Maintenance of neuronal TDP-43 expression requires axonal lysosome transport

Abstract

TDP-43 mislocalization and pathology occurs across a range of neurodegenerative diseases, but the pathways that modulate TDP-43 in neurons are not well understood. We generated a Halo-TDP-43 knock-in human induced pluripotent stem cell (iPSC) line and performed a genome-wide CRISPR interference FACS-based screen to identify modifiers of TDP-43 levels in neurons. A meta-analysis of our screen and publicly available screens identified both specific hits and pathways present across multiple screens, the latter likely responsible for generic protein level maintenance. We identified BORC, a complex required for anterograde lysosome transport, as a specific modifier of TDP-43 protein, but not mRNA, levels in neurons. BORC loss led to longer half-life of TDP-43 and other proteins, suggesting lysosome location is required for proper protein turnover. As such, lysosome location and function are crucial for maintaining TDP-43 protein levels in neurons.

Keywords: BORC; CRISPRi screen; TDP-43; cell biology; human; iNeuron; lysosome; neuroscience.

Figure 1.. Generation of an endogenously HaloTagged TDP-43 induced pluripotent stem cell (iPSC) line.

(A) Schematic of knock into the N-terminus of TDP-43. Cas12a-crRNA ribonucleoparticle (RNP) and HaloTag donor plasmid containing homology arms to TDP-43 were nucleofected into iPSCs. Halo was knocked into a single TDP-43 allele at the N-terminus of the protein, before exon 2, just after the start codon (ATG). (B) Microscopy validation of Halo-TDP-43 localization to nucleus. Halo in pink, ConA lipid dye in green, and Hoechst in blue. Halo signal is indicated in grayscale in left panels. Scale bar 10 µm. (C) TDP-43 western blot of non-targeting (NT) and TDP-43 KD neurons. Halo-TDP-43 is approximately 33 kDa heavier than untagged TDP-43, which corresponds to the molecular weight of Halo. RPL24 is shown as a loading control. TDP-43 KD decreases levels of both tagged and untagged TDP-43 compared to NT control sgRNA. (D) Quantification of western blot shown in (C) shows untagged (left) and HaloTagged (right) TDP-43 is decreased upon TDP-43 KD (n=6). Interestingly, Halo-TDP-43 is expressed at levels about one half as much as untagged TDP-43.

Figure 1—figure supplement 1.. i11w-hT induced pluripotent stem cells (iPSCs) have a normal karyotype.

(A) Karyotype of Halo-TDP-43 iPSCs shows a normal karyotype (44 chromosomes+XY), demonstrating that CRISPR editing did not cause any chromosomal abnormalities. (B) Left blot shows Cas9, Halo-TDP-43, and RPL23 staining (Anti-HA, Anti-Halo, and Anti-RPL23, respectively), demonstrating TDP-43 is HaloTagged. On the right, the blot was stripped and re-probed with anti-TDP-43, showing both HaloTagged TDP-43 and untagged TDP-43.

Figure 2.. CRISPR interference (CRISPRi) screen identifies modifiers of TDP-43 protein levels in induced pluripotent stem cell (iPSC)-derived neurons.

(A) Schematic of FACS screen. Halo-TDP-43 iPSCs were transduced with the dual guide library and selected. After selection, iPSCs were differentiated to neurons via Dox-inducible NGN2 expression. Neurons were FACS-sorted and populations expressing high Halo-TDP-43 and low Halo-TDP-43 were collected, DNA extracted, and sgRNA libraries sequenced. (B) Rank plot of screen results showing genes whose KD increases Halo-TDP-43 levels (left) and genes whose KD decreases Halo-TDP-43 levels, including TDP-43 itself (right). BORC genes are in blue. Non-targeting guides are indicated in gray and cluster in the middle, linear portion of the graph, demarking genes that do not change Halo-TDP-43 levels. (C) GO analysis of screen hits that reduce Halo-TDP-43 levels includes the BORC complex and many mitochondria-associated terms. The false discovery rate (FDR) is from the calculated permutation p-value of 1000 iterations. (D) GO analysis of screen hits that increase Halo-TDP-43 levels includes many hits related to mRNA processing. The FDR is from the calculated permutation p-value of 1000 iterations.

Figure 2—figure supplement 1.. Induced pluripotent stem cell (iPSC) screen shows different hits than neuron screen.

