Viral vector-mediated SLC9A6 gene replacement reduces cerebellar motor and molecular abnormalities in the shaker rat model of Christianson syndrome

Abstract

Background

Christianson syndrome is an x-linked recessive neurodevelopmental and neurodegenerative condition caused by mutations to the SLC9A6 gene encoding NHE6, a sodium-hydrogen exchanger critical for regulating endosomal pH. Using an adeno-associated viral (AAV) vector targeting Purkinje cells (PHP.eB-L7-Slc9a6-GFP), we recently performed functional complementation studies to demonstrate that mutation to the rat Slc9a6 gene causes a Christianson syndrome-relevant cerebellar phenotype in the shaker rat.

Methods

We carried out a longitudinal study evaluating the impact of gene replacement targeting Purkinje cells on ataxia and tremor in the shaker rat. Further, in a smaller follow-up study, we tested administration of AAV9-CAG-hSLC9A6 to determine whether key molecular and motor findings could be replicated with a more clinically relevant viral construct. In both experimental cohorts, we performed molecular studies to evaluate expression of NHE6 and key cerebellar markers.

Results

Administration of either of PHP.eB-L7-Slc9a6-GFP or AAV9-CAG-hSLC9A6 AAV vectors led to significant improvement in both the molecular and motor phenotypes. The abundance of each disease-relevant cerebellar proteins was strongly correlated to motor ataxia, but less so to tremor. Further, while ataxia and tremor were initially strongly correlated early in the disease progression, this relationship weakened over time.

Conclusions

These findings impact future SLC9A6-targeted gene therapy efforts for Christianson syndrome and support gene replacement as a potentially viable therapeutic strategy. Common markers associated with cerebellar degeneration are much more strongly tied to ataxia than to tremor, indicating that ataxia may be more tied related to degenerative processes than tremor.

Graphical Abstract

Graphical Abstract.

Introduction

Christianson syndrome is an X-linked recessive disorder that presents with developmental delays, intellectual disability, cerebellar degeneration, progressive ataxia, and epilepsy, among other symptoms [1]. Christianson syndrome is caused by loss of function in SLC9A6 encoding NHE6, a sodium-hydrogen exchanger that regulates pH in early and recycling endosomes [2]. NHE6 is particularly highly expressed in several brain regions and neuronal types, including cerebellar Purkinje cells, coinciding with extensive Purkinje cell degeneration outside of lobule x [3]. There are currently no treatments or cures for Christianson syndrome, necessitating therapeutic development.

We previously published on the Wistar Furth shaker rat, an x-linked, spontaneous model of cerebellar degeneration, tremor, and ataxia [4–7] and demonstrated that this phenotype was caused by a frame-shifting, early-truncating, loss-of-function mutation in Slc9a6. This established the shaker rat as a spontaneous model of the Christianson syndrome [8]. The shaker rat arose spontaneously from a breeding colony at St Louis University in the 1990s [9–12], and we have since used the model for gene discovery purposes [4, 8], as well as the development of novel deep brain stimulation approaches aimed at reducing both ataxia and tremor [5].

In establishing Slc9a6 as the gene mutated in the shaker rat, we performed functional complementation studies that employed adeno-associated viruses (AAVs) to demonstrate the causality of the identified mutation [8]. Previous work focused primarily on gene discovery and replacement and centered on ataxia, which is a primary motor symptom in Christianson syndrome. In this report, we expand on the motor outcomes described in our previous work using PHP.eB-L7-Slc9a6-GFP AAV targeted gene replacement specifically to Purkinje cells, with emphases on the progression of motor symptoms and the complex relationship between tremor and ataxia.

Clinical evaluations of tremor and ataxia have traditionally occurred in the context of natural disease progression or with symptomatic relief in the case of tremor, simply based on the lack of disease modifying treatments for either symptom or any form of treatment for ataxia [13, 14]. Thus, the preclinical modification of disease progression provides a rare opportunity to study the complex relationship between ataxia and tremor both with and without intervention. In this work, we sought to evaluate several key questions: 1. How does gene replacement targeting both Purkinje cells and more broadly impact a cerebellum-associated motor phenotype associated with Christianson syndrome?, 2. How do underlying cerebellar molecular markers covary with motor symptomology?, and 3. How do tremor and ataxia co-evolve in a model of cerebellar degeneration both with and without disease modification?

Here we show that AAV-mediated gene replacement in Purkinje cells has significant efficacy in reducing both tremor and ataxia in the shaker rat, further supporting gene replacement as a therapeutic strategy for Christianson syndrome. We found that SLC9A6 replacement was disease modifying, for both tremor and ataxia, but in differential manner, modifying the relationship between tremor and ataxia. Thus, tremor and ataxia in the shaker rat may arise from distinct cerebellar mechanisms, or non-cerebellar brain regions may differentially compensate for ataxia and tremor. Finally, we administered a translationally relevant viral construct, AAV9-CAG-hSLC9A6 not encoding GFP, to the shaker rat and again found molecular and motor benefit with broader targeting.

