Gm2a^−/− mice (Liu et al., 1997) were purchased from Jackson Laboratories (Maine, United States). Neu3^−/− (Yamaguchi et al., 2012) mice obtained from the Centre Hospitalier Universitaire Sainte-Justine Research Centre (Montreal, Quebec, Canada) with a kind permission from Professor T. Miyagi, who generated this strain (Yamaguchi et al., 2012). Mice were bred, housed, and maintained at the Animal Care Facility at Queen’s University (Kingston, Ontario, Canada) as previously described (Liu et al., 1997; Yamaguchi et al., 2012). All experimental procedures were approved by the institutional Animal Care Committee in accordance with the Canadian Council for Animal Care.

Gm2a^−/−Neu3^−/− colonies were generated by crossing single knockout mouse models (Neu3^−/− × Gm2a^−/−) to produce F1 generation Gm2a/Neu3 heterozygotes. Heterozygotes (Gm2a^+/− Neu3^+/−) were crossed to produce F2 double knockout progeny (Gm2a^−/-Neu3^−/−) in proportions consistent with Mendel’s Law of Independent Assortment.

Genotyping was performed on DNA extracted from ear notches collected from the mice at or before 21 days of age. DNA digestion was carried out using Q5^® High-Fidelity 2X Master Mix (New England BioLabs Ltd., Ontario, Canada). Samples were then prepared for polymerase chain reactions (PCR). The following primers were used to detect Gm2a and Neu3:

A series of behavioral tests were conducted to characterize potential physical impairments in double knockout mice. Testing began at 8 weeks of age and continued every 4 weeks, until 16 weeks of age. After this period, testing was conducted bi-weekly until endpoint. Frequency of behavioral testing was increased to weekly, starting at 16 weeks of age, to determine the rate of behavioral decline. Two behavioral parameters were tested. These included an Open Field Test (OFT, ActiMot, TSE systems, Berlin, Germany) to evaluate gross locomotor activity (Osmon et al., 2018), and a RotaRod test (RR; IITC Life Sciences, California, United States) to evaluate motor coordination and balance (Osmon et al., 2018).

For the OFT, mice were placed in a bordered arena (40 cm × 40 cm) for 5 min and were allowed to roam freely. Resting time, distance traveled, and speed during this time was recorded (Osmon et al., 2018).

For the RR, mice were placed on elevated rotating cylinders that accelerated from 4 to 40 rotations per minute (rpm) over 5 min. When the mouse fell, it off balanced a beam below them, which signaled the end of the test. Each mouse was tested on the RR apparatus three times per time point, with a minimum of 10 min of rest between each trial. Latency to fall, end rotations per minute (rpm) and distance traveled were recorded (Osmon et al., 2018).

Gm2a^−/-Neu3^−/− mice were euthanized at their humane endpoint and all other cohorts were euthanized between 27 and 30 weeks of age. A humane endpoint is defined as the point at which a mouse loses 15% of its peak weight, or when a mouse is unable to right itself (which indicates a reduced quality of life). Mice are sacrificed by CO₂ asphyxiation according to the Animal Care Committee guidelines. Organs were perfused with phosphate-buffered saline (PBS) and collected thereafter, for future biochemical and histological analysis.

Cardiac puncture on euthanized mice was used to collect blood to assess serum protein levels. Organ perfusion with PBS was carried out following cardiac puncture. Gross organs and CNS tissue were collected for Western blot analysis, ganglioside accumulation assays and histological imaging. Regions of the brain and spinal cord dissected for tissue collection and sectioning are shown in Supplementary Figure 1.

Protein extracts from murine livers were separated by SDS-PAGE and transferred to a nitrocellulose membrane (Bio-Rad Laboratories, California, United States). Non-specific binding sites were blocked with a 5% skim milk tris-buffered saline (TBS) solution for 1 h. Membranes were then incubated with primary antibodies against GM2A (Taysha Gene Therapies, Texas, United States; TSHA081821; 1:1000), α-subunit of β-HEXA (Abcam, Cambridge, United Kingdom; ab189865; 1:1000), β-subunits of β-HEXA (Thermo Fisher Scientific, Massachusetts, United States; 16229-1-AP; 1:1000) or β-actin (Sigma-Aldrich, Missouri, United States; A5316; 1:5000) overnight in 5% skim milk. Membranes were then washed three times in TBS-Tween-20 (TBST) for 10 min prior to incubation with secondary antibody for 1 h at room temperature. Secondary antibodies included rabbit anti-Goat-HRP (Sigma-Aldrich, Missouri, United States; AP106P; 1:5000) for GM2A, and goat anti-mouse-HRP (Sigma-Aldrich, Missouri, United States; A5278; 1:5000) for β-actin. Membranes were washed again in TBST and then incubated with a horse-radish-peroxidase solution (Millipore Sigma, Massachusetts, United States). Membranes were visualized by chemiluminescence (Bio-Rad Laboratories, California, United States). GM2A and β-actin were quantified through densitometry using IMAGE J Software (National Institute of Health, Maryland, United States).

