Inactivation of tumor suppressor genes, whose products normally provide negative control of cell proliferation, contributes to malignant transformation in a variety of cell types. Knudson first proposed that two hits, or mutations, are required for the development of retinoblastoma. His prediction was subsequently supported by the cloning of the retinoblastoma tumor suppressor gene (RB1) and by functional studies of the retinoblastoma protein, Rb. The first mutation of RB1 in cases of retinoblastoma can be either constitutional or somatic, whereas the second mutation is always somatic. In the inherited form of retinoblastoma, the first mutation is present in the germline; an early onset and a high frequency of bilateral disease characterize these cases. In contrast, both mutations in nonhereditary retinoblastoma are somatic.

Like Rb protein, many of the proteins encoded by tumor suppressor genes act at specific points in the cell cycle. For example, the TP53 gene, located on chromosome 17, encodes a 53-kd nuclear protein that functions as a cell cycle checkpoint. As a transcription factor whose expression is increased by DNA damage, p53 blocks cell division at the G1 phase of the cell cycle to allow DNA repair. The TP53 gene also is capable of stimulating apoptosis of cells containing damaged DNA. Targeted disruption of TP53 in the mouse leads to the development of a variety of tumors. Germline mutation of one TP53 allele is found in patients with Li-Fraumeni syndrome who generally inherit a mutated TP53 gene from an affected parent. Patients with Li-Fraumeni syndrome are predisposed to sarcomas, breast cancer, brain tumors, adrenocortical cell carcinoma, and acute leukemia; they have a 50% probability of cancer development by age 30 years.

Another important class of tumor suppressor genes involved in cell cycle control and in the generation of human cancers is the cyclin-dependent kinase (CDK) inhibitors. These proteins negatively regulate the cell cycle by inhibiting CDK phosphorylation of Rb protein and include p15INK4B, p16INK4A, p18INK4C, p19INK4D, p19ARF, p21CIP1, p27KIP1, and p57KIP2. Although carcinogenic roles for the INK4B, INK4C, INK4D, CIP1, KIP1, and KIP2 genes appear to be limited, INK4A is among the most commonly mutated genes in human tumors. The p16INK4A protein is a cell-cycle inhibitor that acts by inhibiting activated cyclin D:CDK4/6 complexes, which play a crucial role in the control of the cell cycle by phosphorylating Rb protein.

Direct evidence linking the INK4A locus to tumorigenesis was provided by the targeted disruption of exon 2 of INK4A in mice. Tumors that developed in mice deficient in INK4A were enhanced by the topical application of carcinogens and ultraviolet light. This locus, however, also encodes a protein from an alternative reading frame, designated p19ARF. Interpretation of INK4A knockout experiments was uncertain because targeted disruption of exon 2 of INK4A also disrupts ARF. Further genetic analysis showed that selective disruption of ARF reproduces the phenotype described for INK4A-null mice; this finding indicates that ARF is a true tumor suppressor gene.

In addition, p19ARF binds to and promotes the degradation of MDM2, the product of the murine double minute 2 gene, and this degradation leads to accumulation of TP53 and to cell cycle arrest. Therefore, it appears that both INK4A and ARF are tumor suppressor genes, acting through either the Rb protein (INK4A) pathway or the TP53 (ARF) pathway in different tumor subsets.

Activation of proto-oncogenes is a common theme in childhood leukemias and solid tumors. Transcription factors (proteins that bind to the regulatory sequences of target genes) compose the largest class of oncogenes identified in pediatric tumors. Oncogenic transcription factors commonly show close homology to proteins with important regulatory functions in primitive organisms; this homology indicates that the biochemical pathways leading to cell transformation are well conserved in nature.

Tumor-specific translocations can oncogenically activate transcription factors by at least 2 mechanisms. In B-progenitor acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), non-Hodgkin lymphoma (NHL), and certain solid tumors, translocations fuse discrete portions of 2 different genes to create chimeric transcription factors with oncogenic properties. Gene fusion is also the mechanism by which tyrosine kinases become activated in ALL. Alternatively, in T- and B-cell acute leukemia, transcription factor genes are dysregulated by their juxtaposition with transcriptionally active T-cell receptor (TCR) or immunoglobulin (IG) genes.