Results

All work was completed in Wistar Furth rats, utilizing the Wistar Furth shaker rat model of Christianson syndrome with wildtype Wistar Furth rats as controls, littermates when possible. Two separate AAV experimental blocks were carried out. First, we administered PHP.eB-L7-Slc9a6-GFP AAV prior to Purkinje cell degeneration and performed longitudinal motor analysis prior to molecular analyses. Second, we administered AAV9-CAG-hSLC9A6 at a similar timepoint and quantified motor performance at a single timepoint prior to molecular analyses. The results are combined into three sub-sections. The first pertains to the effects of AAV-mediated gene replacement in the context of Christianson syndrome-relevant motor dysfunction and molecular changes. The second pertains to the relationship between molecular markers in the cerebellum and motor function. Finally, the third pertains to the relationship between tremor and ataxia. Notably, rats reported in L7–6 studies were previously reported on [8], and some data are repurposed from prior publication for new analyses: motor data from 8, 18, and 25 weeks of age (15% of L7–6-related motor data), as well as western blotting data.

Gene replacement reduces Christianson syndrome-relevant motor dysfunction and molecular changes, whether targeted to Purkinje cells or ubiquitously

A‌AV gene replacement targeted to Purkinje cells reduces gait ataxia

In rats administered PHP.eB-L7-Slc9a6-GFP AAV (n = 8 shaker rats), control PHP.eB-L7-GFP AAV (n = 5 shaker rats, n = 5 wildtypes), or uninjected controls (n = 4 wildtypes) at an average of 35 days of age, we tracked ataxia of gait weekly from 6 to 25 weeks of age. We measured gait ataxia by comparing total distance travelled to net displacement in automatically identified, uncued rapid movements, as in previous work [5, 7, 8]. We performed an ANOVA accounting for time as a variable and found a difference across groups in the context of ataxia, comparing from 10 weeks on, when there was first an ataxia phenotype in untreated shaker rats. In a post-hoc test comparing just shaker rats administered PHP.eB-L7-Slc9a6-GFP AAV to those administered PHP.eB-L7-GFP AAV, we found a significant reduction in ataxia in shaker rats administered the therapeutic AAV (Fig. 1A). We compared shaker rats at individual time points and found that PHP.eB-L7-Slc9a6-GFP AAV reduced gait ataxia significantly at all time points at or after 10 weeks of age compared to PHP.eB-L7-GFP AAV-injected controls, except for 19 weeks of age. Further, we compared the total ataxia burden exhibited by each rat from 10 weeks of age on, finding that this was significantly reduced in shaker rats administered the PHP.eB-L7-Slc9a6-GFP AAV compared to those administered the control PHP.eB-L7-GFP AAV (Fig. 1B).

Figure 1.

PHP.eB-L7-Slc9a6-GFP AAV reduces ataxia. (A) Untreated shaker rats administered a control PHP.eB-L7-GFP AAV progressively become more ataxic, and PHP.eB-L7-Slc9a6-GFP AAV administered at 5 weeks of age significantly reduced gait ataxia. Administration of a control PHP.eB-L7-GFP AAV to wild type rats had no impact on coordination of gait. Note that motor data at weeks 8, 18, and 25 are previously reported [8]. (B) PHP.eB-L7-Slc9a6-GFP AAV decreased cumulative ataxia burden in shaker rats by 49.6%, computed as the sum of ataxia scores above 1.6, as calculated in A. (C) PHP.eB-L7-Slc9a6-GFP AAV delays onset of ataxia in the context of crossing thresholds for mild, moderate, and severe ataxia. (D) PHP.eB-L7-Slc9a6-GFP AAV decreased the maximal ataxia score experienced by shaker rats, as calculated in A.

We compared the onset time for ataxia across groups, given several thresholds. Using the metric described above, wild type rats had average ratios of approximately 1.6, while a threshold of 2.0 represents visually perceivable gait ataxia, 2.4 represents moderate ataxia, and 2.8+ represents severe ataxia. Notably, given that some rats never crossed some or all the ataxia thresholds, we compared the final time points at which an animal hadn’t yet crossed the threshold level of ataxia as a proxy for onset, with a maximum value of 26 weeks, given a final time point of 25 weeks. With ataxia thresholds of each of 2.0, 2.4, and 2.8, we found that PHP.eB-L7-Slc9a6-GFP AAV delayed the onset of ataxia in shaker rats (Fig. 1C). Finally, we evaluated the peak ataxia exhibited by each rat across all time points, finding that PHP.eB-L7-Slc9a6-GFP AAV decreased peak ataxia significantly (Fig. 1D). For all measures in Fig. 1, both untreated and treated shaker rats were significantly more symptomatic than wild type rats, and no differences were found between uninjected wild type rats and those administered the control AAV.

A‌AV gene replacement targeted to Purkinje cells reduces tremor

We tracked tremor in the above animals from 6–25 weeks of age. We performed a repeated measures ANOVA accounting for time as a variable from 12 weeks on, when there was first a significant tremor phenotype and found a difference among groups in the context of tremor. In a post-hoc test comparing just shaker rats administered PHP.eB-L7-Slc9a6-GFP AAV to those administered PHP.eB-L7-GFP AAV, we found a significant reduction in tremor in shaker rats administered the therapeutic AAV (Fig. 2A). Transduction of Purkinje cells with PHP.eB-L7-Slc9a6-GFP AAV yielded numerally lower tremor at all time points 12 weeks of age and on except for the final time point, 25 weeks of age; however, many time points did not individually yield significant differences. The total tremor burden exhibited by each rat from 10 weeks of age on was significantly reduced in shaker rats administered the PHP.eB-L7-Slc9a6-GFP AAV compared to those administered the control AAV (Fig. 2B).

