Quantitative Gait Analysis Using a Motorized Treadmill System Sensitively Detects Motor Abnormalities in Mice Expressing ATPase Defective Spastin

The hereditary spastic paraplegias (HSPs) are genetic conditions in which there is progressive axonal degeneration in the corticospinal tract. Autosomal dominant mutations, including nonsense, frameshift and missense changes, in the gene encoding the microtubule severing ATPase spastin are the most common cause of HSP in North America and northern Europe. In this study we report quantitative gait analysis using a motorized treadmill system, carried out on mice knocked-in for a disease-associated mutation affecting a critical residue in the Walker A motif of the spastin ATPase domain. At 4 months and at one year of age homozygous mutant mice had a number of abnormal gait parameters, including in stride length and stride duration, compared to heterozygous and wild-type littermates. Gait parameters in heterozygous animals did not differ from wild-type littermates. We conclude that quantitative gait analysis using the DigiGait system sensitively detects motor abnormalities in a hereditary spastic paraplegia model, and would be a useful method for analyzing the effects of pharmacological treatments for HSP.


Introduction
Hereditary spastic paraplegias (HSP) are genetically determined subtypes of motor neuron disease that are characterized by dying-back degeneration of the axons of the corticospinal tract [1][2][3][4][5]. In humans, they cause a progressive spastic gait abnormality that can begin from childhood to late adult life. No treatments for the HSPs are available presently, but potential therapeutic approaches, such as inhibition of BMP signaling or manipulation of microtubule dynamics, have been suggested [4,6]. Thus there is a need to develop appropriate model systems and reliable methodologies with which to test therapeutic effects.
Mutations in SPAST/SPG4, which encodes the microtubule severing ATPase spastin, are the most frequent cause of HSP [7]. In North America and northern Europe, mutations in this gene account for up to 40% of autosomal dominant uncomplicated HSP and around 10%

Knock-in mouse generation and breeding
A schematic diagram of the targeting strategy is shown in Fig 1A. We had generated (at Taconic Artemis) mice knocked-in for a mutation in exon 8 (Spast c.1152c<a (p.Asn384Lys), NM_001162870.1, CCDS50181.1) encoding an N384K amino acid substitution, on a C57BL/6 genetic background. Briefly, exons 5-8 of the spast locus were replaced by a positive selection cassette containing the mutation. This cassette contained two positive selection markers (neomycin and puromycin resistance) flanked by flippase (Flp) recognition target F3 or FRT sites, in introns 4 and 7 respectively. It also incorporated loxP sites in introns 4 and 7, providing the possibility of future deletion of these exons by crossing with Cre-recombinase expressing mice, to generate a spastin loss-of-function model. The targeting vector was generated using BAC clones from the C57BL/6J RPCI-23 BAC library and transfected into TaconicArtemis C57BL/ 6N Tac ES cell line. ES clones were analyzed by Southern Blotting to confirm correct recombination and single integration, according to standard procedures. Homologous recombinant clones were used to generate chimeric animals by injection into C57BL/6 blastocysts. Highly chimeric mice were bred to C57BL/6 Flp-Deleter mice, to remove the selective markers. Germline transmission was identified by the presence of black, strain C57BL/6 offspring (G1) and was confirmed by PCR of ear biopsies using the following primer pairs: 2201_31: AAACTGTTCCCAAGGCATCC and 2201_32: GTGTCAGTGTGCATAAGTCATGG, which detect the presence or absence of the loxP site in intron 4, 2084_29: GTCATAGCTGTAACCAACTTCTG and 2084_30: ACCAAACACATGCGAGTGAGG, which amplify an exon 8 sequence containing the mutation. The presence of the mutation was confirmed by sequencing (Source BioScience) (Fig 1B).
Mice were maintained in accordance with United Kingdom and European Union regulations. Animal work for this study was approved by the University of Cambridge Ethical Review Committee and was carried out under project licenses (80/2304 and 70/7888) granted by the United Kingdom Home Office under the Animals (Scientific Procedures) Act 1986. The University of Cambridge is a designated establishment for breeding and for scientific procedures under the Animals (Scientific Procedures) Act 1986.
Mice were housed under temperature-controlled and specific-pathogen-free conditions, with a 12-hour light/dark cycle and free access to drinking water and food. They were maintained on a C57BL/6 background. We interbred heterozygous mice to generate wild-type (termed spast wt/wt ), heterozygous (termed spastin N384K/wt ) and homozygous (termed spastin N384K/N384K ) mice. The cohort of mice analysed in the behavioural studies was made up of groups of spast wt/wt , spast N384K/wt and spastin N384K/N384K littermates. Embryonic or day old mice used to generate primary neurons were sacrificed by decapitation, adult mice were sacrificed by cervical dislocation.

