Self-silencing GFAP missense alleles in familial subclinical Alexander disease: implications for therapy

Introduction

Alexander disease (ALXDRD) is a rare astrogliopathy characterised by white matter abnormalities, ultimately leading to neurodegeneration.¹ ALXDRD patients have been classified by age of onset, MRI findings or both. The most severe form is characterised by early onset of symptoms, while the milder form is marked by later onset and slower progression.²

ALXDRD is caused by monoallelic GFAP variants, with no clear genotype–phenotype correlations, although some are almost exclusively associated with early-onset ALXDRD, while others are identified only in mild or late-onset cases.²

GFAP expression may play a significant role in ALXDRD, potentially modifying disease penetrance and expressivity. While the predominant pathogenic mechanism in early-onset ALXDRD is thought to be the negative dominance of misfolded mutant GFAP, additional mechanisms may underlie milder disease forms, such as the variants affecting splicing or production of alternative GFAP isoforms identified in adult-onset patients.

Interestingly, several asymptomatic carriers showing ALXDRD MRI abnormalities have been reported.³ We explored the functional consequences of the novel c.1127G>T (p.R376L) variant, found in association with asymptomatic or paucisymptomatic ALXDRD patients, concluding that the amino acid change potentially contributes to protein aggregation but the nucleotide change triggers loss-of-function splicing alterations likely serving as a ‘protective’ factor of the disease severity.

Subjects/methods

This study included nine adult asymptomatic patients who underwent genetic testing of the GFAP gene because brain MRIs had suggested potential ALXDRD. Functional studies of GFAP variants included fluorescence microscopy analysis of GFP-fused wild-type or mutant GFAP proteins expressed in human U251-MG astrocytoma cells, and in vitro splicing analysis following transfection of a plasmid-encoded minigene spanning GFAP intron 4 to intron 7 into HEK293T cells. Total RNA was extracted after 24 hours and the mRNA splicing products were sequenced using an NGS-based approach. RNA was also extracted from the sodium-butyrate-treated lymphoblasts derived from one case and two controls, and the resulting cDNA underwent nested-PCR amplification and Sanger sequencing. More detailed methodological information is provided in online supplemental material .

Results

We studied nine asymptomatic adults from six apparently unrelated families with no family history, incidentally diagnosed with ALXDRD by brain MRIs performed for unrelated indications (figure 1A, online supplemental table 1 and figure 1). All subjects were found to carry the novel heterozygous missense variant c.1127G>T, causing the p.R376L amino acid change. The interest in this variant was enhanced by the previous description of three other variants at codon 376 associated with severe ALXDRD (c.1126C>T/p.R376W and c.1126C>G/p.R376G) or milder ALXDRD with late disease onset following head trauma (c.1127G>A/p.R376Q) (figure 1B, online supplemental table 2). To unravel the molecular mechanisms underlying these phenotypic differences, we studied the effects of the four R376 variants at both protein and mRNA level in cellular models.

Figure 1. Functional studies of a novel GFAP missense variant (p.R376L) identified in asymptomatic adults incidentally diagnosed with ALXDRD by brain MRIs revealed an intriguing mechanism whereby a missense mutation affecting protein structure exerts a cis-acting effect on splicing, predicted to cause a significant reduction of mutant protein levels, consistent with the subclinical manifestations. (A) MRIs showing brain abnormalities in GFAP p.R376L patients. T2-weighted hyperintensities and atrophy of the medulla oblongata/upper cervical spinal cord (i, white arrow (patient 2, 4th decade), and ii, black arrow (patient 3, 7th decade)), in the presence, or absence, of further ALXDRD typical characteristics, such as T2 hyperintensity of the hilum of the dentate nuclei (iii, white arrow (patient 6, 6th decade)), supratentorial periventricular white matter abnormalities (iv, (patient 6, 6th decade)), and midbrain peripheral rim of hyperintensity on FLAIR images (inset in i, white arrow (patient 1, 8th decade)). (B) GFAP gene diagram showing positions of the four p.R376 variants. Green, yellow and grey boxes indicate the three nucleotides of the p.R376 codon spanning between exon 6 and exon 7 (cyan boxes). Below the gene diagram, the four variants are indicated as follows: green boxes: c.1126C>G and c.1126C>T at position −2 from the end of exon 6; yellow boxes: c.1127G>A and c.1127G>T at position −1 from the end of exon 6. Intron 6 is indicated as a dark red box. Mutant alleles are red typed. The phenotypes associated with the variants are shown on the right (EO, early onset; LO, late onset). (C) Analysis of GFAP conformation in the presence of the four p.R376 variants. Images show U251-MG cells expressing the different forms of GFAP proteins fused to GFP with DAPI-stained nuclei. Images were acquired using the confocal microscope Zeiss FV3000 at ×60 magnification and merged by ImageJ software. (D) Quantitative analysis of aggregates in the presence of the four p.R376 variants. Bar diagram shows the percentage of cells characterised by GFAP aggregates in the cytoplasm. For each condition, more than 50 cells were evaluated. Values are the mean of five independent experiments±SD. Asterisks indicate statistically significant differences between the samples joined by lanes. Student’s t-test, *p<0.05; **p<0.01. (E) Splicing products generated by the four GFAP p.R376 variants in vitro. Diagram of the different splicing isoforms associated with the variants analysed. NGS-based analyses of the splicing products revealed that the WT construct transcript predominantly underwent correct splicing between exons 5 and 6 and exons 6 and 7 (cyan boxes), with a minor fraction skipping exon 6 (r.907_1127del) and potentially leading to a truncated protein (p.N303Dfs*46). The p.R376G and p.R376W variants exhibited a splicing pattern similar to WT, albeit with a slightly higher percentage of exon 6 skipping. In contrast, the p.R376L and p.R376Q transcripts showed significantly reduced exon 5-6 splicing and minimal exon 6-7 splicing, resulting in a little fraction of the normal protein, and aberrant products characterised by exon 6 skipping or partial retention (47 bp) of intron 6 (dark red box) partial retention (47 bp) of intron 6 (dark red box) that would lead to a truncated protein (p.R376Lfs*1). (F) Percentage of spliced reads for each of the following splice junctions: exon 5–6 (hg38 chr17:44911457-44911671), exon 6–7 (hg38 chr17:44910659-44911235), exon 6–7 through the alternative donor site (DS) of exon 6 (hg38 chr:17 44910659-44911188), and exon 5–7 with skipping of exon 6 (hg38 chr17:44910659-44911671) (n≥3). NGS sequencing of PCR amplified from cDNA of WT and mutant forms of pSPL3 plasmids reveals a relevant amount of aberrantly spliced transcript in both c.1127T (p.R376L) and c.1127A (p.R376Q) variant expressing plasmids. (G) In the table, for each variant, the following information is listed in the columns from left to right: (i) the nucleotide (nt) position on cDNA; (ii) the amino acid (aa) position on the protein; (iii) the position of nucleotide changes at the R376 codon; (iv) the splicing products predicted for each variant; (v) the probability of nonsense mediated decay (NMD) as a quality control mechanism induced by aberrant splicing; (vi) the cellular phenotype resulting from both the extent of aggregation and potential NMD, where green lines and circles represent filaments and aggregates, respectively, and (vi) the clinical phenotype associated with each variant. ALXDRD, Alexander disease.