Figure 2.

PHP.eB-L7-Slc9a6-GFP AAV reduces tremor. (A) Untreated shaker rats administered a control PHP.eB-L7-GFP AAV display substantial tremor, significantly reduced by the administration of PHP.eB-L7-Slc9a6-GFP at 5 weeks of age. Further, the administration of a control PHP.eB-L7-GFP AAV to wild type rats had no impact in the context of tremor. As in Fig. 1A, note that motor data at weeks 8, 18, and 25 are previously reported [8]. (B) PHP.eB-L7-GFP AAV decreased the cumulative tremor burden in shaker rats by 39.7%. (C) PHP.eB-L7-Slc9a6-GFP AAV delayed onset of tremor in the context of crossing thresholds for mild, moderate, and severe tremor. (D) PHP.eB-L7-Slc9a6-GFP AAV decreased the maximal tremor experienced by shaker rats. Notably, as described in the methods, a tremor score of 0.0 represents a typical maximum amount of tremor quantified in wild type rats; therefore, some scores slightly below 0.0 are expected.

We compared the onset time for tremor across groups, given several thresholds. Our normalized numerical scale was designed so that a value of 0 represented no tremor (the maximum 5–8 Hz spectral power a wild type rat would exhibit) while a value of 1 represented the spectral power corresponding to what was qualitatively identified as severe tremor in previous work [5]. Notably, given that some rats never crossed some or all the tremor thresholds, we compared the final time point at which an animal hadn’t yet crossed the threshold level of tremor as a proxy for onset, with a maximum value of 26 weeks, given a final time point of 25 weeks. With a tremor threshold of each of 0.1, 0.25, and 0.5, we found that PHP.eB-L7-Slc9a6-GFP AAV delayed the onset of tremor in shaker rats (Fig. 2C). Finally, we evaluated the peak tremor exhibited by each rat across all timepoints, finding that within shaker rats, PHP.eB-L7-Slc9a6-GFP AAV decreased peak tremor significantly (Fig. 2D). For all measures in Fig. 2, both untreated and treated shaker rats were significantly more symptomatic than wild type rats, and no differences were found between uninjected wild type rats and those administered the control AAV.

A‌AV gene replacement with ubiquitous targeting reduces gait ataxia and molecular markers of cerebellar health

Following the above L7 AAV experiments, we began work with a more translationally relevant viral construct. Compared to the PHP.eB.L7-Slc9a6-GFP AAV, the AAV9-CAG-hSLC9A6 AAV has 4 primary differences: removal of the GFP tag, replacement of the rat Slc9a6 gene with the human SLC9A6 gene, the use of the ubiquitously expressing CAG promoter instead of the Purkinje cell-specific L7–6 promoter, and the use of the more clinically relevant AAV9 instead of PHP.eB. We administered varying doses of AAV9-CAG-hSLC9A6 AAV—ranging from 0.8 to 30 × 10^10 vg—at an average of 34 days of age and measured ataxia and tremor at 14 weeks of age before immediately collecting cerebellar tissue for molecular analyses (n = 23 shaker rats). We found that cerebellar NHE6 expression was significantly increased in rats administered AAV9-CAG-hSLC9A6 AAV (Fig. 3A and C), as was cerebellar CALB1 expression (Fig. 3B and D). AAV9-CAG-hSLC9A6 AAV administration yielded a reduction in ataxia (Fig. 3E). However, while AAV-CAG-hSLC9A6 AAV administration trended towards reduced tremor, this was not significant (Fig. 3F), matching the wide variability in tremor at similar time points in the L7 experiments. Note that molecular outcomes from the above L7 experiments matching those in Fig. 3A-D are not reported as they have previously been published [8].

Figure 3.

AAV9-CAG-hSLC9A6 AAV increased cerebellar NHE6 and CALB1 abundances determined by western blotting and decreased motor ataxia. (A-D) Mirroring findings with PHP.eB-L7-Slc9a6-GFP AAV, administration of AAV9-CAG-hSLC9A6 AAV at ~ 5 weeks of age minimally increased NHE6 expression (A, C), but this minimal increase led to proportionally greater increases in CALB1 expression (B, D). (E) AAV9-CAG-hSLC9A6 AAV administration at 5 weeks of age significantly improved motor ataxia at 14 weeks of age compared to 7 untreated shaker rats (5 administered a control AAV and 2 uninjected). (F) Rats treated with AAV9-CAG-hSLC9A6 trended towards a decrease in tremor at 14 weeks of age, albeit with substantial variability. Note that each data point is an individual rat, and all n-values for statistics are calculated as number of rats, not technical replicates.

