For inherited genetic diseases, fetal gene therapy offers the potential of prophylaxis against early, irreversible and lethal pathological change. To explore this, we studied neuronopathic Gaucher disease (nGD), caused by mutations in GBA. In adult patients, the milder form presents with hepatomegaly, splenomegaly and occasional lung and bone disease; this is managed, symptomatically, by enzyme replacement therapy. The acute childhood lethal form of nGD is untreatable since enzyme cannot cross the blood–brain barrier. Patients with nGD exhibit signs consistent with hindbrain neurodegeneration, including neck hyperextension, strabismus and, often, fatal apnea1. We selected a mouse model of nGD carrying a loxP-flanked neomycin disruption of Gba plus Cre recombinase regulated by the keratinocyte-specific K14 promoter. Exclusive skin expression of Gba prevents fatal neonatal dehydration. Instead, mice develop fatal neurodegeneration within 15 days2. Using this model, fetal intracranial injection of adeno-associated virus (AAV) vector reconstituted neuronal glucocerebrosidase expression. Mice lived for up to at least 18 weeks, were fertile and fully mobile. Neurodegeneration was abolished and neuroinflammation ameliorated. Neonatal intervention also rescued mice but less effectively. As the next step to clinical translation, we also demonstrated the feasibility of ultrasound-guided global AAV gene transfer to fetal macaque brains.
Antibodies against activated microglia (CD68) and astrocytes (glial fibrillary acid protein, GFAP) were used to assess spatiotemporal neuroinflammation and gliosis in the nGD mouse and this was quantified by thresholding image analysis. Unlike wild types (WTs) and heterozygotes, brains of symptomatic post-gestational day (P)12 knockouts exhibited significant and widespread microglial activation (Supplementary Fig. 1c) particularly in the Gi of the brainstem, the VPM/VPL and layer V of the S1BF. Blinded semi-quantitative scoring revealed that no region was spared (Supplementary Table 1). This was confirmed by immunostaining for a second microglial marker, Iba1 (Supplementary Fig. 1f). Significant astrogliosis (Supplementary Fig. 2c) was confirmed by both thresholding analysis and blinded scoring (Supplementary Table 2). At presymptomatic age P8, knockout mouse brains showed significant microglial (Supplementary Fig. 1b) and astrocyte (Supplementary Fig. 2b) activation in the thalamus and brainstem. Blinded scoring (Supplementary Tables 1 and 2) and Iba1 immunohistochemistry (Supplementary Fig. 1e) provided confirmation. Remarkably, even in P1 neonates there was significant activation of astrocytes (Fig. 1a–c) and microglia (Fig. 1d–f) in the brainstem.
Neuronal loss was quantified by stereological assessment of Nissl-stained coronal sections. There was significant neuronal loss in the cortex, thalamus and brainstem of knockouts at P12 (Supplementary Fig. 3b), accompanied by acellular spaces consistent with either large vacuole formation or lipid accumulation as determined by hematoxylin and eosin staining (Supplementary Fig. 3c). In P8 knockouts, there was significant loss of cortical and midbrain neurons (Supplementary Fig. 3a). Neuronal loss correlated with microglial and astrocyte activation and, consistent with findings in a milder nGD model3, there was significant cortical thinning in S1BF (Supplementary Fig. 3d); however, no difference in volume was observed (Supplementary Fig. 3e).
In the absence of functional glucocerebrosidase (GCase), its substrate, glucosylceramide (d-glucosyl-β1-1′-N-acyl-d-erythro-sphingosine), and related sphingolipid species accumulate in both humans and mouse models4. Total glucosylceramide and glucosylsphingosine are elevated >20-fold at birth in the nGD mouse5, but no detailed measurements of different glucosylceramide species and proximal metabolites have been made. We assembled a temporal profile in whole-brain homogenates using liquid chromatography and tandem mass spectrometry. In knockouts, all glucosylceramide species except for C24:0-OH and C24:1-OH were significantly elevated at P1, C18:0 and C16:0 being most abundant (Fig. 1g and Supplementary Fig. 4a); glucosylpsychosine was increased the most (Supplementary Table 3). At P8 and P12, these remained the most elevated in both absolute (Supplementary Fig. 4b,c) and relative (Supplementary Table 3) terms. As expected, the gangliosides GM1a, GD1a, GD1b and GTb remained unchanged in knockouts irrespective of their age (Supplementary Fig. 4d–f).
