TDP-43 proteinopathy alters the ribosome association of multiple mRNAs including the glypican Dally-like protein (Dlp)/GPC6

Amyotrophic lateral sclerosis (ALS) is a genetically heterogeneous neurodegenerative disease in which 97% of patients exhibit cytoplasmic aggregates containing the RNA binding protein TDP-43. Using tagged ribosome affinity purifications in Drosophila models of TDP-43 proteinopathy, we identified TDP-43 dependent translational alterations in motor neurons impacting the spliceosome, pentose phosphate and oxidative phosphorylation pathways. A subset of the mRNAs with altered ribosome association are also enriched in TDP-43 complexes suggesting that they may be direct targets. Among these, dlp mRNA, which encodes the glypican Dally like protein (Dlp)/GPC6, a wingless (Wg/Wnt) signaling regulator is insolubilized both in flies and patient tissues with TDP-43 pathology. While Dlp/GPC6 forms puncta in the Drosophila neuropil and ALS spinal cords, it is reduced at the neuromuscular synapse in flies suggesting compartment specific effects of TDP-43 proteinopathy. These findings together with genetic interaction data show that Dlp/GPC6 is a novel, physiologically relevant target of TDP-43 proteinopathy. Supplementary Information The online version contains supplementary material available at 10.1186/s40478-021-01148-z.


Introduction
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder characterized by the progressive loss of motor neuron function, culminating in death due to respiratory failure [32,79]. Mutations in several dozen genes including SOD1 [87], TDP-43 [45,98], and C9orf72 [19,84] have been implicated in disease pathogenesis, however for > 79% of patients the cause of disease remains unknown [52], suggesting that yet-to-be-uncovered mechanisms significantly contribute to the majority of ALS cases. At the cellular level, several processes have been linked to ALS including neuroinflammatory response [63], cellular metabolism [22,106], RNA processing, axonal transport and protein homeostasis [105]. A hallmark of ALS is the finding that > 97% of patient spinal cord motor neurons exhibit cytoplasmic puncta containing the RNA binding protein TDP-43 (TAR DNA Binding Protein 43) [59,77]. Together with reports that 2-3% of ALS patients harbor TDP-43 mutations [45,77,98,109], these findings suggest a key role for TDP-43 in ALS pathogenesis regardless of etiology.
TDP-43 is a nucleo-cytoplasmic shuttling, DNA/RNA binding protein which is predominantly localized to the nucleus where it regulates transcription [101] and splicing [26,108]. In the cytoplasm, TDP-43 regulates stress granule dynamics [21,25,46,72] as well as axonal and dendritic mRNA localization and translation [1,12,15]. In disease, TDP-43 is depleted from the nucleus, causing splicing defects, derepression of cryptic exons [57] and increased retrotransposon expression [49,61,75,103]. TDP-43 has been shown to induce toxicity through both nuclear loss-of-function and cytoplasmic gain-offunction mechanisms [93,110,113]. When mislocalized to the cytoplasm, TDP-43 associates with a plethora of RNA containing complexes including stress and transport granules [60,89,111], as well as protein complexes devoid of RNA [31,69]. TDP-43 has also been shown to influence the translation of specific mRNAs, both as a negative and a positive regulator [12,14,15,67,76]. Taken together these findings suggest a complex role for TDP-43 in translation regulation, which has been linked to the formation of TDP-43 cytoplasmic puncta in disease and altered protein expression and/or localization of its mRNA targets.
A potential mechanism by which TDP-43 cytoplasmic inclusions dysregulate translation is by altering the ribosomes' access to mRNAs, as suggested by the "ribostasis hypothesis" [83]. Further substantiating this model are recent reports of TDP-43 mediated translation inhibition of mRNAs, including futsch and hsc70-4 mRNAs (in axons) [14,15] and Rac1 mRNA (in dendrites) [12], all of which were identified through candidate approaches. However, the relationship between TDP-43 associated mRNAs and translation in disease has not yet been examined through an unbiased approach. Here we combined RNA immunoprecipitations (RIP) and translating ribosome affinity purification (TRAP) to identify mRNAs that simultaneously satisfy two criteria: 1) are enriched in TDP-43 complexes and thus have the potential to be regulated by TDP-43, and 2) their association with ribosomes is altered in the context of Drosophila models of TDP-43 proteinopathy, consistent with an effect on translation. RIP experiments uncovered candidate mRNA targets linked to neuromuscular junction development and synaptic growth. TRAP uncovered mitochondrial metabolism, proteostasis and wingless signaling as significant components of the "normal" motor neuron translatome. When compared to controls, TDP-43 proteinopathy was found to significantly alter the ribosome association of multiple mRNAs whose cognate proteins play roles in splicing, purine metabolism and mitochondrial electron transport among others.
Here we identify Dally-like protein (Dlp), a glypicantype heparan sulfate proteoglycan (HSPG) that has been described as a regulator of wingless (Wg/Wnt) [120] and Liprin alpha receptor (LAR) mediated signaling at the neuromuscular junction [100] as a candidate target of TDP-43 mediated translation inhibition. We show that dlp mRNA is enriched in TDP-43 complexes and sequestered in insoluble aggregates, consistent with the ribostasis hypothesis. Consistent with this, we find that Dlp protein is significantly reduced at the neuromuscular junction while steady state levels of dlp transcript remain unchanged, further supporting the possibility of translation inhibition in axons and/or at synapses. Surprisingly, within the ventral nerve cord neuropil, Dlp accumulates in puncta, suggesting that TDP-43 proteinopathy affects Dlp in a compartment specific manner. Dlp depletion at the NMJ, but not puncta formation, was also observed with endogenous TDP-43 knockdown suggesting that some Dlp phenotypes are the result of loss of nuclear TDP-43 function while others are the results of toxic cytoplasmic gain of function. Genetic interaction experiments show that dlp overexpression mitigates TDP-43 dependent locomotor deficits, consistent with the notion that Dlp mediates aspects of TDP-43 proteinopathy in vivo. Lastly, GPC6 protein, a human homolog of Dlp, exhibits aggregate-like accumulations while GPC6 mRNA is enriched in insoluble fractions derived from ALS patient spinal cords, mirroring the findings from Drosophila motor neurons. Together, these findings highlight key pathways altered in the motor neuron translatome in the context of TDP-43 proteinopathy and support the notion that altered expression of the glypican Dlp/GPC6 contributes to motor neuron degeneration.

