The mechanism through which nonsense mutations impair gene expression and cause human genetic disease [Mort et al., 2008] consists of premature translation termination, and the synthesis of truncated proteins with loss-of-function features. Moreover, these mutations can trigger nonsense-mediated decay of mRNA (NMD) [Khajavi et al., 2006]. With this as background, they are commonly believed as responsible for null genetic conditions.
However, the mechanism of ribosome readthrough, which consists of misrecognition of the premature stop codon by an aminoacyl-tRNA instead of the termination factors [Rospert et al., 2005], could restore translation impaired by nonsense mutations. While ribosome readthrough is expected to occur at low rate, it could account for minimal full-length protein biosynthesis.
A recent study on a nonsense mutation in LAMA3 gene demonstrated for the first time that this process occurs in a patient in which the full-length protein is detected [Pacho et al., 2011]. Noticeably, LAMA3 encodes a structural protein (laminin-332 α3 chain) that accumulates over time, which magnifies the readthrough effects but not allow for the proper assessment of the mutation-dependent rate.
To date, no data has been reported on secreted proteins with limited half-life such as coagulation factors (from few hours to few days) [Furie and Furie, 1992], which would better reflect the spontaneous ribosome readthrough efficacy in vivo.
Factor IX (FIX) deficiency (Hemophilia B, HB; MIM# 306900) represents an ideal model to address this issue. In fact, nonsense mutations are very frequent in the severe HB forms (20%) [Giannelli et al., 1998] (http://www.hgmd.cf.ac.uk;http://www.isth.org/default/index.cfm/publications/registries-databases/mutations), and the X-linked nature of the disease enables for the study of the expression of a single, mutated allele. FIX protein is secreted in plasma, thus favoring its evaluation in vivo. At variance from most human diseases, residual levels of functional FIX protein, even if not detectable by coagulation laboratory assays [Santagostino et al., 2010], might have pathophysiological implications, thus making the impact of ribosome readthrough extremely relevant. Of utmost importance are the immunological complications upon therapeutical infusion of the defective protein in patients [Bolton-Maggs and Pasi, 2003; DiMichele, 2007]. Hemophilia patients with nonsense mutations [Giannelli et al., 1998; Kemball-Cook et al., 1998] clearly display a lower risk of developing inhibitors than patients with ample gene deletions [Belvini et al., 2005; Ljung et al., 2001; Thorland et al., 1999]. These observations point to the presence of trace levels of full-length proteins even in patients with premature nonsense triplets, but this issue has not been addressed yet.
Here, we investigated four nonsense mutations in F9 gene (MIM# 300746; GenBank accession number K02402.1) differing in position and sequence context, which are candidate determinants of ribosome readthrough [Cassan et al., 2001; Manuvakhova et al., 2000; Namy et al., 2001; Rospert et al., 2005; Tork et al 2004]. Studies in plasma from HB patients and in vitro were undertaken to evaluate their differential impact on mRNA and protein biology.
FIX coagulant activity levels were determined by aPTT-based assays as previously described [Belvini et al, 2005], whereas FIX protein levels were evaluated by ELISA (Affinity Biologicals, Ancaster, Canada).
Western blotting analysis was conducted essentially as described by Pinotti et al. (2002), by exploiting a polyclonal antihuman FIX peroxidase-conjugated antibody (Affinity Biologicals) and the Supersignal® West Femto reagent (Thermo Scientific, Rockford, IL) for detection. To calibrate the assays, serial dilution of pooled normal plasma (PNP) were used.
Plasma samples from severely affected HB patients (Fig. 1A) bearing the p.L103* (PF9*), p.R294* (PF9294*), and p.R298* (PF9298*) nonsense mutations [Belvini et al., 2005; reported in the Human Gene Mutation Database, httxsp://www.hgmd.cf.ac.uk] were exploited to investigate the impact of ribosome readthrough in vivo.
The FIX antigen and activity levels, measured in plasma samples withdrawn after a wash-out period of one week were virtually undetectable in plasma from PF9103* and PF9294*. In plasma from PF9298*, upon a 70 hr wash-out, we revealed traces of circulating FIX antigen (1.5% of PNP) that were associated with measurable FIX activity (1%). Being the half-life of the infused recombinant FIX (BeneFix, Wyeth, Taplow, UK) of ∼20 hr, we cannot rule out the presence of residual FIX protein levels that might complicate the assessment of endogeneous FIX in PF9298* plasma.
The presence of the FIX mRNA in patients was investigated in leukocyte mRNA by reverse transcription followed by polymerase chain reaction (RT-PCR) [Pinotti et al., 1998]. While the housekeeping GAPDH mRNA was detected in all samples, the correct amplified FIX fragments were identified in PF9294* and PF298* only (Fig. 1B, left panel). RT-PCR and sequencing was exploited for a semiquantitative estimate of the mutant versus normal FIX mRNA forms in mutation carriers (Fig. 1B, right panel). The expression of the mutated FIX mRNA was appreciable in carriers of the p.R294X and particularly of the p.R298X mutations, but not of the p.L103* change that, being an early stop codon, was expected to trigger major NMD. These data, obtained at the ectopic mRNA level, do not support major NMD for the p.R294* and p.R298* mutations.