The upregulation of neuroinflammatory markers and glycosphingolipids in P1 knockouts suggested that prophylactic in utero gene therapy might be beneficial. We selected adeno-associated virus serotype 9 (AAV9) as we6, and others7, have observed widespread neuronal expression of green fluorescent protein (GFP), following fetal intracranial injection. We examined the expression profile after fetal intracranial injection of an AAV9 vector encoding GFP driven by a β-glucuronidase (GUSB) promoter8 (Supplementary Fig. 5a). At 16 days of gestation, fetuses received 5 µl of AAV9 vector (5 × 1010 genome copies) into the lateral ventricle. At 30 days of age, strong bilateral fluorescence extended from the olfactory bulbs to the brainstem (Supplementary Fig. 5b). Bilateral homogeneity of expression and extensive spread from prefrontal cortex to cerebellum and brainstem was confirmed by immunohistochemistry (Supplementary Fig. 5c). Many GFP-expressing cells with neuronal morphology were observed in the cortex, striatum, hippocampus, thalamus, cerebellum and, importantly, brainstem (Supplementary Fig. 5d).
Using the same vector configuration, mode and timing of injection, we delivered human AAV9-GBA9 into pups of Gba heterozygous parents. While impossible to genotype in utero, we identified three knockouts after birth by assay of GCase activity in blood spots obtained by superficial temporal vein puncture. They exhibited neither paralysis nor dyskinesia, remaining indistinguishable from WT littermates until being euthanized at 35 days of age. Treated knockouts had substantially diminished microglial and astrocyte activation versus untreated P12 knockouts in the prefrontal cortex, striatum, hippocampus and cerebellum (Fig. 2a,b and Supplementary Fig. 6a,b) and there was a tendency for greater treatment effects in Gi than S1BF (Supplementary Fig. 6d,e).
Lysosomal-associated membrane protein 1 (LAMP-1) is increased in fibroblasts of a patient with nGD10, and was, therefore, selected as a more direct marker of lysosomal compartment size. P12 knockout brains showed strong LAMP-1 expression compared with P35 WT mice. In contrast, staining was similar in treated knockouts and WT mice, with the exception of a few, sparse darkly stained LAMP-1-positive cells in treated knockouts (Fig. 2c and Supplementary Fig. 6c).
Gene therapy prevented neuronal loss in knockouts in the thalamus and Gi (Fig. 2d). Staining with two anti-GCase antibodies indicated that normal enzyme expression had been restored, yet each provided different insights. A carboxy-terminus-specific antibody showed that WT and AAV9-treated knockouts exhibited discrete GCase immunoreactivity within neurons, which was completely absent in untreated knockouts. In contrast, an amino-terminal antibody revealed strong staining of glia rather than neurons in all cortical layers, which was absent in WT brains. After AAV9 treatment of knockouts, staining was observed in neurons in the deeper cortical layers but not in glia (Fig. 2e).
To test long-term efficacy, we injected five knockouts (two males and three females, three separate litters). Consistent with the description of the model2, 7 untreated control knockouts developed forelimb paralysis and then tetraparesis; 5 and 2 mice were euthanized at humane endpoints at 14 and 15 days of age. In contrast, at 70 days of age, all treated knockouts appeared normal and fertile, and 2 were able to produce a litter of 2 symptomatic knockouts (Supplementary Video 1). One treated mouse was euthanized at 125 days after exhibiting stereotypic circling behavior and another at 131 days because of an untreatable eye infection; at which time the remainder were then euthanized (Fig. 2f). GCase activity in brain homogenates of treated mice was not significantly different from WTs (Fig. 2g). Paradoxically, other parameters indicated only partial correction of the disease phenotype. Treated knockouts tended to perform worse on Rotarod (Fig. 2h) and grid walk (Fig. 2i) tests. At 100 days, they weighed significantly less than unaffected littermates (Fig. 2j) and appeared hyperkinetic (Supplementary Video 2). They exhibited significant elevation of the shorter chain glycosphingolipids, glucosylpsychosine and lyso-lactosyl ceramide (Fig. 2k) and GD1a (Supplementary Fig. 6g) but not glucosylceramide (Supplementary Fig. 6h). Microglial activation (Supplementary Fig. 6i) and astrogliosis (Supplementary Fig. 6j) was ameliorated but not normalized and four of the five mice exhibited ventriculomegaly.