RNA isolation
RNA was isolated from the ventral cord lysates, whole larvae inputs, or immunoprecipitations using a Qiagen RNeasy Mini Kit (Qiagen 74104). For fractionation experiments, dissected ventral nerve cords (VNCs) or neuromuscular junctions (NMJ), RNA was isolated using Trizol (Thermofisher Scientific 15596026). With both isolation methods, RNA was eluted into 20 μl of molecular grade water with quality and quantity determined by a nanodrop spectrophotometer, measuring absorbance at 260/280 nm.

RNA Seq
Isolated RNA was prepared for RNA sequencing using the SMART-seq v4 Ultra Low Input RNA kit (Takara Bio USA 634888). The Nextera XT DNA Library Prep Kit (Illumina FC-131-1024) was used to tag, clean, and pool the samples with quality assayed by an Agilent 2100 Bioanalyzer. The cDNA libraries for RIP/TRAP were sequenced by the Beijing Genomics Institute using an Illumina HiSeq 4000 and the Illumina Nextera XT Kit (Illumina 15032350) with 100-bp paired end reads. Human fractionation RNA-seq were performed at the Translational Genomics Institute (TGen) using similar protocols.

RNA-seq analysis
Reads were quality filtered to > Q30 then aligned to the Flybase FB2018_05 genome using MAFFT [85] with differential expression analysis conducted through Deseq2 [64]. In RIP experiments, 9812 and 9954 genes were detected in TDP-43 WT and TDP-43 G298S complexes, respectively. In TRAP experiments, 9711, 9347, and 9669 genes were detected in RpL10 GFP complexes in the control, TDP-43 WT , and TDP-43 G298S models, respectively. To identify the "normal" motor neuron translatome we calculated the Log2FoldChange between RpL10 IP and RpL10 VNCs. Next, to identify alterations resulting caused by TDP-43 proteinopathy to the motor neuron translatomes, we subtracted the Log2FoldChange between RpL10 IP and RpL10 VNCs from the Log2FoldChange between TDP-43 WT or G298S RpL10 IP and TDP-43 WT or G298S RpL10 IP VNCs. Gene ontology (GO) analyses were conducted using David 6.8 [42,43]. Human spinal cord transcriptomic data underwent Deseq2 analysis from counts provided by target ALS, detecting 33,843 genes. For the human fractionation RNA-seq data, reads were aligned to the NCBI GRCh38 genome with GENCODE 29 genome using STAR 2.5.3a [22] then underwent Deseq2 differential expression analysis detecting 24,090 genes. GO plot 1.0.2 [114], STRING 11.0 [102], and the R package "LPS" version 1.0.10 were used to generate bio-informatic figures.

Fractionations
Fractionations of both overexpression and CRISPR Drosophila models were conducted as previously described [14]. In brief, 25 third instar larvae (for the overexpression models) or 1-2 days old 25 homozygous flies (for the CRISPR lines) were homogenized in Trizol (Thermofisher Scientific 15596026). Lysates were then centrifuged at 25,000×g for 30 min. The supernatant became the soluble fraction and the pellet was solubilized in urea buffer (30 mM Tris, 7 M Urea, 2 M Thiourea, 4% CHAPS, 1X Protease Inhibitor Cocktail (Millipore Sigma 11873580001), 0.5 mM PMSF, RNAsin Plus 400 units/ ml (Fischer Scientific PRN2615), pH 8.5) to generate the urea/insoluble fraction. For human fractionation samples, post-mortem tissue (four spinal cord and two cerebellum samples as controls) was homogenized and subjected to the fractionation protocol described above.

Immunofluorescence (Drosophila tissues)
Third instar ventral nerve cords were dissected as previously described [24]. In brief, 3rd instar larvae were placed in saline on a Sylgard dissection plate and the ventral nerve cord was dissected for further processing. Primary antibodies/stains used were 1:5 anti-Dlp (antibody 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 52,242, targets amino acids V523 to Q702) and 1:200 anti-GFP (Rockland 600-102-215). DNA was visualized using Hoechst (1:10,000, Invitrogen H3570). Goat anti-mouse Alexa 568 (1:500 Thermofisher Scientific A-11004) was used as secondary antibody. Samples were imaged using a Zeiss LSM 880 inverted confocal microscope with a 40X oil lens. The number and cumulative area of Dlp puncta was quantified via manual counting and outlining of blinded VNC images.
Third instar larval NMJs were dissected as previously described [24]. In brief, 3rd instar larvae were pinned to Sylgard dissection plates and immersed in saline. The posterior end of the larvae was removed and the larvae were cut down the dorsal midline. Larvae were further pinned laterally to expose the lateral muscles innervated by motor neuron axons. Primary antibodies used were 1:5 anti-Dlp (antibody 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), 1:100 anti-HRP Alexa 647 (Jackson Immuno-research 123-605-021). Goat anti-mouse Alexa Fluor 568 (1:500 Thermofisher Scientific A-11004) was used as secondary antibody. Samples were imaged using a Zeiss LSM 880 NLO upright confocal microscope with a 40X oil lens. Dlp intensity per area was quantified with NIH imageJ v1.52p within four terminal boutons defined by HRP staining. All images were blinded prior to being analyzed.

