantly delayed in mice heterozygous for an inactivating Grb2 mutation. 58 The hypothesis that high GRB2 expression in DFSP contributes to tumorigenesis in DFSP may warrant further investigation. PRKA and GRB2 both act downstream of the PDGF receptor (web figure 4 [triangle] ).

A single case of DFSP (STT3053) was noted to lack PDGFβ overexpression. This finding raises the question whether another signaling gene mutation that might substitute for PDGFβ could be present in this case. A similar situation has been described in gastrointestinal stromal tumors where it was recently reported that PDGFRA activation mutations could substitute for the usual KIT mutations found in these tumors. 58

In summary, DNA microarray analysis of global gene expression patterns and gene copy number changes in DFSP has provided new molecular criteria for the diagnosis of this malignancy, suggested potential clues to its pathogenesis, and identified new candidate molecular targets for therapy.

Footnotes

Address reprint requests to Matt van de Rijn, Department of Pathology, Stanford University Medical Center, 300 Pasteur Drive, Stanford, CA 94305. E-mail: mrijn/at/stanford.edu.

Supported by NCI grant CA85129 and the Howard Hughes Medical Institute.

P. O. B. is an Associate Investigator of the Howard Hughes Medical Institute.

S. C. L. was a recipient of a Dutch Cancer Society Postdoctoral Research Fellowship.

References

1. Bowne WB, Antonescu CR, Leung DH, Katz SC, Hawkins WG, Woodruff JM, Brennan MF, Lewis JJ: Dermatofibrosarcoma protuberans: a clinicopathologic analysis of patients treated and followed at a single institution. Cancer 2000, 88:2711-2720. [PubMed]
2. Suit HD, Spiro IJ, Spear M: Benign and low-grade tumors of the soft tissues: role for radiation therapy. Cancer Treat Res 1997, 91:95-105. [PubMed]
3. Rubin BP, Schuetze SM, Eary JF, Norwood TH, Mirza S, Conrad EU, Bruckner JD: Molecular targeting of platelet-derived growth factor B by imatinib mesylate in a patient with metastatic dermatofibrosarcoma protuberans. J Clin Oncol 2002, 20:3586-3591. [PubMed]
4. Maki RG, Awan RA, Dixon RH, Jhanwar S, Antonescu CR: Differential sensitivity to imatinib of 2 patients with metastatic sarcoma arising from dermatofibrosarcoma protuberans. Int J Cancer 2002, 100:623-626. [PubMed]
5. Simon MP, Pedeutour F, Sirvent N, Grosgeorge J, Minoletti F, Coindre JM, Terrier-Lacombe MJ, Mandahl N, Craver RD, Blin N, Sozzi G, Turc-Carel C, O’Brien KP, Kedra D, Fransson I, Guilbaud C, Dumanski JP: Deregulation of the platelet-derived growth factor B-chain gene via fusion with collagen gene COL1A1 in dermatofibrosarcoma protuberans and giant-cell fibroblastoma. Nat Genet 1997, 15:95-98. [PubMed]
6. O’Brien KP, Seroussi E, Dal Cin P, Sciot R, Mandahl N, Fletcher JA, Turc-Carel C, Dumanski JP: Various regions within the alpha-helical domain of the COL1A1 gene are fused to the second exon of the PDGFB gene in dermatofibrosarcomas and giant-cell fibroblastomas. Genes Chromosomes Cancer 1998, 23:187-193. [PubMed]
7. Shimizu A, O’Brien KP, Sjoblom T, Pietras K, Buchdunger E, Collins VP, Heldin CH, Dumanski JP, Ostman A: The dermatofibrosarcoma protuberans-associated collagen type Iα1/platelet-derived growth factor (PDGF) B-chain fusion gene generates a transforming protein that is processed to functional PDGF- BB. Cancer Res 1999, 59:3719-3723. [PubMed]
