Messenger RNA Transcripts of the Hepatocyte Nuclear Factor-1α Gene Containing Premature Termination Codons Are Subject to Nonsense-Mediated Decay Free

Maturity-onset diabetes of the young (MODY) is a subtype of diabetes characterized by early age of onset (usually <25 years), autosomal dominant inheritance, and a progressive defect in β-cell function (1). In U.K. populations, mutations in the hepatocyte nuclear factor-1α (HNF-1α) gene account for ∼65% of cases (2). Over 120 different HNF-1α mutations have been reported (3), of which ∼30% are nonsense or frameshift and lead to the production of premature termination codons (PTCs) (4). The most common HNF-1α mutation results from the insertion of a C nucleotide in a polyC tract in exon 4 (2,5–7). This mutation, P291fsinsC, accounts for ∼20% of families with HNF-1α mutations (4,8,9).

HNF-1α mutations might produce the MODY phenotype by haploinsufficiency or a dominant-negative mutational mechanism. There is considerable support for haploinsufficiency; a mutation in the HNF-1α promoter that disrupts the HNF-4α binding site results in a phenotype indistinguishable from mutations in the coding region (10). Many mutations when tested in vitro have little transactivation activity and do not act as a dominant negative (9). It has been suggested that some mutations may act through a dominant-negative mutational mechanism. The HNF-1α protein acts as a homodimer (or heterodimer with HNF-1β); therefore, mutant proteins lacking transactivation capability but retaining DNA binding and dimerization might exert a dominant-negative effect through the sequestration of wild-type proteins in inactive dimers. The commonest mutation, P291fsinsC, shows dominant-negative activity when overexpressed in MIN6 mouse insulinoma, C33 human epithelial cervical carcinoma, and INS-1 rat cells in vitro (8,9,11), but not in HeLa cells (12).

It has not been possible to analyze mRNA in patients with diabetes caused by HNF-1α mutations because tissues with high levels of HNF-1α expression (liver, pancreas, kidney, and gut) are inaccessible (13). Highly complex mechanisms occurring in living cells may be poorly represented by in vitro systems. For example, many transcripts harboring PTCs do not produce truncated proteins since they are degraded by the nonsense-mediated decay (NMD) pathway (14). This process is dependent on translation and requires the presence of at least one intron 3′ of the PTC for its action (15). Reductions in the levels of mutant mRNA transcripts in fibroblasts or lymphoblastoid cell lines have been shown to be associated with PTC mutations in the fibrillin (16), β-globin (17), Col1A1 (18), and BRCA1 genes (19). Although only a limited number of human genes have been studied, NMD is likely to be a global process that affects the majority of genes (20). To date, there have been no studies to determine whether HNF-1α mutant mRNA transcripts are subject to NMD.

In a recent publication we used a novel RT-PCR-based method to amplify HNF-1α transcripts from lymphoblastoid cells to show that three splice site mutations resulted in PTCs (21). Interestingly, the mutant transcripts were identified by sequence analysis in only one of three replicate RT-PCRs. This suggested that the mutant mRNA transcripts were less abundant than their wild-type counterparts. We hypothesized that mutant mRNAs harboring PTCs may be subject to degradation by the NMD pathway and sought to develop a quantitative method to investigate this further.

We now report a quantitative real-time PCR method to determine the relative amount of mutant and wild-type mRNA transcripts in an allele-specific manner using a coding single nucleotide polymorphism (cSNP) as a marker. Since amplification of HNF-1α mRNA from peripheral lymphocytes was unsuccessful, lymphoblastoid cell lines were established from diabetic HNF-1α mutation carriers and unaffected nonmutation carriers. All subjects selected were heterozygous for the L459L C>T polymorphism in exon 7, and it was possible to determine from family cosegregation studies which allele was inherited with the mutation. The assay is based on the fluorescent resonance energy transfer principle (Fig. 1A) (22) and relies upon the differential melting points of matched and mismatched probe-to-template duplexes for allele specificity (Fig. 1B). The use of two probes with different emission spectra allows the quantitation of both transcripts in an individual heterozygous for the cSNP. The assay was validated and tested for specificity and was shown to be sensitive and accurate over a wide range of template concentrations (data not shown). No significant differences in the relative abundance of alleles were seen between RNA extractions, reverse transcriptions, or PCR runs in control lines (P = 0.912, 0.796, and 0.222, respectively, as tested pairwise by the Mann-Whitney statistic).