Immunohistochemistry (human post-mortem spinal cords)
Paraffin-embedded post-mortem spinal cord tissue sections were obtained from the Barrow Neurological Institute ALS Tissue Bank and Target ALS Postmortem Tissue Core. All tissue samples were collected after informed consent from the subjects or by the subjects' next of kin, complying with all relevant ethical regulations and approved by the appropriate Institutional Review Boards. Clinical neuropathological diagnoses were made by board certified neuropathologists according to consensus criteria. Subject demographics are listed in Additional file 4: Table S7-1. A total of 16 ALS and 7 controls (7 non-neurologic disease controls) were used in this study. All ALS cases have TDP-43 pathology while the non-neurologic disease controls lack TDP-43 pathology. Tissue sections were processed as previously described [5]. For immunohistochemistry (IHC) all tissue sections were deparaffinized, rehydrated and antigen retrieval performed using Target Antigen Retrieval Solution, pH 9.0 (Dako) in a steamer. Super Block (Scytek), supplemented with Avidin (Vector Labs) was used to block non-specific binding sites for 1 h. All sections were incubated overnight using rabbit polyclonal primary antibody GPC6 protein at 1:600 (Bioss bs-2177R) diluted in Super Block supplemented with Biotin (Vector Labs). A biotinylated goat anti-rabbit IgG (H + L) was used as secondary antibody (Vector Laboratories, BA-1000, 1:200) diluted in Super Block for 1 h. Immunostaining was visualized using the Vectastain Elite ABC reagent (Vector Labs) and Vector ImmPACT NovaRED peroxidase substrate kit (Vector Labs). Slides were counterstained with hematoxylin (Sigma Aldrich). Sections were visualized using an Olympus BX40 microscope. Images were blinded, then motor neurons were outlined and inverted in NIH ImageJ, which was also used to quantify GPC6 protein intensity relative to background. CellProfiler 3.1.9 [73] was used to quantify GPC6 puncta, defined between 3 to 13 pixels in diameter.

qRT-PCR
All cDNA was synthesized using Fisher First Strand cDNA synthesis reagents (Thermofisher Scientific K1641). qPCR reactions were conducted in three biological replicates using Taqman Fast Advanced Master Mix (Thermofisher Scientific 4444556) and conducted on either an ABI 7100 or Analytik Jena 844-00504-4 qTOWER qPCR machine. dlp mRNA was quantified using Taqman assay (Thermofisher Scientific Dm01798597_m1) using GPDH (Thermofisher Scientific Dm01841185_m1) for normalization. Fold change was calculated using the standard ΔΔCT method [81].

Locomotor assays
Third instar wandering larvae were placed on a grape juice agar plate. After being given 1 min to acclimate, the larvae were turned on the ventral side and timed until they uprighted themselves and began the first forward motion [15]. Data points outside of the median turning time ± 1.5IQR within each genotype were identified as outliers and excluded from statistical analyses.

Fluorescent in situ hybridization (FISH)
FISH was performed using RNAscope (ACD Bio). A custom probe generated by ACD bio for dm-dlp and a control probe dm-Gapdh1 were used to detect dlp and Gapdh1 in dissected VNC and NMJs as previously described [115]. Images were acquired using a Zeiss LSM 880 inverted confocal microscope with a 40X oil lens.

Statistical analyses
For RNA seq analyses, significance was determined using the P adjusted value calculation embedded within Deseq2. All other statistical analyses were performed using GraphPad Prism 7.0 using two-tailed analyses. Normality was evaluated using the Kolmogorov-Smirnov test. All data shown are mean ± standard error of the mean.

mRNAs enriched with TDP-43 in motor neurons encode proteins linked to neuronal and synaptic signaling
TDP-43 has been shown to associate with a plethora of RNA targets and regulate various aspects of RNA processing including RNA localization and translation [1,15,76]. These findings suggest that cytoplasmic TDP-43, which is a hallmark of disease, has multiple opportunities to cause alterations in the motor neuron proteome and contribute to pathogenesis. To uncover these alterations, we defined TDP-43 dependent changes in the motor neuron translatome in vivo using Drosophila models of TDP-43 proteinopathy that recapitulate key aspects of the disease including locomotor defects, cytoplasmic aggregates and reduced lifespan [23,24]. We hypothesized that while some of the translational alterations may be an indirect consequence of neurodegeneration, others are directly caused by TDP-43, possibly through mRNA association with, and sequestration into, TDP-43 cytoplasmic complexes. To distinguish between direct versus indirect effects on the translatome we set out to identify mRNAs that are both enriched with TDP-43 and translationally dysregulated in motor neurons in vivo using a combination of RNA immunoprecipitations (RIP) and Translating Ribosomes Affinity Purifications (TRAP) [39,107] in the context of TDP-43 proteinopathy. To this end, we first used Drosophila larvae expressing human TDP-43 protein specifically in motor neurons via the GAL4-UAS system (D42 GAL4 > UAS-TDP-43-YFP) and performed immunoprecipitation experiments to pull down TDP-43 and associated mRNAs (RIP). The mRNAs associated with TDP-43 (WT or mutant G298S) were isolated and subjected to RNA sequencing (see Materials and Methods, and Fig. 1a). We have previously shown that TDP-43 WT and TDP-43 G298S are expressed at comparable levels [24], thus any differences between these transgenics indicate variant specific alterations rather than different expression levels. Bioinformatic analyses identified several mRNAs significantly enriched with TDP-43, as determined by comparing TDP-43 associated mRNAs in a complex with the transcriptome of dissected ventral nerve cords expressing the appropriate TDP-43 variant (TDP-43 WT or G298S , see Fig. 1b,c). Of the approximately 9,900 transcripts detected in ventral nerve cords (Additional file 1: Table S1-1), 1,892 were enriched with, and shared by, both TDP-43 variants (Log2FC > 2, P adj < 0.05). Additionally, about 25% of mRNAs enriched with TDP-43 were unique to TDP-43 WT or TDP-43 G298S (662 mRNAs enriched with TDP-43 WT and 616 mRNAs enriched with TDP-43 G298S , respectively, see Fig. 1d) consistent with previous evidence that wild-type and mutant TDP-43 variants cause toxicity in part through distinct mechanisms (see also Fig. 1e, f ) [14].

