Impact of limb phenotype on tongue denervation atrophy, dysphagia penetrance, and survival time in a mouse model of ALS

Abstract

Current treatments for dysphagia in ALS do not target the underlying tongue weakness and denervation atrophy that is prevalent in spinal and bulbar ALS cases. To address this clinical gap, we studied the low copy number SOD1-G93A (LCN-SOD1) mouse model of ALS to quantify the impact of limb phenotype on tongue denervation atrophy, dysphagia penetrance, and survival time in preparation for future treatment-based studies. Two male LCN-SOD1 breeders and 125 offspring were followed for limb phenotype inheritance, of which 52 (30 LCN-SOD1 and 22 wild-type/WT, both sexes) underwent characterization of dysphagia penetrance (via videofluoroscopic swallow study; VFSS) and survival time at disease end-stage (15-20% body weight loss). From these, 16 mice (8/genotype) underwent postmortem histological analysis of the genioglossus for evidence of denervation atrophy. Results revealed that both breeders displayed a mixed (hindlimb and forelimb) ALS phenotype and sired equal proportions of hindlimb vs. mixed phenotype offspring. Dysphagia penetrance was complete for mixed (100%) versus incomplete for hindlimb (64%) phenotype mice; yet survival times were similar. Regardless of limb phenotype, LCN-SOD1 mice had significantly smaller genioglossus myofibers and more centralized myonuclei compared to WT mice (p < 0.05). These biomarkers of denervation atrophy were significantly correlated with VFSS metrics (lick and swallow rates, p < 0.05) but not survival time. In conclusion, both LCN-SOD1 phenotypes had significant tongue denervation atrophy, even hindlimb phenotype mice without dysphagia. This finding recapitulates human ALS, providing robust rationale for using this preclinical model to explore targeted treatments for tongue denervation atrophy and ensuing dysphagia.

Introduction

Progressive tongue weakness and denervation atrophy are predominant clinical features of ALS, eventually culminating in tongue paralysis that renders swallowing nonfunctional in the majority of cases.^(1-6) Remarkably, tongue weakness is one of the earliest clinical signs, regardless of ALS onset phenotype (bulbar or spinal).^{(7, 8)} The ensuing swallowing impairment (dysphagia) has a cascading negative effect on health outcomes, profoundly impacting quality of life (QOL)⁽⁹⁾ while increasing the risk of death by almost eight fold.⁽¹⁰⁾ It is therefore intuitive to suggest that early treatment of tongue weakness in ALS may significantly improve outcomes. However, existing dysphagia treatments for ALS are largely palliative and do not directly target the underlying hypoglossal motor unit degeneration;^{(7, 11-13)} that is, hypoglossal lower motor neurons (XII LMNs) in the brainstem, their corresponding axons, and the multiple tongue muscle fibers innervated by each axon. Even the two FDA-approved drugs riluzole and edaravone, which are aimed at slowing the progression of ALS, only marginally increase survival by ~2-3 months and have no beneficial treatment effect on tongue weakness and dysphagia.^(14-16)

To address this clinical need, we are leveraging a translational mouse model of ALS (low copy number SOD1-G93A, referred to herein as LCN-SOD1) to accelerate the discovery of targeted therapeutic strategies to beneficially alter hypoglossal motor unit degeneration in this relentlessly progressive and fatal disease. Our preliminary work with this model identified videofluoroscopic evidence of significantly diminished lick rate (i.e., tongue dysmotility), swallow rate, and longer pharyngeal bolus transit times during voluntary drinking at disease end-stage,⁽¹⁷⁾ resembling aspects of oropharyngeal dysphagia in ALS patients.^{(18, 19)} Corresponding XII LMN loss (i.e., hypoglossal motor neuron degeneration) and macroscopic evidence of tongue atrophy (i.e., smaller tongue surface measurements and reduced tongue weight) were also identified, thus providing robust clinico-pathological evidence of hypoglossal motor unit degeneration in this model.⁽¹⁷⁾ However, microscopic changes in the tongue muscle fibers have not yet been investigated for correlation with swallowing function and survival time, which was the primary aim of this study.

As a first step, we chose to investigate the genioglossus, an extrinsic tongue protrusor muscle innervated by the XII nerve.⁽²⁰⁾ Importantly, the genioglossus forms the majority of the tongue body and interdigitates with the other 7 tongue muscles;⁽²¹⁾ thus, it is intricately involved in all tongue movements and can therefore provide insight into overall tongue function/dysfunction in the LCN-SOD1 mouse model of ALS. Here, we focused on quantification of genioglossus myofiber cross-sectional area (CSA), as this measurement is widely used as a histological biomarker of muscle denervation atrophy and treatment response in ALS models.^(22-25) We also assessed other morphological changes commonly associated with denervation atrophy in ALS, including centralized myonuclei, angular myofibers, pyknotic myonuclear clumps, and compensatory hypertrophy.^(25-29) Our primary goal was to determine the impact of ALS phenotype on genioglossus denervation atrophy.

Our prior work with this model identified a heterogeneous pattern of limb involvement (similar to human ALS), which was reduced for simplicity to two ALS phenotypes based on limb involvement at disease end-stage: hindlimb (i.e., only the hindlimbs are affected) and forelimb (i.e., forelimb ± hindlimb paralysis).⁽¹⁷⁾ Here, we extended this work by characterizing the onset and spread of limb deficits across the adult lifespan in LCN-SOD1 mice, and then specifically investigated the effect of limb phenotype on tongue denervation atrophy, dysphagia penetrance, and survival time. We hypothesized that forelimb-affected LCN-SOD1 mice would display more severe genioglossus atrophy and correspondingly more severe dysphagia and shorter survival times compared to mice with only hindlimb involvement. Rationale was based on our previously reported nonsignificant trend for more severe dysphagia in a small sample size (n=5) of forelimb-affected LCN-SOD1 mice.⁽¹⁷⁾ This hypothesis is also in line with the general knowledge that ALS tends to progress by neuroanatomical contiguity in humans, such as between the upper limb and bulbar regions, and that life expectancy is shorter for ALS patients with bulbar involvement.^(30-32) Thus, mice with only hindlimb involvement may be less likely to develop bulbar signs/symptoms, such as dysphagia, which may in turn allow them to live longer.

Finally, we were curious to know if ALS phenotype relative to limb involvement is inherited in this model. That is, can we predict the limb phenotype of LCN-SOD1 offspring based on the phenotype of the affected parent? Doing so would permit allocation of offspring into phenotype-specific treatment groups at weaning (i.e., immediately following genotyping at ~1 month of age), rather than waiting several months for clinical phenotype to become apparent. We envision this predictive “power” would allow us to efficiently achieve balanced group sample sizes as we begin exploring targeted dysphagia treatment strategies aimed at delaying, slowing, or even preventing hypoglossal motor neuron degeneration in this translational mouse model of ALS.

Methods

Animals:

To investigate patterns of ALS phenotype inheritance, we followed the offspring of 2 randomly selected breeder pairs from our LCN-SOD1 colony, which was established in 2013 via the JAX cryorecovery program (The Jackson Laboratory, Bar Harbor, ME) by mating cryorecovered transgenic males [B6SJLTg(SOD1*G93A)dl1Gur/J; stock #002300]^{(33, 34)} with F1 hybrid females [B6SJLF1/J; stock #100012]. The colony has since been maintained by breeding male colony offspring with F1 hybrid females [B6SJLF1/J; stock #100012] purchased from the Jackson Laboratory ~2-3 times per year. The two breeder pairs in this study were allowed to breed undisturbed in ventilated caging in a barrier facility, beginning at 6 weeks of age, with the exception of routine cage changing and weaning of offspring. Following retirement from breeding at 7 months of age, the two male breeders were single-housed for weekly monitoring of body weight and clinical scoring until disease end-stage (described below).

