Altered distribution of ATG9A and accumulation of axonal aggregates in neurons from a mouse model of AP-4 deficiency syndrome

Abstract

The hereditary spastic paraplegias (HSP) are a clinically and genetically heterogeneous group of disorders characterized by progressive lower limb spasticity. Mutations in subunits of the heterotetrameric (ε-β4-μ4-σ4) adaptor protein 4 (AP-4) complex cause an autosomal recessive form of complicated HSP referred to as "AP-4 deficiency syndrome". In addition to lower limb spasticity, this syndrome features intellectual disability, microcephaly, seizures, thin corpus callosum and upper limb spasticity. The pathogenetic mechanism, however, remains poorly understood. Here we report the characterization of a knockout (KO) mouse for the AP4E1 gene encoding the ε subunit of AP-4. We find that AP-4 ε KO mice exhibit a range of neurological phenotypes, including hindlimb clasping, decreased motor coordination and weak grip strength. In addition, AP-4 ε KO mice display a thin corpus callosum and axonal swellings in various areas of the brain and spinal cord. Immunohistochemical analyses show that the transmembrane autophagy-related protein 9A (ATG9A) is more concentrated in the trans-Golgi network (TGN) and depleted from the peripheral cytoplasm both in skin fibroblasts from patients with mutations in the μ4 subunit of AP-4 and in various neuronal types in AP-4 ε KO mice. ATG9A mislocalization is associated with increased tendency to accumulate mutant huntingtin (HTT) aggregates in the axons of AP-4 ε KO neurons. These findings indicate that the AP-4 ε KO mouse is a suitable animal model for AP-4 deficiency syndrome, and that defective mobilization of ATG9A from the TGN and impaired autophagic degradation of protein aggregates might contribute to neuroaxonal dystrophy in this disorder.

Fig 1. Molecular and behavioral characteristics of AP-4 ε KO mice.

(A) Schematic representation of the strategy for generation of the AP-4 ε (AP4E1) KO mouse. Exons are numbered. Exon 3 was deleted by Cre-mediated recombination, resulting in a null allele caused by a frameshift and premature termination of the ORF. The neomycin cassette used for selection, as well as Lac Z reporter sequences, were removed by Flp-mediated recombination. The positions of the PCR primers used for genotyping are indicated. (B) Genotyping of +/+ (WT), -/- (KO) and +/- (heterozygous) mice was done using PCR primers complementary to sequences flanking exon 3. The WT and KO alleles generated 932 and 269 bp fragments, respectively. (C) SDS-PAGE and immunoblot analysis of AP-4 ε and β4 subunits, AP-1 γ1 subunit, and β-tubulin (control) in homogenates of cerebellum and cerebral cortex from two WT and two KO mice. The positions of molecular mass markers (in kDa) are indicated on the left. (D) Representative picture of the clasping response of 5-month-old WT and KO mice held by the tail. Clasping of hind paws in the KO mouse is indicated by red arrowheads. (E) Percentage of WT and KO mice that show clasping at 5 and 8 months (mo) of age. Numbers on top of each bar indicate number of clasping/total mice in each group (5 mo WT: 5 male, 8 female, KO: 4 males, 5 females; 8 mo WT: 4 male, 3 female, KO: 7 male, 3 female). (F) Performance of 3-month-old mice on the rotarod. Numbers of mice used were: 22 WT (12 male, 10 female) and 19 KO (10 male, 9 female). Individual points show the RPM at which each mouse fell from the rod. Bars indicate the mean ± SEM for each group. (G) Grip strength of 3-month-old mice. Numbers of mice used were: 27 WT (13 male, 14 female) and 25 KO (13 male, 12 female). Individual points show the grip strength in Newton (N) and bars indicate the mean ± SEM for each group. (H) Novelty-induced locomotor activity of 3-month-old mice in the open field test. Numbers of mice used were: 12 WT (6 male, 6 female) and 12 KO (6 male, 6 female). Values are the mean ± SEM of the distance traveled every 5 min. The inset describes the cumulative total distance traveled in meters (m) during a total of 30 min. (I) Acoustic startle response of 3-month-old mice subjected to increased noise levels in decibels (dB). Numbers of mice used were: 22 WT (11 male, 11 female) and 22 KO (10 male, 12 female). Values are the mean ± SEM of the startle response in arbitrary units (A.U.). *P<0.05, **P<0.005, ***P<0.0005.

Fig 2. Thin corpus callosum and axonal spheroids in the brain of AP-4 ε KO mice.

