Chemokine-induced recruitment of genetically modified bone marrow cells into the CNS of G_M1-gangliosidosis mice corrects neuronal pathology

Chemokines comprise a family of about 50 small protein ligands that together with their cognate receptors control leukocyte trafficking during immune surveillance and inflammatory-cell recruitment during host defense.^1-3  Besides their known function in the immune system, these molecules have also been implicated in the maintenance of central nervous system (CNS) homeostasis and as a mediator of neuroinflammation.^4-8  Two types of interactions primarily control the activity of chemokines in the CNS. The first involves chemokine immobilization by glycosaminoglycans (GAGs), which play a role in the formation of chemokine gradients and trigger localization of leukocyte subpopulations to the site of infection or injury, present at the endothelial surface and the extracellular matrix.^9,10  The other interaction entails the tight binding of chemokines to their G-protein-coupled receptors on the surface of leukocytes; this binding activates integrins and promotes their adhesion to the endothelium, followed by their penetration across the endothelial layer into the CNS perivascular space.^11-15  In the CNS, chemokines are expressed constitutively at low or negligible levels in astrocytes and microglia, but their expression is rapidly induced by neuroinflammatory conditions^16-18  and is mediated by several proinflammatory cytokines (eg, tumor necrosis factor α [TNF-α], interferon γ [IFN-γ], and tumor growth factor β1 [TGF-β1]).^19-22  CNS inflammation occurs in several neurodegenerative conditions, including those associated with a lysosomal storage disease (LSD),^23,24  and is likely responsible for the recruitment of monocytes and macrophages from peripheral blood. In the murine model of the LSD, G_M2-gangliosidosis recruitment of macrophages into the CNS is mediated by the chemokine macrophage inflammatory protein 1-α (MIP-1α) and has been implicated in the pathogenesis of this disease.²⁵ 

Genetic defects that alter β-galactosidase (β-gal) activity in humans have a devastating effect on the CNS and result in the neurodegenerative LSD G_M1-gangliosidosis.²⁶  The clinical phenotypes of this disease demonstrate a continuum of severity: (1) Infantile G_M1-gangliosidosis progresses rapidly; developmental arrest occurs soon after birth; and symptoms include profound neurologic deterioration, psychomotor retardation, visceromegaly, and cardiac involvement. Death due to infection or cardiac failure occurs normally within the second year of life. (2) Late-infantile/juvenile G_M1-gangliosidosis includes little or no systemic organ involvement, but neurologic problems, ataxia, and seizures develop generally soon after the onset of the symptoms. (3) Chronic G_M1-gangliosidosis manifests as slow, progressive dementia, Parkinsonian features, and extrapyramidal signs. At the cellular level, G_M1-gangliosidosis affects primarily neurons, which become distended and fill with membranous cytoplasmic bodies, but astrocytes and microglia also appear vacuolated. Abnormal accumulation of G_M1-ganglioside (G_M1) and to a lesser extent its asialo-derivative, G_A1, in neurons is the most prominent biochemical feature of this disease. As with other LSDs, there is no effective cure and only palliative treatment is available.

The mouse model of G_M1-gangliosidosis recapitulates the early-onset forms of the disease²⁷  (ie, G_M1 and G_A1 progressively accumulate in nearly every neuron). Tremor, ataxia, abnormal gait, and gradual deterioration of motor function culminate in the paralysis of the hind limbs.²⁷  Using this model, we recently found that G_M1 accumulation in the endoplasmic reticulum (ER) of brain cells and depletion of ER calcium stores stimulate an unfolded protein response (UPR), which in turn, causes neuronal apoptosis.²⁸  The latter process elicits a neuroinflammatory response associated with the activation of inflammatory markers and cytokines, which in turn activate microglia and macrophages at the site of apoptosis.²⁹ 

Here we demonstrate that neurodegeneration and neuroinflammation in G_M1-gangliosidosis mice cause the up-regulation of selected chemokines in specific brain regions. This phenomenon promotes the migration/infiltration of bone marrow cells (BMCs) genetically modified to overexpress β-gal into the CNS, which in turn corrects the neuropathology associated with G_M1-gangliosidosis.