TDP-43 proteinopathy induces a broad range of translational changes in motor neurons
Next, to define the in vivo motor neuron translatome and identify changes in translation induced by TDP-43 proteinopathy (see Materials and Methods, and Fig. 2a) we conducted TRAP experiments [39,107]. To this end, RpL10-GFP was expressed in motor neurons using D42 GAL4 either on its own, or together with TDP-43 WT , or TDP-43 G298S . Following ribosome immunoprecipitations we isolated total RNA and conducted RNA seq, detecting ~ 9,500 mRNAs (9,711 mRNAs for RpL10 controls, 9,347 for TDP-43 WT and 9,669 for TDP-43 G298S , see Materials and Methods, Additional file 2: Table S2-1). We first defined the "normal" motor neuron translatome using the normalized gene counts associated with ribosome immunoprecipitations from RpL10-GFP controls (Additional file 2: Table S2-2). Using David 6.8 [42,43], we determined that genes enriched in the top 10% of the motor neuron translatome normalized counts were related to proteostasis and energy metabolism (Fig. 2b, Additional file 2: Table S2-3), consistent with high levels of protein turnover [2] and energy demands in neurons [94].
To identify changes in translation caused by TDP-43 proteinopathy in motor neurons in vivo, we subsequently compared the mRNAs enriched with ribosomes precipitated from Drosophila larvae expressing RpL10-GFP TDP-43 WT or RpL10-GFP TDP-43 G298S to RpL10-GFP controls after normalizing to the transcriptomes of dissected ventral nerve cords of the appropriate genotype (RpL10-GFP TDP-43 WT , RpL10 GFP TDP-43 G298S or RpL10-GFP). These comparisons identified several genes enriched with or depleted from motor neuron ribosomes in both of the ALS models relative to RpL10-GFP controls All genes associated with TDP-43 (log2FC > 0) are displayed regardless of significance. c) proteinopathy, relative to transcript levels in the ventral nerve cord. d-g GO terms for TDP-43 enriched genes (log2FC > 2, P adj < 0.05) either unique to TDP-43 WT (e) or TDP-43 G298S (f) or shared between genotypes (g) (ΔLog2FC > 1 or < −1). Subsequent GO term analyses using David 6.8 [42,43] revealed that in the context of TDP-43 WT , spliceosome components were translationally dysregulated (Fig. 2c, Additional file 2: Table S2-4), suggesting that TDP-43 WT overexpression causes pathway alterations previously associated with nuclear lossof-function phenotypes [3,18]. Interestingly, in the context of TDP-43 G298S , we found significant alterations in neuronal metabolism (Additional file 2: Table S2-5), including purine processing leading into the pentose phosphate pathway (PPP), which provides a mechanism for countering oxidative stress via increased NADPH production [50]. PPP was found to be down-translated, consistent with our previous studies showing that glucose 6 phosphate dehydrogenase (G6PD), the rate limiting enzyme in PPP is altered in flies and patient tissues with TDP-43 pathology [70] (Fig. 2d). Further validating this finding, overexpression of G6PD mitigated TDP-43 G298S induced locomotor dysfunction (Supplemental Fig. 2-1a). Interestingly, oxidative phosphorylation genes encoding components of the electron transport chain showed both increased and reduced association with ribosomes (Fig. 2e). Recent findings of reduced Complex I activity would suggest that increased translation of Complex I components identified via TRAP may reflect a compensatory mechanism [22,106,117].
Interestingly, genes salient to translation exhibited altered translation (Fig. 2f ). In both the TDP-43 WT and TDP-43 G298S models, ribosomal RNA maturation was down-regulated (Fig. 2g, Additional file 2: Table S2-6) while cytoplasmic translation components were up-regulated relative to the control (Fig. 2h, Additional file 2: Table S2-7). These findings are consistent with a previous study in which patient derived iPSC neurons exhibited a compensatory increase in global translation [104]. Further corroborating our findings with previous studies, we detected increased association with ribosomes for CG6762, the Drosophila ortholog of human SRXN1, (TDP-43 WT , Log2FC = 0.673, P adj = 2.79E−32; TDP-43 G298S , Log2FC = 0.328, P adj = 2.50E−51), one of 14 genes that exhibited increased translation in a human cell model of TDP-43 A315T proteinopathy [76]. Taken together, our ribosomal tagging experiments highlight complex changes in translation, some of which may be direct consequences of TDP-43 proteinopathy while others may reflect compensatory mechanisms.