BCR-abl fusion gene

The first fusion gene described in ALL, bcr-abl, is created by the der(22) of the t(9;22)(q34;q11), which occurs in 3-5% of childhood cases. The t(9;22) is also present in most cases of chronic myelogenous leukemia (CML). This translocation moves the abl proto-oncogene from chromosome 9 into the BCR gene on chromosome 22. In most CML cases and in about half of adult BCR-abl–positive ALL cases, the BCR breakpoints are located in the major breakpoint cluster region. Chimeric messenger ribonucleic acids (mRNAs) transcribed from the BCR-abl fusion gene encode a fusion tyrosine kinase of approximately 210 kd (p210).

In most cases of childhood BCR-abl–positive ALL, the BCR breakpoints are distributed in an upstream breakpoint cluster region; in this location the fusion gene encodes a 190-kd chimeric protein (p190). The tyrosine kinase activity of p210 and p190 is higher than that of abl. Both p210 and p190 localize to the cytoplasm and transform hematopoietic precursors in experimental systems. Their role in leukemic transformation appears to involve multiple signaling pathways.

In pediatric ALL, the t(9;22) is associated with a poor prognosis; allogeneic hematopoietic stem cell transplantation during first remission was formerly believed to be the only curative treatment. Recent evidence, however, indicates that certain subsets of these patients, including those who have low leukocyte counts at the time of diagnosis or whose disease initially responds well to prednisone, can be cured with contemporary chemotherapy.

E2A-PBX1 fusion gene

The t(1;19)(q23;p13.3), which occurs in 25% of ALL cases with a pre-B (cytoplasmic IG-positive) immunophenotype, fuses the transactivation domains of the E2A gene, which encodes a basic helix-loop-helix (bHLH) transcription factor, to PBX1, an atypical homeobox (HOX) gene. Each of the E2A proteins (E12 and E47) contains a bHLH domain that is responsible for sequence-specific DNA binding and protein dimerization. In addition, the amino terminal portion of E2A contains 2 transcriptional transactivation domains. The E2A-PBX1 chimeric protein contains these transactivation domains fused to the homeodomain of PBX1.

E2A-PBX1 binds DNA in a site-specific manner, is a strong transcriptional transactivator, and transforms NIH3T3 fibroblasts in culture. In addition, it induces T-cell lymphomas in transgenic mice and, when introduced into murine bone marrow progenitors by retroviral infection, also induces AML.

The transgenic mouse model also implicates E2A-PBX1 in the induction of apoptosis in lymphoid cells. The binding of both PBX1 and E2A-PBX1 to the consensus PBX1 DNA sequence is stimulated by direct interactions between PBX1 and other HOX proteins. Because HOX proteins appear to direct E2A-PBX1 to DNA sites recognized by HOX:PBX1 complexes, it is likely that E2A-PBX1 interferes with hematopoietic differentiation by disrupting gene expression that is normally regulated by HOX proteins. In this regard, E2A-PBX1 can induce the aberrant expression of developmentally regulated genes when it is expressed in fibroblasts.

Surprisingly, B-cell precursors, the target of E2A-PBX1 in human leukemias, cannot be transformed in culture. Instead, the inducible expression of E2A-PBX1 in B cell progenitors induces TP53-independent apoptosis, suggesting that the in vivo leukemogenic potential of E2A-PBX1 may depend on cell type-specific resistance to apoptosis.

Molecular characterization of the t(1;19) has led to the development of reverse transcriptase–polymerase chain reaction (RT-PCR) assays that detect the E2A-PBX1 chimeric transcript. These assays can detect E2A-PBX1 fusions in patients for whom the results of cytogenetic studies were normal or in whom studies were unsuccessful, and they can identify patients whose cells contain the t(1;19) but lack the fusion gene.

With intensified chemotherapy, event-free survival (EFS) estimates that were once about 50% are now closer to 80%; this increase suggests that the adverse prognostic impact of this fusion can be overcome with chemotherapy that is more effective.

MLL fusion genes

The MLL gene, located at band 11q23, is altered in approximately 80% of infant ALL cases, 3% of ALL cases involving older children, and 85% of secondary AML cases that arise in patients who have been treated with topoisomerase II inhibitors. MLL encodes a 431-kd protein that contains 3 AT hook domains at the N-terminus, 2 central zinc finger domains, a region with homology to DNA methyltransferases, and a C-terminal region that shows high homology to the Drosophila trithorax protein. In Drosophila, trithorax regulates a variety of homeotic genes and is required for normal development. In human leukemias, 11q23 translocations cluster in an 8.5-kb region of MLL and fuse the N-terminal portion of MLL, containing the AT hook and methyltransferase domains, to over 25 different proteins.

