Truncated mutants of beta-glucosidase 2 (GBA2) are localized in the mitochondrial matrix and cause mitochondrial fragmentation

The enzyme β-glucosidase 2 (GBA2) is clinically relevant because it is targeted by the drug miglustat (Zavesca®) and because it is involved in inherited diseases. Mutations in the GBA2 gene are associated with two neurological diseases on the ataxia-spasticity spectrum, hereditary spastic paraplegia 46 (SPG46) and Marinesco-Sjögren-like syndrome (MSS). To establish how GBA2 mutations give rise to neurological pathology, we have begun to investigate mutant forms of GBA2 encoded by disease-associated GBA2 alleles. Previously, we found that five GBA2 missense mutants and five C-terminally truncated mutants lacked enzyme activity. Here we have examined the cellular locations of wild-type (WT) and mutant forms of GBA2 by confocal and electron microscopy, using transfected cells. Similar to GBA2-WT, the D594H and M510Vfs*17 GBA2 mutants were located at the plasma membrane, whereas the C-terminally truncated mutants terminating after amino acids 233 and 339 (GBA2-233 and -339) were present in the mitochondrial matrix, induced mitochondrial fragmentation and loss of mitochondrial transmembrane potential. Deletional mutagenesis indicated that residues 161–200 are critical for the mitochondrial fragmentation of GBA2-233 and -339. Considering that the mitochondrial fragmentation induced by GBA2-233 and -339 is consistently accompanied by their localization to the mitochondrial matrix, our deletional analysis raises the possibility that that GBA2 residues 161–200 harbor an internal targeting sequence for transport to the mitochondrial matrix. Altogether, our work provides new insights into the behaviour of GBA2-WT and disease-associated forms of GBA2.


Cell culture
U2OS, SH-SY5Y, and HeLa cells (ATCC) were maintained in DMEM supplemented with 10% fetal calf serum. Primary rat hippocampal cells were obtained and cultured as described [31]. Animal use for this purpose was approved by the Dalhousie University Committee on Laboratory Animals (approval number . Rats were sacrificed by anesthesia with isofluorane (5% v/v) followed by CO2 asphyxiation. Hippocampi were dissected from E18 Sprague-Dawley rat embryos, incubated with 0.03% trypsin for 15 minutes, and dissociated using a fire-polished Pasteur pipette. Cells were then plated on coverslips coated with 0.1% (w/v) poly-Llysine (Peptides International) at a density of 3-6x10 3 cm -1 in Neurobasal Medium (Invitrogen) supplemented with 2% B-27 (Invitrogen), 50 μM glutamine, 25 μM glutamate, 5% fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. After 4 hours of plating, the medium was replaced with serum-free Neurobasal Medium supplemented with B-27 and 0.5 mM glutamine. One-third of the medium was replaced on a weekly basis until transfection.
To indirectly detect APEX2-tagged proteins via fluorescence microscopy, cells grown on glass coverslips were incubated with biotin-phenol, briefly exposed to hydrogen peroxide, quenched and washed as described [33], then fixed, permeabilized, and washed as described above, and incubated with Alexa594-conjugated streptavidin at 1:1000 (Invitrogen S11227).

Electron microscopy
Electron microscopy using the APEX2 tag was performed as described [34]. At 48 hours of post-transfection, cells expressing APEX2 constructs were briefly washed with cacodylate buffer containing 2% glutaraldehyde at 37˚C, fixed with cacodylate/glutaraldehyde for 1 hour on ice, washed with ice-cold 100 mM sodium cacodylate buffer, pH 7.4 (5 x 2 min), quenched with 20mM glycine in sodium cacodylate buffer (10 min), and washed again with cacodylate buffer (5 x 2 min). Cells were incubated in 0.5 mg/ml diaminobenzidine (from freshly prepared 10x stock in 100 mM HCl) and H 2 O 2 (0.03%) in cacodylate buffer for 25 minutes, washed with cacodylate buffer, fixed with freshly prepared 2% osmium tetroxide (OsO 4 ) for 30 minutes, washed with ice-cold milli-Q water several times, and incubated with 2% uranyl acetate solution overnight. Cells were gently scraped off the culture dish and pelleted, dehydrated through a graduated series of acetone, and embedded in Epon Araldite resin, which was cured at 60˚C for 48 hours. Ultrathin sections were cut using a Reichert-Jung Ultracut E Ultramicrotome with a diamond knife, placed on 300 mesh copper grids which, and stained with 2% uranyl acetate and lead citrate. Sections were viewed using a JEOL JEM 1230 Transmission Electron Microscope at 80kV and imaged using a Hamamatsu ORCA-HR digital camera.