Each breeder sired multiple litters totaling 125 offspring: Breeder 1: 59 offspring from 6 litters (30 males, 29 females); and Breeder 2: 66 offspring from 7 litters (38 males, 28 females). Offspring were weaned at 19-21 days of age and genotyped by quantitative real-time polymerase chain reaction (qPCR) of tail snips using our standard protocol.⁽¹⁷⁾ From the multiple litters, a total of 44 mice (i.e., just over 1/3 of the total offspring) were randomly selected to obtain a 50:50 ratio of LCN-SOD1 (n=22) and nontransgenic wild-type (WT) controls (n=22) of either sex for experimental testing. The remaining offspring (n=81) were allocated to colony breeding (males only), sentinels (for colony quality control purposes), or other studies that included end-stage phenotyping. Table 1 summarizes the group allocation of offspring for each breeder.

Table 1:

Breeders, Offspring, and Group Allocation of Offspring for Research Purposes

BREEDERS	Offspring		GROUP ALLOCATION OF OFFSPRING								TOTAL COUNT
			This Study		Breeders		Sentinels		Other Studies
	WT	LCN	WT	LCN	WT	LCN	WT	LCN	WT	LCN
#1	25	34	8	13	0	1	7	9	10	11	59
#2	34	32	14	9	0	2	7	8	13	13	66
Group Count	59	66	22	22	0	3	14	17	23	24	125
TOTAL COUNT	125		44		3		31		47

NOTE: WT = wild-type, non-transgenic mice; LCN = low copy number SOD1-G93A transgenic mice.

Following weaning, mice were group housed (based on sex) in ventilated microisolation cages under our lab’s standard vivarium conditions: ambient temperature (20-26 °C), humidity (30–70%), light cycle (12-hour standard), water bottle (filtered tap water, pH adjusted to 3.5), food pellets (Purina 5008 Mouse Diet), and enrichment materials (hut, running wheel, dental treats, and nestlet). Veterinary staff provided daily healthcare monitoring and routine surveillance for common rodent illnesses. All procedures were approved by our Institutional Animal Care and Use Committee and performed within our AAALAC-accredited academic institution, in accordance with National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Identification of Disease End-stage – Humane Study Endpoint

The endpoint criterion for this study was when transgenic mice reached 15% reduction from maximum body weight.⁽¹⁷⁾ At this advanced disease stage, transgenic mice display paralysis of one or more limbs but are able to independently forage for food and water in the home cage. Beginning at 6 weeks of age, body weight (grams, g) was recorded weekly for the 44 offspring using a digital scale (VIC-412, Acculab; Danvers, MA); data collection for the 2 male breeders began at 7 months. To avoid diurnal variations, weighing was performed at approximately the same time each day. When mice reached 10% weight loss, disease progression became accelerated and more frequent weighing (up to daily) was necessary to detect when mice crossed the 15% weight loss threshold. At that timepoint, individual transgenic mice and age-matched controls underwent behavioral testing, followed by euthanasia (typically within 1-2 days following behavioral testing to permit confirmation of test results) for postmortem assessment of the tongue (described below). Due to the required overnight water restriction for VFSS testing (described below), mice were typically closer to 20% weight loss at the time of euthanasia yet still able to independently forage for food and water. Thus, disease end-stage for this study corresponded with 15-20% body weight loss. Importantly, the pre-water restriction body weight (i.e., the night before VFSS testing) was used for statistical analyses in order to avoid confounding of body weight data due to water restriction. However, the age (months) at which mice were euthanized served as the survival age to correspond with histological data.

ALS Phenotyping based on Limb Involvement

Starting at 6 months of age, the 44 offspring underwent weekly clinical scoring to detect and monitor hindlimb and/or forelimb impairment over time;^{(17, 35, 36)} data collection for the 2 male breeders started at 7 months. Individual mice were assessed while freely exploring the home cage, while walking on a paper-lined benchtop, and during the hindlimb extension reflex when quickly lifted by the tail. Each limb was independently scored (by the same evaluator) relative to several phenotypic markers based on visual examination of posture (e.g., curled toes) and range of motion (e.g., retracted limb). Only the onset and end-stage clinical scores were analyzed to characterize patterns of disease onset and spread for each mouse relative to limb involvement. This information was then used to establish a phenotype classification scheme for use in statistical analysis for heritability, VFSS, survival, and histological data.

Videofluoroscopic Swallow Study (VFSS) at Disease End-stage to Characterize Dysphagia Penetrance

At disease end-stage, the 44 offspring (22 WT and 22 LCN-SOD1) underwent videofluoroscopic swallow study (VFSS) testing using our standard protocol and custom equipment.^{(17, 37-39)} Each mouse was tested only once to eliminate any potential confounding effects of serial X-ray exposure on normal muscle and nervous tissue biology. Prior to VFSS testing, mice were behaviorally conditioned to the test environment and equipment (without X-rays or contrast agent) once per week for approximately 6-8 weeks to minimize stress on test days, thus ensuring optimal behavior and reliable test data. Following an overnight (12-16 hour) water restriction to motivate participation, mice underwent VFSS testing in our miniaturized fluoroscope (The LabScope; Glenbrook Technologies, Newark, NJ) designed for use with rodents. A custom test chamber (14.4 x 5.0 x 5.0 cm polycarbonate tube with removable end-caps) was used to maintain freely-behaving mice in the lateral plane while voluntarily drinking our standard liquid contrast solution (Omnipaque, GE Healthcare, 350 mg iodine/mL; diluted to a 25% solution with deionized water and 3% chocolate flavoring) from a bowl positioned near the chamber floor. The chamber was secured to a custom remote-controlled platform that was adjusted as needed to maintain the head/neck region of each mouse in the fluoroscopy field of view. The bowl was filled/re-filled remotely using a custom syringe delivery system. To minimize radiation exposure, the fluoroscope was manually activated (~28 kV, ~0.2 mA) only when the mouse was actively drinking from the bowl, which was visualized in real-time via a webcam (Logitech, HD Pro C920) positioned above the chamber. Approximately one minute of active drinking per mouse was digitally recorded at 30 frames per second (fps) and saved as AVI files. After testing, mice were immediately returned to the home cage with unlimited access to food, water, and standard enrichment materials.

Prior to video analysis, three 2-second episodes of uninterrupted drinking were spliced from the raw video for each mouse using Pinnacle Studio (version 14; Pinnacle Systems, Inc., Mountain View, CA), with five frames added to each end to provide contextual information as needed for analysis. The video clips were subsequently analyzed frame-by-frame using our prototype VFSS analysis software, JawTrack™.⁽³⁹⁾ This software provides an interactive interface for semi-automated jaw motion tracking, manual labeling of swallow-related event locations based on bolus flow, and automatic computation of several VFSS outcome measures (metrics) based on event marker location; VFSS metrics are operationally defined in Table 2. Results are automatically displayed graphically and numerically in synchrony with the video clip to facilitate user verification and manual correction of the data. Using this interface, all clips were independently analyzed by two trained reviewers, and all discrepancies for event marker type and location were resolved by consensus before exporting the data (Excel) as averaged VFSS metrics for use in statistical analysis.

Table 2:

VFSS Metrics Quantified using JawTrack™ Software

VFSS Metrics	Operational Definitions	Units
Lick Rate	Number of jaw open/close cycles in each second of a 2-second video clip, converted to a rate (licks/second), then averaged.	#/s
Inter-lick Interval	Time between successive lick cycles throughout a 2-second video clip.	ms
Swallow Rate	Number of swallows in each second of a 2-second video clip, converted to a rate (swallows/second), then averaged.	#/s
Inter-swallow Interval	Time between successive swallow pairs throughout a 2-second video clip, then averaged.	ms
Lick-swallow Ratio	Number of jaw open/close cycles between each successive swallow pair throughout a 2-second video clip, then averaged.	n/a
Pharyngeal Transit Time	Bolus flow time through the pharynx for each successive swallow, then averaged. Start frame: “rest frame” immediately preceding visible transfer of the bolus from the vallecula (swallow trigger point). End frame: when the bolus tail enters the proximal esophagus.	ms

NOTE: All video clips depicted uninterrupted drinking behaviors, beginning with a swallow event (i.e., rest frame immediately preceding bolus flow from the vallecula). Units: s=second, ms = millisecond, # = number, and n/a = not applicable.

