Natural history study and statistical modeling of disease progression in a preclinical model of myotubular myopathy

ABSTRACT Generating reliable preclinical data in animal models of disease is essential in therapy development. Here, we performed statistical analysis and joint longitudinal–survival modeling of the progressive phenotype observed in Mtm1−/y mice, a reliable model for myotubular myopathy. Analysis of historical data was used to generate a model for phenotype progression, which was then confirmed with phenotypic data from a new colony of mice derived via in vitro fertilization in an independent animal house, highlighting the reproducibility of disease phenotype in Mtm1−/y mice. These combined data were used to refine the phenotypic parameters analyzed in these mice and improve the model generated for expected disease progression. The disease progression model was then used to test the therapeutic efficacy of Dnm2 targeting. Dnm2 reduction by antisense oligonucleotides blocked or postponed disease development, and resulted in a significant dose-dependent improvement outside the expected disease progression in untreated Mtm1−/y mice. This provides an example of optimizing disease analysis and testing therapeutic efficacy in a preclinical model, which can be applied by scientists testing therapeutic approaches using neuromuscular disease models in different laboratories. This article has an associated First Person interview with the joint first authors of the paper.

, gastrocnemius (GAS) and quadriceps (QUAD) muscles relative to body weight, from 5 week old wildtype and Mtm1 -/y mice. Unpaired t-test with Welsh's correction performed. (B) Correlation analysis of TA muscle mass/body weight ratio, relative to disease severity score (DSS) for 5 week old Mtm1 -/y (red) and wild type mice (black). Line of best fit and 95%CI highlighted in black for all mice (r= -0.7241; p < 0.0001), line of best fit for Mtm1 -/y alone (red line). Muscle mass presented alone (GAS-C; TA-E) or relative to body weight (GAS-D; TA-F), for wildtype and Mtm1 -/y mice injected with 6.25, 12.5 or 25mg/kg DYN101-m targeting murine Dnm2 reduction. (G) TA/ body weight ratio represented relative to DSS for wild type (black), Mtm1 -/y mice injected with 6.25 (orange), 12.5 (green), 25mg/kg (blue) DYN101-m. 5 week old Mtm1 -/y mice from (B) reproduced here for comparison purposes only (red dots, shading), and are not included in statistical analysis. Data represented as a violin plots (A,C-F), individual mouse data shown. Statistical analysis: (A) Unpaired t-test with Welsh's correction; Spearman correlation tests performed for (B), (G) Ordinary 1-way ANOVA followed by Dunnett's multiple comparisons test performed for (C)-(F). *p<0.05, **p<0.01, ***p<0.001,, ****p<0.0001.  Table S1. Previous disease severity scoring system (Tasfaout, Buono et al. 2017). A scoring system was set up to evaluate the clinical evolution of six centronuclear myopathy features. Difference of body weight between Mtm1 -/y versus WT littermate, ability to perform the hanging test, walking manner, presence or absence of ptosis and kyphosis and breathing difficulties (frequency and amplitude evaluation based on clinical observations) are recorded and a score of 0, 0.5 or 1 is given to each clinical readout. The sum represents the DSS. The higher the DSS, the more severe the phenotype, minimum 0 (healthy mouse), maximum 6 (severely affected mouse) (Tasfaout, Buono et al., 2017).  = min ( , ) and = ( ≤ ) = � 1 0 The longitudinal measurements and time to event were jointly measured via a latent bivariate process, which was realized independently in each subject The longitudinal sub-model is defined as: The observed response scaled between 0 and 1 ( = 1, … , = 1, … , ), is defined as : ~ � , �; where the beta distribution is defined by the parameters and , which are defined, respectively, by the mean and "the sample size" of the distribution as follows: The parameter is estimated from the data, and the mean is defined as a mixed model with logitlink function, where T is a constant to center time: The equation of the mean contains a random effect on the intercept 1 =~(0, 2 ). In turn, the survival sub-model is then defined via a Weibull survival model: Where expresses the induced association. A negative value of gamma implies that the larger value of the DSS Score the smaller the probability to stay in the study. A positive value would however imply the opposite. For simplicity in estimation, k = 1 in all estimations of the model indicating a constant failure rate. The model for the evolution of weight over time is similar to the one proposed here over.