Finding evidence of therapeutic benefit with L7- and CAG-based AAVs in this and previous [8] work, we took advantage of the variable nature of therapeutic benefit to ask fundamental questions: how do molecular markers co-vary with motor dysfunction, and how are ataxia and tremor interrelated in the context of Christianson syndrome?

The abundance of key cerebellar health markers is strongly tied to ataxia, but less so to tremor

In the context of AAV9-CAG-hSLC9A6 AAV, NHE6 and CALB1 expression are correlated with ataxia, but not tremor

As shown in Fig. 3 above, all shaker rats administered AAV9-CAG-hSLC9A6 AAV exhibited stronger NHE6 expression than uninjected shaker rats. However, in the context of those treated with AAV9-CAG-hSLC9A6 AAV, viral dose was not correlated with any of NHE6 expression, CALB1 expression, ataxia, or tremor (Supplemental Fig. 2). However, NHE6 was correlated with CALB1 (Fig. 4A) and ataxia (Fig. 4B). Further, CALB1 was highly correlated with ataxia (Fig. 4C). Interestingly, tremor was not significantly correlated with either NHE6 (Fig. 4D) or CALB1 (Fig. 4E).

Figure 4.

Following AAV9-CAG-hSLC9A6 AAV administration, ataxia, NHE6, and CALB1, are interrelated, but NHE6 and CALB1 are not correlated with tremor at 14 weeks of age. (A) NHE6 abundance following viral administration of AAV9-CAG-hSLC9A6 AAV positively correlates with CALB1 expression, suggesting a rescue of Purkinje cells. (B) NHE6 abundance is likewise negatively correlated with gait ataxia. (C) CALB1 is highly negatively correlated with gait ataxia. (D-E) While NHE6 and CALB1 abundance were correlated with ataxia and ataxia is correlated with tremor, NHE6 and CALB1 are not significantly correlated with tremor. Grey shaded regions represents 95% confidence intervals.

In the context of PHP.eB-L7-Slc9a6-GFP AAV, expression of NHE6 and key cerebellar proteins is correlated with gait ataxia, but not with tremor

Following the above analyses in the context of AAV9-CAG-hSLC9A6, we returned to the analysis of PHP.eB-L7-Slc9a6-GFP AAV to evaluate the relationship between motor outcomes and various cerebellar health-relevant expression levels. In a previous report on our PHP.eB-L7-Slc9a6-GFP AAV work, we showed that each of NHE6, CALB1, PCP2, and RGS8 were significantly reduced in shaker rats compared to wild type or Slc9a6 heterozygous rats, with Slc9a6 significantly reduced from wild type rats to heterozygotes [8]. Pooling all animals in the L7 AAV study with western blotting data (n = 11), we find that each of these metrics is significantly correlated with late-stage coordination, averaged from 21–25 weeks of age (Fig. 5A). Notably, these relationships held even when evaluating only within shaker rats, except for NHE6, which trended towards significance.

Figure 5.

In PHP.eB-L7-Slc9a6-GFP AAV experiments, cerebellar molecular markers—Reported in prior publication⁸—Are correlated with gait ataxia. (A) NHE6, CALB1, PCP2, and RGS8 are significantly correlated with late-stage ataxia (21–25 weeks). (B) Similar correlations are not found with late-stage tremor. Grey shaded regions represents 95% confidence intervals.

Late-stage tremor, averaged from 21–25 weeks of age, did not significantly correlate with any of these metrics when pooling all animals (Fig. 5B). When evaluating only within shaker rats (n = 7), trending relationships weakened further.

Tremor and ataxia share a complex relationship

The relationship between tremor and ataxia weakens over time

In the context of AAV9-CAG-hSLC9A6 AAV experiments, we found that at 14 weeks of age—the single time point of motor recordings—tremor and ataxia were highly correlated (Fig. 6A). Thus, we longitudinally analyzed the relationship between tremor and ataxia in all shaker rats in PHP.eB-L7-Slc9a6-GFP AAV experiments. We analyzed the relationship between tremor and ataxia from 12 weeks on—in other words, after significant symptom onset in both symptoms. In the initial weeks following symptom onset—12 to 15 weeks—ataxia was highly correlated with tremor. However, the correlation between ataxia and tremor decreased over time (Fig. 6B).

Figure 6.

(A) In the context of AAV9-CAG-hSLC9A6 AAV experiments, ataxia is strongly correlated with tremor at 14 weeks. (B-C) In the context of PHP.eB-Slc9a6-GFP AAV experiments, the correlation between peak tremor and late-stage ataxia is stronger than the correlation between tremor and ataxia at any given timepoint. (B) Ataxia and tremor are initially strongly related following symptom onset, but this correlation significantly decreases over time. (C) Peak tremor occurred prior to 20 weeks in all but 2 animals; however, the correlation between peak ataxia and late-stage ataxia (21–25 weeks) was higher than the correlation between tremor and ataxia at any individual timepoint. (D) Endpoint NHE6 expression and ataxia were negatively correlated at all timepoints. This negative correlation became stronger over time. (E) NHE6 and tremor were negatively correlated at all timepoints; however, the correlation between NHE6 and tremor did not strengthen over time. Grey shaded regions represents 95% confidence intervals.