Postmortem Histological Assessment of the Genioglossus for Evidence of Denervation Atrophy

For histological analysis of the genioglossus muscle, a subsample of the 44 mice was randomly selected to permit comparisons between genotypes (i.e., a 50:50 mix of WT and LCN-SOD1) as well as ALS phenotypes (i.e., equal group sample sizes). Only mice that drank during VFSS were included to permit correlation analyses. Within two days after VFSS testing, mice were euthanized with an overdose of sodium phenobarbital and transcardially perfused with 0.9% NaCl (saline) followed by 4% paraformaldehyde (PFA; in 0.1 M phosphate buffered saline, pH ~ 7.4). The entire tongue was carefully removed and post fixed in 4% PFA for 2 days at 4 °C with rotation. The genioglossus muscle was then dissected from the tongue under a dissecting microscope (Leica M80 Stereo Microscope) and divided into anterior, middle, and posterior sections (Fig. 1), based on the directionality of the muscle fibers. Although the left and right sides were not distinguishable after dissection, both sides remained attached at the midline septum to produce a bilateral sample for processing. Additionally, the geniohyoid muscle remained attached to the posterior genioglossus (whenever possible) to serve as a reference point for microscopy imaging.

Fig. 1. Genioglossus dissection for histological analysis.

Lateral (a) and inferior (b) views of a representative whole mouse tongue with labeled structures. (c) Partial dissection of the genioglossus from the other tongue muscles (except the geniohyoid) to demonstrate technique. (d) The dissected genioglossus prior to its division into anterior, middle, and posterior sections, with the geniohyoid included for anatomical reference during histological analysis. (e) Separation of the genioglossus into anterior, middle, and posterior (with geniohyoid) sections prior to paraffin processing and histological staining. Black boxes denote the portion of the genioglossus sample containing cross-sectional myofibers; tissue outside the boxes denotes the tongue blade that serves as a base for vertical orientation of the samples during paraffin embedding. Scale bars (3 mm) and/or US penny (19 mm diameter) are shown for size reference. GG = genioglossus; GH = geniohyoid.

The tissue samples were paraffin processed and embedded using our lab’s standard protocol which included dehydration in a graded series of ethanol solutions, clearing in citrisolv, and paraffin processing and embedding. Tissue blocks underwent serial-sectioning (10 μm cross-sections) using an automated microtome (Leica RM 2155, Germany). Ribbons containing 4-5 sections were collected onto positively charged (silane-coated) slides, totaling 15-20 slides per mouse. The sections were viewed via light microscope (Leica DM750) to pre-identify 10-12 slides per mouse for subsequent immunohistochemistry (IHC) or hematoxylin and eosin (H&E) staining. The selection criteria included bilateral posterior genioglossus samples with circular-shaped cross-sectional myofibers throughout the entire sample. Slides meeting these criteria were subsequently baked for 2 hours at 55 °C and stored in slide boxes at room temperature until needed for staining.

IHC Staining and Analysis:

Three to 5 slides per mouse underwent laminin IHC for visualization of the basement membrane of individual myofibers, closely following our established protocol⁽⁴⁰⁾ that included a pepsin antigen retrieval step (0.5% pepsin in 5 mM HCL for 10 minutes at 37°C) and utilized a Shandon Sequenza Immunostaining Center (Thermo Fisher Scientific, Waltham, MA). The primary antibody solution consisted of 5% NDS, 1XPBS, 0.1% Triton, and an anti-laminin antibody (rabbit polyclonal, 1:100, Millipore Sigma, Burlington, MA). The secondary antibody solution consisted of 5% NDS, 1X PBS, 0.1% Triton, and secondary antibody (donkey anti-rabbit Alexa-Fluor 488, 1:1000; Molecular Probes, Eugene, OR). Slides incubated without primary and secondary antibodies served as negative controls. Stained slides were coverslipped using Prolong Gold Antifade Mountant (Molecular Probes) to prevent fluorescence quenching, and subsequently air-dried flat in the dark before storing at 4°C in slide boxes pending fluorescence microscopy.

The laminin-stained genioglossus sections were viewed via the 10X objective of an epifluorescence microscope (Leica DM 4000, Germany) for subjective grading and ranking of the staining quality (excellent, good, fair, or poor) for each slide. The 2-3 best slides from each mouse were subsequently imaged at 5X using Leica imaging software (Leica Application Suite v4.3) and saved as TIF files. Within each image, two non-overlapping regions of interest (ROIs) were added – one on each side of the midline septum, spaced equidistant, and positioned at approximately the same location in each section (Fig. 2a). For each mouse, two sections with the best fitting bilateral ROI boxes were imaged at 20X (Fig. 2b), converted to 8-bit resolution, and thresholded between 15-30^{(40, 41)} to permit semi-automatic outlining of laminin-positive myofibers using ImageJ software.⁽⁴²⁾ Outlining errors were manually corrected using a digital tablet and stylus (Wacom CTH460 Bamboo Pen and Touch Tablet, Portland, Oregon) before generating myofiber count and CSA data (Fig. 2c). To avoid including blood vessels and other irrelevant structures, analysis was limited to CSA values between 40 and 1500 μm² and a circularity greater than 0.3.^{(40, 43)} Of the two sections per mouse, the one with the highest myofiber count bilaterally was chosen as the representative data for use in statistical analysis. This approach ensured that each myofiber was counted only once per mouse, thus eliminating double counting bias that may confound histological findings.

Figure 2. Posterior genioglossus region of interest (ROI) for cross-sectional area (CSA) quantification following laminin staining.

(a) Representative 5X overview image of a bilateral posterior genioglossus sample with labeled components. The two ROIs per sample are denoted by white boxes (613 X 458 μm) flanking the midline septum (dashed white line); each ROI captured approximately half of the tissue sample area for each side of the posterior genioglossus. (b) Representative 20X image of ROI #2, with laminin staining (green) of the basement membrane surrounding each myofiber. (c) Myofiber CSA quantification using ImageJ software. Scale bars (yellow lines) = 200 μm.

H&E Staining and Analysis:

Three to four slides per mouse underwent Gill’s H&E staining using our standard protocol⁽¹⁷⁾ for qualitative assessment of morphological biomarkers of muscle denervation atrophy in ALS: centronucleation, angular myofibers, pyknotic myonuclear clumps, and compensatory hypertrophy (Table 3).^(25-29) Stained slides were cover slipped using Permount (Fisher Scientific) and air dried flat at room temperature before proceeding with microscopy. Slides were viewed at 20X using a brightfield microscope (Leica DM4000) to identify the 3 best representative genioglossus sections per mouse, based on subjective grading. Next, two independent reviewers manually scored each representative section at 40X relative to the occurrence of each biomarker in the field of view: 0 = none, 1 = few (≤5), and 2 = marked (>5). Scoring discrepancies were resolved by group consensus. For each mouse, the modal score for each biomarker was used in statistical analysis. TIF images were acquired at 40X using Leica imaging software (Leica Application Suite v4.3).

Table 3:

Morphological Biomarkers of Muscle Denervation Atrophy

Biomarker	Operational Definition
Centronucleation	Myonuclei become centrally (rather than peripherally) located in individual myofibers; hallmark of regenerating myofibers.
Angular myofibers	Polygonal-shaped myofibers; indicative of denervation.
Pyknotic myonuclear clumps	Clusters of myonuclei with little or no visible cytoplasm (i.e., chromatin condensation); indicative of severe muscle fiber atrophy.
Compensatory hypertrophy	Surviving/unaffected myofibers become enlarged (often with clear centers or visible “targets”) to compensate for nearby degenerating/atrophic myofibers.

