Point mutations in the RUNX1/AML1 gene: another actor in ... - Nature

Oncogene (2004) 23, 4284–4296

& 2004 Nature Publishing Group All rights reserved 0950-9232/04 $30.00 www.nature.com/onc

Point mutations in the RUNX1/AML1 gene: another actor in RUNX leukemia Motomi Osato*,1 1

Institute of Molecular and Cell Biology, Oncology Research Institute, National University of Singapore, Singapore, 30 Medical Drive, Singapore 117609

The RUNX1/AML1 gene is the most frequent target for chromosomal translocation in leukemia. In addition, recent studies have demonstrated point mutations in the RUNX1 gene as another mode of genetic alteration in development of leukemia. Monoallelic germline mutations in RUNX1 result in familial platelet disorder predisposed to acute myelogenous leukemia (FPD/AML). Sporadic point mutations are frequently found in three leukemia entities: AML M0 subtype, MDS-AML, and secondary (therapy-related) MDS/AML. Therapy-related leukemias resulting from anticancer treatments are not uncommon, and the incidence of RUNX1 point mutations appears comparable to the incidence of the t(8;21) AML M2 subtype and the inv(16) AML M4Eo subtype. Half of the point mutations in M0 cases are biallelic, although the frequency varies with ethnicity. Most of the RUNX1 mutations are clustered in the Runt domain and result in defective DNA binding but active b-subunit binding, which is consistent with three-dimensional structural findings and may explain the dominant inhibitory effects. Unlike the classical tumor suppressor genes requiring biallelic inactivation, haploinsufficient RUNX1 is apparently leukemogenic. However, RUNX1 abnormalities per se are insufficient to cause full-blown leukemia. Intensive investigation of cooperating genetic alterations should elucidate leukemic mechanisms. Oncogene (2004) 23, 4284–4296. doi:10.1038/sj.onc.1207779

located on the breakpoint of chromosomal translocation t(8;21) in 1991. Since then, leukemia research on AML1/ RUNX1 has been exclusively concentrated on investigating t(8;21) and resultant AML1(RUNX1)/ ETO(MTG8); t(3;21), RUNX1/EVI1; t(12;21), TEL (ETV6)-RUNX1; and inv(16), PEBP2b(CBFb)/ MYH11. In 1999, Osato et al. (1999) and Song et al. (1999)reported a RUNX1 point mutation in sporadic and familial myeloid leukemia, respectively. Subsequent studies have identified point mutations in acute myeloid leukemia (AML), predominantly in the M0 subtype, in different ethnic groups (Preudhomme et al., 2000; Langabeer et al., 2002; Matsuno et al., 2003; Roumier et al., 2003a, b; Silva et al., 2003). Until recently, mutations were believed to cluster within the Runt domain. However, mutations in the C-terminal region, outside the Runt domain, have also been identified, predominantly in MDS-AML (Harada et al., 2004). Finally, RUNX1 point mutations turned out to occur with comparable frequency to the first mode of RUNX1 alterations, namely, t(8;21), inv(16), and t(12;21), in secondary MDS/AML (Harada et al., 2003). Accordingly, the RUNX1 point mutation is fully recognized as the second mode of leukemogenesis in RUNX leukemia. In this review, I summarize the currently available cases of point mutations and discuss the leukemogenesis mediated by RUNX1 þ / status.

Keywords: RUNX; AML1; PEBP2; point mutation; familial leukemia Frequency and distribution of the point mutations

Introduction Since the discovery of the Philadelphia chromosome, t(9;22), in chronic myelogenous leukemia in 1960, leukemia researchers and physicians have paid much attention to the investigation of chromosomal translocations and resultant chimeric genes. It was also the case for RUNX1/AML1-related leukemia (also known as CBF leukemia, hereafter referred to as RUNX leukemia). Miyoshi et al. (1991) identified AML1 as a gene *Correspondence: Motomi Osato; E-mail: [email protected]

The RUNX1 gene consists of 10 exons (exons 1–6, 7A, 7B, 7C, and 8) and two promoters followed by the distinct initiation codon in exon 1 and exon 3. The proximal promoter-driven transcript starts in exon 3 and encodes 453 amino acids (aa) (full length 453), whereas the distal promoter-driven transcript from exon 1 encodes 480 aa (full length 480). Full length 480 contains an additional 32 aa instead of 5 aa from the full length 453 in its N-terminal end. Thus, both full lengths share exons 3–6, 7B, and 8 and nearly the complete coding sequence. The Runt domain, a highly conserved 128 aa protein motif responsible for both DNA binding and heterodimerization with the b-subunit PEBP2b/ CBFb spans aa 50–178 in full length 453 and is located on exons 3–5. All RUNX1 mutation-screening studies

Point mutations in the RUNX1/AML1 gene M Osato

4285

covered exons 3–5, and most included intron sequences flanking the exon. A few studies screened exons 1 and 2 as well, and half of the studies did not cover exons 6–8 encoding the C-terminal moiety where the transactivation domain spanning aa 291–371 is located. All types of human leukemia, except chronic lymphocytic leukemia (CLL), have been extensively investigated. In addition, a significant number of cell lines, including those established from solid tumors, as well as healthy volunteers were also included in the mutation screenings (Table 1)

(Langabeer et al., 2002; Vegesna et al., 2002; Harada et al., 2003; Taketani et al., 2003). A total of 102 cases out of the 1786 tested had a RUNX1 alteration including a point mutation and gene deletion. Most of the nucleotide changes resulted in amino-acid changes with clear functional defects as shown in Figure 1. However, some mutations failed to show any apparent functional alterations, and the nucleotide change may have been constitutive or leukemia specific. Nucleotide changes that do not lead to amino-acid change

Table 1 Frequency of mutations in the RUNX1/AML1 gene in various cancer patients and healthy volunteers Tumor

Subtype

Cases

Mutation

%

AML

M0 M1 M2 M2 unclassified M2 with t(8;21) M2 without t(8;21) M3 M4 M4 unclassified M4 with inv(16) M4 without inv(16) M5 M6 M7 Hypoplastic Unclassified

