Establishment of a novel mouse model expressing GR50
Mouse models expressing GR>80 display various disease phenotypic profiles13,14. However, it remains unclear whether lower repeat lengths confer a measurable toxicity, in line with the C9 expansion penetrance pattern27-29. To identify lower pathogenic GR lengths for in vivo evaluation, we first assessed the survival of GR-expressing cells in vitro. Previously, our lab has demonstrated length-dependent toxicity for different DPRs, including GR, expressed in cortical and motor neurons up to 7 and 5 days in vitro (DIV), respectively9. Within these time frames, we previously saw reduced survival in cortical neurons expressing GR of lengths greater than 100 repeats, as well as differing toxicity profiles between cortical and motor neurons. Interestingly, we later reported that heterologous expression of the DPR GA50 caused a delayed toxicity in cortical and motor neurons under observation for up to 14 DIV despite its lower repeat length16. This led us to question if this phenomenon held true for GR as well: if expressing GR with a lower number of repeats in cells observed over an extended time frame would also confer an overt toxic profile. To this end, we expanded our window of observation of GR-expressing cells to a total of 14 DIV. Cortical neurons and motor neurons were transiently transfected with a combination of eGFP or three different dipeptide lengths: GR25-, GR50-, and GR100-GFP, with a synapsin-driven cell-filling td-Tomato construct9. Cortical neurons (CNs) were transfected when deemed fully mature at DIV7, while motor neurons (MNs) were transfected at DIV5 as per our published paradigm9 (Supplementary Fig. 1). Cells were visualized daily for 14 days and those that were double-positive for GFP and td-Tomato fluorescence signal, indicating neuron-specific transfection, were assessed for survival. Indeed, with this expanded window of observation, a length dependent toxicity is observed in both cortical (Fig. 1C) and motor neurons (Fig. 1D), with GR100-eGFP-expressing cells showing the most rapid and robust reduction in survival. GR50-GFP-expressing cortical and motor neurons also display reduced survivability when followed over this extended time frame. This outcome is reflected in the loss of cortical and motor neurons over time, as shown in Fig. 1A and 1B. GR50-GFP-expressing cortical and motor neurons display a significantly higher risk of death (HR 1.455 for CNs, 1.529 for MNs) than cells expressing GR25-GFP (HR 0.9553 for CNs, 0.9883 for MNs) but not as robust as GR100-GFP-expressing cells (HR 2.441 for CNs, 2.062 for MNs) (Tables 1A and 1B). We concluded that GR25 did not display a significantly different toxic profile compared to the eGFP control group, whereas GR100 showed rapid and robust neurotoxicity as also shown by HR values in both cortical and motor neurons. Based on this in vitro analysis we identified GR50 as a candidate dipeptide length for in vivo exploration with the rationale that its expression in vivo would indeed elicit chronic toxicity, but not so robustly that it would possibly prohibit a detailed longitudinal analysis and a temporal separation of different disease-related phenotypes.
To determine whether GR50 expression could cause C9 ALS/FTD-related phenotypes in vivo, we employed a strategy that utilized a Flexible Accelerated Stop Tetracycline Operator (FAST) cassette system, which allows for multiple genetic manipulations under a single cassette23. The FAST cassette is inserted at the ROSA26 locus in C57BL/6 mice, a well-characterized safe-harbor site used for gene insertion24,30. This cassette relies on the ROSA26 promoter25,26; together, this combination facilitates ubiquitous expression of the FAST cassette transgene in mice. The ROSA26 promoter is a low-to-mild expression driver in comparison to other vertebrate promoters, including EF1a31. The FAST cassette includes a floxed STOP codon upstream of an ATG-driven FLAG-GR50-eGFP, or in the case of the control, FLAG-eGFP (Fig. 2A). When crossed with a constitutively active Cre mouse line that expresses Cre-recombinase driven by the ubiquitous promoter CAG (CAG-Cre), the STOP codon is excised, facilitating expression of the downstream FLAG-GR50-eGFP or FLAG-eGFP. In choosing this specific promoter system, we kept in mind that although our system expressed one DPR of one length, the C9orf72 expansion itself is ubiquitous in nature32. We also chose to use the ATG start codon ahead of FLAG in both the GR50-GFP and control-GFP mice so as not to rely on RAN-translation for protein expression. Furthermore, the GR50-encoding sequence consists of randomized alternative codons rather than the GGCCGG repeat sequence, eliminating the potential pathogenic effects of the formation of GC repeat-rich RNA transcripts. This design allowed us to assess the consequences of GR50 expression without intentionally introducing confounding LOF or other GOF disease mechanisms.
