Novel Biomarkers of Human GM1 Gangliosidosis Reflect the Clinical Efficacy of Gene Therapy in a Feline Model

Abstract

GM1 gangliosidosis is a fatal neurodegenerative disease that affects individuals of all ages. Favorable outcomes using adeno-associated viral (AAV) gene therapy in GM1 mice and cats have prompted consideration of human clinical trials, yet there remains a paucity of objective biomarkers to track disease status. We developed a panel of biomarkers using blood, urine, cerebrospinal fluid (CSF), electrodiagnostics, 7 T MRI, and magnetic resonance spectroscopy in GM1 cats—either untreated or AAV treated for more than 5 years—and compared them to markers in human GM1 patients where possible. Significant alterations were noted in CSF and blood of GM1 humans and cats, with partial or full normalization after gene therapy in cats. Gene therapy improved the rhythmic slowing of electroencephalograms (EEGs) in GM1 cats, a phenomenon present also in GM1 patients, but nonetheless the epileptiform activity persisted. After gene therapy, MR-based analyses revealed remarkable preservation of brain architecture and correction of brain metabolites associated with microgliosis, neuroaxonal loss, and demyelination. Therapeutic benefit of AAV gene therapy in GM1 cats, many of which maintain near-normal function >5 years post-treatment, supports the strong consideration of human clinical trials, for which the biomarkers described herein will be essential for outcome assessment.

GM1 gangliosidosis is a fatal, untreatable neurodegenerative disease of children and adults. Gray-Edwards and colleagues demonstrate >5-year survival in AAV-treated GM1 cats and describe novel minimally invasive biomarkers for use in clinical trials. Analyses of blood, cerebrospinal fluid, electroencephalography, MRI, and MRS are included from GM1 cats and human patients.

Introduction

GM1 gangliosidosis (GM1) is a fatal neurodegenerative disease that affects individuals of all ages.¹ GM1 is one member of a class of >70 distinct disorders known as lysosomal storage diseases (LSDs), which have a combined occurrence of 1:7,700 live births.2, 3 GM1 is caused by a deficiency of β-galactosidase (β-gal, EC 3.2.1.23), the first enzyme in the catabolism of complex glycosphingolipids, ultimately resulting in storage of GM1 ganglioside and its asialo derivative (GA1).4, 5 Although a clinical continuum, GM1 gangliosidosis can be classified into three types (I–III), determined by the rate of disease progression and age of onset. These often correlate with the level of residual β-gal activity (i.e., the least residual activity often causes the most severe disease).⁶ The most common disease forms affect children and are designated as infantile and juvenile onset (type I and type II, respectively), while the least common, chronic, and most variable phenotype occurs in adults (type III). A late-infantile classification has been used to distinguish patients whose disease severity is intermediate to the infantile and juvenile forms. In all cases, GM1 gangliosidosis is characterized primarily by progressive deterioration of the nervous system, though other organs are also involved. Survival ranges from 2 years for the infantile-onset phenotype to a normal lifespan in some adult-onset patients.¹

Though mice are invaluable in the study of pathogenesis and therapeutic discovery, they do not always accurately reproduce human disease,7, 8, 9 so it is desirable to test potential therapies in a model more similar to humans.10, 11, 12, 13, 14 Feline GM1 gangliosidosis was first described in 1971¹⁵ and faithfully recapitulates late-infantile or juvenile GM1 gangliosidosis. Over the last 35 years, the feline gangliosidosis model has been rigorously studied and is an ideal model for long-term evaluation of novel therapies. We have developed an adeno-associated viral (AAV)-mediated gene therapy that exhibits extraordinary efficacy in GM1 animal models,16, 17, 18 including GM1 cats in which a >4-fold survival increase has been reported.¹⁹ In that study, extensive analysis of the brain and spinal cord was performed in cats treated short term (16 weeks) or long term (humane endpoint). Widespread distribution of β-gal activity ranged from normal to 6.6-fold normal, with corresponding reduction of storage material to normal levels in most regions of the CNS. Approximately half of the treated cats from the original report¹⁹ remain healthy, with mild or no clinical signs of disease at 5.7 years of age or older (untreated lifespan of 8.0 ± 0.6 months). To best evaluate storage in these valuable, long-lived cats, we used targeted lipidomics in cerebrospinal fluid (CSF) samples to measure >30 lipids, including GM1 gangliosides, and found dramatic reductions reported in a companion manuscript. Although valuable, lipidomics analysis required a liquid chromatography-tandem mass spectrometry assay only available in specialized laboratories. Thus, in the current study, we identified objective biomarkers of disease progression using rapid, clinically available tests that track disease severity in feline GM1 gangliosidosis and correlate with improved phenotypes after gene therapy. Many of the biomarkers also reflect disease severity in human GM1 patients and would therefore be useful for assessment of disease status and therapeutic response in future clinical trials.

