Results
Mutational screening ruled out mutations in the SPG4 and SPG7 genes in Pt2, due to the presence of spastic tetraparaparesis; in the HTT gene in Pt1, due to the subtle onset of symptoms consistent with an affective disorder, together with cognitive dysfunction; and in the UBQLN2 and C9orf72 genes, recently associated to ALS/FTD, in both.
The MRI features were consistent with AOAD, but no mutation was detected in the nine exons encoding the prevalent (alpha) isoform of GFAP (GFAP-α, NP_002046.1; Figure 2A). All of the known mutations associated with Alexander’s disease have so far been found in this isoform, [17] which is the only one analyzed by standard screening. However, exome-NGS revealed a heterozygous variant (c.1289G > A, p.R430H) in the alternative GFAP exon 7A (Ex7A) in both patients (Figure 2B). Ex7A is part of the transcript encoding the GFAP-ϵ isoform (NP_001124491.1), which differs from GFAP-α in the last 35 amino acids. A third isoform, GFAP-κ (NP_001229305.1), which contains a unique exon 7B, has also been identified (Figure 2A) [18]. The c.1289G>A nucleotide change was absent in the healthy mother and in all other tested family members. DNA samples from I-1 and I-3, fathers of Pt1 and Pt2, respectively, were unavailable. Haplotype analysis of the GFAP genomic region by SNPs array in the available family members confirmed that the father of Pt1 was different from that of Pt2 and of his siblings, whilst Pt1 and Pt2 share the same maternal allele (Additional file 4). Since the likelihood that the same rare variant (<0.01%) may occur independently in the two patients is negligible, the most probable hypothesis is that the mutation was transmitted by descent to both Pt1 and Pt2 by maternal germinal mosaicism, a mechanism that can also explain the healthy status of the mother. Since blood was the only source of DNA available from the mother, somatic mosaicism affecting other tissues of this subject cannot be excluded, as recently found in an AD patient with atypical infantile clinical presentation and essentially normal MRI features [19]. However, we think that the latter hypothesis is unlikely, since no trace of mutation could be detected by an ad hoc RFLP analysis carried out in the mother’s DNA (not shown) and, in contrast with the case reported by Flint et al. [19], this lady is now 87 years old and well.
Figure 2  Characterization of the GFAP c.1289G>A/p.R430H mutation. A: Schematic representation of the exonic structure of different GFAP isoforms. Dotted lines indicate the termination codons. The arrows indicate the position of the c.1289G>A variant (Note that in GFAP-κ the c.1289G>A mutation is part of the 3′-UTR). B: Electropherograms of GFAP exon 7A region containing c.1289G>A variant, in patients 1 and 2 (Pt1, Pt2) and in their mother (I-2). C: The histogram displays the percentages of cells transfected with GFP-GFAP-ϵwt (green bars) or GFP-GFAP-ϵR430H (purple bars), classified in filamentous pattern (F), cytoplasmic aggregates on a filamentous pattern (F + A), cytoplasmic aggregates with no filamentous pattern (A). Scale bars represent 15 μm. A total of 324 cells for GFP-GFAP-ϵwt and 285 for GFP-GFAP-ϵR430H, from 3 independent experiments, were blindly analyzed by two different operators. ANOVA test for interaction p = 0.001. In contrast with a p.R430C SNP (rs 78994946), reported with a frequency of 1% in dbSNP, the p.R430H change found in our patients is absent in both dbSNP and the Exome Variant Server (EVS) database, which contains >10000 alleles (≈7000 of European origin). These data are compatible for p.R430H being a deleterious mutation (Additional file 5).
GFAP is an intermediate filament (IF) protein expressed mainly by astrocytes and ependymocytes. Recent data suggested that GFAP-ϵ was unable to form filaments by itself but it could participate to the formation of the GFAP network by interacting with GFAP-α [20]. Hence we analyzed the IF meshwork in human astrocytoma U251-MG cells, constitutively expressing both GFAP-α and GFAP-ϵ, by expressing GFP-tagged wt and mutated GFAP-ϵ (GFP-GFAP-ϵwt vs. GFP-GFAP-ϵR430H). Cells were assigned to three patterns: [14] (i) exclusively filamentous pattern (F), (ii) cytoplasmic aggregates on a filamentous pattern (F + A), (iii) cytoplasmic aggregates with no filamentous pattern (A). The expression of GFP-GFAP-ϵwt led to a distribution among the three groups similar to that reported for GFP-GFAP-αwt[14] (Figure 2C) indicating no intrinsic damaging effect of recombinant GFP-GFAP-ϵwt in our experimental conditions. Contrariwise, expression of mutant GFP-GFAP-ϵR430H produced significant decrease in F (43% vs. 58%; test t p = 0.002) and increase in A (22% vs. 15%; test t p = 0.009) cells (Figure 2C), with a distinct distribution in the three patterns compared to GFP-GFAP-ϵwt expressing cells (ANOVA test for interaction p = 0.001). Notably, the expression of GFP-tagged GFAP carrying the R430C variant (GFP-GFAP-ϵR430C) led to a distribution amongst the three different patterns similar to that obtained with GFP-GFAP-ϵwt, i.e. non-significant (ANOVA test for interaction p = 0.333). These results indicate that GFAP-ϵR430H is inefficiently incorporated, and is likely to perturb the GFAP network in GFAP-expressing astrocytoma cells, whereas the GFAP-ϵR430C variant is functionally wt, but we cannot exclude the possibility that variations in the level of expression contributed to this result.