A modified Folch extraction was used as described in Osmon et al. (2018). Briefly, mid-sections of the brain were taken to be representative of whole murine brain. Each mid-section was sonicated and transferred to a glass tube (UltiDent Scientific Inc., Quebec, Canada). Next, a series of methanol and chloroform dilutions were added to the sample and then flowed through a column (Phenomenex, California, United States) to collect gangliosides. Samples were eluted and added to a thin layer chromatography plate (TLC; MilliporeSigma, Massachusetts, United States) for development. The plate was placed in a tank containing chloroform, methanol, and calcium chloride. The plate was developed in the tank for several hours to allow individual gangliosides to separate, and then sprayed with a drying solution made up of orcinol and sulfuric acid. Individual gangliosides were quantified by densitometry by comparing the intensity of the GM2 bands against an internal control (GD1A) using IMAGE J software (National Institute of Health, Maryland, United States). A TLC Monosialoganglioside mixture (MJS Biolynx Inc., Ontario, Canada) and Gm2a^−/− mid-section samples were used as controls. GD1A was used as an internal control as it is a constant presence in mice, throughout adulthood (Yu et al., 2012).

β-HEXA activity assays were performed as previously described (Tropak et al., 2004, 2007). Briefly, serum was collected via cardiac puncture and then diluted 1:35 in citrate phosphate buffer (CP Buffer). A portion of the serum was heat inactivated (HI) at 52°C to destabilize β-HEXA, leaving only β-HEXB active in these samples. Next, the substrate 4-methylumbelliferyl-2-acetamido-2-deoxy-β-D-glucopyranoside (MUG; Toronto Research Chemicals, Ontario, Canada) was added to both normal and HI serum samples. Β-galactosidase activity was used as an internal control, which was measured using 4-methylumbelliferyl-β-D-galactopyranoside (MUGal) (MilliporeSigma, Massachusetts, United States).

MUG and MUGal activity were assessed by incubating serum samples at 37°C in a 3.2 mM MUG or MUGal substrate solution. After 1 h, enzyme activity was stopped with 0.1 M 2-amino-2-methyl-1-propanol (AMP). MUG and MUGal activity were estimated based on a 4-methylumbelliferone (4-MU) standard curve, which ranged from 0.73 nM to 130 μM. Activity was detected by colorimetric intensity at an excitation wavelength of 365 nm and an emission wavelength of 450 nm on a Varioskan^™ microplate reader (Thermo Fisher Scientific, Massachusetts, United States). HexA activity was calculated by subtracting HI sample activity (representative of HexB activity) from total MUG activity (all hexosaminidases), after being normalized to MUGal concentrations.

Portions of the mid- and caudal-sections of the brain were fixed in 4% paraformaldehyde (PFA) for 24–48 h and then transferred to 70% ethanol for another 24–48 h. Samples were then transferred to PBS and shipped to The Center for Phenogenomics (Toronto, Ontario, Canada) for slide preparation and staining with hematoxylin and eosin, or with an anti-GM2 antibody (a gift from Kyowa Hakko Kirin Co. Ltd., Tokyo, Japan). Ganglioside accumulation was quantified using QuPath Software (Belfast, Ireland).

All statistical analyses were performed with GraphPad PRISM 9.3.1. Statistical testing for all behavioral modalities were performed using two-way repeated measures analysis of variance (ANOVA) with Tukey post hoc test to compare cohorts across all timepoints. Ganglioside accumulation and hexosaminidase activity were analyzed by a one-way ANOVA with a Tukey post hoc test for multiple pairwise comparisons. A Kaplan Meier curve was used to present the survival patterns of each cohort, which was analyzed with a Log-Rank (Mantel-Cox) test to assess significant survival differences. A nonlinear regression (curve fit) was used to determine if there was a correlation between symptom onset and age of death.