For all but one time point after 18 weeks of age, no significant relationship was found between tremor and ataxia. Peak tremor was reached prior to 20 weeks of age in all but 2 of the 13 shaker rats in the study; however, when comparing peak tremor to late-stage ataxia (weeks 21–25), we found that peak tremor was highly correlated with eventual late-stage ataxia, achieving a stronger correlation than that between tremor and ataxia at any individual time point (Fig. 6C).

NHE6 expression and ataxia become correlated over time, but the correlation between NHE6 and tremor does not change over time

The degree to which endpoint NHE6 expression, measured following motor experiments in post-mortem tissues, predicted ataxia increased over time (Fig. 6D). Assuming that AAV-mediated NHE6 endpoint expression reflects the survival of transduced Purkinje cells, these data indicate that while there is some variability in the timing of ataxia onset in the absence of NHE6, endpoint NHE6 expression is highly correlated with the eventual end-stage ataxic burden.

Interestingly, such a relationship did not hold for tremor, and a trend was found that post-experimental NHE6 was better at predicting earlier-stage tremor than later stage tremor (Fig. 6E).

Discussion

Christianson syndrome is characterized by intellectual disability, epilepsy, behavioral abnormalities, and progressive motor dysfunction driven by cerebellar degeneration. Because SLC9A6 mutations cause loss of NHE6 function, viral vector-mediated gene replacement may be an appropriate therapeutic approach for Christianson syndrome. Here, we demonstrate that multiple viral constructs are capable of decreasing motor impairments and cerebellar protein expression, and with both constructs, relatively minimal NHE6 expression was required for substantial motor and molecular benefit. These major benefits driven by minimal NHE6 expression share resemblance with case reports in human patients in which even a minimal degree of functional NHE6 expression is associated with a large decrease in symptomatic burden [15]. Further, NHE6 loss of function leads to an over-acidification of the endosome [16], with Christianson syndrome mutations decreasing endosomal pH by 0.1–0.2 [17]. Low pH enhances viral release from the endosome [18], and this potentially may allow Christianson syndrome therapeutics using relatively low viral titers. Ultimately, the combination of our results, human findings, and the potential for decreased titers due to endosomal over-acidification all create reasons for optimism regarding viral gene replacement as a therapeutic strategy for Christianson syndrome. Indeed, treated rats in longitudinal studies had an average overall ~ 50% reduction in total ataxia burden and a ~ 40% reduction in total tremor burden.

While our results are promising, motor dysfunction is only one component of Christianson syndrome symptomology, necessitating substantial additional study. Quite clearly, study of efficacy in the context of Christianson syndrome symptomology including seizures and intellectual disability is critical, and further additional future steps will also be critical. This may include, but not be limited to, safety and toxicity study, optimization of promoter selection, evaluation of the relationship between dose timing and efficacy, route of administration study, and studies aimed at characterizing the distribution of NHE6 in the cerebellum and elsewhere following dosing. Indeed, balancing of endosomal pH, the implications of any given percentage of recovery of NHE6 in any brain region would depend on how that NHE6 expression is distributed. For example, 20% restoration evenly across all Purkinje cells would very likely result in a quite different outcome in the context of both safety and efficacy than 200% restoration in 10% of Purkinje cells and 0% in all others. While certain studies, such as dose timing or promoter selection might be ideal to carry out in rodents, such models may be less relevant in the context of final safety and toxicity or route of administration studies, necessitating eventual final work in larger species. Finally, considerable future work will be possible in the context of mechanistic links between Purkinje cell protection and motor rescue.

All results in this manuscript should be interpreted specifically in the context of Christianson syndrome; however, our findings may be interesting within a broader discussion of cerebellum-driven motor symptoms. Cerebellar dysfunction leads to numerous movement symptoms, including ataxia and tremor, and the relationship between tremor and ataxia is complex. Tremor is associated with numerous spinocerebellar ataxias [19–21] and gait/balance issues are associated with multiple forms of tremor [22–24]. In some degenerative ataxias, the presence of specific forms of tremor may be a predictive factor in the rate of ataxia progression, albeit with heterogeneity. For example, SCA1 and SCA6 patients with postural tremor experience slower progression of ataxia, while SCA2 patients with postural tremor experience faster progression of ataxia [19]. In some disorders that regularly feature both tremor and ataxia such as Fragile X-associated Tremor/Ataxia Syndrome, either tremor or ataxia can onset first with a similar likelihood [25]. It has been proposed that essential tremor may represent an intermediate form of spinocerebellar degeneration [26–29]. However, others have reported that Purkinje cell loss is not a feature of essential tremor [30, 31], generating debate [32, 33]. Increased understanding of the link between tremor and ataxia—as well as that between tremor and cerebellar degeneration—is needed, both for tremor patients and ataxia patients across a variety of conditions, including Christianson syndrome.

