Movement-area nerve cell death linked to LINE-1 activity in a mouse model of TDP-43 neurodegeneration involving both internal and external cell signals

We used transgenic mouse models that have previously been described to yield modest levels of human TDP-43 (hTDP-43-WT & hTDP-43-Q331K) expression (1.5x), and gradual appearance of neuro-pathological impacts [33]. To establish a baseline of the neurological phenotypes in our hands, we first quantified hindlimb clasping and latency to fall on the accelerated rotarod (Fig 1). As has been previously reported [33], we observe an age-dependent appearance of motor deficits that are more severe and occur early in life in the hTDP-43-Q331K transgenic animals than in the hTDP-43-WT transgenic animals. The hTDP-43-Q331K animals did not show significant differences in hindlimb clasping scores compared to the non-transgenic animals at 1.5 months (Fig 1b). However, they developed significant hindlimb clasping defects by three months of age (Fig 1c) (p = 0.0006 for the non-transgenic vs hTDP-43-Q331K comparison and p < 0.0001 for the hTDP-43-WT vs hTDP-43-Q331K comparison. Kruskal-Wallis test and Dunn’s multiple comparison tests were used to calculate p values with * p ≤ 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). This deficit persisted throughout their lifespan (Fig 1c–1f). The hTDP-43-WT mice have previously been reported to show no significant differences from the non-transgenic littermates in hindlimb clasping at 10 months of age [33]. Our findings concur with this conclusion. However, we additionally examined hindlimb clasping at 15 months and 17 months (Fig 1f–1g) and observed that the hTDP-43-WT mice showed significantly higher hindlimb clasping scores at 17 months when compared with non-transgenic littermates (p = 0.0022 using the Mann-Whitney test to calculate p values and ** representing p < 0.01). In terms of latency to fall, the hTDP-43-Q331K animals exhibited shorter latency (p = 0.0305 for the non-transgenic vs hTDP-43-Q331K comparison using ordinary one-way ANOVA) on the accelerated rotarod beginning at three months (Fig 1h and 1i), consistent with previously reported motor deficits for this genotype [33]. In contrast, the hTDP-43-WT mice did not show significant differences from the non-transgenic animals on the accelerated rotarod through 17 months (Fig 1j–1m). These results are similar to those seen in our double-heterozygous crosses to the L1-EGFP reporter mouse line (hTDP-43-Q331K/L1-EGFP, hTDP-43-WT/L1-EGFP, L1-EGFP controls) (S1 Fig; and see below). Thus, we observe early-onset motor deficits in hTDP-43-Q331K mice and late-onset deficits in hTDP-43-WT mice. These phenotypes are consistent with previous reports and further establish a late-onset motor dysfunction in the hTDP-43-WT mice [33].

Next, we examined changes in MC transcriptional profiles over a time course in hTDP-43-WT and hTDP-43-Q331K compared to non-transgenic control animals. We extracted total RNA from MC at 1.5, 3, 6, 10, and 15 months and performed bulk RNA sequencing using the NextSeq 500 platform. Given the literature that implicates RTEs in TDP-43-related neurodegeneration [12–15,18–22,34], we analyzed the data using a well-established computational pipeline that also permits examination of repetitive reads, such as those that derive from RTE expression [35]. The features were then computationally clustered based on Spearman correlation to group genes and RTEs with similar trends.

Overall, we observed a broad differential expression of RTEs and genes in the MC, with the most dramatic effects manifesting transiently at 3 months in the hTDP-43-Q331K and at 15 months in hTDP-43-WT animals when compared to non-transgenic littermates (S2 and S3 Figs). In each hTDP-43-Q331K and hTDP-43-WT genotype, these transient changes in expression of many cellular genes, RTEs, and ERVs occur mainly at the time when the behavioral deficits first appear, or just prior to that. These effects are clearly visualized via heatmaps that we generated using combined lists of genes and RTEs that are differentially expressed across all comparisons while accounting for age, utilizing a generalized linear model (Fig 2). Although RNA sequencing has not previously been performed for this mouse model, the differential expression identifies many cellular transcripts that have previously been seen in other models of neurodegeneration as well as in human tissues ([13,30,36–49]; S2–S13 Figs).

When examining effects on RTEs and ERVs, across the genotypes and ages, the differentially expressed elements break out into two main categories. First, there are a group of RTEs/ERVs whose expression increases with age irrespective of genotype (S12 Fig) (p_adj = 2.37E-08, odds ratio = infinite). This is consistent with a prior literature that establishes age-dependent derepression of many RTEs and ERVs [34,50–55]. Second, there are a set of RTEs/ERVs that are expressed at higher levels in the context of our neurodegeneration models. This group of elements show elevated expression in MC at the 3-month age in the hTDP-43-Q331K animals and at the 15-month age in the hTDP-43-WT animals. We also note that this group of RTEs and ERVs whose expression is induced in the context of TDP-43 neurodegeneration also have relatively high expression in all three genotypes in young animals that are just 1.5 months old compared to later ages (Fig 2a; (padj = 9.92E-76, odds ratio = 13.298)). This is consistent with the possibility that this set of elements that are impacted by neurodegeneration also are expressed during development [56–58] and exhibit residual RTE transcripts early in life.