FVB Glb1^+/+ mice (2-6 months) were used as BMC donors, and Glb1^-/- mice (3-4 weeks) of the same background were used as recipients.²⁷  All procedures were done in accordance with the US Public Health Service Policy on the Human Care and Use of Laboratory Animals.

Stromal-cell-derived factor 1α (SDF-1α) and SDF-1β expression in the mouse was determined by reverse transcription of RNA samples by real-time polymerase chain reaction (PCR). TaqMan Pre-developed Assay Reagent Kits (Applied Biosystems, Foster City, CA) were used to analyze total RNA from cerebellum. Oligotex mRNA Purification Kit (Qiagen, Valencia, CA) was used to analyze polyadenylated (polyA⁺) RNA from spinal cord. Values were normalized against glyceraldehyde phosphate dehydrogenase (GAPDH) expression, which was amplified simultaneously.

SDF-1α protein level was measured by enzyme-linked immunosorbent assay (ELISA) on cerebellar homogenates. Tissue samples were homogenized in antiprotease buffer containing 1 × phosphate-buffered saline (PBS) with Protease Inhibitor Cocktail Tablets (Boehringer Mannheim, Indianapolis, IN). Debris was removed by centrifugation and the aqueous extract stored at -70°C until the time of analyses. Total protein concentration was determined using a bicinchoninic acid reagent (Pierce Chemical, Rockford, IL), and the level of SDF-1α protein was detected using a sandwich-type immunoassay kit (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions.

RNA from midbrain, brain stem, and cerebellum were purified and analyzed for β-chemokine expression via RNAse protection assay (BD Biosciences, San Jose, CA) according to the manufacturer's instructions. We used a multiprobe DNA template for the following chemokines: lymphotactin/CXC chemokine ligand 1 (Ltn/CXCL1), regulated upon activation, normal T cells, expressed and secreted (RANTES)/CC chemokine ligand 5 (CCL5), eotaxin/CCL11, MIP-1α/CCL3, MIP-1β/CCL4, MIP-2 CXC ligand 2 (CXCL2), IFN-inducible 10-kDa protein/CXCL10 (IP-10), monocyte chemoattractant protein-1/CCL2 (MCP-1), and T-cell activation antigen (Ag)/CCL1 (TCA-3). L32 and GAPDH served as housekeeping genes. RNAse protection assay (RPA) analysis was performed on 20 μg total RNA and hybridized with probes labeled with [α-³²P] uridine triphosphate (UTP). After digestion of ssRNA, the RNA pellet was solubilized and resolved on a 7.5% sequencing gel. Controls included the probe set hybridized to tRNA only, appropriate control RNA, which serves as integrity control for the RNA sample, and a yeast tRNA as a background control. Individual bands were visualized and quantified by a PhosphoImager (Molecular Dynamics, Piscataway, NJ) and analyzed using Image-Quant v1.1 software (Molecular Dynamics).

Cerebrospinal fluid (CSF) was collected from Glb1^+/+ mice and Glb1^-/- mice, cytospun onto a slide, and stained with May-Grünwald-Giemsa for total cell counts. CSF was collected from the cisternea magna of the subarachnoid space (fourth ventricle) and peripheral blood by cardiac puncture. For the transmigration assay, leukocytes from peripheral blood were counted on a hemocytometer. Filtered chemotaxis medium containing the CSF (35 μL) was placed in the lower transwell chamber. Blood cells (1 × 10⁶ cells/100 μL) were placed in the upper chamber; chambers were separated by a 3-μm pore filter (Falcon; Becton Dickinson, San Jose, CA). After incubation for 4 hours at 37°C in 5% CO₂ atmosphere, the cells that had migrated to the lower chamber were recovered, cytospun onto a microscope slide, stained with May-Grünwald-Giemsa, and counted. Different cell types were distinguished according to their morphologic characteristics and their relative number was calculated.