Statistics

Investigators involved with data collection were blinded to genotype until the study database was created in IBM SPSS Statistics 24, which was used to perform all statistical analyses. Basic summary statistics, boxplots, and Q–Q plots revealed that all VFSS, survival, and histological variables of interest violated the assumptions of normality and homogeneity of variance; therefore, nonparametric tests (Mann-Whitney U, Kruskal–Wallis, Wilcoxon signed-rank, one-sample binomial test, and Spearman’s rank correlation) were used for statistical analyses as appropriate. For all variables, outliers were identified and re-checked for accuracy, but not removed from the dataset. Averaged values were used for VFSS analysis, whereas raw data were used for survival, phenotype inheritance, and histological analyses. Two-tailed p-values were reported, with p values of ≤0.05 considered statistically significant. Bonferroni adjusted p-values were used for pairwise comparisons.

Results

ALS Phenotypes & Heritability

As expected, all mice carrying the mutant SOD1-G93A transgene developed limb involvement (i.e., LCN-SOD1 group, n=22), whereas mice without the transgene retained normal limb function (i.e., WT group, n=22). For the LCN-SOD1 mice, weekly clinical scoring revealed two distinct anatomical sites of onset [i.e., hindlimb (n=19) versus forelimb (n=3)] and three general patterns of anatomical disease spread [i.e., hindlimb to hindlimb (n=14), hindlimb to forelimb (n=5), and forelimb to hindlimb (n=3)] affecting one or both sides, as shown in Fig. 3. None of the offspring developed a purely forelimb phenotype (i.e., forelimb onset, without progression to involve the hindlimbs). For simplicity (and to increase group sample sizes for statistical analyses), we classified the hindlimb-hindlimb pattern as “hindlimb” phenotype, and the other two patterns involving the hindlimbs and forelimbs as “mixed” phenotype. This dichotomous phenotypic classification scheme was then used to explore differences in swallowing function, survival, and genioglossus atrophy (described below).

Figure 3. ALS phenotype classification system based on pattern of limb onset and progression.

This scatterplot shows the pattern of disease onset and spread relative to limb involvement in individual mice (n=44). Note that all mice that did not develop limb deficits were WT controls (n=22), confirmed by genotyping. LCN-SOD1 mice (n=22; also confirmed by genotyping) displayed variable patterns of limb involvement, which were classified into two categories for this study: hindlimb (n=14) and mixed (n=8). Interestingly, all of the forelimb onset mice progressed to involve the hindlimbs, resulting in no mice with a purely forelimb phenotype at disease end-stage.

This same phenotype classification scheme was applied to the two male breeders for heritability analysis. Both male breeders displayed a mixed ALS phenotype (i.e., hindlimb and forelimb involvement) at disease end-stage. Importantly, both breeders sired offspring of either hindlimb (Breeder 1: n=10; Breeder 2: n=4) or mixed (Breeder 1: n=3; Breeder 2: n= 5) phenotype, demonstrating that limb phenotype is not predictable for offspring of mixed phenotype breeders. However, for a more comprehensive assessment of phenotype heritability, we included the other 81 offspring that were allocated to other research purposes (see Table 1) but were monitored for end-stage phenotype. The phenotype distribution of the 125 offspring is detailed in Table 4, which shows the relatively equal distribution of hindlimb and mixed phenotype offspring for each breeder: Breeder #1: 16 hindlimb, 12 mixed; Breeder #2: 13 hindlimb, 16 mixed. A one-sample binomial test confirmed that hindlimb and mixed phenotypes occurred with equal probabilities (p > 0.05). The phenotype distribution also was similar between male and female offspring, which was confirmed using a Pearson Chi-Square test: X² (1, n = 57) = 2.959, p = 0.085. Note that 10 LCN-SOD1 mice (i.e., 7 from Breeder #1 and 3 from Breeder #2) were classified as “unknown” phenotype because they did not reach disease end-stage due to a variety of non-ALS health issues that typically affect 5-10% of our colony (e.g., dermatitis, urogenital issues, circling behaviors, etc.), necessitating euthanasia at a younger age. Also of note, there were no purely forelimb phenotype mice sired by either of the two mixed phenotype breeders.

Table 4:

Limb Phenotype is not Predictable for LCN-SOD1 Offspring from Mixed Phenotype Breeders

BREEDERS (Mixed Phenotype)	WT		LCN-SOD1						TOTAL COUNT
			Hindlimb		Mixed		Unknown
	M	F	M	F	M	F	M	F
#1	9	15	6	10	8	4	6	1	59
#2	13	21	5	8	9	7	1	2	66
Group Count	22	36	11	18	17	11	7	3	125
TOTAL COUNT	58		29		28		10

NOTE: WT = wild-type, non-transgenic mice; LCN = low copy number SOD1-G93A transgenic mice; M = male; F = female.

Dysphagia Penetrance based on VFSS Metrics

Of the 44 mice that underwent VFSS testing at disease end-stage, 6 mice (3 WT and 3 mixed phenotype) did not drink and were therefore excluded from the study. For the remining 38 mice (19 WT and 19 LCN-SOD1), VFSS videos were analyzed by the two trained reviewers using the interactive interface of our JawTrack™ software (Fig. 4). Jaw maximum open/close marker locations required manual correction approximately 5% of the time (either by adding omitted markers, deleting extraneous markers, or adjusting markers by 1 frame in either direction), thus demonstrating high (~95%) jaw tracking accuracy by our prototype software compared to manual ground truth data (i.e., frame-by-frame validation by the two reviewers). Reviewer agreement was ~98% for all event markers (i.e., location of maximum jaw open/close position, swallow onset, and PTT end frame) before consensus, providing additional validation of the raw data and derived VFSS metrics.

Figure 4. Quantification of VFSS metrics using JawTrack™.

Representative radiographic images of a WT (a) and LCN-SOD1 (d) mouse voluntarily drinking liquid contrast from a bowl, with jaw tracking markers positioned on the upper (yellow) and lower (blue) jaw via our JawTrack™ software. The swallow trigger point is depicted by bolus accumulation in the vallecular space. Note the similar drinking posture for both mice, except for a visibly obvious curled forepaw for the LCN-SOD1 mouse. Representative jaw tracking plots with labeled events of interest are shown for a WT (b) and LCN-SOD1 (c) mouse. Green and red dots depict jaw maximum open versus closed position, respectively. The purple shaded boxes depict the 2-second region of interest for analysis, starting with a swallow event (pink line). Manually added event markers include “swallow onset” (pink line) and “end of pharyngeal transit” (blue line). The 6 VFSS metrics were automatically quantified using event marker data; representative metrics are labeled: ILI = inter-lick interval (ms); ISI = inter-swallow interval (ms); PTT = pharyngeal transit time (ms); LSR: lick-swallow ratio (number of licks between 2 successive swallows). The two unlabeled VFSS metrics include lick rate (number of licks/second) and swallow rate (number of swallows/second). Within each purple region, all possible measurements for each of the 6 VFSS metrics were averaged; this process was repeated for 3 separate video clips (30 frames per second) per mouse, and the averaged values for each VFSS metric across the 3 clips were used for statistical analyses. Of note, the jaw open/close distance for the representative LCN-SOD1 mouse appears to be diminished; however, this metric remains to be validated. Radiographic calibration marker (black line) = 10 mm; C2 = 2^nd cervical vertebra; ms = milliseconds.

