| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
|
FJ
EXPRESS SUMMARY ARTICLE The Full-length version of this article is also available, published online January 3, 2005 as doi:10.1096/fj.04-2104fje. |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,*,1,2
,1,*
,*
* Department of Neurogenetics,
Department of Anatomy and Embryology,
Bioinformatics Laboratory, and
Department of Human Genetics, Academic Medical Centre, University of Amsterdam, Amsterdam, The Netherlands
2Correspondence: Department of Human Genetics, Academic Medical Centre, University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands. E-mail: g.j.schaaf{at}amc.uva.nl
SPECIFIC AIMS
The primary aim of this study was to increase the knowledge of the basic biology of rhabdomyosarcoma (RMS), the most common soft tissue sarcoma in children. Because the response of RMS patients to treatment varies considerably, there is a need for good diagnostic and/or prognostic markers. We present the first full transcriptome analysis of RMS as a rich source for the identification of genes contributing significantly to the development and maintenance of RMS.
PRINCIPAL FINDINGS
Two public SAGE libraries derived from healthy skeletal muscle samples were compared with four RMS libraries we generated from three embryonal RMS (ERMS) samples and one alveolar RMS (ARMS) sample using an adapted statistical test (log likelihood ratio test or G-test). A fetal muscle SAGE library was generated to study its similarity to the RMS transcriptome and determine whether the differentially expressed genes identified by the G-test are similarly regulated in fetal muscle tissue and RMS compared with normal muscle.
1. Statistical comparison of multiple SAGE libraries
Here, we describe for the first time the modification and use of the log likelihood ratio test (G-test) in analyzing SAGE data. In contrast to other widely used tests (Z-test, Chi-squared test), the G-test allowed us to test the overall null hypothesis that a set of more than two libraries are all the same and to test specific differences between individual libraries and between two (or more) groups of libraries. The test procedure is flexible and can be tuned to different library comparison formats, so that even a large number of libraries can be easily and reliably analyzed. The G-test will be made available as a computer program upon E-mail request (biolab-services{at}amc.uva.nl; subject: Gtest).
2. Identification of differentially expressed genes
The G-test identified in a simultaneous comparison of subsets of SAGE libraries of normal muscle, embryonal (ERMS), and alveolar (ARMS) RMS, 403 tags with significant different abundances; 259 could be annotated, representing 251 genes. The remaining 144 tags await identification.
The G-test identified 175 genes as differentially expressed between two normal muscle and four RMS libraries (NvsR genes). Our automated literature mining procedure showed that 108 of these genes had not been associated with RMS and could be designated "novel RMS/muscle" genes; 48 were designated as "muscle-associated" and 19 as "RMS" genes. Gene Ontology (GO) analysis identified 34 biological processes and 20 molecular functions that were significantly enriched with differentially expressed genes (containing in total 114 NvsR genes; Fig. 1
).
|
Most of the NvsR genes down-regulated in RMS, such as desmin and troponin T1, were grouped by GO analysis into "muscle development" and "muscle contraction." Functional classes related to growth and cell cycle progression were represented by NvsR genes that are up-regulated in RMS. A considerable number of NvsR genes are involved in "actin reorganization," including actinin (ACTN2 and ACTN3), troponin (TNN1), cofilin (CFL1), nebulin-anchoring protein (NRAP), and fascin 1 (FSCN1). Several significant functional classes relate to cell proliferation and include insulin growth factor 2 (IGF2), midkine (MDK), ribosomal protein S4 (RPS4X), and S-phase kinase-associated protein 2 (SKP2).
The up- or down-regulation of 56 of 69 (81%) differentially expressed genes was confirmed using custom microarrays in an extended dot blot approach, strengthening results from the SAGE analysis.
In the comparison of the ARMS and ERMS samples, 91 genes were identified as differentially expressed (AvsE genes). Our automated literature mining procedure showed that 60 genes could be designated "novel RMS/muscle" genes, 21 "muscle-associated," and 10 "RMS" genes. GO analysis identified 11 biological processes and 10 molecular functions significantly enriched with differentially expressed genes (in total, 45 AvsE genes).
Most prominent among the significant GO classes were those involved in "RNA binding," "cell adhesion," "muscle development," and "glucose metabolism." Microarray analysis confirmed the up- or down-regulation of 26 of 34 (76%) differentially expressed genes in the ARMS vs. ERMS samples.
