Introduction
Characterized by extensive overlap in cellular pathology, genetic mutations, and molecular markers, amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD) comprise a neurodegenerative disease continuum affecting multiple subsets of neurons in the spinal cord, motor cortex, cingulate cortex, and frontal and temporal lobes [
1‐
4]. Notably, nearly 97% of ALS cases [
5] and approximately 45% of FTD cases [
6] exhibit neurodegeneration with ubiquitin-positive cytoplasmic inclusions that contain transactive response (TAR) DNA binding protein 43 (TDP-43). A number of point mutations in
TARDBP, the gene encoding TDP-43, have been identified as causative of ALS [
7‐
9] while others have been linked to FTD [
10,
11]. Although mutations in at least 20 other genes have been associated with ALS/FTD, the majority of ALS/FTD cases are sporadic [
12].
Substantiating its significance as a marker of pathology, TDP-43 has also been identified as a component of cytoplasmic aggregates in dementias other than FTD, including Alzheimer’s disease (AD) [
13‐
15], AD with Lewy body dementia (AD/LBD) [
16], Parkinsonism-dementia complex (PDC) [
17], and hippocampal sclerosis [
18]. Up to 57% of AD cases exhibit TDP-43 positive intracellular inclusions [
19]; notably, AD patients with TDP-43 proteinopathy show accelerated disease progression with more severe cognitive impairment [
14,
20]. In addition to the presence of TDP-43 proteinopathy, dementias such as FTD and AD share TDP-43 associated cellular pathology, including dystrophic neurites [
21]. Thus, despite differences in underlying genetic risk factors and specific neural populations affected, a large proportion of dementias exhibit overlapping TDP-43 pathology, providing a common entry point for understanding shared mechanisms of neurodegeneration [
22,
23].
Wild-type TDP-43 is localized primarily to the nucleus where it regulates RNA transcription [
24] and splicing [
25‐
29] necessary for nervous system development and function. TDP-43 also shuttles between the nucleus and cytoplasm and has been shown to play a role in axonal and dendritic mRNA localization [
30‐
32], stress granule dynamics [
33‐
35] and translation [
36‐
38]. During severe or prolonged cellular stress and in disease, TDP-43 exits the nucleus and associates with ubiquitin-positive cytoplasmic inclusions [
39]. Indeed, cellular dysfunction and subsequent degeneration associated with TDP-43 pathology have been attributed to both loss-of-function of nuclear TDP-43 [
27] and cytoplasmic gain-of-function [
40] whereby cytoplasmic inclusions sequester mRNAs and perturb protein homeostasis [
39].
Following the identification of TDP-43 as a protein of pathological interest in numerous neurodegenerative diseases, model organisms have been critical to understanding the normal biological functions of TDP-43 and how disruption of these functions leads to disease. Despite having diverged from a common ancestor with vertebrates over 600 million years ago [
41], the fruit fly (
Drosophila melanogaster) genome contains a homolog of
TARDBP, namely
TBPH, which displays sequence similarity and functional overlap with human TDP-43 [
42,
43].
In addition to extensive overlap in neurogenetic homology, circuit-level conservation in the central nervous system has also been demonstrated between invertebrates and vertebrates [
44‐
46], making invertebrate models relevant for understanding diseases of cognition. Indeed, flies display several behaviors affected in human disease [
47] including associative learning [
48], long-term memory [
49], sleep [
50], social interactions [
51,
52], and addiction [
53] that can serve as system-level readouts of molecular- and/or cellular-level dysfunction. Models of TDP-43 proteinopathy in
Drosophila have recapitulated the locomotor phenotypes, motor neuron dysfunction, and shortened lifespan [
43,
54,
55] observed in ALS and highlighted degeneration in central brain structures [
54] reminiscent of FTD. Beyond simply simulating disease states, these fly models have identified translational targets [
37,
38,
56] and genetic modifiers of TDP-43 proteinopathy [
57‐
60], and exposed systemic effects on metabolic pathways [
61,
62].
While TDP-43 proteinopathy has been extensively studied in motor neurons and glia as a model of ALS, dementia modeling based on TDP-43 proteinopathy in flies has been limited. This is in part because the mushroom bodies (MBs) of the fly brain, which were originally described as essential to odor learning [
63] have only recently been recognized as being required for multi-modal and context-dependent learning [
64,
65] including place learning [
66]. The MBs also regulate sleep [
67,
68], satiety [
69], social behavior [
70] and gate behaviors attributed to other brain regions such as decision-making [
71] and aggression [
72]. Therefore, the MBs share functional overlap with regions of the vertebrate cortex [
73]. Existing fly models of intellectual disability [
74] and Alzheimer’s disease (AD; [
75] have either focused on, or identified phenotypes in the MBs of the fly brain, suggesting that this structure is appropriate for modeling cognitive disorders.
