Codon optimized TDP-43 gene was synthesized from DNA2.0 (ATUM, Newark, CA, United States of America). The complete sequence of the codon optimized TDP-43 is provided in the Supplementary Figure S3. Drosophila kozak sequence (ATCAAC) was added upstream of the start codon for the TDP43 gene. These constructs were subcloned into the Not1-Xba1 site of the modified fly upstream activation sequence (UAS) expression (pEx-UAS) and glass (pEx-gl) vectors (Exelixis, San Francisco, CA, United States of America). The expression vectors containing the CO-TDP-43 gene were then microinjected into the flies to obtain transgenic flies (BestGene, Chino Hills, CA, United States of America). We generated multiple transgenic fly lines as noted in Supplementary Table S1. We selected UAS-TDP-43_6 line for our subsequent experiments with various GAL4 lines and the gl-CO-TDP-43_4 line for subsequent experiments using the eye as a model. The expression of non-CO-TDP-43 is driven in the fly eye by the glass multimer reporter, GMR-GAL4 on the X-chromosome (Freeman, 1996). All transgenic lines, both codon-optimized and non-codon optimized, express human wild-type TDP-43. Flies expressing human codon wild-type TDP-43 using the UAS promoter were obtained from Dr. Fen-Biao Gao (University of Massachusetts, Worcester, MA, United States of America) (Lu et al., 2009). SevEP-GAL4 driver (expressed in R7 and R8 photoreceptor neurons) was recombined with UAS-TDP-43CO to obtain stable transgenic flies expressing w¹¹¹⁸;SevEP-GAL4,UAS-TDP-43CO/CyO; +. The GMR-GAL4 on the X-chromosome was placed in trans to the gl-TDP-43CO line to generate GMR-GAL4;gl-TDP-43CO/CyO transgenic flies. The following stocks were obtained from Bloomington Drosophila Stock Center (Bloomington, Indiana University, IN, United States of America): w¹¹¹⁸;UAS-LacZ (BDSC 3955), w¹¹¹⁸,GMR-myr-mRFP (BDSC 7121), y¹,w¹¹¹⁸;Sp/CyO; eGFP-ATG5 (recombined from BDSC 59848), y¹,w¹¹¹⁸;UAS-GFP-mCherry-ATG8 (BDSC 37749), GMR-GAL4(X) (eye specific) (BDSC 79572), w¹¹¹⁸;SevEP-GAL4 (R7 and R8 in photoreceptor cells) (BDSC 5793), w¹¹¹⁸,beadex ^MS1096-GAL4 (wing driver) (BDSC 8860), w¹¹¹⁸;Scabrous-GAL4 (sensory organ precursor and wing discs driver) (BDSC 6479), and y¹,w^*; CCAP-GAL4 (driver expressed in CCAP/bursicon neurons in ventral nerve cord and subesophageal ganglion in adult brain) (BDSC 25685). Eq-GAL4 (bristle driver) and y¹,w¹¹¹⁸;Rh1-GAL4/CyO (expressed in R1-R6 photoreceptor cells) were obtained from Dr. Hugo J. Bellen (Baylor College of Medicine, Houston, TX). All crosses were set and flies were maintained at room temperature (22°C) in standard D. melanogaster Jazzmix medium (Applied Scientific, Fisher Scientific, Pittsburgh, PA, United States of America).