Human Glb1 cDNA was inserted 5′ of the internal ribosomal entry site in the MSCV-GFP vector.^30,31  An ecotrophic GP⁺ E86-derived producer-cell line was used.

BMCs were isolated from femurs and tibias of Glb1^+/+ mice and lysed with Gey solution, and nucleated cells were counted on a hemocytometer. BMCs (2 × 10⁶ cells/mL) were prestimulated in Iscoves medium containing 20 ng/mL mouse interleukin-3 (IL-3), 50 ng/mL human IL-6, and mouse stem-cell factor (R&D Systems) for 48 hours and then transduced with a high titer of MSCV-β-gal-GFP or MSCV-GFP for 48 hours on retronectin-coated plates (Takara, Kyoto, Japan). Transduced cells were analyzed by flow cytometric analysis on a FACScalibur system (Becton Dickinson). Recipient mice (Glb1^-/-, Glb1^+/+, PPCA^-/-) were irradiated (850 cGy) to eliminate endogenous hematopoietic precursors. Genetically modified BMCs (2-6 × 10⁶ cells) overexpressing β-gal and the green fluorescent protein (GFP) marker (MSCV-β-gal-GFP) or only the GFP marker (MSCV-GFP) were then injected into the tail vein of recipients 24 hours after irradiation. For secondary transplants, BMCs were collected from 2 Glb1^-/- mice that received a transplant 12 weeks earlier. BMCs (3-9 × 10⁶ cells) collected from these animals were 13% and 35% positive for GFP, and were transplanted into irradiated Glb1^-/- mice. Analyses were performed 1, 3, 6, and 9 months after primary and secondary bone marrow transplantations (BMTs).

Blood samples (20 μL) were obtained 1, 3, 6, and 9 months after BMT. Fluorescence activity cell sorting (FACS) analyses of erythrocyte, platelet, and lymphocyte content were done. The level of β-gal activity was measured, as detailed elsewhere.³²  The protein concentration was determined using bicinchoninic acid reagent (Pierce Chemical).

Mouse brains were either formalin-fixed and paraffin-embedded or flash-frozen in liquid nitrogen and then sectioned for immunohistochemistry. The following primary antibodies and dilutions were used: rabbit anti-human β-gal (1:80), rabbit anti-SDF-1β (1:100; Torrey Pines Biolabs, San Diego, CA), rabbit anti-G_M1-ganglioside (1:500; US Biological, Swampscott, MA), mouse Alexa Fluor anti-glial fibrillary acidic protein (GFAP) (1:100; Molecular Probes, Eugene, OR), rat anti-F4/80 (1:100; Serotec, Raleigh, NC), rabbit anti-GFP (1:300; Molecular Probes), and mouse anti-NeuN (1:300; Chemicon, Temecula, CA). The following secondary antibodies and dilutions were used: Alexa Fluor 488 and 594 goat antirabbit (1:500; Molecular Probes), fluorescein rabbit antirat (1:100; Vector, Burlingame, CA), biotinylated rabbit antirat (1:100; Vector), biotinylated goat antirabbit (1;300; PharMingen, San Diego, CA), Alexa Fluor rabbit antigoat (1:200; Molecular Probes), Alexa Fluor 594 goat antimouse (1:500; Molecular Probes), and Alexa Fluor 594 rabbit antirat (1:300; Molecular Probes).

Total lipids were extracted from mouse brain homogenates. The lipid extract corresponding to 100 μg protein was separated on a thin-layer chromatography plate (Silica gel 60 Å; Whatman, Clifton, NJ). Gangliosides were visualized by resorcinol spray and heating. G_M1-ganglioside (TRB Pharma, Campinas, Brazil) was used as the standard.