Here we describe the development of a novel fly model of dementia based on TDP-43 proteinopathy induced by specific over-expression of wild-type or mutant human TDP-43 in a well-defined subset of MB neurons. Our model recapitulates key cellular and behavioral characteristics of human dementias with TDP-43 pathology, including age-dependent loss of nuclear TDP-43, axonal degeneration, and working memory and sleep deficits that parallel those observed in dementia patients. Using RNA immunoprecipitations we show that in MBs, TDP-43 associates with several mRNAs, a subset of which are unique to MBs while others are shared with mRNA targets previously identified in motor neurons [
38]. Interestingly, among the latter, we identified
dlp mRNA, encoding the heparan sulfate proteoglycan (HSPG) Dally-like protein (Dlp)/GPC6, which we previously found to be a target of TDP-43 in fly models of ALS and is altered in human ALS spinal cords [
38]. Here we report that
dlp mRNA is enriched in TDP-43 complexes in MBs while Dlp protein is decreased within axons during aging, consistent with TDP-43 dependent axonal transport and/or translation deficits in an age-dependent manner. Notably, Dlp overexpression in MBs rescues TDP-43 dependent deficits in working memory consistent with Dlp being a physiologically significant target of TDP-43 in the MB circuit. These findings demonstrate that TDP-43 proteinopathy in MBs causes dementia-like phenotypes that are mediated at least in part by Dlp/GPC6, a regulator of the Wg/Wnt signaling pathway. Further substantiating the link between Dlp/GPC6 and TDP-43 pathology, we found
GPC6 mRNA to be altered in FTD patient brains. Specifically, neuronal nuclei expressing
STMN2 and
KALRN cryptic exons associated with TDP-43 nuclear depletion show an upregulation of
GPC6 mRNA when compared with nuclei that retain TDP-43 as evidenced by the presence of canonically spliced
STMN2 and
KALRN junctions. Lastly, we report a number of candidate targets of TDP-43 that show overlap between our fly models and patient brains, suggesting these models will be useful for studying molecular pathyways underlying distinct neuronal vulnerabilities across the spectrum of TDP-43 proteinopathies.
Materials and methods
Drosophila genetics and maintenance
Flies were maintained at 25 °C in 12-h light–dark cycle with 25–30% humidity. Specific information on
Drosophila lines used in this study and specific experiments where each line was employed can be found in the Key Resources table (Additional file
1). For aging and lifespan studies, newly eclosed virgin male and female flies were collected and maintained on standard fly cornmeal/molasses media refreshed weekly. Flies harboring
UAS TDP-43::YFP and
UAS dlp transgenes for co-overexpression of TDP-43 and Dlp were generated using standard genetic recombination techniques. The presence of
UAS TDP-43::YFP was confimed by YFP expression when crossed with the pan-neuronal driver, elav GAL4 while Dlp overexpression was confirmed using RT-qPCR (
dlp OE = 2.45 FC,
TDP-43WT::YFP;
dlp OE = 3.09 FC,
TDP-43G298S::YFP;
dlp OE = 1.71 FC compared to
w1118 controls; Additional file
1: Fig. S6-supplement 1b).
Western blotting
One to three days old flies were collected from
SS01276 crossed with (1)
w1118; TDP-43WT::YFP, (2)
w1118; TDP-43WT::YFP; dlp OE, (3)
w1118; TDP-43WT::YFP mCD8::RFP, and (4)
w1118 in triplicate. Heads (N = 15 for each genotype) were decapitated and homogenized in 100 µl 2X Laemmli Sample Buffer (BIO-RAD 1610737) containing 5% of 2-Mercaptoethanol (Sigma-Aldrich M3148). The homogenized protein samples were boiled for 5 min in the digital Heat Block (Benchmark), and spun for 1 min. Supernatants were collected and 10 µl of protein sample was loaded in each well of precast Mini-PROTEAN TGX 4–20% gradient Gel (BIO-RAD 4561096). Following SDS-PAGE, the proteins were transferred to Nitrocellulose membrane (BIO-RAD 1620215). After transfer, the membrane was blocked in 5% nonfat milk in PBST (PBS and 0.1% TWEEN20) and incubated overnight at 4 °C with primary antibodies mouse anti-GFP Living Color (Cell Signaling Technology 2955, 1:1000) to detect TDP-43 YFP and rabbit anti-Beta-Actin (Cell Signaling Technology 4967S, 1:1000) as loading control, followed by washes in TBST and incubation with secondary antibody (IRDye® 800CW, Goat anti-Mouse D21115-25, 1:10,000) and (IRDye®680RD Goat anti-Rabbit D21207-05, 1:10,000) for 1 h at RT. The blot was imaged using a LICOR scanner (Odyssey®DLx) and protein bands intensities were quantified with LICOR “Image Studio Lite” (Additional file
1: Fig. S7-supplement 2).
