The distribution of MLL breakpoints correlates with outcome in infant acute leukaemia

Authors

  • Mariana Emerenciano,

    1. Paediatric Haematology-Oncology Program, Research Centre, Instituto Nacional de Câncer, Rio de Janeiro, Brazil
    Search for more papers by this author
  • Claus Meyer,

    1. Institute of Pharmaceutical Biology/ZAFES/Diagnostic Centre of Acute Leukaemia (DCAL), Goethe-University of Frankfurt, Frankfurt/Main, Germany
    Search for more papers by this author
  • Marcela B. Mansur,

    1. Paediatric Haematology-Oncology Program, Research Centre, Instituto Nacional de Câncer, Rio de Janeiro, Brazil
    Search for more papers by this author
  • Rolf Marschalek,

    1. Institute of Pharmaceutical Biology/ZAFES/Diagnostic Centre of Acute Leukaemia (DCAL), Goethe-University of Frankfurt, Frankfurt/Main, Germany
    Search for more papers by this author
  • Maria S. Pombo-de-Oliveira,

    Corresponding author
    • Paediatric Haematology-Oncology Program, Research Centre, Instituto Nacional de Câncer, Rio de Janeiro, Brazil
    Search for more papers by this author
  • The Brazilian Collaborative Study Group of Infant Acute Leukaemia

    Search for more papers by this author
    • The Brazilian Collaborative Study Group of Infant Acute Leukaemia members are listed in Appendix I as co-authors.

Correspondence: Maria S. Pombo-de-Oliveira, Paediatric Haematology-Oncology Program, Research Centre, Instituto Nacional de Câncer - INCA, Rua André Cavalcanti, 37. Rio de Janeiro/RJ – Brasil. E-mail: mpombo@inca.gov.br

Summary

Acute leukaemia in early childhood - and mainly infant leukaemia (IL) – is characterized by acquired genetic alterations, most commonly by the presence of distinct MLL rearrangements (MLL-r). The aim of this study was to investigate possible correlations between clinical features and molecular analyses of a series of 545 childhood leukaemia (≤24 months of age) cases: 385 acute lymphoblastic leukaemia (ALL) and 160 acute myeloid leukaemia (AML). The location of the genomic breakpoints was determined in a subset of 30 MLL-r cases. The overall survival of the investigated cohort was 60·5%, as determined by the Kaplan-Meier method. Worse outcomes were associated with age at diagnosis ≤6 months (< 0·001), high white blood cell count (= 0·001), and MLL-r (= 0·002) in ALL, while children with AML displayed a poorer outcome (= 0·009) regardless of their age strata. Moreover, we present first evidence that MLL-r patients with poor outcome preferentially displayed chromosomal breakpoints within MLL intron 11. Based on the literature, most MLL-r IL display a breakpoint localization towards intron 11, which in turn may explain their worse clinical course. In summary, the MLL breakpoint localization is of clinical importance and should be considered as a novel outcome predictor for MLL-r patients.

Acute leukaemia (AL) is one of the most common malignancies of early childhood, and is associated with a high incidence of early death during the perinatal period. In some cases, AL is present at birth and neonates die shortly thereafter, while other newborns develop normally for a few months and clinical and haematological problems appear later (Ross et al, 1994; Biondi et al, 2000). It is possible to correlate these differences in AL progression with variable genetic markers that are associated with the age of individual AL patients. MLL gene rearrangements (MLL-r) are a remarkable example: they constitute the most frequently identified genetic factor correlated with the development of acute lymphoblastic leukaemia (ALL) or acute myeloid leukaemia (AML) in early infancy (≤12 months of age). Despite advances in most other age groups, the prognosis of infants remains poor (Pieters et al, 2007; Balgobind et al, 2009). Therefore, MLL-r is also a hallmark for epidemiological studies in infant leukaemias (ILs). Molecular studies have demonstrated that these critical molecular lesions occur in utero in early haematopoietic precursors (Ford et al, 1993; Greaves & Wiemels, 2003), and that maternal exposures during pregnancy may be associated with infant and childhood leukaemias (Alexander et al, 2001; Pombo-de-Oliveira & Koifman, 2006).

The extraordinary diversity and distribution of MLL-r in ALL versus AML, and in infants versus paediatric or adult patients, suggest different associated biological mechanisms that may also reflect diverging prognostic responses. Extensive analyses at the molecular level have allowed the identification of 71 different MLL fusion partners to date (Meyer et al, 2009; Marschalek, 2011). Efforts have been made to associate immunobiological patterns with different fusion partner genes and age at diagnosis (Jansen et al, 2007; Mann et al, 2007); however, it remains unclear which molecular mechanisms or which exposures to external poisons contribute to the biological differences between MLL-r in children of different age strata. Some questions are less explored, such as (i) the predictive value of MLL-r in children aged between 13 and 24 months (ii) whether interethnic differences and/or differences in environmental exposure would influence the frequency of MLL-r-associated leukaemias; and (iii) whether differences in genetic susceptibility have pharmacogenomic implications. In this study, we analysed a series of patients enrolled in the Brazilian Collaborative Study Group of Infant Acute Leukaemia (BCSGIAL) in order to evaluate the frequencies of different MLL translocations and their association with demography, clinical and laboratory findings, and overall survival (OS).

Materials and methods

Subjects

Patients were gathered from 24 cities in the South, Southeast, Northeast, and Middle West regions of Brazil, covering both urban and rural areas, except the Amazon region. The BCSGIAL has addressed questions related to maternal environmental substance exposures during pregnancy, immunological factors, and genetic susceptibility associated with MLL-r leukaemia in Brazilian children (Pombo-de-Oliveira & Koifman, 2006; Zanrosso et al, 2010; Gonçalves et al, 2012). Additionally, the BCSGIAL studies include an accurate diagnosis of biomarkers to adjust treatment strategies and follow-up for children (Emerenciano et al, 2006; Pombo-de-Oliveira et al, 2009).