Three tests were used to ascertain neurologic function: motor coordination and balance, open-field activity, and walking pattern. Glb1^+/+ mice, Glb1^-/- mice, and Glb1^-/- mice that underwent transplantation were tested during the same session to minimize variability. The ability to maintain balance was tested on a standard rotarod apparatus (Stoelting, Wood Dale, IL).³³  Motor and exploratory behaviors were assessed in an acrylic open arena.³⁴  Gait abnormalities were determined by painting the mouse paws with nontoxic, washable paint and then placing the mice in a corridor (10 cm × 10.5 cm × 81.5 cm) lined with white paper.³⁵ 

Data are expressed as mean ± SD and were evaluated by a one-way analysis of variance (ANOVA) followed by the Tukey test for pairwise comparisons. The analyses were performed using the SPSS/PC+ statistical software (SPSS, Chicago, IL) and the mean differences were considered statistically significant when P values were less than .05.

We first tested the cytokine-mediated up-regulation of SDF-1, one of the most potent chemoattractant molecules whose expression is induced in several neurodegenerative and neuroinflammatory conditions.³⁶  Real-time PCR and ELISA analyses of Glb1^-/- brain regions revealed up-regulation of SDF-1α and SDF-1β isoforms; the highest levels of the SDF-1α and SDF-1β mRNAs (Figure 1A) and SDF-1α protein (Figure 1B) were seen in the cerebellum of 3-month-old mice and, to a lesser extent, 4-month-old mice. Analyses of other chemokines during disease progression demonstrated up-regulation of numerous β-chemokines in specific regions of the brain and spinal cord; their activation also peaked at 3 and 4 months of age. The relative increase in mRNA levels varied depending on the chemokine, the CNS region, and the age of the animal. At 3 months, MIP-1β and TCA-3 had the highest relative expression in the cerebellum, and MIP-1β and MIP-1α were highest in spinal cord (Figure 1C); at 4 months, MIP-1α, MIP-1β, and IP-10 were the highest in brain stem, and MIP-1α and MIP-1β were the highest in the cerebellum and spinal cord (Figure 1D). The level of chemokine expression in PPCA^-/- (protective protein/cathepsin A) mice, a model of the LSD galactosialidosis,³⁷  which is associated with storage of glycoproteins rather than glycolipids, was comparable with that of wild-type mice (data not shown).

Under normal conditions, some blood-born cells transiently enter the CSF and are released back into the circulation, unless specific stimuli provoke their retention. During CNS inflammation, T cells and monocytes are retained in the subarachnoid space via signals released by the chemokine receptors.³⁸  These migrating cells scavenge invaders and necrotizing tissue debris, thereby helping to repair tissue damage and promote healing.³⁹ 

At 3 months, the CSF of Glb1^-/- mice contained more white blood cells (WBCs) than did that of Glb1^+/+ mice. Instead, the number of WBCs in the CSF of PPCA^-/- was similar to that of wild-type mice (Figure 1E). Therefore, leukocyte retention in the CSF compartment of Glb1^-/- mice is apparently disease specific; this phenomenon usually results from synergistic effects of multiple chemokines.

To verify these results, we performed transmigration assays using CSF from 3-month-old Glb1^-/- mice and Glb1^+/+ mice as a source of chemoattractants and freshly prepared peripheral WBCs as target cells. The number of WBCs that migrated toward the Glb1^-/- CSF was double the number that migrated toward Glb1^+/+ CSF (Figure 1F). May-Grünwald-Giemsa staining of migrating cells revealed that most transmigrated cells were monocytes (78%). The number of other WBCs (lymphocytes, 12%; neutrophils, 10%) did not differ substantially between Glb1^-/- and Glb1^+/+ CSF samples.