When crossed with CAG-Cre mice, the offspring that were heterozygous for both CAG-Cre and either FLAG-GR50-GFP (GR50-GFP) or FLAG-eGFP (control-GFP) were assessed. To ensure that our transgenic mice expressed the expected proteins of interest, we performed western blots using antibodies specific for GR or GFP on cortex and spinal cord tissue homogenates from 3-month-old mice. When a GR-specific antibody is used, we detect GR bands at approximately 37 kDa in the GR50-GFP mice, but not control-GFP animals (Fig. 2B, upper panel). This banding pattern is consistent with the one seen in HEK cell lysate transiently transfected with FLAG-GR50-GFP that we used as reference. When the same blot was stripped and re-probed, a GFP specific band is detected in both GR50-GFP and control-GFP mice at different heights as expected (Fig. 2B, lower panel). Control-GFP mice express a band at approximately 25 kDa while the band seen in GR50-GFP mice remains at approximately 37 kDa. Because the ROSA26 promoter confers ubiquitous expression, we also probed for GR or GFP expression in other CNS and non-CNS tissues of GR50-GFP and control-GFP mice (Supplementary Fig. 2). Homogenate from the cortex, cerebellum, spinal cord, and liver successfully show GR and GFP expression in GR50-GFP mice (Supplementary Fig. 2A) and positive GFP expression in control-GFP mice (Supplementary Fig. 2D). Interestingly, while GR50-GFP mice express relatively equal protein amounts across the neuronal tissues assessed, GR expression in the liver of GR50-GFP mice is significantly higher (Supplementary Fig. 2B, C). This higher liver expression is not readily observed in control-GFP mice, though equal expression across neuronal tissues is observed (Supplementary Fig. 2E). Nevertheless, liver tissue appeared healthy without visible signs of hepatotoxicity (Supplementary Fig. 2F).
In addition to understanding the expression pattern of GR50 across neuronal and non-neuronal tissues, we also wanted to understand the cell-type specific expression of GR50 across the CNS. We explored cell-type specificity by relying on the GFP fluorescence of GR50 in our mouse model, which we confirmed is expressed in CNS regions (Fig. 2C). We identified cells positive for both GR50-GFP and either s100b, iba1, or NeuN for astrocytes, microglia, and neurons, respectively (Fig. 2D). Throughout cortical layers, GR50-GFP is expressed in approximately 90.76% of neurons, 31.56% of microglia, and 31.20% of astrocytes (Fig. 2E). When visualizing transverse sections of lumbar spinal cord, GR50-GFP is expressed on average in 69.34% of neurons, 26.25% of microglia, and 9.42% of astrocytes (Fig. 2F).
Multiple groups have described the varied localization pattern of GR both in vitro9 and in vivo13,14,22, where GR exists in both a diffuse and aggregated form and can localize to both the nucleus and cytoplasm of an expressing cell. A detailed analysis of GR50-GFP expressing cells throughout the CNS in our GR50-GFP mice at 3 months of age identified three types of localization patterns, similar to classes described in a previous study14 (Fig. 2G-I). In our mouse, cells expressing GR50-GFP diffusely and localized to the cytoplasm (Type I cells) comprise of 73.13% of GR50-GFP expressing cells in the cortex and 89.02% in the lumbar spinal cord. Type II cells, while also demonstrating a diffuse, cytosolic localization, also show evidence of perinuclear accumulation as visualized in the nuclear folds, and comprise of 19.40% of GR50-GFP expressing cells in the cortex and 2.44% in the lumbar spinal cord. The remaining 7.46% of cells in the cortex and 8.54 in the cytosol are type III, consisting of aggregated GR50-GFP in the cytosol. While we did not reliably detect an aggregated, nuclear GR50-GFP signal, it is of note that endogenous poly-(GR) is expressed primarily in the cytoplasm of expressing CNS cells in patient tissue22,33,34, and the rare occasion of nuclear detection in patient tissue seems to be an artifact of a-specific antibody crossreactivity22. These localization patterns and frequencies differ from those observed in vitro (Supplemental Fig. 1C-E), as primary cortical and motor neurons transiently transfected with FLAG-GR50-GFP display the following localization parameters: Type A cells express a diffuse, cytosolic localization pattern much like type I cells in vivo, and comprise of 26.13% of primary cortical neurons and 25.13% of primary motor neurons. Most primary cortical and motor neurons are type B, consisting of a diffuse, cytosolic localization with evidence of nuclear aggregates, at 58.79% and 58.97%, respectively. Type C cells, much like type III cells in vivo, consist of cytosolic aggregates and is seen in 7.04% of primary cortical neurons and 8.21% of motor neurons. Type D cells consist of only nuclear aggregates and are observed in 8.04% of primary cortical neurons and 7.69% of primary motor neurons.