Statistical analyses
Statistical analyses were conducted in R [v. 4.1.2, R Core
76] and Rstudio [v. 2021.09.0 + 351, Rstudio
77] unless specified otherwise in methods. The tidyverse [
78] and ggpubr [
79] packages were used for summary statistics and graphics. Other analysis-specific packages are cited in the corresponding Methods section. While data for the mutant TDP-43
G298S model is presented as a supplement to our main findings, statistical analyses were conducted simultaneously for both TDP-43 genotypes and p-values were corrected for multiple comparisons that included both genotypes. For any given analysis when males and females did not differ statistically or sample sizes were low (
e.g., histological preparations), they were pooled to increase power. Where necessary, outliers were removed prior to hypothesis testing. When data did not fit the assumptions of parametric models and non-parametric analyses were used, hypothesis testing proceeded by first using a Kruskal-Wallace to test for a difference among groups, followed by pairwise comparisons using the Wilcoxon Rank Sum Test with p-values adjusted for multiple comparisons using the false discovery rate method [
80]. Specific statistical methods are described for each assay. Summary statistics for each figure can be found in the Additional file
2.
Mushroom body morphological analyses
Mushroom body morphology was evaluated using membrane-targeted RFP (mCD8 RFP) driven by
SS01276 [
81]. Adult brains were dissected in cold HL-3 saline [
82], fixed for 60 min in 4% paraformaldehyde, then rinsed in phosphate buffered saline (PBS, 3X), permeabilized in PBS with 0.25% Triton X-100 (PBST), and blocked in PBST plus 5% normal goat serum (Sigma-Aldrich 566,380) 2% bovine serum albumin (Sigma-Aldrich A5611) for 45 min prior to antibody labeling. YFP was detected by incubating brains overnight at 4 °C with a mouse monoclonal anti-GFP FITC antibody (1:300, Rockland 600-302-215). mCD8 RFP was visualized using native fluorescence. All brains were mounted on slides with the ventral side containing the mushroom body lobes (MBLs) facing the coverslip. For cell body imaging, brains were additionally incubated in Hoechst (1:10,000, Invitrogen H3570) for 10 min and mounted on slides with the dorsal side containing the calyx facing the coverslip.
Image acquisition and analysis
TDP-43 cytoplasmic localization and MBN cell loss: Images were acquired using a Zeiss 880 Laser Scanning Confocal inverted microscope with a Plan-Apochromat 63x/1.4 oil DIC M27 lens. TDP-43 YFP signal intensity was quantified per unit area in the nucleus and cell body. Hoechst was used to define the nuclear boundaries, while total cellular TDP-43 signal was measured by tracing a boundary around YFP signal in the entire cell (Additional file
1: Fig. S1-supplement 3). To quantify the total number of MBNs expressing TDP-43 YFP, all nuclei were counted from optical sections 2 μm apart to ensure nuclei were counted only once. We tested for differences in TDP-43 nuclear to cellular ratio and MBN number by age, sex, and genotype using Analysis of Variance (ANOVA) performed on a linear model (
lm function in R; R Core Team 2017). Residuals were normally distributed (Shapiro–Wilks test for normality,
P = 0.890 and
P = 0.395 for ratio and MBN number, respectively) and we therefore continued with pairwise comparisons using Tukey’s Honest Significant Difference corrected for multiple comparisons.
Signal intensity: All images were acquired using a Zeiss 880 Laser Scanning Confocal inverted microscope with a Plan-Apochromat 63×/1.4 oil DIC M27 lens. For mCD8 RFP fluorescence intensity, brains were imaged using either an Alexa Fluor 568 or a DS Red filter set with pinhole adjusted to 1 AU and 3 µm optical section thickness. For measuring anti-GFP FITC fluorescence intensity, brains were imaged using the FITC filter set with the pinhole adjusted to 1 AU and 0.5 µm optical section thickness. Fluorescence intensity was measured in FIJI as the integrated density over the sample area in axons by manually tracing α/β or γ lobes and subtracting adjacent background signal. First, a polygon was traced free-hand over the visible area of each lobe (α/β or g). This polygon was then dragged to an adjacent area of the brain to obtain a mean fluorescence intensity for the background in each section. In instances where the polygon was shaped in such a way that it could not be dragged to adjacent brain area, a comparably-sized polygon was drawn in the adjacent brain region. Intensity was measured in at least three sections to obtain mean fluorescence intensity in each channel (RFP or FITC) for both α/β and γ lobes of each brain. To test for differences in signal intensity within a genotype across age time points, values of signal intensity were normalized to mean signal intensity of young (1–3 days old) brains of the same genotype and lobe. Statistical analysis of change in fluorescence intensity with age was performed by comparing middle aged (~ 30 days) or old (~ 60 days) flies and young (~ 1 day) flies from the same genotype using the Wilcoxon Rank Sum test. Summary statistics can be found in Additional file
2: Table S2a.
YFP particle size: The Analyze Particles function in FIJI was used to identify YFP puncta in MB lobes. Images were first contrast enhanced using a brightness/contrast enhancement (saturated pixels set to 0.3% and normalized) followed by thresholding using the Bernsen method in the Auto Local Threshold function with a 25–50 pixel radius depending on image quality. Particles were considered puncta when they were 0.25–3.0 µm
2 in size and showed 0.25–1.0 circularity. To measure mean particle size in each sample, a polygon was drawn over the visible portion of each lobe in at least three sections for each lobe. From these subsamples we calculated a mean particle size for each lobe of each brain. Summary statistics can be found in Additional file
2: Table S2B.