Loss of MLL function has been studied using gene knockout techniques. MLL heterozygous mice are small at birth, demonstrate retarded growth, and display anemia and thrombocytopenia. MLL-deficient mice die in utero and fail to express specific HOX genes. Although gene knockout experiments suggest that the loss of one MLL allele may contribute to leukemogenesis, proof that MLL fusions contribute directly to leukemogenesis was derived from chimeric mice that express MLL-AF9 under the control of normal MLL transcriptional elements. After a latency period of 4-12 months, AML develops with great frequency in chimeric mice whose cells express MLL-AF9, and the leukemic phenotype is similar to that of patients carrying the t(9;11).

In contrast, leukemia does not develop in mice whose cells express a truncated MLL gene; this finding suggests that the fusion protein is essential for tumorigenesis. These experiments not only demonstrate that chromosomal translocations are directly involved in tumor development but also they provide a model system for studying other MLL fusion genes.

MLL rearrangements confer a dismal prognosis on infants with ALL; long-term EFS rates are approximately 20%. A subset of these patients, however, particularly those whose disease responds well to initial chemotherapy, have a relatively favorable outcome.

TEL-AML1 gene fusion

The t(12;21) is detected by karyotyping in fewer than 0.05% of ALL cases. Molecular techniques, however, have demonstrated that the TEL-AML1 fusion gene, created by the t(12;21), is present in approximately one fourth of childhood ALL cases. In the resulting chimeric protein, the helix-loop-helix (HLH) domain of TEL is fused to the DNA-binding and transactivation domains of AML1.

TEL and AML1 are involved in a variety of other leukemia-associated translocations. TEL originally was cloned as a fusion of TEL with the gene encoding the platelet-derived growth factor receptor b (PDGFRb); this fusion was caused by the t(5;12) in chronic myelomonocytic leukemia. AML1 is the DNA-binding component of the AML1:CBF transcription factor complex, which is the most frequent target of myeloid-associated translocations, including t(8;21), t(3;21), and inv(16).

The TEL-AML1 has been proposed to transform cells by interfering with AML1-mediated expression of HOX genes involved in lymphopoiesis. In this regard, fusion of TEL to AML1 converts AML1 from an activator to a repressor of transcription; this repression is dependent on the HLH dimerization motif of TEL. The leukemogenic properties of TEL-AML1 (and the other TEL fusions) also may involve disruption of the normal TEL pathway as TEL-AML1 forms heterodimers with and inactivates TEL.

Although the targets of TEL are unknown, the role of TEL in normal development has been examined by the targeted disruption of TEL in mouse embryos. TEL-deficient mice die at approximately day 11 of embryogenesis because of defective yolk sac angiogenesis and apoptosis of neural and mesenchymal cells; this finding establishes TEL as an important regulator of embryologic development.

TEL-AML1 expression is associated with an excellent prognosis; EFS estimates approach 90%. Recent results indicate a 10-year cumulative incidence of relapse of only 9% ± 5% for patients whose cells are positive for TEL-AML1. Thus, TEL-AML1 expression identifies a large, but previously unrecognized, subset of patients with B-precursor ALL who have a favorable outcome.

Activation of myc in B-cell ALL

B-cell ALL is characterized by the presence of surface IG, morphology characteristic of the French-American-British (FAB) classification L3, and translocations involving the myc gene on chromosome 8, band q24.

Approximately 80% of B-cell cases contain the t(8;14)(q24;q32), in which myc is translocated to the IG heavy chain gene locus. Nearly all of the remaining cases contain the t(2;8)(p12;q24) or the t(8;22)(q24;q11), in which either the k (located at band 2p12) or l (located at band 22q11) light chain gene is translocated to a region that is adjacent to myc. All 3 translocations lead to increased myc expression.