Mitochondrial membrane potential
U2OS cells seeded on 35 mm glass-bottom culture dishes (Ibidi USA Inc., Fitchburg, WI; 100,000 cells/dish) were transfected to co-express mClover2 and different forms of GBA2. At 48 hours post-transfection, the culture medium was replaced with FluoroBrite DMEM (Gibco, Grand Island, NY) containing 10% FBS and 200 μM L-glutamine, supplemented with tetramethyl-rhodamine methyl ester (TMRM, Invitrogen, Carlsbad, CA; final concentration of 25 nM) with or without FCCP (Sigma; 10 μM). Cells were incubated for 15 minutes at 37˚C and imaged using a Zeiss LSM 510 Meta confocal microscope. Green channel images (mClover2 fluorescence) were used to draw contours around individual transfected cells, which were copied to corresponding red channel images and used as regions of interest to determine integrated TMRM intensities using ImageJ software, corrected for background intensity. For each experimental condition, 30 cells were analyzed, in triplicate.

Statistics
Quantitative data were analyzed for statistical significance by ANOVA with Tukey's multiple comparisons test using GraphPad Prism 8.3.0 software.

Results
To extend our previous characterization of wild-type and mutant forms of GBA2 [22], we compared the cellular location of GBA2-WT with those of the missense D594H mutant, the frameshift M510Vfs � 17 mutant, and the C-terminally truncated mutants terminating at residues 233 and 339 (GBA2-233, and GBA2-339) ( Fig 1A). We expressed FLAG-tagged GBA2-WT under the control of the cytomegalovirus immediate-early promoter (CMV; Fig  1B) in HeLa, SH-SY5Y, and primary rat hippocampal cells, and in U2OS cells, and detected the protein using an anti-FLAG antibody via western blotting ( Fig 1E) and immunostaining. GBA2-WT-FLAG was widely distributed throughout the cell in HeLa, SH-SY5Y, and primary rat hippocampal cells (S1A Fig, S1B Fig and S1C Fig) and in U2OS cells (Fig 2A). In another approach to localize GBA2-WT, we used the APEX2 tag (Fig 1C), which is an engineered form of soybean ascorbate peroxidase [33]. APEX2-tagged proteins traffic to many different cellular locations, depending on the protein that carries the APEX2 tag, and are transported to the same locations as the endogenous versions of these proteins [34]. APEX2 biotinylates nearby proteins when activated with hydrogen peroxide and provided with biotin-phenol [33]. Biotinylated proteins can then be detected with fluorescently labeled streptavidin. Following this approach, GBA2-WT-APEX2 ( Fig 1F) displayed a wide distribution (S2A Fig), similar to GBA2-WT-FLAG (Fig 2A). APEX2 also can convert diaminobenzidine (DAB) into an insoluble osmiophilic polymer that can be directly observed via electron microscopy [34]. Upon being oxidized by APEX2, DAB remains tightly localized to its site of production [35]. We expressed APEX2-tagged GBA2-WT and exposed the cells to diaminobenzidine. GBA2-W-T-APEX2 stained the plasma membrane ( Fig 2B). Cells expressing GBA2-WT-FLAG and GBA2-WT-APEX2 displayed mitochondrial networks containing tubules of various lengths that closely resembled those of non-transfected cells (S1 Fig and Fig 2A).

