MLLT1 YEATS domain mutations in clinically distinctive Favourable Histology Wilms tumours

Abstract

Wilms tumour is an embryonal tumour of childhood that closely resembles the developing kidney. Genomic changes responsible for the development of the majority of Wilms tumours remain largely unknown. Here we identify recurrent mutations within Wilms tumours that involve the highly conserved YEATS domain of MLLT1 (ENL), a gene known to be involved in transcriptional elongation during early development. The mutant MLLT1 protein shows altered binding to acetylated histone tails. Moreover, MLLT1-mutant tumours show an increase in MYC gene expression and HOX dysregulation. Patients with MLLT1-mutant tumours present at a younger age and have a high prevalence of precursor intralobar nephrogenic rests. These data support a model whereby activating MLLT1 mutations early in renal development result in the development of Wilms tumour.

Figure 1. MLLT1 variant verification.

(a) Base-pairs (242) of genomic DNA were amplified and the product was detected by capillary electrophoresis. Eight representative samples with three containing additional larger bands identifying insertion mutations (lanes 2, 5 and 6) are shown. Samples with novel bands were confirmed by Sanger sequencing. (b) The relative expression of the mutant and wild-type alleles was determined using RT–PCR and primers designed to amplify exons 4 and 5 of the MLLT1 gene (NM_ 005934) with detection by capillary electrophoresis. The normal allele is represented by a peak at ∼175 bp (top panel). Two samples show expression of a shorter allele corresponding to deletion mutations (second and third panels) and one sample shows expression of a longer allele (bottom panel) corresponding to an insertion mutation. It should be noted that the mutant alleles are expressed at approximately the same level as the normal alleles. (c) Samples with abnormal bands identified following PCR amplification were confirmed by Sanger dideoxy sequencing. Shown are three samples including the reference sequence on top (NM_005934.3), a 6-nucleotide deletion in the middle (c.335_340delCGCCCG) and a 9-nucleotide duplication on the bottom (c.343_351dupAACCACCTG). The base-pair positions have been aligned to show that these affect the same domain.

Figure 2. Unsupervized analysis of gene expression data.

Non-negative Matrix Factorization (NMF) analysis of 75 FHWT resulted in 6 clusters with the highest cophenetic correlation (0.95) after k=2. Five of six MLLT1 mutant tumours with available gene expression data occurred in NMF cluster 3, and two were accompanied by CTNNB1 mutations. This cluster also contained four tumours with mutation or small segment deletion of WT1, all of which also had either a mutation of CTNNB1 or small segment deletion or mutation of WTX. This cluster also contained a substantial number of tumours with retention of imprinting of 11p15 (including all MLLT1-mutant tumours). The sequencing and copy number data is provided in Supplementary Table 1. The predicted membership in gene expression subsets previously reported is provided. (Subsets 1 and 2 were low risk tumours and are therefore not represented in TARGET, Subsets 3 and 4 are highlighted in green and blue, respectively, and Subset 5 in white). Illustrated at the bottom are the expression patterns of genes of interest that were highly significantly differentially expressed between MLLT1-mutant and wild-type tumours.

Figure 3. Quantitative RT–PCR for HOTTIP and HOXA13.

(a) FHWT samples with MLLT1 mutations (n=6; red) and randomly selected MLLT1-wild-type FHWT (n=10; blue) were evaluated for HOTTIP and HOXA13 expression levels by RT–PCR normalized to endogenous GAPDH levels. A FHWT lacking MLLT1 mutation was used as the reference sample for calculating the relative quantitation (RQ) value. Samples were run in duplicate and the error bars represent the s.e.m. of the RQ values. The data are presented in log scale. Note that for tumours that lack expression a bar is not visible. (b) HEK293 cells transfected with empty vector (Ctl), wild-type MLLT1 (WT), a 1:1 ratio of wild-type MLLT1 and either p.117_118insNHL (upper panels) or p.112_114PPV>K (lower panels), or only mutant p117_118insNHL or p.112_114PPV>K is shown. RNA was isolated 24, 48 and 72 h after transfection, and the expression levels of HOXA13 and HOTTIP were normalized to endogenous GAPDH levels. The ddCt value for the control sample at 24 h was used as the reference sample for calculating the RQ value in each sample for each gene. Three independent experiments were performed for each transfection and the error bars represent the s.e.m. of the RQ values. (c) HEK293 cells transfected with wild-type MLLT1 or mutant MLLT1 (p117_118insNHL). Lysates were collected at 24, 48, and 72 h for western blotting with a MLLT1 (left) or beta-actin (antibody). The mutant protein was not significantly differentially expressed compared with the wild-type protein at any of the time points evaluated. Results for p.112_114PPV>K likewise demonstrated no significant difference in levels of protein expression.

Figure 4. MLLT1 computational sequence and structure analysis.

(a) The sequences of the MLLT1 and AF9 YEATS domains are provided. Residues forming the YEATS domain are underlined in red. Identical residues are highlighted in blue. The secondary structural elements, including the β strands and intervening loops described for AF9 (ref. 19) are provided. The recognition sites for H3K9ac by AF9 are indicated with a black diamond; residue differences at positions 107 and 111 that are also recognition sites for H3K4 by AF9 are indicated with a red diamond. Arrows indicate the MLLT1 mutations identified in the current study. (b) Homology model of the MLLT1 YEATS domain (cyan), superimposed on AF9 YEATS (orange) bound to H3K9ac (yellow stick model; PDB 4TMP). The motifs in which the indels occur are shown in blue and magenta letters. The β1, β7, β8 strands and L8 loop are designated (see text). (c) Amino acid changes between AF9 and MLLT1 in the L8 loop affect the recognition of the N-terminal region of acetylated lysine motifs. Residue differences (H>N) between AF9 and MLLT1 at positions 107 and 111 are highlighted in the superimposition of the AF9 crystal structure (PDB 4TMP; orange) with our homology model of MLLT1 YEATS (magenta). (d) The MLLT1 mutations are predicted to specifically affect the L8 loop. Computational homology models are coloured cyan (wild-type MLLT1), magenta (p.111_113NPP>K), green (p.112_114PPV>L) and blue (p.117_118insNHL). p.117_118insHHL is predicted to produce the same conformation and phenotype as p.117_118insNHL, because N/H117 are solvent exposed at the back of β8. (e) Effect of MLLT1 mutations on binding H3K9ac, measured by ITC, performed in triplicate. Shown are integrated heats resulting from binding of H3K9ac to MLLT1 wild-type (black), p.111_113NPP>K (magenta), p.112_114PPV>L (green) and p.117_118insNHL (blue). Heats resulting from peptide being titrated into buffer are in grey. Inlay: ITC data adjusted to baseline and offset for better visibility.