Duchenne muscular dystrophy (DMD) is an X-linked neuromuscular disorder that is characterized by progressive muscle weakening and wasting. About 1 in 5,000 newborn boys are affected by this disease (Ricotti et al., 2016). DMD patients lack the protein dystrophin, which connects the cytoskeleton with the extracellular matrix, and thereby protects muscle fibers from damage during contractions. In the absence of dystrophin, chronic muscle damage and inflammation, and impaired regeneration lead to replacement of muscle tissue by connective and adipose tissue (Deconinck and Dan, 2007). Most DMD patients lose ambulation before their teens and die in their third to fourth decade of life due to respiratory and cardiac complications (Aartsma-Rus et al., 2006).

DMD is caused by frame-shift or nonsense mutations in the DMD gene, which is the largest known gene in the human genome. It contains 2.2 Mb of genomic DNA and 79 exons that make up the coding sequence for full-length isoforms. There are seven promoters spread over the DMD gene that give rise to several dystrophin isoforms (Aartsma-Rus et al., 2006). Full-length dystrophin isoforms (Dp427c, Dp427m, and Dp427p) are lost in all DMD patients and, depending on the position of the mutation, one or multiple shorter isoforms are lacking as well. In muscle, only the full-length isoform Dp427m is expressed, while the majority of the other isoforms (Dp427c and p, Dp140, Dp71 and Dp40) are primarily expressed in the brain (Aartsma-Rus et al., 2006; Doorenweerd et al., 2017).

It is therefore not surprising that the absence of dystrophin in the brain also has implications on brain development, behavior and intellect. Many DMD patients have learning difficulties and delayed developmental milestones. There is a higher incidence of comorbidities with neuropsychiatric and behavioral disorders such as autism-spectrum disorder, attention-deficit hyperactivity disorder, obsessive-compulsive disorder, anxiety, and depression (Hinton et al., 2006; Waite et al., 2012; Banihani et al., 2015; Ricotti et al., 2016). DMD patients often have an intellectual disability, and an IQ of one standard deviation below the mean of the general population (Hinton et al., 2006; Banihani et al., 2015).

The most commonly used DMD mouse model is the mdx mouse. This strain lacks the full-length dystrophin isoforms due to a spontaneous point mutation in exon 23 of the Dmd gene. In contrast to DMD patients, muscle pathology of the mdx mouse is not severe and they have a near normal life expectancy (Chamberlain et al., 2007). Yet, it has been demonstrated that also mdx mice show emotional defensive behavioral abnormalities and impaired long-term memory retention due to the absence of dystrophin (Vaillend et al., 2004; Vaillend and Chaussenot, 2017; Comim et al., 2019). An interesting finding is that mdx mice show a freeze response upon fixation or a stress stimuli (i.e., foot-shock), which is rarely seen in wild-type mice (Sekiguchi et al., 2009; Vaillend and Chaussenot, 2017; Comim et al., 2019).

A study by Remmelink et al. investigated performance in several cognitive domains in female mdx mice. In an automated home-cage (PhenoTyper) setup, a relatively strong and distinct impairment in reversal learning (RL) following a discrimination learning (DL) task was observed (Remmelink et al., 2016). The aim of this study was to test whether this particular phenotype can also be found in mdx males, or if dystrophin deficiency affects behavior differently in male and female mice. Interestingly, we could not reproduce these abnormalities in cognitive flexibility in neither mdx males nor females. Since the automated home-cage system avoids human-animal interaction and ensures a standardized housing and testing environment, we conclude that this irreproducibility must be attributable to sources of variation that acted prior to cognitive testing. Therefore, we investigated both studies in detail and report here a number of factors that may have contributed to irreproducibility. This finding highlights that behavioral tests are subject to variation, identified potential sources of variation, and supports the notion that behavioral studies ideally deliberately incorporate potential sources of variation in order to test reproducibility and hence the robustness of the findings.

Mice (n = 15 per gender per genotype) were tested in three batches starting with the PhenoTyper home-cages at an age of 18 ± 1.3 weeks and the DLB/Y-maze at an age of 21 ± 2.2 weeks. From the spontaneous behavior and DL/RL data, one mdx male was excluded by automatic quality control since the mouse was sleeping outside of the shelter and the data file of one mdx female lost. One wild-type female was found dead before the start of the experiments. One mdx male and one female were found dead before the start of the DLB and Y-maze test. While mdx mice are more sensitive to stress, it seems unlikely to be the cause of death since the mdx mice were found during the resting period and did not undergo extensive handling during this period.