Antibodies
The following antibodies were used: Mouse monoclonal anti-MAP2 (abcam ab11268), mouse monoclonal anti-acetlylated tubulin (Sigma-Aldrich T7541)), rabbit polyclonal anti-spastin (spastin86-340, raised in house to amino acids 86-340 of spastin [24,25]), rabbit monoclonal anti-GAPDH (Cell Signalling 2118, clone 14C10), mouse monoclonal anti-Tau-5 antibody (Abcam ab80579). Spastin N384K/N384K knock-in mouse generation and validation. A) Schematic diagram of strategy used to generate spastin N384K allele. Numbering refers to spastin exons. Puro = puromycin resistance cassette, Neo = neomycin resistance cassette, Flp = flippase, FRT = flippase recognition target, UTR = untranslated region. B) DNA sequencing of PCR product generated using primers designed to amplify fragment of spastin exon 8 from genomic DNA of a wild-type mouse and spastin N384K/N384K littermate. C) Immunoblotting versus spastin of lysate or pellet fraction from foetal brain tissue (E17), from mice of the Homogenate was centrifuged at 5000rpm for 15 min and supernatant centrifuged at 49000rpm for 1 hour. Protein concentration was quantified using BCA protein assay (Pierce) and total spastin levels were analysed in 30μg of lysate protein using SDS-PAGE. Pellet fractions were resuspended in 300μl sample buffer and 15μl was analyzed using SDS-PAGE, with equal lane loading confirmed by Coomassie staining. Gels were blotted onto PVDF membrane and blocked in 5% non-fat dried milk in TBS-Tween. Membrane was incubated with primary antibodies and HRP-conjugated secondary antibodies in blocking buffer and HRP-signal detected with Super-Signal 1 West Pico Chemiluminescent Substrate using X-Ray film (Fuji Medical). Immunoblot bands were quantified using ImageJ software.

Primary Neuron Culture
Cortical neurons were obtained from E17 mouse embryos or day old pups and cultured on Poly-D-Lysine coated glass coverslips in Neurobasal medium (Gibco/Invitrogen) with B27 supplement (Gibco/Invitrogen) and L-glutamine (Sigma-Aldrich). After seven days in culture neurons were fixed in 3.7% formaldehyde and processed for immunofluorescence microscopy.

Immunofluorescence microscopy
To quantify axonal swellings, formaldehyde fixed neurons on coverslips were permeabilised with 0.1% TX100 in PBS and blocked in 5% fetal calf serum in PBS for 1 hour. Coverslips were incubated with primary antibodies and Alexa Fluor 1 conjugated secondary antibodies in blocking buffer and mounted onto glass slides using Prolong 1 Gold antifade reagent with DAPI (Molecular Probes). Neurons were visualized using a Zeiss Axioimager Z2 Wide-field upright microscope and axonal swellings were quantified per 250 nuclei. Axonal swellings were identified as acetylated tubulin positive, MAP2 negative regions on a single axon which were greater than twice the diameter of the axon.