The set of RTE and ERV elements that exhibit high expression at 1.5 months of age in control animals but low expression at older ages are largely the same subset that are significantly increased in the MC at 3 months in the hTDP-43-Q331K and at 15 months in the hTDP-43-WT transgenic animals (p_adj = 1.03E-15, odds ratio = 9.934). In the two TDP-43 transgenic groups, the most dramatic effects on both cellular transcripts (S4–S7 and S13 Figs) and RTEs/ERVs occur transiently, around the time the neurological defects emerge. For example, compared with non-transgenic littermates, the hTDP-43-Q331K animals show an overall elevation of many different RTEs, ERVs and cellular transcripts at the 3-month age (Figs 2a and S2), which correlates with the time-point when neurological defects occur in this genotype (Fig 1). Unlike the neurological effects, which progress and persist, these effects on the expression of RTEs and ERVs are transient. Although the increase in overall expression of many RTEs and ERVs at 3 months is largely transient, there are subsets of elements that show continued elevation across all timepoints (Fig 2c).

In contrast with hTDP-43-Q331K, the hTDP-43-WT animals showed no significant differences in RTE/ERV expression when compared to non-transgenic littermates at the 1.5, 3, 6, and 10-month time points. However, at 15 months, the hTDP-43-WT animals demonstrated a significant increase in 908 RTEs and ERVs (Fig 2c). Principal component analysis plots for significantly expressed RTEs and genes (S8–S10 Figs) show a clear separation of the hTDP-43-WT biological replicates from the other groups in PC1 (S8 Fig (i)). The hTDP-43-Q331K also shows a separation from the other samples with one outlier. The biological replicates for the 3-month and 15-month time points (across all genotypes) show clear separation from all the other time points on PC3 of the principal component analysis (S8 Fig (vii)).

The effects of hTDP-43-Q331K at 3 months and hTDP-43-WT at 15 months on RTE and ERV expression are quite broad, impacting each of the major clades of RTEs. Because LINE elements are derepressed in postmortem brain samples from human patients [13,19] and because a subset of LINE elements are replication competent, we next examined the effects of TDP-43 transgenes on the active subset of mouse LINE elements. We found that the overall pattern of differential expression of active mouse LINE elements matches that seen across the other types of elements: expression of active LINE elements increased in the MC of 3-month-old hTDP-43-Q331K animals and was significantly elevated at 15 months in the hTDP-43-WT animals (Fig 2b). These effects were similar across both evolutionarily recent and more ancient transposable elements (Fig 2a and 2c).

LINE-1 elements have previously been shown to be derepressed in postmortem cortical neurons that exhibit mislocalization of TDP-43 in FTD subjects [19]. In our RNA sequencing experiments, we observed that active LINE-1 elements are transiently upregulated around the time of neurological symptom onset, at 3 months in hTDP-43-Q331K and at 15 months in hTDP-43-WT (Fig 2b). Because both mouse and human genomes contain fully active copies of LINE-1 elements, their expression may result in retrotransposition. To investigate this possibility, we used an established L1-EGFP indicator cassette that expresses GFP only after the transgenic LINE reporter element has passed through an RNA intermediate and reinserted as a cDNA into a de novo chromosomal location [59,60]. We crossed the hTDP-43-Q331K and the hTDP-43-WT animals to the L1-EGFP reporter line to generate double-heterozygous offspring.

We performed immunohistochemistry in the MC of the hTDP-43-Q331K/L1-EGFP and hTDP-43-WT/L1-EGFP double transgenic animals, as well as their L1-EGFP littermates, to examine the presence of GFP-positive cells. For each genotype, we then imaged GFP-positive cells across a time-course. As expected from previous reports [60], we observed sparse GFP labeling in the L1-EGFP littermates of a relatively small number of neurons and glial cells in MC across all ages and genotypes (Figs 3a–3e and S14, S15). This is presumably the result of rare de novo retrotransposition of LINE-1 elements that can take place in neuronal and glial lineages during the last rounds of cell division in normal development. In the L1-EGFP animals, we see very few GFP-positive cells per section in MC across all ages. By contrast, in the hTDP-43-WT/L1-EGFP and hTDP-43-Q331K/L1-EGFP genotypes, we see spatially defined groups of many GFP-positive neurons and glia in the MC. Importantly, this effect has a stochastic onset, appearing in very few animals at 1.5 months of age, but in a higher fraction of animals at the 3- and 6-month timepoints (Fig 3a–3c and S1 Table). Because these large, spatially defined groups of L1-EGFP-labeled cells are rarely seen in the hTDP-43-Q331K/L1-EGFP and the hTDP-43-WT/L1-EGFP genotypes at 1.5 months of age but are more frequently detected in 3- and 6-month-old animals, we presume that they are due to independent retrotransposition events taking place after development. Such large groups of GFP-labeled cells are not observed in the L1-EGFP single transgenic genotype at any age.