We used an ex vivo gene therapy approach to investigate whether increased chemokine expression in Glb1^-/- mice would facilitate the recruitment of BMCs into the CNS and whether this approach could be a potential treatment for G_M1-gangliosidosis. Wild-type BMCs were transduced with either MSCV-β-gal-GFP or MSCV-GFP viral preparations. We first determined that MSCV-β-gal-GFP-transduced BMCs expressed and secreted high levels of the corrective enzyme, which was readily taken up by β-gal-deficient human fibroblasts and restored enzyme activity (Supplemental Figure S1A-B, available at the Blood website; click on the Supplemental Materials link at the top of the online article). For BMT experiments, the transduction efficiency of BMCs with the MSCV-β-gal-GFP vector ranged from 25% to 89%, and the percentage of stem-cell antigen 1-positive (Sca-1⁺) and cKit⁺ hematopoietic stem cells (HSCs) expressing GFP and β-gal ranged from 4.15% to 7.64%. The transduction efficiency of BMCs with the MSCV-GFP vector ranged from 18.40% to 67.37%. We used 4 to 6 Glb1^-/- recipients for each transplantation in 15 independent experiments. BMCs, transduced with either MSCV-β-gal-GFP or MSCV-GFP (mock transplantations), were transplanted into a total of 52 lethally irradiated 3- to 4-week-old Glb1^-/- mice: 43 received MSCV-β-gal-GFP-transduced BMCs and 9 received MSCV-GFP-transduced BMCs. Wild-type (n = 5) and PPCA^-/- (n = 3) mice that underwent transplantation were used as BMT controls (Supplemental Table S1). For long-term studies, 11 Glb1^-/- mice received a secondary bone marrow (BM) transplant with retrovirally marked BMCs collected from 2 mice that received a primary transplant 3 months earlier. One mouse expressed GFP in 15% of its cells, and the other in 33% of its cells. In both cohorts of treated mice (primary and secondary), GFP-expressing cells of all lineages were detected in peripheral blood samples up to 6 months after primary transplantation (Figure 2A) and up to 9 months after secondary BMT (Figure 2B); this finding indicated efficient engraftment.

GFP expression in peripheral blood was paralleled by an increased level of β-gal activity in most visceral organs (P < .05 relative to untreated mice, post hoc Tukey test). The higher activities were found in the hematopoietic tissues such as the spleen and bone marrow (Figure 2C-D). Immunohistochemical analyses using anti-human β-gal and anti-GFP antibodies identified a substantial number of cells expressing both markers in most of the systemic organs (Supplemental Figure S2). Although Glb1^-/- mice showed mild disease in the visceral organs with no overt morphologic changes,²⁷  these data suggested sustained engraftment of BMCs and long-term expression of the transgene.

After both primary and secondary BMTs, β-gal activity was significantly higher in the brain and spinal cord of treated Glb1^-/- mice than it was in untreated or mock-treated animals (P < .05, post hoc Tukey test) (Figure 2E). The high increment of β-gal activity measured in the cerebellum and brain stem suggested that regions with strong activation of chemokines had preferentially recruited BMCs. We therefore compared the CNS morphology and number of β-gal⁺ cells in brain and spinal cord sections from animals that received genetically modified MSCV-β-gal-GFP BMCs with those that received MSCV-GFP BMCs (mock). The CNS of Glb1^-/- mice that received the therapeutic transgene appeared to have corrected the characteristics associated with the lack of β-gal (ie, lysosomal vacuolation), but in those that underwent mock transplantation the neuropathology persisted (Figure 3A); although the number of GFP⁺ BM-derived cells was comparable in the 2 groups of mice that underwent transplantation (Figure 3C). The degree of correction varied among treated animals and reflected the percentage of transduced BMCs expressing β-gal that had engrafted and were efficiently recruited to different brain regions. Immunolabeling with anti-β-gal antibody revealed sustained expression of the protein particularly in the thalamus, cerebellar Purkinje cells, and the brain stem (Figure 3B), but also in the cortex, pontine nucleus, medulla oblongata, and hippocampus (Supplemental Figure S3). This preferential and efficient recruitment of BM-derived cells into the brain was not observed in either of the BMT control mice (wild-type and PPCA^-/-) that were subjected to the same treatment (data not shown). Many β-gal⁺ cells also expressed GFP and the microglia marker F4/80 (Figure 4A-B), which confirmed their hematopoietic origin. In fact, GFP-expressing microglia-like cells were identified throughout the brain, particularly in regions with the highest expression of chemokines (ie, cerebellum, brain stem, and thalamus) (Figure 3C). Numerous β-gal-expressing cells were also positive for the neuronal marker NeuN (Figure 4C), a finding that suggested efficient metabolic cooperation between BM-derived cells (ie, monocytes) that migrated into the CNS and the host's neurons. Although not all neurons were devoid of vacuolation, the extent of clearance of lysosomal storage paralleled the percentage of engrafted BMCs overexpressing β-gal. Many neurons displayed the typical punctuated pattern of lysosomes even at the 6- to 9-month time points after BMT (Figure 3B and Supplemental Figure S3). These results confirmed that β-gal-transduced HSCs had successfully engrafted.