185 66 102 5 28 69 37 74 21 18 50 67 4 52 2 30

39 5 2 0 0 2 1 4 2 0 2 1 0 1 0 0

21.0 7.5 1.9 0.0 0.0 2.8 2.7 5.4 9.5 0.0 4.0 1.4 0.0 1.7 0.0% 0.0%

619

53

8.6%

34 81 138 53 56

1 1 22 8 1

2.9 1.2 15.9 15.0 1.8

Subtotal

362

33

9.1

ALL CML MPD

227 122 24 30 2 17 1

3 2 1 0 0 1 0

1.3 1.6 4.1 0.0 0.0 5.8 0.0

7 10 27

0 0 0

0.0 0.0 0.0

105 100 25 108

1 8 0 0

0.9 8.0 0.0 0.0

338

9

2.6

1786

102

5.8

Subtotal MDS

CMMoL RA/RARS RAEB/RAEBt MDS leukemia Unclassified

ET PV, MF

PNH DS TAM Thrombocytopenia Cell lines

Myeloid Lymphoid Solid tumors

Healthy

Japanese British French Dutch

Subtotal Total

Variationa 1:H58N, 1:L29S, 1:S21syn, 1: I87synb

1: I87syn

P157syn 1: G42R, 1: I87syn, T101syn 1: G42R

2: I87syn

1: H58N

1: G42R 3: S21syn, 5: L29S

6: L29S, 4: I87syn, 4: S21syn, 3: G42R, 2: H58N, 1: T101syn, 1: P157syn

a The variation category contains synonymous changes and mutations without functional abnormality. The number of the case is indicated ahead of the colon. bThe mutation is named according to the ‘Recommendation for a nomenclature system for human gene mutations’ (Antonarakis et al., 1998), for example, X stands for termination codon; syn, synonymous change. Abbreviations: AML, acute myeloid leukemia; MDS, myelodysplastic syndrome; CMMoL, chronic myelomonocytic leukemia; RA, refractory anemia; RARS, refractory anemia with ringed sideroblasts; RAEB, refractory anemia with excess of blasts; RAEBt, refractory anemia with excess of blast in transformation; ALL, acute lymphoid leukemia; CML, chronic myeloid leukemia; MPD, myeloproliferative disorder; ET, essential thrombocytemia; PV, polycytemia vera; MF, myelofibrosis; PNH, paroximal nocturnal hemoglobinurea; DS TAM, Down syndrome transient abnormal myelopoiesis

Oncogene

Point mutations in the RUNX1/AML1 gene M Osato

4286

(synonymous) may generate an abnormal splicing donor or acceptor site, resulting in a nonfunctional protein. These enigmatic cases are listed as ‘variation’ in Table 1. The numbers of mutations in Table 1 include variation cases. Most of the mutations are clustered within the Runt domain (Figure 1); N-terminal region, 9 (8.6%); Runt domain, 84 (80.8%); C-terminal region, 11 (10.5%). The total number of mutations (Table 1 and Figure 1) differs (104 vs 102) because the compiled mutants in Figure 1 include 11 cases from FPD/AML, which are not listed in Table 1. Conversely, gene deletion is not included in Figure 1, while a biallelic case in Table 1 provides two distinct mutations simultaneously. Functional consequence, haploinsufficiency Entire gene deletions and early truncated-type mutations (frameshift or nonsense mutations) were found with significant frequency. Most of the missense mutations also lost their DNA binding and transcriptional activities (Figure 1, right-middle column). Since the majority of cases are heterozygous, haploinsufficiency appears to be a basis for pathogenesis. This notion is most strongly supported by the whole gene deletion in FPD/AML. RUNX1 þ / status is likely to provide the cells with a growth advantage to overcome ‘selective pressure’ in the subject. However, if this were the only mechanism, mutations would be distributed throughout the gene without ‘hot spots’. Therefore, clustering in the Runt domain suggests that a mechanism other than genuine haploinsufficiency is important for leukemogenesis. Dominant negative effect by the retained b-binding activity In the Runt domain, eight regions or ‘hot spots’ were affected more than twice (Figure 1). Two different mechanisms are probably at least partly responsible. First, the primary sequence of particular sites seems to be preferably targeted by the mutation mechanism. R139, R174, and R177 are encoded by triplets containing CpG dinucleotides, which are extremely prone to transition-type mutations by a mechanism facilitated by methylation of the cytidine residue (Rodenhiser et al., 1997; Yoshida et al., 2003). Second, certain mutation sites are considered to provide cells a stronger growth advantage under ‘selective pressure’. Mapping of the missense mutations strongly supports this notion. Three regions recurrently affected by missense mutations exclusively correspond to three loop-containing regions responsible for DNA binding: bP1–loop(bP1–bB), bE’–loop(bE’–bF), and bP2–loop(bP2–C-terminus) (Figure 2) (Nagata et al., 1999; Osato et al., 2001; Tahirov et al., 2001). Consequently, these mutants are defective only in DNA binding, and not in packing or heterodimerization (Figure 1), which may explain the dominant negative effects of the RUNX1 mutants. In Oncogene

cotransfection assays, missense and some nonsense mutants showed inhibitory effects towards transactivation mediated by wild-type RUNX1. These mutants bound to PEBP2b/CBFb more efficiently than wild-type RUNX1 (Imai et al., 2000; Michaud et al., 2002). In addition, PEBP2b protects RUNX1 from ubiquitin– proteasome-mediated degradation (Huang et al., 2001). Thus, preferential binding of mutant RUNX1 proteins to PEBP2b may in turn increase protein half-life and hence further augment the dominant negative effects. In the right-end column of Figure 1, a general correlation can be seen between the retained b-binding activity and the inhibitory potential of each mutant. This correlation is more clearly indicated by the comparison of the distribution of missense mutations between RUNX1 and RUNX2 (Figure 2). Germline mutation of RUNX2 causes the bone disease cleidocranial dysplasia (CCD) (Mundlos et al., 1997). The missense mutations of RUNX2 distributed throughout the Runt domain affected the overall folding architecture in the cases marked by asterisks in Figure 2, resulting in impairment of both DNA- binding and heterodimerization activity (packing defect). The mutations underlined in Figure 2 showed defects in b binding alone. In clear contrast, no such mutations resulting in defective packing or b binding have been found in RUNX1 so far. (Some reported cases are mapped to the residues responsible for packing (suspected cases), but functional analyses have not yet been performed. No suspected case with defect in b binding alone has been reported.) Conversely, mutations mapping to the first DNA interacting region, bP1-loop (bP1-bB), are seen in RUNX1 but not in RUNX2 (Figure 2) (Yoshida et al., 2003). These RUNX1 mutants, R80C, K83N, and K83E showed strong affinity for the b-subunit and hence a strong dominant negative effect, suggesting that haploinsufficiency is a dominant mechanism in RUNX2 mutants, whereas a dominant negative effect is important in RUNX1 mutants. Sequestration of the b-subunit may be one important mechanism of the dominant negative effect caused by the RUNX1 mutants.