(A) MRI shows thinner corpus callosum (arrowheads and green highlights) in the brain of a 9-month-old KO relative to an age-matched WT mouse. Results are representative of three mice per group. (B) Coronal brain sections stained with H&E show ~55% thinner corpus callosum in a 9-month-old KO mouse relative to an aged-matched WT mouse. Bars: 500 μm. The thickness of the corpus callosum in comparable sections is indicated. Results are representative of two mice per group. (C,D) H&E staining of coronal sections of brains from 9-month-old mice show the appearance of both weakly and strongly stained spheroids (arrowheads) in deep cerebellar nuclei (DCN) of KO but not WT mice. The numbered panels are four-fold magnifications of the boxed areas. Bars: 20 μm. (E,F) Immunohistochemical staining for calbindin (green) in sections from the cerebellum shows rare axonal spheroids in the proximal axons of Purkinje neurons (E) and numerous spheroids in distal axons of Purkinje neurons in DCN (F) of KO mice but not WT mice. Spheroids are indicated by arrowheads. Bars: 50 μm. (G) Immunohistochemical co-staining for calbindin (green) and LAMP1 (red) of sections from the cerebellum show that Purkinje cells spheroids in DCN from KO mice are negative for LAMP1 (arrowheads). Bars: 50 μm. (H) Immunohistochemical co-staining for the neurofilament protein NFH (green) and LAMP1 (red) of sections from midbrain white matter tracts show spheroids containing both proteins (arrowheads) in KO mice. Bars: 50 μm. In G and H, single-channel images are shown in inverted grayscale. In E-H, DAPI (blue) was used to stain the nuclei.

Fig 3. Distribution of glutamate receptors in Purkinje neurons of the cerebellum from WT and AP-4 ε KO mice.

(A-C) Sections of the cerebellum from WT and AP-4 ε KO mice were co-immunostained with antibodies to calbindin (red) and δ2R (AB_2571601 from Frontier Institute Co., ltd) (A), GluA2 (SAB4501295 from Sigma) (B), or GluA1 (AB1504 from Millipore) (C). Staining for δ2R was done on sections treated with pepsin and staining for GluA1 on sections treated at 85°C with sodium-citrate buffer, in both cases for antigen retrieval. Similar results were obtained using samples that were not treated with high temperature sodium-citrate buffer (S5C and S5D Fig) and using a different antibody to GluA1 (antibody ab31232 from Abcam) (S5A and S5B Fig). Mo: molecular layer; Gr: granular layer, Pc: Purkinje cell layer, DCN: deep cerebellar nucleus. DAPI (blue) was used to stain the nuclei. Bars: 50 μm. Examples of spheroids are indicated by arrowheads. Single-channel images are shown in inverted grayscale. Notice the presence of δ2R and GluA2, and the absence of GluA1, in calbindin-positive spheroids in the DCN of AP-4 ε KO.

Fig 4. Accumulation of ATG9A at the TGN of AP-4 μ4 mutant patient fibroblasts.

(A) Skin fibroblasts from one control individual and two patients homozygous for mutations in the AP4M1 gene encoding AP-4 μ4 [7] were analyzed by SDS-PAGE and immunoblotting for the ε subunit of AP-4, ATG9A, and α-tubulin (loading control). The positions of molecular mass markers (in kDa) are indicated on the left. (B,C) Co-immunostaining for endogenous ATG9A (green) and AP-4 ε (red) (B) or GM130 (red) (C) of the fibroblasts mentioned in A. Bars in B and C: 20 μm. DAPI (blue) was used to stain the nuclei. (D) Quantification of the distribution of ATG9A relative to the nucleus in fibroblasts from control (n = 9), patient 1 (n = 7) and patient 2 (n = 7) using ImageJ with the Radial Profile plugin. Values are the mean ± SEM of fluorescence intensity relative to the total cells intensity in each group. Notice the concentration of ATG9A at the TGN and its depletion from the peripheral cytoplasm in the patients’ fibroblasts.

Fig 5. Accumulation of ATG9A at the TGN in neurons from AP-4 ε KO mice.

(A) Co-immunostaining of hippocampal neurons from WT and KO mice for ATG9A (green) and GM130 (red). Single-channel images are shown in inverted grayscale. Outlines of the neurons are indicated by dashed lines. Images on the rightmost column are 3-fold-magnified views of the neuronal soma from the merged images. Bars: 10 μm. (B) Rescue of the normal distribution of endogenous ATG9A by transfection of plasmids encoding HA-tagged ε and PM-RFP into hippocampal neurons from KO mice. Neurons were stained for endogenous ATG9A (green) and GM130 (white), and for the HA epitope (blue). The six panels on the right are single-channel, inverted grayscale, magnified views of the boxed areas 1 and 2 on the left. Bar: 20 μm. Notice the concentration of ATG9A at the TGN and concomitant depletion from the periphery in the AP-4 ε KO neurons, and the reversal of this phenotype in the HA-ε-rescued neurons.