Results

Treatment Groups and Survival Information

GM1 cats were treated with an intracranial injection of an AAV vector encoding feline β-galactosidase and were subdivided into two cohorts by capsid, AAV1 and AAVrh8, as previously described.¹⁹ Half of AAV-treated GM1 cats in both cohorts from the original report are still healthy, with mild or no clinical disease (Movie S1), and mean survival of all treated cats is 4.4 ± 2.0 years, or >6.7 times greater than untreated controls (Movie S2). Mean survival of the five treated cats that remain alive is 6.4 ± 0.5 years, with a range of 5.7–7.0 years. A current Kaplan-Meier survival curve is shown in Figure 1A. AAV-treated GM1 cats were assessed using serial hematologic and chemistry profiles in blood and CSF, plus brain MRI and magnetic resonance spectroscopy (MRS), to develop biomarkers of disease progression over years. Biomarkers identified in the cats were then evaluated for translatability to infantile, late-infantile, and juvenile GM1 gangliosidosis patients.

Figure 1.

GM1+AAV Cat Survival and CSF Biomarkers in GM1 Cats and Humans

(A) Kaplan-Meier curve showing survival of GM1 cats after AAV gene therapy. Ongoing GM1 cats treated by AAV1 or AAVrh8 display survival up to or beyond 60 months of age compared to a mean survival of 8.0 ± 0.6 months in untreated GM1 cats. (B) Aspartate aminotransferase (AST) was significantly elevated in untreated GM1 cats (n = 4; p = 0.01) compared to normal cats (n = 16; ages ranging from 4 months to 5 years). CSF was evaluated at 16 weeks in the GM1+AAVrh8 cohort (n = 4) and GM1+AAV1 cohort (n = 3). In long-term (LT) follow-up of animals with minimal clinical disease (2–3 years post-treatment; n = 5 per serotype), AST levels were significantly reduced compared to untreated levels (p = 0.01). When treated animals reached the clinical humane endpoint (n = 3 from AAV1 and n = 2 from AAVrh8), AST levels were intermediate to normal and untreated levels. (C) CSF AST levels were increased in two late-infantile GM1 patients compared to control patients who did not have lysosomal storage disease (n = 13). CSF AST was near normal levels in almost all juvenile patients (n = 14), except for one (GSL020) who, like the late-infantile patients, had accelerated neurologic decline at the time of testing. (D) LDH was significantly elevated in the untreated GM1 cat (n = 4) compared to normal (n = 16; p = 0.003), and gene therapy normalized LDH levels both at 16 weeks (AAVrh8, n = 4; AAV1, n = 3) and long term (n = 5 per group; p = 0.01). (E) LDH in CSF from late-infantile GM1 patients (n = 2) was significantly increased compared to the juvenile cohort (n = 14). Normal LDH values were not measured. *p < 0.05 and **p < 0.01 from normal control cats, ^ƚp < 0.05 and ^Ŧp < 0.01 from untreated GM1 cats at humane endpoint, ^⋄⋄p < 0.01 from juvenile GM1 patients.

CSF Evaluation

CSF aspartate aminotransferase (AST) and lactate dehydrogenase (LDH), enzymes representing cellular damage or leakage, were significantly elevated in GM1 cats at their humane endpoint (p = 0.01 and p = 0.004, respectively) (Figures 1B and 1D). Sixteen weeks after AAV gene therapy with either AAV1 or AAVrh8 vectors, AST and LDH were reduced to normal levels, and they remained so during long-term follow-up of cats with attenuated disease progression (1–3 years of age). However, in five treated cats that progressed to the humane endpoint, AST levels were intermediate to normal and GM1 cats (Figure 1B). In addition, the two AAV-treated animals with large LDH elevations had increased seizure activity near the time of death (Figure 1D). In humans, AST was elevated in late-infantile GM1 patients compared to normal and juvenile GM1 patients (Figure 1C). LDH also was elevated in late-infantile versus juvenile GM1 patients (Figure 1E). Though normal human CSF LDH was not measured, high LDH levels in late-infantile patients are noteworthy compared to the previously published reference range of 3–17 U/L.²⁰ For the single juvenile GM1 patient with profound elevations of both AST and LDH, the CSF sample had been collected at a time of unusually accelerated neurologic decline (Figures 1C and 1E).

Blood Chemistry

Routine blood chemistry analyses were performed at the following ages: 5–12 weeks (pre-symptomatic and pre-treatment), 19–31 weeks (moderate to severe disease), 50–80 weeks, and >100 weeks. Untreated GM1 cats do not survive beyond 19–31 weeks of age. Similar to CSF, serum from untreated GM1 cats showed AST elevations at all time points (Figure 2A). With gene therapy, serum AST levels were normalized by 50–80 weeks in both AAV1 and AAVrh8 cohorts and remained normal beyond 100 weeks of age. Serum AST was further evaluated as a surrogate biomarker for CSF. AST levels were greater in CSF than in serum in untreated GM1 cats, whereas the inverse relationship was found for AAV-treated GM1 cats and normal controls (Figure 2B). Nevertheless, AST levels in serum were reflective of those in CSF. In human GM1 patients, serum AST also correlated with disease severity, with significantly higher values in infantile and late-infantile patients than in juvenile patients (Figure 2C). In individuals with serial samples collected over time (Figure 2D), AST increased longitudinally in each infantile patient and each GM1 cat, suggesting that it reflects disease progression in any given subject in these cohorts. However, the ability of AST to chart disease severity in individual patients was less clear in the late-infantile group (n = 3) (Figure 2D). In contrast to results in CSF, serum LDH was not elevated in GM1 cats (data not shown).

Figure 2.