(A) Rank plot of iPSC screen results showing genes whose KD increases Halo-TDP-43 levels (left) and genes whose KD decreases Halo-TDP-43 levels, including TDP-43 itself (right). BORC genes are indicated in blue. Non-targeting (control) guides indicated in gray. (B) Hypergeometric test (HGT) analysis of screen hits that decrease Halo-TDP-43 levels in neurons also shows many mitochondrial terms, as was seen in the GO analysis. (C) HGT analysis of screen hits that increase Halo-TDP-43 levels in neurons shows many RNA metabolism hits. (D) GO analysis of hits that decrease Halo-TDP-43 levels in iPSCs shows many ribosome and translation-related terms, a striking difference from the hits found in neurons. The false discovery rate (FDR) is from the calculated permutation p-value of 1000 iterations. (E) GO analysis of hits that increase Halo-TDP-43 levels in iPSCs shows DNA synthesis, spliceosome, and DNA repair terms. The FDR is from the calculated permutation p-value of 1000 iterations. (F) HGT analysis of hits that decrease Halo-TDP-43 levels in iPSCs shows translation, ribosome, and nonsense-mediated decay (NMD)-related terms. The FDR is from the calculated permutation p-value of 1000 iterations. (G) HGT analysis of hits that increase Halo-TDP-43 levels in iPSCs shows DNA-related terms. The FDR is from the calculated permutation p-value of 1000 iterations.

Figure 3.. Meta-analysis of CRISPR interference (CRISPRi) screens identifies novel and common genes.

(A) Comparison between CellRox and Halo-TDP-43 meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively. (B) Comparison between Liperfluo and Halo-TDP-43 meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively. Comparison between Tau and Halo-TDP-43 meta-analysis z-scores with novel hits highlighted. BORC genes are highlighted in light blue. Correlation is indicated in blue with R and p-value at top left of graph. (C) Comparison between Tau and Halo-TDP-43 meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively. (D) Comparison between Liperfluo and CellRox meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively. (E) Comparison between CellRox and Tau meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively. (F) Comparison between Tau and Liperfluo meta-analysis z-scores with novel hits highlighted. Correlation is indicated in blue with R and p-value at top left of graph. BORC genes are highlighted in blue, novel hits from Liperfluo, CellRox, and Tau screens as compared to our Halo-TDP-43 screen are indicated in green, orange, and pink, respectively.

Figure 3—figure supplement 1.. Meta-analyses of Halo-TDP-43 and published CRISPR interference (CRISPRi) FACS screens.

(A) GO analysis of genes that negatively modulate readouts for CellRox and Halo-TDP-43 screens shows DNA and signaling related terms. The false discovery rate (FDR) is from the calculated permutation p-value of 1000 iterations. (B) GO analysis of genes that positively modulate readouts for CellRox and Halo-TDP-43 screens shows translation-related terms. The FDR is from the calculated permutation p-value of 1000 iterations. (C) GO analysis of genes that negatively modulate readouts for Liperfluo and Halo-TDP-43 screens shows terms related to ubiquitination, neddylation, and lipids. The FDR is from the calculated permutation p-value of 1000 iterations. (D) GO analysis of genes that positively modulate readouts for Liperfluo and Halo-TDP-43 screens shows adenosine, acetylation, and autophagy-related terms. The FDR is from the calculated permutation p-value of 1000 iterations. (E) GO analysis of genes that negatively modulate readouts for Tau and Halo-TDP-43 screens shows terms related to ubiquitination, neddylation, and lipids. The FDR is from the calculated permutation p-value of 1000 iterations. (F) GO analysis of genes that positively modulate readouts for Tau and Halo-TDP-43 screens shows adenosine, acetylation, and autophagy-related terms. The FDR is from the calculated permutation p-value of 1000 iterations.

Figure 4.. Validation of modulators of TDP-43 levels in induced pluripotent stem cell (iPSC)-derived neurons.