To test the efficacy of gene therapy later in development, newborn (P0–P1) knockouts received intracerebroventricular injection of 5 μl AAV9 vector (5.0 × 1010 genome copies). Previous studies have demonstrated the ability of intravenous AAV9 to cross the blood–brain barrier of neonatal mice11,12 and non-human primates11 resulting in efficient expression in neural cells and viscera13. Therefore, we also tested whether an intravenous administration of the AAV9 vector (4 × 1011 genome copies) to knockouts could ameliorate both visceral and brain pathology. Both treated cohorts were rescued from neonatal death, exhibited neither paralysis nor dyskinesia and remained indistinguishable from WTs until being euthanized at 55 days of age (Fig. 3a). There were no significant differences in weight (Fig. 3b) or motor coordination (Fig. 3c) between treated knockouts and WTs.
There was significant supraphysiological expression of GCase in the cortical S1BF region, CENT2 of the cerebelleum and Gi, and physiological levels in the hippocampal CA1–CA2 and thalamic VPM/VPL regions (Fig. 3d). There was no significant increase in astrogliosis in the CA1–CA2, VPM/VPL and Gi regions of intravenously injected knockouts at 55 days (Fig. 3e); however, there was a significant increase in the S1BF, VPM/VPL and CENT2 regions. There was no significant increase in microglial activation between the treated mice in any region except S1BF (Fig. 3f). Furthermore, the untreated 14-day-old knockout nGD mice showed a significantly higher amount of staining compared to the older intravenously treated mice in the VPM/VPL and Gi regions. There was no significant increase in the lysosomal marker LAMP-1 in any region except Gi (Fig. 3g). There was no significant loss of neurons in the Gi of knockout mice treated with intravenous gene therapy (Fig. 3h). However, there was a significant loss in the cortical S1BF and thalamic VPM/VPL regions. The cortex was significantly thinner in intravenously injected knockouts (Fig. 3h). Despite overall strong GCase staining in the S1BF region and the cerebral cortex neuroinflammation, neuronal loss and subsequent cortical thinning were still evident.
After both intracerebroventricular and intravenous injection, GCase was significantly elevated in visceral organs (Fig. 3i), suggesting vector escape from the brain after intracerebroventricular administration. Intravenous gene therapy normalized enzyme activity in all visceral organs except the heart, where activity was supraphysiological. Intravenous gene therapy prevented splenomegaly (Fig. 3j) and Gaucher cell infiltration in spleen, liver and lungs. In contrast, intracerebroventricular injection failed to prevent Gaucher cell infiltration in spleen (Fig. 3k), liver (Fig. 3l) and lungs (Fig. 3m), as confirmed by immunohistochemistry. Therefore, while intracerebroventricular gene therapy is effective in treating the lethal neuropathology, it does not ameliorate the visceral manifestations. However, intravenous administration effectively limits both neurological and visceral pathology.
Separate groups were assessed for long-term efficacy. All treated mice survived to at least 180 days and were indistinguishable from age-matched WTs (Fig. 3n). Spleens from intracerebroventricular injection cohorts were significantly heavier than those receiving intravenous injection (Fig. 3p). Large Gaucher cells were present in the spleens following intracerebroventricular, but not intravenous injection (Fig. 3o).
We next asked whether in utero gene delivery could achieve widespread expression in the larger, gyrencephalic primate brain by performing ultrasound-guided delivery of AAV9 vectors to mid-gestation fetal macaques. Cerebrospinal fluid flows from the lateral ventricles through the ventricular system to the central canal of the spinal cord and the subarachnoid cisterns. Therefore, to achieve maximum vector spread, we identified the lateral cerebral ventricles of two day-58 (0.4G) macaque fetuses by transabdominal ultrasound. Vector was injected into the right lateral ventricle under ultrasound guidance. scAAV9-GUSB-GFP was administered at a concentration of 6.6 × 1012 vector genomes (vg) ml−1 over 60 s. Correct delivery was confirmed by observation of transient (10 min) ventricular swelling (Fig. 4a). Both macaques presented normally at birth. There was extensive and bilateral rostro-caudal expression of GFP (Fig. 4b) with numerous GFP-positive cells of neuronal morphology in the superior frontal and middle temporal gyri, thalamus, hippocampus, uvula and Gi (Fig. 4c). Most other brain regions contained immunopositive cells except for, particularly, the caudate nucleus and putamen (Supplementary Fig. 7). No discrete cells were observed in unstained brain (Figs. 4b,c and Supplementary Fig. 7).