To ascertain whether retrovirally transduced BMCs that migrated into the CNS would reverse or retard G_M1 accumulation, we measured the amount and distribution of G_M1 in different brain regions of the treated mice and compared it with that in age-matched wild-type mice and untreated mice or Glb1^-/- mice that underwent mock transplantation (Figure 5). Thin-layer chromatography revealed a clear decrease in the G_M1 content in brain stem and cerebellum of treated mice, although the G_M1 level in total brain extracts, comprising the hindbrain, midbrain, and forebrain regions, was not substantially different from that in untreated littermates (Figure 5A). The reason for the latter finding was due to the wide heterogeneity of these areas when analyzed together. In fact, immunofluorescence staining of brain sections with anti-G_M1 antibody demonstrated a clear reduction of storage in other brain nuclei, such as the hippocampus, of treated Glb1^-/- mice compared with mice that underwent mock transplantation (Figure 5B). The decrease in G_M1 levels in each brain region correlated with the relative increase in β-gal activity (Figure 5A, bottom).

Evidence of successful correction of CNS pathology was further provided by the results of RNAse protection assays, which showed decreased levels of chemokine expression at 3 and 4 months after BMT, a time that corresponded to peak expression of several chemokines in untreated mutant mice (Figure 6A-B). The chemokines that displayed the highest activation in Glb1^-/- mice (ie, MIP-1α, MIP-1β, IP-10, and TCA-3) responded best to the treatment. In addition, real-time PCR analyses revealed substantial down-regulation of SDF-1α and SDF-1β mRNAs in the cerebella of treated mice (Figure 6C), and ELISA results revealed an equivalent reduction in SDF-1α protein (Figure 6D).

Immunohistochemical labeling with anti-SDF-1β antibody confirmed that considerably fewer SDF-1β⁺ cells were present in the thalamus of treated Glb1^-/- mice than in mutant mice that underwent mock transplantation (Figure 7A). SDF1β-expressing cells were identified as astrocytes by coimmunostaining with anti-GFAP antibody (available as Supplemental Figure S3). Thus, transplantation of retrovirally transduced BMCs expressing the therapeutic enzyme appeared to reverse chemokine activation. The neuroinflammatory response was also reduced, as demonstrated by fewer F4/80⁺ microglia in the thalamus (shown as example) of treated mice, while Glb1^-/- mice that underwent mock transplantation still exhibited numerous spheroid microglia, a morphologic characteristic of their activated status⁴⁰  (Figure 7A). Using an anti-GFAP antibody, we also found that treated Glb1^-/- mice had greatly attenuated astrogliosis particularly in the thalamus (Figure 7A), which is one of the brain nuclei most affected by this phenotypic abnormality in G_M1-gangliosidosis mice.²⁹  Overall, these findings indicate decreased neuroinflammation in the treated Glb1^-/- mice.

Accumulation of G_M1 in the ER of Glb1^-/- neurons activates a UPR that ultimately results in neuronal apoptosis mediated by the proapoptotic proteins CHOP (CCAAT/enhancer-binding protein [C/EBP]-homologous protein) and caspase-12.²⁸  We therefore tested whether the decreased G_M1 content in the brains of treated Glb1^-/- mice would affect the expression of these UPR effectors. Real-time PCR analyses of brain regions and spinal cord of treated Glb1^-/- mice showed that the level of CHOP mRNA was practically normalized, and that of caspase-12 was drastically reduced 3 months after transplantation (Figure 7B-C).