GBA2 mutants
Similar to GBA2-WT, the D594H and M510Vfs � 17 mutants were present throughout the cell (S3 Fig). By contrast, GBA2-233 was present in puncta that overlapped with a mitochondrial marker in HeLa, SH-SY5Y, and primary rat hippocampal cells (S1B Fig, S1D Fig and S1F Fig) and in U2OS cells ( Fig 3A); occasionally, the punctate pattern of GBA2-233 was accompanied by an overall staining. As U2OS cells are very flat and wide, and their mitochondria lie in one focal plane without overlapping much, we used this cell line to further investigate the impact of the truncated GBA2 mutants of mitochondria.
GBA2-233 colocalized to a much higher degree with mitochondria compared to GBA2-WT ( Fig 4A). Also, the majority of cells expressing GBA2-233 displayed very short mitochondria appearing as puncta of various sizes (Fig 3A). GBA2-233-APEX2 (Fig 1C and 1F) also localized to mitochondria with a fragmented phenotype, as detected with fluorescently labeled streptavidin (S2B Fig). At the ultrastructural level, GBA2-233-APEX2 stained the mitochondrial matrix ( Fig 3B). To establish whether GBA2-233 was imported into mitochondria and altered mitochondrial morphology because of its high expression level, we expressed this mutant at a lower level using the mouse stem cell virus long-terminal repeat (MSCV LTR) as promoter (Fig 1D and 1E). Under these conditions, GBA2-233 was also present in mitochondria and caused mitochondrial fragmentation (S4B Fig), albeit in a smaller proportion of transfected cells compared to expression driven by the CMV promoter (Fig 4B and 4C). Likewise, the SPG46-associated GBA2 mutant terminating at residue 339 (GBA2-339, Fig 1A and 1B) colocalized to a high degree with mitochondria (Figs 4A and 5A), was present in the mitochondrial matrix (Fig 5B), and caused mitochondrial fragmentation in the majority of cells, both when expressed at higher and lower levels (Fig 4B and 4C). Further, our results show that the mitochondrial localization of GBA2-233 and GBA2-339 and the corresponding mitochondrial fragmentation is independent of their C-terminal tags, as FLAG-tagged and APEX2-tagged forms of these proteins behave in a similar fashion.

Mitochondrial transmembrane potential
Considering that the mitochondrial transmembrane potential is a prerequisite for the ability of mitochondria to fuse [36], we evaluated whether the mitochondrial fragmentation described above was associated with a reduction in the transmembrane potential. To this end, we used cDNA constructs coding for mClover2, a green fluorescent protein, in tandem with either GBA2-WT or a disease-associated GBA2 mutant, separated by the ribosome-skipping T2A sequence (Fig 6A). To assess the mitochondrial transmembrane potential, we used the fluorescent indicator tetramethylrhodamine methyl ester (TMRM), which is a lipophilic compound carrying a delocalized positive charge. TMRM is therefore membrane-permeant, enters live cells, and distributes across cell membranes in a voltage-dependent manner, accumulating in negatively charged mitochondria in proportion to the potential across the inner membrane [37][38][39]. Non-transfected cells and cells expressing mClover2 together with GBA2-WT, GBA2-D594, or GBA2-M510Vfs � 17 accumulated TMRM in their mitochondria to similar levels (Fig 6C and 6D, left panels), which was prevented by co-incubating the cells with the mitochondrial uncoupling agent carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) (Fig 6C and 6D, right panels). By contrast, cells expressing mClover2 and GBA2-233 accumulated TMRM at much lower levels, comparable to cells co-incubated with FCCP (Fig 6C and  6E). These results show that GBA2-233 caused the mitochondria to lose their transmembrane potential.