The sustained, widespread gene expression from a single vector injection is supported by a recent study in non-human primate studies and clinical trials. At least 15 years of AAV-mediated transgene expression has been demonstrated in a non-human primate model of Parkinson’s disease14. Proof-of-concept has already been provided for single-dose sustained long-term AAV expression in liver-directed gene therapy for the treatment of hemophiliac patients15.
Widespread gene delivery to the central nervous system by fetal intracranial injection offers the best therapeutic potential for inherited early onset neurodegenerative diseases affecting the whole brain. Delivery to the developing central nervous system may be especially important since neurons have limited regenerative capacity. This is evidenced by the fact that neonatal vector delivery was less effective than fetal intervention in averting neuronal loss. Of several studies on in utero gene therapy in gene-disrupted mice16, none involved models where pathology is already evident at birth and fetal intervention is the only tractable approach. Recent gene therapy clinical trials for mucopolysaccharidosis type IIIa and spinal muscular atrophy17 indicate that the youngest patients derived most therapeutic benefit18.
Normalization of GCase activity in the brain did not completely abolish pathology. Knockouts receiving intracerebroventricular injection developed visceral disease. Hypermetabolism, previously identified in patients with Gaucher19, may account for their lower body mass. However, this would not explain the failure to normalize brain glycosphingolipids, nor prevent long-term microglial and astrocyte activation and ventriculomegaly. AAV gene delivery follows a Poisson distribution20. Therefore, individual neurons may receive variable gene dosage, with some remaining un-transduced, as evidenced by heterogeneous staining in treated knockout brains (Fig. 2e).
Human fetal gene therapy necessitates accurate prenatal diagnosis. From cell-free fetal DNA in the maternal circulation, founder mutations underlying Gaucher disease can be detected non-invasively21. Concerns over weak genotype–phenotype correlation can be mitigated: several lysosomal storage diseases, including Gaucher disease, have been diagnosed by enzyme assay of chorionic villi samples22. Currently, prenatal screening is usually restricted to a family with a history of Gaucher disease or parents of an affected child, although advances in non-invasive prenatal diagnosis may help identify most cases.
Perinatal gene transfer has been shown to reduce or avoid an immune response to either a protein (absent in an animal model and potentially also in some patients), or to a viral capsid protein23,24. Immune tolerance may permit vector re-administration. The presence of GCase in the skin of the knockout nGD mice precluded us from investigating this specifically. Nevertheless, immune tolerance towards a non-mammalian protein following fetal or neonatal gene delivery has been observed in non-human primate studies25 and in mice this allowed effective re-administration of vector26. Avoidance of an immune response following gene therapy is important. Non-human primates seropositive for AAV9 are less efficiently transduced using an AAV9 vector27. Prednisolone was used in a recent AAV9 gene therapy clinical trial for spinal muscular atrophy to manage anti-vector immune responses17. While this effectively managed the immune response, in utero administration of vector may pre-empt exposure to WT AAV and thus development of anti-capsid immunity.
This study highlights the potential of fetal gene therapy for neonatal lethal neurodegenerative diseases; the non-human primate data offer a clinically feasible protocol for its implementation. This, and other preclinical fetal gene therapy studies, address the suggested requirements of the NIH recombinant DNA advisory committee28, namely safety and efficacy in relevant models for progression to clinical trials.
Mouse procedures and welfare were approved by the University College London Animal Welfare and Ethical Review Board (AWERB) and in accordance with project and personal licenses granted by the UK Home Office and the Animal (Scientific Procedures) Act of 1986. The mouse model for acute neuronopathic Gaucher disease used in this study has been previously described together with all genotyping procedures and primer sequences2. Mice heterozygous for mutation in the Gba gene and carrying the Cre recombinase gene on a C57BL/6 background were outbred onto WT CD1 mice (Charles River) for five generations. The CD1 strain allows for larger litter size and excellent maternal care of pups. The colony was then maintained using breeding pairs heterozygous for the Gba mutation and carrying the Cre-recombinase gene. Macaque procedures and welfare were approved by and performed in accordance with the Institutional Animal Care and Use Committee (IACUC) at the National University of Singapore and Singapore Health Services Pte., Ltd (IACUC 2009-SHS-512). All in vivo studies performed in macaques were conducted at the SingHealth Experimental Medicine Centre that has been accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC). Potential dams were screened for pre-existing anti-AAV9 antibodies and seronegative females were time-mated. Pregnancies were confirmed, dated and monitored by ultrasound before and after vector delivery.
