BET1 variants establish impaired vesicular transport as a cause for muscular dystrophy with epilepsy

Abstract

BET1 is required, together with its SNARE complex partners GOSR2, SEC22b, and Syntaxin-5 for fusion of endoplasmic reticulum-derived vesicles with the ER-Golgi intermediate compartment (ERGIC) and the cis-Golgi. Here, we report three individuals, from two families, with severe congenital muscular dystrophy (CMD) and biallelic variants in BET1 (P1 p.(Asp68His)/p.(Ala45Valfs*2); P2 and P3 homozygous p.(Ile51Ser)). Due to aberrant splicing and frameshifting, the variants in P1 result in low BET1 protein levels and impaired ER-to-Golgi transport. Since in silico modeling suggested that p.(Ile51Ser) interferes with binding to interaction partners other than SNARE complex subunits, we set off and identified novel BET1 interaction partners with low affinity for p.(Ile51Ser) BET1 protein compared to wild-type, among them ERGIC-53. The BET1/ERGIC-53 interaction was validated by endogenous co-immunoprecipitation with both proteins colocalizing to the ERGIC compartment. Mislocalization of ERGIC-53 was observed in P1 and P2's derived fibroblasts; while in the p.(Ile51Ser) P2 fibroblasts specifically, mutant BET1 was also mislocalized along with ERGIC-53. Thus, we establish BET1 as a novel CMD/epilepsy gene and confirm the emerging role of ER/Golgi SNAREs in CMD.

Keywords: BET1; GOSR2; SNARE; epilepsy; muscular dystrophy.

Figure 1. Clinical characteristics of P1

1. Left quadriceps muscle biopsy performed at age 10 months (i–iii) demonstrating a dystrophic process with variation in fiber size (black arrow), increased connective tissue (blue arrow), and rare internalized nuclei (white arrow) on hematoxylin and eosin (H&E) staining (i) and Gömöri trichrome staining (ii) and type 1 fiber atrophy on nicotinamide adenine dinucleotide (NADH) staining (iii). Muscle biopsy in an unaffected (control) individual (iv–vi). Scale bar: 100 μm.

2. Western blot data (n = 1) with quantification demonstrated reduced α‐Dystroglycan (α‐DG) in P1’s muscle compared to control.

3. Western blot data with quantification demonstrated reduced α‐Dystroglycan in fibroblasts from P1 and P2 compared to control (c), although not to the extent of the reduction observed in fibroblasts from a patient with a confirmed α‐dystroglycanopathy (α‐DG pt) due to biallelic pathogenic variants in LARGE (n = 1).

4. Brain MRI findings in P1 at 28 months of age (top) revealing normal appearing cortex, cerebellum, and pons on T1‐weighted images and T2‐weighted image. There is evidence of mild thinning of the corpus callosum (arrow). Brain MRI in an unaffected (control) individual at 30 months of age (bottom) demonstrating normal‐sized corpus callosum.

Source data are available online for this figure.

Figure EV1. Genetic studies in P1

1. Pedigree of P1 with corresponding BET1 Sanger chromatogram depicting the paternally inherited c.202G>C; p.(Asp68His) missense variant and a maternally inherited c.134delC; p.(Ala45Valfs*2) frameshift variant (NM_005868), consistent with recessive inheritance.

2. Sashimi plots comparing fibroblast RNA sequencing reads in P1 (red) and two control muscle samples (blue and green) at the exons 3–5 of BET1.

3. Immunofluorescence images of BET1 (red) and GOSR2 (green) in fibroblasts from P1, P2, and control. In P1, BET1 immunoreactivity is reduced. Scale bar: 25 μm.

4. In the overlay, BET1 colocalization with GOSR2 and Syntaxin‐5 is reduced in P2 compared to controls while BET1 colocalization with SEC22b is unchanged. BET1 colocalization was determined using Pearson’s correlation coefficient (PCC) on regions of interest (ROIs) drawn around cells (GOSR2 P2: n = 56; C: n = 86, P = 0.0025) (Syntaxin‐5 P2: n = 27; C: n = 32 P = 0.0006) (SEC22b P2: n = 27; C: n = 30, P = 0.2093). (*P ≤ 0.05 Mann–Whitney U‐test). Data represented are mean ± SEM.

Figure 2. RNA sequencing analysis of the BET1 c.202G>C; p.(Asp68His) missense variant

1. Schematic depiction of BET1 indicating the position and conservation of the variants identified in this study. TMD = transmembrane domain.

2. Sashimi plots comparing muscle RNA sequencing reads in P1 (red) and two control muscle samples (blue and green) at the exons 3–5 of BET1.

3. Schematics of normal BET1 exon 3–4 splicing (top) and the various aberrant splice products associated with the paternally inherited p.(Asp68His) variant in P1. Splice variant one utilizes normal splicing generating a full‐length BET1 protein with the Asp68His substitution. Splice variant two skips exon 4 to non‐coding transcript (NR_133908) exon 5, resulting in a shift of the reading frame and a premature termination codon after 18 amino acids. Splice variant three utilizes a cryptic splice acceptor within exon 4, deleting the first 34 nucleotides of exon 4, and producing again a frameshift and a premature termination. The maternally inherited frameshift variant c.134delC introduces a premature stop codon in exon 2 and is predicted to induce nonsense‐mediated decay (NMD).