Disease Models & Mechanisms • Supplementary information
The main difference lies in the equation for the mean. Indeed, the equation is assumed to be a linear gaussian model that evolves a the square root of time rather than a beta model.
The disease severity data from individual mice are included only until the point of death. The death of the mouse is then factored into the joint longitudinal survival model for disease severity scores. The above model used here allows the expected range to be presented following the correction for limited sample size, variability, and survival differences. The output is presented as the line of best fit (black line) and prediction intervals (shaded zone). A correct model fit occurs when 95% of the observed individual data is contained, for each time point, in the intervals depicted. Individual data from all living mice is shown as an overlay of the model to confirm this point on each graph, and thus the validity of the model. Tasfaout, Buono et al 2017, and subsequently optimized in Buono et al (associated manuscript).

Objective
The goal of this SOP is to detail the procedure of allocating a disease severity score (DSS) in mice. The DSS is a parameter which is used for the evaluation of the disease severity of the myopathic phenotype in Mtm1 -/y mice. The calculation is based on 4 parameters: difference of body weight, hanging test ability, kyphosis and walking difficulties. The minimal score is 0 (no myopathic phenotype) and the maximal score is 5 (characterizing a very severe myopathic phenotype).

Protocol 1) Body weight
Measure body weight, to 2 decimal places. The score for this parameter (score between 0 and 1) is the difference of a Mtm1 -/y mouse body weight from week n to n+1: Table 1: Body weight to score conversion Body weight SCORE X ≥ 0.25g 0 -0.25g > X > 0.25g 0.5 X ≤ -0.25g 1 X(g) = (body weight week n+1(g)) -(body weight week n(g))

2) Hanging test ability
This test must be done one mouse at a time: a. Take one type 3 cage and one grid. b. Place mouse in the middle of the grid c. Turn the grid upside down. The suspending animal should hold on to the grid in order to avoid falling.
• Test must be set up at a certain height, around 40 cm, for mouse to not being influenced to jump. d. Prevent the mouse from turning over to the other side of the grid or at pellets food and bottle place by barring with the hand (without touching the mouse) or by using the lid of a transport box. e. The latency to fall will be measured three times (60 seconds each) for each mouse, with a minimum interval of 10 minutes between trials.
• The latency time measurements begin when the mouse is hanging free on the wire and end with the animal falling to the cage underneath the wire or grid. • If performing the whole body hanging test for the first time, mouse can fall as soon as grip is turned upside down. If this is the case, the first assay can be considered as familiarization of the mouse with the testing conditions and will not be considered as one of the three assays of the test. Only time (seconds) for the three next trials will be reported in the dedicated table (Table 2). • If a mouse falls for any other reason that muscle strength default (eg if a mouse is not willing to do the test, voluntary jump from the grip, or falls because of your hands…), this will not be considered as a trial. This must be recorded in the note section in the dedicated table (Table 2) • A mouse should normally explore the grid. If a mouse stays in place and does not explore the grip. This must be recorded in the note section in the dedicated table (Table 2) f. Time (seconds) when mouse falls should be reported in the dedicated table (Table 2).
• 60 seconds is the maximum time allowed

3) Kyphosis
This parameter reflects the curvature of the spine. The score is noted as: Score 0: no curvature of the spine Score 0.5: mild curvature of the spine Score 1: severe curvature of the spine

4) Walking difficulties
This parameter reflects the ability to use hindlimbs and the ability of mice for walking. The score is noted as:

Calculation of DSS
Disease Severity Score (DSS) is the sum of the scores of the 4 parameters: body weight, hanging test ability, kyphosis and walking difficulties. The maximal DSS is 5.
Note: always refer to ethics application for humane endpoints.

Data Collection
All the data concerning the DSS will be reported in the dedicated table (table 3).

Body weight
Score 0-1 Difference in body weight between Mtm1 -/y mouse from week n to n+1.

Kyphosis
Score 0-1 Curvature of the spine Score 0: no curvature of the spine Score 0.5: mild curvature of the spine Score 1: severe curvature of the spine

Walking difficulties
Score 0-1 Ability to use hindlimbs Score 0: normal use of hindlimbs Score 0.5: splayed use of hindlimbs Score 1: loss of use of hindlimbs