Using the averaged, consensed VFSS metric data, sex differences within each phenotype [19 WT: 5 male, 14 female; 14 hindlimb: 6 male, 8 female; and 5 mixed: 3 male, 2 female] were explored using the Mann-Whitney U test, which revealed no significant difference between males and females for each of the 6 VFSS metrics (p > 0.05). Therefore, data from both sexes were combined within each phenotype and VFSS metric for all subsequent statistical analyses. Differences in VFSS metrics between LCN colony phenotypes (WT, hindlimb, and mixed) were investigated using the Kruskal–Wallis test. Results showed that 5 of the 6 VFSS metrics were significantly different between phenotypes: lick rate [X²(2) = 23.548, p < 0.0001], inter-lick interval [X²(2) = 23.123, p < 0.0001], swallow rate [X²(2) = 11.383, p = 0.003], inter-swallow interval [X²(2) = 9.959, p = 0.007], and pharyngeal transit time [X²(2) = 16.753, p < 0.0001]. The nonsignificant VFSS metric was lick-swallow ratio (p > 0.05), a measure not dependent on time. For the 5 statistically significant VFSS metrics, Bonferroni post hoc comparisons revealed that both hindlimb and mixed LCN-SOD1 phenotypes were significantly different from WT controls (p < 0.05). Specifically, both LCN-SOD1 phenotypes had significantly slower lick (hindlimb: p < 0.0001; mixed: p = 0.004) and swallow rates (hindlimb: p = 0.044, 1-sided; mixed: p = 0.006), longer inter-lick (hindlimb: p < 0.0001; mixed: p = 0.004) and inter-swallow (hindlimb: p = 0.049; mixed: p = 0.023) intervals, and longer pharyngeal transit times (hindlimb: p = 0.001; mixed: p = 0.016) compared to controls. Within LCN-SOD1 mice, none of the VFSS metrics were significantly different between hindlimb and mixed phenotypes (p < 0.05), suggesting that both LCN-SOD1 phenotypes develop a similar dysphagia profile. These VFSS results are summarized in Fig. 5, which shows the WT normative range for each VFSS metric and the corresponding cut-off values for identifying deficits in LCN-SOD1 mice (lick rate: <8.5 Hz; inter-lick interval: >120 ms; swallow rate: <1.5 Hz; inter-swallow interval: >870 ms; and pharyngeal transit time: >100 ms). Not shown is lick-swallow ratio (i.e., the only non-significant VFSS metric relative to genotype), which had a normative cut-off value of >6.8.

Figure 5. Videofluoroscopic evidence of dysphagia in both ALS phenotypes.

The data distribution for the five statistically significant VFSS metrics is visually displayed as boxplots (median, quartiles, and whiskers, mean = X) with adjacent dotplots (individual data points) for the 3 phenotypes (WT, hindlimb, and mixed). Compared to WT controls, both hindlimb and mixed LCN-SOD1 phenotypes had significantly altered swallowing function at disease end-stage, characterized by significantly slower lick (a) and swallow (c) rates, and longer inter-lick (b) and inter-swallow (d) intervals. The red dashed line demarcates the normative boundary for each swallow metric, with the red arrow indicating the direction of the deficit for LCN-SOD1 mice relative to WT values. Note that lick rate had the highest penetrance (9/14 hindlimb mice and 4/5 mixed phenotype mice). Also note that none of the 5 VFSS metrics were significantly different between hindlimb and mixed phenotypes. Data points outside the whiskers are considered mild outliers (filled diamonds); there were no extreme outliers. Group samples sizes: WT (n=19), LCN-SOD1 (n= 19; 14 hindlimb and 5 mixed phenotypes). Significance levels: p > 0.05 (ns nonsignificant), p ≤ 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****).

Spearman rank correlation analysis was performed on the 5 significant VFSS metrics for the combined LCN-SOD1 phenotypes (hindlimb and mixed) to determine the strength and direction of the association between pairs of rank-ordered variables. Results revealed a significant negative association between the following metrics: lick rate and inter-lick interval (rs = −0.981, p < 0.0001), and swallow rate and inter-swallow interval (rs = −0.938, p < 0.0001). These findings were expected because frequency and time period are inversely proportional to each other and thus are correlated by definition. In this case, faster lick rates correspond with shorter time between successive licks (i.e., inter-lick intervals), and faster swallow rates correspond with shorter time between successive swallows (i.e., inter-swallow intervals). Lick and swallow rates were not significantly correlated (p > 0.05), and pharyngeal transit time was not correlated with any other VFSS metric (p > 0.05).

Given the small sample size of the mixed phenotype group (n=5), we explored whether a larger sample size would alter the above findings. To do this, we identified 8 additional mixed phenotype offspring serving as colony sentinels and proceeded with end-stage VFSS testing, which increased the mixed phenotype sample size to 13 mice to more closely match the hindlimb group (n=14). When the Kruskal–Wallis test was repeated with this larger sample size, 2 of the 6 VFSS metrics emerged as significantly different between hindlimb (n=14) and mixed (n=13) LCN-SOD1 phenotypes: swallow rate (p = 0.038) and inter-swallow interval (p = 0.039, 1-sided). Specifically, swallow rate was slower and inter-swallow interval was longer for mixed compared to hindlimb phenotype LCN-SOD1 mice, as shown in Fig. 6.

Figure 6. Increasing the mixed phenotype sample size reveals novel VFSS findings.

Adjacent boxplots and dotplots show the data distribution of the two swallow-based VFSS metrics that were significantly different after increasing the mixed phenotype group sample size. Specifically, mixed phenotype mice had significantly slower swallow rates (a) and longer inter-swallow intervals (b) compared to hindlimb phenotype mice. Data points outside the whiskers are mild outliers (filled diamonds). The red dashed line and arrow indicate the normative boundary (relative to WT values) and “deficit direction” for each swallow metric. Group samples sizes: hindlimb (n=14) and mixed (n=13). Asterisk denotes statistical significance (p ≤ 0.05); X = mean.

Table 5 summarizes the penetrance of each VFSS metric deficit relative to limb phenotype and sex for this larger LCN-SOD1 sample size, rank ordered from highest to lowest penetrance. Note the moderate to high penetrance of lick rate and inter-lick interval deficits, which was highest for mixed phenotype LCN-SOD1 females. Moreover, mixed phenotype females were the only group with deficits in all 6 VFSS metrics (i.e., 60% penetrance), followed by 5/6 metrics for mixed phenotype males (39% penetrance) and hindlimb phenotype females (27% penetrance). Hindlimb phenotype males had the lowest penetrance (25%) -- only 3 of the 6 VFSS metrics were affected (lick rate, inter-lick interval, and inter-swallow interval). A Kruskal-Wallis test revealed a significant difference in ranks between phenotypes relative to the frequency of deficits in swallow rate [X²(1) = 6.790, p = 0.009], inter-swallow interval [X²(1) = 6.215, p = 0.013], and lick-swallow ratio [X²(1) = 4.870, p = 0.027] but not lick rate, inter-lick interval, and pharyngeal transit time (p > 0.05). Thus, 3 VFSS metric deficits (swallow rate, inter-swallow interval, and lick-swallow ratio) occurred more frequently for mixed phenotype LCN-SOD1 mice, whereas 3 VFSS metrics deficits (lick rate, inter-lick interval, and pharyngeal transit time) had a similar frequency of occurrence between hindlimb vs. mixed phenotypes.

Table 5:

Ranking of VFSS Metric Deficit Penetrance in LCN-SOD1 Mice

VFSS Metric Deficit (normative cut-off values)	LCN-SOD1				TOTAL PENETRANCE (VFSS deficits)
	Hindlimb (n=14)		Mixed (n=13)
	M	F	M	F
Lick Rate (<8.5 Hz)	4/6 (67%)	5/8 (63%)	2/3 (67%)	10/10 (100%)	21/27 (78%)	1
Inter-lick Interval (>120 ms)	4/6 (67%)	4/8 (50%)	2/3 (67%)	9/10 (90%)	19/27 (70%)	2
Inter-swallow Interval (>870 ms)	1/6 (17%)	1/8 (13%)	1/3 (33%)	7/10 (70%)	10/27 (37%)	3
Swallow Rate (<1.5 Hz)	0/6 (0%)	1/8 (13%)	1/3 (33%)	6/10 (60%)	8/27 (30%)	4
Lick-swallow Ratio (>6.8)	0/6 (0%)	0/8 (0%)	1/3 (33%)	3/10 (30%)	4/27 (15%)	5
Pharyngeal Transit Time (>100 ms)	0/6 (0%)	2/8 (25%)	0/3 (0%)	1/10 (10%)	3/27 (11%)	6
TOTAL PENETRANCE (Groups)	9/36 (25%)	13/48 (27%)	7/18 (39%)	36/60 (60%)	Rank Order (highest to lowest)
	4	3	2	1

NOTE: LCN = low copy number SOD1-G93A transgenic mice; M = male; F = female; n = group sample size. penetrancePenetrance of each VFSS metric deficit is indicated by cell color: white = 0% (no penetrance); light gray = 10-33% (low); dark gray = 50-70% (moderate penetrance); black = 90-100% (high penetrance).