While DMD is primarily characterized by progressive muscle wasting, the cognitive and behavioral implications of the loss of brain dystrophin isoform(s) have been investigated more frequently in recent years. A greater understanding of the pathological consequences of a loss of dystrophin isoforms in the brain of DMD mouse models like the mdx mouse, will facilitate the development of therapeutic approaches to treat the DMD brain. In this study, we compared behavioral and cognitive abnormalities between male and female mdx mice.

Learning disabilities are common in DMD patients, with incidences ranging from 18.6% to 44% in various cohorts studied (Banihani et al., 2015; Ricotti et al., 2016). Several tests have been designed to study learning abnormalities in mice. In our study, we investigated DL and RL in a PhenoTyper home-cage. We did not find any DL abnormalities in mdx mice regardless of their gender. This is in accordance with our previous study in female mdx mice (Remmelink et al., 2016). In line with this, other studies have not found abnormalities in DL, assessed in operant procedure tasks in which mdx mice received a food-reward for correct actions (Lewon et al., 2017; Dickson and Mittleman, 2019). However, it was found that mdx mice actually outperform wild-type mice after a food deprivation period hinting to an additional motivation to perform for a food reward (Lewon et al., 2017). Acquisition learning has also been studied in maze-based assays such as the Morris water maze in which mdx mice either did not show any difference in initial learning to find the platform (Sesay et al., 1996; Vaillend and Chaussenot, 2017) or showed an initial reduced performance during the assay but did not show overall impaired learning compared to wild-type mice (Chaussenot et al., 2015). In the Barnes maze (Remmelink et al., 2016) and the Radial maze (Chaussenot et al., 2015) mdx and wild-type mice's learning abilities were also comparable.

We previously observed that female mdx mice have impaired cognitive flexibility (Remmelink et al., 2016). In that study, four out of the nine female mdx mice included did not reach the 80% criterion during the RL phase of the DL/RL test. Furthermore, in this cohort we also found impaired flexibility in the RL phase of the Barnes maze test (Remmelink et al., 2016). In contrast, other studies in which serial reversal learning was either assessed in a food-reward based task and/or in a water maze or radial maze, did not observe impaired flexibility in mdx mice (Chaussenot et al., 2015; Dickson and Mittleman, 2019). In our current study, we did not detect a RL deficit in male mdx mice, nor could we reproduce our earlier findings of a deficit in the RL task in female mdx mice. As the task is fully automated and executed by the mice at night without human interference, we are puzzled about this discrepancy. We here discuss differences between our previous study, our current study and other studies that might have an influence on the discrepancies in behavioral abnormalities seen in the mdx mouse model.

Our previous results might have been affected by a small sample size. However, post-hoc power calculations showed that the power of that study was larger than 80%, which is typically accepted as a sufficiently powered animal study. Adding to the discrepancy, in the current study we observed a decrease in general activity and subtle differences in sheltering behavior of female mice during the days of spontaneous behavior, which was not detected in the previous study. Meanwhile we have performed the DL/RL tasks in well-powered n = 25 mdx mice (on a C57BL/6J genetic background), as part of another study, and again did not find RL deficits (manuscript in preparation). Taken together, the fact that two independent studies and our own follow up studies using the DL/RL task did not reproduce these findings show that in the mdx mouse reversal learning is unaffected and that our previous observation could have been caused by idiosyncratic factors.

In an attempt to identify these factors, we scrutinized the experimental protocols of the current and previous study. We could not identify obvious differences in the hygiene level of the mice. There were however differences in the breeding scheme and housing conditions. In the current study mice were bred from homozygous wild-type or mdx lines while in the previous study we had heterozygous breeding lines that resulted in wild-type, heterozygous and homozygous mdx mice. We also noticed that mice were shipped from the breeding location to the test location at a younger age in the previous study (8–11 weeks vs. 8–16 weeks in the present study). Due to the implementation of an additional health screening step in the test facility, mice were housed for an additional 4 weeks in IVC cages, whereas in the previous study mice were housed in open cages upon arrival at the testing facility. There were differences in the number of animals housed per cage. Male mice were housed individually and the number of females per cage was 2–3 mice per cage while in the previous study female mice were housed with 2 mice per cage. The age at which mice were subjected to the DL/RL task was 13 weeks in the previous study, and 18 weeks in the present study. Finally, in the previous study the home-cage testing protocol had an additional 2-day light spot avoidance test in between the 3 days of spontaneous behavior and the DL/RL task, which was not present in the current study. A comparison of the behavioral tests done in the current study and previous study can be found in Supplementary Table 3. In summary, we identified a number of factors that may affect mouse behavior and thereby an mdx phenotype.