A fraction of TDP-43 associated mRNAs are altered in their association with ribosomes in the context of TDP-43 proteinopathy
To distinguish between translational changes that may be caused directly by TDP-43 versus compensatory alterations, we compared genes altered in their association with ribosomes to those enriched in TDP-43 complexes, in the context of TDP-43 proteinopathy. These analyses show that approximately 1/3 of the genes depleted from ribosomes were also enriched in TDP-43 complexes (Fig. 3a, Additional file 3: Table S3-1, 30.6% for TDP-43 WT , 37.6% for TDP-43 G298S ) suggesting that these may be direct targets of translation inhibition. GO term analyses of these TDP-43 enriched, TRAP depleted genes using David 6.8 [42,43] identified pathways that have previously been associated with ALS including GPCR signaling [28,97], NMJ Development [28,97], Autophagy [6,116], ER Organization [112], Immune Response [65] and Oxidation Reduction [95] among others ( Fig. 3b-e, 3: Table S3-2,3). Interestingly, a similar proportion (Fig. 3f, Additional file 3: Table S3-4, TDP-43 WT 27.6%, TDP-43 G298S 34.9%) of the genes enriched with ribosomes in the context of TDP-43 proteinopathy were also associated with TDP-43 complexes (Fig. 3f, Additional file 3: Table S3-4), consistent with previous finding that TDP-43 can also function as a positive regulator of translation [76]. GO term analyses identified pathways such as Membrane Potential Stability and Transmembrane Transport-K+ ( Fig. 3g-j, Additional file 3: Table S3-4) comprising genes related to membrane excitability, a process known to be altered in ALS [74]. We note that although GO term analysis is useful to identify translational alterations affecting multiple components of the same pathway, the genes encompassed by these GO terms represent only a fraction of genes enriched and translationally dysregulated in the context of TDP-43 (Fig. 3a, f, 13.7% for TDP-43 WT , 12.4% for TDP-43 G298S ). This suggests that in addition to specific pathways, a plethora of individual genes might represent salient targets of TDP-43 mediated translational inhibition or activation.
(See figure on next page.) Fig. 2 Translational Alterations Induced by TDP-43 Proteinopathy. a Experimental schematic for RNA immunoprecipitations of human TDP-43, specifically from the motor neurons of third instar larvae. b STRING clusters from genes in the top 10% of normalized counts associated with RpL10-GFP in w 1118 larvae (control motor neuron translatomes). c Altered ribosome association of spliceosome genes in TDP-43 WT relative to the control. d Altered ribosome association of purine metabolism genes in TDP-43 G298S relative to control. e Altered ribosome association of oxidative phosphorylation genes in TDP-43 G298S relative to control. f Altered ribosome association of translation associated genes in TDP-43 G298S relative to control. g, h GO terms for RpL10 depleted genes (log2FC < −1) (g) and RpL10 enriched genes (log2FC > 1) (h) shared between both TDP-43 WT

dally-like protein (dlp) mRNA, a glypican involved in wingless (Wg/Wnt) signaling is sequestered in insoluble complexes in Drosophila models of TDP-43 proteinopathy
Additional analyses of the "normal" motor neuron translatome using STRING [102] identified an enrichment in Wg/Wnt signaling genes (Fig. 4a), with the top 5% of genes translated in motor neurons containing Wg/Wnt signaling genes at a rate 61% greater than that of the entire data set. This is consistent with the established role of Wg/Wnt signaling in motor neurons and at the neuromuscular junction [47]. Interestingly, Wg/Wnt signaling has been shown to be dysregulated in ALS [11,34,35], although its mechanisms remain poorly understood, in part due to the complexity of Wg/Wnt signaling and cross-talk with other pathways. Given these observations, we queried the pathway for candidate targets of TDP-  Table S4-1). dlp encodes a glypican, a member of the heparan sulfate proteoglycan (HSPG) family, localized to the plasma membrane via a GPI anchor or secreted in the extracellular matrix.
Next, to determine whether dlp mRNA is insolubilized as predicted by the ribostasis hypothesis, we conducted subcellular fractionations and quantified dlp mRNA in the soluble and insoluble/urea fractions from third instar larvae expressing TDP-43 in motor neurons. These experiments showed that dlp mRNA is significantly insolubilized in the context of TDP-43 G298S (Fold change = 1.99, P value = 0.0256, Fig. 4d) but not in the context of TDP-43 WT overexpression. Whether dlp mRNA is insolubilized in the same complex as TDP-43 G298S or in a different complex remains to be determined.
Further substantiating this finding, dlp mRNA was insolubilized in a TDP-43 G294A (Fold change = 2.51, P value = 0.034, Fig. 4e) but not in the TDP-43 WT CRISPR fly model of ALS in which the endogenous Drosophila TDP-43 (TBPH) gene has been replaced with human TDP-43 [9]. Taken together these results indicate that dlp mRNA is a novel TDP-43 candidate target that becomes insolubilized in the context of TDP-43 proteinopathy and are consistent with previous reports of Wg/Wnt alterations in disease.