Table 6 shows the VFSS metric deficits for each of the 27 LCN-SOD1 mice, of which 21 had a deficit in at least 1 VFSS metric; this equates to an overall dysphagia penetrance of 78% for this model. Importantly, 5 of the 6 LCN-SOD1 mice without a lick rate deficit also did not have any other VFSS metric deficits; one LCN-SOD1 mouse had deficits in swallow rate, inter-swallow interval, and lick-swallow ratio, but not lick rate. Note that mixed phenotype mice had 2-5 deficits per mouse, whereas hindlimb mice had greater variability (1-5 deficits per mouse).

Table 6:

VFSS Metric Deficits for Individual LCN-SOD1 Mice

LCN- SOD1 Mice	VFSS Metric Deficits						Total Deficits
	Lick Rate (<8.5 Hz)	Inter-lick Interval (>120 ms)	Swallow Rate (<1.5 Hz)	Inter- swallow Interval (>870 ms)	Lick- swallow Ratio (>6.8)	Pharyngeal Transit Time (>100 ms)
1	0	0	0	0	0	0	0	0
2	0	0	0	0	0	0	0
3*	0	0	0	0	0	0	0
4*	0	0	0	0	0	0	0
5*	0	0	0	0	0	0	0
6	1	0	0	0	0	0	1	1
7*	1	1	0	0	0	0	2	2
8	1	1	0	0	0	0	2
9	1	1	0	0	0	0	2
10	1	1	0	0	0	0	2
11	1	1	0	0	0	0	2
12*	1	1	0	0	0	0	2
13*	1	1	0	0	0	0	2
14*	1	1	0	0	0	0	2
15	1	1	0	0	0	0	2
16	1	1	0	0	0	0	2
17	1	1	0	0	0	1	3	3
18	1	1	0	1	0	0	3
19	0	0	1	1	1	0	3
20	1	0	1	1	1	0	4	4
21	1	1	0	1	0	1	4
22	1	1	1	1	0	0	4
23	1	1	1	1	0	0	4
24	1	1	1	1	0	0	4
25	1	1	1	1	0	1	5	5
26*	1	1	1	1	1	0	5
27	1	1	1	1	1	0	5
Total	21	19	8	10	4	3	65

NOTE: 0 = no deficit (white cells) and 1 = deficit (gray cells) relative to WT normative cut-off values for each VFSS metric. Yellow cells = mixed phenotype mice. Asterisks indicate individual mice allocated to histological assessments: hindlimb (3 mice without dysphagia and 1 with dysphagia) and mixed (4 mice with dysphagia).

Survival Time

At disease end-stage (i.e., 15-20% body weight loss), all mice retained the ability to forage for food and water in the home cage. Compared to WT controls, body weight was significantly reduced for hindlimb (p < 0.0001) and mixed (p = 0.001) phenotype mice (Kruskal-Wallis test with Bonferroni pairwise comparisons). However, both LCN-SOD1 phenotypes (hindlimb and mixed) had similar body weights and percent body weight loss at end-stage VFSS testing (p > 0.05, Mann-Whitney U test). Survival age (i.e., age at euthanasia; mean = 1.89 days after VFSS testing) was explored using the Mann-Whitney U test (with and without the 8 additional mixed phenotype mice), which revealed no significant difference between hindlimb and mixed LCN-SOD1 phenotypes (p > 0.05). The WT control group was excluded from survival analysis, as they were age-matched with end-stage LCN-SOD1 mice for the study endpoint; the typical lifespan of WT controls is ~2 years. The average survival age for the combined hindlimb and mixed phenotypes was 8.7 months (with and without the 8 additional mixed phenotype mice), which is consistent with prior reports for this model.^{(17, 34, 44)} Sex differences within each phenotype were explored using the Mann-Whitney U test, which revealed no significant difference in survival age between males and females (p > 0.05). These results are graphically summarized in Fig. 7. The association between survival age and lick and swallow rates (i.e., non-correlated VFSS metrics) were explored for LCN-SOD1 mice using the Spearman rank correlation test, which failed to detect any significant linear relationships (p > 0.05).

Figure 7. Survival age is similar between ALS phenotypes and sexes.

Adjacent boxplots and dotplots show the data distribution of survival age relative to ALS phenotype (hindlimb versus mixed) and sex. No significant differences were identified between phenotypes [hindlimb (median = 8.65; mean = 8.70) and mixed (median = 8.77; mean = 8.68)] or sexes [hindlimb male (median = 8.65; mean = 8.72); hindlimb female (median = 8.65; mean = 8.69); mixed male (median = 9.40; mean = 9.50); and mixed female (median = 8.45; mean = 8.43)]. Group sample sizes: hindlimb (n=14; 8 males and 6 females) and mixed (n=13; 10 males and 3 females). ns = nonsignificant (p > 0.05); X = mean.

Denervation Atrophy of the Posterior Genioglossus

From the 38 mice with VFSS data (19 WT, 14 hindlimb, and 5 mixed), 16 were included in histological assessment of the posterior genioglossus. The limiting factor was the mixed phenotype group with only 5 mice, one of which was excluded from histological analysis due to processing artifacts. The remaining 4 mixed phenotype mice were included along with 4 randomly selected hindlimb mice for equal representation of both LCN-SOD1 phenotypes in statistical analysis. Table 6 denotes the 8 LCN-SOD1 mice (asterisks) allocated to histological assessments: hindlimb phenotype = 3 mice without dysphagia and 1 with dysphagia; mixed phenotype = 4 mice with dysphagia. In addition, 8 WT mice were randomly selected for statistical comparisons with the 8 LCN-SOD1 mice (4 hindlimb and 4 mixed phenotype).

IHC Quantification of Genioglossus Myofiber CSA:

One bilateral posterior genioglossus section (i.e., 2 images, one from each side) from each of the 16 mice was used in statistical analysis. The number of myofibers per image ranged from 332 to 585 (WT = 332 - 585; LCN-SOD1 = 335 - 531), which was not significantly different between genotypes (Mann-Whitney U = 132; p > 0.05). Myofiber counts were standardized across the 16 mice by randomly selecting 325 myofibers per image,^{(24, 45)} which was just below the bottom of the range for all 18 mice combined. CSA data from both sides were combined for each mouse, resulting in 650 myofibers per mouse (i.e., 5,200 myofibers per genotype) for subsequent statistical analyses.

Myofiber size differences between phenotypes (8 WT, 4 hindlimb, and 4 mixed) were explored using the Kruskal-Wallis test, which revealed a statistically significant difference between groups [X²(2) = 215.048, p < 0.0001]. Based on Bonferroni post hoc comparisons, genioglossus myofiber CSA was significantly reduced for hindlimb (p < 0.0001) and mixed (p < 0.0001) phenotype mice compared to WT controls; however, there was no significant difference between the two LCN-SOD1 phenotypes (p > 0.05). Therefore, data from hindlimb and mixed phenotype mice were combined for all subsequent statistical analyses, which focused on genotype comparisons (WT versus LCN-SOD1).

Myofiber CSA values ranged from 40.4 to 514.58 μm² for WT mice (n=8; median = 139.50 μm², IQR = 115.63 μm²) and from 40.4 to 501.29 μm² for LCN-SOD1 mice (n=8; median = 117.69 μm², IQR = 102.48 μm²). As shown in Fig. 8, the CSA distribution was non-normal for each genotype (p < 0.001; Kolmogorov-Smirnov test of normality) and could be readily classified into small (≤100 μm²), medium (>100 but ≤200 μm²), and large (>200 μm²) myofiber sizes for statistical comparisons. Independent samples Mann-Whitney U tests revealed that compared to WT mice, LCN-SOD1 mice had a significantly higher percentage of small-sized myofibers (U = 1,334,675; p < 0.0001) and a correspondingly lower percentage of large-sized myofibers (U = 645,697; p = 0.002); however, the distribution of medium-sized fibers was essentially the same between genotypes (p > 0.05). Notably, LCN-SOD1 mice had a 36.6% increase in small-sized myofibers compared to WT mice.