Tissue genotyping
Genomic DNA was purified from mouse ear biopsies using DNeasy 1 Blood and Tissue kit (Qiagen). PCR was performed using primer pairs 2201_31 and 2201_32 on an Applied Biosystems GeneAmp 1 9700 PCR system. PCR products were analysed on a 1.5% agarose gel and genotype was identified by the presence or absence of bands at 235bp (wt/wt), 235 and 427bp (wt/N384K), 427bp (N384K/N384K).

Modified Shirpa Analysis
The following behaviours chosen from the modified Shirpa analysis were recorded for each mouse at 1 year old: body position, tremor, palpebral closure, coat appearance, presence of whiskers, presence of defecation, transfer arousal, locomotor activity in arena by measuring number of squares entered with all 4 feet in 3 seconds, tail elevation, touch escape, positional passivity, trunk curl, limb grasping and evidence of biting [26].

DigiGait analysis
DigiGait equipment was purchased from Mouse Specifics, Inc (Boston) and analysis was carried out as previously described [27]. Briefly, each mouse was placed on a motor-driven genotypes indicated. The histogram shows corresponding densitometry of the spastin bands in the lysate (n = 3 per genotype, p-values generated by paired t-tests). GAPDH immunoblotting or Coomassie staining serve to verify equal loading of lanes. Size bars indicate molecular weight in kD.
doi:10.1371/journal.pone.0152413.g001 treadmill with a transparent treadmill belt and imaged from beneath with a high-speed digital video camera. A minimum of three seconds of movie is required for digigait analysis, and mice that could not run at the chosen speed for that duration were deemed to have failed to complete the test. Colour images were converted to their binary matrix equivalents and the areas of the moving paws relative to the belt and camera were calculated throughout each stride. This was used to generate a dynamic gait signal of the paw placement relative to the treadmill belt and camera. Each limb's gait signal was used to calculate the stride duration (time duration of one complete stride for the paw under analysis). This was broken down into subcomponents of stance duration (the time duration when the paw is in contact with treadmill) and swing duration (the time duration when the paw is above the walking surface and not in contact with the belt). Stance duration further comprises the braking phase (time duration from initial paw contact with the treadmill to maximum paw contact) and propulsion duration (time duration from maximum paw contact to lifting from treadmill). Stride width was defined as the perpendicular distance between the centroids of each set of axial paws during peak stance. Gait symmetry was measured as the ratio of forelimb stepping frequency to hind limb stepping frequency.

Statistical Analysis
Statistical analyses were carried out using GraphPad Prism 5.01 for Windows (GraphPad Software, San Diego) statistical software. Specific statistical tests used are described in the relevant figure or table captions. Briefly, for behavioral analysis, one-way (to analyse effects of a single variable, i.e. genotype) or two-way (to analyse effects of two variables, i.e. gender and genotype) ANOVA was performed to compare effects across 3 groups (spast wt/wt , spast N384K/wt or spast N384K/N384K ). In certain cases ANOVA was supplemented with Bonferroni post-test for comparison of individual pairs of genotypes. Fischer's exact test was used to analyse the contingency table of digigait success or failure by genotype. Differences in spastin immunoblot band densitometry were assessed by paired t-tests.

Generation of spastin knock-in mice
We used a gene targeting approach to generate a mouse model on a C57BL/6 background, which was knocked in for a mutation (1152c<a) in exon 8 of murine spastin. This mutation encodes an N384K missense change in the spastin protein, corresponding to the human disease-causing missense mutation N386K [8]. The presence of the mutation was confirmed by sequencing ( Fig 1B). Spastin protein was expressed in brain tissue from spastin wt/wt , spastin N384K/wt and spastin N384K/N384K animals ( Fig 1C).

