The Tumor Suppressor Gene hCDC4 Is Frequently Mutated in Human T-Cell Acute Lymphoblastic Leukemia with Functional Consequences for Notch Signaling

T-cell acute lymphocytic leukemia (T-ALL) is an aggressive malignant disease of T-cell precursors, accounting for 10% to 15% of pediatric and 25% of adult ALL cases. The pathogenetic background of T-ALL has mainly been studied in the context of acquired chromosomal translocations resulting in aberrant expression of a number of genes encoding transcription factors important in mammalian development (1). The molecular characterization of a rare t(7;9)(q34;q34.3) translocation that juxtaposes the COOH terminus of the transmembrane receptor Notch1 to the T-cell receptor β (TCRB) locus led to a detailed analysis of Notch1 deregulation in T-ALL (2). This translocation, leading to the expression of a truncated constitutively active form of Notch1 (2), was later shown to be a potent inducer of T-ALL in murine models (2). A more prevalent role for Notch1 in the pathogenesis of T-ALL was established following the discovery of frequent gain-of-function mutations in NOTCH1 (1, 3). The majority of mutations cluster in regions of the gene encoding the heterodimerization (HD) domain of NOTCH1, but the proline, glutamate, serine, and threonine (PEST) domain also harbors frequent truncating mutations (1–3).

Notch signaling is tightly regulated by a number of Notch-associated proteins (1) including the human homologue of SEL-10, hCdc4 (also known as Ago and Fbxw7), which has been shown to negatively regulate the intracellular domain (ICD) of Notch1 by mediating its ubiquitin-dependent proteolysis (4). In addition, Cdc4 knockout mice die in utero at embryonic day 10.5, manifesting elevated Notch ICD levels and marked abnormalities in vascular development (5, 6).

hCdc4 is a member of the F-box family of proteins, which are interchangeable substrate recognition subunits of SCF ubiquitin ligases (7). In the case of hCdc4, substrate recognition is mediated through interaction with eight adjacent WD40 protein binding motifs located in the COOH terminus (8). hCdc4 has also been shown to mediate the ubiquitin-dependent proteolysis of several other proteins important in regulating cellular division including cyclin E (9, 10), c-Myc, c-Jun, and Aurora-A (7). The role of hCdc4 as a negative regulator of several oncoproteins has suggested that it may function as a tumor suppressor in human malignancies. In support of this, mutations of hCDC4 have been detected in several human tumor types (11, 12), and inactivation of hCdc4 has been shown to lead to an increased genetic instability (11). Interestingly, although Cdc4 heterozygous mice do not develop spontaneous tumors (5, 6), heterozygous loss of both Cdc4 and p53 causes predisposition to development of a variety of epithelial tumors in mice (13).

In this report, we show that hCDC4 is frequently mutated in leukemic cells from children with T-ALL and that these mutations are associated with decreased Notch1 ubiquitylation and increased Notch1 protein stability. Finally, we find that patients with hCDC4 and/or NOTCH1 mutations have a favorable outcome compared with patients without alterations in these two components.

Patients and cell lines. Samples of cryopreserved lymphoblasts from 26 children with T-ALL and 20 patients with ALL of B-precursor phenotype, treated according to high-risk protocols prepared by the Nordic Society of Pediatric Hematology and Oncology (NOPHO; see Table 2 and ref. 14) at the Karolinska Hospital, Stockholm, Sweden, were included in this study (Table 1). The patients were treated according to NOPHO-ALL 86 for patients diagnosed 1988-1991 (n = 4), NOPHO ALL-92 for patients diagnosed 1992-2000 (n = 14), and NOPHO ALL-2000 for patients diagnosed from 2001 onwards (n = 8; ref. 14). Normal cells were obtained from peripheral blood or bone marrow following remission induced by chemotherapy. Leukemic and normal cells were purified as previously described (15). Approval was obtained from the Institutional Review Board for these studies. Malignant T-ALL and pre–B-ALL and other cell lines (Table 1) were cultured as described (9, 11, 16).