Expression studies in an astrocytoma cell line transfected with plasmids encoding the four GFAP p.R376 variants showed that all variants affected the intermediate filament network and induced aggregate formation, although to different extents (p.R376L<p.R376Q<<p.R376G/W) (figure 1C-D, online supplemental figure 2). This is consistent with in silico predictions of the impact of the R376 variants on GFAP secondary structure (online supplemental figure 3).

To explore other possible mechanisms underlying the mild pathogenicity of the p.R376L variant, we performed minigene-based splicing analysis of the four different exon 6 variants located at position −1 (c.1127G>T/A, p.R376L/Q) or −2 (c.1126C>G/T, p.R376G/W) from exon 6 end (figure 1B, online supplemental figure 4). Sequencing of cDNA products showed normal mRNA processing for variants producing p.R376W and p.R376G but aberrant splicing for variants producing p.R376L and p.R376Q, resulting in both the skipping of mutated exon 6 and partial inclusion of intron 6 (figure 1E-F, online supplemental figure 5), results also confirmed in patient-derived lymphoblasts (online supplemental figure 6). In silico analysis predicted that the proteins produced, if any, would be truncated (online supplemental figure 7). More detailed results are provided in online supplemental material.

Discussion

We investigated the functional consequences of the novel c.1127G>T (p.R376L) variant, found in a group of patients clinically presenting asymptomatic or paucisymptomatic ALXDRD, and compared the effects of this variant to those of three other nucleotide changes impacting the same R376 codon, each associated with different ALXDRD severities. Although the major effects were observed for the p.R376G and p.R376W variants, associated with severe ALXDRD, both the p.R376L and the p.R376Q, associated with subclinical phenotypes, exhibited significant disruption of the filament network and aggregate formation levels.

Consistent with in silico predictions, we demonstrated a significant effect on splicing of both the p.R376L and p.R376Q variants but normal splicing for the p.R376G and p.R376W mutations.

Altogether, these results suggest the hypothesis that ALXDRD severity in the cases associated with the p.R376L and p.R376Q variants is influenced by two counterbalancing key factors: the propensity of the mutant GFAP protein to aggregate and the effect of the mutations on GFAP splicing. Thus, the absence or minimal severity of the disease in patients with the p.R376L variant would be due to the low levels of mutant protein production caused by aberrant splicing. Conversely, the p.R376G and p.R376W variants, which do not alter protein production, would result in normal levels of mutant GFAP proteins that are highly aggregation-prone, thus generating more severe forms of ALXDRD.

Consistent with this hypothesis, an ALXDRD patient was reported to carry a splicing-affecting variant resulting in the deletion of GFAP residue E207 (p.E207del).⁴ Interestingly, the patient exhibited a very late onset (eighth decade) and a considerably milder phenotype in comparison to prior cases with missense mutations at p.E207, highlighting the possibility that the impact of the mutation might be mitigated by reduction in mutant protein levels due to aberrant splicing, as illustrated here with the cases of p.R376L and p.R376Q.

In conclusion, our study uncovers an intriguing mechanism in which a missense mutation affecting protein structure exerts a cis-acting splicing effect, predicted to cause a significant reduction of mutant protein levels. This discovery sheds light on the relationship between genotype and phenotype in ALXDRD, emphasising the critical importance of assessing potential splicing effects in conjunction with amino acid changes in the GFAP gene. These effects can markedly decrease protein levels, altering the manifestation of the disease to the extent of rendering the subject asymptomatic (figure 1G). Therefore, our findings serendipitously corroborate the rationale for using ASO therapies in treating ALXDRD,⁵ as these treatments similarly aim to diminish the accumulation of damaging proteins. In addition, these findings may extend beyond ALXDRD, suggesting broader applicability in understanding and treating other genetic disorders with similar missense-mutation-splicing interplay.

The funders had no role in the study design; the collection, analysis and interpretation of data; the writing of the report; and the decision to submit the article for publication.