CSF and Blood-Based Biomarkers of GM1 Gangliosidosis in Cats and Humans

Blood-based biomarkers were measured at 5–12 weeks, 19–31 weeks, 50–80 weeks, and 100+ weeks in normal cats, GM1 cats, and GM1+AAVrh8- and GM1+AAV1-treated cats. Biomarkers were also tested in human patients. See the symbol legend at the bottom of the figure for cohort designations. (A) Blood AST levels were elevated in the untreated GM1 cat and normalized after AAV gene therapy. (B) AST abnormalities in CSF were also found in serum. *p < 0.05 and **p < 0.01 from CSF within a cohort. (C) Serum AST levels in each cohort of human GM1 patients (with trend lines shown). (D) Serum AST levels in individual human patients (left panel) or cats (right panel) over time are connected by solid lines, demonstrating that AST correlates with disease progression in any given GM1 cat or infantile patient. Utility of serum AST to chart disease progression in late-infantile patients is less clear (left panel). Other parameters altered in serum included calcium (E and F), creatinine (G and H), and albumin (I and J). Yellow horizontal lines indicate age-normalized levels in human patients. Cat numbers for serum: 5–12 week, 19–31 week, 50–80 week, and 100+ week cohorts, respectively, are as follows: normal, n = 16, 25, 11, and 15; GM1, n = 7, 18, 0, and 0; AAVrh8, n = 8, 12, 10, and 17; AAV1, n = 7, 7, 6, and 18. Cat numbers for CSF in (B): normal, n = 16; GM1, n = 4; GM1+AAV, n = 5 per cohort. Human numbers: infantile, n = 7; late infantile, n = 11; juvenile, n = 42. Statistics for all cat data except (B): *p < 0.05 and **p < 0.01 from normal control cats, ^ƚp < 0.05 and ^Ŧp < 0.01 from untreated GM1 cats at humane endpoint. Statistics for human data: **p < 0.01 from infantile, ^ƚp < 0.05 from late infantile, ^Ŧp < 0.01 from late infantile, ^⋄⋄p < 0.01 from juvenile GM1 patients.

Hypocalcemia was present in GM1 cats (Figure 2E) and in infantile GM1 patients relative to the median age-matched normalized value (yellow line in Figure 2F). Hypocalcemia was not present in late-infantile or juvenile GM1 patients. Calcium was normalized in AAV-treated cats during 50–80 weeks, becoming slightly but significantly elevated after >100 weeks (p < 0.05) even after calcium concentrations were corrected for hypoalbuminemia (data not shown). Creatinine levels were reduced in untreated GM1 cats at all ages and in late-infantile and juvenile patients relative to infantile patients (Figures 2G and 2H). Gene therapy produced varying degrees of normalization in long-term follow-up of treated cats (Figure 2G). Hypoalbuminemia was present in untreated GM1 cats at both early and late time points, with amelioration after AAV treatment (Figure 2I). Hypoalbuminemia in human GM1 patients became more pronounced with disease severity, ranging from no reduction in juvenile patients to the greatest reduction in infantile patients (Figure 2J).

Alkaline phosphatase (ALP) was almost double normal values in untreated GM1 cats during 19–31 weeks (p < 0.01) (Figure S1A). ALP also was statistically higher in infantile GM1 patients than in late-infantile and juvenile patients (p < 0.01) (Figure S1B), perhaps due to a loss of bone mass secondary to decreased ambulation. After gene therapy with either vector in cats, ALP was above normal at earlier ages but gradually decreased to near-normal levels by the >100 week time point.

Some chemistry parameters trended differently in human patients and cats yet showed significant changes in treated animals. For example, serum alanine aminotransferase (ALT) was within normal limits in all cohorts of GM1 patients (Figure S1D) but was elevated in GM1 cats (Figure S1C). AAV treatment in cats significantly reduced ALT levels at the 19- to 31-week time point, and ALT remained stable until 100+ weeks. Total plasma protein was reduced in infantile and late-infantile GM1 patients compared to juvenile patients (Figure S1F) yet was within normal limits in untreated GM1 cats (Figure S1E). Sustained hyperproteinemia was found in long-term follow-up of both AAV cohorts.

In some instances, chemistry abnormalities in GM1 cats were not corrected by gene therapy. For example, in the GM1 cat, the inflammatory marker globulin²¹ was elevated during 19–31 weeks and remained unchanged by treatment with either vector (Figure S2A). In addition, a prominent reduction in the albumin-to-globulin ratio was present in GM1 animals despite therapy (Figure S2B). Reduction of albumin, a negative acute phase protein,²¹ and increased globulin may indicate systemic inflammation.

Hematologic Parameters

No significant differences were identified in the complete blood count (CBC) in human GM1 infantile, late-infantile, or juvenile patients. However, mean platelet volume (MPV), which was elevated in untreated GM1 cats, declined to normal levels in AAV-treated cats (Figure S2C). White blood cell (WBC) number was elevated in untreated and treated GM1 cats during 19–31 weeks, with incomplete correction thereafter in cats treated by gene therapy (Figure S2D). Differential WBC counts revealed a lymphocytopenia in untreated GM1 cats that appeared to be corrected by long-term therapy (Figure S2E). Conversely, neutrophil number was elevated in untreated GM1 cats and was refractory to AAV treatment (Figure S2F).