Mutation detection. Genomic DNA was prepared from leukemic and normal cells and derived cell lines using a commercial kit according to the manufacturer's instructions (Qiagen). Mutation analysis of NOTCH1 and hCDC4 was done as previously described (3, 12). PCR primers and single-strand conformational polymorphism (SSCP) conditions are available on request.

Notch1 stability and activity analysis. Preparation of protein extracts, electrophoresis, and Western blotting have been described (9, 12). For protein stability experiments, cells were treated with 100 μg/mL cycloheximide (Biomol) and 5 mmol/L EDTA for 15 min where indicated (17). NOTCH1 and hCDC4 expression constructs (4, 18), transient short interfering RNA (siRNA) transfections, hCDC4 expression analysis, and NOTCH1 gene reporter assays have been described (4, 18). In vitro mutation of hCdc4 Arg⁴⁶⁵, Arg⁴⁷⁹, and Arg⁵⁰⁵ was conducted using the in vitro mutagenesis QuickChange kit (Strategene). Antibodies to MYC (anti-9E10, Santa Cruz Biotechnology), hemagglutinin (anti-HA, Santa Cruz Biotechnology), and the cleaved, activated (ICD) form of Notch 1 (Cell Signaling) were used as recommended by the manufacturers. Tubulin antibodies (Sigma) were used as a control for loading. Horseradish peroxidase–conjugated antirabbit and antimouse (Cell Signaling) antibodies were used as secondary antibodies. Ectopically expressed hCdc4 was detected with a rabbit polyclonal antibody raised against the WD40 domains of hCdc4 (9).

Statistics. The significance of correlations between hCDC4 mutations, NOTCH1 mutation, and clinical variables was determined using Fisher's exact test (Table 2). The Kaplan-Meier method was used to generate graphs and estimate event-free survival. Differences in the prognosis between groups were evaluated using the log-rank test.

PCR-SSCP analysis was used to screen the entire coding region of hCDC4 in pediatric leukemic T- and B-lineage ALLs (12). No band shifts were observed in 20 B-lineage ALL leukemia samples analyzed (data not shown). However, several aberrantly migrating bands, indicative of mutations, were detected in 8 of 26 T-ALL specimens (Fig. 1A and data not shown). Sequence analysis showed that the specimens contained mutations that resulted in missense substitutions at codons Arg⁴⁶⁵ (three patients; exon 8), Arg⁴⁷⁹ (four patients; exon 9), or Arg⁶⁸⁹ (one patient; exon 11) in hCdc4 (Fig. 1A; Table 1). Codons Arg⁴⁶⁵ and Arg⁴⁷⁹ are localized to the surface of the β-propeller structure of hCdc4 (8), which is essential for recognition of phosphorylated cyclin E1 in mammalian cells (9–11) and the cyclin-dependent kinase inhibitor Sic1 in yeast (8). Arg⁶⁸⁹ is localized in the eighth β-propeller blade of hCdc4 (8) and is conserved in human, mouse, worm, and flies. This particular mutation has not been previously described in human cancers and it remains to be shown if it affects substrate binding or if it may destabilize the blades of the propeller in hCdc4. Importantly, hCDC4 mutations were absent in cells from matching nonmalignant leukocytes obtained from the same patients, indicating that the mutations were of somatic origin (data not shown). Together, these data show that >30% of patients with pediatric T-ALL harbor hCDC4 mutations, the highest mutation frequency reported in the hCDC4 gene in any human tumor type to date.

Additionally, we did mutational analysis of hCDC4 in a panel of cell lines derived from various hematologic malignancies, particularly of T-cell origin (Table 1; ref. 16). Nine malignant T-cell lines were screened for mutations, and in four of those, Jurkat, JM, CCRF-CEM, and P12-ISHIKAWA, hCDC4 was mutated (Table 1). Three T-cell lines (Jurkat, JM, and P12) had missense mutations that changed Arg⁵⁰⁵ of hCdc4, whereas CCRF-CEM had a mutation at Arg⁴⁶⁵ (Table 1), as previously reported (10). Thus, similar amino acid changes at critical arginine residues in the binding pocket of hCdc4 occur in both primary leukemic T cells and T-cell lines (Table 1). No hCDC4 mutations were identified in any of the other leukemia or lymphoma cell lines, with the exception of the multiple myeloma cell line U266-1984, which harbored an Arg⁵⁰⁵ missense mutation (Table 1).