But the presence of such large groups of labelled cells is a transient effect that is no longer detected at 10- and 15-months (Fig 3d and 3e and S1 Table). For example, at 3 months, the hTDP-43-Q331K/L1-EGFP animals exhibited significantly higher numbers of GFP-positive cells per section compared to their L1-EGFP littermate controls (C2 control in Fig 3b; p = 0.0406. Box-Cox transformation was done with a λ = 0.041 followed by One-Way ANOVA and student t-tests post hoc to compute values with * p ≤ 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001) (Fig 3b). The hTDP-43-WT/L1-EGFP animals also displayed significantly higher GFP-counts per section than the L1-EGFP littermate controls (C1 control in Fig 3; p = 0.0216) per section but with fewer cells compared to that seen in the hTDP-43-Q331K/L1-EGFP animals. The hTDP-43-WT/L1-EGFP exhibit relatively lower subject-to-subject variability in numbers of GFP expressing cells than the hTDP-43-Q331K/L1-EGFP animals, with a normal distribution (Std Dev = 5.740 for the hTDP-43-WT vs Std Dev = 11.88 for the hTDP-43-Q331K cohorts) (Fig 3b).

This effect on GFP-positive cells per section was also seen at 6 months age for both the hTDP-43-Q331K/L1-EGFP and the hTDP-43-WT/L1-EGFP animals, with significantly higher numbers of GFP-positive cells per section (p = 0.0157 in the hTDP-43-Q331K/L1-EGFP vs L1-EGFP littermates (C1) and p = 0.0410 for the hTDP-43-WT/L1-EGFP vs L1-EGFP-littermate comparison; C2) in the MC (Fig 3c). Notably, at 10 and 15 months, the number of GFP-positive cell counts per section returned to control levels (Fig 3d, 3e). Thus, both the hTDP-43-Q331K and the hTDP-43-WT animals exhibit higher numbers of cells expressing the L1-EGFP retrotransposition reporter in the MC at 3 and 6 months, but the presence of L1-EGFP-labeled cells becomes undetectable at later ages. The hTDP-43-Q331K/L1-EGFP animals exhibit these LINE-1 retrotransposition events in the MC at a time that correlates with the onset of motor dysfunction and closely follows the timing of the derepression of RTEs seen in the RNA sequencing experiments (Figs 1 and 2). In contrast, the hTDP-43-WT/L1-EGFP animals display LINE-1 retrotransposition events that occur much earlier than the onset of motor dysfunction (Figs 1 and 3b) and prior to when we observe RTE derepression in bulk tissue RNA sequencing (Fig 2). In addition to the MC, we also quantified GFP-positive cells in the striatum, nucleus accumbens, and agranular insular cortex (S14 and S16–S18 Figs). Significant increases in number of cells that exhibited LINE-1 retrotransposition events were observed in all these regions at 3 or 6 months in both genotypes (S15 Fig).

Although we observe an increase in the number of GFP-labeled cells in both the hTDP-43-Q331K/L1-EGFP and hTDP-43-WT/L1-EGFP animals at 3 and 6 months of age, we noted that the distribution of labeled GFP cells along the anteroposterior axis of the MC was quite different between the two genotypes. For example, at the 6-month age (Fig 4), the L1-EGFP-labeled cells in hTDP-43-WT/L1-EGFP animals are primarily restricted to the anterior portion of the MC. But in the hTDP-43-Q331K/L1-EGFP animals, we observe large groups of GFP-positive cells with a continuous anteroposterior distribution. We also noted that the GFP label in the more posterior sections of the hTDP-43-Q331K/L1-EGFP group included many cells with morphology characteristic of pyramidal neurons (Fig 4), whereas the more anterior sections appeared to contain a mix of cells with morphology indicative of both neurons and astrocytes (see below).

Having established that the hTDP-43-Q331K/L1-EGFP and the hTDP-43-WT/L1-EGFP animals exhibit increased numbers of cells with LINE-1 retrotransposition events in the MC, we next investigated if there were cell-type-specific differences in LINE-1 retrotransposition within the spatially clustered L1-EGFP expressing cells in the two double heterozygous lines. In order to quantify cell types within these regions, we defined a cluster of cells within a given brain region as a group of GFP-labeled cells whose count was three standard deviations higher than the mean GFP-labeled cell counts of that region in the L1-EGFP control. Although multiple brain regions exhibited GFP-positive cells (Figs 3 and S14, S15), the MC was the only region that exhibited such defined clusters.