A Western blot was performed to confirm the absence of GM2A protein and to determine the effect of knocking out Neu3 on the expression of the α- and β-subunits in the brain (Figure 2A). As expected, GM2A protein was absent in Gm2a^−/− and Gm2a^−/-Neu3^−/− mice and unaffected in wild-type and Neu3^−/− mice (Figures 2A,B). Expression of the α-subunit was not impacted by knocking out Neu3 in Gm2a^−/− mice, and there were relatively equal levels of the α-subunit between all cohorts (Figure 2C). Expression of the β-subunit in Gm2a^−/-Neu3^−/− mice was significantly lower than wild-type mouse brains (Figure 2D). A similar trend was seen in Gm2a^−/− mice, which had reduced expression of the β-subunit compared to wild-type and Neu3^−/− mice (Figure 2D). To determine the correlation between expression of each subunit and β-HEXA activity, a hexosaminidase activity assay was conducted on mouse serum (Figure 2E). Here, we saw no significant differences in β-HEXA activity in any of the cohorts, regardless of the slightly altered expression levels in the subunits observed in some of the cohorts.

Gm2a^−/−Neu3^−/− mice were generated by interbreeding singly knocked out Neu3 (Yamaguchi et al., 2012) and Gm2a (Liu et al., 1997) mice. Kncokout of Gm2a and Neu3 was confirmed by PCR (Supplementary Figures 2A,B). As discussed above, Gm2a^−/− mice exhibit mild neurological and behavioral phenotypes, including normal lifespans (Liu et al., 1997); whereas Neu3^−/− mice are relatively asymptomatic with respect to neurological and behavioral phenotypes (Seyrantepe et al., 2010, 2018). Consistent with previous reports (Liu et al., 1997), we confirm that Gm2a^−/− mice have a normal life span (Deschenes et al., 2023). However, Gm2a^−/−Neu3^−/− mice had significantly shortened lifespans compared to Gm2a^−/− counterparts (Figure 3A). Double knockout mice lived an average of 27.0 ± 1.5 (mean ± SD) weeks, compared to Gm2a^−/− mice, which lived a relatively normal lifespan that averaged 91.9 ± 9.8 weeks. These results demonstrate that Neu3 gene in Gm2a gene deficient backgrounds reduce lifespans to an extent that may be representative of a more severe form of GM2 gangliosidosis, than the previously utilized animal model (Gm2a^−/− mice).

The newly developed mouse strain exhibited a significantly reduced average lifespan of 27 ± 1.5 weeks (range from 24.43 to 30.0 weeks), compared to Gm2a^−/− mice (91.9 ± 9.8 weeks) (Figure 3A). Gm2a^−/-Neu3^−/− mice exhibited a range of physical impairments indicative of disease manifestation (Supplementary Figure 3B). A positive correlation between the onset of symptoms of disease manifestation and age of death was observed (Supplementary Figure 3A). Depending on symptom severity, Gm2a^−/−Neu3^−/− mice lived, on average, for 1.5–3 weeks after the appearance of the first signs of the disease. Moreover, the earlier the symptoms appeared, the earlier the humane endpoint was reached (see Methods: Euthanizations). The predominant symptoms indicative of disease included tremulous movement and ataxic gait (Supplementary Figure 3B); these began as early as 21 weeks of age in some mice, while others were relatively asymptomatic until 28 weeks of age. In addition, Gm2a^−/−Neu3^−/− mice began to lose weight around 20–22 weeks of age with at least 66.7% of them losing over 15% of their peak body weight at humane endpoint (Figures 3B,C). Other symptoms included seizures (4.2%), eye infections (16.7%), and death (8.3%) although these were of substantially less predominance (Supplementary Figure 3B). Four Gm2a^−/−Neu3^−/− mice experienced hindlimb paralysis around 9–12 weeks of age; however, they have not been included in any of the analyses as we were unable to confirm the cause. Although, this is a classic symptom observed in the SD mouse model (Phaneuf et al., 1996).

Coordination and balance were assessed on a RR apparatus which measures the time that elapses before a mouse falls off an elevated rotating cylinder. Three RR parameters were assessed, including latency to fall, distance traveled, and end RPM. All cohorts tested on this apparatus performed comparably at approximately 8 weeks of age but started to deviate at around 12 weeks of age (Figures 4A,E,I). After 12 weeks of age, Gm2a^−/−Neu3^−/− mice exhibited a downward trend in performance for all three RR parameters assessed. These trends became significant for all 3 parameters by 20–24 weeks of age and remained significant at later time points (Figures 4B–D,F). By 24 weeks of age, most of the Gm2a^−/−Neu3^−/− mice were incapable of staying on the rotating cylinder at all. This trend was consistent across all other RR parameters. This progressive decrease in performance on RR with the Gm2a^−/-Neu3^−/− mice was not observed in other cohorts, suggesting that knocking out Neu3 in Gm2a^−/− mice imparts coordination and balance defects that are absent in Gm2a^−/−. In addition, Gm2a^−/− mice appear to perform similarly to the wild-type cohort, which opposes a previous paper that characterizes this model (Liu et al., 1997).