In turn, altered interactions between the myc protein and several other transcription factors are thought to lead to lymphoid transformation. Normally, myc dimerizes with the MAX transcription factor, which also can form heterodimers with MAD and Mxi1. myc:MAX dimers activate gene expression, whereas MAD:MAX dimers interact with the Sin3A protein to repress transcription. Overexpression of myc as a result of the t(8;14) or related translocations leads to increased levels of myc/MAX heterodimers relative to MAD:MAX, ultimately causing transformation by the activation of unknown target genes.

Although B-cell ALL does not respond well to the conventional chemotherapy used to treat childhood B-precursor ALL, outstanding results (EFS estimates of nearly 90%) have been obtained with treatments designed for Burkitt lymphoma, which emphasize cyclophosphamide and the rapid rotation of antimetabolites in high dosages. B-cell leukemia is, therefore, the first form of ALL to be treated by separate protocols designed specifically for its unique features.

Activation of transcription factor genes in T-cell ALL

Recurring translocations in T-cell ALL often involve the transcriptionally active sites of the TCRb locus (7q34) or the TCRa and TCRd locus (14q11); these translocations lead to dysregulated expression of transcription factor genes. Like the translocations identified in B-cell ALL, these rearrangements may result from mistakes in the normal recombination process involved in the generation of functional antigen receptors.

Disruption of the core binding factor (CBF) complex

The AML1:CBFb transcription factor complex, also known as CBF, is the most common translocation target in human leukemia. It is disrupted in approximately 30% of AML cases and 25% of ALL cases. AML1 is a member of the runt family of transcription factors and possesses DNA-binding, transactivation, and protein-protein interaction properties. Its DNA-binding affinity increases when it forms heterodimers with CBFb, which does not interact directly with DNA.

Knockout experiments have demonstrated that both AML1 and CBFb are essential for definitive hematopoiesis; these findings suggest that the AML1:CBFb complex regulates genes essential for normal blood cell development. The AML1:CBFb complex is disrupted by the t(8;21)(q22;q22) in approximately 40% of AML cases with FAB type M2. The t(8;21) creates an AML1-ETO fusion gene, whose protein product includes the runt homology domain of AML1 fused to ETO.

Like AML1-CBFb, AML1-ETO binds DNA and interacts with CBFb; however, AML1-ETO dominantly represses normal AML1-mediated transcriptional activation through interactions with the nuclear corepressor complex. ETO interacts directly with the nuclear corepressors N-CoR and Sin3A, forming a complex that recruits histone deacetylase (HDAC).

The AML1-ETO/N-CoR/Sin3A/HDAC complex leads to deacetylation of histones, alteration of chromatin structure, and active repression of AML1 target genes. Through these actions as a dominant negative protein complex, expression of AML1-ETO in the developing mouse produces a phenotype that is lethal to embryos and is similar to that caused by the loss of AML1. AML1-ETO also causes additional abnormalities in hematopoiesis that may represent preleukemic events.

CBF is also disrupted by the inv(16) and the t(16;16), which occurs in about 15% of AML cases and generally is associated with myelomonocytic differentiation and the presence of abnormal bone marrow eosinophils (FAB subtype M4Eo). As a result, the 5' portion of CBFb is joined to part of the smooth muscle myosin heavy chain gene (MYH11); this joining results in a chimeric CBFb-MYH11 protein.

This fusion protein binds to AML1 and transforms fibroblasts in vitro. Like AML1-ETO, CBFb-MYH11 also interferes with the normal transcriptional transactivation capacity of AML1-CBFb, in this case by binding and sequestering AML1 into inactive complexes.

Expression of this fusion protein in mice produces a phenotype similar to that of AML1-ETO mice, with abnormalities in early hematopoiesis. Recent data show that treatment of these mice with chemical mutagens produces a high frequency of AML; this finding suggests that cooperating genetic events are required for leukemic transformation by CBFb-MYH11.

PML-RARa fusion gene

Most cases of acute promyelocytic leukemia (APL, AML-M3) are associated with a balanced translocation that involves the retinoic acid receptor-alpha (RARa) gene at band 17q21 and the PML gene at band 15q21. RARa is a ligand-dependent transcription factor that interacts directly with DNA to regulate many genes, whereas PML is a tumor suppressor that plays a role in apoptotic pathways. Normally, RARa binds DNA as a heterodimer with RXR and represses transcription by recruiting the N-CoR/Sin3A/HDAC corepressor complex, much like AML1-ETO.