The lower the RUNX1 activity, the higher the leukemogenicity How does the dominant negative effect mentioned above relate to leukemogenesis? Although the limited Figure 1 Distribution and functional consequences of RUNX1 gene mutations. The mutations (bold) and disease subtypes are compiled to the left of the corresponding amino acid indicated at the center. Amino acids in the Runt domain in dark green in the green bar representing exons are listed in a consecutive manner, while those outside the Runt domain are not. Missense, termination, and frameshift mutations are highlighted by yellow, dark gray, and gray, respectively. The functional consequences, DNA binding, b-subunit binding, and dominant negative effects are listed on the right. þ indicates function retention. Normal and hypernormal functions are in pink and red, respectively. Dominant negative effects are shown in light blue or dark blue in the right-end column. The strong cases are in dark blue

Point mutations in the RUNX1/AML1 gene M Osato

4287

Oncogene

Point mutations in the RUNX1/AML1 gene M Osato

4288

insufficiency may be a common underlying mechanism in RUNX leukemias. Acquired trisomy 21 underscores this notion. RUNX1 þ / mutant alleles appear to be duplicated frequently. This finding may also lend supporting evidence to the hypothesis that reduced RUNX1 activity increases leukemogenic status. This argument will be discussed later.

Mutation status – disease phenotype in sporadic leukemias Biallelic mutation and AML M0 subtype

Figure 2 Structural mappings of the missense mutations detected in leukemia or the bone disease CCD. The RUNX1 mutants in leukemia (rectangular outline, orange) are mapped exclusively to the three regions involved in DNA binding, whereas the RUNX2 mutations in CCD (blue) are distributed all over the Runt domain except for the loop (bP1-bF). Mutations associated with packing (asterisk) and b-binding (underline) defects are found only in CCD. Adapted from Nagata et al. (1999)

number of individuals and minimal clinical information evaluated prohibit a robust conclusion, families with mutations acting simply via haploinsufficiency have a lower incidence of leukemia than families with mutations acting in a dominant-negative fashion. In two of the largest FPD/AML pedigrees, leukemia occurred at a markedly higher rate in the family with the strong dominant negative K83E mutation (57%) compared to the pedigree with a complete deletion of RUNX1 (24%) (Table 2) (Song et al., 1999; Michaud et al., 2002). Thus, low RUNX1 activity appears to correlate with high leukemogenicity. The hypothesis may be expanded to some extent to leukemogenesis induced by chimeric genes. The majority of current findings indicate that a major function of chimeric genes is a transdominant inhibition of wild-type RUNX1, resulting in severely decreased RUNX1 activity in affected cells. Currently, more than 10 partner genes fused with RUNX1 have been identified in chromosomal translocations. Although each chimeric gene may have its own leukemogenic mechanism, RUNX1 Oncogene

RUNX1 point mutations were most frequently found in the AML M0 subtype (Table 1). In addition, 22 of the 39 (56.4%) M0 cases showed the biallelic type of mutation (Table 3). Conversely, only two of the 80 RUNX1 mutation cases in other leukemia subtypes were the biallelic type. Therefore, it is clear that RUNX1 point mutations of the biallelic type are tightly associated with the AML M0 subtype. Mice heterozygous for the RUNX1-MTG8 or PEBP2b(CBFb)MYH11 chimeric genes showed defects in definitive hematopoiesis similar to Runx1/ mice, suggesting that the chimeric genes can act as strong transdominant inhibitors of the remaining normal Runx1 allele (Castilla et al., 1996; Okuda et al., 1996; Yergeau et al., 1997). Therefore, the leukemogenic potential of biallelic RUNX1 mutants is considered equivalent to those of chimeric genes. On the other hand, t(8;21), inv(16), and biallelic point mutations are related to the distinct M2, M4Eo, and M0 subtypes of myeloid leukemias, respectively. The different disease phenotypes are quite interesting. The contribution of each partner gene, thought to be considerable, may also be explained by the different residual activities of RUNX1 in the affected adult hematopoietic cells. Conditional Runx1 knockout mice showed apparent defects in hematopoiesis (Ichikawa et al., 2004), whereas hematopoiesis in conditional heterozygous RUNX1-ETO knockin mice was nearly normal (Higuchi et al., 2002), suggesting that Runx1/ induces more severe effects than the RUNX1-ETO chimeric gene in adult hematopoiesis. The M0 subtype shows minimally differentiated morphology, while the M2 subtype is characteristically well differentiated. Therefore, the differing residual activities of RUNX1 seem to be important. In addition, influence on the activities of other RUNX family genes must be considered because RUNX1-MTG8 or PEBP2b (CBFb)-MYH11 chimeric genes could inhibit RUNX2 and RUNX3 as well, while RUNX1/ status is a defect in RUNX1 alone. Careful study will resolve this intriguing issue. Of note, the biallelic-type mutation is strictly confined to the status with two copies of RUNX1. Biallelic mutation in trisomy 21 is not associated with the AML M0 subtype, but is associated with myeloid malignancies other than M0. This phenotypic difference will be discussed later.

Point mutations in the RUNX1/AML1 gene M Osato

4289 Table 2 Pedigree

Mutation

RUNX1 mutations in familial leukemia: FPD/AML

Mechanism

Leukemia incidence

Song et al. (1999), Dowton et al. (1985), Michaud et al. (2002) 1 Deletion Haploinsufficiency 7/29 (24%)

2

V91fsX101

Haploinsufficiency

1/5 (20%)

3 4 5 6

R177X R174X R174Q R139Q

Haploinsufficiency Dominant negative+ Dominant negative+ Dominant negative+

4/9 (44%) 4/10 (40%) 4/9 (44%) 1/4 (25%)

Michaud et al. (2002), Gerrard et al. (1991), Li et al. (2004) 1 K83E Dominant negative++

8/14 (57%)

Age of leukemia onset

Subtype of leukemia

Cytogeneticsa

8 10

AMoL

1:7, +marker Lymphosarcome

24 52

CML Lymphocytic lymphoma Leukemia MDS-AML

48 62 (mean 34) Other 10 cases suffered from cancers 1: 5q 41

MDS–AML M1

1:+8 41 75 6 23 7 (mean 30.4)

MDS/AML, AML Juvenile CMLAML AMMoL AMMoL Three cases: details not available

2

R135fs

Haploinsufficiency

4/9 (44%)

3

Y206X

Dominant negative+

3/13 (23%)

3:AML, 1:pernicious anemia 2:AML, 1:leukemia

?