Adult retina and imaginal eye discs from third instar larvae were dissected and fixed in 4% paraformaldehyde for 1 h on ice. Adult retina was washed in 0.5% PTX for 3 h to reduce autofluorescence. The tissues were blocked in 0.8% PBS + Triton-X + BSA for 2 h and incubated with primary antibody overnight at 4°C. The tissues were incubated in secondary antibody for 2 h at room temperature, washed in 0.1% PBS + Triton-X and mounted on glass slides with Vectashield (Vector Laboratories, Burlingame, CA, United States). Tissues were stained with the following antibodies: mouse monoclonal anti-TDP-43 antibody (1:500, Abcam, Cambridge, MA, United States), rabbit polyclonal anti-TDP-43 antibody (1:500, Proteintech, Chicago, IL,United States), rat monoclonal anti-Elav (1:20, DSHB, University of Iowa, Iowa City, IA, United States), mouse monoclonal anti-GFP (1:400, Millipore, Billerica, MA, United States), Alexa Fluor 633-conjugated Phalloidin (1:30, Invitrogen, Grand Island, NY, United States), Alexa Fluor 488 conjugated chicken anti-rat (1:400, Invitrogen, Grand Island, NY, United States) Alexa Fluor 568 conjugated goat anti-rabbit (1:400, Invitrogen, Grand Island, NY, United States) and Alexa Fluor 568 conjugated goat anti-mouse (1:400, Invitrogen, Grand Island, NY, United States).

Overexpression of CO-TDP-43 in the fly eye was used to measure total protein levels by immunoblotting. Approximately 50 fly heads were decapitated and homogenized for 1 min in homogenization buffer (10 mM Tris-HCl, 0.8 M NaCl, 1 mM EGTA, pH 8.0% and 10% sucrose) along with 1X PhosSTOP phosphatase and 1X cOmplete protease buffer (Roche Applied Science, Indianapolis, IN, United States). The homogenized samples were centrifuged at 4 °C for 15 min at 18,000 g. The supernatant was collected and equal parts of the supernatant and Laemmle sample loading buffer with β-mercaptoethanol (Bio-Rad, Hercules, CA, United States) was added for each sample. Following a brief pulse centrifugation, samples were loaded on 4%–20% SDS-PAGE gels (Bio-Rad, Hercules, CA, United States) for electrophoresis. For higher molecular weight species detection, the fly heads were homogenized in 1X PBS along with the same protease and phosphatase inhibitors. Non-reducing sample loading buffer (Nupage sample buffer, Life Sciences, Grand Island, NY, United States) was added to the supernatant without β-mercaptoethanol. The blots were blocked in 5% milk, incubated with primary antibodies overnight at 4 °C, washed in 1X TBS + Tween, and incubated with secondary antibody for 1 h at room temperature. The following antibodies were used: Mouse monoclonal anti-TDP-43 antibody (1:1,000, Abcam), Rabbit Polyclonal anti-TDP43 antibody (1:1,000, Proteintech), Rabbit monoclonal anti-Phospho-TDP43 (1:500, Ser 409/410, Proteintech), Mouse monoclonal anti-tubulin antibody (1:1,000, DSHB, University of Iowa), Mouse monoclonal anti-beta Actin antibody (1:5,000, Santa Cruz Biotechnology), secondary anti-mouse IgG-HRP (1:2000, GE Healthcare), and anti-rabbit IgG-HRP (1:5,000, Sigma).

For Lysotracker staining, imaginal eye discs from the third instar larvae were dissected in 1X PBS solution without fixative. The eye discs were then stained with 100 nM LysoTracker Red DND-99 (Invitrogen) for 2 min, followed by a 1-min wash in 1X PBS. The tissues were mounted on a glass slide with a drop of 1X PBS solution; no Vectashield was added. The coverslip was sealed with nail polish and visualized immediately using a confocal microscope. The z-stack images were analyzed using the ImageJ software (Schneider et al., 2012).

ERG was recorded in 1-day old flies using the same methods as previously described (Fabian-Fine et al., 2003; Williamson et al., 2010). Briefly, flies were glued on glass slides using Elmer’s non-toxic glue. Both the reference and recording electrodes were made of glass pipettes filled with 3 M KCl. The light stimulus was computer-controlled using white light-emitting diode system (MC1500; Schott), and was provided in 1-s pulses. The data was recorded using Clampex software (version 10.1; Axon Instruments) and measured and analyzed using Clampfit software (version 10.2; Axon Instruments).