Histological and immunohistochemical analyses
Experimental cohorts of mice were euthanized by terminal transcardial perfusion using phosphate-buffered saline (PBS), with the exception of P1 mice euthanized by decapitation. Macaque brains and spinal cords were collected following euthanasia using isoflurane and pentobarbitone, cardiac puncture and perfusion of 1% paraformaldehyde.
For free-floating immunohistochemistry-based analyses, the brains were removed and placed into 4% paraformaldehyde solution and stored for 48 h at 4 °C before being cryoprotected in 30% sucrose for a minimum of 24 h at 4 °C. The brains were then serial sectioned along the rostro-caudal axis at 40-μm-thick sections using a freezing microtome (Thermo Fisher Scientific). Immunohistochemical staining was used to detect gene expression. Endogenous peroxidase activity was quenched by incubating sections in 1% H202 in Tris-buffered saline (TBS) solution for 30 min. Following three washes in TBS, sections were incubated for 30 min in a solution of 15% normal serum in TBS-T (0.3% Triton X-100 in TBS). Subsequently, sections were incubated at 4 °C overnight in primary antibody in TBS-T/10% serum. After rinsing in TBS, the sections were incubated for 2 h in secondary IgG diluted in TBS-T/10% serum. Sections were then rinsed in TBS and incubated for 2 h in a 1:1,000 solution of Vectastain avidin–biotin solution (ABC; Vector Labs). The sections were rinsed in TBS and incubated in 0.05% solution of 3,3′-diaminobenzidine (DAB) containing 0.01% H202. The staining reaction was stopped by adding ice-cold TBS. The sections were mounted onto chrome-gelatine-coated Superfrost-plus slides (VWR) and left to air-dry overnight. The sections were dehydrated in 70% ethanol, and cleared in xylene for 30 min before being coverslipped using DPX mounting medium. This protocol was used with the following combinations of primary antibodies, blocking serum and secondary antibodies: rat anti-mouse CD68 (1:100; AbD Serotec, MCA1957)12, rabbit serum and rabbit anti-rat (1:1,000; Vector Laboratories Inc.); mouse anti-GFAP (1:500; Millipore, MAB3402), goat serum and goat anti-mouse (1:1,000; Vector Laboratories Inc.); rabbit anti-LAMP-1 (1:200; Abcam, AB24170), goat serum and goat anti-rabbit (1:1,000; Vector Laboratories Inc.); mouse anti-GBA (N terminus) (1:150; Origene, TA803361), goat serum and goat anti-mouse (1:1,000; Vector Laboratories Inc.); rabbit anti-GBA (C terminus) (1:1,000; Sigma-Aldrich, G4171), goat serum and goat anti-rabbit (1:1,000; Vector Laboratories Inc.); rabbit anti-GFP (1:10,000; Abcam, AB290), goat serum and goat anti-rabbit (1:1,000; Vector Laboratories Inc.), respectively. Brain sections were Nissl-stained using cresyl violet. The sections were incubated in 0.05% cresyl violet in 0.05% acetic acid for 30 min at 60 °C and then washed in deionized water. The sections were differentiated through ascending concentrations of alcohol before being cleared in xylene and coverslipped with DPX.
For histology and immunohistochemical staining on paraffin-embedded tissue sections, mice were euthanized by transcardial perfusion as described above. Brain fixed in 10% buffered formol-saline was embedded in paraffin wax. Microglial activation was examined using an ionized calcium-binding adaptor molecule 1 (Iba1) antibody (Wako Pure Chemical Industries, 019-19741). Immunohistochemistry was performed on automated immunohistochemistry staining machines (Ventana Medical Systems Inc.) using proprietary secondary detection reagents (Ventana Medical Systems Inc.) before development with 3′3-diaminobenzedine tetrachloride as the chromogen. Conventional methods were used for Harris hematoxylin and eosin staining. Sections were mounted on chrome-gelatine-coated slides and air-dried overnight. The sections were stained, protected from light, with filtered 0.1% Mayer’s hematoxylin (Sigma-Aldrich) for 10 min. The slides were rinsed in distilled water for 5 min and consequently dipped 12 times in 0.5% eosin solution. The sections were quickly washed in distilled water and subsequently dehydrated for 30 s in rising concentrations of ethanol (50%, 70%, 95%, 100%). The slides were finally incubated in Histo-clear for 30 min and coverslipped with DPX mountant medium.