Data collection and laboratory procedures were evaluated and approved by the Ethics Committee of all participating hospitals. Data analysis was approved by the Research Ethics Committee, Instituto Nacional de Câncer (studies CEP #005/06 and #024/10; CONEP # 707/2010).

Leukaemia diagnosis

Diagnostic specimens of bone marrow aspirates and/or peripheral blood cells from children in early childhood (≤24 months of age) were assigned to this study from January 2000 to January 2011. Samples from participants with ALL or AML were diagnosed prior to any treatment. The diagnosis was first established through morphological examinations of lymphoid and myeloid cells according to standard criteria. Depending on the immunophenotype, AL was categorized as precursor B-lineage ALL (B-I, B-II, B-III), T-ALL (T-I, T-II, T-III, T-IV), or AML (as defined by the World Health Organization classification and revised in Bain (2006). The immunophenotype was assessed by flow cytometry as outlined in detail previously (Emerenciano et al, 2011) using a FACS Calibur (Becton, Dickinson and Company, San Diego, CA, USA) and the CellQuest software package (Becton, Dickinson and Company). Subsequently, specimens were screened for somatic genetic abnormalities commonly observed in childhood ALL (ETV6-RUNX1, TCF3-PBX1, BCR-ABL1, STIL-TAL1, TLX3, MLL-r) and in AML (RUNX1-RUNX1T1, CBFB-MYH11, PML-RARA, MLL-r) (van Dongen et al, 1999). Childhood leukaemia associated with Down syndrome, monosomy 7, Fanconi anaemia, Bloom syndrome, Ataxia telangiectasia, neurofibromatosis, myelodysplasia, and children older than 24 months were excluded.

Detection of MLL rearrangements

Detection of an MLL-r was performed by conventional cytogenetics, reverse transcription polymerase chain reaction (PCR), and/or by fluorescence in situ hybridization (FISH; LSI MLL Dual Color Break Apart Rearrangement Probe, Vysis Inc., Downers Grove, IL, USA) as previously described (Emerenciano et al, 2006). Long distance inverse PCR (LDI-PCR) was used to identify the MLL fusion partners and the respective breakpoints. Briefly, 1 μg of genomic DNA was digested and the resulting DNA fragments were self-ligated. This re-ligated DNA was used for the subsequent LDI-PCR analysis. PCR amplimers were purified from the gel and subsequently sequenced to obtain chromosomal breakpoint information (Meyer et al, 2005).

Statistical analysis

The children recruited in the study received different chemotherapy treatments in accordance with the Brazilian ALL treatment protocol group (GBTLI-ALL93, GBTLI-ALL99), Berlin-Frankfurt-Munster (BFM), or the INTERFANT-based protocols (Schrappe et al, 2000; Pieters et al, 2007; Brandalise et al, 2010). Therefore, it was not possible to evaluate each biomarker with respect to the different treatment protocol strategies. To compare the frequency of MLL-r between distinct groups (for example, age strata, ALL versus AML patients, Whites versus Non-whites) the χ2-test (two-sided) was used (or Fisher's Exact Test when expected values were less than five). P-values ≤0·05 were considered statistically significant. The disease risk associated with MLL-r occurrence in ALL or AML patients was determined according to demographic and laboratory findings by calculating odds ratios (ORs) with 95% confidence interval (CI) and unconditional logistic regression to adjust for potential confounding variables. A multivariate logistic regression model (method enter) was used to analyse associations between MLL genomic breakpoint status (intron 11 vs. others) and established demographic-clinical parameters [i.e. age at diagnosis and white blood cell (WBC) counts]. OS was determined in months from the diagnosis to outcome (death, alive, or last follow-up) using the Kaplan-Meier method. Patients lost to follow-up were censored at their date of last known contact. The differences between AL survival distributions were compared using the log-rank test. The multivariate Cox proportional hazard regression method was used to determine the independent prognostic factors influencing OS. All statistical analyses were performed using the Statistical Product and Services Solutions statistical package, version 18·0 (SPSS Inc, Chicago, IL, USA).

Results

Demography and immunophenotyping

A total of 545 children (385 ALL and 160 AML) were included in this study. They were first analysed by correlating age at diagnosis with leukaemia subtypes and MLL status (Table 1). IL patients (≤12 months old) and AL patients aged 13–24 months were matched regarding demography. The majority of samples (55%) displayed high WBC counts ranging from 1·1 −965·0 × 109/l.

Table 1. Demographic and clinical characteristics of patients according to age at diagnosis
 n (%) P
All cases n = 545≤6 months= 1237–12 months n = 15413–24 months n = 268
  1. ALL, acute lymphoblastic leukaemia; AML, acute myeloid leukaemia.

  2. a

    As defined by the mother of the index.

  3. b

    Includes two Japanese descendants, two Indians and six without information.

  4. c

    Corresponds to the place of birth of the index.

  5. d

    In 8 ALL patients (2·1%), it was not possible perform the immunophenotype sub-classification, given that 3 patients were aged ≤12 months and 5 were 13–24 months of age.

  6. e

    MLL gene status was indeterminate in 116 cases (21·3%): 46 in the age group ≤12 months and 70 in the 13–24 months group.