We quantified the levels of mutant mRNA transcripts relative to wild-type transcripts for three mutations predicted to lead to the formation of PTCs at codons 171, 315, and 466 (mutations R171X, P291fsinsC, and I414G415ATCG→CCA), two missense mutations (G207D and R229P), and two unaffected controls. P values were determined by comparison of the median ΔCt value of mutated cell lines with the median ΔCt value obtained from the control cell lines by Mann-Whitney U test. Using control cell lines with the L459L polymorphism, we demonstrated that both alleles were expressed at similar levels (P = 0.156). All three nonsense mutant transcripts showed reduced expression (to 40%, P = 0.009; <0.01%, P ≤ 0.0001; and 6%, P = 0.001 for R171X, I414G415ATCG→CCA, and P291fsinsC, respectively) compared with the normal transcript (Table 1). The mutant transcript levels were increased by treatment with the NMD inhibitor cycloheximide by 1.8-fold (P = 0.05), 3.7 × 10⁵-fold (P ≤ 0.0001), and 81-fold (P = 0.003) for R171X, I414G415ATCG→CCA, and P291fsinsC, respectively, when compared with the solvent control (Table 2). These results indicate that the observed reduction in mutant allele frequency is due to NMD.

The variable levels of NMD seen in HNF-1α (mutant allele present at <0.01–40% of wild type) are comparable with those reported for PTC-containing transcripts in the fibrillin and BRCA1 genes where reductions from <10 to 50% of the normal allele have been described (16,19). The reduced fibrillin mutant mRNA levels were classified as severely depleted (<10% of wild type), moderately depleted (10–29% of wild type), or mildly depleted (30–50% of wild type), as determined by pulse-chase analysis (16). By this classification scheme, the R171X nonsense mutation would be described as mildly depleted and the I414G415ATCG→CCA and P291fsinsC frameshift mutations as severely depleted. The efficiency of NMD may be affected by both the type of stop codon, the surrounding sequence, and differential read-through (23).

The missense mutation G207D showed no reduction in mutant transcripts that were present at ∼93% of the normal allele (P = 0.546). The transcript carrying the missense mutation R229P showed an unexpected reduction to 52% of the normal allele (P = 0.03). This reduction was not restored by cycloheximide, indicating that the instability is independent of NMD. Reduced protein levels were also observed for the missense mutants Y122C, R159Q, K205Q, and P447L in in vitro transient transfection experiments (9). The mechanisms resulting in RNA instability are not yet clear, but missense mutations can sometimes cause alterations in mRNA folding that leads to mRNA instability (recently reported for the dopamine receptor (24) or may disrupt sequences involved in stabilization or polyadenylation (25).

In the absence of pancreatic tissue samples from patients with HNF-1α gene mutations, we have exploited the illegitimate transcription of the HNF-1α gene in lymphoblastoid cell lines derived from patients. This approach has previously been used to study other genes that are also expressed in inaccessible tissues, e.g., BRCA1 (19). Confirmation of our results using pancreatic mRNA from mutation carriers would be desirable but is clearly not possible, so we cannot be certain that the same degree of NMD occurs in the pancreas.

The advantage of using lymphoblastoid cells over engineered cell lines, transient transfections, or transgenic mice is that the mutations are present in the heterozygous state and the genomic structure of the gene is intact. Previous reports showed that the overexpression of P291fsinsC inhibited the endogenous activity of HNF-1α in C33 human epithelial cervical carcinoma, INS-1 rat, and MIN6 mouse insulinoma cells in vitro (8,9,11). However, this dominant-negative effect was not observed at 50:50 ratios of mutant-to-wild-type protein, and the absence of introns in the cDNA constructs may preclude the degradation of PTC mutations by NMD. Transgenic mice expressing P291fsinsC in pancreatic β-cells develop progressive hyperglycemia (11) and provide a closer model for HNF-1α mutations in humans than the homozygous knockout mouse that exhibits liver and kidney dysfunction (13) in addition to diabetes (26). However, the expression levels of HNF-1α protein in the transgenic mice cannot be directly compared with humans since the cDNA constructs were present in between 5 and 20 copies per genome and are intronless (27). In keeping with there not being a significant dominant-negative effect in vivo, there is no evidence to suggest a more severe phenotype in patients with diabetes caused by mutations that have a dominant-negative effect in vitro (8,12).

The previous reports of mutations in the HNF-1α gene promoter and dimerization domain provide evidence that haploinsufficiency can cause MODY (10). PTC mutations are also likely to result in haploinsufficiency as a consequence of a reduction in mRNA level due to NMD. We would suggest that even the common HNF-1α mutation P291fsinsC, which has a dominant-negative effect in vitro (8,9,11), may result in diabetes as the result of haploinsufficiency because insufficient truncated protein would be translated in order to have a dominant-negative effect in vivo.

In conclusion, our study utilized ectopic transcription in lymphoblastoid cell lines to show that premature truncating codon mutations are subject to NMD in vivo. This is likely to be an important disease mechanism in the 30% of nonsense, frameshift, and splicing mutations that result in a PTC. Haploinsufficiency may be the common mutational mechanism for the majority of HNF-1α mutations. However, final proof will require the analysis of mRNA expression in legitimately expressing tissues from patients with HNF-1α mutations.