GR50-expressing mice display sex-dependent histopathological differences.
We then investigated if GR50 expression could confer neurotoxicity in our mouse model. We assessed the consequences of said expression in CNS tissues from GR50-GFP mice compared to control-GFP mice. One of the most accepted pathological hallmarks in C9 ALS/FTD is the mis-localization of the nuclear protein TDP-43 to the cytoplasm in affected cells35. Importantly, the nuclear-to-cytoplasmic mis-localization of the ALS-relevant pathogenic marker TDP-43 is not robustly observed in GR50-GFP mice throughout the CNS at 12 months of age (Supplementary Fig. 4). We also did not observe any reduction in survival between sexes and genotypes in the test cohort over the course of 12 months (Fig. 3A). Neither female nor male GR50-GFP mice show evidence of robust neuron loss in the cortex as measured by NeuN expression compared to control-GFP mice between 3 and 12 months of age (Supplementary Fig. 3A, B). Female and male GR50-GFP mice also show no evidence of specific layer V motor neuron loss (Supplementary Fig. 3A, B).
While GR50-GFP females show no sign of overt neuron loss in the lumbar spinal cord at the same time-points as measured by NeuN expression (Supplementary Fig. 3D), there is a slight but significant reduction at 12 months in GR50-GFP males compared to control-GFP males (Supplementary Fig. 3E, F). These observations led us to assess cell-type specific differences in expression between GR50-GFP and control-GFP mice. As loss of motor neurons along the corticospinal tract is a hallmark of ALS pathogenesis, we investigated if motor neurons specifically were lost in GR50-GFP mice. When visualizing the ventral horn of lumbar spinal cord, we see a reduction in choline acetyltransferase (ChAT)-positive motor neurons in GR50-GFP female mice at 6, 9, and 12 months compared to control-GFP females (Fig. 3B). Interestingly, male GR50-GFP and control-GFP mice do not differ in the number of ChAT-positive motor neurons in the ventral horn at the same timepoints (Fig. 3C).
Sex-dependent differences are also seen at the level of astrocyte activation in CNS regions (Fig. 3D). Female GR50-GFP mice show increased glial fibrillary acidic protein (GFAP) activation in the ventral horn grey matter of the lumbar spinal cord at 3 months of age, prior to the time-points at which ChAT-positive cells were visibly lost (Fig. 3E). The cortex of female GR50-GFP mice expresses a higher GFAP signal compared to control-GFP females at later timepoints (Fig. 3E). Male GR50-GFP mice do not have increased GFAP activation in the lumbar ventral horn grey matter nor in the cortex compared to control-GFP males (Fig. 3F). Interestingly, between 3 and 12 months, we observe variable GFAP intensity in the spinal cord of both GR50-GFP and control-GFP females. Although this was not expected, dynamic GFAP signal along the course of disease has been observed in other neurodegenerative diseases36.
In addition to changes in GFAP expression and motor neuron counts, both males and females demonstrate increased evidence of the pro-apoptotic marker cleaved caspase-3 activity within the lumbar spinal cord as measured by cytosolic puncta accumulation over time37-39, correlating with some of the neuron loss observed in the spinal cord (Fig. 3G, H). However, cleaved caspase-3 accumulation is not seen in the cortex of male and female mice (Supplementary Fig. 3G, H).
Differences in sciatic nerve function absent of demyelination are observed in female GR50-expressing mice.
We next assessed if the histopathological consequences of GR50 expression along with the observed sex-specificity translated into functional consequences reminiscent of ALS progression. As motor unit dysfunction and loss correlates with ALS progression, we recorded the compound muscle action potential (CMAP) along the sciatic nerve in GR50-GFP and control-GFP mice as a measure of motor unit integrity and function40 (Fig. 4A). Waveform amplitude was measured as an indication of the number of functional nerve fibers and degree of muscle bulk, while the nerve conduction velocity was used as evidence of myelination state, as reduction in nerve conduction velocity may indicate a loss of signal-insulating myelin41.