Western blotting with a specific polyclonal antihuman FIX antibody and a very sensitive chemioluminescent substrate were optimized to investigate FIX molecules in plasma (Fig. 1C). A band corresponding to full-length FIX (∼60kD) was observed in PF9298*, and upon film overexposure, in PF9294*. This band was undetectable in PF9103*'s plasma, which provided us with an internal negative control and validated the specificity of the antibody used.
Expression studies in Human Embryonic Kidney (HEK293) cells were conducted to investigate the occurrence of readthrough over the different nonsense triplets, its rate, and to corroborate findings in patients' plasma. At variance from the previous studies that exploited reporter genes [Manuvakhova et al., 2000; Pacho et al., 2011], we inserted the nonsense mutations into the full-length human FIX cDNA (Fig. 2A) and investigated the proteins secreted in medium, thus favoring the comparison of truncated and full-length molecules, and, therefore, evaluation of readthrough effects in the proper nucleotide and protein context.
Expression vectors, created by site-directed mutagenesis [Pinotti et al., 1998; Pinotti et al., 1996] of the human FIX cDNA cloned into the pCMV5 vector, were transiently transfected in HEK293 as described [Furlan Freguia et al., 2004]. To calibrate the assays, serial dilution of recombinant FIX-wt were exploited.
In this experimental model, the recombinant rFIX-294* and rFIX-298* molecules were secreted at appreciable level (3.1 ± 1.1% and 2.5 ± 0.7% of rFIX-wt, respectively), whereas the rFIX-103* antigen was undetectable in medium (Fig. 2B, ELISA). Coagulant assays did not reveal any appreciable FIX coagulant activity in conditioned media.
Western blotting in medium from the rFIX-294* or rFIX-298* expressing cells revealed a large proportion of truncated FIX forms (Fig. 2C, upper panel), which were not detected in patients' plasma (Fig. 1C) very likely because of removal from circulation. Combination of ELISA and Western blot analysis indicates that truncated molecules were secreted with lower efficiency than rFIX-wt, which is consistent with the essential role of the FIX carboxyl-terminal region for secretion [Kuraki et al., 1997].
Noticeably, overexposure of films also showed a form compatible with full-length FIX, with an intensity of approximately 0.2% of rFIX-wt (Fig. 2C, lower panel). Most importantly, the band corresponding to full-length FIX was absent, even upon overexposure, in medium from cells transfected with the pFIX-103* or the gutted pCMV5, thus emphasizing the specificity of the signal detected. The differential readthrough efficiency might have contributed to the levels of the mutant mRNA in mutation carriers (Fig. 1B).
It is worth noting that ribosome readthrough often leads to insertion of aminoacids other than the natural one at the nonsense mutation position [Rospert et al., 2005], thus potentially impairing the intracellular protein biosynthesis and vanishing the effects of readthrough itself. However, the poor conservation of arginine 294 and arginine 298 among serine-proteases [Greer, 1990] would make tolerable most substitutions at these positions, in agreement with the observed residual FIX biosynthesis.
Altogether, these findings in patients' plasma and with recombinant proteins support the occurrence of ribosome readthrough over the p.R294* and p.R298* nonsense mutations and the synthesis of traces of full-length FIX. These data support the investigation of the readthrough-mediated full-length FIX synthesis, as additional determinant of clinical phenotype, and particularly of anti-FIX antibody development.
Interestingly, the efficiency through which the nonsense mutations underwent readthrough (i.e., p.R298* = p.R294* >>>p.L103*) in the proper protein context was roughly consistent with the score (Fig. 2A) derived from reporter gene assays [Manuvakhova et al., 2000].
To provide additional insights into the key role of the nonsense triplet sequence context for readthrough, we also expressed and investigated the naturally occurring p.R162* mutation that, based on the above mentioned score, is predicted to undergo readthrough with a three or four times higher efficiency than the p.R294* and p.R298* mutations (Fig. 2A). The secreted rFIX-162* antigen levels were 5 ± 0.3% of rFIX-wt (Fig. 2B) and, accordingly, the band corresponding to full-length FIX was much more appreciable (∼0.5% of rFIX-wt) than that from the other nonsense variants (Fig. 2C vs. 2D, lower panels). This observation suggests a direct relationship between the amount of full-length proteins and of truncated forms, the large majority of secreted molecules.
Noticeably, the efficient readthrough over the F9 p.R162* mutation, associated with severe HB, was consistent with that over the LAMA3 p.R943* nonsense mutation [Pacho et al., 2011], which displays an identical sequence context (TTGACT) and a mild clinical phenotype. Differently from FIX, a circulating enzyme with a one-day half-life, the structural features of the LAMA3 encoded protein would permit its accumulation over time in the basement membrane and render its function less vulnerable to amino acid substitutions introduced by readthrough. These observations underscore the interplay of the nucleotide sequence, the determinant of ribosome readthrough, and structure–function features of proteins in determining the relationships between nonsense mutations and clinical phenotypes.
In conclusion, these data from both patients' plasma and recombinant proteins obtained in a panel of selected F9 nonsense mutations demonstrate for the first time, for a secreted protein with short half-life, a gradient of spontaneous and productive ribosome readthrough, consistent with its predicted determinants.