In this study, we found frequent inactivation of MLL3 by a frameshift mutations in microsatellite deficient CRCs and no DNA methylation at the MLL3 loci in any colon samples.
For mutation analysis, we screened 8 CRC cell lines using PCR and sequencing of all 59 coding exons. We found mutations in 5 out of the 8 cell lines (63.0%). MLL3 has a poly-A (A)9
tract within the coding sequence of exon 38 that is included in the processed transcript; homozygous frameshift mutations were found in RKO and HCT116, while heterozygous mutations were found in the microsatellite unstable cell lines SW48 and LoVo. These are all microsatellite unstable cell lines. Additionally, we found that SW48 and DLD1 have separate somatic mutations in other coding regions of MLL3 (, ). These results of mutation analysis could confirm using cDNA (data not shown). Next, we analyzed the 3 somatic mutation regions (c.8382delA (frameshift), c.10313G>A (p.G3438D), c.13630C>T (p.R4544W)) in an initial set of 72 primary CRCs and found frameshift mutations within the (A)9
tract in 10 samples (MSI-H, 22.9% (8/35); MSS, 5.4% (2/37)) (). We then analyzed 9 somatic mutation regions including the 3 sites we found and 6 sites found by Sjoblom et al. 
in a separate set of 54 primary CRCs and 21 healthy patient samples (). We found frequent mutations within the (A)9
tract in microsatellite unstable CRCs (28.6%, 4/14, see examples in ) but no other mutations in these samples. Thus, overall, we found mutations in 19/134 cases analyzed (14%). These mutations are detectable in a broad range of the coding sequence, with clustering in the poly(A) tract, confirmed by our analysis of microsatellite unstable tumors.
MLL3 mutation analysis in 8 colorectal cancer cell lines and 72 samples of primary colorectal tumors.
To study DNA methylation, we first noted that MLL3 is transcribed from two separate promoters, one of which results in a truncated version of the protein (). The promoter for the larger transcript contains a previously unrecognized CpG island, but one is not present in the truncated form. We analyzed the methylation of this CpG island using quantitative bisulfite- pyrosequencing (examples in ). However, we found that the CpG island of MLL3 area is highly homologous (~92%) to a CpG island which overlaps the promoter of a pseudo- gene on chromosome 22 (psiTPTE22, pseudo-gene of transmembrane phosphatase with tensin homology on chromosome 22: NR_001591). Indeed the bisulfite pyrosequencing assay was able to amplify this region as well indicating the possibility of false positives. We found dense methylation in all 8 cell lines examined by pyrosequencing (HCT116, 74%: RKO, 66%: LoVo, 77%: SW48, 50%: CaCo2, 79%: SW480, 65%: DLD1, 72%: SW620, 74%), and a high degree of methylation in primary CRCs (45 out of 54 examined or 83.3%, ). Methylation of this CpG island in cancer was not associated with common clinicopathologic features including age, gender, location and clinical stage. A measurable degree of methylation was present in the adjacent normal appearing mucosa of most patients analyzed, suggesting that this locus could be a target of age-related methylation 
. Indeed, in healthy appearing normal colon mucosa samples, we found a strong age-related methylation of this CpG island (r
DNA methylation analysis using quantitative bisulfite pyrosequencing in primary CRCs and MLL3 relative expression, protein analysis in different CRC cell lines.
We next sought to better understand the DNA methylation of the two homologous loci by performing bisulfite direct sequencing assays which could discriminate against the two loci. Because of the psiTPTE22 gene is 10 base pairs smaller than MLL3 gene in that region (). Interestingly, methylation of the MLL3 gene ranged from 0–5% in normal mucosa and CRC cell lines except for RKO (14.7%). However the, psiTPTE22 gene was highly methylated in colon samples, both in CRC cell lines and primary tumors. Additionally, methylation of the psiTPTE22 loci was associated with age related methylation in normal colon mucosa ().
Methylation analysis by bisulfite direct-sequencing.
To examine the expression profile of MLL3, we used qPCR (quantitative polymerase chain reaction) and three different Taqman probes to cover the full length transcript (NM 170606) and the truncated transcript (NM 021230) (). As shown in , the full length transcript (Probe A) is substantially down-regulated in 5 out of 6 cell lines examined, suggesting a genetic etiology for this silencing (). By contrast, the two other probes (Probe B and C), which detect both truncated and full-length transcripts, demonstrate minimal down-regulation (or even up-regulation) in these cell lines. Collectively, the data show that somatic mutations particularly frameshift mutations in cancer silences the full length transcript while leaving the truncated transcript intact.
We next analyzed two different size of protein of MLL3 in cells (both wild type/frame shift mutation). Protein analysis was conducted by western blot to determine whether these cells can produce the appropriate MLL3 protein (). MLL3 has a truncated form in the 3′ end. Thus, we analyzed the MLL3 protein level using antibodies (this antibody was designed from center boundary of MLL3 (Sigma-Aldrich, Cat. #SAB1300328 (St. Louis, MO)). It will detect only long isoform of MLL3 protein product.
We found the appropriate band in CaCo2 and SW480, wild type of MLL3, and LoVo and SW48, heterozygous for the frameshift mutation, by MLL3 antibody. In contrast, there was no detectable band in RKO and HCT116, both homozygous for the frameshift mutation. These results correlated with the gene expression levels and protein analysis of MLL3 ().