8. Greco A, Fusetti L, Villa R, Sozzi G, Minoletti F, Mauri P, Pierotti MA: Transforming activity of the chimeric sequence formed by the fusion of collagen gene COL1A1 and the platelet derived growth factor b-chain gene in dermatofibrosarcoma protuberans. Oncogene 1998, 17:1313-1319. [PubMed]
9. Nishio J, Iwasaki H, Ohjimi Y, Ishiguro M, Isayama T, Naito M, Kaneko Y, Kikuchi M: Supernumerary ring chromosomes in dermatofibrosarcoma protuberans may contain sequences from 8q11.2-qter and 17q21-qter: a combined cytogenetic and comparative genomic hybridization study. Cancer Genet Cytogenet 2001, 129:102-106. [PubMed]
10. Nishio J, Iwasaki H, Ohjimi Y, Ishiguro M, Isayama T, Naito M, Iwashita A, Kikuchi M: Overrepresentation of 17q22-qter and 22q13 in dermatofibrosarcoma protuberans but not in dermatofibroma: a comparative genomic hybridization study. Cancer Genet Cytogenet 2002, 132:102-108. [PubMed]
11. Kiuru-Kuhlefelt S, El Rifai W, Fanburg-Smith J, Kere J, Miettinen M, Knuutila S: Concomitant DNA copy number amplification at 17q and 22q in dermatofibrosarcoma protuberans. Cytogenet Cell Genet 2001, 92:192-195. [PubMed]
12. Naeem R, Lux ML, Huang SF, Naber SP, Corson JM, Fletcher JA: Ring chromosomes in dermatofibrosarcoma protuberans are composed of interspersed sequences from chromosomes 17 and 22. Am J Pathol 1995, 147:1553-1558. [PMC free article] [PubMed]
13. Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, Boldrick JC, Sabet H, Tran T, Yu X, Powell JI, Yang L, Marti GE, Moore T, Hudson J, Jr, Lu L, Lewis DB, Tibshirani R, Sherlock G, Chan WC, Greiner TC, Weisenburger DD, Armitage JO, Warnke R, Staudt LM: Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000, 403:503-511. [PubMed]
14. Garber ME, Troyanskaya OG, Schluens K, Petersen S, Thaesler Z, Pacyna-Gengelbach M, van de Rijn M, Rosen GD, Perou CM, Whyte RI, Altman RB, Brown PO, Botstein D, Petersen I: diversity of gene expression in adenocarcinoma of the lung. Proc Natl Acad Sci USA 2001, 98:13784-13789. [PMC free article] [PubMed]
15. Perou CM, Sorlie T, Eisen MB, van de Rijn M, Jeffrey SS, Rees CA, Pollack JR, Ross DT, Johnsen H, Akslen LA, Fluge O, Pergamenschikov A, Williams C, Zhu SX, Lonning PE, Borresen-Dale AL, Brown PO, Botstein D: Molecular portraits of human breast tumours. Nature 2000, 406:747-752. [PubMed]
16. Nielsen TO, West RB, Linn SC, Alter O, Knowling MA, O’Connell JX, Zhu S, Fero M, Sherlock G, Pollack JR, Brown PO, Botstein D, van de Rijn M: Molecular characterisation of soft tissue tumours: a gene expression study. Lancet 2002, 359:1301-1307. [PubMed]
17. Allander SV, Illei PB, Chen Y, Antonescu CR, Bittner M, Ladanyi M, Meltzer PS: Expression profiling of synovial sarcoma by cDNA microarrays: association of ERBB2, IGFBP2, and ELF3 with epithelial differentiation. Am J Pathol 2002, 161:1587-1595. [PMC free article] [PubMed]
18. Khan J, Wei JS, Ringner M, Saal LH, Ladanyi M, Westermann F, Berthold F, Schwab M, Antonescu CR, Peterson C, Meltzer PS: Classification and diagnostic prediction of cancers using gene expression profiling and artificial neural networks. Nat Med 2001, 7:673-679. [PMC free article] [PubMed]