Dally-like protein expression during aging in MBs: Brains from flies 1–3 days or 50–55 days were dissected out and the tissue fixed, permeabilized and blocked as described above. Brains were then incubated overnight at 4 °C in an anti-Dally-like protein antibody at 1:5 in block (13G8 developed by Phil Beachy, obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, targets amino acids V523 to Q702), rinsed 3X in 0.1% PBST and incubated in goat anti-mouse Alexa Flour 647 antibody overnight at 4 °C (1:300, ThermoFisher A32728). Images were acquired using a Zeiss 880 Laser Scanning Confocal inverted microscope with a Plan-Apochromat 63×/1.4 oil DIC M27 lens. Signal intensity was traced from maximum intensity projections of the MBLs and normalized to background intensity. Dlp signal intensity declined with age in TDP-43 flies, so the mCD8 RFP channel was used to trace the lobes in aged brains, ensuring the correct lobe area was used for signal intensity measurements.
GAL4 dilution effect: To test whether UAS-driven expression of a second transgene reduces UAS-driven TDP-43
WT expression, in addition to western blots, we also measured mean pixel intensity from maximum intensity projections of the horizontal lobes (β and γ combined in maximum intensity projection) in two genotypes with
SS01276-driven transgenes,
OR-R;TDP-43WT::YFP and
w1118; TDP-43WT::YFP mCD8::RFP (Additional file
1: Fig. S7-supplment 3).
Behavioral assays
Y-maze: To measure changes in working memory, we employed a y-maze assay described by Lewis, Negelspach, Kaladchibachi, Cowen and Fernandez [
83]. We tracked the two-dimensional movement of individual
Drosophila placed in small symmetrical y-mazes using video recordings for 10 one-minute trials following two minutes of acclimation time in the maze. Observations took place between 08:45 and 11:00 and 14:00 to 16:00 when flies were most active. Male and female flies were run separately and the locations of genotypes in the seven by seven maze array were randomized for each set of trials. Assays were conducted in a dark room with each maze lit uniformly from below with a white LED array and capped with a clear Plexiglas coverslip. The overall movement of flies and visits to three unique arms consecutively were quantified using the Noldus Ethovision software. Spontaneous alternation behavior was measured by scoring three consecutive arm entries in a sliding window for each of ten trials. The alternation score was calculated as alternations divided by alternation attempts. Movement and alternation data were summed for each fly during each one-minute trial and then averaged over 10 trials for statistical analyses. Flies originating from three biological replicates were pooled for analysis (Additional file
2: Table S3). The large samples sizes permitted the use of parametric tests, and we assessed differences in distance moved or alternation score across genotypes using ANOVA on a linear mixed-effects model [lmer function in R;
84] that included replicate as a fixed effect. Flies that performed no alternations naturally did not have a percent alternation score and were removed from subsequent analysis. Males and females were analyzed separately. Flies were additionally assessed for the presence of bias for specific arms, calculated as a ratio of the mean number of entries into each arm over mean total entries and tested for significant variation from 0.333 using the Wilcoxon Signed Rank test. Interestingly, while females showed no arm biases, males overexpressing
TDP-43WT and OR-R controls showed a bias against arm B (
TDP-43WT,
\(\overline{x }\) = 0.307 ±
\(0.08\),
P < 0.0001; OR-R,
\(\overline{x }\) = 0.316 ± 0.13,
P = 0.0038).
TDP-43WT also showed a bias for arm A (
\(\overline{x }\) = 0.352 ± 0.076,
P = 0.0033), while OR-R controls showed a bias for arm C (
\(\overline{x }\) = 0.347 ± 0.113,
P = 0.048). Although these analyses suggest that arm biases exist in males and are not unique to TDP-43 overexpression, a single replicate was used to test whether Dlp over-expression could rescue alternation deficits seen in TDP-43 overexpressing flies therefore hypothesis testing proceeded with non-parametric tests as described in
Statistical analyses.
Sleep: Adult flies from the three age time points were monitored individually using
Drosophila Activity Monitors [DAMs; Trikinetics, Waltham, MA;
85]. Flies were placed in monitoring tubes with a small amount of food the day before they reached two to three, 31–33, or 61–63 days old. Sleep and activity monitoring began at 12 AM following placement of flies in the sleep incubator, allowing flies at least six hours prior to the start of the experiment to acclimate to the tubes. Sleep and locomotor activity data were collected at one-minute intervals for three days and analyzed using ShinyR-DAM [
86]. In
Drosophila, sleep is defined at five minutes of inactivity [
50] and here sleep data were processed using the ShinyR DAM program (Cichewicz and Hirsh 2018), which provided average sleep, sleep bout number and duration, and activity to sleep bout length ratio. Comparison of these variables by age and genotype were conducted separately for male and female flies. Data from two replicates were pooled (> 15 flies per genotype × replicate; sample sizes in Additional file
2: Tables S4A and S4B).