The adult eye, wing and bristle pictures were taken with a Nikon AZ100 M microscope equipped with a Nikon DS-Fi1 digital camera (Nikon Instruments, Melville, NY, United States). Extended depth of focus (EDF) and volumetric images were taken using the Nikon NIS-Elements AR 3.0 software as previously described (Ambegaokar and Jackson, 2011). The scanning electron microscopy (SEM) images were taken using JSM-6510LV SEM (JEOL United States, Peabody, MA, United States). The confocal images were taken with a Zeiss LSM 510 UV META laser scanning confocal microscope using 40X water and 63X oil-immersion high-resolution objectives. These images were analyzed using the LSM Image Browser and NIH ImageJ software (Schneider et al., 2012).

The SEM and light-microscopy images were quantitated by image-J. Briefly, the area of the eye surfaces were measured by NIH image-J software and paired t-test was performed between the Control vs non-CO-TDP-43 vs CO-TDP-43 eyes for statistical analysis. Western Blot data comparing non-CO-TDP-43 and CO-TDP-43 was estimated by paired t-test. The histograms represent mean ± SEM and are analyzed using Graphpad prism software. Quantification of Lysotracker staining was performed using the NIH Image J software (Schneider et al., 2012). The measurements and histograms represent mean ± SEM and plotted using Microsoft Excel and SigmaPlot (version 10.1) software. Statistical analysis was performed using one-way ANOVA with Bonferroni’s correction and paired Student’s t-test with two-tailed distributions of equal variance.

The codon-optimized TDP-43 fly lines are available upon request.

We generated multiple codon-optimized CO-TDP-43 transgenic fly lines to investigate TDP-43 mediated neurodegeneration. We utilized both the yeast GAL4/UAS binary system (Brand and Perrimon, 1993) and a glass (gl) promoter direct fusion construct specifically generated to study TDP-43 mediated effects on the fly retina (Figures 1A, B). In addition, we also used another eye promoter, Sevenless (SevEP-GAL4), that only expresses in a subset of photoreceptor neurons (R7 and R8) and cone cells (Therrien et al., 1999). To highlight the robust effect observed in our CO-TDP-43 lines, we compared the phenotypes to a previously reported human TDP-43 transgenic line, which we denote as non-CO-TDP-43 (Lu et al., 2009; Choksi et al., 2014).

Heterozygous expression of CO-TDP-43 using the gl promoter caused depigmentation, roughness, disruption of polarity, and loss of inter-ommatidial bristles (Figures 1F, J). The CO-TDP-43 expressed using SevEP-GAL4 showed a similar but milder phenotype of the eye (Figures 1E, I). In comparison to the CO-TDP-43 flies, the non-CO-TDP-43 transgenic flies (Figures 1D, H) did not show a robust eye phenotype and appeared to be similar in morphology to the wild-type control flies (Figures 1C, G). Interestingly, the eye phenotype observed with heterozygous gl-CO-TDP-43 flies were age dependent. At 1-day post-eclosion, CO-TDP-43 exhibited a mild phenotype (Figure 1L) that worsened by 10-days (Figure 1F and Supplementary Figure S4). In contrast, flies with two copies of the CO-TDP-43 transgene showed a strong phenotype at 1-day post-eclosion, with apparent necrotic patches or hyperpigmentation (Figure 1M white arrowheads, compared to control in Figure 1K). In addition, the CO-TDP-43 flies with either one or two copies of the TDP-43 transgene showed less eye volume than wild-type control flies at 1-day post-eclosion (Figures 1O, P compared to control in Figure 1N). We were unable to observe these phenotypic differences in flies aged longer than 1-day post-eclosion as flies with two copies of the CO-TDP-43 transgene do not survive a day or two after eclosion. We also found that overexpression of CO-TDP-43 using GMR-GAL4 driver led to pupal lethality at 18°C and 25°C (Supplementary Table S1), with some escapers at 18 °C that showed necrotic patches (Supplementary Figure S1). Taken together, our results showed that CO-TDP-43 transgenic flies have a more robust eye phenotype indicative of neurodegeneration in retinal cells compared to non-CO-TDP-43 transgenic flies.