19. Khan J, Simon R, Bittner M, Chen Y, Leighton SB, Pohida T, Smith PD, Jiang Y, Gooden GC, Trent JM, Meltzer PS: Gene expression profiling of alveolar rhabdomyosarcoma with cDNA microarrays. Cancer Res 1998, 58:5009-5013. [PubMed]
20. Allander SV, Nupponen NN, Ringner M, Hostetter G, Maher GW, Goldberger N, Chen Y, Carpten J, Elkahloun AG, Meltzer PS: Gastrointestinal stromal tumors with KIT mutations exhibit a remarkably homogeneous gene expression profile. Cancer Res 2001, 61:8624-8628. [PubMed]
21. Pollack JR, Sorlie T, Perou CM, Rees CA, Jeffrey SS, Lonning PE, Tibshirani R, Botstein D, Borresen-Dale AL, Brown PO: Microarray analysis reveals a major direct role of DNA copy number alteration in the transcriptional program of human breast tumors. Proc Natl Acad Sci USA 2002, 99:12963-12968. [PMC free article] [PubMed]
22. Fletcher CD, Dal Cin P, De W, I, Mandahl N, Mertens F, Mitelman F, Rosai J, Rydholm A, Sciot R, Tallini G, van den Berghe H, Vanni R, Willen H: Correlation between clinicopathological features and karyotype in spindle cell sarcomas: a report of 130 cases from the CHAMP study group. Am J Pathol 1999, 154:1841-1847. [PMC free article] [PubMed]
23. Mitelman F, Johansson B, Mertens F: Mitelman Database of Chromosome Aberrations in Cancer (2003). CGAP Website. 11–26-2002, http://cgap.nci.nih.gov/Chromosomes/
24. Gollub J, Ball CA, Binkley G, Demeter J, Finkelstein DB, Hebert JM, Hernandez-Boussard T, Jin H, Kaloper M, Matese JC, Schroeder M, Brown PO, Botstein D, Sherlock G: The Stanford Microarray Database: data access and quality assessment tools. Nucleic Acids Res 2003, 31:94-96. [PMC free article] [PubMed]
25. Eisen MB, Spellman PT, Brown PO, Botstein D: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci USA 1998, 95:14863-14868. [PMC free article] [PubMed]
26. Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci USA 2001, 98:5116-5121. [PMC free article] [PubMed]
27. Aiba S, Tabata N, Ishii H, Ootani H, Tagami H: Dermatofibrosarcoma protuberans is a unique fibrohistiocytic tumour expressing CD34. Br J Dermatol 1992, 127:79-84. [PubMed]
28. Natkunam Y, Rouse RV, Zhu S, Fisher C, van de Rijn M: Immunoblot analysis of CD34 expression in histologically diverse neoplasms. Am J Pathol 2000, 156:21-27. [PMC free article] [PubMed]
29. Gloghini A, Rizzo A, Zanette I, Canal B, Rupolo G, Bassi P, Carbone A: KP1/CD68 expression in malignant neoplasms including lymphomas, sarcomas, and carcinomas. Am J Clin Pathol 1995, 103:425-431. [PubMed]
30. Chleq-Deschamps CM, LeBrun DP, Huie P, Besnier DP, Warnke RA, Sibley RK, Cleary ML: Topographical dissociation of BCL-2 messenger RNA and protein expression in human lymphoid tissues. Blood 1993, 81:293-298. [PubMed]
31. Siegmund D: Confidence sets in change-point problems. Int Statist Rev 1988, 56:31-48.
32. Kempson RL, Fletcher CDM, Evans HE, Hendrickson MR, Sibley RK: Tumors of the soft tissues. Rosai J Sobin LH eds. Atlas of Tumor Pathology 2001:pp 130-148 Armed Forces Institute of Pathology, Washington DC.
33. Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD, Sallan SE, Lander ES, Golub TR, Korsmeyer SJ: MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 2002, 30:41-47. [PubMed]