We performed CMAP recordings in a cohort of male and female GR50-GFP and control-GFP mice at two different time-points corresponding with the initial onset of motor neuron loss in female GR50-GFP mice (3 versus 6 months old). In males, both genotypes show no difference in the amplitude when stimulated from the distal site (Fig. 4B) at both the 3-month and 6-month time-point, comparable with the lack of specific motor neuron loss observed in male mice. Distal amplitude is reduced in 3-month-old GR50-GFP females compared to control-GFP females, suggestive of distal axonal dysfunction at this age (Fig. 4B, D). At 3 months, however, female mice do not demonstrate significant differences in neuromuscular junction (NMJ) integrity, as shown by the degree of presynaptic synaptic vesicle 2 (SV2) and postsynaptic a-bungarotoxin signal overlap (Fig. 4F, G). In both male and female GR50-GFP and control-GFP mice, no difference in nerve conduction velocity is observed (Fig. 4C), suggesting that demyelination did not contribute to observed deficits.
Interestingly, when the same cohort of female GR50-GFP mice underwent 6-month CMAP recordings, the deficit in distal amplitude is no longer observed and is comparable to control-GFP females (Fig. 4B, E). This observation could suggest a developmental delay as opposed to a pure distal axonal deficit, or motor neuron compensation since we report a reduction in ChAT-positive neurons in 6-month-old female mice. When a separate cohort of GR50-GFP and control-GFP females was evaluated at 12-months-old, there was still no difference in either the proximal or distal amplitudes nor the nerve conduction velocity, nor was there a difference in NMJ innervation state (Supplemental Fig. 5).
Mice expressing GR50 display sex-dependent differences in gait dynamics.
When examining GR50-GFP and control-GFP animals, it was difficult to discern the genotype as the mice appear indistinguishable. This observation, combined with the functional and histopathological differences observed led us to hypothesize that there may be subtle, but measurable, locomotor differences that could be quantified. Impaired gait is one locomotor measurement that may not be so obvious to the experimenter unless the gait discrepancies were overt. Changes in gait dynamics have been observed in ALS patients in conjunction with their decline in motor function42,43, therefore we considered this an appropriate metric to measure in our mouse model to identify a more subtle relevant phenotype.
Gait analysis was performed on GR50-GFP and control-GFP animals using the Digigait quantitative gait assessment apparatus. The mice briefly ran on a clear treadmill to gather video footage of their steps. This footage was then analyzed for over 40 measurable gait indices, or features, associated with each of the four limbs. Because we aimed to understand if an overall difference in the gait of GR50-GFP and control-GFP mice could be measured rather than a change in a particular feature, we performed a principal component analysis (PCA) as a dimension reduction tool to identify trends and groupings among the genotypes. A PCA performed on Digigait data generated from the male and female mice at 12-months-old again shows a sex and genotype-specific difference in gait dynamics. The underlying variance of GR50-GFP females naturally separates from control-GFP only in the hindlimb gait features, indicative of a distinct difference in the hindlimb gait dynamics of GR50-GFP females compared to controls (Fig. 5A). Males, however, do not demonstrate a natural separation in variance for any limbs, suggesting that gait differences could not be detected between male GR50-GFP and control-GFP animals (Fig 5B).
Mice expressing GR50 display a subtle motor and behavioral deficit.
Finally, we sought to determine if GR50 expression translates to a measurable motor and behavioral phenotype in mice. Prior to behavioral assessments, animals were weighed to determine if potential weight loss should be considered a confounding factor. While female mice do not show a difference in weight over time between genotypes, GR50-GFP males display increased weight gain compared to control-GFP males at 12-months-old (Fig. 6A). Mice were subject to motor assessment using an accelerated rotarod performance task. Both male and female GR50-GFP mice have a reduced latency to fall compared to control-GFP animals (Fig. 6B). While rotarod performance is affected in GR50-GFP expressing animals, neither male nor female hindlimb grip strength show significant differences at 3, 6, and 9 months of age (Fig. 6C). Nevertheless, while sex-dependent histopathological and functional differences are observed, both sexes expressing GR50-GFP display a locomotive deficit.
As C9orf72-ALS lies on a genetic and clinical spectrum with FTD, we also tested whether expression of GR50 correlated with a measurable cognitive impairment. We relied on the open field assessment to identify both changes in locomotion and evidence of general anxiety behavior. Male GR50-GFP mice do not significantly differ from control-GFP males in the total distance traveled in the chamber, their average speed in the chamber, or the time they spent in the center of the chamber at 3, 6, and 9-months-old (Fig. 6D-G). Female GR50-GFP and control-GFP mice do not significantly differ in their average speed, though GR50-GFP females trend lower at those same timepoints (Fig. 6I). GR50-GFP expressing females travel a shorter total distance in the chamber at 9-months-old than control-GFP females (Fig. 6H). At 3-months-old, GR50-GFP female mice spend less time in the center zone of the chamber, evidence of increased anxiety compared to control-GFP females of the same age (Fig. 6J, K).