Figure 8. End-stage LCN-SOD1 mice have a significant shift toward small-sized genioglossus myofibers.

The CSA values for the 5200 myofibers per genotype are shown as side-by-side histograms with corresponding myofiber size distribution percentages, categorized according to small, medium, and large-sized fibers. Note the non-normal distribution (gray dashed line = normal curve) for both genotypes, as well as the significantly higher percentage of small-sized myofibers (≤100 μm²) and correspondingly lower percentage of large-sized myofibers (>200 μm²) for LCN-SOD1 mice (n=8) compared to WT mice (n=8). Notably, LCN-SOD1 mice had 36.6% (40.3 - 29.5 / 29.5 X100) more small-sized (≤100 μm²) genioglossus myofibers compared to WT controls. Significance levels: p > 0.05 (ns = nonsignificant), p < 0.01 (**), and p < 0.0001 (****). CSA = cross-sectional area.

As a final step, associations between genioglossus myofiber CSA (median value), survival age, and lick and swallow rates (i.e., non-correlated VFSS metrics) were explored for LCN-SOD1 mice using the Spearman rank correlation test. A significant positive correlation was identified between myofiber CSA and lick rate (r_s = 0.454, p = 0.039, one-sided), and between CSA and swallow rate (r_s = 0.452, p = 0.039, one-sided), such that mice with slower lick and swallow rates also had smaller genioglossus myofiber CSAs. However, no significant linear relationship was detected between myofiber CSA and survival age (p > 0.05); thus, LCN-SOD1 mice with smaller genioglossus CSAs did not necessarily have the shortest survival times (data not shown).

H&E Quantification of Additional Morphological Biomarkers of Genioglossus Muscle Atrophy:

Mann-Whitney U tests revealed significant differences between genotypes relative to centralized myonuclei (U = 52.00, p = 0.038) but not the other three H&E-based biomarkers of myogenic atrophy (angular myofibers, pyknotic myonuclear clumps, and compensatory hypertrophy; p > 0.05). Specifically, all mice, regardless of genotype, had centralized myonuclei, whereas neither WT or LCN-SOD1 mice had grouped myofiber atrophy, pyknotic myonuclear clumps, or compensatory hypertrophy. Of note, angular fibers appeared to be more prevalent in LCN-SOD1 (75.0% of samples) compared to WT (37.5% of samples) mice; however, this difference was not statistically significant (p > 0.05). Interestingly, none of the H&E-based biomarkers of myogenic atrophy were correlated with genioglossus myofiber CSA (p > 0.05; Spearman rank correlation test).

Fig. 9 shows the frequency distribution of centralized myonuclei relative to phenotype. A Kruskal-Wallis test revealed a significant difference in ranks between phenotypes [X²(2) = 24.477, p < 0.001], with WT mice having the lowest ranking for centralized myonuclei, followed by hindlimb then mixed phenotype mice (WT = 22.88, hindlimb = 38.19, mixed = 46.06). Whereas the majority (7/8, 88%) of WT mice had “few” occurrences of centralized myonuclei, there was one WT mouse with a “marked” score. For transgenic mice, the frequency of “few” versus “marked” scores was 50:50 for hindlimb-affected mice, whereas 100% of mixed phenotype mice had “marked” centralized myonuclei.

Figure 9. Centralized myonuclei are more abundant in end-stage LCN-SOD1 mice.

(a) All mice, regardless of phenotype (WT, hindlimb, mixed), had evidence of centralized myonuclei indicative of ongoing myofiber remodeling in both healthy and disease conditions. However, myofiber remodeling was “upregulated” in LCN-SOD1 mice, particularly those with mixed phenotype, as indicated by the higher frequency of “marked” scores for LCN-SOD1 mice compared to WT controls. (b & c) Representative 40X images of H&E-stained genioglossus sections, with centralized myonuclei indicated by black arrows. Note the representative WT mouse in (b) had few (n=2) centralized myonuclei, whereas the representative mixed phenotype LCN-SOD1 mouse in (c) had marked (n=12) centralized myonuclei. Also note that each myofiber had only one centralized nucleus; there were no instances of multiple centralized nuclei per myofiber. (d) A magnified (100X) image of the black dashed box in (c) clearly shows centrally (black arrow) versus peripherally located myonuclei. Scale bar (black lines) = 50 μm. Group sample sizes: WT = 8, hindlimb = 4, mixed = 4. WT = wild-type/nontransgenic control mice; Hindlimb = LCN-SOD1 mice with only hindlimb involvement; Mixed = LCN-SOD1 mice with a combination of hindlimb and forelimb involvement.

Discussion

This preclinical study examined the impact of limb phenotype (hindlimb versus mixed) on tongue denervation atrophy, dysphagia penetrance, and survival time in the LCN-SOD1 mouse model of ALS. Three main findings were identified: 1) both LCN-SOD1 phenotypes had significant tongue denervation atrophy at disease end-stage, 2) dysphagia penetrance was incomplete for hindlimb (64%) versus complete for mixed (100%) phenotype mice, and 3) survival time was similar between hindlimb and mixed phenotype LCN-SOD1 mice. Thus, our hypothesis that LCN-SOD1 mice with forelimb involvement (i.e., mixed phenotype) would display more severe genioglossus atrophy and correspondingly more severe dysphagia and shorter survival times compared to mice with only hindlimb involvement did not hold completely true. While dysphagia was indeed more apparent in forelimb-affected mice (i.e., complete penetrance of the dysphagia phenotype), they did not necessarily have more severe tongue denervation atrophy or shorter survival times compared to hindlimb phenotype mice. Instead, we found that even hindlimb phenotype mice without dysphagia had significant tongue denervation atrophy, characterized by significantly smaller genioglossus myofibers and more abundant centralized myonuclei. This finding provides relevant translational insight into tongue weakness in human ALS, which is one of the earliest clinical signs, regardless of ALS onset phenotype (bulbar or spinal).^{(7, 8)} Thus, this preclinical model of ALS with clinico-pathological features of dysphagia is ideally suited for exploration of targeted treatments to mitigate tongue denervation atrophy and ensuing dysphagia in ALS.

A secondary goal of this study was to determine if ALS phenotype relative to limb involvement is heritable in this model. Given that the SOD1 mutation has an autosomal dominant pattern of inheritance in our mouse model as well as ALS patients,^{(46, 47)} we were curious to know if limb phenotype is also passed on from parent to offspring. That is, do LCN-SOD1 offspring inherit the same limb phenotype as the affected parent? Currently, we must patiently wait several months for accurate phenotype classification of offspring based on the anatomical pattern of limb onset and spread from clinical disease onset to end-stage (i.e., 15-20% body weight loss). An unfortunate consequence of this “waiting game” is unbalanced representation of ALS phenotypes at the end of the research study, as was the case here, which is known to diminish statistical power.^{(48, 49)} Therefore, we chose to necessarily expend additional time and resources to sufficiently increase our under-represented mixed phenotype group sample size to closely match that of the hindlimb phenotype group for improved statistical outcomes. The caveat of this approach is it violates the fundamental principle of animal research that advocates using as few animals as possible.⁽⁵⁰⁾ Thus, we were hopeful to overcome this dilemma in our future studies by accurately predicting each offspring’s limb phenotype based solely on the phenotype of the affected parent.