Pediatric NHL can be divided into 3 main categories, as follows: small noncleaved cell (SNCC) lymphoma, lymphoblastic lymphoma, and large cell lymphoma (LCL). SNCC lymphomas have a mature B-cell phenotype and demonstrate the same translocations as those seen in B-cell ALL. The molecular genetics of B-cell ALL are described above. Most cases of lymphoblastic lymphoma are of T-cell origin and have the same type of genetic change seen in T-cell ALL, also described above. This section focuses on the genetics of LCL.

Anaplastic large-cell lymphoma (ALCL) is a subtype of LCL that commonly is characterized by a T-cell phenotype, expression of the CD30 antigen, and an aggressive clinical profile with peripheral adenopathy and skin involvement. Because unambiguous morphologic or immunophenotypic criteria are lacking, diagnosing this entity can be difficult. Compounding the problem are similarities between ALCL and Hodgkin disease, which can lead to an erroneous diagnosis.

Effective clinical management of rhabdomyosarcoma, the family of tumors including Ewing sarcoma and primitive neuroectodermal tumor (PNET), and neuroblastoma depends on unequivocal diagnosis that can guide the selection of specific therapy. However, each of these tumors may first be seen as a soft-tissue mass with the appearance of undifferentiated small round cells. Although immunohistochemical analysis can help in the diagnostic workup for these tumors, this method has limitations. Recently, molecular diagnostic techniques have had an important role in ensuring diagnostic accuracy. The identification of molecular alterations has important prognostic and therapeutic implications.

Rhabdomyosarcoma

Most alveolar rhabdomyosarcomas contain 1 of 2 recurring translocations, namely, the common t(2;13)(q35;q14) or the rare t(1;13)(p36;q14). Both translocations disrupt the FKHR gene, which encodes a widely expressed transcription factor. The t(2;13) fuses part of the PAX3 transcription factor gene to FKHR, encoding a Pax3-Fkhr chimeric protein, whereas the t(1;13) creates a Pax7-Fkhr fusion. In vitro, these fusion proteins can function as transcriptional transactivators and can contribute to transformation.

RT-PCR assays have been developed to detect the chimeric transcripts resulting from these fusion events. Such tests are both specific and sensitive, enabling the detection of transcripts in as few as one tumor cell per 100,000 normal cells and identifying transcripts in cases that are not amenable to standard cytogenetic analysis.

Clinically, tumors expressing Pax7-Fkhr are associated with favorable features, and the prognosis for patients with these tumors is better than that of patients with Pax3-Fkhr–positive tumors.

Ewing sarcoma and the PNET family of tumors

More than 90% of Ewing tumors are characterized by the EWS-FLI1 fusion gene formed by the t(11;22) or by variant EWS fusions caused by the t(21;22) or the t(7;22). The t(11;22) produces a chimeric transcription factor that includes the transcriptional transactivation domain of EWS fused to the DNA binding domain FLI1; this factor is presumed to function by the aberrant activation of target genes. RT-PCR and fluorescence in situ hybridization assays for this fusion have been useful in distinguishing Ewing sarcoma from other small round cell tumors.

The precise t(11;22) breakpoint location has recently been demonstrated to have possible prognostic significance. Two studies suggest that the more common type of breakpoint (designated type I) is associated with a favorable outcome.

Neuroblastoma

Patients older than 1 year and those with tumor cell metastases have a poorer prognosis than other patients with neuroblastoma; these clinical features have been used to guide the selection of therapy. The identification of genetic alterations in this disease was recently recognized to greatly improve risk assessment.

In contrast to sarcomas, which are characterized by genetic alterations that produce chimeric transcription factors, neuroblastoma is characterized by gene amplification, tumor suppressor inactivation, and alterations in gene expression.

Amplification of the myc oncogene, located on chromosome 2, band p24, occurs in about one fourth of tumors and is associated with advanced stage and rapid disease progression. In addition, myc amplification is a powerful predictor of outcome independent of stage and age and is therefore a factor used to assign patients to more intensive therapies. Loss of heterozygosity of the short arm of chromosome 1 is also associated with an unfavorable outcome, a finding suggesting that a tumor suppressor gene may be located in this region. In contrast, hyperdiploid tumors in infants with neuroblastoma respond favorably to standard therapy, whereas diploid tumors require more intensive treatment.