Figure 3. Analysis of BET1 expression and localization in patient fibroblasts and ER‐to‐Golgi transport in BET1 knockdown HeLa cells

1. Immunoblot of patient fibroblast cell lines (P1 and P2) and three independent controls (C1‐3). Immunoblots were probed with the indicated primary antibodies. Quantification of BET1 protein levels normalized to GAPDH. Data represented are mean ± SD of four independent biological experiments. Significant differences of P1 BET1‐levels were checked compared to each control (***P < 0.001, one‐way‐ANOVA followed by Tukey’s post hoc test).

2. Subcellular localization of BET1 in control and patient fibroblasts by indirect immunofluorescent confocal microscopy. Representative immunofluorescence images of BET1 (red) and cis‐Golgi marker GM130 (green). Scale bar: 20 µm. Colocalization of BET1 with GM130 was determined using Pearson’s correlation coefficient (PCC) measured per cell (right panel; C1: n = 63, C2: n = 50, C3: n = 49 and P2: n = 33) (***P ≤ 0.001, one‐way ANOVA was followed by Tukey’s post hoc test). Means are shown ± SEM.

3. Representative immunofluorescence images of BET1 (red) and ER marker PDI (green). Scale bar: 20 µm. Right panel shows PCCs for BET1 versus PDI (C1: n = 61, C2: n = 42, C3: n = 59 and P2: n = 32 of two biological replicates). (***P ≤ 0.001, one‐way ANOVA was followed by Tukey’s post hoc test). Means are shown ± SEM.

4. HeLa cells stably expressing the pC4‐reporter were transfected with control siRNA (ctrl) or two independent siRNAs (#1 and #2) directed against BET1 (siRNA #1 is shown in figure EV 2B). 2d post‐transfection the pC4‐secretion assay was performed for 0 and 10 min post‐induction of secretion. Representative images are shown (upper panel ctrl and lower panel #2 siRNA, respectively. Scale bar: 20 µm. Right panels represent magnifications of dashed areas.

5. Quantification of colocalization using the Pearson's correlation coefficient (PCC) of the pC4‐construct (GFP) and the cis‐Golgi marker GM130 (0 min: ctrl n = 63, #1 n = 56, #2 n = 64; 10 min: ctrl n = 104, #1 n = 126, #2 n = 110). Shown are means ± SEM. Colocalization is significantly reduced by BET1 knockdown (siRNA #1 and #2) compared to control transfected cells. Significant differences of PCC (pC4:GM130) after 10 min induction with solubilizer were analyzed with a one‐way‐ANOVA followed by Dunnett's post hoc test. ***P ≤ 0.001. 0 min: ctrl 0.265 ± 0.016, #1 0.268 ± 0.017, #2 0.264 ± 0.018; 10 min: ctrl 0.798 ± 0.009, #1 0.726 ± 0.011, #2 0.715 ± 0.011.

6. Quantification of knockdown efficiency of BET1 in HeLa‐pC4 cells by immunoblots. Data represented are mean ± SD of three biological replicates. ***P < 0.001 compared to ctrl cells, one‐way‐ANOVA followed by Dunnett’s post hoc test).

Source data are available online for this figure.

Figure EV2. Normal SNARE complex partners in BET1 fibroblasts with reduced colocalization at the cis‐Golgi in transfected cells

1. Quantification of protein levels of BET1 SNARE complex partners and the novel interaction partner ERGIC‐53 in fibroblasts. Values are normalized to GAPDH. No significant differences between the control and patient fibroblasts were detected. Data represented are mean ± SD of four biological replicates. Significant differences were analyzed by one‐way ANOVA followed by Tukey's post hoc test.

2. Immunofluorescence images of BET1 knockdown HeLa pC4 cells transfected with BET1 siRNA #1 10 min after induction of ER‐to‐Golgi transport with solubilizer. Scale bar: 20 µm.

3. HeLa pC4 cells were transfected with control siRNA (ctrl) or siRNA against BET1 (#1 and #2) and protein levels of BET1 and ER‐to‐Golgi complex SNAREs SEC22b, Syntaxin‐5, and GOSR2 were analyzed by immunoblot. *residual SEC22b signal after immunoblot stripping.

4. Quantification of EV2C. Data represented are mean ± SD of three biological replicates. Significant differences to ctrl cells were analyzed by one‐way ANOVA followed by Dunnett’s post hoc test. ***P ≤ 0.001.

5. Table showing a spectral count‐based summary of three independent AP‐MS experiments with significant hits shown in Fig 5A. Hits are sorted by the difference of BET1‐WT and BET1‐Ile51Ser. Hits which do not fulfill the criteria to have a mock‐transfected spectral count of two or less are colored in grey.