Fig 6. Accumulation of ATG9A at the TGN in different brain regions from AP-4 ε KO mice.

Immunostaining of sections from the cerebral cortex, cerebellar cortex, hippocampus, and spinal cord of WT and AP-4 ε KO mice using an antibody to endogenous ATG9A (green). Nuclei were stained with DAPI (blue). Images on the second and fourth columns are magnifications of the boxed areas on the first and third columns, respectively. Bars: 50 μm (zoom-out view) and 10 μm (magnification). Notice the brighter ATG9A staining at the TGN in neurons from AP-4 ε KO mice.

Fig 7. Increased levels of ATG9A and normal levels of other autophagy proteins in brain and neurons from AP-4 ε KO mice.

(A) Homogenates of the olfactory bulb (OB), cerebral cortex, cerebellum (Cerebell), hippocampus (Hippo), midbrain and brain stem from WT and AP-4 ε KO mice were analyzed by SDS-PAGE and immunoblotting with antibodies to the proteins indicated on the right. The positions of the I and II forms of LC3B are indicated. The positions of molecular mass markers (in kDa) are indicated on the left. (B) Quantification of ATG9A levels from experiments as in A. Values are the mean ± SEM from four determinations. *P<0.05, **P<0.005. Notice the increased levels of ATG9A in some brain regions of AP-4 ε KO vs. WT mice. (C) Quantification of ATG5 levels from experiments as in A. Values are the mean ± SEM from four determinations. Notice that ATG5 levels are not significantly changed in AP-4 ε KO vs. WT mice. (D) Cortical neurons in primary culture from WT and AP-4 ε KO mice were incubated in the absence (n.t.) or presence of 100 nM bafilomycin A1 (Baf) for 5 h. Total proteins extracts were analyzed by SDS-PAGE and immunoblotting for LC3B, SQSTM1, and β-tubulin. The positions of the I and II forms of LC3B are indicated. The positions of molecular mass markers (in kDa) are indicated on the left. (E-G) Quantification of LC3B-II to LC3B-I ratio (E), total LC3B (F) and total SQSTM1 (G) relative to β-tubulin from experiments such as that in D. Values are the mean ± SD from three determinations. *P<0.05. Notice the significant increase of SQSTM1 after incubation with bafilomycin in both WT and AP-4 ε KO cortical neurons. Also notice that AP-4 ε KO does not affect LC3B processing and autophagic flux.

Fig 8. Increased accumulation of mutant huntingtin aggregates and decreased mobility of autophagosomes in the axon of AP-4 ε KO hippocampal neurons.

(A) Imaging of cultured hippocampal neurons from WT and AP-4 ε KO mice co-transfected with plasmids encoding aggregation-prone HTT103Q-GFP (green) and cyan fluorescent protein (CFP) together with either LC3B-mCherry (mCh) (red) (top two rows), or ATG9A-mCherry (red) (bottom row). Single-channel images are shown in inverted grayscale. Arrowheads indicate axonal aggregates containing both GFP-and mCherry-labeled proteins. Bars: 20 μm. (B) Magnified and straightened axon terminals and dendrites from WT and AP-4 ε KO neurons shown in A. (C) Quantification of the number of HTT103Q-GFP foci with LC3B-mCherry or ATG9A-mCherry per 100 μm of axon. Values are the mean ± SD from 10 neurons. **P<0.005, ***P<0.0005. Notice the increased number of axonal HTT103Q-GFP and LC3B-mCherry aggregates in KO relative to WT neurons, and the suppression of this increase by overexpression of ATG9A-mCherry. (D) Straightened 50 μm segments of axon terminals from DIV8 WT and AP-4 ε KO hippocampal neurons expressing LC3B-GFP (top) and corresponding kymographs (bottom). Vesicles moving to the right or the left of the top panel represent anterograde or retrograde movement, respectively. Lines with negative or positive slopes in the kymographs (bottom) correspond to anterograde or retrograde movement, respectively. Note the retrogradely moving LC3B-GFP vesicles in the WT axon, and the absence of moving vesicles in the AP-4 ε KO axon. (E) Quantification of the number of anterograde, retrograde and static LC3B-GFP particles in the distal axon of WT and AP-4 ε KO neurons. The number of moving and static particles is expressed as a percentage of the total number of events in each kymograph. Values are the mean ± SEM from nine WT and five AP-4 ε KO neurons. *P<0.05, **P<0.005, ***P<0.0005.