4/7 (57%)

33

AML M1

33?

RAEB-M1 1:leukemia, 1: RAEB

1/5 (20%)

31

AML M2

1/3?

ALL

NA

Buijs et al. (2001) 1 D171Y

Walker et al. (2002) 1 A107P

?

Gabbeta et al. (1996), Sun et al. (2004) 1 V91fsX101 Haploinsufficiency Total

12

42/117 (35.8%)

1:5, 7

1:7

1:92,XXYY,add(7) (q31),add(7)(q31) 1:+21/ +8

1:del(7)(q22)

4: 7, 2: +8, 2: 5, 1: +21 among 15 available

a The number of the case is followed by a colon and the cytogenetic results. Abbreviations: AMMoL, acute meylomonocytic leukemia; NA, not available; for other abbreviations, see Table1 footnote

Entire distribution including the C-terminal moiety and MDS-AML/secondary leukemia The RUNX1 point mutations are also frequently found in MDS and subsequent overt leukemia (referred to in this review as MDS-AML) (Tables 1 and 4). The mutations are characteristically distributed, as they were also detected in the C-terminal region which is rarely targeted in other types of leukemias (Harada et al., 2004). Since this distribution was only recently identified, C-terminal muta-

tions were overlooked in a number of previous studies. Half of the studies did not investigate exons 7B and 8 in the C-terminal region. With regard to MDS-AML, the incidence covering both exons 3–6 and exons 3–8 were evaluated less frequently than those of exons 3–8 alone (Table 4). Thus, C-terminal mutations may be underestimated in MDS-AML. However, C-terminal mutations still seem rare in AML, since C-terminal mutations were not identified in the study with a relatively large number of cases. Therefore, RUNX1 point mutations including the Oncogene

4290

Oncogene

Table 3 RUNX1 mutations in AML M0 subtype Cases Mutation (%)a 8 (16%)

59

16 (27%)

41

5 (12%)

Variation

Screened region

Method

Ethnicity Functional Allelea analysis

Cytogeneticsa

Other mutationa

Comment

Reference

1:H58N

ex3–8

SSCP, NIRCA

Japanese Yes

2: bi-, 6: monoallelic

1: complex [+13, +8] 2: 20q–

Trend of higher WBC

Matsuno et al. (2003) and Osato et al. (1999)

ex3–5

SSCP

French

15: biallelic

5: normal 4: complex

5: FLT3 ITD or TK2 (63%) (10 of 49 tested had mutation) 4 FLT3 ITD (25%) (13 of 57 tested had mutation)

Higher WBC, Greater marrow blast Higher CD33

Roumier et al. (2003a, b), Preudhomme et al. (2000)

1:L29S,

No

ex3–5

SSCP

British

No

ex3–5

LOH

Dutch

No

1:S21syn 1 7

4 13 185

a

6 (35%)

2 (50%) 2 (15%) 39 (21.1%)

1:I87syn

ex3–6 ex3–8

SSCP SSCP

Japanese No Japanese Yes

1: biallelic with S21syn 5: abnormal [3: +13, 1:7] 4: bi-, 2: monoallelic 2: monoallelic 2: monoallelic 22/39 (56.4%): biallelic

[2: del(3), 2:del(5), 2:7, 1: der(13;13),1:21] [23 of 40 tested showed complex abnormalities] Lower incidence of complex karyotype? 1: normal

Higher IgH/TCR rearrangement

Langabeer et al. (2002)

1: normal 4:abnormal [1:3, 1: 7,1:+13 , 1: 21] 1: normal, 1: +7 2: normal No C/EBPa mutation 5: +13, 4: 7, 2: 20q, 1: +8 among 26 available

Silva et al. (2003)

Pediatric cases 1: secondary AML

Taketani et al. (2003) Harada et al. (2004)

The numbers of the cases are indicated ahead of the colon. Abbreviations: SSCP, single-strand conformation polymorphism; NIRCA, nonisotopic RNA cleavage assay: LOH, loss of heterozygosity; FLT3 ITD, FLT3 internal tandem duplication; TK2, mutation in tyrosine kinase domain 2; WBC, white blood cell. Mutation nomenclature, see Table 1, footnote

Point mutations in the RUNX1/AML1 gene M Osato

51

a

Point mutations in the RUNX1/AML1 gene M Osato

4291 Table 4 Tumor

Subtype

MDS

CMMoL

Cases Mutation Screened Method regiona 4

0

RA/RARS RAEB

46 41

1 6

RAEBt MDS leukemia

35 34

15 5

ex3–8

SSCP

MDS

MDS leukemia

Functional Allelea analysis

Cytogeneticsb

Commentb Reference

Japanese

Yes

19: normal

16: Harada et al. secondary (2004) MDS/ AML

26: monoalelic

13: abnormal [4: +8, 3: 7, 1: 5]

6

0

ex3–8

NIRCA

MDS

Unclassified

12

0

ex1–8

MDS

CMML

1

0

RAEB, RAEB-t

5

1

6

1

CMMoL RA RARS RAEB RAEBt MDS leukemia

27 13 4 23 27 0 94

CMMoL RA RARS RAEB RAEBt MDS leukemia

MDS

NA

MDS

MDS leukemia

Total

NA 33

Total

Total MDS

Ethnicity

160 26 (1: I87I, 2: G42R, T101syn, P157sync)

Total

MDS

RUNX1 mutations in MDS

Total

Japanese

Yes

—

—

—

Osato et al. (1999)

Direct German sequencing

—

—

—

5/7 cases alone

Ferrari et al. (2001)

ex1–8

SSCP

Spanish

No

1: monoallelic

1: normal

FLT3: WT

Carnicer et al. (2002)