Gender
Male302 (55·4)65 (52·8)89 (57·8)148 (55·2)0·710
Female243 (44·6)58 (47·2)65 (42·2)120 (44·8) 
Race/skin coloura
White347 (63·7)87 (70·7)94 (61·0)166 (61·9)0·036
Non-White188 (34·5)31 (25·2)57 (37·0)100 (37·3) 
Otherb10 (1·8)5 (4·1)3 (1·9)2 (0·7) 
Region of Brazilc
South56 (10·3)11 (8·9)21 (13·6)24 (9·0)0·688
Southeast232 (42·6)58 (47·2)62 (40·3)112 (41·8) 
Northeast171 (31·4)36 (29·3)49 (31·8)88 (32·8) 
Middle-West80 (14·7)18 (14·6)22 (14·3)40 (14·9) 
Leukaemia subtype
ALL385 (70·6)87 (70·7)100 (64·9)198 (73·9)0·152
AML160 (29·4)36 (29·3)54 (35·1)70 (26·1) 
ALL subtyped
Pro-B167 (44·3)71 (82·5)54 (55·1)42 (21·8)<0·001
c-ALL178 (47·2)14 (16·3)35 (35·7)129 (66·8) 
Pre-B9 (2·4)0 (0·0)1 (1·0)8 (4·1) 
Pro-T23 (6·1)1 (1·2)8 (8·2)14 (7·3) 
AML subtype
M05 (3·1)1 (2·8)2 (3·7)2 (2·9)0·279
M1-M221 (13·1)6 (16·7)4 (7·4)11 (15·7) 
M35 (3·1)0 (0·0)0 (0·0)5 (7·1) 
M4-M598 (61·3)23 (63·8)37 (68·5)38 (54·3) 
M6-M731 (19·4)6 (16·7)11 (20·4)14 (20·0) 
MLL gene statuse
MLL rearranged176 (41·0)59 (59·0)63 (48·1)54 (27·3)<0·001
MLL wild-type253 (59·0)41 (41·0)68 51·9)144 (72·7) 

Leukaemia subtype was successfully classified in 537 out of 545 (98·5%) samples; the immunophenotype of 8 ALL patients could not be analysed. B precursor-ALL (= 354) and T-ALL (= 23) comprised 91·9% and 5·9% of ALL cases, respectively. The B precursor-ALL series included 167 cases displaying a CD10 negative pro-B phenotype (B-I), while the remaining cases with CD10 positive had common ALL (B-II, = 138) or pre-B ALL (B-III, = 9) phenotypes. T-ALL cases included T-I (= 13), T-II (= 5), and T-III (= 5) subtypes (Table 1).

AML cases (= 160) represented 29·4% of the total series; the M4-M5 subgroups accounted for the majority of these cases (61·3%), while 19·4% were M6-M7, 13·1% were M1-M2, 3·1% were M0, and 3·1% were M3.

MLL rearrangements and fusion partner genes

In 429 out of 545 children with AL (78·7%), the presence or absence of a rearranged MLL allele was successfully analysed. Table 1 shows that 41% of the patients investigated displayed a rearranged MLL allele (= 176). In addition, MLL-r were significantly more frequent among patients aged ≤6 months than among patients aged >12 months at diagnosis (< 0·001). Most of the MLL-r cases displayed a pro-B ALL phenotype (56·2% of the overall samples and 75·6% of the ALL samples). MLL-r were detected in 27·3% (= 54) of patients older than 12 months. Of these 54 patients, 37 had ALL (68·5%) and 17 had AML (31·5%). Of patients younger than 12 months with MLL-r, 77% had ALL and 23% had AML. In 80 cases, other recurrent chromosomal alterations frequently found in childhood leukaemia were detected. These aberrations were considered mutually exclusive to MLL-r. Thirty-six cases had inappropriate biological material to perform the analysis.

As summarized in Supplementary Table 1, IL were more prone to harbouring an MLL-r, while children older than 12 months displayed less rearrangements (OR 0·26; 95% CI 0·15–0·43). This association was statistically significant when infant ALL cases were compared to childhood ALL cases (OR 0·13; 95% CI 0·06–0·24). Vice-versa, no such association was evident among children of the same age strata and diagnosed with AML (OR 1·15; 95% CI 0·43–3·03). Females were more likely than males to present with MLL-r (OR 1·55; 95% CI 1·05–2·29); this increased risk for females was higher in AML patients (OR 2·20; 95% CI 1·04–4·64) and not statistically significant in ALL patients. When analysed by place of birth, only ALL patients from the Northeast region exhibited a three-fold higher of MLL-r (OR 3·25; 95% CI 1·36–7·73) compared to ALL patients from the South or AML patients. Among patients with MLL-r, a higher WBC was more strongly associated with ALL (OR 3·20; 95% CI 1·94–5·29) than with AML (OR 2·19; 95% CI 1·05–4·57).

The precise translocation partner gene (TPG) of the rearranged allele was elucidated at the molecular level for 148 patients (84·1%). For 28 patients, no sufficient amount of genomic DNA material was left for further molecular analysis. Three TPG types accounted for 77·8% (= 137) of all MLL-r leukaemia samples: 85 patients (48·3%) were diagnosed with an MLL-AFF1, 27 (15·3%) with an MLL-MLLT3, and 25 (14·2%) with an MLL-MLLT1 fusion gene. The remaining 11 cases (6·2%) were diagnosed with the following MLL TPGs: MLLT10 (= 4), ELL (= 3), MLLT4 (= 1), EPS15 (1p32; = 1), NEBL (3q21; = 1), and SEPT6 (17q23; = 1) (Table 2). The MLLT3 partner gene was more frequently observed in males (63·0%, vs. 37·0% in females); the other TPGs presented with equal gender distributions. The presence of the MLL-AFF1 fusion gene was strongly associated with a pro-B ALL immunphenotype. The co-existence of MLL-AFF1 and AFF1-MLL fusion transcripts was observed in 71·1% of t(4;11) cases and, in 80·7% of these cases, a high WBC was observed (= 0·029). Patients diagnosed with both an MLL-r and T-ALL subtype were exclusively associated with the MLL-MLLT1 fusion gene. In AML, MLLT3 was the most common TPG, accounting for 44·5% of all AML-positive cases. All MLL-r patients were associated with a high WBC count. In case of MLL-MLLT1 fusions, a greater frequency of hyperleucocytosis was observed.