Erythrocytopenia was present throughout the lifespan of untreated GM1 cats, with correction in AAV-treated cats after >100 weeks (Figure S2G). Similarly, hemoglobin concentration was reduced in untreated GM1 cats but was normal after >100 weeks in cats that received gene therapy (Figure S2H). Almost half of GM1 red blood cells have a spiculated appearance (echinocytosis) during 19–31 weeks, versus ∼8% in normal cats (Figures S2I–S2K), as previously reported in feline GM2 gangliosidosis.²² Intracranial AAV gene therapy had no effect on echinocytosis (Figure S2). One representative normal blood smear is shown because no differences were noted in normal cats over time.

Urinary Glycosaminoglycans

β-gal cleaves not only GM1 ganglioside but also peripheral tissue substrates such as keratan sulfate,²³ a glycosaminoglycan (GAG) that is excreted in urine. Urinary GAGs were significantly above normal in GM1 cats at humane endpoint (p < 0.01). Urinary GAG levels in AAV-treated cats remained significantly higher than in control cats but lower than in untreated GM1 cats (p < 0.05) (Figure S3).

Electroencephalography

The unsedated electroencephalogram (EEG) in GM1 patients is characterized by a slowing of rhythmic activity (Figure 3A) compared to that of age-matched normal controls (Figure 3B). The late-infantile GM1 patient shown, GSL013, exhibited moderate to severe cerebral dysfunction characterized by continuous, irregular, and often semi-rhythmic delta activity with only mild amounts of theta activity. The posterior dominant rhythm is weak with poor formation, sustainability, and regulation and is lost in some areas. In cats, sedated to facilitate data collection, the untreated GM1 EEG (Figure 3C) is characterized by symmetrical slowing of rhythmic activity predominated by slow delta waves superimposed with occasional alpha and theta waves compared to the normal cat (Figure 3D). The normal cat control exhibits occasional small sharp spikes, which are likely benign epileptiform transients associated with sedation (Figure 3D). After AAV gene therapy, cat 9-1356 had independent multifocal sharp or spike discharges but did not show slowing as exhibited by the untreated GM1 cat (Figure 3E). In another AAV-treated cat (8-1435), sharp spikes were markedly reduced and the EEG took on a more normal appearance (Figure 3F). Sharp spikes in 8-1435 may have been due at least partially to sedation, as in the normal control. Unfortunately, EEGs are not possible in awake cats. Both AAV-treated cats exhibited intermittent seizure activity and were being treated with antiepileptic medications but were otherwise healthy at the time of analysis. Similar EEG findings were noted in another AAV-treated, asymptomatic cat that did not have a history of seizures (9-1424), demonstrating that in some cases, EEG abnormalities were subclinical (data not shown).

Figure 3.

EEG Tracings of GM1 Cats and Humans and the Effect of AAV Gene Therapy

The late-infantile GM1 patient GSL013 exhibits rhythmic slowing of brain waves (A) compared to an age-matched normal control (B). Similarly, the GM1 cat exhibits rhythmic slowing (C) compared to normal controls (D). Between 4.5 and 5 years after AAV gene therapy, rhythmic slowing is largely ameliorated in treated cats (E and F). However, epileptiform activity is prominent in one AAV-treated cat (9-1356) just before endpoint (E). Representative images are shown from multiple EEGs: GM1 patient, n = 11; normal patient, n = 3; GM1 cat, n = 3; normal cat, n = 2; GM1+AAV1, n = 1; GM1+AAVrh8, n = 2.

7 T MRI

Normal cat T2-weighted brain MRI shows hypointense (dark) white matter compared to the surrounding hyperintense (light) gray matter, as seen in cerebral cortex or thalamus (Figure 4A) and deep cerebellar nuclei (DCN) (Figure 4E). Untreated GM1 cats at humane endpoint exhibit a blending of gray and white matter intensities (isointense), especially apparent in DCN, with more severe areas of hyperintense white matter compared to cortical gray matter (Figures 4B and 4F, black arrow). In addition, cortical atrophy is noted in the untreated GM1 cat, as shown by increased CSF (bright white) surrounding the brain (Figure 4B, black arrowhead outlined in white, and Figure 4F). Approximately 5 years after AAV gene therapy, MRI intensities of gray and white matter remain largely preserved (Figures 4C, 4D, 4G, and 4H). Representative MRIs from two treated cats show hyperintensities at the thalamic injection sites (black arrowhead), with no demonstrated clinical sequelae. In previous studies, T2 hyperintensities were found loosely associated with regions of granular, eosinophilic neurons, which were attributed to enzyme overexpression in feline models of both GM1 and GM2 gangliosidosis.11, 19 In addition, one cat (8-1435) exhibits an area of hypointensity at the site of a thalamic hemorrhage during surgery (white arrowhead), from which the cat made a complete clinical recovery within 7 days (Figure 4D). The thalamic hypointensity is expected to be an area of residual tissue damage, though confirmation awaits postmortem evaluation. The same cat also has bilateral hyperintensities in the ventral aspect of the brainstem approximately located at the olivary nucleus (Figure 4H, white arrow), again without any known clinical effect. Such T2 hyperintensities can indicate numerous pathologies, including primary demyelination or hypertrophic olivary degeneration due to atrophy of the cerebellar cortex or dentate nucleus.²⁴

Figure 4.