Quantitative analysis of histological and immunohistochemical staining
The quantification of immunohistochemical staining within discrete regions of the brain was conducted by a user blinded to the experimental cohorts. Forty non-overlapping RGB images were taken at ×40 magnification through four consecutive sections within each region of interest. During image capture, all camera and microscope settings were maintained. Regional areas of immunoreactivity were analysed by optical segmentation using Image Pro-Plus (Media Cybernetics). The foreground immunostaining was defined by averaging of the highest and lowest signals and the threshold was set and remained unchanged as each consecutive image was analysed. The data are presented as the mean percentage area of immunoreactivity per field for each discrete region. Stereological counts of neurons by optical fractionation and measurement of cortical thickness and volume were performed using Stereo Investigator (MBF Biosciences)29. The grid size and the counting frame used for analysing different brain regions were as follows: S1BF 150 × 150 μm, 50 × 50 μm; VPM/VPL 175 × 175 μm, 50 × 50 μm; Gi 100 × 100 μm, 50 × 50 μm. Cells were counted using a 100× objective. Efficient sampling was estimated by a coefficient of error between 0.05 and 0.130. Four sections for each brain were analysed and the average values of cell counting were used in the calculations.
The mean thickness of the S1BF cortical region was estimated by using the Cavalieri vertical sections principle31. The length of ten parallel consecutive lines intersecting perpendicularly the cerebral cortex, traced from the somatosensory barrel cortical layer 1 to the corpus callosum, was measured. Three sections of the midbrain region per each brain were analysed and the average values were reported.
Mouse brains were homogenized in water using an Ultraturax T25 probe homogenizer (IKA). Protein concentrations were determined using bicinchoninic acid (BCA) assay and 200 µg was analyzed for glucosylceramide, glucosylpsychosine and lyso-lactosyl ceramide accumulation. The substrates were extracted in 1 ml of chloroform/methanol (2:1 v/v) containing 80 µg ml−1 of d4-C16:0-glucosylceramide internal standard that was synthesized in house32. The samples were shaken for 30 min at room temperature before the addition of 200 µl of PBS for phase separation. After a 10 min centrifugation at 16,000g, the upper and lower phases were collected and 5 µl of each was injected into the ultraperformance liquid chromatography–tandem mass spectrometry system. Glycosphingolipid reference standards (Matreya) were also analyzed to confirm analyte identity.
The samples were injected onto a Waters ACQUITY UPLC system operated in partial loop mode and separated on a Waters ACQUITY UPLC BEH C18 column (130 Å, 1.7 µm, 2.1 mm × 50 mm) under the following gradient conditions: 0.00–0.20 → 80% A; 0.20–5.00 → 0.1% A; 5.00–9.00 → 0.1% A; 9.01–11.00 → 80% A, where mobile phase A was double-distilled H2O (ddH2O) with 0.1% FA; phase B was methanol and the flow rate was 0.65 ml min−1. Column and sample temperatures were kept at 40 °C and 10 °C, respectively. Weak wash solvent was ddH2O [0.1% FA] and strong wash solvent was acetonitrile/methanol/isopropanol/ddH2O (1:1:1:1 v/v). The eluting analytes were detected on a Waters XEVO TQ-S triple quadrupole mass spectrometer that was equipped with the electrospray ion source and operated in multiple reaction monitoring and positive ion mode (see Supplementary Table 4 for multiple reaction monitoring details) with the tune page parameters set to achieve the maximum sensitivity for glycosphingolipids as described previously33. The data were processed with MassLynx v4.1.
Glycosphingolipids (GSLs) from brain homogenates were extracted with chloroform/methanol (1:2, v/v) overnight at 4 °C and further purified using solid-phase C18 columns (Telos, Kinesis). GSLs were dried down under nitrogen and treated with either Cerezyme (Genzyme) to obtain glucose from GlcCer, or ceramide glycanase (prepared in house from the medicinal leech Hirudo medicinalis/verbena) to obtain oligosaccharides from other GSLs. Liberated glucose and free glycans were then fluorescently labeled with anthranillic acid (2AA). Excess 2AA label was removed using DPA-6S SPE columns (Supelco). Purified 2AA-labeled glucose and 2AA-labeled glycans were separated and quantified by normal-phase high-performance liquid chromatography34. The normal-phase high-performance liquid chromatography system consisted of a Waters Alliance 2695 separations module and an in-line Waters 2475 multi λ-fluorescence detector set at excitation λ360 nm and emission λ425 nm. The solid phase used was a 4.6 × 250 mm TSK gel-Amide 80 column (Anachem). Results were normalized to protein content determined using BCA assay.