Figure 4. In silico characterization of the SNARE complexes bearing the Asp68His and the Ile51Ser variant

1. Representation of the SNARE complex with Asp68 highlighted (top). BET1^Asp68 is involved in a hydrogen bond with SEC22b^Lys169, with the most probable conformation of His68 in the mutant variant (bottom).

2. Stability of hydrogen bonds between BET1^Asp68 and BET1^Asp72 respectively was estimated using molecular dynamics simulation in three replicas. Parameter occupancy could be more than 100% if several stable hydrogen bonds could be established by residue. Bars represent average fraction of the simulated time the residue is involved in a hydrogen bond(s). Data represented are mean ± SD.

3. Sequence conservation of the corresponding regions of BET1 and SEC22b animal proteins. The relevant residues involved in hydrogen bond network are highlighted by color, hydrogen bonds are shown.

4. Representation of the SNARE complex with Ile51 (upper) and Ser51 (lower) residues highlighted.

5. Sequence conservation of the corresponding regions of BET1 and SEC22b and the relevant residues involved in the hydrophobic patch formed by Ile51 BET and SEC22b^Ile148.

Figure 5. ERGIC‐53 is a novel interaction partner of BET1 with impaired interaction of Ile51Ser BET1 and ERGIC‐53

1. Volcano plot of AP‐MS data using HA‐BET1 or Ile51Ser.

2. Co‐IPs using HA‐tagged constructs expressed in HEK293 cells. As baits HA‐tagged proteins were immunoprecipitated with anti‐HA beads. IP samples and cell lysates were subjected to immunoblotting with α‐HA, α‐ERGIC‐53, α‐BET1, α‐SEC22b, α‐GOSR2, and α‐GAPDH antibodies, respectively.

3. Quantification of three independent replicates of Co‐IP experiments. Protein levels of the Co‐IP samples were normalized to BET1‐HA‐levels. Data represented are mean ± SD of three biological replicates. Significant differences were checked by one‐way ANOVA followed with Tukey's post hoc test. *P ≤ 0.05.

4. Endogenous Co‐IP of BET1 and ERGIC‐53. Immunoblots of HEK293T cell lysates (+lysate) or lysis buffer (−lysate) incubated with (+) or without (−) BET1 antibody followed by precipitation with protein A/G beads. ERGIC‐53 signal was only present in samples with lysate and antibody, whereas * denotes unspecific signals from beads used for precipitation. As a positive control, GOSR2 antibody was used.

5. Subcellular localization of ERGIC‐53 in fibroblasts by indirect immunofluorescent confocal microscopy. Representative immunofluorescence images of ERGIC‐53 (green) and GM130 (red) in fibroblasts from P1, P2, and a control cell line. Scale bar: 20 µm. PCC (right panel) for GM130 versus ERGIC‐53 (C1: n = 74, C2: n = 95, C3: n = 59, P1: n = 72 and P2: n = 76). (***P ≤ 0.001, one‐way ANOVA was followed by Tukey’s post hoc test). Shown are means ± SEM. Immunofluorescence images shown are one example of a total of two biological replicates.

6. Colocalization of ERGIC‐53 and BET1 in control and P2‐derived fibroblasts. Representative immunofluorescence images of ERGIC‐53 (green) and BET1 (red) in fibroblasts from P2 and a control cell line. Scale bar: 20 µm. PCC (right panel) for BET1 versus ERGIC‐53 (C1: n = 72, C2: n = 46, C3: n = 56 and P2: n = 37). (***P ≤ 0.001, one‐way ANOVA was followed by Tukey’s post hoc test). Data represented are mean ± SEM of two independent experiments.

7. Immunofluorescence images of control fibroblasts (C1) and patient fibroblasts (P1 and P2) treated with Brefeldin A for 40 min and Golgi recovery for indicated time points. Golgi reconstitution was analyzed by immunofluorescence microscopy of the cis‐Golgi marker GM130. Cells displaying an intact Golgi are marked with a plus (+). Scale bar: 20 µm.

8. The number of cells with an intact Golgi was divided by the number of total cells per image and plotted against time (three biological replicates). Dots are means ± SD. For quantification, a sigmoidal fit was performed with the software GraphPad Prism 7. Dashed lines indicate extrapolated plateaus. LogIC50 ± 95% confidence interval: C1 = 59.38 ± 2.31 min, P1 = 72.89 ± 1.65 min, P2 = 68.57 ± 1.929 min. Significant differences were assumed when 95% confidence interval do not overlap.

Source data are available online for this figure.

Figure EV3. The effect of BET1 variants on yeast growth

BET1 Leu77Ile (Leu51Ile) and Leu77Ser (Leu51Ser) show similar growths compared to BET1 wild type. Temperature‐sensitive strain of S. cerevisiae was transformed with wild‐type BET1, the two different variants, and a mock plasmid. At 24°C, growth was detected for all variants and wild type, including the negative control. Only the negative control showed an impairment of growth at restrictive temperatures (30°C and 33°C).