Ultra-High-Field T2-Weighted MRI of GM1 Cats

Normal cats have distinct regions of hypointense (dark) white matter and hyperintense (light) gray matter in the cortex or thalamus (A) and cerebellum (E), with DCN hypointensity especially prominent. Indistinct regions of white and gray matter (isointense) develop in untreated GM1 cats, perhaps most apparent in the DCN (F). Some regions of cortical white matter even become hyperintense to (lighter than; black arrow) gray matter, and cortical atrophy is noted by the increase in CSF (bright white; black arrowhead outlined in white) outlining the brain due to diminished gyri and enlarged sulci (B). Five years after gene therapy, gray and white matter intensities are largely restored in AAV-treated cats (C, D, G, and H) and thalamic hyperintensities (black arrowhead) are common (C and D), though not universal (i.e., not all treated cats had hyperintensities). Cat 8-1435 exhibits an area of hypointensity at the site of a thalamic hemorrhage during surgery (white arrowhead), from which the cat made a complete clinical recovery. In addition, bilateral hyperintense areas in the brainstem near the olivary nucleus are noted in one treated cat (8-1435; white arrow; H). Shown are representative images of several MRI scans: normal, n = 4; GM1, n = 6; GM1+AAVrh8, n = 3.

Magnetic Resonance Spectroscopy

Spatial evaluation of therapeutic efficacy was performed using MRS in six brain regions (voxels) in normal, GM1, and GM1+AAVrh8 cats 3–5 years after treatment (Figure 5). Metabolites analyzed include the glial marker myoinositol (INS), neuronal marker N-acetylaspartate (NAA) alone or with N-acetylglutamate (+NAAG), demyelination indicators glycerophosphocholine and phosphocholine (GPC+PCh), metabolism markers creatine and phosphocreatine (Cr+PCr), and the neurotransmitter glutamate and its precursor glutamine (Glu+Gln). The cerebellum was the most severely affected voxel in untreated GM1 cats, with five of six metabolites showing significant differences from normal (p ≤ 0.05). NAA and GPC+PCh were abnormal for GM1 cats in both early (4 months) and late (8 months) disease stages (p < 0.01). The remaining metabolites, Ins, NAA+NAAG, and Cr+PCr, were abnormal in GM1 cats at humane endpoint, and Glu+Gln trended similarly (p = 0.052). In AAV-treated cats >3 years post-injection, all abnormal metabolites were corrected partially or completely.

Figure 5.

Single-Voxel Magnetic Resonance Spectroscopy of GM1 Cats

MRS was performed in six voxels: thalamus (A), corona radiata (B), temporal cortex (C), parietal cortex (D), occipital cortex (E), and cerebellum (F) in normal cats at 4 months, 8 months, and 2–5 years (n = 4 per cohort); untreated GM1 cats at 4 months (n = 6) and at 8 months (n = 6); and GM1+AAVrh8 cats at 3–5 years (n = 3). Metabolites analyzed include the glial cell marker myoinositol (INS), neuronal marker N-acetylaspartate (NAA), demyelination indicators glycerophosphocholine and phosphocholine (GPC+PCh), NAA+N-acetylglutamate (NAA+NAAG), metabolism markers creatine and phosphocreatine (Cr+PCr), and glutamate and glutamine (Glu+Gln). *p < 0.05 and **p < 0.01 from normal control cats, ^ƚp < 0.05 from untreated GM1 cats at humane endpoint. Error bars are SD.

Metabolite concentrations from the cerebellar voxel correlated with clinical signs. NAA was positively correlated with clinical function (R² = 0.63), in that higher levels of NAA predicted improved clinical status (Figure S4A). Inverse correlations to neurologic function were found for GPC+PCh (R² = 0.46) and Ins (R² = 0.59) (Figures S4B and S4C).

The most common metabolite changes across all voxels in untreated GM1 cats were GPC+PCh, elevated in all regions, and NAA, decreased in four of six regions. Gene therapy restored GPC+PCh to normal levels in all but one region (parietal cortex), perhaps reflecting the myelin preservation that was also apparent on anatomical MRI scans (Figure 4). NAA was stabilized or slightly increased in all brain regions >3 years after treatment, although levels trended toward a significant reduction in the cerebellum (p = 0.062) (Figure 5). Post-treatment Glu+Gln were significantly decreased in two of six brain regions, including thalamus (p < 0.01) and corona radiata (p < 0.05), and trended toward a reduction in the cerebellum (p = 0.056), making them the metabolites most resistant to long-term normalization. However, because Glu+Gln were significantly reduced only in treated GM1 cats (not in untreated cats), a direct effect of AAV treatment on these metabolites cannot be ruled out. Over the six brain regions analyzed, 16 of 36 metabolite measurements were abnormal in untreated GM1 cats. After gene therapy, most metabolite abnormalities were corrected (Figure 5).