Neuromotor abilities progressively worsen in Glb1^-/- mice and culminate in ataxia, tremor, and total paralysis of the hind limbs.²⁷  Behavioral testing demonstrated that Glb1^-/- mice have an impaired walking pattern: their gait was slow, and they tended to walk in small, labored, uncoordinated movements. Stride strength was significantly reduced, and the paw print length was increased. In contrast, long-term-treated Glb1^-/- mice walked faster, with increased stride length and improved coordination. On a rotarod test of balance, the performance of treated Glb1^-/- mice was significantly better than that of untreated mutant mice (Figure 7D). Also in an open-field test of locomotor and exploratory activity, treated mice were significantly more active than untreated mutant mice, and the treated mice left the center square of the open field significantly quicker (Figure 7E). In addition, at 6 months after BMT, treated mice showed increased rearing behavior, which indicated improved strength of the hind limbs.

The CNS is concealed behind the blood-brain barrier (BBB), which impedes the migration of immune cells and the diffusion of plasma proteins,^41,42  giving the CNS an immunologically privileged status. The functional integrity of the BBB is altered in inflammatory diseases of the CNS where leukocytes migrate across the BBB and the brain parenchyma^16,43  under control of chemical messengers such as proinflammatory cytokines, chemokines, and adhesion molecules.^22,44-46  Proinflammatory cytokines stimulate astrocytes, microglia, and endothelial cells to produce chemokines,⁴⁷  which in turn activate matrix metalloproteinases that degrade extracellular matrix proteins and disrupt the endothelial tight junctions of the microvessels in the brain.⁴⁸ 

Here we provide evidence that the neuroinflammation associated with the neurodegenerative condition in G_M1-gangliosidosis mice provokes the up-regulation of selected chemokines in specific brain regions. One of these chemokines is SDF-1, a potent chemoattractant for microglia that has been implicated in controlling the formation of glial scars and the restoration of the BBB after injury.⁴⁹  SDF-1 participates in the recruitment of mononuclear cells into the CNS, and its intrathecal production is elevated in several neuroinflammatory diseases.⁵⁰  We have also detected activation of numerous β-chemokines, specifically in the brain stem and cerebellum of Glb1^-/- mice. Although the reasons for the time-dependent up-regulation of individual chemokines during the progression of the disease remain unclear, this regulatory pattern may relate to the differential stages of the neuroinflammatory response in mice of different ages. For example, MCP-1 mediates the early response to chemical-induced brain damage,⁵¹  and it is the earliest and the most activated chemokine in young Glb1^-/- mice. In contrast, MIP-1α, MIP-1β, and IP-10 are produced by glia and infiltrating leukocytes in chronic neurodegenerative diseases^19,52,53 ; the activity of these chemokines peaks in older Glb1^-/- mice and persists during the later stages of disease. Finally, the number of monocytes in the CSF of Glb1^-/- mice was elevated, which is suggestive of chemokine-mediated recruitment and retention of peripheral blood cells in the CSF compartment.

The up-regulation of chemokines in the Glb1^-/- mice suggests that this disease condition could create a microenvironment within the CNS that would favor the response to ex vivo gene therapy. Ex vivo gene therapy using virally transduced BMCs expressing the therapeutic enzyme is potentially feasible for treating LSDs, because engrafted BMCs can infiltrate and repopulate organs, including the CNS,⁵⁴  and cross-correct enzyme-deficient cells. This procedure has been implemented with promising results in other animal models of LSDs.^23,55-58  It is becoming increasingly clear, however, that individual LSDs respond differently to the same therapeutic paradigm, most likely because of the nature of the enzyme defect, the accumulated products, and their combined effect on cell metabolism.