To determine the identity of labeled cells within each cluster, we performed immunohistochemistry with EGFP together with either the pan-neuronal marker Neurotrace-Blue, the astrocyte marker S100β, or the microglial marker Iba1. We observed in both hTDP-43-Q331K/L1-EGFP and hTDP-43-WT/L1-EGFP genotypes that at 3 and 6 months, all MC GFP-positive cells were labeled with either Neurotrace-Blue or S100β (Fig 5a and 5b), indicating that the cell types labeled by the LINE-1 reporter are neurons and astrocytes. We did not find any GFP-positive cells labeled with Iba1 in the MC of any of the animals, indicating that microglia are not likely a site of L1-EGFP retrotransposition.

We also quantified the total number of GFP-positive cells per cluster. We found no significant difference in the numbers of L1-EGFP-positive neurons per cluster between the hTDP-43-WT/L1-EGFP and the hTDP-43-Q331K/L1-EGFP animals at 3 months. However, at 6 months, there are significantly more L1-EGFP-positive neurons per cluster in the hTDP-43-Q331K/L1-EGFP animals compared to the hTDP-43-WT/L1-EGPP animals (p = 0.0027 using ordinary one-way ANOVA with Sidak’s multiple comparisons test). We also observed that the number of LI-EGFP-positive neurons per cluster in the MC of the 6-month-old hTDP-43-Q331K/L1-EGFP animals was also significantly higher than that observed in both genotypes at 3 months (p = 0.0068 for the 6-monthhTDP-43-Q331K/L1-EGFP and 3-monthhTDP-43-WT/L1-EGFP comparison and p = 0.0128 for 6-monthhTDP-43-Q331K/L1-EGFP and 3-month hTDP-43-Q331K/L1-EGFP comparison) (Fig 5c). In contrast to the hTDP-43-Q331K/L1-EGFP animals, which show an increase in numbers of GFP-labeled neurons per cluster between 3 and 6 months, the hTDP-43-WT/L1-EGFP animals show no significant differences between those ages (Fig 5c). There were no significant differences in the number of GFP-positive astrocytes per cluster between 3 and 6 months for either genotype (Fig 5d).

We next examined the spatial distribution of the clusters of cells that exhibited LINE-1 retrotransposition events across the anteroposterior axis in MC and compared the distribution across brain hemispheres. To do so, we examined the organization of GFP-positive clusters in 14 sections at identical bregma locations in hTDP-43-Q331K/L1-EGFP and hTDP-43-WT/L1-EGFP animals. Each section was 50 microns thick, with a 150-micron interval between sections. We adhered to the cluster definition used previously to determine the presence or absence of clusters in the MC. We first determined if an animal had a defined cluster. In animals that passed this total cell count cluster threshold, we determined if subsequent anterior-to-posterior sections contained defined clusters and if they appeared in the ipsilateral or contralateral MC from the previous section cluster (Fig 6a. pseudo-left, RED or pseudo-right, GREEN designator, see methods). Section-wise cluster thresholds were based on the L1-EGFP mean+3SD for that specific bregma anterior-posterior coordinate from L1-EGFP littermate controls. The sections within each animal that did not pass the section-wise cluster are marked as yellow (Fig 6a, see legend).

At 3 and 6 months, the hTDP-43-Q331K/L1-EGFP mice typically displayed a single large unilateral cluster that spanned multiple sections along the anteroposterior axis of the MC. In contrast, the hTDP-43-WT/L1-EGFP mice typically exhibited multiple smaller clusters that were restricted to just a few sections. These animals often exhibited small clusters on both sides of the brain within a given section (Fig 6a). The distribution of these clusters of L1-EGFP-labeled neurons also differed over time. At 3 months of age, the hTDP-43-Q331K/L1-EGFP genotype exhibited clusters of L1-EGFP labeling, typically in the anterior portion of the MC. At 6 months of age, however, clusters of neurons that are positive for the L1-EGFP reporter are more uniformly distributed along the anteroposterior axis in the hTDP-43-Q331K/L1-EGFP animals. This was not the case with the hTDP-43-WT/L1-EGFP animals, which continued to exhibit smaller clusters of L1-EGFP labeled cells, often restricted to the anterior region, with no significant changes in size or distribution between 3 and 6 months (Fig 6b). We identified significant effects of bregma coordinates, age, and genotype (p < 0.0001, as determined by a two-way ANOVA mixed-effects model on log-transformed data) on the spatial distribution of clusters of L1-EGFP-labeled cells in the MC of the mutant and wild-type forms of TDP-43. Together, the findings are consistent with the idea that the majority of cells exhibiting L1-EGFP retrotransposition in the hTDP-43-Q331K/L1-EGFP animals are part of a single, spatially contiguous cluster of cells. This cluster of cells typically appears first in the anterior regions of the MC but extends to the posterior regions by 6 months of age. By contrast, the distribution of L1-EGFP-labeled cells in the hTDP-43-WT/L1-EGFP animals indicates the presence smaller clusters of cells, which likely occurred independently at different locations along the anterior-posterior axis.