As discussed above, tremor/ataxia relationships are highly variable across varying cerebellum-related disorders, and this reported work is specific to Christianson syndrome. However, it is interesting to consider recent findings that have suggested that cerebellar tremor and ataxia arise from distinct and dissociable mechanisms in a variety of other models featuring cerebellar abnormalities [34]. Given this context in other models, it is interesting that each molecular marker analyzed was correlated with late-stage ataxia, and most were correlated with peak tremor—which occurred relatively early in symptomatic progression in nearly all cases—but none were significantly correlated with late-stage tremor. Further, the degree to which each of the reported molecular markers predicted gait ataxia increased over time, while their relationship with tremor either remained unchanged or weakened over time. Finally, while peak tremor—typically at early time points—was related to eventual end-stage ataxic burden with a significant relationship between tremor and ataxia at early time points, this direct correlation degraded with age. Thus, it may be worthwhile to investigate whether ataxia and tremor are dissociable in the context of the shaker rat model of Christianson syndrome, in agreement with the above-mentioned study.

Several potential caveats and limitations should be addressed. First, and most obviously, we did not observe dose dependence in the context of any of our molecular or motor outcomes in CAG AAV studies. Notably, this is likely tied to higher doses not associated with stronger NHE6 expression. Thus, we suspect that these findings were related to the differential volumes of injection at different doses, based on previous results showing that increased fluid volume leads to decreased transduction [35]. Indeed, we did not dilute lower doses to match volumes, and it is our hypothesis that doing so in future studies would, in fact, lead to dose-dependent relationships. Second, we note that statistical assumptions were violated in several correlation analyses, with several distributions (NHE6 and tremor in L7 studies) failing a Lilliefors test for normality. While some of these distributions would have met requirements if evaluated on a log scale, a log scale-based evaluation of, say, tremor would not be meaningful; regardless, conclusions would only change minimally if log scales were used for analysis. Third, we note that sample sizes are small for molecular outcomes in L7 studies in particular, as some tissues were used for alternate experiments outside the scope of this work. Thus, future replication in larger cohorts would be beneficial. Fourth, it should be noted that while motor benefits in terms of ataxia are strongly correlated with cerebellar markers, functional studies are not specifically included to link motor benefits to cerebellar changes, particularly in the context of AAV experiments with the CAG promoter. While prior work with the L7–6 promoter has shown strong specificity in Purkinje cells [36]—not directly studied by us—the CAG promoter is ubiquitous in nature, and it is possible that some motor benefits occur through non-cerebellar mechanisms.

Finally, we acknowledge a lack of biodistribution study in this and prior work. While we have previously published a small amount of histological images in the context of anti-CALB1 staining in PHP.eB-L7-Slc9a6-GFP rats, showing an increase in CALB1 specifically in Purkinje cells in representative examples [8], future work is critical in the context of biodistribution. Indeed, it is important to understand the distribution of NHE6 in Purkinje cells and other cells. Unfortunately, it has been our experience that commercially available antibodies carry substantial non-specific binding in rat tissues, negating their use for histology. However, others have characterized custom antibodies that function for histology in mouse and human cell models [37], and thus, one important future direction involves dosing of AAVs in NHE6-null mice to evaluate distribution of NHE6 following transduction.

Christianson syndrome is a complex neurodevelopmental and neurodegenerative disorder caused by loss of function in the SLC9A6 gene. In this work, we found strong evidence that viral gene therapy can reduce the ataxic component and increase related cerebellar protein levels in Christianson syndrome, but much work remains in the context of other symptoms, dose timing, and further optimization. This work presented the ability to critically evaluate the complex relationship between tremor and ataxia under the rare context of disease-modifying therapy. While the findings may be specific to Christianson syndrome, they may add to the complex discussion of tremor and ataxia.

Materials and methods

All animal work was approved and performed under University of Utah Institutional Animal Care and Use Committee protocols 19-10 011 and 2029. A total of 50 animals are reported on herein, sample sizes derived through power analyses based on the most variable outcome measure. Rats in L7-6 studies were further analyzed beyond the context of results reported in prior work [8]: specifically, motor data from weeks 8, 18, and 25 were reported in prior publication and make up 15% of motor data reported in the context of L7-6 in this study, while L7-6 western blotting data are repurposed in these new analyses. Rats were housed 2–5 per cage under a standard 12:12 h light cycle with free access to food and water. We maintain the shaker colony in the Wistar Furth inbred background (Envigo RMS LLC).

Genotyping methods

All rats were genotyped as previously described [8]. Briefly, we sought to identify all possible genotype outcomes via PCR using a forward primer of 5′-AAGACATGGCTGTGGCTCGG-3′ and a reverse primer of 5′-AGCTAGGGGACAGGGGTCCG-3′, generating a 363 base pair amplicon. The indel was confirmed via Sanger sequencing utilizing the same above-mentioned forward primer.