Animals that received primary and secondary transplants of marked BMCs showed persistent GFP expression in peripheral blood and increased enzyme activity in the systemic organs and the CNS. This finding suggests that gene transfer had efficiently targeted long-term repopulating stem cells. The nuclei most affected by G_M1-gangliosidosis,^27,28  namely the thalamus, pontine nucleus, cerebellum, and brain stem, were also the sites in which chemokines were up-regulated and the neuronal phenotype was most corrected after transplantation. In these areas, we also found a consistent, elevated number of β-gal⁺ BMCs that also expressed GFP and microglia-specific markers. The restoration of enzyme function in the brain of transplant recipients led to a dramatic decrease in G_M1-ganglioside storage and was accompanied by attenuation of the neuroinflammatory response. Most of the disease-activated chemokines, including SDF-1, MIP-1α, MIP-1β, and IP-10, were down-regulated in response to therapy. Remarkably, the levels of transcription of CHOP and caspase-12, 2 components of the UPR pathways responsible for neuronal apoptosis in Glb1^-/- mice,²⁸  were also drastically reduced after BMT. Finally, in the brain of treated mice, β-gal-expressing cells persisted long term (up to 9 months after secondary BMT), indicating continuous recruitment of BMCs from circulating precursors. This finding makes it unlikely that radiation-induced tissue damage before BMT was responsible for the recruitment of correcting cells into the CNS.

Normalization of the molecular effectors of CNS pathogenesis was paralleled by a strikingly improved gross appearance of the mice at a time when untreated Glb1^-/- animals are severely affected by the disease. Tremor, ataxia, rigidity, and inability to walk and eat improved in animals that underwent BMT 6 to 9 months earlier. Overall performance of treated mice improved on behavioral tests assessing motor function, balance, coordination, and exploratory activity. Together, these results indicate that CNS function was, at least in part, rescued by ex vivo gene therapy. It is important to reiterate, however, that not all mice that received transplants experienced the same extent of correction of their neuronal pathology. Differences in neuronal correction may reflect those in the engraftment of HSCs and in stem-cell-specific proviral integration sites, which may influence the transcriptional activity of the provirus.^59,60  Curiously, the BMCs with higher transduction efficiency before transplantation usually engrafted better and persisted longer after transplantation into Glb1^-/- mice.

These studies underscore the crucial role of chemokine upregulation in the CNS in response to BMC-mediated therapy of neurodegenerative and neuroinflammatory conditions associated with G_M1-gangliosidosis. Since PPCA^-/- mice did not show altered levels of chemokines in their CNS, this response seems to be characteristic of only some LSDs and underlines differences in pathogenesis that may ultimately impact the efficacy of therapy. Other glycolipid storage diseases, such as Krabbe disease and G_M2-gangliosidosis, also involve up-regulation of these molecular effectors.^23-25  In the Twitcher mouse, preferential recruitment of hematopoietic cells to the demyelinating areas of the affected brain is most likely caused by activation of MCP-1 and IL-10; in the G_M2-gangliosidosis mouse, time-dependent up-regulation of MIP-1α facilitates the migration of blood cells into the brain parenchyma.²⁵ 

We observed the most extensive improvement of the disease phenotype in treated Glb1^-/- mice at the same time points and brain regions that develop the highest levels of chemokine upregulation, namely at 3 to 4 months after transplantation. We believe that the pattern of activation of chemokines coincides or immediately follows the time-dependent progression of neuronal-cell death and neuroinflammation characteristic of this disease. These results directly implicate chemokines in the recruitment of therapeutic BMCs into the brain in G_M1-gangliosidosis and possibly other LSDs. If the latter assumption is correct, the positive response to normal BMT in G_M2-gangliosidosis mice and the efficient engraftment of lentivirus-transduced BMCs in the mouse model of metachromatic leukodystrophy can also be attributed to up-regulation of chemokines at specific CNS sites.^25,57  Further study of CNS chemokines and leukocyte recruitment into the CNS may lead to improved therapeutic strategies for G_M1-gangliosidosis and similar neurodegenerative disorders.

We wish to thank Gerard Grosveld for continuous support; Taylor Walker and Elida Gomero for animal maintenance; William J. Martin for help in the collection of the CSF and blood samples; Huimin Hu for help with the immunohistochemistry; Ann Marie Hamilton-Easton and Richard Ashmun for FACS analyses; Tommaso Nastasi and Erik Bonten for technical suggestions; and Angela McArthur and Charlette Hill for editing and formatting the article. R.S. thanks Roberto Giugliani and Janice Coelho for encouragement and support in preparation of her thesis.