Self-complementary AAV vectors
The human GBA cDNA was generated by polymerase chain reaction, verified by DNA sequencing and cloned into the Topo plasmid vector (Invitrogen). A 1.6-kb fragment (containing the open reading frame of the human GBA cDNA and bovine growth hormone polyadenylation signal sequence) was then generated by digesting the plasmid with the restriction enzymes NcoI and SpeI and subcloned into a shuttle plasmid containing the scAAV2 inverted terminal repeats and the 0.4 kb GUSB promoter35 (Supplementary Fig. 8). Recombinant AAV9 serotype vectors encoding GBA (scAAV9-GUSB-GBA) were generated by the standard triple plasmid transfection method as described previously36. Briefly, cell lysates of transfected 293 cells were clarified by centrifugation and then purified by subjecting the preparations to two cycles of density gradient ultracentrifugation using cesium chloride. The purified recombinant vectors were resuspended and stored in 10 mM sodium phosphate buffer, pH 7.3, containing 180 mM sodium chloride and 0.001% pluronic F68.
AAV administration to mice and macaques
For marker gene studies, we injected CD1 outbred mice. nGD mice were backcrossed onto the CD1 background, to maximize litter sizes and minimize the risk of cannibalization.
We have previously described the administration of gene delivery vectors to the brains of fetal mice37 and intracerebroventricular injections to neonates has been described in an earlier study38. We have previously described intravenous administration to neonatal mice12. Briefly, pregnant mice carrying pups at 16 days of gestation were anesthetized using isoflurane inhalation anesthesia and a midline laparotomy was performed to expose the uterus. Five microliters of AAV vector (5 × 1010 genome copies) was administered to all fetuses per dam via a transuterine injection targeting the anterior horn of the lateral ventricle of the left hemisphere of the brain. The laparotomy was sutured and mice were provided topical and systemic analgesia and allowed to recover in a warm chamber. The same dose was administered via intracerebroventricular injection to P0 neonates. The intracerebroventricular injections were directed to the anterior horn of the lateral ventricle. The injection site was identified at 2/5 of the distance from the lambda suture to each eye. P0–P1 mice were anesthetized on ice for 30–60 s. A 33-gauge needle (Hamilton) was inserted perpendicularly at the injection site to a depth of 3 mm and 5 μl of vector was slowly administered. The pup was allowed to recover and placed back into the cage. Forty microliters of AAV vector (4 × 1011 genome copies) was administered to P0 neonates by injection of the superficial temporal vein. At P0–P1, pups were anesthetized on ice for 30–60 s and intravenous injections were performed via the superficial temporal vein using a 33-gauge needle (Hamilton). Once the needle was slowly removed, gentle pressure was applied to the injection site. When the pup fully recovered, it was returned to the dam.
Cynomolgus macaques (Macaca fasicularis) were used for non-human primate studies. Macaques underwent time-mating and pregnancy was confirmed through ultrasound evidence of fetal pole and fetal heart activity. Dating of pregnancy was performed as previously reported39, and IUGT was performed at 0.4G. After induction of anesthesia via sevoflurane, a 25-G Quincke needle (Becton-Dickenson) was used to target the lateral ventricle closest to the anterior maternal abdomen under continuous ultrasound imaging. Vector was injected as a slow bolus (200–350 μl to deliver a total dose of 1.3 × 1012 vg and 2.3 × 1012 vg in each fetus respectively) with ventricular swelling observed as evidence of correct delivery. The needle was removed immediately and the fetus monitored for a further 15 min. A male fetus was delivered naturally and a female fetus was delivered by cesarean section and hand-reared as previously reported. Euthanasia and organ collection were performed on days 0 and 6 respectively under ketamine and isoflurane, via cardiac puncture and exsanguination as previously described40.