A‌AV production

PHP.eB-L7-Slc9a6-GFP and PHP.eB-L7-GFP AAV vectors were produced as previous described [8]. Briefly, we PCR-amplified coding sequences from a cDNA library made from wild type rat cerebellar RNA with the forward primer of 5′-TTTATGGCTGTG GCTCGGCGCGGCTGG-3′ and a reverse primer of 5′-TTTCGGCTGGACTGTGTCTTGTGTCATC-3′. We cloned the PCR product into a TOPO vector (Thermofisher, Cat# K457502). We re-amplified the rat Slc9a6 cDNA from the TOPO vector and cloned into pAAV-L7–6-GFP-WPRE [36] (Addgene, Plasmid #126462, gifted from Hirokazu Hirai) to generate pAAV-L7-6-Slc9a6-GFP. We generated recombinant AAVs in the University of Utah Drug Discovery Core Facility by co-transfecting HEK-293 T cells with either pAAV/L7–6-Slc9a6-GFP or the control pAAV-L7–6-GFP, in addition to pHelper (Stratagene, La Jolla, CA, USA) and pUCmini-iCAP-PHP.eB [38] (Addgene, plasmid #103005, gifted from Viviana Gradinaru), followed by viral particles purification, concentration, and genomic titer determination.

AAV9-CAG-hSLC9A6 AAVs were produced as follows: human cDNA sequence for SLC9A6 (hSLC9A6) (Accession: NM_001042537; Version: NM_001042537.2) corresponding to the NHE6-v1 splice variant (See Supplemental Fig. 1 regarding the selection of 701 aa NHE6-v1 vs. 669 aa NHE6-v0) was derived from the NCBI DNA database and used to design primers to PCR-amplify the coding sequences from a cDNA library. The primers set was as follows: SLC9A6-F (BamHI): 5’-AATTGGATCCGCCACCATGGCTC GGCGCGGCTGGCGGC-3′ and SLC9A6-R (EcoRI): TATCGAATTCTTAGGCTGGA CCATGTCTCGT. To generate hSLC9A6 AAV expression plasmid, pAAV-CAG-GFP plasmid (Addgene plasmid #37825, gifted from Edward Boyden) was modified by deleting GFP with BamHI and EcoRI restriction enzymes. The amplified hSLC9A6 PCR product was directly cloned into GFP pre-deleted pAAV-CAG-GFP plasmid at BamHI and EcoRI sites, designated as pAAV-CAG-hSLC9A6. The construct was verified by sequencing. Recombinant AAV particles [pAAV-CAG-GFP (control) and pAAV-CAG-hSLCA6] were generated by Signagen Laboratories, MD, USA (https://signagen.com/).

A‌AV administration

We administered 18 rats 2.26 × 10¹¹ vg of the control PHP.eB-L7-GFP AAV or PHP.eB-L7-Slc9a6-GFP AAV at 33–37 days of age (average 35 days): three male wild type rats, two female heterozygotes, and thirteen shaker rats, specifically six males and seven females hemizygous and homozygous, respectively, for the Slc9a6 mutation. Two male wild type and two female heterozygotes were kept as uninjected controls. Rats were grouped randomly and AAV administration surgeries were performed as previously described [8]. Briefly, we provided analgesia using 0.1 ml bupivacaine local to the incision and 0.1 mg/kg carprofen subcutaneously daily for 3 days. We anesthetized and maintained rats on 2% isoflurane. We opened to the scalp and dried and performed craniotomies targeting the right lateral ventricle. We waited three min post craniotomy and injected 10 μl of AAV at a rate not exceeding 2 μl/min. After five minutes wait post injection, we withdrew the needle at or slower than 1.0 mm/min. Rats were given a 5-day minimum recovery period.

Further, we administered 24 additional rats – 16 hemizygous males and 8 homozygous females – with AAV9-CAG-hSLC9A6 AAV at 28–41 days of age (average 34 days) by the same procedures with a co-injection of AAV9-CAG-GFP. While ages at injection were more variable in the CAG experiment, average age at injection across groups was insignificantly different by one-way ANOVA (f [5, 23]=1.29, P = 0.31, η² = 0.264). In comparison to L7 AAVs administered at only one dose, we administered AAV9-CAG-hSLC9A6 AAV to 6 groups of 4 rats at the following titers: 8.0 × 10⁹, 1.7 × 10¹⁰, 3.6 × 10¹⁰, 8.0 × 10¹⁰, 1.7 × 10¹¹, 3.0 × 10¹¹ vg. Regardless of the AAV9-CAG-hSLC9A6 AAV dose, all rats were administered 1.74 × 10^10 vg AAV9-CAG-GFP. Notably, one rat from the 1.7 × 10¹¹ vg group died 7 weeks post injection, prior to any analyses; therefore, this group is reported with 3 rats.

Note that an L7-6 icosahedral icon is used on all figures referencing experiments carried out with PHP.eB-L7-Slc9a6-GFP AAV, while a CAG icosahedral icon is used on all figures referencing experiments carried out with AAV9-CAG-hSLC9A6 AAV. When both are included, the appropriate icon is applied to relevant sub-figures.

Motor analyses

Tremor and ataxia were computed as previously described [5–8] using stratified randomized order of recording. Animals in the L7 experiment were recorded from for 30 minutes weekly from 6 to 25 weeks of age, while animals in the CAG experiment and additional uninjected shaker rats were recorded from for 30 min at one time point: 14 weeks of age. Rats were placed in the force plate actometer [39–41] chamber for 30 min without any cue or task. The original force plate actometer was rewired through a National Instruments Card and spontaneous ambulation was recorded through center of mass tracking at 1000 Hz using National Instruments LabVIEW.