0 0 0 0 0 0 0

ex3–5

SSCP

French

—

—

—

Preudhomme et al. (2000)

2 18 0 2 5 10 37

1 0 0 0 0 1 2

ex3–6

SSCP

Japanese

Yes

2: monoallelic

NA

Imai et al. (2000)

14

1

ex3–6

Direct sequence

NA (USA) No

NA

NA

Song et al. (1999)

3

2

ex3–6

SSCP

Japanese

1: bi, 1: monoallelic

1: +21, 1: 7q

Taketani et al. (2003)

30 2

0

No

4: +8, 4: 7/7q among 28 available

Total all studies CMMoL 34 RA/RARS 81 RAEB, RAEBt 138 MDS leukemia 53 Unclassified 56 Total 362

1 1 22 8 1 33

2.9% 1.2% 15.9% 15.0% 1.8% 9.1%

CMMoL 5 RA/RARS 46 RAEB, RAEBt 81 MDS leukemia 40 Unclassified 12 Total 184

0 1 22 5 0 28

0.0% 2.1% 27.2% 12.5% 0.0% 15.2%

exon 3–8 studies 1: P157syn 1: G42R, 1: I87syn, 1: T101syn, 1: G42R

a The screened exons are indicated. bThe numbers of the case are indicated ahead of the colon. cNomenclature of the mutations, see Table 1, footnote. Abbreviations: NA, not available; abbreviations for diseases, see Table 1, footnote

Oncogene

Point mutations in the RUNX1/AML1 gene M Osato

4292 Table 5 Subtype

AML M0 subtype MDS-AMLa Secondary MDS/ AMLb

AML M2 subtype

AML M4 subtype AML M4Eo subtype

Cases

Estimated frequency of leukemia with RUNX1 point mutation

Variation % of mutation in the category

% of the category in AML

% of mutation in AML

% of the category in leukemia

% of mutation in leukemia

Prognosis

185

39

21.1%

3–9% (6%)

0.6–1.8% (1.2%) 1.8–3.0%

0.4–0.5%

poor

121 42

27 15

15.2% 35.7%

— —

— —

7–10% 5–10%

1.1–1.5% 1.8–3.6%

poor poor

% of t(8;21) in M2 40%

% of M2 in AML 25–30%

% of t(8;21) in AML 12–15%

% of AML in leukemia 30–45%

% of t(8;21) in leukemia 4.2–6.3%

prognosis of t(8;21) leukemia good

% of inv(16) % of M4 or % of inv(16) in M4 or M4Eo M4Eo in AML in AML 20% 20% 7–12%

% of AML in leukemia 30–45%

% of inv(16) in leukemia 3.0–4.5%

prognosis of inv(16) leukemia good

95%

—

—

—

% of t(12;21) in childhood ALL 20–25%a

% of ALL in leukemia 10–13%a

% of t(12;21) in leukemia 1.0–1.5%

5%

Childhood ALL

prognosis of t(12;21) leukemia good

Calculation of the incidence was carried out based on several textbooks and public databases from CDC, WHO, and National Cancer Institute (NIH). aMDS-AML stands for MDS and subsequent AML. bSecondary MDS/AML represents therapy-related MDS or AML. Abbreviations for diseases, see Table 1, footnote

C-terminal moiety appear to be found predominantly in MDS-AML. Harada et al. (2003) from Hiroshima, one of the two cities suffered from atomic bomb, reported two further implications of RUNX1 mutations in secondary MDS or AML (referred to in this review as MDS/AML), including therapy-related and atomic bomb-related MDS/AML. Firstly, the relationship to radiation seems important in terms of the social issue. In addition to the victims in the above two cities, an estimated 220 000 US soldiers and a significant number of local residents and passengers have been irradiated in the aboveground nuclear weapon test sites at Nevada, Bikini atoll, and others in the late 1940s to 1950s. Many of these people are thought to have succumbed to MDS/AML caused by RUNX1 point mutation. Secondly, the involvement in therapy-related MDS/ AML seems to have more general impact. With the improvement of treatments for primary cancer, the incidence of therapy-related MDS/AML continues to increase. Therapy-related MDS/AML accounts for roughly 10–20% of all new cases of MDS and AML in standard hospitals. Therefore, the incidence of secondary MDS/AML with RUNX1 mutation is comparable to those of t(8;21) AML M2 subtype and inv(16) AML M4Eo subtype (Table 5). Considering its poor prognosis, clinical implication of RUNX1 point mutation in secondary MDS/AML would be rather more significant than we expected. Leukemic incidence rates were based on several textbooks and public databases from CDC, WHO, and National Cancer Institute (NIH). CLL accounts for 10–30%, which may explain the relatively low incidences of t(8;21), inv(16), and t(12;21). Oncogene

Acquired (not congenital) trisomy 21 and non-MO myeloid malignancy Preudhomme et al. (2000) first reported the relationship between trisomy 21 and RUNX1 point mutation (Table 6). Sporadic cases of myeloid but not lymphoid malignancy with trisomy 21 were associated with a high frequency of RUNX1 point mutations. Taketani et al. (2002) subsequently surveyed RUNX1 mutation in congenital cases of trisomy 21, that is, the Down syndrome-related hematopoietic diseases. However, point mutation in congenital cases appears to be rare except for the enigmatic H58N variation, which has nearly normal function. Allelic analysis of sporadic cases with acquired trisomy 21 showed the following combination: 2–3 mutant vs 0–1 wild type. The cooccurrence of mutation and trisomy 21 should be associated with the allelic combination with one mutant and two WT combination. However, this combination has never been found. These findings suggest that trisomy 21 increases the copy number of the mutant allele, which already exists in the status of two copies of RUNX1. In other words, trisomy 21 is thought to be a secondary change to RUNX1 þ / status. Trisomy 21 as a second hit will be discussed later. As mentioned earlier, biallelic RUNX1 mutations are tightly liked to the AML M0 subtype. However, interestingly, acquired þ 21 carrying a RUNX1 point mutation has never been observed in the AML M0 subtype, even in cases with biallelic mutations (Table 6). This ‘discrepancy’ is not explained by the differing residual activities of RUNX1, as half of the M0 cases with RUNX1 mutation are monoallelic. Leukemic phenotype may be determined by an as yet undetermined mechanism.