Table 2. Demographical and laboratorial characteristics of patients according to MLL rearrangements and specific fusion partners
 MLL rearranged (%)
All (n = 176) AFF1 (n = 85)MLLT3 (n = 25)MLLT3 (n = 27)Othera(n = 39)
 TotalbTranscripts verifiedc (n = 45) 
  MLL-AFF1 MLL-AFF1/AFF1-MLL  
  1. ALL, acute lymphoblastic leukeamia; AML, acute myeloid leukaemia; WBC, white blood cell.

  2. a

    *Includes MLLT10 (n = 4), ELL (n = 3), MLLT4 (n = 1), EPS15 (1p32; n = 1), NEBL (n = 1), SEPT6 (n = 1) and unidentified MLL translocation partner gene (n = 28).

  3. b

    Total of all patients with AFF1 as MLL translocation partner gene.

  4. c

    Patients with enough biological material were further investigated for the presence of both reciprocal MLL-AFF1 and AFF1-MLL transcripts and the differences were statistically tested.

Age at diagnosis (months)
≤6 months59 (33·5)33 (38·8)7 (53·8)13 (40·6)10 (40·0)6 (22·2)10 (25·7)
7–12 months63 (35·8)28 (71·8)3 (23·1)12 (37·5)10 (40·0)9 (33·4)16 (41·0)
13–24 months54 (30·7)24 (28·2)3 (23·1)7 (21·9)5 (20·0)12 (44·4)13 (33·3)
 = 0·623 
Gender
Male88 (50·0)39 (45·9)6 (46·1)12 (38·7)12 (48·0)17 (63·0)20 (51·3)
Female88 (50·0)46 (54·1)7 (53·9)19 (61·3)13 (52·0)10 (37·0)19 (48·7)
 = 0·048 
Race/skin colour
White104 (59·1)48 (56·5)8 (61·5)19 (61·3)18 (72·0)15 (55·6)23 (59·0)
Non-White68 (38·6)35 (41·2)4 (30·8)12 (38·7)7 (28·0)12 (44·4)14 (35·9)
Other4 (2·3)2 (2·4)1 (7·7)-0 (0·0)0 (0·0)2 (5·1)
  P = 0·280  
Region of Brazil
South13 (7·4)4 (4·7)1 (7·7)1 (3·3)1 (4·0)3 (11·1)5 (12·8)
Southeast65 (36·9)28 (32·9)2 (15·4)8 (25·8)11 (44·0)8 (29·6)18 (46·2)
Northeast75 (42·6)38 (44·7)8 (61·5)13 (41·9)13 (52·0)11 (40·7)13 (33·3)
Middle-West23 (13·1)15 (17·6)2 (15·4)9(29·0)0 (0·0)5 (18·5)3 (7·7)
  P = 0·520  
Leukaemia subtype
ALL131 (74·4)72 (84·7)11 (84·6)27 (87·1)23 (92·0)15 (55·5)21 (53·8)
AML45 (25·6)13 (15·3)2 (15·4)4 (12·9)2 (8·0)12 (44·5)18 (46·2)
  P = 0·048  
WBC count (×109/l)
<5051 (29·0)24 (28·2)7 (22·6)6 (19·3)4 (16·0)12 (44·4)11 (28·2)
≥50125 (71·0)61 (71·8)6 (19·4)25 (80·7)21 (84·0)15 (55·6)28 (71·8)
  P = 0·029  

MLL breakpoint distribution

The analysis of genomic breakpoints within the MLL breakpoint cluster region was performed in a subset of 55 MLL-r with available biological material and successfully determined in 30 cases. The correlation with age at diagnosis, gender, leukaemia subtype, and WBC count is shown in Fig 1. The breakpoint distribution according to age displays a preference for recombination events affecting MLL intron 11 in children ≤6 months at diagnosis (83·0%). Conversely, children aged 7–12 months presented with a preference for recombination events occurring in MLL intron 9 (63·6%). These differences were highly significant (= 0·005). Male patients presented with a similar distribution between MLL intron 9 (35·0%) and intron 11 (40·0%), while female patients displayed recombination events remarkably more frequently in MLL intron 11 (70·0%). ALL patients had chromosomal breakpoints within intron 11 in 11 of 20 (55·0%) cases, while AML cases had breakpoints in MLL intron 9 and intron 11 at similar frequencies (40%). A breakpoint within intron 11 was associated with a higher WBC count at diagnosis (= 0·047). The multivariate analysis showed that the risk of having the MLL breakpoint affecting intron 11 is increased for patients ≤6 months [(OR 22·9; 95% CI 2·2–237·4), = 0·009] and for patients with high WBC [(OR 13·2; 95% CI 1·1–152·9), = 0·039].

Figure 1.

MLL breakpoints distribution according to (A) age at diagnosis (months) (B) gender (C) leukaemia subtype (D) white blood cell count (×109/l). ALL, acute lymphoblastic leukaemia; AML, acute myeloid leukaemia.