Both the RNA sequencing and L1-EGFP reporter studies described above indicate that RTE expression and L1-EGFP retrotransposition occur transiently around the onset of motor deficits. Although the motor symptoms persist into later ages, the L1-EGFP reporter-labeled cells appear in large clusters in the MC around the time of motor symptom onset and then are no longer detected at later timepoints. The transient nature of the presence of EGFP-labeled cells was unexpected because activation of the L1-EGFP reporter depends on retrotransposition through an RNA intermediate and insertion of a de novo cDNA copy in which a constitutive CMV promoter is fused to the EGFP cassette.

One possible explanation for the absence of L1-EGFP-labeled cell clusters at later timepoints is the death of the neurons and glia in which LINE-1 retrotransposition events take place. To test this possibility, we performed immunohistochemistry using the TUNEL assay in conjunction with DAPI, to label all nuclei, and GFP to label neurons or glia in which the LINE-1 indicator cassette reveals that a retrotransposition event has occurred. We found that L1-EGFP-positive neurons were often positive for TUNEL. This was true in all three genotypes, regardless of the presence of the TDP-43 wild-type or mutant transgenes (Fig 7a). 75% of GFP-positive neurons in L1-EGFP, 79% of those in hTDP-43-Q331K/L1-EGFP mice and 52% in hTDP-43-WT/L1-EGFP mice were positive for TUNEL (Fig 7c). Thus, neurons in which the L1-EGFP reporter indicates a retrotransposition event has taken place are highly likely to undergo programmed cell death, irrespective of whether the animal contains the human TDP-43 transgene. In contrast, GFP-positive astrocytes with LINE-1 retrotransposition events did not exhibit TUNEL staining, suggesting a cell-type-specific effect of LINE-1 retrotransposition-associated programmed cell death (Fig 7b). But while the correlation between LINE-1 retrotransposition and TUNEL label in neurons did not differ across genotypes, the numbers of such neurons that are labeled by the L1-EGFP reporter are far higher in the hTDP-43-Q331K animals (Figs 3–6).

We also noticed that in the hTDP-43-Q331K/L1-EGFP animals, there was TUNEL labeling in many cells that were not labeled with GFP. This suggested the possibility that cells that exhibit LINE-1 retrotransposition events driven by hTDP-43-Q331K may exhibit cell non-autonomous effects leading to death of nearby neurons or glia. We therefore quantified TUNEL labeling in cells that were GFP-negative as a function of distance from clusters of L1-EGFP-positive cells. To accomplish this, we tiled confocal images over a 300µmX900µm area of MC tissue, beginning near the edge of positive clusters of L1-EGFP-labeled cells in a hTDP-43-Q331K/L1-EGFP and the hTDP-43-WT/L1-EGFP animals. We then quantified the TUNEL labeling within 100µmX100µm bins, allowing us to calculate the percentage of GFP-negative nuclei that were positive for TUNEL in each bin, as a function of distance from the edge of the cluster of GFP-labeled cells (Fig 7d).

In the L1-EGFP control mice, the GFP-negative cells located in bins closer to the rare GFP-positive cells showed no significant difference in TUNEL staining when compared to the bins that were further away (Fig 7e). Thus, in animals that do not express hTDP-43-Q331K or hTDP-43-WT, the percentage of nuclei undergoing programmed cell death was similar in regions that were near or far from the cells with LINE-1 retrotransposition. By contrast, in the hTDP-43-Q331K/L1-EGFP and hTDP-43-WT/L1-EGFP mice, we observed many GFP-negative cells labeled with TUNEL near GFP clusters, with the percentage of TUNEL-positive nuclei decreasing as a function of distance from the GFP-positive cell cluster (Fig 7e). We found that both the TDP-43 genotype (p < 0.0001) and the distance (p = 0.0008) from the L1-EGFP retrotransposition cluster had significant effects on the percentage of nuclei that were TUNEL-positive. Both hTDP-43-Q331K/L1-EGFP and hTPD43-WT/L1-EGFP showed significantly higher numbers of TUNEL-positive, GFP-negative nuclei compared to L1-EGFP controls at a distance from L1-EGFP clusters, with hTDP-43-Q331K/L1-EGFP having significantly higher numbers of TUNEL labeled cells than hTDP-43-WT/L1-EGFP or L1-EGFP controls at each distance up to 700 µm (Fig 7e). Beyond 700 µm, we did not observe significant differences in the number of cells undergoing programmed cell death across the three genotypes. Thus, we demonstrate that L1-EGFP-negative cells that are near clusters of L1-EGFP-labeled cells also undergo programmed cell death, and that the likelihood of such cell death falls off as a function of distance from the cluster of cells in which the L1-EGFP reporter has replicated.