Lifespan
Newly eclosed virgin males and females were separated and placed on fresh food, then transferred weekly into new food vials. The number of surviving flies was counted every other day for 100 days. Survival analysis and plots were generated using R packages
survival [
87] and
survivminer [
88].
mRNA targets of TDP-43
RNA Immunoprecipitations: Equal numbers of male and female flies, aged one to three days, were pooled in Eppendorf tubes and flash frozen in liquid nitrogen. A minimum of 500 flies expressing YFP, TDP-43WT YFP, or TDP-43G298S YFP were collected for each of three biological replicates. Frozen flies were then transferred to 50 mL conical tubes and heads were separated from the bodies using four rounds of vortexing and flash freezing with liquid nitrogen. A sieve was used to filter out the bodies and isolated heads were collected into tubes containing lysis beads (Next Advance Green lysis beads). Heads were homogenized (Next Advance Bullet Blender) in one mL fresh lysis buffer (DEPC water, 50 mM HEPES buffer pH 7.4, 0.5% Triton X-100, 150 mM NaCl, 30 mM EDTA), protease inhibitors (cOmplete™, Mini, EDTA-free Protease Inhibitor Cocktail, Millipore Sigma 11873580001), and RNAsin Plus 400 u/mL (Fischer Scientific PRN2615), centrifuged for 10 min at 10,000 rpm, and the lysate was collected. A portion of the lysate was saved for the protein input and RNA input samples. For protein input, the lysate was mixed with 2X Laemmli buffer, boiled at 95–100 °C for 10 min, and stored at − 20 °C. For RNA input, TRIzol reagent (Thermofisher 15596026) was added to the lysate and the sample was stored at − 80 °C. High affinity GFP-Trap magnetic agarose beads (ChromoTek) were added to the remaining lysate and rotated end-over-end for 90 min at 4 °C to allow for binding of YFP. The beads were separated from the lysate with a magnet and washed three times with fresh wash buffer (DEPC water, 50 mM HEPES buffer pH 7.4, 0.5% Triton X-100, 150 mM NaCl, 30 mM EDTA). The beads were resuspended in wash buffer and split into two tubes for the protein IP and RNA IP samples. For the protein IP sample, 2X Laemmli buffer was added to the beads, samples were boiled at 95–100 °C for 10 min, and the beads were removed with a magnet. Western blots (see above) were performed to ensure that the immunoprecipitated complexes contained TDP-43 before processing the RNA IPs. For the RNA IP sample, TRIzol reagent was added to the beads, the solution was pipetted up and down for 60 s, and the beads were removed with a magnet.
RNA-Seq: RNA was quantified by nanodrop to bring it within range for ribogreen quantification. RNA was also checked using Agilent Tapestation High Sensitivitiy RNA screentape. One ng total RNA was used for SMART-Seq HT PLUS (Takara Bio USA, Inc. Cat # R400748) following manufacturer’s protocol. Determination of cDNA quality and quantity was determined via Agilent Tapestation High Sensitivity D5000 Screentape and Qubit dsDNA HS assay for input into library amplification. Libraries were quantified by Agilent Tapestation High Sensitivity D1000 Screentape and Kapa Library Quantification kit for Illumina platforms. Libraries were pooled equimolarly, pools were quantified by Agilent Tapestation High Sensitivity D1000 Screentape and Kapa Library Quantification kit and loaded on the NovaSeq 6000 S4 flowcell and sequenced to 101 × 11 × 11 × 101 cycles. Trimmed fastqs were aligned to the Dmel genome with STAR v2.6.1d [
89]. Aligned reads were counted with featureCounts v1.6.3 [
90] using the genome annotation. Files produced for individual samples were aggregated into a single.txt file using python. Differential expression was quantified using Deseq2 [
91]. We included over-expression of cytoplasmic YFP in MBNs as a control for IPs, however the expression levels of YFP alone were far greater than cytoplasmic TDP-43 YFP (data not shown), making it difficult to assess the role of cytoplasmic YFP alone. To ensure that candidate mRNA targets showing the greatest enrichment were indeed specific in their association with TDP-43, we subtracted YFP control Log2FC values from the Log2FC values of targets recovered in each of our models. Indeed, in this YFP-subtracted analysis
futsch was recovered as enriched with TDP-43 in both variants (Fig.
6a, Additional file
1: Fig. S6-supplment 1a) and the majority of mRNAs that showed high enrichment in the original analysis were retained in the YFP-subtracted analyses (Fig.
6a vs 6d, Additional file
1: Fig. S6A vs S6C).
Functional annotation: The Database for Annotation, Visualization, and Integrated Discovery (DAVID) was used to functionally annotated genes enriched with TDP-43 [
92,
93]. To assess genes over-represented in our disease model data sets, we used lists of genes significantly enriched with each of our TDP-43 models (differential gene expression lists available in source data).
RNA Extraction for Validation: RNA input and RNA IP samples, stored in TRIzol reagent (Thermofisher 15596026) at − 80 °C, were allowed to thaw completely on ice and chloroform was added to each sample. Samples were shaken briefly to mix, left to incubate for three minutes at room temperature, and then centrifuged at 12,000 rpm for 15 minutes at 4 °C to allow for separation of the aqueous and organic phases. The RNA-containing aqueous phase was collected and molecular grade isopropyl alcohol was added to each sample. Samples were incubated for 10 min at room temperature for RNA precipitation and then centrifuged at 12,000 rpm for 10 minutes. The pelleted RNA was washed with 75% ethanol (200 proof ethanol & HyPure water) and left to dry in a fume hood. RNA was resuspended in HyPure water and the concentration and 260/280 absorbance ratio were measured using Nanodrop.