Survival analyses

OS was 73·3% for ALL patients and 45·4% for AML patients. Mean follow-up time from diagnosis was 75·1 months (95% CI 67·2–83·1). Follow up data was available for 469 patients of which 270 were alive (57·6%) and 199 died (42·3%). Of the 199 deceased patients, 150 deaths (73·4%) occurred in the first 12 months of treatment, and 115 deaths (57·8%) occurred in the first 6 months after diagnosis. Of the 270 living patients, 75% have been followed for at least 36 months. Variables associated with significant poorer outcomes were age at diagnosis ≤6 months (< 0·001), high WBC count (= 0·048), AML subtype (= 0·009), MLL-r (= 0·002), and MLL breakpoint region (= 0·018) (Table 3; Fig 2). However, when leukaemia subtypes were analysed separately, a high WBC count and MLL-r conferred a significantly worse prognosis only for children with ALL, and not for those with AML. OS has also been associated with MLL TPGs. Although not statistically significant, AFF1 and MLLT3 presented with a poor prognosis in the analysis including all patients (= 0·986). The univariate analyses of OS according to TPG were also performed, separating MLL-AFF1 versus other TPG in ALL cases and MLL-MLLT3 versus other TPG in AML cases, but no statistically significant difference was observed.

Table 3. Univariate analysis of overall survival (OS) according to AL subtypes
 All cases ALL AML 
OSa (95% CI) P b OSa (95% CI) P b OSa (95% CI) P b
  1. OS, overall survival; 95% CI, 95% confidence interval; WBC, white blood cell count; ALL, acute lymphoblastic leukaemia; AML, acute myeloid leukaemia; TPGs, translocation partner genes; ND, not determined; cIncludes MLLT10 (n = 4), ELL (n = 3), MLLT4 (n = 1), EPS15 (1p32; n = 1), NEBL (n = 1), SEPT6 (n = 1) and unidentified MLL TPG (n = 31).

  2. a

    Median survival.

  3. b

    Mantel-Cox (log-rank) test.

Age at diagnosis (months)
≤610·7 (7·1–14·2)<0·00112·2 (8·6–15·7)<0·0016·5 (1·0–12·0)0·016
7–1237·3 (16·0–58·6) 42·1 (4·1–80·2) 29·0 (8·6–49·3) 
13–2463·0 (52·9–73·2) 67·1 (55·6–78·6) 30·9 (21·3–40·5) 
WBC (x109/l)
<5072·1 (64·1–86·2)0·04885·3 (72·1–98·5)0·00113·1 (2·6–23·6)0·337
≥5024·5 (8·0–41·0) 23·1 (4·5–41·7) 29·0 (22·3–45·3) 
MLL status
Rearranged14·2 (10·1–18·2)0·00216·1 (7·9–24·2)0·0027·3 (1·8–12·7)0·087
Germline57·3 (51·3–68·4) 72·1 (63·9–74·7) 21·5 (5·1–37·9) 
MLL TPGs
AFF1 13·8 (7·3–20·3)0·98614·1 (9·9–18·3)0·9207·3 (2·7–11·8)0·857
MLLT1 34·2 (0·0–70·6) 34·2 (0·0–85·9) 2·2 (0·0–11·7) 
MLLT3 17·3 (10·4–24·1) 21·1 (13·6–28·5) 3·6 (1·4–5·8) 
Otherc12·2 (8·8–15·5) 12·2 (7·9–24·2) 4·8 (0·0–18·5) 
MLL breakpoint region
Intron 116·8 (0·0–14·5)0·0186·8 (0·0–14·9)0·0856·6 (0·0–16·5)0·091
Other42·1 (15·2–69·1) 21·1 (13·4–28·8) 61·1 (22·3–99·9) 
t(4;11) transcripts (ALL cohort)
MLL-AFF1 2·3 (0·5–11·8)0·2061·0 (0·6–3·2)0·161 
MLL-AFF1/AFF1-MLL 8·2 (0·7–17·3) 9·0 (0·8–21·6)  
Figure 2.

Kaplan-Meier overall survival (OS) curves for infant patients according to (A) age at diagnosis (months) (B) acute leukaemia subtype (ALL [acute lymphoblastic leukaemia] vs. AML [acute myeloid leukaemia]) (C) white blood cell (WBC) count (×109/l) (D) MLL status (E) MLL breakpoint region, and (F) gender (M, male; F, female) and MLL status. Symbols represent censored cases.

The univariate analyses of OS according to age group in AL are shown in Table 4. Of the patients aged ≤12 months, variables with significant poorer outcomes were high WBC count (= 0·026), MLL-r (= 0·047), and MLL breakpoint region (= 0·052). When patients older than 12 months were analysed, only MLL-r conferred a significantly worse prognosis. Because an increased risk to harbour an MLL-r was observed among female patients, we analysed whether the combined status of gender and MLL affected the OS. There was a statistically significant difference between the groups (P < 0·001), with the worst survival observed in females with MLL-r (Fig 1F). This difference was also significant when analysing the different subgroups: ALL, AML, younger than 12 months, and aged 13–24 months (data not shown).

Table 4. Univariate analysis of overall survival according to age groups in AL patients
 ≤12 months13–24 months
OSa(95% CI) P b OSa (95% CI) P b
  1. WBC, white blood cell count; ALL, acute lymphoblastic leukaemia; AML, acute myeloid leukaemia; TPGs, translocation partner genes; ND, not determined; cIncludes MLLT10 (n = 4), ELL (n = 3), MLLT4 (n = 1), EPS15 (1p32; n = 1), NEBL (n = 1), SEPT6 (n = 1) and unidentified MLL TPG (n = 31).