TDP-43 associated pathology in ALS patients have been linked to significant neuronal loss and early axonal atrophy in sensory nerves (Heads et al., 1991; Mochizuki et al., 2011). For example, Vaughan and colleagues reported that the pathogenic A315T mutation in TDP-43 affects neurite growth and decreased dendritic branching of sensory neurons (Vaughan et al., 2018). In fact, previously reported Drosophila neurodegeneration models showed that overexpression of the neurotoxic ataxin-1 mutant in sensory precursors using the scabrous-GAL4 (sca-GAL4) driver leads to loss of bristles in the adult fly (Tsuda et al., 2005). Similarly, we previously showed that misexpression of fly dVAP33, a gene linked to ALS, using sca-GAL4 leads to loss of notal macrochaetae (Ratnaparkhi et al., 2008). To further investigate the phenotypic effects of CO-TDP-43 on sensory precursor cells of the wing and notum, we used multiple wing and bristle drivers to misexpress TDP-43 protein, including beadex ^MS1096 -GAL4 (bx ^MS1096 -GAL4), sca-GAL4, equate-GAL4 (eq-GAL4), and CCAP-GAL4. We found that non-CO-TDP-43 transgene expressed using bx ^MS1096 -GAL4 led to viable adults with shriveled wings, with some flies having wings that were either necrotic or had areas of hyperpigmentation (Figure 2B). In contrast, CO-TDP-43 flies using the same driver exhibited a more severe phenotype, with pharate adults and very small and severely malformed wings with necrotic or hypermelanized patches (Figure 2C, compared to control in Figure 2A). Interestingly, unlike the previously reported model of ALS, neither non-CO-TDP-43 nor CO-TDP-43 had any effect on macrochaetae (bristles) on the notum when misexpressed using sca-GAL4 driver. Instead, the CO-TDP-43 transgenic flies produced pharate adults with necrotic wings that were unable to expand (Figure 2F, compared to non-CO-TDP-43 in Figure 2E and control in Figure 2D). Since we failed to see an effect of TDP-43 on macrochaetae using sca-GAL4, we used another bristle-specific driver, eq-GAL4, to misexpress non-CO and CO-TDP-43 in the fly notum (Tang and Sun, 2002). While both control and non-CO-TDP-43 flies showed normal macrochaetae formation (Figures 2J, K), CO-TDP-43 showed a dramatic loss of defective notal macrochaetae (Figure 2L). Furthermore, Vanden Broeck and colleagues previously showed that both up and downregulation of fly dTDP-43 cause selective apoptosis in the crustacean cardioactive peptide (CCAP)/bursicon neurons (Vanden Broeck et al., 2013). Loss of CCAP/bursicon neurons have been shown to cause pupal lethality with escapers that show wing expansion defect phenotypes (Park et al., 2003). Upon expression of non-CO-TDP-43 in the CCAP/bursicon neurons using CCAP-GAL4, we observed a similar wing expansion defect in adults (Figure 2H). Misexpression of CO-TDP-43 in CCAP/bursicon neurons resulted in smaller, necrotic and swollen wings compared to control flies (Figure 2I and Figure 2G, respectively). In summary, these results suggest that misexpression of CO-TDP-43 in flies leads to smaller wings with abnormal morphology and macrochaetae irregularities compared to misexpression of non-CO-TDP-43.