We were disappointed to find that limb phenotype inheritance was only chance level (~50%) for LCN-SOD1 mice. Specifically, both randomly selected male breeders displayed a mixed ALS phenotype (i.e., hindlimb and forelimb involvement) at disease end-stage and sired equal proportions of mixed versus hindlimb phenotype offspring. Whether this same chance level inheritance pattern holds true for hindlimb phenotype breeders remains unknown and is therefore a focus of our ongoing research efforts with this model. Nonetheless, our finding that both LCN-SOD1 phenotypes similarly developed tongue denervation atrophy suggests that pre-determination of limb phenotype is unnecessary for investigations exploring targeted treatment strategies to delay, slow, reverse, or even prevent hypoglossal motor neuron degeneration in this translational mouse model of ALS. This distinction, however, becomes crucial for behavioral-based dysphagia studies and those intent on identifying clinico-pathological correlations, as we have shown here that dysphagia penetrance indeed differs between phenotypes (mixed = 100%; hindlimb = 64%). These informative results will therefore guide appropriate research design of our future studies with this model.

Another unexpected finding was that survival times were similar between hindlimb and mixed phenotype LCN-SOD1 mice, which we attribute to our relatively early humane end-point criterion coinciding with 15-20% body weight loss. Importantly, mice at this stage are still ambulatory and able to forage for food and water in the home cage, although limb involvement is apparent. We strategically avoided using higher weight loss thresholds (e.g., 20 or 30%) or loss of the righting reflex (i.e., inability to return to sternal recumbency, typically within 5-30 seconds after placing the mouse on its side), based on our unpublished pilot work showing that mice at this more advanced disease stage cannot reach the food hopper and/or water bottle spout and quickly become moribund, necessitating euthanasia. While it is logical to speculate that relocating the food and water to (or nearer to) the cage floor would significantly extend survival to permit further characterization of phenotype differences in this model, we chose to reserve this interventional approach (i.e., compensatory strategy) for a future study rather than potentially confounding the results of this current study.

Another important, unexpected finding from this study was that one of the 6 hindlimb phenotype LCN-SOD1 mice without a lick rate deficit displayed other signs of dysphagia (i.e., reduced swallow rate, inter-swallow interval, and lick-swallow ratio), as shown in Table 6. Based on this finding, we cannot simply rely on lick rate to identify mice with versus without dysphagia. This is unfortunate, as we were hoping to incorporate lickometer testing into our future studies to provide a cost-effective and non-invasive screening tool to quickly identify and track LCN-SOD1 mice with dysphagia over time. However, we now have compelling rationale for continued use of VFSS as our primary diagnostic tool for accurate dysphagia detection in this model. Importantly, the identified videofluoroscopic signs/symptoms of dysphagia (e.g., slower lick and swallow rates, longer inter-lick and inter-swallow intervals, and longer pharyngeal transit times) in LCN-SOD1 mice mimic aspects of oropharyngeal dysphagia in ALS patients,^{(18, 19)} further demonstrating the translational value of this model in ALS research.

Study limitations

This foundational work has several notable limitations that are appropriately guiding our ongoing dysphagia research efforts. First, reasons for the increased number of smaller sized (<100 μm²) myofibers in the posterior genioglossus of LCN-SOD1 mice remain unclear. A similar shift to smaller sized myofibers was also shown for limb and diaphragm muscles in ALS mouse models, which was predominantly attributed to myofiber type transition; in this case, myosin heavy chain (MHC) type IIb (i.e., larger-sized, fast-twitch, fast fatigable) to type IIa (i.e., smaller-sized, slower-twitch, more fatigue resistant) fibers.^{(51, 52)} However, MHC expression patterns in the tongue (and craniofacial/bulbar muscles in general) are distinctly different from other skeletal muscles in the body in health and disease conditions,^{(53, 54)} a distinction that also holds true for mice.⁽⁵⁵⁾ Thus, what is generally known for the highly studied limb, abdomen, and respiratory skeletal muscles cannot be directly applied to the relatively less studied bulbar muscles, including the tongue. We are therefore proceeding with a variety of experimental approaches (e.g., IHC and RNA-sequencing) with LCN-SOD1 mice to understand what is driving the genioglossus myofiber size shift in this model – degenerative versus regenerative/remodeling processes (or a combination thereof). While our corresponding finding of significantly more abundant centralized myonuclei in LCN-SOD1 mice points toward an active degeneration/regeneration cycle,^(25-29) the lack of correlation between centronucleation and CSA is quite intriguing and suggests alternative or more complex processes may be at play. Investigating this discord may unmask molecular biomarkers of bulbar muscle denervation atrophy in ALS that may also be candidates for targeted dysphagia treatment interventions. Moreover, this ongoing pre-clinical work may shed light on the active role of bulbar muscles in ALS disease pathogenesis, which remains largely unknown.

Second, we have thus far characterized dysphagia only at disease end-stage in this model, at both behavioral and histological levels. Before proceeding with treatment studies, however, it is essential to understand the timing of dysphagia onset and rate of progression in this model, and how these may be correlated with limb deficits. Toward this goal, clinico-pathological studies are underway across the adult lifespan in this model. We expect this foundational step may provide abundant biomarkers (behavioral, cellular, and molecular) of dysphagia that may in turn become candidate treatment targets as well as objective indicators of treatment effectiveness.

Third, we identified five possible limb phenotypes describing the anatomical pattern of clinical disease onset and progression in this model (see Fig. 3). However, one of the possible forelimb phenotypes (forelimb onset without progression to the hindlimbs) was not observed in the mice included in this study. Importantly, we have identified rare occurrences of purely forelimb affected mice in our LCN-SOD1 colony, thus providing evidence that this ALS phenotype does exist but has a much lower penetrance than the others. Additionally, the other two possible forelimb-involved phenotypes (forelimb onset that progressed to involve the hindlimbs, and hindlimb onset that progressed to involve the forelimbs) had small sample sizes, which necessitated combining them into a single “mixed” phenotype group representing forelimb-affected mice either with or without concomitant hindlimb involvement. This approach resulted in a dichotomous ALS phenotype classification scheme representing mice with only hindlimb versus mixed (forelimb ± hindlimb) phenotype mice for statistical analyses. Remarkably, this classification scheme was sufficient to determine that mice with forelimb involvement are particularly susceptible to developing bulbar dysfunction, in this case, dysphagia. Indeed, 100% of the mice in the mixed phenotype group had videofluoroscopic evidence of dysphagia, compared to only 64% of hindlimb phenotype mice. This finding is in line with the increasingly reported, although poorly understood, pattern of contiguous regional symptom spread in human ALS (e.g., between the upper limb and bulbar regions).^(30-32)

Our ongoing efforts are therefore focused on increasing the sample sizes of each of the four limb phenotypes identified thus far in this model to determine if additional clinically relevant distinctions emerge relative to dysphagia. The ensuing knowledge has translational implications for human ALS, which also has a heterogeneous pattern of limb involvement that renders phenotypic classification quite challenging. Importantly, while the pathogenic mechanisms in ALS remain largely unknown, neuroanatomic propagation of the pathology is suspected to be a primary component.^{(32, 56, 57)} This hypothesis underscores the potential scientific contribution of the LCN-SOD1 mouse model in elucidating the pathogenesis of ALS.

Conclusions

We have significantly extended our prior work with the LCN-SOD1 mouse model of ALS by demonstrating novel evidence of tongue denervation atrophy that correlates with dysphagia (i.e., slower lick and swallow rates). Remarkably, both ALS phenotypes (i.e., hindlimb versus mixed) had significant tongue denervation atrophy compared to WT controls, even hindlimb phenotype mice without dysphagia. Dysphagia penetrance was incomplete (64%) for hindlimb phenotype mice but complete (100%) for the mixed phenotype group, thus rendering forelimb involvement a more reliable predictor of dysphagia in this model. Moreover, dysphagia diagnosis was based on videofluoroscopic quantification of oral motility/kinematic and pharyngeal bolus flow measures adapted from human VFSS testing, thus highlighting our translational intent. The impactful findings of this foundational study set the stage for utilizing this preclinical LCN-SOD1 model to accelerate the discovery of personalized medicine strategies for dysphagia in ALS.