Axonal swellings in primary neurons from knock-in mice
In light of the primary neuron axonal swelling phenotype observed in previous spastin mouse models, we examined whether similar swellings were present in primary neurons derived from spastin N384K/N384K mice. We saw prominent neurite swellings in spastin N384K/N384K primary neurons and the affected neurites never labeled for the somatodendritic marker MAP2 but labeled with the axonal marker tau, confirming them as axons (Fig 2A-2C). Spastin N384K/wt neurons did not develop significantly more axonal swellings than wild-type littermates ( Fig  2B). These results demonstrate the presence of a cell-autonomous neuronal pathology, and are consistent with findings in other spastin mouse models. Furthermore, they demonstrate no Gait abnormality in spastin N384K/N384K mice Heterozygous mutant mice bred normally, and both heterozygous and homozygous mutants showed no obvious defects in size or morphology. However, at 1 year old, spastin N384K/N384K mice were significantly lighter than heterozygous or wild-type mice (Fig 3A). We characterized the locomotor phenotype of a cohort of 1 year old male and female littermate spastin N384K/wt and spastin N384K/N384K animals, as well as littermate spast wt/wt mice. No obvious phenotypic differences between the different genotypes were detected on modified primary SHIRPA examination. We also examined the maximum running speed of the animals at 1 year using an accelerating treadmill, with treadmill speeds increasing by 1cm/s at 10s intervals, beginning from 5cm/s. Again, we found no significant effect of genotype.
We next used the DigiGait analysis system to quantify gait parameters while the mice were running on a treadmill. This system images the ventral side of mice as they move on a motorized transparent treadmill belt and quantifies spatial and temporal gait indices [27]. These include the duration for a complete stride (stride duration), as well as individual components of the stride, such as the duration of swing and stance phases (see Methods section for a full description of the gait parameters analysed). As many rodent gait parameters are influenced by running speed [28], we used a constant speed of 18cm/s. This running speed was chosen to optimize the number of animals able to complete the testing, as at higher speeds increasing numbers of mice of all genotypes were not able to keep up with the treadmill, most likely because of their relatively advanced age. Even so, some animals were unable to run at 18cm/s and so were excluded from analysis, but the proportion excluded did not significantly differ by genotype (Table 1). Using 2-way ANOVA, we found significant effects of genotype, but not of gender in a number of basic hind limb gait metrics (Table 2 and Fig 3). In particular, the spastin N384K/N384K mice had a significantly lengthened mean stride duration, and the increase was almost entirely due to an elongation of the swing phase of gait ( Table 2, Fig 3B-3D and S1-S4 Movies). These changes altered the profile of a typical step in the spastin N384K/N384K mice, compared to the other two genotypes (Fig 3E). Mean stride length was also significantly increased in the spastin N384K/N384K mice, consistent with the need to maintain a constant speed with a slower stride (Table 2 and Fig 3F).
To analyse whether there was evidence that the N384K mutation acted in a dominant fashion, we next analysed the key gait metrics that were influenced by genotype, specifically to determine whether heterozygous mutant animals differed significantly from homozygous mutant or wild-type animals. As the 2-way ANOVA analysis demonstrated no effect of gender on these parameters, we pooled data from males and females, and compared the effect of each genotype against the other two. While we saw significant differences between wild-type and spastin N384K/N384K animals, and between spastin N384K/wt and spastin N384K/N384K , we saw no significant difference between spastin wt/wt and spastin N384K/wt mice (Table 3). We thus saw no dominant effect of the mutation on the gait phenotype.

