Autosomal dominant retinitis pigmentosa mutations in inosine 5'-monophosphate dehydrogenase type I disrupt nucleic acid binding

Abstract

Two mutations of IMPDH1 (inosine 5'-monophosphate dehydrogenase type I), R224P and D226N, have recently been found to cause adRP (autosomal dominant retinitis pigmentosa). IMPDH1 catalyses the rate-limiting step in guanine nucleotide biosynthesis and also binds single-stranded nucleic acids. In the present paper, we report the biochemical characterization of the adRP-linked mutations, R224P and D226N, and a potentially pathogenic mutation, V268I. The adRP-linked mutations have no effect on enzyme activity, protein stability or protein aggregation. These results suggest strongly that the mutations do not affect enzyme activity in vivo and thus do not perturb the guanine nucleotide pool. The R224P mutation changes the distribution of enzyme between the nucleus and cytoplasm. This effect was not observed with the D226N mutation, so the relevance of this observation to disease is unclear. In contrast, both mutations decrease the affinity of nucleic acid binding and both fail to co-immunoprecipitate RNA. These observations suggest that nucleic acid binding provides a functional assay for adRP pathogenicity. The putative adRP-linked mutation V268I also disrupts nucleic acid binding, which suggests that this mutation is indeed pathogenic.

Figure 1. Structure of IMPDH1

(A) Location of adRP-linked mutations on IMPDH1. A monomer of the human IMPDH1 crystal structure (Protein Data Bank code 1JCN) is shown with the catalytic domain in grey and the subdomain in blue. The active site is modified by the inhibitor 6-Cl-IMP, shown in purple. Arg²²⁴ and Asp²²⁶, the residues where mutations cause adRP, are shown in red. Val²⁶⁸, the residue linked to a potentially disease causing mutation, is shown in green. Ala²⁸⁵, the residue linked to a non-pathogenic mutation, is shown in orange. Note that some portions of the subdomain are disordered in the crystal structure. (B) Oligomerization of wild-type and mutant IMPDH1 analysed by gel filtration. Recombinant IMPDH1 was applied to a Superose 6HR column, eluted with 10 mM Tris/HCl, pH 8, 100 mM KCl and 0.5 mM DTT at a flow rate of 0.4 ml/min at 4 °C. The chromatograms show the elution of wild-type (wt) IMPDH1 (blue), A285T (cyan), R224P (green), D226N (red) and V268I (purple), and molecular-mass standards of 670, 450, 240 and 45 kDa (black). A tetramer of IMPDH1 has a molecular mass of 220 kDa. mAU, milli-absorbance units

Figure 2. Localization of wild-type and mutant H1GFP

(A) H1GFP (84 kDa) expression following transfection into HeLa cells was compared with endogenous IMPDH (55 kDa) by Western blot with anti-GFP antibodies and anti-IMPDH serum (26 h post-transfection shown). Additional experiments confirm that detection is linear within this range (results not shown). (B) Live fluorescence localization of wild-type (wt), R224P or V268I H1GFP protein in HeLa cells shown as GFP fluorescence. (C) The ratio of nuclear to cytoplasmic H1GFP. As described in the Experimental section, the mean GFP fluorescence intensities for the nucleus and cytoplasm were determined for >20 cells of each type. The results are means±S.D., and those significantly different from wild-type are denoted by an asterisk (*P<1×10⁻⁵ and 0.001 for R224P and V268I respectively).

Figure 3. AdRP-linked mutations decrease affinity for a random pool of ssDNA and increase the fraction bound in a filter-binding assay

The fraction bound was determined by various amounts of IMPDH1 (shown as concentration of tetramer) incubated with a labelled pool of random ssDNA sequences as described in the Experimental section. Results are representative of three experiments, and the best fit of a simple binding model is represented by the lines (eqn 3).

Figure 4. AdRP-linked mutations decrease the association of IMPDH with RNA in vivo

HeLa cells transfected with H1GFP were cross-linked with formaldehyde and immunoprecipitated with anti-GFP antibody. (A) Coomassie-Blue-stained SDS/PAGE (10% gels) of the immunoprecipitate demonstrating that similar amounts of H1GFP are precipitated in each sample. Arrows denote H1GFP (84 kDa), GFP (29 kDa) and antibody (Ab; 50 and 75 kDa). (B) The immunoprecipitates were treated with phosphatase followed by [γ-³²P]ATP and polynucleotide kinase to label nucleic acids. The filter-bound radioactivity is shown relative to untransfected cells (control). The ³²P-labelled immunoprecipitates were treated with either RNase or DNase. Asterisks mark samples significantly different from wild-type (P<0.02 for both control and RNase). (C) RNA co-immunoprecipitated with GFP, wild-type H1GFP, A285T, R224P, D226N and V268I. Samples were treated as described in (B), with the addition of a proteinase K digestion. ³²P-Labelled co-immunoprecipitated nucleic acid was analysed on a 6% denaturing polyacrylamide gel with ssDNA of known molecular mass as general markers. ATP denotes a mock labelling reaction in the absence of immunoprecipitate. wt, wild-type.