Point mutations in the RUNX1/AML1 gene M Osato

4293 Table 6 Trisomy 21 and RUNX1 mutation Disease

Acquired/congenital Case Mutation

AML, ET, CML ALL Down syndrome Down syndrome

Acquired Acquired Congenital Congenital

13 11 3 46

%

Screened region Ethnicity Allele

5a 38.4% ex3–5 0 0.0% ex3–5 0 0.0% ex3–5 b 2.1% ex3–6 1 (H58N)

French French French Japanese

2–3 mutants/0–1 WT — — heterozygous

Reference Preudhomme et al. (2000) Preudhomme et al. (2000) Preudhomme et al. (2000) Taketani et al. (2002, 2003)

Disease subtypes of the cases with mutation are as follows: aM1 post atypical CML, M2 (two cases), ET, atypical CML, bTAM. Disease abbreviations, see Table 1, footnote

Familial leukemia, FPD/AML Heterozygous germline mutation in the RUNX1 gene causes familial leukemia, familial platelet disorder with predisposition to acute myelogenous leukemia (FPD/ AML). A total of 12 pedigrees have been identified so far, mainly in individuals of European descent (Table 2). FPD alone, before the onset of AML, is not fatal but significantly inhibits blood clotting, which could be life threatening after surgery, injury, dental treatment, or childbirth. The incidence of leukemia among affected individuals varies from 20 to 50% (average 35%), and is highest in familial leukemia syndromes including Fanconi anemia, Bloom syndrome, and Kostman syndrome. The affected subjects seem to develop myeloid malignancies throughout their life span, but not later in life as described in some previous reports. Leukemia developed before the age of 10 years in four of 11 cases. These features of FPD/AML were compared with a similar dominant hereditary disease, Huntington disease (HD). The subjects of HD are healthy until disease onset at 30–50 years of age and die 10–15 years later. In addition to the low penetrance and/or low severity of symptoms, late onset is a key factor for hereditary disease, since patients often have children before disease symptoms appear. On the other hands, FPD/AML seems detrimental to the subjects in any factors of hereditary disease. How can such a pedigree full of drawbacks exist? FPD/AML may persist as a disease due to the easy-mutability of RUNX1. As discussed before, RUNX1 indeed contains easy-mutable primary sequences: triplets containing CpG dinucleotides extremely prone to transition-type mutations facilitated by methylation of the cytidine residue (Rodenhiser et al., 1997; Yoshida et al., 2003). This mechanism could work particularly well in germ cells. The FPD/AML pedigree may readily disappear in the human population, while a ‘de novo’ pedigree could be continually generated with comparable frequency. Consistent with this argument, of CCD cases caused by RUNX2 mutation, approximately half are the ‘de novo’ and not the familial type (Zhou et al., 1999). Irradiation readily generates CCD-like phenotypes in mice due to Runx2 alterations. RUNX2 shares the above noted easymutable sequences. In the case of RUNX1, healthy British subjects have two kinds of variations: L29S and S21syn with 5 and 3% frequency, respectively (Langabeer et al., 2002). Japanese subjects also have several recurrent enigmatic variations: G42R, I87syn, and

H58N. Interestingly, G42R was identified in both leukemia patients and healthy volunteers (Harada et al., 2003; Taketani et al., 2003). These variations, probably including polymorphisms, might reflect the above noted easy-mutability of RUNX1. As mentioned above, FPD/AML is extremely rare worldwide. Therefore, this clinical entity had been overlooked by physicians until recently. Bone marrow transplantaion was carried out from an affected sibling with mild FPD symptoms to a patient suffering from AML (Buijs et al., 2001). The recipient succumbed to leukemia again, clearly from the hematopoietic cells derived from the donor. Genetic alterations cooperating with RUNX1 þ / status As mentioned earlier, RUNX1/ status is frequently found in the AML M0 subtype. Thus, RUNX1 may appear to act as a classical tumor suppressor gene, requiring inactivation of two alleles according to Knudson’s two hits theory. However, biallelic inactivation is not required in all cases of leukemogenesis; in fact, the overall frequency of biallelic cases is relatively low (27 of 102 cases, or 26%). Biallelic cases of FPD/ AML have never been reported. Therefore, RUNX1 þ /  per se appears to be leukemogenic. The dominant negative effect associated with missense mutations might play a role in leukemogenesis, but not to the same extent as gene inactivation. In any case, heterozygous mutation appears to be a step towards leukemogenesis. Therefore, a mutation in the wild-type RUNX1 on the remaining allele is considered to be a second hit in addition to RUNX1 þ / status. Furthermore, acquired but not congenital trisomy 21 is also thought to be a second hit. As discussed earlier, the link between þ 21 and RUNX1 mutation is interpreted as follows; a mutation takes place in the two copy status and the mutant allele is duplicated in accordance with þ 21. In both RUNX1/  and þ 21 cases, acquirement of another mutant allele is therefore thought to be an additional genetic alteration cooperating with RUNX1 þ / status (Figure 3). In line with þ 21, cytogenetics helps to identify second hits. Patients with RUNX1 point mutations show the following concurrent abnormalities; 12 cases showed 7; 6, þ 8; 5, þ 13 (Tables 2–4). Loss of chromosome 7 is suspected to be a strong predisposing factor for the second hit in RUNX1 þ /. Oncogene

Point mutations in the RUNX1/AML1 gene M Osato

4294

FLT-3 mutations were frequently observed even in monoallelic RUNX1 mutations in M0 patients (Table 3) (Matsuno et al., 2003; Roumier et al., 2003a, b). Hence, the FLT-3 mutation is also a cooperating genetic alteration (Figure 3). FLT-3 mutation results in a constitutively active form known to stimulate cell proliferation. Therefore, differentiation blockade by RUNX1 alteration and growth stimulation by FLT3 mutation would synergistically contribute to a substantial fraction of AML M0 subtype cases. Interesting enough, FLT-3 mutations are known to be rare in the other RUNX leukemias, t(8;21) and inv(16). Certain RUNX leukemias seem to cooperate with distinct second hits. Elucidation of this mechanism might shed light on development of RUNX leukemias.