Mouse behavioral studies
Mice underwent foot-fault testing (adapted from ref. 41) and Rotarod testing (adapted from ref. 42) at P35, P73 and P100. The operator was blinded to the genotype of the animals throughout the testing period. Briefly, foot-fault testing involved placing a mouse on a stainless-steel grid (41 × 25 cm) with a mesh size of 1 × 1 cm, elevated 30 cm above the bench surface. The number of times a limb fell through the grid (a foot-fault) within 30 s was recorded. Rotarod testing employed a commercially available Rotarod (Harvard Apparatus). Mice were placed on the rotating drum under continuous acceleration (from 4–40 r.p.m. over 5 min), with the latency to fall from the rod, and the speed at which this occurred, measured. Any passive rotations were also noted.
Stereoscopic fluorescence microscopy
GFP expression in various organs was visualized with a stereoscopic fluorescence microscope (MZ16F; Leica). Representative images were captured with a digital microscope camera (DFC420 Leica) and image analysis software (LAS 4.2; Leica). As a result of the differences in the sizes of the organs, intensity of signal and distribution of cells of interest, the images captured were optimized for exposure length and brightness in each case.
GCase enzyme assay
GCase activity was determined using the established synthetic substrate, 4-methylumbelliferone-β-glucopyranoside protocol as previously described43. Frozen tissue samples were homogenized with distilled water on ice and the total protein concentration was measured using BCA assay. Samples were incubated with the substrate for 2 h at 37 °C. The reaction was stopped with 1 M glycine buffer, pH 10.4. Fluorescence of the standard 1 nM 4-methylumbelliferone and the samples was read (FluoStar Optima Plate Reader; excitation wavelength: 360 nm, emission wavelength: 450 nm). The enzymatic activity (nmol h−1 μg−1) was calculated.
Sample sizes, statistics and randomization
Since the genotypes of injected mice could not be ascertained until after birth, the initial sample size was determined by the number of knockout pups in the uteri of three randomly selected pregnant dams. For most analyses, one-way or two-way ANOVA was performed with either Bonferroni’s or Tukey’s multiple comparison. The F statistic of ANOVA is tolerant of violations in normality, for which we did not test. Nevertheless, mass spectrometry data, which were clearly not normally distributed, were converted by log transformation before ANOVA. Each experiment was performed only once. Behavioral and semi-quantitative scoring were blinded to the investigator.
Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.
All data generated or analyzed during this study are included in this published article (and its Supplementary Information). Raw data are available online at https://doi.org/10.5281/zenodo.1246085. These data are also available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
S.N.W., A.A.R. and J.D.C. received funding from UK Medical Research Council grant G1000709. S.N.W. received funding from MR/N019075/1 and MR/P026494/1 and SPARKS (17UCL01). A.A.R. and S.N.W. received funding from MRC grants MR/R015325/1 and MR/N026101/1 and NC3Rs grant NC/L001780/1. G.M., A.A.R. and S.N.W. received funding from the UK Gauchers Association. A.A.R. received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 666918 (BATCure), Action Medical Research (GN2485), MRC grant MR/M00676X/1, Asociación Niemann Pick de Fuenlabrada, Niemann-Pick UK, Niemann Pick Research Foundation and the NBIA Disorders Association. S.M.K.B. and S.N.W. received funding from ERC grant ‘Somabio’ 260862. C.N.Z.M. received salary support from the Singapore Ministry of Health’s National Medical Research Council NMRC/TA/0003/2012 and NMRC/CSA-INV/0012/2016. M.H. is supported by Parkinson’s UK grant H-1501. F.M.P. is a Royal Society Wolfson Research Merit Award holder and a Wellcome Trust Investigator in Science. F.M.P. and D.A.P. were supported by the Mizutani Foundation for Glycoscience. J.K.Y.C. is funded by Singapore’s Ministry of Health’s National Medical Research Council NMRC CSA/043/2012, CSA(SI)/008/2016 and CIRG/1459/2016. J.D.C. received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 666918 (BATCure), the Batten Disease Support and Research Association (US) and Batten Disease Family Association (UK). K.M. received funding from the Peto Foundation. S.B. was supported by the National Institute of Health Research (NIHR) UCLH/UCL Biomedical Research Centre. We thank R. Baker, M. Choolani, N. Johana, N. Wen, B. Warburton, S. Richards, T. O’Mahoney, G. Sturges O. Woolmer, E.-H. Davies, T. Collin-Histed, A. Mehta and D. Hughes. Most of all, for guidance, mentorship and inspiration, we thank C. Coutelle.