The robustness of the external phenotypes observed with CO-TDP-43 prompted us to examine the protein expression levels of the TDP-43 transgene in these flies. We next used multiple gl direct fusion CO-TDP-43 lines to examine TDP-43 expression levels in the fly eye. Compared to the GMR-GAL4 driven non-CO-TDP-43 transgenic flies, the CO-TDP-43 flies showed a 2-fold increase in monomeric total TDP-43 protein in multiple lines (Figures 3A, B and Supplementary Figure S2B, lane 5, 6 and 7). Next, we compared the TDP-43 protein levels between the CO-TDP-43 and non-CO-TDP-43 lines using the SevEP driver. In this case, the CO-TDP-43 protein levels were higher than the non-CO-TDP-43 lines, but this enhancement was not statistically significant (Figures 3A, C). Possibly because TDP-43 is expressed only in a subset of photoreceptor neurons. The line showing the highest increase in protein levels, one of the gl direct fusion CO-TDP-43 lines (Supplementary Figure S2B, lane 5), also demonstrated a robust eye phenotype (Figure 1) and was therefore used in subsequent experiments. Predictably, compared to the gl-CO-TDP43 line, the recombinant line using SevEP-GAL4 to overexpress CO-TDP-43 did not show an increase in total TDP-43 expression, since it is only expressed in a subset of photoreceptor neurons (Supplementary Figure S2B, lane 8). Next, we assessed the levels of TDP-43 phosphorylated at Serine 409/410 residues since phosphorylation of these residues are associated with TDP-43 oligomerization and toxicity (Gao et al., 2018). We found high levels of phospho-TDP43 in our CO-TDP-43 lines compared to non-CO-TDP-43 (Figures 3D, E).

As observed in patients with ALS and FTLD, high-molecular weight toxic species of TDP-43 have been detected in transgenic flies overexpressing TDP-43 containing pathogenic variants (Miguel et al., 2011; Chang and Morton, 2017). For example, we previously reported high-molecular weight species of TDP-43 in flies overexpressing disease-associated TDP-43 Q331K mutations (Choksi et al., 2014). To investigate if we are able to detect these high-molecular weight oligomeric species in our codon optimized lines, we used the gl direct fusion line and the SevEP-GAL4 driven CO-TDP-43 lines. As expected, we detected higher molecular weight species in SDS-PAGE under non-denaturing conditions in both lines tested, with increased levels in the gl driven CO-TDP-43 line that were absent in the non-CO-TDP-43 flies and surprisingly in SevEP-CO-TDP-43 lines. Interestingly, we also observed a 35 kD truncated fragment only in the codon optimized flies (Figure 3F). These bands represent the previously reported caspase cleaved C-terminal fragment that is considered to be the toxic component of TDP-43 aggregates (Liu et al., 2014; Chiang et al., 2016).

The distinctive pathology of TDP-43 mediated neurodegeneration involves its mislocalization to the cytoplasm and loss of normal nuclear expression (Neumann et al., 2006; Lee et al., 2012). Therefore, we further investigated the localization of CO-TDP-43 in neuronal cells. Figures 3G, L represents the schematic of the eye discs and the area imaged from apical and posterior views respectively. When co-stained with Elav and TDP-43, the eye discs showed higher nuclear and cytoplasmic expression of TDP-43 in both CO-TDP-43 lines (gl and SevEP-GAL4 driven) compared to non-CO-TDP-43 flies (Figures 3I, N vs. Figures 3J, O, K, P). In particular, the gl driven CO-TDP-43 flies formed punctate structures in the cytoplasm resembling cytoplasmic aggregates compared to the other transgenics (Figures 3J, O, compared to control Figures 3H, M). Overall, these observations indicate that a higher level of TDP-43 protein has the propensity to form protein aggregates via a gain-of-function mechanism, similar to other neurodegenerative proteins such as Tau, Aβ, and Alpha-synuclein.

Based on severe retinal phenotypes observed with toxic aggregates of TDP-43 protein, we further investigated the internal cellular morphology of the photoreceptor neurons. We utilized another eye-specific driver, Rh1-GAL4, which is expressed in R1–R6 neurons starting in late pupal stage and persisting throughout adulthood (Chyb et al., 1999). Unlike the GMR-GAL4, SevEP-GAL4 or gl direct fusion lines, this driver allowed us to examine adult-onset expression of TDP-43. In 7-day post-eclosion CO-TDP-43 flies (Figure 4B), we observed a degenerative phenotype in the adult retina marked by the loss of rhabdomere structures and vacuolization compared to the control flies (Figure 4A). Comparatively, using the gl direct fusion CO-TDP-43 line, we observed the degenerative phenotype as early as 1-day in post-eclosion flies. The gl-CO-TDP-43 flies (Figure 4D) exhibited an altered morphology of the photoreceptor neurons, which appeared to be flattened and had a disruption in rhabdomere separation compared to control flies (Figure 4C) when visualized in the tangential view of the adult retina. Examination of the longitudinal view of the adult retina showed a marked reduction in thickness and shorter photoreceptor length compared to controls (marked by white lines in Figures 4E, I). In addition, we found that these photoreceptor neurons were accompanied by large vacuolar structures, and co-staining with Elav revealed that TDP-43 was localized both in the nucleus and in the cytoplasm (Figures 4I–L) compared to control flies (Figures 4E–H).