Abnormal gait parameters can be detected at 4 months old
Having characterised gait parameters in the 1-year old cohort of spastin-HSP mice, we tested whether gait abnormalities could be detected at a younger age by analyzing gait metrics in a separate cohort of mice at 4 months old. Of note, in wild-type mice at this age, key gait parameters including stride duration, stance duration and swing duration were very similar to those previously obtained by Digigait in a group of 3 month old C57BL/6 mice running at a similar  Table 2. To facilitate interpretation, p-values for genotype are indicated on the plots by * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.  To correct for multiple testing, the alpha value for significance was calculated by dividing 0.05 by 8, using a highly conservative assumption of 8 independent tests (many of the tests are in reality related), to give α = 0.00625. spast wt/wt n = 10 (4 male, 6 female), spastin N384K/w t n = 16 (7 male, 9 female), spastin N384K/N384K n = 10 (6 male, 4 female). speed [29]. However, we found a significant increase in stride duration in spastin N384K/N384K mice versus littermate controls (Table 4 and Fig 4A). This was not restricted to the swing phase of gait, as stance phase duration was significantly increased (Table 4 and Fig 4B and 4C). Thus overall, the gait profile was not affected by genotype (Fig 4D). Finally, as with the older mice,  there was a significant increase in stride length in spastin N384K/N384K compared to wild-type and heterozygous littermates, consistent with the need to run at the same speed with a slower stride duration (Table 4 and Fig 4E). We also analysed, as described above for the 1 year old mice, whether there was any dominant effect of the mutation in heterozygous mice, but found no significant difference between wild-type and heterozygous mice for any gait parameter analysed.

Discussion
We generated a novel genetically modified mouse knocked-in for a disease-associated missense mutation that results in expression of an ATPase-defective form of spastin, thus modelling a class of spastin mutations that has not previously been analysed in mammalian in vivo systems.
In contrast to our model, in which enzymatically dead spastin is expressed, previous mouse models have involved loss of the spastin protein, caused either by a deletion of exons 5-7, or defective splicing of exon 7 [20,21]. Nevertheless, like these earlier models, in our model homozygous mutant animals had axonal swellings in cortical neuronal cultures (which contain the target cells of HSP) and a gait abnormality. The gait abnormality was evident on quantitative gait testing, but was subtle as it was not detected on modified Shirpa testing. Many gait metrics change with running speed, and so to allow direct comparison of metrics between animals, gait was assessed at a constant speed [28]. At both 4 months and one year old we found that the spastin N384K/N384K mice had a symmetrical gait, but that they had a significantly longer stride duration than control mice.
Step length was also increased in the spastin N384K/N384K mice, and this is consistent with findings in one of the knock-out models [20]. The identification of quantifiable gait abnormalities as early as 4 months old in the spastin N384K/N384K mice will be useful in assessing the therapeutic response to future pharmacological agents designed to treat HSP. Indeed, some gait metrics differed by genotype significantly in 4 month old mice but not in 1 year old mice (e.g. stance duration). With the caveat that we examined relatively few 4 month old animals, so that we may by chance have under-estimated the inter-animal variability in gait metrics at this age, we speculate that this may be because degenerative changes affecting gait parameters may be more prevalent in the older animals, so increasing the variability of results (compare for example the variability in wild-type mouse results for stride duration in Fig 3B versus Fig 4A). This would tend to obscure subtle differences between control and homozygous knock-in mice.
In contrast to the situation in humans, where heterozygous spastin mutations are sufficient to cause HSP, we saw no frank axonal swellings or convincing locomotor phenotype in mice heterozygous for the spastin N384K-expressing allele, although in the 4 month old animals there was a trend towards several gait metrics for heterozygous animals being intermediate between wild-type and homozygous mutant values. While mice may require increased dosage of mutant alleles to drive development of neurodegenerative diseases within their short (compared to humans) lifespan [30], significant effects on axonal swelling or gait phenotypes might have been expected in heterozygous mice if the mutation acted via a dominant negative or gain-of-function pathological mechanism. Thus while not formally excluding them from playing a role in human patients, our results provide no strong support for these mechanisms of disease acting in our mouse model.

Conclusions
We have developed a novel mouse model expressing an ATPase-defective form of spastin. Homozygous mutant mice develop a subtle but reliably detectable gait abnormality. Our results highlight the utility of quantitative gait analysis using treadmill systems in identifying gait abnormalities in models of hereditary spastic paraplegia.
Supporting Information S1 File. Minimal data set. (XLSX) S1 Movie. Representative example of ventral imaging produced by the Digigait system of a 1 year old wild type mouse running on a transparent treadmill at 18cm/s.