  2. a

    Median survival.

  3. b

    Mantel-Cox (log-rank) test.

WBC (x109/l)
<5037·3 (16·9–57·7)0·02685·4 (71·6–99·3)0·815
≥5012·3 (8·5–16·1) 64·7 (53·5–75·9) 
Leukaemia
ALL19·7 (8·6–30·8)0·12887·1 (75·6–98·6)0·073
AML12·5 (1·0–24·0) 40·9 (31·3–50·5) 
MLL status
Rearranged12·3 (7·7–16·9)0·04746·8 (32·8–60·8)0·020
Germline29·0 (9·4–48·6) 67·4 (56·6–78·3) 
MLL TPGs
AFF1 12·3 (7·5–17·1)0·79148·2 (27·4–69·1)0·840
MLLT1 14·2 (8·7–25·2) 26·2 (5·8–46·5) 
MLLT3 16·4 (0·9–39·8) 49·2 (24·5–73·9) 
Otherc10·8 (1·3–20·3) 77·9 (37·6–118·2) 
MLL breakpoint region
Intron 116·8 (0·0–14·9)0·0520·6 (0·0–2·3)0·222
Other21·1 (0·0–44·6) 12·6 (5·6–15·6) 
t(4;11) transcripts (ALL cohort)
MLL-AFF1 6·9 (1·4–18·7)0·8300·5 (0·2–1·9)0·010
MLL-AFF1/AFF1-MLL 8·2 (3·4–16·4) 13·7 (0·2–27·1) 

In the investigated cohort (ALL and AML), multivariate analysis revealed that age at diagnosis (= 0·003; HR 0·74, 95% CI 0·60-0·90), WBC count (= 0·049; HR 1·40, 95% CI 1·06–1·97), AL subtype (ALL vs. AML, = 0·007; HR 1·59, 95% CI 1·14-2·23), MLL status (= 0·034; HR 1·61, 95% CI 1·05–2·51), and MLL breakpoint region (= 0·019; HR 4·95, 95% CI 1·30–18·85) conferred adverse risk. Likewise, Cox regression analysis in ALL patients demonstrated that age at diagnosis (= 0·036; HR 0·76, 95% CI 0·59–0·97), WBC count (= 0·002; HR 1·99, 95% CI 1·28–3·10), and MLL status (= 0·020; HR 1·93, 95% CI 1·12–3·37) were identified as independent prognostic factors. No independent risk factors could be identified for the investigated AML cases.

Discussion

In this series of early childhood leukaemia, the frequency of MLL-r separated the cohort into 74·4% ALL and 25·6% AML, in agreement with several previous reports (Borkhardt et al, 2002; Greaves, 1999; Harrison et al, 2010; Jansen et al, 2007; Pieters et al, 2007; Pui et al, 1995; Sam et al, 2011; von Neuhoff et al, 2010). It also corroborates the findings of a recent report (Sam et al, 2011), in which younger age was predominantly associated with ALL rather than with AML. The observation that MLL-r IL patients are diagnosed earlier than MLL germline patients has been demonstrated by others (Jansen et al, 2007; Pieters et al, 2007; Pui et al, 1995; Sam et al, 2011; Sorensen et al, 1994). The molecular epidemiological approach in several genetic studies has raised the concept that most, if not all, childhood leukaemia cases originate in utero (Greaves & Wiemels, 2003). Despite these findings, the underlying mechanism that determines the latency of leukaemia development remains to be resolved. Increased apoptotic rates due to the presence of certain MLL fusion proteins in non-permissive cell types may be a potential mechanism that explains the likelihood of developing a leukaemia (reviewed in Marschalek, 2011); however, the knowledge concerning the tolerance, lineage restriction, and other potential mechanisms deriving from different MLL-r alleles remains limited. For example, children with T-ALL are usually diagnosed at older ages. However, in this same series of IL cases, the existence of T-ALL was identified. The availability of such a rare epidemiological set of cases prompted us to conduct a molecular screening of leukaemia- and clone-specific biological markers, which could be helpful in understanding the critical events leading to T-ALL in such a short latency time (Mansur et al, 2010).

In agreement with earlier studies (Ross & Robison, 1997; Sam et al, 2012), the increased likelihood of harbouring an MLL-r in females was observed here. This risk was increased in female patients with AML. It has been postulated that an elevated risk of IL in females would suggest that there is a hormonal trigger associated with female and higher risk (Ross et al, 1997). Indeed, while testing the potential role of intrauterine and perinatal factors associated with IL risk, we have demonstrated a greater association between maternal use of oestrogen during pregnancy and IL risk (OR 8·76; 95% CI 2·85–26·93) (Pombo-de-Oliveira & Koifman, 2006). Motivated by this epidemiological data, Schnyder et al (2009) performed an in vitro model of translocation susceptibility and determined that higher concentrations of oestrogens (estradiol-E2 and 4-OH-E2) might indeed be implicated in the occurrence of MLL-r. In addition, we describe here that 66·7% of female patients had the breakpoints located in intron 11. Thus, the reported higher incidence of MLL-r in females may account for the higher probability of recombinant events in intron 11, which also correlated with younger age at diagnosis, ALL subtype, and higher WBC count.