To test whether additional genes could influence phenotype expression, 18 genes with variants in Pt2 were prioritized by the Endeavour software, [21] using “training genes” associated with MND (Additional file 6). The highest score was achieved by HDAC6, on chromosome Xp11.23, encoding a member of the histone deacetylase family (NP_006035.2); Pt2 was hemizygous for a c. 2566C>T/p.P856S, variant, whereas Pt1, II-6 and II-7 were wt, and the mother, I-2, was heterozygous (Figure 3A). Whilst the variants in the other genes were all relatively frequent SNPs and/or present also in Pt1 (Additional file 6), the P856S change was absent in all available databases, including EVS. The amount of HDAC6 transcripts was similar in fibroblasts from Pt2 vs. Pt1 or control subjects, indicating that neither HDAC6 expression nor stability is severly affected by the mutation (Figure 3B). However, acetylated alpha-tubulin, a HDAC6 substrate, [22] was consistently increased (Figure 3C); treatment of fibroblasts with tubacin, a selective HDAC6 inhibitor, clearly increased the acetylation of alpha-tubulin, confirming the specificity of this assay to detect impaired HDAC6 activity (Additional file 7).
Figure 3  Characterization of the HDAC6 c.2566C>T/p.P856S variant. A: Electropherograms of HDAC6 exon 25 region containing the c.2566C>T variant, in patients 1 and 2 (Pt1, Pt2) and their mother (I-2). B: Levels of HDAC6 transcript, normalized to that of the endogenous GAPDH cDNA, in controls (Ct; mean of five subjects) and patients 1 and 2 (Pt1, Pt2), obtained from 3 independent experiments. Two-tailed Student’s t-tests showed no significant differences: Pt2 vs. Ct p = 0.811; Pt2 vs. Pt1 p = 0.813; Pt1 vs. Ct p = 0.896. C: Exemplifying Western-blot analysis of fibroblast lysates from control subjects (Ct1, Ct2) and patients 1 and 2 (Pt1, Pt2), using antibodies against acetylated α-tubulin (upper panel), α-tubulin (middle panel) and α-GAPDH, as loading control (lower panel). The graph represents the ratio acetylated α-tubulin/α-tubulin obtained by densitometric analysis from 3 independent experiments: 100% corresponds to the mean value of four control subjects. Two-tailed Student’s t-test Pt2 vs Ct p = 0.002, Pt1 vs Ct p = 0.33. D: Immunocytochemistry on fibroblasts from a control subject (Ct) and patients 1 and 2 (Pt1, Pt2), using antibodies against α-tubulin and acetylated α-tubulin. Scale bars are reported on the right for each row. E: Percentages of multilobated nuclei in control, Patient1 and Patient2. A total of 15 digital images (at least 600 cells for each patient) representative of the whole sections were collected and analyzed for each sample; the arrow indicates a typical multilobated nucleus. Two-tailed Student’s t-test between Control vs. Pt1 showed no significant differences (p = 0.4970); Pt2 vs. Control p = 0.000099; Pt2 vs. Pt1 p = 0.000018 (both highly significant). Densitometric analysis of immunoreactive bands from three independent experiments, showed that the ratio acetylated α-tubulin/α-tubulin was significantly augmented to 213% in Pt2, compared to the mean value of four control subjects, but was unchanged (87%) in Pt1 (Figure 3C). Moreover, immunocytochemical staining showed abnormal clumps of acetylated α-tubulin in the perinuclear region of Pt2 fibroblasts (Figure 3D). Interestingly HDAC6P856S fibroblasts showed a significantly higher number of multilobated nuclei, compared to control cells, which could be consequent to altered physical connection between nuclear membrane and cytoskeletal network (Figure 3E). Taken together these results suggest dysregulation of the microtubule-organizing center (MTOC), associated with reduced HDAC6 activity [23].