All relevant ethical regulations and animal procedures were reviewed and approved by Stonybrook University Division of Laboratory Animal Resources and the Stony Brook University Institutional Animal Care and Use Committee (approval #1133421).

C57BL/6N mice, B6N.Cg-Tg (Prnp-TARDBP*Q331K)103 Dwc/J Stock # 017933, and B6N.Cg-Tg (Prnp-TARDBP) 96Dwc/J stock # 017907 [33] transgenic mice were purchased from Jackson Laboratories. The human TDP-43 cDNA with or without the Q331K mutation is inserted under the murine prion promoter with an N-terminal myc tag allowing it to be expressed in the central nervous system [33]. The single-copy insert SN1 L1-EGFP animals were a gift from Dr. Jef D. Boeke and Dr. Fred Gage [79]. Briefly, a retrotransposition-competent endogenous mouse LINE-1 element promoter was utilized with codon-optimized mouse L1 ORF1 and ORF2 sequences and and EGFP-based retrotransposition indicator cassette.The EGFP sequence with an intron in the middle is inserted in an inverse fashion under the pCMV promoter in the 3’UTR of the LINE-1 element. Upon transcription, the intron is spliced out. The EGFP is expressed under the pCMV promoter upon reverse transcription and insertion. Thus, the mouse model expresses GFP upon a copy-and-paste event. The hTDP-43-Q331K Tg and the hTDP-43-WT Tg lines were crossed with the L1-EGFP reporter line to create the double transgenic lines. All protocols were approved by IACUC. The animals were housed in the maximum isolation facility and then transferred to the conventional facility prior to the experiments. The cages, bedding and food were sterilized. Food and water were provided ad libitum. A light-dark cycle of 12h was used (7:00 on, 19:00 off). The temperature was maintained between 20–26 °C. For all experiments, mice were used at 1.5, 3, 6, 10, 15, and 17m time points. All cohorts were age-matched and mixed by sex.

Animals were weaned at 21 days, ear tagged, and a tail clip was taken for genotyping. Genomic DNA extraction was done using homemade buffers. Briefly, 600µl of TNES buffer (Tris-HCL, pH 7.5 = 50ml; 5M NaCl = 80ml; 0.5M EDTA, pH = 8.0 = 20ml; 20% SDS = 30ml; fill up to 1000ml with MilliQ) and 20µl of 10mg/ml proteinase-K were added to each tail clip. This was incubated in a 55 °C water bath overnight. The samples were then neutralized with 180µl, 5M NaCl. After vigorously shaking the samples for a minute, the samples were spun down at 14000 rpm for 20 minutes at room temperature. 600µl of the supernatant was transferred to a fresh tube, and 800µl of ice-cold 100% ethanol was used for precipitating the DNA. The tubes were inverted 10 times and centrifuged for 10 minutes at 14000 rpm. The supernatant was discarded, and the samples were centrifuged again for 2 minutes at 14000 rpm. The remaining supernatant was discarded, and the pellet was dried for 10 minutes and resuspended in 150µl of TE buffer. The genomic DNA was diluted 1:20 in MilliQ water and used for the genotyping PCR (Tables 1–3).

Hindlimb clasping and latency to fall on the accelerated rotarod were used as measures of motor deficits in the mice. Hindlimb clasping is a marker of disease progression in several neurodegenerative disease models, such as ALS, Huntington’s disease, and cerebellar ataxias. Mice with lesions in the MC, cerebellum, spinal cord, and basal ganglia show hindlimb clasping [80]. Hindlimb clasping is defined as the retraction of the hindlimbs upon tail suspension. Wild-type animals generally spread out their hindlimbs in anticipation of a surface to mount on. The mice were scored on a severity score of 0–3 [81]. Score 0 for when hindlimbs are consistently splayed. Score 1 for when one hindlimb is partially retracted 50% of the time. Score 2 for when both hindlimbs are partially retracted 50% of the time. Score 3 for when both hindlimbs are completely retracted 50% of the time. Testing was always conducted in the afternoon from 15:00–16:00, for all the animals. The accelerated rotarod assay was done in the afternoon from 16:00–18:00 for all the animals. The animals had 1 day of habituation, 1 day of training, and 3 days of testing. Animals were habituated for five minutes on the first day. The mice are placed at a 45° angle onto the rotarod, which is accelerated at a set ramp. The latency to fall from the rotarod is measured. The animals fall onto a soft cushion to avoid injuries from impact. The training consists of three trials, each with a 30-minute inter-trial interval. Trial 1 was done at 0 rpm for 60s. Trials 2 and 3 are done at 0–4rpm for 60s. The testing stage consists of three days, each with three trials. The latency to fall is measured for each animal. Each trial was done at 4–30 rpm for 300s with an inter-trial interval of 30 minutes. The final latency to fall was calculated as an average of all trials over the three days of testing [33].