Dlp protein levels are altered in the neuromuscular system in the context of TDP-43 proteinopathy
A possible outcome of dlp mRNA enrichment with TDP-43 complexes, insolubilization and decreased association with ribosomes is inhibition of Dlp protein synthesis. To test the impact of TDP-43 proteinopathy on Dlp protein expression we conducted immunofluorescence experiments in the larval neuromuscular system. We examined Dlp expression at the neuromuscular junction (NMJ) and the ventral nerve cord (VNC), which comprises motor neuron cell bodies and compactly packaged neurites, particularly dendrites [29] (Fig. 5a-z). These experiments showed that overexpression of TDP-43 WT or TDP-43 G298S , and RNAi knock-down of TBPH (TBPH RNAi ) are sufficient to deplete Dlp protein within NMJ boutons relative to the w 1118 control (TDP-43 WT , decreased 35.7%, P value = 0.00920; TDP-43 G298S , decreased 37.2%, P value = 0.00860; TBPH RNAi , decreased 42.5%, P value = 0.00920; Fig. 5m). While no major changes in Dlp expression were detected in the motor neuron cytoplasm within the VNC (data not shown), striking aggregate-like structures were observed in the neuropil when either TDP-43 WT or TDP-43 G298S were overexpressed (see Fig. 5n-y). Quantification shows that both TDP-43 WT and TDP-43 G298S proteinopathy cause a statistically significant increase in Dlp granule number (TDP-43 WT , Fold Change = 4.99, P value = 1.77E−4; TDP-43 G298S , Fold Change = 4.24, P value = 8.49E−4) and cumulative granule area relative to the w 1118 controls. In contrast, TBPH RNAi did not cause significant Dlp granularity phenotypes (TBPH RNAI , Fold Change = 2.03, P value = 0.262, Fig. 5n-z and Additional file 5: Figure S5-1a).
These results indicate that TDP-43 proteinopathy causes altered Dlp protein expression, more specifically a reduction at synaptic terminals and aggregate like structures in the VNC neuropil. To investigate whether these alterations are post-transcriptional, as suggested by the RIP and TRAP experiments, we used fluorescence in situ hybridization (RNAScope) to examine dlp mRNA localization in the VNC and at the NMJ. Since we were not able to detect a specific signal, possibly due to low levels of expression (data not shown) we performed RT-qPCR experiments using dissected VNCs and larval NMJ preparations and found no significant changes in dlp mRNA levels between TDP-43 proteinopathy models and controls (Additional file 5: Figure S1b-c). Although the VNC and NMJ qPCR samples contain multiple cell types, which affects our ability to detect mRNA differences caused by TDP-43 overexpression specifically in motor neurons, these experiments collectively suggest that Dlp expression is likely regulated by TDP-43 post-transcriptionally, possibly through a combination of translation and axonal transport.

dlp is a modifier of TDP-43 dependent locomotor deficits
Next, to test whether Dlp is an effector of TDP-43 proteinopathy, we overexpressed dlp using the motor neuron specific driver D42 GAL4 on its own and in the context of TDP-43 WT or TDP-43 G298S and measured its effect on TDP-43 induced locomotor defects. Using larval turning assays to measure locomotor function we found that although dlp overexpression (dlp OE ) on its own resulted in a higher turning time (15.3 ± 1.01 s compared to w 1118 controls 8.95 ± 0.46 s, P value = 2E−6), in the context of either TDP-43 WT or TDP-43 G298S disease models, it caused a lower turning time, indicating a rescue of locomotor dysfunction (11.30 ± 0.56 s compared to 13.40 ± 0.69 s for TDP-43 WT alone, P value = 0.041; 12.50 ± 0.83 s compared to 16.40 ± 1.16 s for TDP-43 G298S alone, P value = 6.4E−3, see Fig. 6a). Interestingly, although TBPH RNAi knock-down also caused locomotor defects (18.75 ± 1.69 s compared to 8.95 ± 0.46 s for w 1118 , P value = 1.66E−07), dlp overexpression in this context is not sufficient to mitigate TBPH RNAi induced locomotor dysfunction (16.13 ± 1.72 s compared to 18.75 ± 1.69 s for TBPH RNAi alone, P value = 0.21, see Fig. 6a).
Consistent with dlp OE mitigating TDP-43 induced locomotor defects, dlp knockdown by RNA interference (dlp RNAi ) enhanced locomotor deficits in both ALS models based on TDP-43 overexpression (16.9 ± 1.21 s for TDP-43 WT dlp RNAi compared to 14.00 ± 0.76 s for TDP-43 WT alone, P value = 0.017; 29.30 ± 2.56 s for TDP-43 G298S dlp RNAi compared to 16.10 ± 0.73 s for TDP-43 G298S alone, P value = 1.9E−4, see Fig. 6b). In contrast, dlp RNAi did not modify TBPH RNAi induced locomotor dysfunction (15.6 ± 1.00 s for TBPH RNAi , dlp RNAi compared to 17.90 ± 1.15 s for TBPH RNAi alone, P value = 0.219, see Fig. 6b), suggesting that locomotor defects caused by TBPH RNAi are likely due to a different mechanism than those caused by TDP-43 overexpression, which appear to be mediated at least in part by Dlp. Surprisingly, dlp R-NAi knock-down caused higher larval turning times on its own (17.5 ± 0.76 s compared to 8.64 ± 0.45 s for attp40 controls, P value = 1.4E−5, Fig. 6b) similar to dlp OE . These data suggest that finely tuned levels of Dlp in motor neurons are required for proper locomotor function.
We next asked whether the rescue of TDP-43 induced locomotor dysfunction by dlp OE could be explained by Quantification of Dlp granule number in the neuropil. N = 10 for w 1118 , 10 for TDP-43 WT , 9 for TDP-43 G298S , 6 for TBPH RNAi . Genotypes and stainings, as indicated. Scale bars: a 10 μm, n 50 μm. v 22 μm. Significance determined using the Holm Sidak's multiple comparison test. *P value < 0.05, **P value < 0.01, ***P value < 0.001. Error bars represent SEM restoration of Dlp protein levels at the NMJ. To address this, we evaluated Dlp levels within synaptic boutons when dlp was overexpressed on its own, in a w 1118 background (dlp OE ) or co-overexpressed with TDP-43 WT , TDP-43 G298S or TBPH RNAi (Fig. 6c-n). First we noticed that dlp OE on its own caused a visible increase (3.88 times higher, P value = 1.654E−6) in Dlp levels at the NMJ compared to w 1118 controls (Additional file 5: Figure S6-1a-g) therefore all samples with dlp OE were imaged at a lower gain than those without, in order to avoid saturation. While this prevented us from quantitatively comparing samples with and without dlp OE , we were able to evaluate Dlp levels within each set of samples (Figs. 5,  6). These experiments show that Dlp levels in terminal boutons are equal to, or greater than those observed in the dlp OE alone (TDP-43 WT OE . Genotypes and stainings, as indicated. Scale bars: c 10 μm. *P value < 0.05, **P value < 0.01, ***P value < 0.001. Error bars represent SEM P value = 1.00). Together with our findings that dlp OE mitigates TDP-43 proteinopathy but not TBPH RNAi induced locomotor defects, these results indicate that Dlp levels at the NMJ are critical in the context of TDP-43 proteinopathy but not TDP-43 loss, and support a cytoplasmic gain of function rather than a nuclear loss of function mechanism. It is surprising that dlp OE in the context of TDP-43 WT causes a statistically significant increase in Dlp levels at the NMJ; this suggests a potentially complex regulatory relationship between Dlp and TDP-43 proteinopathy that warrants further investigation.
To further probe the relationship between TDP-43 induced neurodegeneration, Dlp, and Wg/Wnt signaling we used genetic interaction approaches between TDP-43 and the Frizzled 2 (Fz2) receptor, an established Dlp interactor [120]. These experiments showed that while overexpression of wild-type Fz2 had no significant effect on TDP-43 induced locomotor dysfunction (Additional file 5: Figure S6-1h), RNAi knock-down or overexpression of a dominant negative form of Fz2 caused significant larval lethality in the context of TDP-43 proteinopathy (< 10 larvae alive among > 200 total larvae, data not shown). These findings suggest that Wg/ Wnt signaling may already be compromised at the NMJ in TDP-43 proteinopathy, albeit the precise mechanism remains to be determined.