Current problems Region for mutation screening As discussed previously, half of the screening studies did not cover C-terminal exons 7B and 8 and therefore overlooked C-terminal mutations in MDS-AML. Cterminal screening is unlikely to reveal information about AML but is recommended. Likewise, exons 1 and 2 might be affected in particular types of leukemia. Theoretically, the promoter and untranslated regions could also be targeted. Variations Significant numbers of variations are reported, as shown in Table 1 and Figure 1. Functional analyses

Figure 3 Leukemogenesis from Runx1 þ /. Biallelic mutations of RUNX1 (RUNX1/) are tightly liked to the AML M0 subtype. However, acquired þ 21 carrying a RUNX1 point mutation (RUNX1 þ //) has never been observed in M0, despite cases with biallelic mutations. Mutation likely takes place in the two copy status (RUNX1 þ /) and the mutant allele is duplicated in accordance with þ 21, resulting in myeloid malignancies other than M0. FLT-3 mutations were frequently found in M0 cases with RUNX1 point mutations. These genetic alterations are mapped based on such observations Oncogene

have not yet been performed in all the cases. Each variation must be studied to determine if it is a mutation or a polymorphism. Analyses of polymorphisms might help to characterize susceptibility to leukemia. Dominant negative effect by C-terminal mutants in MDS-AML In the reporter assay, all C-terminal mutants found in MDS-AML showed inhibitory effects in a transdominant manner, like the missense mutants in the Runt domain (Harada et al., 2004). However, similar C-terminal RUNX2 mutations in CCD patients retain partial transactivation activity and show much milder phenotypes compared to the standard CCD phenotypes, which are caused by simple gene deletions (Yoshida et al., 2002). This discrepancy may be due to the distinct cell context or promoter used for the assay. However, a correlation with MDS but not AML might indicate that C-terminal mutants have milder inhibitory effects than Runt domain mutants. The biological assay, but not the reporter assay, may address this issue. Cooperative genetic alterations FLT-3 mutation and þ 21 are thought to be second hits in MDS/AML with RUNX1 mutations. High frequency of chromosome 7 loss suggests inactivation of the unknown gene on chromosome 7 plays a synergistic role with RUNX1 mutations. However, the overall frequency of these mutations does not account for all cases. Furthermore, to induce a leukemia, three to five genetic alterations are considered to be required based on current evidence. Therefore, a number of genetic alterations must be identified to understand the whole leukemogenic mechanism of RUNX leukemia. My laboratory recently established a mouse system that appears to be quite useful. Myeloid leukemia-prone mice mimicking FPD/AML were generated by backcrossing Runx1 þ / onto the BXH2 strain, in which retroviral insertional mutagenesis occurs spontaneously with the eventual development of myeloid leukemia in over 90% of the cases (Li et al., 1999). The proviral integration sites in the leukemias provide powerful genetic tags for disease gene identification. Therefore, collaborating oncogenes with Runx1 þ / would be readily detected. Loss-of-function of RUNX1 is widely believed to be the common underlying mechanism of RUNX leukemia. RUNX1 þ / status is considered to be the minimum alteration required for several forms of RUNX leukemias. Therefore, the mouse model for FPD/AML mentioned above should be useful for searching for cooperative genetic alterations in the other RUNX leukemias, t(8;21) and inv(16). These genetic alterations identified by the retroviral tagging system might explain the leukemogenic mechanisms not only of RUNX1 point mutations but also of other RUNX leukemias.

Point mutations in the RUNX1/AML1 gene M Osato

4295

Future directions Why is RUNX1 þ / status leukemogenic? This question seems to be the most fundamental question not only for RUNX1 point mutation but also all RUNX leukemias. Determining the means by which RUNX1 þ / is leukemogenic may not be possible by the currently used approaches to evaluate differentiation, proliferation, and apoptosis at the cellular level and transcriptional control at the molecular level. A novel approach is warranted. For example, alteration of

‘stemness’ should be investigated in RUNX1 þ / status. Stemness could provide cells with a time window sufficient for sequential genetic alterations. Intrinsic molecular functions relevant to leukemogenesis, other than transcriptional control, might be identified by novel approaches. Acknowledgements I thank Drs Hamish S Scott and Hironori Harada for sharing unpublished data. This study was supported by A*STAR (Agency for Science, Technology and Research), Singapore.

References Antonarakis SE and Nomenclature Working Group. (1998). Hum. Mutat., 11, 1–3. Buijs A, Poddighe P, van Wijk R, van Solinge W, Borst E, Verdonck L, Hagenbeek A, Pearson P and Lokhorst H. (2001). Blood, 98, 2856–2858. Carnicer MJ, Nomdedeu JF, Lasa A, Bellido M, Aventin A, Baiget M and Sierra J. (2002). Leukemia, 16, 2329–2332. Castilla LH, Wijmenga C, Wang Q, Stacy T, Speck NA, Eckhaus M, Marin-Padilla M, Collins FS, Wynshaw-Boris A and Liu PP. (1996). Cell, 87, 687–696. Dowton SB, Beardsley D, Jamison D, Blattner S and Li FP. (1985). Blood, 65, 557–563. Ferrari T, Weber B, Pils S, Harbott J and Borkhardt A. (2001). Ann. Hematol., 80, 72–73. Gabbeta J, Yang X, Sun L, McLane MA, Niewiarowski S and Rao AK. (1996). Blood, 87, 1368–1876. Gerrard JM, Israels ED, Bishop AJ, Schroeder ML, Beattie LL, McNicol A, Israels SJ, Walz D, Greenberg AH, Ray M and Israels LG. (1991). Br. J. Haematol., 79, 246–255. Harada H, Harada Y, Niimi H, Kyo T, Kimura A and Inaba T. (2004). Blood, 103, 2316–2324. Harada H, Harada Y, Tanaka H, Kimura A and T I. (2003). Blood, 101, 673–680. Higuchi M, O’Brien D, Kumaravelu P, Lenny N, Yeoh EJ and Downing JR. (2002). Cancer Cell, 1, 63–74. Huang G, Shigesada K, Ito K, Wee HJ, Yokomizo T and Ito Y. (2001). EMBO J., 20, 723–733. Ichikawa M, Asai T, Saito T, Yamamoto G, Seo S, Yamazaki I, Yamagata T, Mitani K, Chiba S, Hirai H, Ogawa S and Kurokawa M. (2004). Nat. Med., 10, 299–304. Imai Y, Kurokawa M, Izutsu K, Hangaishi A, Takeuchi K, Maki K, Ogawa S, Chiba S, Mitani K and Hirai H. (2000). Blood, 96, 3154–3160. Langabeer SE, Gale RE, Rollinson SJ, Morgan GJ and Linch DC. (2002). Genes Chromosomes Cancer, 34, 24–32. Li FQ, Person RE, Takemaru K, Williams K, Meade-White K, Ozsahin AH, Gungor T, Moon RT and Horwitz M. (2004). J. Biol. Chem., 279, 2873–2884. Li J, Shen H, Himmel KI, Dupoy AJ, Lagaespada DA, Nakamura T, Shaughnessy JD Jr, Jenkins NA and Copeland NG. (1999). Nat. Genet., 23, 348–353. Matsuno N, M O, Yamashita N, Yanagida M, Nanri T, Fukushima T, Motoji T, Kusumoto S, Towatari M, Suzuki R, Naoe T, Nishii K, Shigesada K, Ohno R, Mitsuya H, Ito Y and Asou N. (2003). Leukemia, 17, 2492–2499. Michaud J, Wu F, Osato M, Cottles GM, Yanagida M, Asou N, Shigesada K, Ito Y, Benson KF, Raskind WH, Rossier C, Antonarakis SE, Israels S, McNicol A, Weiss H, Horwitz M and Scott HS. (2002). Blood, 99, 1364–1372.