We took an n = 3 for the hTDP-43-Q331K Tg, hTDP-43- WT Tg, and non-transgenic littermates. The age groups used for MC tissue collection and RNA extraction were 1.5-, 3-, 6-, 10-, and 15-month animals. All cohorts were mixed-sex groups. Animals were perfused with ice-cold PBS on ice. The MC was dissected on ice and flash-frozen in 100% ethanol and dry ice. The tissue samples were stored at -80 degrees for a few days before being shipped for total RNA extraction and library prep to Azenta. Sequencing was done on the Next-Gen 550 platform. The sequence configuration was Illumina, 2x150bp, with an estimated output of about 350 million raw paired-end reads per lane. Illumina adapters were removed using cutadapt (v4.4) and aligned to the mm39 genome using STAR (v2.7.10b) [82]. Genes and TE were quantified using TE transcripts (v1.2.3) using RefSeq gene annotations (Apr 2023 release) [83] and UCSC RepeatMasker annotations [84]. Differential analysis was performed using DESeq2 (v1.40.2) [85]. Heatmaps were generated with variance stabilizing transformed counts using the NMF R package (v0.27) [86].

We divided the differentially-expressed RTEs in non-trangenic animals into five arbitrary categories as shown in S5 Table: (i) strongly age-dependent- which showed significant increases in RTEs in at least 3 out of the 4 time points when compared to 15 month-old non-transgenic animals along with no significant decreases in RTEs in the 1.5 month-old animals compared to the 15-month-old animals; (ii) mildly age-dependent- which showed significant increases in RTEs in at least one of the 4 time point comparisons and no significant decreases in RTEs; (iii) 1.5-month peak- which showed significant decreases in RTEs in the 15 month-old non-trangenic animals compared to the 1.5-month-animals but not with the 3-month-old animals; (iv) early expressors- which show significant decreases in RTEs in the 15-month-old animals when compared to the 1.5- and 3-month time points but not in the older time points; and (v) strongly anti-age dependent- significant decreases in RTEs in 15 month-old non-transgenic animals when compared to 1.5-, 3-, and 6-month time points. We observed a significant overlap (p_adj = 1.03E-15, odds ratio = 9.9339) in the significantly upregulated RTEs in the 15-month-old hTDP-43-WT and any upregulated RTEs in the 3-month-old hTDP-43-Q331K animals. We decided to use the significantly upregulated RTEs in the 15-month-old hTDP-43-WT animals as the final set of RTEs that are upregulated as a result of TDP-43-related neurodegeneration. We then compared this “RTEs associated with neurodegeneration” list to see if it was enriched in any of the five categories of RTEs expressed in the non-transgenic animals (S5 Table).

The animals were anesthetized using Isoflurane. The heart was exposed, PBS was injected into the left ventricle, and the right atrium was punctured. The animal was transcardially perfused with PBS for five minutes until the liver and heart cleared. The animal was then perfused with 4% PFA in 1X PBS for five minutes. The animal was decapitated, and the brain was extracted. The brain was transferred to 4% PFA and incubated at 4⁰C overnight. It was then transferred to 30% sucrose in 1X PBS and incubated on a shaker at 4⁰C for 2–3 days until the sample sank completely. The brain was then coated with OCT in a brain mold and flash-frozen in dry ice and 100% ethanol. The brain was stored at -80⁰C until it was cryosectioned. The brain samples were sectioned at a thickness of 50µm and stored as free-floating sections in 1X PBS and stored at 4⁰C. The samples were stored in 15% glycerol in 1X PBS for long-term storage at -20⁰C. Standard IHC protocols were used for staining. The sections were blocked in 3% normal donkey serum (NDS- Sigma D9663-10ML) and 0.3% Triton-X 100 in 1X PBS for 3 hours at room temperature on a shaker. The sections were then incubated in the primary antibodies (Table 4) in 1% NDS and 0.1% Triton X- 100 in 1X PBS at 4⁰C overnight on a shaker. The sections were then washed in 0.1% PBST thrice for 15 minutes on a shaker at room temperature. The secondary antibody was diluted in 1% NDS and 0.1% Triton X- 100 in 1X PBS. The sections were incubated in the secondary antibody dilution for 3 hours at room temperature in the dark with gentle shaking. The sections were then washed three times for fifteen minutes in 0.1% PBST, followed by a final 5-minute wash in PBS. The sections were mounted in Flouromount-G mounting media with or without DAPI from Southern Biotech (Cat# 0100–01 and Cat# 0100–20 respectively), coverslipped, and were allowed to dry for one day before imaging.