Besides cognitive flexibility and activity, we also studied anxiety-related behavior in mdx mice in the DLB assay. While mice have a natural tendency to explore new spaces, this is hampered in the presence of stressors like bright light (Bourin and Hascoet, 2003). Mice with increased anxiety will therefore show more aversion to brightly lit new spaces than less anxious mice. The DLB is based on this principle and thus a good measure for increased anxiety. In our study, mdx mice showed a slight decrease in exploration of the new brightly lit compartment, but this decrease was not significant compared to the wild-types. This is in contrast to our previous study in which we observed an aversion toward the lit area in mdx females (Remmelink et al., 2016). This discrepancy cannot be explained by a difference in experimental setup since those are the same. Anxiety has only sporadically been studied in the DLB; one group found that mdx males showed increased anxiety-related behavior compared to wild-types (Vaillend and Chaussenot, 2017). However, in one of their previous studies, which used a different DLB setup, they did not see this decrease in time spent in the light compartment (Vaillend et al., 1995). Anxiety has also been studied in the elevated plus-maze in which mice can explore a maze that has two enclosed arms opposite of each other and two open arms. A decrease in exploration of the open arms is indicative for increased anxiety, which was not found in mdx males (Sekiguchi et al., 2009; Vaillend and Chaussenot, 2017). Taken together, the underlying cause of discrepancies in findings within and between distinct anxiety tests remains unclear and should be further investigated. Both nature and strength of the anxiogenic stimulus and intrinsic novelty-seeking motivations could be underlying factors.

The Y-maze was used to assess short-term spatial working memory. Rodents have a natural curiosity to explore new areas and thus the spontaneous alternation percentage in the Y-maze gives an indication on the functionality of short-term memory. Mice with a good working memory have a high spontaneous alternation percentage, or a bias toward exploring new arms located to the left or right of them, whereas low spontaneous alternation percentage indicates a poor working memory (Kraeuter et al., 2019). In our study, we observed that all groups performed above chance levels. While the Y-maze requires minimal animal handing due to its continuous nature, it might not be the best experiment to detect (subtle) abnormalities in the spatial working memory of mice (Hughes, 2004; Deacon and Rawlins, 2006). In our previous study, we assessed spatial working memory in the T-maze and observed no differences between mdx and wild-type females during a two-day period (Remmelink et al., 2016). Vaillend et al. also performed the T-maze test, using a longer protocol; one sample trial on the first day, two consecutive trials on the second day followed by a final trial either 6 or 24 h later (Vaillend et al., 1995). They observed no abnormalities in the spontaneous alternation percentage at 6 h, however mdx mice performed at chance levels while wild-types performed above chance levels at 24 h (Vaillend et al., 1995). Long-term, but not short-term spatial working memory might thus be primarily affected in mdx mice.

An outstanding question is which tests are most suitable to study the behavioral abnormalities and to detect efficacy of treatments targeting the brain of mdx mice. Ideally, these would be robust and sensitive tests, providing a large therapeutic window and being executed in a similar manner (using identical protocols) within and between labs to allow direct comparisons between outcomes. Generation of Standard operating procedures (SOPs) for a selection of such behavioral tests, as currently available for muscle functionality outcomes on the TREAT-NMD Alliance website (https://treat-nmd.org/research-overview/preclinical-research/experimental-protocols-for-dmd-animal-models/) could facilitate this and improve test reproducibility. Discrepancies described in our current study highlight the need of well-powered studies. This also holds true for tests that are performed in a fully-automated fashion without human interference. Some of the behavioral phenotypes in mdx mice appear to be subtle and sensitive to so far unidentified factors that have an impact on mouse behavior prior to the actual testing of the mice. The level of standardization of the present set of studies was obviously not sufficient to reach reproducible results. Given these subtle phenotypes, it is tempting to advocate for standardization of external conditions such as housing, handling etc. This could improve test reproducibility, however, given the uncertainty about which factors to standardize, further standardization might not be successful. Differences in the number of mice housed together can be found over the different studies discussed. Often male mdx mice are housed together with 2–5 littermates (mdx and wild-types) (Sekiguchi et al., 2009; Chaussenot et al., 2015; Vaillend and Chaussenot, 2017; Dickson and Mittleman, 2019), other times mdx and wild-type mice are caged separately but still socially with other male mice (Vaillend et al., 1995). In our study we decided to cage male mice individually. The reasoning behind this comes from observed aggression in male mice and since our mice are individually housed for an extensive period in the PhenoTyper cages, it is deemed more suitable to house male mice individually to reduce the potential aggressive behavior toward cage mates before and after the behavioral testing. Mdx mice have been shown to have an increased amount of aggressive encounter with wild-type mice after a period of single housing (Miranda et al., 2015). It is possible that the single housing influenced the mouse behavior and resulted in the observed discrepancy.