Disparities in motor function were much more apparent between cohorts when evaluated by the OFT, with differences in the Gm2a^−/−Neu3^−/− mice beginning as early as 8 weeks of age (Figures 5A,E,I). For the OFT parameters evaluated—mean speed, distance traveled, and resting time—the performance of Gm2a^−/− and Neu3^−/− mice were similar, indicating that neither single gene deficiency by itself significantly impacts motor function. Remarkably however, Gm2a^−/−Neu3^−/− mice exhibited significantly reduced scores for the mean speed and distance traveled parameters of the OFT when compared to all the cohorts (Figures 5B–D,F). Correspondingly, Gm2a^−/−Neu3^−/− mice resting time was significantly longer than the other cohorts (5B-5D). These differences became significant at 8–12 weeks of age and persisted through all time points. Furthermore, motor impairments in Gm2a^−/−Neu3^−/− mice grew progressively worse as indicated by their OFT motor scores, which increasingly diverged from the scores of other cohorts over time (Figures 5A,E,I). These data clearly demonstrate that double knockout mice have impaired gross locomotor activity beginning very early in life.

GM2 is a breakdown product of GM1 in the ganglioside catabolic pathway, which then continues to be metabolized in a stepwise fashion into glucosylceramide (GlcCer). This process involves the intermediary products GM3 and LacCer (Figure 1). Importantly, a branching pathway from GM2 to GA2 and then LacCer has also been identified as an alternative catabolic route for GM2 breakdown (Sango et al., 1995; Figure 1). Hence, GlcCer is a key convergence point for two GM2 breakdown pathways and its levels are an important marker of total GM2 breakdown activity.

GM1, GM2, GM3 and GlcCer were extracted and quantified using TLC. Quantitative analysis of glycolipids was carried out on the mid-section of brains, which are taken as representative of the whole murine brain. The brains were collected at humane endpoints for Gm2a^−/−Neu3^−/− mice and at 26–30 weeks of age for all other cohorts. Interestingly, we were unable to detect GM2 accumulation in Neu3^−/− brain tissue indicating that GM2 catabolic efficiency is unimpacted by knocking out the NEU3-mediated route of GM2 breakdown (Figures 1, 6A), likely because GM2 is more efficiently catabolized by β-HEXA/GM2A-mediated pathway. These results were consistent with a previous study (Seyrantepe et al., 2018). In contrast, appreciable GM2 accumulation was detected in the brains of Gm2a^−/− mice, suggesting that GM2 breakdown is impaired, and confirms that GM2A co-factor activity for the β-HEXA enzyme is involved in GM2 degradation. However, knocking out Neu3 in Gm2a^−/− backgrounds resulted in a striking increase (approximately 3- to 4-fold) in GM2 levels relative to Gm2a^−/− mice. This demonstrates that significantly less GM2 is degraded in Gm2a^−/-Neu3^−/− mice than in Gm2a^−/− mice (Figure 6A). This also indicates that the NEU3-mediated pathway may be compensating for impaired GM2 breakdown in Gm2a^−/− mice, and that in the absence of this compensation, GM2 accumulation is significantly higher. The levels of other gangliosides quantified remain unchanged in all cohorts (Figures 6B,C). Importantly, the levels of GM2 accumulation in the double knockouts parallel the severity of GM2 accumulation observed in a bona fide model of infantile SD cohort of age-matched Hexb^−/− mice (Figure 6A).

In a separate experiment, the progression of GM2 accumulation in double knockout mice was also assessed over five different time points: 8, 12, 16, 20 and 26 weeks of age (Figure 6D). There were no significant differences seen between any of the cohorts at any of the timepoints. It is possible that this experiment was not sufficiently powered to see differences in ganglioside levels throughout their lifetime.

Histological analysis of murine brain sections supported our GM2 accumulation findings (Figure 7). The cortex (Figure 7A), cerebellum (Figure 7B) and pons (Figure 7C) in Gm2a^−/−Neu3^−/− mice display large accumulations of GM2 with no buildup visible in other cohorts. Accumulation in the cortex and pons is uniform across the entire section, whereas GM2 is almost exclusively built up in the white matter of the cerebellum (Figure 7). While GM2 is more abundant in the gray matter, it also has a major role in the development of myelin (Kraĉun et al., 1984; Sipione et al., 2020). Vacuolization is also visible in all histological brain sections of Gm2a^−/−Neu3^−/− mice but is most notable in the cerebrum and pons. These vacuoles appear to be larger and more abundant in all knockout cohorts compared to wild types.