Finally, expression of neurotrophin receptors is highly correlated with both biologic and genetic features. For example, high TRKA expression is correlated with a lack of myc amplification and a favorable outcome. TRKB, however, is more commonly expressed in higher-stage tumors that also show myc amplification.

Current risk classification schemes rely on both clinical and biologic factors in an attempt to provide the appropriate intensity of therapy for each group of patients.

Osteosarcoma

In contrast to Ewing sarcoma and rhabdomyosarcoma, recurring translocations and fusion oncogenes have not been identified in osteosarcoma. Instead, inactivation of tumor suppressor genes likely plays a role in the development of this tumor. Patients with germline mutations of either TP53 or RB1 are at increased risk of developing osteosarcoma, and loss of heterozygosity (LOH) at the sites of these genes (17p and 13q) is a frequent finding in tumors. In addition, 3q and 18q are common sites of LOH in osteosarcomas, suggesting that tumor suppressor genes located in these regions may be inactivated. Recently, increased expression of the growth factor HER2 has been associated with a poor response to chemotherapy and a worse outcome in osteosarcoma, providing both a prognostic marker and potential therapeutic target.

Brain tumors

A variety of tumor suppressor genes is implicated in the development of childhood brain tumors, including TP53 in brainstem gliomas and the PTEN gene in glioblastoma multiforme. The best-studied tumor, however, is medulloblastoma, a primitive neuroectodermal tumor that arises in the cerebellum and is the most common brain tumor in children. Although most tumors arise sporadically, medulloblastoma also occurs in patients with Turcot syndrome and in those with Gorlin syndrome. The latter is characterized by developmental anomalies, radiation sensitivity, basal cell carcinoma, a propensity to develop medulloblastoma, and germline mutations in the PTC gene.

Basal cell carcinomas from patients with Gorlin syndrome often demonstrate loss of the second PTC allele, suggesting that PTC functions as a tumor suppressor gene. In addition, one allele of PTC is occasionally mutated in sporadic medulloblastomas, implicating the PTC pathway in tumorigenesis. Interestingly, mice heterozygous for PTC also develop medulloblastoma, but the tumors retain one functional allele of PTC, indicating that haploinsufficiency of this gene is sufficient for oncogenesis.

Wilms tumor

Although more than 95% of Wilms tumor cases are sporadic, this disease also can occur in the context of congenital anomalies or as part of a familial predisposition syndrome. Patients with congenital anomalies or a family history often have bilateral tumors and are diagnosed at an earlier age, indicating the germline loss of a tumor suppressor gene in these children. Syndromes associated with Wilms tumor include the Beckwith-Wiedemann overgrowth syndrome, the Denys-Drash syndrome of renal failure and genitourinary (GU) anomalies, and the WAGR syndrome (Wilms tumor, aniridia, GU anomalies, and mental retardation). Cytogenetic studies of patients with WAGR syndrome and sporadic Wilms tumor demonstrated the importance of the 11p13 band in the development of Wilms tumor. This led to the cloning of the WT1 tumor suppressor gene. WT1 encodes a transcription factor that is important in normal kidney development and functions as a classic tumor suppressor.

However, mutations of WT1 are detected in a minority of sporadic Wilms tumor cases, suggesting that other genes are involved in the development of this disease. Aberrant expression of genes located at 11p15, such as H19, IGF2, and p57, as well as other loci, also are likely involved in tumorigenesis.

Children with the Beckwith-Wiedemann syndrome (BWS) are predisposed to Wilms tumor; these children are also at increased risk to develop hepatoblastoma, neuroblastoma, and rhabdomyosarcoma. In addition to predisposing persons to cancer, BWS is characterized by prenatal and postnatal gigantism, abdominal wall defects, macroglossia, and hemihypertrophy. BWS is usually sporadic, but autosomal dominant transmission has been reported as well. Both sporadic and hereditary forms have alterations of band 11p15. This fact initially supported the hypothesis of a “BWS” gene in this region. However, it now seems likely that BWS is caused by an imbalance in the expression of several genes in this region rather than by the disruption of a single gene.

Characterization of the genes situated at translocation breakpoints in childhood tumors has provided new insights into the mechanisms of malignant transformation. These observations have also allowed the development of molecular diagnostic assays, which have had a tremendous impact on the treatment of childhood cancer.