Furthermore, if a particular phenotype in mice can only be detected under very specific conditions, the question remains how generalizable any conclusions generated in such a model can be. Here we used mdx mice, the most used mouse model for DMD (Yucel et al., 2018). This model is however less severely affected than DMD patients and thus does not fully recapitulate human disease progression. Although muscle pathology is less severe, behavioral abnormalities such as increased stress and freezing response have been detected in this strain (Sekiguchi et al., 2009; Miranda et al., 2015; Comim et al., 2019; Razzoli et al., 2020). Other mouse models for DMD are available and could be interesting to study behavioral and learning abnormalities. The mdx^4cv mouse model lacks multiple dystrophin isoforms and has a lower number of revertant fibers, fibers containing spontaneously restored dystrophin, than mdx mice. Although the muscle pathology appears to be very similar to that of the mdx and still less severe than DMD patients, the absence of multiple dystrophin isoforms makes it an interesting model to assess the effect of loss of one or more dystrophin isoforms on cognition and learning abilities (Yucel et al., 2018). Loss of multiple isoforms have been implied to result in a higher incidence of comorbidity with intellectual and emotional disabilities in DMD patients (Banihani et al., 2015; Ricotti et al., 2016). Another interesting mouse model for DMD is the D2-mdx mouse model. This mouse model was obtained by crossing mdx mice onto a DBA/2J background and resulted in a more severe muscle pathology compared to mdx mice (van Putten et al., 2019). Little is known about the effect of loss of full-length dystrophin on the behavioral phenotype of these mice.

To conclude, we do not see differences in spontaneous behavior and cognitive flexibility in mdx mice regardless of their gender. Since behavioral differences between mdx and wild-type mice seem to be subtle, large sample sizes, the search for more sensitive tests and the availability of standard operating procedures optimized for behavioral testing in mdx mice should be encouraged.

Raw data is publicly available at public.sylics.com.

The animal study was reviewed and approved by Animal Ethics Committee of VU University and performed according to Dutch regulation for animal experimentation, and in accordance with EU Directive 2010/63/EU.

MvP, AA-R, and ML contributed to the study conception and design. Experiments were performed by MvP, ML, and BK. Data was analyzed by SE, MvP, ML, and BK. Data was discussed by SE, AA-R, MvP, and ML. The manuscript was written by SE and MvP. All authors contributed to the reviewing and editing of the manuscript.

ML and BK are full-time employees of Sylics (Synaptologics B.V.), a privately owned company providing services using automated home-cages. AA-R does not disclose anything related to this work. For full transparancy she discloses being employed by LUMC which has patents on exon skipping technology, some of which has been licensed to BioMarin and subsequently sublicensed to Sarepta. As co-inventor of some of these patents AA-R is entitled to a share of royalties. AA-R further discloses being ad hoc consultant for PTC Therapeutics, Sarepta Therapeutics, CRISPR Therapeutics, Summit PLC, Alpha Anomeric, BioMarin Pharmaceuticals Inc., Eisai, Astra Zeneca, Santhera, Audentes, Global Guidepoint and GLG consultancy, Grunenthal, Wave and BioClinica, having been a member of the Duchenne Network Steering Committee (BioMarin) and being a member of the scientific advisory boards of ProQR, Eisai, hybridize therapeutics, silence therapeutics, Sarepta therapeutics, and Philae Pharmaceuticals. Remuneration for these activities is paid to LUMC. LUMC also received speaker honoraria from PTC Therapeutics and BioMarin Pharmaceuticals and funding for contract research from Italpharmaco and Alpha Anomeric. Project funding is received from Sarepta Therapeutics. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.