RT-qPCR: RNA extracted from input and IP samples was used as template RNA for cDNA synthesis by reverse transcription. Total RNA used in the cDNA synthesis reactions was normalized across samples. cDNA was synthesized using the Fisher First Strand cDNA synthesis kit (Thermofisher Scientific K1641). qPCR reactions were prepared in a 96-well qPCR plate, in triplicates, using Taqman Fast Advanced Master Mix (Thermofisher Scientific 4444556) and Taqman probes for Dally-like protein (dlp) (Thermofisher Scientific Dm01798597_m1) and Gpdh1 (Thermofisher Scientific Dm01841185_m1). qPCR was conducted on the qTOWER (Analytik Jena 844-00504-4) qPCR machine. Delta Ct was calculated as the difference in Ct values (dlp—gpdh1). To determine enrichment of dlp mRNA in the IPs, ΔΔCt was calculated as the difference in Delta Ct values (IP—Input). Fold change was calculated as 2^(− ΔΔCt).
Dally-like protein target validation: Brains from flies one to three days or 50–55 days were dissected out and the tissue fixed, permeabilized and blocked as described above. Brains were then incubated overnight at 4 °C in an anti-Dally-like protein antibody at 1:5 in block (13G8 developed by Phil Beachy, obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, targets amino acids V523 to Q702), rinsed 3X in 0.1% PBST and incubated in goat anti-mouse Alexa Flour 647 antibody overnight at 4 °C (1:300, ThermoFisher A32728). Images were acquired using a Zeiss 880 Laser Scanning Confocal inverted microscope with a Plan-Apochromat 63×/1.4 oil DIC M27 lens. Signal intensity was traced from maximum intensity projections of the MBLs and normalized to background intensity. Dlp signal intensity declined with age in TDP-43 flies, so the mCD8 RFP channel was used to trace the lobes in aged brains, ensuring the correct lobe area was used for signal intensity measurements.
snRNA seq analyses from FTD patient brains: Log normalized gene counts tables from [
94] were filtered to neuronal nuclei containing either a
STMN2 or
KALRN cryptic exon or a
STMN2 or
KALRN canonical junction, as defined in [
94]. Human orthologs of Drosophila genes were identified using the DIOPT tool [
95].
GPC6 gene expression was compared between neuronal nuclei containing
STMN2 and
KALRN cryptic exons and neuronal nuclei containing
STMN2 and
KALRN canonical splice junctions using the independent t-test function (ttest_ind) in python. Violin plots were generated between neuronal nuclei containing a cryptic exon and neuronal nuclei containing a canonical exon using the scanpy violin function.
Discussion
Neurodegenerative diseases have historically been classified by shared symptomology and post-mortem pathology. More recently, the integration of molecular markers and genetic testing has uncovered evidence that cell-type susceptibility and dysregulation of specific molecular pathways can interact to produce similar clinical presentations across different disease etiologies [
133]. These findings highlight convergent pathomechanisms during the degeneration process [
3,
22] and suggest inroads for better understanding of these complex diseases.
A common pathological feature of several neurodegenerative diseases including ALS, FTD, and other dementias is the nuclear depletion of the RNA binding protein TDP-43 accompanied by the accumulation of cytoplasmic, ubiquitinated foci [
13‐
16,
18,
134]. Since dementia is often confirmed post-mortem by the presence of TDP-43 pathology, animal models play an important role in elucidating the molecular mechanisms underlying TDP-43 proteinopathy and its role in neurodegeneration. Among these, fly models have proven their utility in this research landscape in part because the availability of powerful molecular and genetic tools allowing multiple hypotheses to be pursued in parallel on a large scale. Indeed, fly models of TDP-43 proteinopathy based on loss of function or overexpression in the entire nervous system [
55,
135,
136], or in subsets of neurons including associative regions of the central brain [
54], motoneurons [
43] or the retina [
137] have revealed a broad repertoire of altered pathways that have subsequently been confirmed in patient tissues [
37,
38,
56].
It has been previously shown that overexpression of TDP-43 in all MBNs using the driver line OK107 GAL4 (RRID:BDSC_854) caused age-dependent cell loss and axonal degeneration in MBs, however degeneration severity varied widely across nearly isogenic individuals [
54]. This variability could be in part attributed to driver line “leakiness” as OK107 GAL4 drives expression beyond MBs, in optic lobe neuropils [
138,
139] and neurosecretory cells [
140,
141]. To develop a robust, MB specific model of TDP-43 driven dementia in
Drosophila melanogaster, we leveraged a split GAL4 driver line with limited expression to a subset of Kenyon cells (γ, α/β MBNs). Notably, this model exhibits nuclear depletion and cytoplasmic accumulation of TDP-43, accompanied by age-dependent axonal degeneration and cell loss, all of which recapitulate key aspects of disease pathology. Importantly, the onset of FTD relevant behavioral symptoms is detectable prior to widespread degeneration, as observed in human disease. Additionally, this model identifies both novel, MB specific, and motor neuron-shared mRNA candidate targets that have previously been associated with TDP-43 pathology. Of the latter, here we chose to focus on the glypican Dlp, a Wg/Wnt signaling regulator. We show that Dlp is a functional target of TDP-43 in the MB circuit that mediates, in part, TDP-43 dependent working memory deficits in the fly model.