34. Pomeroy SL, Tamayo P, Gaasenbeek M, Sturla LM, Angelo M, McLaughlin ME, Kim JY, Goumnerova LC, Black PM, Lau C, Allen JC, Zagzag D, Olson JM, Curran T, Wetmore C, Biegel JA, Poggio T, Mukherjee S, Rifkin R, Califano A, Stolovitzky G, Louis DN, Mesirov JP, Lander ES, Golub TR: Prediction of central nervous system embryonal tumour outcome based on gene expression. Nature 2002, 415:436-442. [PubMed]
35. Sorlie T, Perou CM, Tibshirani R, Aas T, Geisler S, Johnsen H, Hastie T, Eisen MB, van de Rijn M, Jeffrey SS, Thorsen T, Quist H, Matese JC, Brown PO, Botstein D, Eystein LP, Borresen-Dale AL: Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci USA 2001, 98:10869-10874. [PMC free article] [PubMed]
36. Beer DG, Kardia SL, Huang CC, Giordano TJ, Levin AM, Misek DE, Lin L, Chen G, Gharib TG, Thomas DG, Lizyness ML, Kuick R, Hayasaka S, Taylor JM, Iannettoni MD, Orringer MB, Hanash S: Gene-expression profiles predict survival of patients with lung adenocarcinoma. Nat Med 2002, 8:816-824. [PubMed]
37. Singh D, Febbo PG, Ross K, Jackson DG, Manola J, Ladd C, Tamayo P, Renshaw AA, D’Amico AV, Richie JP, Lander ES, Loda M, Kantoff PW, Golub TR, Sellers WR: Gene expression correlates of clinical prostate cancer behavior. Cancer Cell 2002, 1:203-209. [PubMed]
38. Dhanasekaran SM, Barrette TR, Ghosh D, Shah R, Varambally S, Kurachi K, Pienta KJ, Rubin MA, Chinnaiyan AM: Delineation of prognostic biomarkers in prostate cancer. Nature 2001, 412:822-826. [PubMed]
39. van de Rijn M, Rouse RV: CD34: a review. Appl Immunohistochem 1994, 2:71-80.
40. Candia AF, Hu J, Crosby J, Lalley PA, Noden D, Nadeau JH, Wright CV: Mox-1 and Mox-2 define a novel homeobox gene subfamily and are differentially expressed during early mesodermal patterning in mouse embryos. Development 1992, 116:1123-1136. [PubMed]
41. Candia AF, Wright CV: Differential localization of Mox-1 and Mox-2 proteins indicates distinct roles during development. Int J Dev Biol 1996, 40:1179-1184. [PubMed]
42. Mankoo BS, Collins NS, Ashby P, Grigorieva E, Pevny LH, Candia A, Wright CV, Rigby PW, Pachnis V: Mox2 is a component of the genetic hierarchy controlling limb muscle development. Nature 1999, 400:69-73. [PubMed]
43. Leimeister C, Externbrink A, Klamt B, Gessler M: Hey genes: a novel subfamily of hairy- and enhancer of split related genes specifically expressed during mouse embryogenesis. Mech Dev 1999, 85:173-177. [PubMed]
44. Steidl C, Leimeister C, Klamt B, Maier M, Nanda I, Dixon M, Clarke R, Schmid M, Gessler M: Characterization of the human and mouse HEY1, HEY2, and HEYL genes: cloning, mapping, and mutation screening of a new bHLH gene family. Genomics 2000, 66:195-203. [PubMed]
45. Bridge JA, Neff JR, Sandberg AA: Cytogenetic analysis of dermatofibrosarcoma protuberans. Cancer Genet Cytogenet 1990, 49:199-202. [PubMed]
46. Mandahl N, Heim S, Willen H, Rydholm A, Mitelman F: Supernumerary ring chromosome as the sole cytogenetic abnormality in a dermatofibrosarcoma protuberans. Cancer Genet Cytogenet 1990, 49:273-275. [PubMed]