Discussion

In a previous report, intracranial gene therapy with AAV1 or AAVrh8 vectors extended the lifespan of GM1 cats >4.7-fold, with many treated cats displaying only mild disease signs at the time of publication. Five AAV-treated GM1 cats from the previous work remain alive at 5.7–7.0 years of age, or >8-fold beyond the untreated lifespan, with minimal signs of disease (Movie S1). The only obvious neurologic abnormality in some treated cats is seizure activity, well controlled by medication. With such clear efficacy from AAV therapy in animal models, sensitive and noninvasive biomarkers of disease progression are needed for human clinical trials. In this study, we identified quantitative biomarkers that changed with disease progression in GM1 cats and human patients. In addition, because clinical symptoms and assessments remain of primary importance, we correlated biomarkers with clinical disease severity in GM1 cats treated with gene therapy.

CSF is a minimally invasive reservoir for CNS-derived biomarkers and for this reason was selected for analysis. In the CSF of GM1 cats and human late-infantile patients, AST and LDH elevations correlated with the severity of clinical neurologic disease. Reported decades ago and reexamined recently as biomarkers of the gangliosidoses,22, 25, 26, 27 AST and LDH had not been evaluated in human GM1 patients. Thought to leak into the CSF after cell damage and death, AST and LDH were elevated in the CSF of late-infantile patients, but not in most juvenile patients, suggesting that cellular “leakiness” is a function of disease severity.25, 26, 27, 28, 29 This hypothesis is supported by the juvenile GM1 patient GSL020, whose substantial AST and LDH elevations above other patients in the cohort (Figures 1C and 1E) occurred during a period of accelerated disease progression. In addition, high LDH levels were found in the CSF of two treated cats (Figure 1D) that had recurrent focal and generalized seizures just before humane endpoint, perhaps pointing to advanced cytotoxicity either preceding or as a result of seizure activity. CSF LDH was a good indicator of disease progression only in animals free of seizures during the period of sample collection. Including cats with breakthrough seizures in the regression analysis abolished the correlation (R² decreased from 0.64 to 0.08) (Figure S5). AST levels were unaffected by seizure activity. Overall, then, elevations of AST and LDH in the CSF of untreated GM1 cats were significantly reduced after AAV gene therapy at both short-term (16 weeks) and long-term follow-up, suggesting their utility as indicators of efficacy in human clinical trials.

In peripheral blood samples from cats and humans, AST levels correlated with disease severity but LDH was unchanged. Thus, emphasis was placed on AST as a potential biomarker for human clinical application (Figures 2A–2D). Further analysis supported the use of serum AST as a biomarker because it (1) was elevated in the peripheral blood of infantile and late-infantile patients, (2) increased in any given GM1 cat or infantile patient over time, (3) was reduced to normal levels in the peripheral blood of GM1 cats after AAV treatment, and (4) followed similar trends as AST in the CSF of cats. However, serum AST was not elevated in juvenile-onset patients, and though mean values were above normal in late-infantile patients overall, levels decreased in individual patients over time. Even so, serum AST may be a valuable marker in an appropriate subpopulation of GM1 patients, especially if CSF samples are unavailable. Other biochemical parameters (ALT, serum albumin, and urinary GAGs) were altered in both patients and GM1 cats, and GM1+AAV cats responded partially. Due to the non-specific nature of these peripheral markers, the origin of their perturbations remains unknown. However, from a clinical perspective, alterations in blood parameters may represent hepatosplenomegaly, muscle atrophy, and/or progressive cachexia, which are well-known sequelae of GM1 gangliosidosis and other LSDs.1, 30, 31, 32, 33, 34 Moreover, the apparent effect of intracranial AAV gene therapy on markers that likely reflect peripheral disease may be due to leakage of vector to the periphery and enzyme expression in liver.¹⁹

EEG alterations are found in most metabolic disorders with a neurologic component, because the EEG reflects complex cerebral events. Mild EEG changes in early-stage GM1 gangliosidosis patients are reported to deteriorate suddenly,35, 36 perhaps after reaching a critical threshold of storage material or pathology.37, 38 In the current work, slowing of rhythmic activity was found in both the moderately diseased GM1 patient and the untreated GM1 cat. Three to five years after AAV gene therapy, rhythmic slowing was largely corrected in GM1 cats, but sharps and spikes indicative of epileptic events were apparent. Therefore, it is possible that ectopic neurites and associated synapses already present at the time of AAV injection (2–3 months of age) persist throughout life, causing abnormal EEG sharps or spikes in treated GM1 cats. In addition, though storage is cleared from most areas of the treated cat brain, it persists in focal sites due to incomplete distribution of β-gal,¹⁹ and advanced pathology in such areas may contribute to seizure activity.

Ultra-high-field MRI allowed visualization of previously indiscernible structures and detailed regional evaluation after AAV gene therapy. A 7 T MRI revealed preservation of brain architecture 3–5 years after AAV gene therapy, as would be expected from the clear clinical benefit. Relative intensities of gray and white matter remained within the normal range in most areas, and cortical atrophy was minimal after gene therapy. MRI pathology has been reported in untreated GM1 human patients,39, 40, 41 and MRI scoring, as well as degree of brain atrophy, worsens over time and correlates with the decline in clinical metrics, language, and ambulation.³⁹ Therefore, anatomical MRI should be a valuable tool to track CNS disease progression after gene therapy.