Neurodegeneration in ALS/FTD is thought to be driven by both loss of nuclear function, as some nuclei become devoid of TDP-43, and cytoplasmic gain of toxic function, as evidenced by TDP-43 accumulation in cytoplasmic puncta. While thus far invertebrate models of TDP-43 proteinopathy have not faithfully recapitulated TDP-43 mislocalization [e.g.,
98,
142], here we found that TDP-43 over-expression in MBNs more closely resembles disease pathology. Indeed, TDP-43 displays normal localization in the juvenile stage but becomes depleted from the nucleus and mislocalized to the cytoplasm in young adults. Although changes in TDP-43 localization could reflect developmental regulation [
143], our findings that the number of cells with nuclear TDP-43 depletion is reduced in young (1–3 days old) compared to old (50–53 days old) adult flies, concommitant with an overall reduction in MB neuron numbers, suggests that TDP-43’s shift in distribution from the nucleus to the cytoplasm is toxic.
Axonal degeneration is evident only in older adult flies and differentially affects MBNs that form the α/β and γ lobes, with the latter showing the greatest degeneration severity. This is particularly interesting because MBNs that form the γ lobe are embryonic in origin and therefore older than the pupal-born MBNs that form the α/β lobes [
99,
144,
145], highlighting the aging component of neurodegeneration. It is also possible that the differences in the onset and effects of TDP-43 proteinopathy we observed in γ lobe MBNs may be due to their remodeling during metamorphosis [
99], which may increase their vulnerability to TDP-43. Taken together, these findings parallel differential susceptibility observed in human cortical neurons [
146,
147]. Given the availability of split GAL4 driver lines for subsets of MBNs [
97] future studies could focus on better understanding of what drives this susceptibility by modeling TDP-43 proteinopathy in various neuronal subpopulations within the MB circuit. It will be interesting to combine these powerful genetic tools with single cell RNA seq efforts in patients and flies in order to pinpoint specific cell sub-types that may be more vulnerable.
Frontotemporal lobar degeneration with TDP-43 proteinopathy (FTLD-TDP) is pathologically classified based on the cortical layers affected, the number of dystrophic neurite, and the extent of neuronal cytoplasmic inclusions [
21]. This pathology is also commonly observed in Alzheimer’s disease and dementia with Lewy bodies cases that exhibit TDP-43 proteinopathy, although different brain regions are selectively affected [
14,
19]. Interestingly, we find that neuronal cytoplasmic inclusions are absent in juveniles, appear first in young adult brains and increase in size during aging. This neuropathology precedes observations of dystrophic neurites, which are first visible around middle age, increase in severity during aging, and ultimately degenerate. This age-dependent pathology observed in the fly brain parallels that observed in human disease [
148] and provides a platform for studying the progression of neurodegeneration in the genetically tractable
Drosophila model.
In humans, behavioral symptoms of FTD precede widespread degeneration by years suggesting that behavioral deficits arise prior to neuronal pathology, in the absence of complete loss of nuclear TDP-43 function or cytoplasmic accumulation [
149]. Therefore, we chose to focus primarily on working memory and sleep deficits using young adult flies, in which a subset of MBNs show nuclear depletion accompanied by widespread TDP-43 mislocalization to the axonal cytoplasm while MBN axons still appear largely intact. At this young age, we found both working memory and sleep fragmentation phenotypes, albeit the latter were only significant in males. We also assessed these behaviors in aged flies, however the distinct age-dependent deficits exhibited by the controls themselves confounded the detection of TDP-43 specific phenotypes.
Behavioral variant frontotemporal dementia (bvFTD) is diagnosed based on often subtle behavioral symptoms including a lack of empathy, increased apathy, dysinhibition and deficits in executive function [
150]. FTD was classically distinguished from AD using a perceived lack of memory deficits, however recent patient studies and meta-analyses indicate that memory deficits are common in bvFTD and at times indistinguishable between AD and bvFTD [
133,
151], challenging the validity of the exclusion of episodic memory deficits in the clinical diagnoses of bvFTD [
152]. Although sleep symptoms are notoriously variable and understudied across different FTD diagnoses, the most common sleep disturbances observed in bvFTD patients are sleep fragmentation and daytime sleepiness that may or may not be accompanied by insomnia [
103,
104,
153]. Our findings of limited sleep fragmentation and daytime sleepiness highlight potential differences between the fly model and human disease presentation. That said, the limited yet specific deficits caused by TDP-43 OE in MBNs provide a robust measurable sleep disruption that can serve as an organism-level output for future molecular or pharmacological intervention experiments.