Miyoshi H, Shimizu K, Kozu T, Maseki N, Kaneko Y and Ohki M. (1991). Proc. Natl. Acad. Sci. USA, 88, 10431–10434. Mundlos S, Otto F, Mundlos C, Mulliken JB, Aylsworth AS, Albright S, Lindhout D, Cole WG, Henn W, Knoll JHM, Owen MJ, Mertelsmann R, Zabel BU and Olsen BR. (1997). Cell, 89, 773–779. Nagata T, Gupta V, Sorce D, Kim W-Y, Sali A, Chait BT, Shigesada K, Ito Y and Werner MH. (1999). Nat. Struct. Biol., 6, 615–619. Okuda T, van-Deursen J, Hiebert SW, Grosveld G and Downing JR. (1996). Cell, 84, 321–330. Osato M, Asou N, Abdalla E, Hoshino K, Yamasaki H, Okubo T, Suzushima H, Takatsuki K, Kanno T, Shigesada K and Ito Y. (1999). Blood, 93, 1817–1824. Osato M, Yanagida M, Shigesada K and Ito Y. (2001). Int. J. Hematol., 74, 245–251. Preudhomme C, Warot-Loze D, Roumier C, Grardel-Duflos N, Garand R, Lai JL, Dastugue N, Macintyre E, Deni C, Bauters F, Kerckaert JP, Cosson A and Fenaux P. (2000). Blood, 96, 2862–2869. Rodenhiser DI, Andrews JD, Mancini DN, Jung JH and Singh SM. (1997). Mutat. Res., 373, 185–195. Roumier C, Eclache V, Imbert M, Davi F, MacIntyre E, Garand R, Talmant P, Lepelley D, Lai JL, Casanovas O, Magnadie M, Mugneret F, Bilhou-Naberra C, Valensi F, Radford I, Mozziconacci MJ, Arnoulet C, Duchayne E, Dastugue N, Comillet P, Daliphard S, Gamache F, Boudjerra N, Jouault H, Fenneteau O, Pedron B, Berger R, Flandrin G, Fenaux P and Preudhomme C. (2003a). Blood, 101, 1277–1283. Roumier C, Fenaux P, Lafage M, Imbert M, Eclache V and Preudhomme C. (2003b). Leukemia, 17, 9–16. Silva FP, Morolli B, Storlazzi CT, Anelli L, Wessels H, Bezrookove V, Kluin-Nelemans HC and Micheline G-G. (2003). Oncogene, 22, 538–547. Song W-J, Sullivan MG, Legare RD, Hutchings S, Tan X, Kufrin D, Ratajczak J, Resende IC, Haworth C, Hock R, Loh M, Felix C, Roy D-C, Busque l, Kurnit D, Willman C, Gewirtz AM, Speck NA, Bushweller JH, Li FP, Gardiner K, Poncz M, Maris JM and Gilliland DG. (1999). Nat. Genet., 23, 166–175. Sun L, Mao G and Rao AK. (2004). Blood, 103, 948–954. Tahirov TH, Inoue-Bungo T, Morii H, Fujikawa A, Sasaki M, Kimura K, Shiina M, Sato K, Kumasaka T, Yamamoto M, Ishii S and Ogata K. (2001). Cell, 104, 755–767. Taketani T, Taki T, Takita J, Ono R, Horikoshi Y, Kaneko Y, Sako M, Hanada R, Hongo T and Hayashi Y. (2002). Leukemia, 16, 1866–1867. Oncogene

Point mutations in the RUNX1/AML1 gene M Osato

4296 Taketani T, Taki T, Takita J, Tsuchida M, Hanada R, Hongo T, Kaneko T, Manabe A, Ida K and Hayashi Y. (2003). Genes Chromosomes Cancer, 38, 1–7. Vegesna V, Takeuchi S, Hofmann WK, Ikezoe T, Tavor S, Krug U, Fermin AC, Heaney A, Miller CW and Koeffler HP. (2002). Leuk. Res., 26, 451–457. Walker LC, Stevens J, Campbell H, Corbett R, Spearing R, Heaton D, Macdonald DH, Morris CM and Ganly P. (2002). Br. J. Haematol., 117, 878–881. Yergeau DA, Hetherington CJ, Wang Q, Zhang P, Sharpe AH, Binder M, Marin-Padilla M, Tenen DG,

Oncogene

Speck NA and Zhang DE. (1997). Nat. Genet., 15, 303–306. Yoshida T, Kanegane H, Osato M, Yanagida M, Miyawaki T, Ito Y and Shigesada K. (2002). Am. J. Hum. Genet., 71, 724–738. Yoshida T, Kanegane H, Osato M, Yanagida M, Miyawaki T, Ito Y and Shigesada K. (2003). Blood Cells Mol. Dis., 30, 184–193. Zhou G, Chen Y, Zhou L, Thirunavukkarasu K, Hecht J, Chitayat D, Gelb BD, Pirinen S, Berry SA, Greenberg CR, Karsenty G and Lee B. (1999). Hum. Mol. Genet., 8, 2311–2316.