We also tested whether ethnicity was associated with the likelihood of acquiring an MLL-r. We were able to demonstrate that children defined as non-White had an increased likelihood of AL with an MLL-r. Aware of the pitfalls of separating the Brazilian mixed population by skin colour (Pena et al, 2011), we analysed the patients according to Brazilian regions. Non-Whites are more prevalent in the Northeast of Brazil, which presented a greater likelihood of MLL-r. This result partly contradicts the findings described by Sam et al (2012), which claimed that Blacks had a lower probability of MLL-r. However, because of the differences in ethnicity definition among Brazilians, it is difficult to draw conclusions from this comparison. Altogether, both our and the former studies suggest that differences in genetic susceptibility and the variability of environmental exposures may explain these findings. Furthermore, infants can also be indirectly exposed to toxins through maternal exposure and the physiological immaturity of the xenobiotic system. We have previously investigated the effect of genotype of enzymes that act in the detoxification phase of the xenobiotic system. We were able to identify a four-fold increased risk for developing IL when NAT2 displays a slow acetylating profile. The combination of maternal and fetal slow acetylating profiles was one of the most important risk factors for IL and early childhood leukaemia (Zanrosso et al, 2010). We also demonstrated an association between maternal pesticide exposure (postulated to cause DNA alterations) during the periconceptional period or during early pregnancy and the development of childhood leukaemia (Pombo-de-Oliveira & Koifman, 2006; Ferreira et al, 2012).

In our study, patients born in the Northeast presented with a higher risk for MLL-r; however, the frequency of MLL TPGs MLL-AFF1 (46%), MLL-MLLT1 (18%) and MLL-MLLT3 (16%) among these patients was similar to those of patients from other areas. Therefore, we conclude that the frequency of specific MLL TPGs in Brazilian IL resembles the findings that were obtained worldwide. Our data shows an expressive percentage of MLL-r cases in children older than 12 months, 30·6% (54 out of 176 rearranged cases). Regarding the prognosis value of MLL TPGs, a major observational study of infant ALL did not identify any association between type of MLL-r and outcome (Pieters et al, 2007). This is consistent with our findings in ALL cases. In paediatric AML, in which the frequency of MLL-r varies between 15% and 20% the outcome was not significantly different when comparing the different MLL TPGs, as we observed a dismal OS for all of them. The prognostic significance of MLL TPGs in AML remains uncertain. Nevertheless, recent studies have contributed to future risk stratification. A study evaluating incidence and risk of chromosomal abnormalities in 720 children (0 to 16 years) registered on Medical Research Council AML trials did not show a significantly improved outcome of patients with t(9;11) when compared with other MLL subgroups (Harrison et al, 2010). Another study analysed the cytogenetic data of a large cohort treated nearly uniformly in AML-Berlin-Frankfurt-Münster (BFM) 98 and observed that children with t(9;11) and additional aberrations and children with MLL-r other than t(9;11) and t(11;19) had an unfavourable outcome (von Neuhoff et al, 2010). A large retrospective analysis observed that patients with a t(1;11) showed independent favourable outcome and several other rearrangements were identified as predictors of poor clinical outcome, including t(6;11)(q27;q23), t(10;11)(p11·2;q23), t(4;11)(q21;q23), and t(10;11)(p12;q23) (Balgobind et al, 2009). In both AL subtypes, few investigations had examined the aetiology and prognosis of children aged between 13 and 24 months, taking into account that a prenatal origin for these MLL-r is also plausible (Alexander et al, 2001; Pombo-de-Oliveira & Koifman, 2006). We could verify that MLL-r is an independent prognostic factor for ALLs conferring poor outcome even for the patients aged 13–24 months.

Finally, we identified a strong association between breakpoint distributions within MLL-r cases and clinical course. Breakpoints in MLL intron 11 were described as associated with infant-ALL and therapy-related leukaemia (Reichel et al, 2001). There is also a known higher frequency of breakpoints within intron 11 in paediatric ALL (Meyer et al, 2009). To our knowledge, this is the first time that a breakpoint-dependent impact of MLL-r on clinical progression – synonymously reflected by initial high WBC count and young age at diagnosis – has been identified in a clinical study. Although the survival analyses according to MLL breakpoints was jeopardized due to the number of available cases, the OS for patients with intron 11 breakpoints was significantly poorer compared to other MLL-r introns. This might be explained by the fact that MLL exons 11–16 encode the PHD domain (PHD1, PHD2, and PHD3). Chromosomal breakpoints within MLL intron 11 will disrupt the structure of the first PHD finger, and thus, similarly cause mis-folding of the remaining domain due to alternative Cys/His-pairings (see Fig 3). Only breakpoints within MLL introns 9 and 10 will not affect the three PHD finger domains. Mis-folding of these important domains is related to two different mechanisms: PHD2 is able to confer dimerization of MLL, while PHD3 is able to bind to cyclophylin 33 (CYP33) (Fair et al, 2001). Binding of CYP33 to PHD3 causes a regulatory switch of the MLL protein, because CYP33 binds to HDAC1/2, CtBP, and the Polycomb-group proteins HCP2 and BMI1. This causes a switch of a promoter-activating MLL complex to a promoter-repressing MLL complex, as explored in detail elsewhere (Rössler & Marschalek, 2012). It is therefore a plausible hypothesis that breakpoints in MLL intron 11 result in fusion proteins that exert different biological functions on the affected cells, most likely eliminating a negative regulatory function normally exerted by the wild type MLL complex. By contrast, breakpoints that occur in other introns of the MLL breakpoint cluster region should not have such an impact. This difference may explain our observed clinical variations in MLL-r patients with a breakpoint in MLL intron 11.

Figure 3.