To investigate the physiological functions of these photoreceptor neurons, we used electroretinogram (ERG) recordings to measure the functionality of active photoreceptor neurons by measuring their response to light stimulus (Dolph et al., 2011). In fact, Drosophila neurodegeneration models overexpressing Tau and Alpha-synuclein exhibited degenerative pathology in the fly retina along with neuronal dysfunction, as detected by ERG recording (Chouhan et al., 2016). Using this technique, we investigated whether CO-TDP-43 misexpression affects neuronal functionality compared to wild-type and GMR-RFP controls (Figures 4M–O). The gl-CO-TDP-43 flies demonstrated a reduction in both the amplitude of ERG in “on transient” and evoked depolarization at 1-day post-eclosion (Figures 4P, Q). These effects were not observed with either control. Taken together, these results strongly suggest that CO-TDP-43 misexpression causes structural and functional degenerative phenotypes in the adult retina.

An autophagic dysfunction has been implicated in many neurodegenerative diseases, including ALS (Wong and Cuervo, 2010; Brady et al., 2011; Sasaki, 2011). Therefore, the presence of large vacuolar structures in the adult retina of CO-TDP-43 flies (see Figure 4) led us to investigate if these vacuoles could possibly be a representation of autophagic intermediates. We performed live imaging of larval eye discs using LysoTracker to detect lysosomes and other acidic organelles, such as autophagosomes, that typically increase in number and/or size during the later stages of autophagy. Figure 5A represents the schematic of the eye discs and the area imaged. Upon CO-TDP-43 misexpression, we detected a significantly larger number of acidic punctae compared to control flies (Figure 5C, white arrows, compared to control in Figure 5B; Figure 5D), suggesting increases in autophagosomes due to elevated levels of TDP-43.

To further characterize the large vacuoles, we coexpressed CO-TDP-43 and a tagged autophagy protein, Atg5-GFP, which is responsible for the formation of the autophagosomes. We found that these vacuoles were positive for both Atg5 and TDP-43 (Supplementary Figures S5A–D). During autophagy, autophagosomes merge with lysosomes to become autolysosomes and are acidified to degrade proteinaceous waste materials (Zhang et al., 2013). To determine if the autophagosomes observed were mature and functional autolysosomes, we used the Atg8-mCherry-GFP tandem reporter to assay the relative acidity of the autophagosome/autolysosomes. Atg8-mCherry-GFP is a tandem reporter that detects Atg8, which is localized in autophagic intermediates, and a pH-sensitive GFP that only emits a signal at a neutral pH (Filimonenko et al., 2007). This is a useful tool to help understand whether the large vacuoles observed in our CO-TDP-43 flies were autophagic as well as acidic, which is characteristic of autophagic intermediates. We coexpressed CO-TDP-43 with the tandem reporter and found that the larger punctae were positive for Atg8-mCherry, while only a subset of the relatively smaller punctae were stained with GFP, indicating non-acidic compartments (Supplementary Figures S5E–G). In contrast, the majority of the punctae that were larger in size were only fluorescent for Atg8-mCherry (Supplementary Figures S5E–G), indicating more acidic and mature autolysosomes. These results suggest that misexpression of CO-TDP-43 leads to increased acidic lysosomal vacuoles that are indicative of autophagic dysfunction.