MRS has been used to evaluate human GM1 gangliosidosis at lower field strengths41, 42 and most recently at 3 T.³⁹ Because our clinical rating scale for GM1 cats was based largely on symptoms that involve the cerebellum, correlation between clinical scoring and cerebellar MRS was encouraging. Also noteworthy was the reduction of NAA in untreated GM1 cats early in the disease process (4 months of age) and in an asymptomatic GM1 cat years after gene therapy (9-1424). Reductions of NAA have been reported in both juvenile and late-infantile GM1 patients before symptom onset, and NAA reduction correlated with clinical scoring metrics, with late-infantile patients declining 10 times faster than juvenile patients.³⁹ The demyelination marker GPC+PCh was significantly higher in the cerebellum of GM1 cats at 4 months of age and became elevated in all brain voxels by the humane endpoint of 8 months, correlating well with gray-to-white matter changes indicative of myelin abnormalities on MRIs. With gene therapy, all brain voxels except the parietal cortex had normal levels of GPC+PCh, and gray-to-white matter intensities were largely preserved. Reductions in Glu+Gln have been reported in several neurodegenerative disorders, including Alzheimer’s disease,43, 44 multiple sclerosis,⁴⁵ amyotrophic lateral sclerosis,⁴⁶ and frontotemporal dementia,⁴⁷ but this is the first report of Glu+Gln reduction in a LSD. Lin et al.⁴⁴ have suggested that glutamate reductions are secondary to neuronal loss—possibly due to excitotoxicity⁴⁸—or to a generalized defect in the neurotransmitter pool,⁴⁶ with high GM1 concentrations inhibiting glutamate release.⁴⁹ When AAV-treated cats reach humane endpoint, postmortem samples will provide a unique opportunity to evaluate MRS measurements and the histopathologic correlates that they are reported to represent.

While numerous candidate biomarkers were uncovered in this work, a few stood out as most promising for future clinical trials. Measured directly in the brains of living subjects, NAA and GPC+PCh were the most commonly elevated metabolites and were normalized by gene therapy. Regression analyses demonstrated convincing correlations between metabolite levels and clinical status across all cat cohorts (Figure S4). In addition, CSF levels of LDH (in the absence of seizures) and AST (regardless of seizure status) were robust, were easily measured in clinical samples, and correlated well with CNS disease status in untreated and treated GM1 cats (Figure S5). AST and LDH are elevated in patients with gangliosidosis, other storage diseases, and CNS disorders such as stroke,22, 25, 26, 27 further supporting their utility in GM1 clinical trials. The striking efficacy of AAV gene therapy for GM1 gangliosidosis in animal models, including dramatic improvements in quality of life and ≥6-fold increases in survival, suggests that such therapy may be effective in GM1 patients. Although tracking of disease progression through clinical assessment will remain critically important, sensitive biomarkers are important tools for translation of gene therapy into the clinic.

Materials and Methods

Animals and Surgery

All animal procedures were approved by the Auburn University Institutional Animal Care and Use Committee. Bilateral injection of the thalamus and deep cerebellar nuclei were performed as previously described.¹⁹ GM1 cats were treated at 1.3–3 months of age, before symptom onset. Clinical assessments were made bimonthly on all cats and performed by a veterinarian, and video footage (Movies S1 and S2) was retrospectively analyzed by a second, independent investigator. Clinical rating scores were based on the following clinical signs, with a normal score of 10 points and subtraction of 1 point for each symptom acquired: slight tremors, overt tremors, hindlimb weakness, wide-based stance, ataxia, occasional falling, limited ambulation, spastic front legs, spastic hind legs, and inability to ambulate. With the personnel available, it was not possible for evaluators to be completely blinded to treatment status, although their observations were performed independently.

AAV Vectors

AAV vectors were produced as previously described.19, 50 Transgene expression was promoted by a hybrid element consisting of the cytomegalovirus (CMV) enhancer and chicken beta actin promoter (CBA). The woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) was included.

Cat Blood and Cerebrospinal Fluid Analysis

Cats were sedated with dexmedetomidine, and blood was collected at predetermined time points. Complete blood counts were performed on blood containing EDTA as an anticoagulant using an ADVIA 120 Hematology System (Siemens Medical Solutions). Serology was performed on plasma samples (heparinized) using a Cobas C311 chemistry analyzer (Roche Hitachi). For quantification of echinocytes, five 40× photomicrographs were taken of Diff-Quick-stained blood smears.

CSF was collected from the cerebromedullary cistern while under general anesthesia using dexmedetomidine and ketamine and maintained using isoflurane. Samples were immediately centrifuged at 100× g for 5 min to remove any blood contamination. All samples underwent one freeze-thaw cycle (−80°C) and were analyzed using a Cobas C311 chemistry analyzer (Roche Hitachi).

Human Blood and Cerebrospinal Fluid Analysis

Blood and CSF were collected for clinical and research indications from 21 patients with GM1 gangliosidosis under National Human Genome Research Institute protocol 02-HG-0107 with parental consent. Control CSF was obtained from banked samples from children whose CSF was collected with for indications other than LSDs. Human blood parameters were normalized to published mean normal values for the appropriate age range to account for developmental-related change of parameters,⁵¹ and normal levels were set to 1 (represented as a yellow line in figures).