Western blot analysis of lysates prepared from T-ALL cell lines with hCDC4 mutations indicated that these cells had elevated Notch1 ICD protein levels, as compared with T-ALL cell lines without mutation in either Notch1 or hCdc4 (Supplementary Fig. S1A). To specifically address whether elevated level of Notch1 is due to reduced degradation of the Notch1 ICD, we did cycloheximide-chase experiments in cell lines with specific genetic alterations in hCDC4 and NOTCH1. Indeed, cell lines with either NOTCH1 and/or hCDC4 mutations exhibited a reduced rate of Notch1 ICD degradation compared with cell lines without alterations in these genes (Fig. 1B; Table 1). These results were also recapitulated in T-ALL cells treated with the γ-secretase inhibitor DAPT to avoid potential differences in the rate of ICD production among these cell lines (Supplementary Fig. S1B). We next analyzed Notch1 ICD degradation in human embryonic kidney HEK 293, breast cancer MCF-7, and colon carcinoma HCT 116 cells following Notch1 activation by EDTA treatment. Notch1 ICD was rapidly degraded in all cell lines, consistent with the wild-type NOTCH1 and hCDC4 genotypes in these cells (Fig. 1B and data not shown). In contrast, HCT 116 hCDC4-negative cells (11) exhibited a clear stabilization of Notch1 ICD compared with the parental HCT 116 cells (Fig. 1B). Notably, silencing of hCDC4 using siRNA duplexes also significantly reduced the rate of turnover of Notch1 ICD in both MCF-7 and HEK 293 cells (Fig. 1C and data not shown), consistent with the proposed role of hCdc4 in degradation of Notch1 ICD. As expected, Notch1 ICD was rapidly degraded in pre–B-ALL cells that do not carry hCDC4 and/or NOTCH1 mutations (data not shown). Finally, to directly address whether the specific hCdc4 arginine mutants identified in T-ALL tumors affect the formation of multiubiquitin chains on Notch1 ICD, we did in vivo ubiquitylation assays in hCdc4-negative cells reconstituted with either a wild-type hCdc4 allele or with Arg⁴⁶⁵ and Arg⁵⁰⁵ mutant versions of hCdc4. As shown in Fig. 1D, each of the mutant alleles was severely deficient in multiubiquitylating Notch1 ICD as compared with wild-type hCdc4.

Given the presence of mutations in both NOTCH1 and hCDC4 in T-ALL cell lines (Table 1), we next explored whether both genes are simultaneously mutated also in primary T-ALL samples. As can be seen in Table 1, leukemic samples from eight patients contain NOTCH1 mutations in the HD domain alone (four patients), the PEST domain alone (one patient), or both domains (three patients). Two patient samples with NOTCH1 mutations were also found to contain hCDC4 mutations (Table 1). In concordance with previous studies (3), mutations in NOTCH1 were not observed in any of the B-ALL cases (data not shown). A total of 11 NOTCH1 mutations were detected in 8 of 26 T-ALL specimens (Table 1), a mutation frequency comparable with that previously described (19, 20). The NOTCH1 mutations identified in this study were similar to previously reported (3, 20) “hotspots” within the HD domain and truncating insertions and deletions in the PEST domain (Table 1). Interestingly, one tumor (patient #26) was found to harbor a NOTCH1 missense mutation that results in a Thr to Met substitution at position 2,484 in the PEST domain (Table 1), and point mutations in the PEST domain that are not frameshift or nonsense mutations are known to be infrequent in T-ALL (3, 20). Several hCdc4 target substrates, including cyclin E1 and c-Myc, contain a so-called hCdc4 phosphodegron motif (CPD), which is essential for interaction with SCF^hCdc4 and their subsequent ubiquitylation (8). Of note, the amino acids surrounding Thr²⁴⁸⁴ of Notch1 resemble a consensus CPD motif, suggesting that the T2484M PEST mutation might prevent its recognition by hCdc4. However, several truncating PEST domain mutations occur downstream of this motif (Table 1), suggesting that multiple CPD motifs may be present in Notch1 ICD and regulate its degradation (21).