Schematic structure of the MLL PHD domain. PHD Fingers are composed of amino acid sequences Cys(2)Cys(X)Cys(2)Cys(4)His(2)Cys(Y)Cys(2)C (numbers represent spacing amino acids). Cysteine residues are displayed as light green spheres, while histidine residues are displayed as dark green spheres. Smaller light blue spheres indicate zinc ions. (A) The position of the MLL wild type PHD domain is displayed. Between PHD2 and 3, a normal zinc finger (ZF) is present. (B) The situation after deleting MLL exon 11 encoded amino acids is displayed. Assuming that the consecutive cysteine and histidine residues are still able to fold into distinct structures, PHD1 and PHD2 are changed into non-PHD fingers and ZF1, a portion of extended PHD3, is missing. This may cause an inability to bind to CYP33.

In summary, we have shown for the first time that MLL intron 11 recombination events are associated with early disease onset and high WBC count, indicating that a disruption of the PHD domains may be associated with a worse clinical course. This new knowledge should be validated by additional studies. Finally, if this information can be verified, both the presence of an MLL-r itself and the location of the breakpoint should be taken into account in clinical practice.

Acknowledgements

This investigation was supported by the Brazilian National Research Council (CNPq) and Instituto Nacional de Câncer-Fundação Ary Frauzino. The project was partially funded by grants from INCT-Controle do Cancer (CNPq #573806/2008-0, FAPERJ#E026/170·026/2008 and FAPERJ#E-26/110·823/2012). MSPO has been supported by CNPq research scholarships (#309091/2007) and FAPERJ (#E026/101·562/2010). ME has been supported by CAPES/PROBRAL#345/10. RM is supported by grant 107819 from the German Cancer Aid and grant DKS 2011·09 from the German Paediatric Cancer Foundation.

Authors' contributions

ME, RM, and MSPO contributed to study design. ME, CM, MBM, RM, and MSPO contributed to drafting of the manuscript. ME and CM performed the molecular studies. ME and MBM contributed to the statistical analysis. MSPO contributed to the conception of the study, data collection, and critical analysis of the data. All co-authors of the Brazilian Collaborative Study Group of Infant Acute Leukaemia contributed with clinical and demographical data. LCST contributed to the statistical analysis.

Conflict of interest

The authors have no competing interests.

Appendix I

The Brazilian Collaborative Study Group of Infant Acute Leukaemia listed as co-authors and Brazilian Institutions

Co-author Institution
Luiz Claudio Santos ThulerInstituto Nacional de Câncer (INCA) and Universidade Federal doEstado do Rio de Janeiro (UNIRIO), Rio de Janeiro, RJ
Alessandra Faro, Camilla Andrade, Caroline BarbieriPrograma de Hematologia e Oncologia Pediátrico, INCA, Rio de Janeiro, RJ
Isis Quesado Magalhães, José Carlos CordobaHospital de Apoio e Base de Brasília, DF (n = 41)
Nubia Mendonça, Flavia N.Serafim Araujo, Lilian Maria Burlachini, Maria D DoreaSociedade de Oncologia da Bahia-Salvador, BA (n = 38)
Silvia Brandalise, Vitória R. PinheiroCentro Infantil Dr. Domingos Boldrini, Campinas, SP (n = 37)
Gilberto Ramos, Joaquim C.Aguirre NetoHospital das Clinicas-Belo Horizonte, MG (n = 30)
Nilma Pimentel de BritoHospital Aristidez Maltez-Salvador, BA (n = 27)
Eny G. Carvalho, Ana M.Marinho da Silva, Jozina M. de Andrade AgarenoHospital Martagão Gesteira-Salvador, BA (n = 21)
Maria Lucia Marinho LeeInstituto de Oncologia Pediátrica - São Paulo, SP (n = 21)
Jane Dobbin, Reinaldo Dal Belo, Alexandre ApaHC1-Instituto Nacional de Cancer -Rio de Janeiro, RJ (n = 21)
Rosania Baseggio, Marcelo Santos de Souza, Atalla MnayjariHospital Regional Rosa Pedrossian - Campo Grande, MS(n = 20)
Andrea Gadelha Nobrega, Eloisa C. E. FialhoHospital Napoleão Laureano- João Pessoa, PB (n = 19)
Terezinha Marques SallesHospital Universitário Oswaldo Cruz CEON, Universidade Federal de Pernambuco (n = 17)
Elaine da Costa, Adriana Martins de Souza, Soraya Rouxinol, Marcelo LandInstituto Pediatria Puericultura Martagão Gesteira -Rio de Janeiro, RJ (n = 16)
Virginia Maria CóserHematology/Oncology Department, Hospital Universitário de Santa Maria (n = 10)
Patricia Carneiro de Brito, Loretta S.C. Oliveira, Cesar Bariani,Hospital Araujo Jorge -Goiania, GO (n = 14)
Mara A. D. PianovskiHospital de Clinicas-Curitiba, PR (n = 14)
Thereza C. S. LafayetteHospital Universitário de Santa Maria, RS (n = 14)
Imarui Costa, Denise BousfieldHospital Infantil Joana de Gusmão -Florianópolis, SC (n = 11)
Teresa Cristina CardosoHospital Manoel Novais Sta Casa de Misericórdia- Itabuna, BA (n = 10)
Fernando de Almeida Wernerck, Monica LanksznerHospital dos Servidores do Estado- Rio de Janeiro, RJ (n = 10)
Luis Fernando Lopes, Wellington MendesHospital AC Camargo-São Paulo, SP n = 9)
Alejandro M. Arancibia, Renato Melaragno, Gustavo NevesHospital Santa Marcelina-São Paulo, SP (n = 9)
Lilian Maria CristofaniInstituto da Criança Professor Pedro de Alcantara-São Paulo, SP (n = 8)
Renata S. Carvalho GurgelHospital Universitário Alcides Carneiro-Campina Grande, PB (n = 5)

Ancillary