We used the TUNEL Assay Kit (Abcam ab66110) to investigate cell death in the brain tissue. The 50µm brain sections, which were perfused, PFA-fixed, and OCT-embedded, were first blocked and permeabilized using 3% normal donkey serum (NDS - Sigma D9663-10ML) and 0.3% Triton X-100 in 1X PBS. This step was performed for 3 hours at room temperature on a shaker, instead of using proteinase K. Next, the tissue sections were incubated in DNA labeling solution with 0.1% PBST for 48 hours at 4°C while being shaken. When co-labeling with other primary antibodies, such as GFP, the appropriate dilution of the antibody was added together with the DNA labeling solution. After this, the tissue was washed three times with 0.1% PBST for fifteen minutes each on the shaker. Following the washes, the sections were incubated with the anti-BrdU antibody and the secondary antibody against GFP for 1 hour at room temperature in the dark, with shaking. The sections were then washed three times in 0.1% PBST, with each wash lasting 15 minutes, followed by a final 5-minute wash with PBS. Finally, the sections were mounted using Fluoromount-G mounting media with DAPI from Southern Biotech (Cat# 0100-20), coverslipped, and allowed to dry for one day before imaging. For the experiment, control brain tissue was treated with 3000 U/ml of DNase I for 10 minutes at room temperature and served as a positive control.

The Olympus slide scanner was used to image all sections at 10X magnification for quantifying GFP-positive cells. The imaging parameters were kept consistent across different rounds of imaging. A 20µm thick stack was imaged, consisting of 20 optical sections with a 1 µm intervals. The Zeiss LSM 800 confocal microscope was used to study the morphology of various cell types in the GFP-positive cell population using the 20X, 40X and 63X magnifications. Additionally, the Zeiss Airy Scan 63X was used to image the TUNEL staining in the GFP-positive cells. The Zeiss LSM 800 confocal microscope was used at 63X to generate images utilized for the TUNEL distance vs. probability plots.

We developed a pipeline for the image processing of GFP-positive cell quantification. All images were captured using the same four channels: Neurotrace-blue at 405nm, GFP at 488nm, an empty red channel to correct for background staining, and GFAP at 647nm. Maximum intensity projections were obtained for the green and red channels. The red channel was subtracted from the green channel to eliminate any nonspecific staining. The final image was then thresholded and the number of GFP-positive cells in the MC, striatum, nucleus accumbens, and agranular insular cortex was counted using ImageJ. Boundaries for each section were defined using templates from the Allen Mouse Brain Atlas. The total number of GFP-positive cells was calculated based on the equivalent sections from all the animals.

We examined the organization of clusters in 14 sections at identical bregma locations across different animals. Each section was 50 microns thick, with a 150-micron interval between sections. We defined a cluster as a group of cells with LINE-1 retrotransposition events, indicated by GFP expression, which was three standard deviations above the mean GFP-positive cell counts found in the MC of the L1-EGFP littermates. To assess the presence or absence of a cluster, we analyzed total counts across 14 sections. All animals represented on the lateralization map had total GFP counts within the MC that exceeded the cluster threshold. We calculated thresholds for each section based on the mean L1-EGFP cell counts specific to that bregma location. Subsequently, we created a lateralization map that reports the presence, size, and lateralization of clusters. Since we were working with free-floating brain tissue, we utilized the asymmetry between the hemispheres caused by cryosectioning to label the hemispheres as pseudo-left or pseudo-right, and then we tracked the presence of the cluster in each hemisphere. Each row in the map represents an animal, while each column represents a section at the specified bregma location.

To calculate the probability of cell death at various distances from the cell clusters exhibiting LINE-1 retrotransposition events, we conducted imaging from the edge of the GFP-positive clusters or cells. Imaging was performed at 63X on the Zeiss LSM 800 confocal microscope, utilizing a 9 x 3 tile matrix (totaling 27 tiles), which covered a rectangular area of 900 μm x 300 μm. We captured images in the DAPI, GFP, and TUNEL channels. The image was divided into 100μm bins, and the number of TUNEL-positive nuclei was determined in each bin from the raw Z-stack data (Fig 7d). We set an arbitrary threshold of 3 puncta per nucleus for it to be considered TUNEL-positive. We quantified the number of TUNEL-positive nuclei per bin and calculated the probability as follows: (number of TUNEL-positive nuclei/ Total number of nuclei counted).