47. Pedeutour F, Coindre JM, Sozzi G, Nicolo G, Leroux A, Toma S, Miozzo M, Bouchot C, Hecht F, Ayraud N: Supernumerary ring chromosomes containing chromosome 17 sequences: a specific feature of dermatofibrosarcoma protuberans? Cancer Genet Cytogenet 1994, 76:1-9. [PubMed]
48. Paris PL, Albertson DG, Alers JC, Andaya A, Carroll P, Fridlyand J, Jain AN, Kamkar S, Kowbel D, Krijtenburg P-J, Pinkel D, Schroder FH, Vissers KJ, Watson VJ, Wildhagen MF, Collins C, van Dekken H: High-resolution analysis of paraffin-embedded and formalin-fixed prostate tumors using comparative genomic hybridization to genomic microarrays. Am J Pathol 2003, 162:763-770. [PMC free article] [PubMed]
49. Kallioniemi OP, Kallioniemi A, Piper J, Isola J, Waldman FM, Gray JW, Pinkel D: Optimizing comparative genomic hybridization for analysis of DNA sequence copy number changes in solid tumors. Genes Chromosomes Cancer 1994, 10:231-243. [PubMed]
50. Jacobs JJ, Keblusek P, Robanus-Maandag E, Kristel P, Lingbeek M, Nederlof PM, van Welsem T, van de Vijver MJ, Koh EY, Daley GQ, van Lohuizen M: Senescence bypass screen identifies TBX2, which represses Cdkn2a (p19(ARF)) and is amplified in a subset of human breast cancers. Nat Genet 2000, 26:291-299. [PubMed]
51. Leszczynski D, Joenvaara S, Foegh ML: Protein kinase C-α regulates proliferation but not apoptosis in rat coronary vascular smooth muscle cells. Life Sci 1996, 58:599-606. [PubMed]
52. Wang XY, Repasky E, Liu HT: Antisense inhibition of protein kinase Cα reverses the transformed phenotype in human lung carcinoma cells. Exp Cell Res 1999, 250:253-263. [PubMed]
53. Ihn H, Tamaki K: Mitogenic activity of dermatofibrosarcoma protuberans is mediated via an extracellular signal related kinase dependent pathway. J Invest Dermatol 2002, 119:954-960. [PubMed]
54. Yuen AR, Sikic BI: Clinical studies of antisense therapy in cancer. Front Biosci 2000, 5:D588-D593. [PubMed]
55. Nemunaitis J, Holmlund JT, Kraynak M, Richards D, Bruce J, Ognoskie N, Kwoh TJ, Geary R, Dorr A, Von Hoff D, Eckhardt SG: Phase I evaluation of ISIS 3521, an antisense oligodeoxynucleotide to protein kinase C-α, in patients with advanced cancer. J Clin Oncol 1999, 17:3586-3595. [PubMed]
56. Cheng AM, Saxton TM, Sakai R, Kulkarni S, Mbamalu G, Vogel W, Tortorice CG, Cardiff RD, Cross JC, Muller WJ, Pawson T: Mammalian Grb2 regulates multiple steps in embryonic development and malignant transformation. Cell 1998, 95:793-803. [PubMed]
57. Heldin CH, Ostman A, Ronnstrand L: Signal transduction via platelet-derived growth factor receptors. Biochim Biophys Acta 1998, 1378:F79-F113. [PubMed]
58. Heinrich MC, Corless CL, Duensing A, McGreevey L, Chen C-J, Joseph N, Singer S, Griffith DJ, Haley A, Town A, Demetri GD, Fletcher CDM, Fletcher JA: PDGFRA activating mutations in gastrointestinal stromal tumors. Science 2003, 299:708-710. [PubMed]

Articles from The American Journal of Pathology are provided here courtesy of American Society for Investigative Pathology

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • Compound
    Compound
    PubChem Compound links
  • EST
    EST
    Published EST sequences
  • GEO DataSets
    GEO DataSets
    GEO DataSet links
  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles
  • Substance
    Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...