Because one function of hCdc4 is the regulation of ubiquitylation and degradation of Notch1 ICD, we also wanted to examine the influence of the hCdc4 arginine mutants on Notch1 signaling. Introduction of a synthetic Notch1-sensitive reporter construct into HCT 116 hCDC4-negative cells, together with a wild-type hCdc4 construct, caused a significant reduction in Notch1 activity (Fig. 2A; Supplementary Fig. S2A). In contrast, no suppression in activity was observed in cells transfected with any of the three hCdc4 arginine mutants (Arg⁴⁶⁵, Arg⁴⁷⁹, or Arg⁵⁰⁵; Fig. 2A). Similar data were obtained in U2OS cells (Fig. 2B), showing that these mutants cannot prevent the immediate Notch1 downstream response. Two patients carried mutations both in Arg⁴⁷⁹ of hCDC4 accompanied by an HD mutation or an HD mutation plus PEST domain truncation in NOTCH1 (Table 1). A similar mutation pattern was also observed in T-ALL cell lines (Table 1). Whether NOTCH1 and hCDC4 mutations elevate Notch1 signaling directly in a synergistic fashion (as shown for NOTCH1 HD and PEST mutations; ref. 3) is not yet known. Conflicting data exist about the prognostic effect of NOTCH1 mutations in T-ALL (19, 20). Similar to the findings of Breit et al. (20), our data indicate that NOTCH1 mutations predict a favorable long-term survival in children with T-ALL (Table 2). Interestingly, the analysis including either Notch1 and/or its negative regulator, hCdc4, shows an even stronger association with favorable outcome (Table 2; Fig. 2C). If these data are confirmed in a larger study, the combined mutational status of NOTCH1 and hCDC4 could potentially serve as an important tool for refined risk stratification in T-ALL. Although T-cell disease is a high-risk criterion in many protocols for the treatment of childhood ALL, this risk factor has been questioned and is in some protocols modified by the combination of other risk factors and/or treatment response. NOTCH1 and hCDC4 analyses may in the future be helpful in the selection of T-cell ALL patients for less intensive therapy.

In conclusion, this study lends support to the notion that deregulated Notch signaling is important for T-ALL development and broadens this concept to directly include mutations not only in the NOTCH1 receptor but also in a negative regulator of Notch, the hCDC4 gene. The finding that two genes in the Notch1 signaling pathway are both frequently mutated in T-ALL emphasizes the importance of Notch1 signaling in T-cell leukemogenesis. A more thorough understanding of the causes and consequences of Notch1 deregulation may also open up new vistas for future therapeutic interventions and allow for a more refined risk stratification in this highly aggressive disease.

Grant support: The Swedish Cancer Society, the Children Cancer Foundation, the Swedish Research Council, Karolinska Institute Foundations, Gustaf the Vth Jubilee fund, the Cancer Society of Stockholm, the Foundation for Strategic Research (CEDB), the European Union (EuroStemCell), and the University of California Tobacco-Related Disease Research Program. Drs. N. von der Lehr and S. Akhondi were supported by fellowships from the Children Cancer Foundation and the Swedish Institute, respectively.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank B. Vogelstein and K. Kinzler (ref. 11; Johns Hopkins University, Baltimore, MD, and Howard Hughes Medical Institute, Chevy Chase, MD) for kindly providing hCDC4-deficient HCT 116 cells and parental control cells, Steve Reed (The Scripps Research Institute, La Jolla, CA) for kindly providing hCdc4-specific antibodies, and A.C. Björklund and M.T. Dagnell for technical support.