The expression of GPC6, a human Dlp ortholog is altered in patient spinal cords
To validate our findings from Drosophila in human patient derived tissues, we evaluated the expression and sub-cellular distribution of the Dlp human ortholog GPC6 [41] in lumbar spinal cord from ALS patients compared to non-neurological controls using immunohistochemistry. These experiments showed that GPC6 appears increased and more granular in ALS spinal cords compared to controls (Fig. 7, Additional file 4: Figure  S7-1, for patient demographic information see Additional file 4: Table S7-1). Quantification of granule numbers showed that significantly more aggregate-like GPC6 puncta were present in ALS patient motor neurons compared to controls (Fold Change = 6.56, P value = 0.0020, see Fig. 7c). Interestingly, the increased granular appearance of GPC6 in ALS spinal cords resembles the Dlp protein levels and localization observed in the Drosophila VNCs overexpressing TDP-43 ( Fig. 5n-z). Additionally, similar to our findings in the fly models of TDP-43 proteinopathy (Fig. 4d, e), soluble vs insoluble fractionations of post-mortem patient tissue revealed that the mRNAs of both dlp orthologs GPC4 and GPC6 were enriched in the insoluble fraction of spinal cords with TDP-43 pathology relative to the insoluble fraction of occipital lobe control tissue from the same patients (N = 4 patients, GPC4 Log2FC = 2.55, P adj = 9.62E−17; GPC6 Log2FC = 1.06, P adj = 0.0122, Additional file 5: Table S7-2). Together, these findings show that GPC4/6 alterations in patient tissues resemble those identified in the Drosophila models of TDP-43 proteinopathy and suggest that glypican function may also be altered at neuromuscular synapses in ALS, consistent with a recent report of GPC4/6 reduction in SOD1 mice [7].
Here, using a Drosophila model of TDP-43 proteinopathy that recapitulates key features of the human disease including locomotor dysfunction and reduced lifespan [23,24] we identified TDP-43 dependent changes in the motor neuron translatome. These experiments identified both pathways previously associated with ALS (e.g., translation, mitochondrial function [8,96]) and novel targets of TDP-43 mediated translational inhibition including the Wg/Wnt signaling regulator dlp and the glutathione metabolism pathway. Although glutathione dysregulation has been shown to be deficient both in patient tissue [118] and patient derived neuronal cultures [17], it has not previously been identified as a direct consequence of TDP-43 pathology. Surprisingly, genes related to cytoplasmic translation are enriched in the ALS motor neuron translatomes relative to the RpL10 controls, possibly reflecting a compensatory mechanism whereby degenerating neurons upregulate global translation in response to increased cellular stress or to mitigate cytoplasmic TDP-43′s inhibitory effect on protein synthesis [88].
In addition to identifying motor neuron specific alterations to the translatome induced by TDP-43 proteinopathy, we identified dlp as a novel target of TDP-43. dlp is translated into a GPI anchored glypican that interacts with the Wg/Wnt pathway receptor Fz2, serving as both a co-factor and competitive inhibitor for substrate binding [120]. RNAi knockdown or overexpression of a dominant negative form of Fz2 resulted in lethality prior to the third instar stage in the context of both TDP-43 WT and TDP-43 G298S disease models suggesting that the Wg/ Wnt signaling pathway may be compromised by TDP-43 proteinopathy and consistent with published expression studies in patient tissues [34,35]. Although overexpression of wild-type Fz2 did not mitigate TDP-43 dependent locomotor defects as would be expected based on the lethality caused by Fz2 loss of function, it is possible that cell autonomous (i.e., motor neuron) and non-cell autonomous (i.e., glia, muscle) aspects of Wg/Wnt signaling [47] may be at play and confound our genetic interaction experiments. Alternatively, Wg/Wnt signaling and TDP-43 proteinopathy act in parallel pathways. Indeed, Dlp could be an effector of TDP-43 proteinopathy through interactions with multiple pathways including the hedgehog [119] and hippo [4] pathways that Dlp function has also been implicated in. Our findings that both dlp OE and dlp RNAi cause locomotor dysfunction suggest that Dlp levels must be exquisitely regulated in motor neurons. Indeed, Dlp overexpression mitigates but does not fully rescue TDP-43 dependent locomotor defects, perhaps due to an inability to restore wild-type Dlp levels. An alternative explanation for this result is that TDP-43 regulates additional mRNA targets.
We show that although TDP-43 WT and TDP-43 G298S both associate with dlp mRNA, and correlate with dlp depletion from ribosomes, the severity of some phenotypic alterations was variant dependent. While overexpression of dlp mRNA was sufficient to rescue locomotor dysfunction and Dlp protein levels were altered in the context of both TDP-43 variants, dlp mRNA was insolubilized only in the TDP-43 G298S model. Additionally, different pathways were shown to have altered translation between the TDP-43 WT and TDP-43 G298S models. Differential results have previously been observed between the TDP-43 WT and TDP-43 G298S models [14], which are likely the result of increased TDP-43 G298S stability [58] and granule viscosity relative to TDP-43 WT [36].