3. The RMS transcriptome resembles the transcriptome of fetal muscle
Hierarchical clustering showed that the fetal muscle transciptome is more similar to that of RMS than to that of normal muscle (Fig. 2
); 86% of the NvsR genes, including NRAP, RPS4X, and IGF2, were coordinately regulated in fetal muscle and RMS compared with normal muscle. Only a few genes (including FGFR4, NOTCH2, UBE2C, UHRF1, and YWHAB) were up-regulated exclusively in RMS.
|
SIGNIFICANCE AND CONCLUSIONS
Our results present the first complete transcriptome analysis of RMS compared with normal and fetal muscle. The stepwise approach of statistical analysis, followed by literature mining and gene ontology analysis, reduces the initial plethora of the expressed genes to a manageable number of new candidate genes that may contribute to the development and/or maintenance of RMS, grouped in relevant gene ontology (GO) classes. The next step, to verify the consistency of the observed differential expression of the genes in a larger set of RMS tumor samples, is now in progress in our laboratory.
Our literature-mining procedure revealed that the majority of the G-test-identified genes had not previously been associated with RMS, revealing this approach as ideally suited for hypothesis-generating research, providing a rich template for focused follow-up studies. However, the literature mining and GO analysis approach requires solid knowledge of the genes submitted. Some tags with significantly different proportions could not reliably be annotated. Regularly updating the annotation of the tags or empirical elucidation of the gene identity of these tags allows discovery of real, novel genes.
The marked GO class "actin reorganization" and "cell growth and proliferation" are crucial during metastasis and growth of tumors, and are tightly regulated during normal myogenic development. These results therefore corroborate the association between undifferentiated myogenic precursor cells and RMS development. Aberrant expression of genes involved in these processes may contribute to the malignant nature of RMS. Fascin is involved in "actin reorganization" and "cell growth and proliferation." Strong FSCN1 expression in several tumors correlates with poor clinical prognosis and would be an interesting candidate to screen as a prognostic marker in RMS.
The close association between myogenic precursor cells and RMS is further substantiated by expression analysis of our fetal muscle SAGE library. Hierarchical clustering showed that the fetal muscle expression profile was much more like that of RMS than that of normal muscle. Coordinated differences in the expression of differentially expressed NvsR genes in fetal muscle and the RMS library vs. normal muscle libraries indicate a high degree of similarity between fetal muscle and RMS transcriptomes. Such a large-scale gene comparison between fetal muscle and RMS had not been described before. The high degree of similarity between fetal muscle and RMS expression profiles may reflect and ensure the undifferentiated myogenic nature of RMS. Genes exclusively up-regulated in RMS—FGFR4, NOTCH2, UBE2C, UHRF1, and YWHAB—may contribute to the failure of RMS cells to complete normal skeletal muscle development and progress to an alternative fate. Increased expression of these "exclusive RMS" genes stimulates (myogenesis-specific) proliferation and so inhibits myogenic differentiation. UBE2C, UHRF1, and other genes differentially expressed between normal skeletal muscle and RMS (including SKP2 and UBE2I) are involved in the ubiquitin-dependent proteolytic degradation of cell cycle control proteins. The relative abundance of these genes may indicate that deregulated ubiquitin-dependent control of cell cycle progression contributes to RMS development.
Differentially expressed genes in the significantly enriched functional classes are of primary importance, but groups of genes with a small (not significant) change in gene expression of functionally related genes may have a major effect on disease progress. Unfortunately, our approach will not identify such groups of genes. On the other hand, when "novel RMS" genes are grouped into functional classes with "known" RMS or normal muscle-associated genes, clues are obtained for their RMS or muscle-specific functions or processes.
The G-test proved an extremely useful tool to fully exploit the information contained in SAGE libraries. Genes identified form a solid basis for further exploration of the biology and pathology of RMS and normal myogenic development.
|
FOOTNOTES
To read the full text of this article, go to http://www.fasebj.org/cgi/doi/10.1096/fj.04-2104fje;
This article has been cited by other articles:
|
|
 |
|
  P. P. Breitfeld and W. H. Meyer Rhabdomyosarcoma: New Windows of Opportunity Oncologist, August 1, 2005; 10(7): 518 - 527. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