Cat Electroencephalography

Cats were sedated with dexmedetomidine, and electroencephalography was performed using a Neurofax 1200 series EEG (Nihon Kohden). Electrode placement was modified from the International Federation in Electroencephalography and Clinical Neurophysiology’s standardized 10-20 electrode placement system.⁵² Electrodes were placed under the skin above the right and left frontal, central, and occipital brain areas, with the reference electrode (Cz) placed on the vertex of the skull. Only impedances less than 5 KOhms were accepted, and the electrodes were referenced to Cz and by hemisphere. The ground electrode was placed on the left thorax adjacent to the heart.

Human Electroencephalography

A 21-channel digital EEG with time-locked video and single-lead electrocardiography (ECG) was performed for clinical and research indications on 21 patients with GM1 gangliosidosis under National Human Genome Research Institute protocol 02-HG-0107 with parental consent. The electrodes were placed according to the international 10-20 system of electrode placement.⁵² The EEG was reviewed using at least three montages: bipolar, referential, and transverse.

MRI and MRS

Cats were anesthetized as described earlier for CSF collection. MRI and MRS data were acquired on a 7 T MAGNETOM scanner (Siemens Healthcare) at 4 and 8 months for untreated GM1 cats (n = 4 and n = 6, respectively); 4 months, 8 months, and 2–5 years for normal controls (n = 4 per time point); and 3–5 years for the GM1+AAVrh8 group (n = 3). A 32-channel head coil (Nova Medical) was used for all scans. Anatomical coronal images were acquired using 3D MPRAGE (magnetization-prepared rapid gradient echo) with 0.5 mm isotropic resolution and TR/TE of 1,910/2.5 ms, followed by 2D axial T2 turbo spin echo (TSE) images with TR/TE of 5,450/12 ms and a resolution of 0.25 × 0.25 × 1 mm³. Single voxel spectroscopy (SVS) was then acquired using a PRESS (point resolved spectroscopy) sequence optimized for 7 T with TE/TR = 30/5,000 ms, 64 averages, and variable pulse power and optimized relaxation delays (VAPOR) water suppression. Shimming was performed using FASTESTMAP⁵³ or a glucocorticoid response element (GRE) shim sequence (Siemens Healthcare), followed by manual shimming if needed. The resulting full width at half maximum (FWHM) of the unsuppressed water peak was typically 16 Hz. Optimization of RF pulse amplitudes and of the water suppression scheme was performed before acquiring each spectrum. The unsuppressed water signal was obtained and used for eddy current correction and for quantification of metabolites. Using high-resolution 3D MRI images, voxels were positioned in the thalamus (7 × 6 × 8 mm), corona radiata (7 × 5 × 8 mm), parietal cortex (7 × 6 × 8 mm), temporal lobe (7 × 6 × 8 mm), occipital cortex (6 × 6 × 5 mm), and cerebellum (7 × 7 × 8 mm). MRI data were analyzed with EFilm 3.2 software (Merge Healthcare). MRS data were processed with the LC model and internal water scaling (http://www.s-provencher.com/lcmodel.shtml).

Urinary GAGs

Urinary GAGs were measured as previously described⁵⁴ in normal, GM1 cats at humane endpoint (∼8 months), and GM1 cats after AAV gene therapy (AAVrh8, n = 2; AAV1, n = 2).

Statistics

Statistical analyses were performed using Microsoft Excel and statistical analysis software (SAS; SAS Institute). Two-sided, paired Student’s t test assuming unequal variances was used for statistical comparisons between groups. The values p < 0.05 and p < 0.01 are indicated from normal (single and double asterisks, respectively) and untreated GM1 cats (single-crossed and double-crossed lines, respectively). For MRS, only spectra with a Cramér-Rao minimum variance bounds of <30 were included in analysis. Regression analyses (R²) were used to correlate clinical signs with metabolite concentrations. It was not possible to perform an independent validation of biomarkers, because no other source of GM1 cats exists and the numbers of human GM1 patients were small.

Study Approval

All animal studies were performed in accordance with the Auburn University Institutional Animal Care and Use Committee. Human studies were performed in accordance with National Human Genome Research Institute protocol 02-HG-0107 with parental written informed consent.

Author Contributions

H.L.G.-E. was the primary author on the manuscript and performed most cat data analyses and interpretation. D.S.R., S.E.T., Y.L.L., J.J., G.G., and C.J.T. (P.I.) enrolled GM1 patients and performed, analyzed, and interpreted human CSF, blood, and electrodiagnostic data. J.L.S. analyzed cat blood data. A.N.R. and A.K.J. performed anesthesia and cat surgeries and collected and analyzed body fluid samples. A.S.M. performed echinocyte counts and participated in their interpretation. P.W.C. assisted in the interpretation of blood-based biomarkers. N.S., R.J.B., and T.S.D. assisted in MRI and MRS data collection and analysis. A.R.T. and D.C.S. analyzed cat EEGs. V.J.M., A.M.B., B.L.B., N.R.C., and H.J.B. assisted in study design, biomarker development, and evaluation of therapeutic benefit. M.S.-E. designed and produced the AAV vectors for this project and participated in study design. D.R.M. was responsible for overall study design and data interpretation, managed all research studies described herein, and assisted in authorship.

Conflicts of Interest

N.S. is an employee of Siemens Healthcare. D.R.M. and M.S.-E. are shareholders in Lysogene (Neuilly-sur-Seine, France).