In both ALS models, analyses of the whole VNC, which contains motor neuron soma and neurites, show that Dlp forms distinct puncta whereas staining at the NMJ revealed a significant reduction in Dlp protein. These findings suggest that TDP-43 dependent translational inhibition may be localized to axons and NMJs, which can explain the relatively low magnitude depletion of dlp mRNA from ribosomes using TRAP, an approach that reflects translational changes in the entire motor neuron rather than specific compartments. It is possible that the synapse specific alteration in Dlp levels could also result from transport deficits, trapping NMJ-bound dlp mRNA/ Dlp protein in the soma or neurites. Although our qPCR data show no change in steady state dlp mRNA levels, we cannot eliminate the possibility of mRNA and/or protein transport defects leading to the accumulation of Dlp in the VNC and reduction at the NMJ.
Interestingly, TBPH RNAi was sufficient to deplete Dlp from the NMJ, but not induce significant neuropil puncta, suggesting that the two phenotypes can occur independently of each other and might result from different mechanisms. Loss of nuclear function versus cytoplasmic gain of function remains an open question, with both mechanisms likely contributing to TDP-43 proteinopathy [53,110]. Dlp depletion at the NMJ in the context of TBPH RNAi suggests that this phenotype is the result of nuclear TDP-43 function whereas the neuropil puncta are likely the result of toxic gain of function as they are significantly increased only in the context of TDP-43 proteinopathy and not in the context of endogenous TDP-43 loss of function. Furthermore, the fact that restoring Dlp expression at the NMJ is not sufficient to mitigate TBPH RNAi induced locomotor defects while it improves TDP-43 WT and TDP-43 G298S dependent phenotypes suggests that Dlp levels at the NMJ are critical in the context of TDP-43 proteinopathy and that other factors may be at play in the loss of function caused by TBPH RNAi knock-down. Collectively, our findings suggest a model in which the mislocalization of Dlp found in flies and patient spinal cords results from a combination of translation inhibition and transport defects as previously Fig. 8 Dlp is altered in the context of TDP-43 proteinopathy. Dlp mRNA is enriched in TDP-43 complexes, depleted from ribosomes and sequestered in insoluble/urea complexes. Taken together, these findings and our observations of Dlp protein being reduced at the NMJ and accumulating in puncta within cell bodies and neuropils suggest multiple cellular defects including local translation inhibition at synapses, axonal trafficking and endomembrane trafficking deficits reported for Futsch/MAP1B [15] (see Fig. 8 for model). Lastly, although more experiments are needed to determine the nature and composition of Dlp puncta we found in the VNC neuropil, we speculate that they may reflect a defect in intracellular trafficking and an accumulation of endomembrane compartments that could include endosomes [62], the endoplasmic reticulum [112] or the Golgi [38], which have been linked to TDP-43 proteinopathy. Identifying the subcellular compartment that these puncta associate with will be critical to understanding how Dlp misexpression relates to TDP-43 induced toxicity.
Our findings that GPC6, a Dlp ortholog, is increased in ALS spinal cord motor neurons, insolubilized in patient spinal tissues, and has puncta resembling the Dlp granules observed in the Drosophila VNC, further highlights the power of the fly models to predict pathological changes in disease. Interestingly, the formation of heparan sulfate proteoglycan containing puncta has previously been observed in other neurodegenerative disease including Alzheimer's, Parkinson's and SOD1 mediated ALS (reviewed in [66]) however this has not been previously reported in the context of TDP-43 proteinopathy, nor has a mechanism been proposed. More recently, genetic knockdown of heparan sulfate-modifying enzyme hse-5 in C. elegans mitigated TDP-43 induced deficits in synaptic transmission [55], further suggesting a link between TDP-43 pathology, neurodegeneration and heparan sulfate proteoglycans such as Dlp/GPC6. Lastly, in the last year, genome wide association studies have identified GPC6 as a risk factor multiple sclerosis [80] and for Alzheimer's in African Americans [51] further suggesting a role for GPC6 in maintaining neuronal function.

Conclusions
Here we report several novel translational consequences of TDP-43 proteinopathy including effects on the spliceosome, metabolic pathways and the translational machinery itself. Our combined RIP and TRAP approach can distinguish between direct consequences of TDP-43 proteinopathy and compensatory changes in translation, which can inform future therapeutic strategies. Our identification of the glypican dlp as a target of TDP-43 proteinopathy highlights specific strategies including restoration of synaptic specific protein expression, intracellular transport as well as Wg/Wnt and other Dlp linked signaling pathways in ALS.