Discussion
Previously, we have investigated the enzymology and pharmacology of GBA2-WT in vitro and in vivo [1,2,4]. We have also established that nonsense and missense mutations in the GBA2 gene associated with cerebellar ataxia/spastic paraplegia abolish most of its enzyme activity [22]. Here we have examined the cell biology of wild-type and mutant forms of GBA2, using confocal and electron microscopy. Our data show that overexpressed GBA2-WT is located at the plasma membrane. The plasma membrane staining we observed by electron microscopy in GBA2-WT-APEX2-transfected cells closely resembles that of cells expressing a plasma membrane-targeted form of APEX2 [33]. Our findings align with those of Boot et al. [40], who observed GFP-GBA2 and GBA2-GFP at the plasma membrane of COS-7 cells. Also, many of the fluorescence microscopy images of Körschen et al. [41] show a wide distribution of differently-tagged forms of murine Gba2 throughout transfected HEK293 cells (eGFP-Gba2,   Gba2-eGFP, and Gba2-HA), partially overlapping with the ER and the Golgi. Further, these authors show a wide distribution of endogenous Gba2-WT in murine hippocampal neurons. By contrast, Yildiz et al. [5] show a reticular immunostaining for endogenous Gba2 in COS cells, which would suggest an ER localization. It thus appears that overexpressed GBA2-WT possibly has a different cellular location (plasma membrane) compared to endogenous GBA2-WT (ER). This possibility needs to be addressed.
We have further established that C-terminally truncated GBA2 mutants (GBA2-233 and GBA2-339) are efficiently imported into the mitochondrial matrix, causing mitochondria to lose transmembrane potential and adopt a fragmented morphology. Using the APEX2-tag, we observed at the ultrastructural level that GBA2-233 and GBA2-339 are located in the mitochondrial matrix. Martell et al. [34] showed a similar mitochondrial matrix staining in COS-7 cells expressing a form of APEX targeted to the mitochondrial matrix via a canonical N-terminal mitochondrial targeting sequence. By contrast, a protein domain located at the matrix-side of the inner mitochondrial matrix (the mitochondrial calcium uniporter N-terminally tagged with APEX) gave a much more limited DAB staining that left most of the matrix relatively electron-transparent [34].
In addition, we have identified amino acids 160-200 as the domain instrumental for mitochondrial fragmentation in cells overexpressing GBA2-339. We consistently observed that, in cells expressing GBA2-233 and GBA2-339, mitochondrial fragmentation is accompanied by mitochondrial localization of GBA2-233 and GBA2-339. On this basis, it is not unlikely that amino acids 160-200 of GBA2 are also responsible for the mitochondrial localization of GBA2-233 and GBA2-339, harboring an internal mitochondria-targeting domain. Internal targeting sequences have previously been identified in a small number mitochondrial matrix proteins [42] and can be activated in various ways [43], raising the possibility that GBA2-WT might-under certain conditions-be imported into the mitochondrial matrix. This possibility remains to be experimentally addressed, but would align with an earlier study identifying rat Gba2 in the proteome of brain mitochondria [44]. Moreover, GBA2 tagged with green fluorescent protein co-purified with six mitochondrial proteins, including the matrix protease AFG3L2 and the mitochondrial ribosomal proteins MRPL44 and MRPS22 [45,46]. Mutations in the genes coding for AFG3L2 and its binding partner paraplegin are associated with a form of spinocerebellar ataxia (SCA28) [47] and a form of hereditary spastic paraplegia (SPG7) [48], respectively.
Compared to human GBA2, the localization of murine Gba2 has not been investigated at the ultrastructural level, so it is not known whether a minority of murine Gba2 is present in mitochondria. However, C-terminally truncated forms of murine Gba2 corresponding to GBA2-233 and GBA2-339 are not localized in mitochondria, but are widely distributed throughout the cell, similar to Gba2-WT and GBA2-WT [17]. Our results thus show that Cterminally truncated mutants of human GBA2 behave differently compared to corresponding mutants of murine Gba2. The basis of this disparity remains to be established. Alternatively, considering that the effects of pharmacologically inhibiting Gba2 in inbred mice greatly depend on their genetic background [14,15], the possibility remains that murine Gba2 has a more pronounced mitochondrial role in mice with a genetic background distinct from that of the Gba2-deficient mice studied thus far.
Finally, the question is to what extent our observations apply to SPG46 and MSS patients. Considering that GBA2-233 and GBA2-339 cause loss of mitochondrial transmembrane potential and mitochondrial fragmentation in cultured cells, it is not likely that this also happens in patient cells. Persistent mitochondrial fragmentation and subsequent lack of ATP and of neuronal transmembrane potential are incompatible with neuronal activity and may be non-viable. It is more plausible that, in SPG46 patients homozygous for the c.700C>T (p. R234 � ) and c.1018C>T (p.R340 � ) mutations, the expression of GBA2-233 and GBA2-339 is attenuated through nonsense-mediated decay, a process that recognizes mRNAs carrying premature stop codons, targeting them for degradation [49]. Altogether, the pathology of these patients is likely due to a deficit of the activities of GBA2-WT, similar to patients carrying other mutations in GBA2. GBA2-233-APEX2 were incubated with biotin-phenol and briefly exposed to hydrogen peroxide, which activates the peroxidase activity of APEX2. Biotinylated proteins were detected with Alexa594-conjugated streptavidin (red) while mitochondria were stained with anti-TOMM20 (green). Scale bar, 20 μm. advice on using the APEX2 tag for electron microscopy, to Mary Ann Trevors for expert assistance with electron microscopy sample preparation, and to Stefan Krueger for providing primary rat neuronal cells.