We imported center of mass data into MATLAB and analyzed straightness of gait during auto-identified rapid movements as a measure of gait ataxia by computing a ratio of distance travelled to displacement. While the minimal value achievable on this ratio is 1.0, minor sway is always present in even healthy animals during walking, and some rapid movements involve rounding of corners. Wild type rats average ~ 1.6 on this metric with a tight standard deviation—indeed more than 98.5% of wild type recordings at more than 10 weeks of age yielded averages between 1.4 and 1.8. On the other hand, severely ataxic rats average more than 2.8 [5, 8], providing substantial ability to quantify improvement in ataxia. Notably, because ratios of less than 1.0 are not possible and wild type values are typically above 1.5 in these reported data, most y-axes start from 1.5 when visualizing straightness of gait/ataxia.

Tremor was quantified through basic Fourier analyses. We computed Fourier transforms of position over 5-second intervals, then averaged over the recording. Previously [5], we had averaged Fourier transforms of all age-matched wild type rats and computed tremor for the given rat relative to age-matched wild type rats specifically. However, given that tremor does not significantly change over time in wild type averages [5], we now more simply integrate the area above a generalized wild type curve averaged across all ages. Tremor strength peaks at ~ 5 Hz, with typical increases in power from 3 to 8 Hz. Therefore, we conservatively quantify tremor based on a 2–10 Hz band. Finally, we normalize on a scale in which 0.0 is approximately the maximum tremor score ever present for a wild type rat and 1.0 corresponds to severe tremor. Notably, we have previously determined that heterozygous females do not exhibit quantifiable tremor and do not make measurably less coordinated movements than wild type females. It is not possible to breed both homozygous and wild type females in the same litter; further, given an interest in litter matching across groups, all wild type animals were male, but what was labelled as ‘wild type’ was a combination of wild type male and heterozygous female.

Tissue collection and protein analyses

Tissue collection took place post motor data collection. In 11 of the 22 rats in the L7 cohort, we retained the left half of the cerebellum for western blot analyses, split evenly across groups, whereas half of the cerebellum was preserved from all animals in the CAG motor experiment. Note that the 11 rats excluded from tissue collection in the L7 cohort were selected via a stratified random sample to enable other experimental endpoints outside the scope of this report. For CAG analyses, we also included an additional 2 age-matched uninjected shaker rats and 4 age-matched uninjected wild type rats that did not have motor recordings made. Tissue was flash frozen and homogenized for protein extraction. Protein western blots were run as previously described [8, 42], with protein quantification performed relative to beta-actin (ACTB) or glyceraldehyde-3-phostphate dehydrogenase (GAPDH).

The following primary antibodies were used: NHE6 [(1:5000), Abcam, Cat# ab137185], GAPDH [(1:5000), Cell Signaling, rabbit mAb #2118], ACTB [(1:30000), Sigma-Aldrich mouse mAb #A3854)], Calbindin 1 (CALB1) [(1:5000), Sigma-Aldrich, mouse mAb #C9848], Receptor of G Protein Signaling 8 (RGS8 [(1:5000), Novus Biologicals, rabbit pAb #NBP2–20153], and Purkinje Cell Protein 2 (PCP2) [(1:3000), Santa Cruz, mouse mAb #sc-137 064]. Secondary antibodies used were HRP-Goat Anti-Rabbit IgG (H + L) [(1:5000), Jackson ImmunoResearch Laboratories, Cat# 111-035-144] and HRP-Horse Anti-Mouse IgG (H + L) [(1:5000), Vector Laboratories, Cat# PI-2000].

Statistical analyses

Statistical comparisons were made via several methods, the associated test reported with each p-value. First, we used two-sample Student’s t-tests, where n is the number of animals. T-tests were performed one-tailed when previous published data directly supported a hypothesis and two-tailed otherwise. Two-tailed tests are noted as such, and other t-tests should be assumed to be one-tailed. Second, we performed repeated measures ANOVA tests with appropriate post-hoc testing. Finally, we performed Pearson correlation testing to quantify p-values from the correlation coefficient r. Notably, for correlation testing, significant correlations are labeled in blue, and insignificant correlations are labeled in red. Further, when correlations are made from individual animals, data points are left unconnected; however, when a data point is generated from a group of animals in correlation analyses across time points, data points are connected. All p-values are reported in Supplementary Table 1, along with effect sizes: Hedges’ d for t-tests and η² for ANOVA tests, and Pearson’s r for correlation analyses.

Author contributions

Collin James Anderson (Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Resources, Software, Visualization, Writing—original draft, Writing—review & editing), Karla P Figueroa (Conceptualization, Investigation, Writing—review & editing), Sharan Paul (Investigation, Methodology, Visualization, Writing—review & editing), Mandi Gandelman (Investigation, Writing—review & editing), Warunee Dansithong (Investigation, Methodology, Writing—review & editing), Joseph A Katakowski (Conceptualization, Writing—review & editing), Daniel R Scoles (Resources, Supervision, Writing—review & editing), Stefan-M Pulst (Conceptualization, Resources, Supervision, Writing—review & editing).

Conflict of interest statement

The authors declare no competing interests.