In addition to observed behavioral deficits, flies with TDP-43 proteinopathy show an approximately 10% reduction in median lifespan. This may be roughly comparable to human patients where FTD exerts a subtle effect on lifespan. In humans, mean age at diagnosis is 61.9 for early onset disease (< 65 years) with mean survival after diagnosis around 8 years [
106]. Although the effect on survival could be caused by socioeconomic status and access to care, the lifespan of FTD patients is reduced in comparison with the general population in most countries were clinical data are collected [e.g., 78.8 years in the United States
154]. Effects on lifespan may be far more severe than a simple comparison of life expectancy would indicate, as Loi et al. [
107] report an increase in mortality risk for FTD patients when compared with age-matched controls. From the perspective of a fly model, it is somewhat surprising to see a lifespan effect given that MBNs comprise a small fraction of total neurons in the fly brain and are often thought to be largely dispensable for viability [
63]. There is evidence that lifespan is regulated at least in part by the α/β MBNs [
109], a subset of MBNs included in our model, however loss of key MB proteins such as DCO [
155] or over-expression of vertebrate TAU in MBNs result in severe memory deficits without affecting lifespan [
156]. Although lifespan reduction could be due to secondary effects such as food consumption or other factors, our model reproduces an important, subtle characteristic of FTD and provides additional evidence that specific populations of MBNs may play a role in lifespan regulation in flies.
Sequestration of mRNAs is one the mechanisms by which TDP-43 is thought to contribute to neurodegeneration and identifying candidate mRNA targets has helped better understand disease etiology [
39,
157]. Although our model is based on TDP-43 overexpression, we note that among the mRNAs enriched with TDP-43 in MBNs we found 12 of 15 physiological targets of TDP-43 previously identified bioinformatically as potential targets of
Drosophila TDP-43 (
i.e., TBPH) based on UG-richness [
158]. In future studies, it will be interesting to explore the relationship between splicing and translational targets of TDP-43. Furthermore, the overlap and specificity of some targets for MNs versus MBNs may provide insights into shared mechanisms and neuronal vulnerability across neurodegenerative diseases exhibiting TDP-43 proteinopathy.
Functional analyses of the mRNAs associated with TDP-43 identify numerous cellular pathways including Wg and Hippo that have been previously implicated in sleep regulation in flies [
159]. Additionally, our findings of dopamine receptors as candidate mRNA targets is consistent with findings that loss of mesocortical dopaminergic tracts and dopamine receptors in the frontal lobes could contribute to the behavioral symptoms in FTLD [
160].
For functional validation we chose to focus on
dlp, an mRNA target of TDP-43 that we previously identified in the ALS model of TDP-43 proteinopathy, and a known regulator of Wg signaling. Indeed, Wg signaling is dysregulated in ALS [
161,
162], mRNAs associated with Wg/Wnt pathway are enriched in FTD patient frontal cortices [
115] and WNT1 and Granulin (GRN), an FTD linked gene, have been shown to regulate each others’ expression in human neuronal progenitors [
163]. Despite these connections, the mechanisms by which the Wg/Wnt pathway is implicated in TDP-43 based pathophysiology in FTD or other dementias remains poorly understood. Using genetic interactions, we found that
dlp OE in MBNs mitigates TDP-43 dependent working memory deficits, as evidenced by improved alternation in the Y maze assay. Taken together, these results support the notion that Wg/Wnt signaling is altered in TDP-43 associated neurodegeneration and modulating its activity via Dlp mitigates behavioral deficits caused by TDP-43 proteinopathy. Further substantiating these findings, we found that the expression of
GPC6 mRNA, a human ortholog of
dlp, is altered in FTD patient brains [
94], specifically in neuronal nuclei that exhibit the molecular signature of TDP-43 depletion (
i.e., cryptic exon inclusion in specific TDP-43 transcriptional targets), consistent with a link between Dlp/GPC6 and TDP-43 pathology. We speculate that this increase in
GPC6 mRNA in neurons with TDP-43 nuclear depletion may reflect a compensatory upregulation caused by a cytoplasmic decrease in protein levels that we detected in
Drosophila MBN axons. Indeed, the same study by Gittings, Alsop et al. [
94], identifies an increase in the expression of genes related to oxidative phosphorylation, ATP synthesis and energy metabolism in the cryptic exons containing cells isolated from FTD patients, suggesting a compensatory mechanism to counteract documented reductions in ATP synthesis and overall mitochondrial function [reviewed in
164]. Although it is also possible that Dlp/GPC6 are differently regulated in flies and humans, the link to TDP-43 proteinopathy is preserved, highlighting the ability of the
Drosophila models to identify functionally relevant disease targets. Interestingly, recent GWAS studies identified
GPC6 as a risk factor for AD in African Americans [
116,
117]. In future studies it will be interesting to see how different mRNA targets, beyond
dlp/GPC6 mitigate specific phenotypic aspects of FTD and identify additional targets across the spectrum of TDP-43 proteinopathies.