(A) Schematic of hit validation. As a first validation, Halo-TDP-43 iPSCs were transduced with single sgRNAs against hit genes, and Halo levels were analyzed by microscopy, generating a list of Halo-validated hits. Then, some genes were selected from the Halo-validated hits, and untagged iPSCs were transduced with the same virus, and TDP-43 protein levels were assessed via TDP-43 immunofluorescence microscopy, identifying hits not affected by a HaloTag. (B) Representative images of Halo-TDP-43 live-cell imaging with BORCS6 KD. sgRNA plasmids contain a cytoplasmic BFP, enabling identification of cells expressing the sgRNA. For BORC genes, iPSCs were transduced with a lysosome marker (LAMP1-mApple) to ensure functional BORC KD through depletion of neuritic lysosomes. Scale bar represents 20 µm. (C) Quantification of BORC KD Halo-TDP-43 microscopy. All BORC genes tested (S1–S8) showed statistically significant decreases in Halo-TDP-43 levels compared to a non-targeting (NT) guide, indicated by blue dots. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (D–G) Summary graphs of screen, Halo live-cell imaging, and TDP-43 immunofluorescence imaging fold changes for (D) BORC KD. (E) Ubiquitin-associated gene KD. (F) m6A-associated gene KD, and (G) mitochondria gene KDs. Color indicates strength of log₂fold change for each condition, circle size indicates –log₁₀p-value; values below 1.3 are not significant, corresponding to a p-value of 0.05. No circle indicates the gene was not tested in that experiment (i.e. not all genes were tested by immunofluorescence).

Figure 4—figure supplement 1.. Validation of neuron screen results shows many genes alter TDP-43 levels.

(A) Quantification of miscellaneous KD Halo-TDP-43 microscopy. OGFOD1 KD increases Halo-TDP-43 levels while UCHL5 KD, ZYFVE26 KD, UQCRQ KD, and PITRM1 KD decrease Halo-TDP-43 levels compared to a non-targeting (NT) guide, indicated by dark blue dots. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (B) Quantification of ubiquitin-associated gene KD Halo-TDP-43 microscopy. MAEA, NAE1, UBA3, UBC, CUL1, and SKP2 KD all increase Halo-TDP-43 levels compared to an NT guide, indicated by orange dots. NEDD8 KD decreases Halo-TDP-43 levels due to a strong survival phenotype. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (C) Quantification of m6A-associated gene KD Halo-TDP-43 microscopy. CNOT3, METTL14, MTLL4, METTL21A, METTL3, KIAA1429, YTHDF2, and ZC3H13 KD all increase Halo-TDP-43 levels compared to an NT guide, indicated by magenta dots. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (D) Quantification of mitochondria-associated gene KD Halo-TDP-43 microscopy. MFN2 KD increases Halo-TDP-43 levels while DTYMK and PMPCB KD decrease Halo-TDP-43 levels compared to an NT guide, indicated by dark blue dots. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (E) Quantification of TDP-43 immunofluorescence (IF) microscopy. TDP-43 KD significantly decreases TDP-43 levels by IF. ZFYVE26, UQCRQ, MAEA, NAE1, UBC, SKP2, CNOT3, METTL14, METTL4, METTL3, KIAA1429, YTHDF1, and ZC3H13 KD all decrease TDP-43 IF levels (although not all decrease Halo-TDP-43 levels), while UBA3 and METTL21A KD increase TDP-43 IF levels, indicated by lime dots. N=12 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (F) Quantification of MFN2 KD TDP-43 immunofluorescence microscopy. MFN2 KD does not significantly change untagged TDP-43 levels compared to an NT guide, indicated by teal dots, while TDP-43 KD does significantly decrease TDP-43 IF levels. N=6 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph.

Figure 5.. Endogenous validation of TDP-43 level modifiers in induced pluripotent stem cell (iPSC)-derived neurons.

(A) Quantification of BORC KD TDP-43 immunofluorescence microscopy. Both BORC genes tested (S2 and S6) show significant decreases in untagged TDP-43 levels compared to a non-targeting (NT) guide, indicated by blue dots. N=4 wells per genotype, 9 images per well (small gray dots). Significant p-values indicated on graph. (B) Representative images of TDP-43 immunofluorescence imaging with BORCS6 KD. sgRNA plasmids contain a cytoplasmic BFP, enabling identification of cells expressing the sgRNA. Scale bar represents 20 µm. (C) Schematic of pLentiCRISPR plasmid. pLentiCRISPR enables lentiviral delivery of a plasmid expressing an sgRNA of interest under a U6 promoter and Cas9 under an Ef1alpha promoter. This enables knockout of a gene of interest targeted by an sgRNA. (D) Representative images of BORCS7 KO TDP-43 IF with lysosomes stained with anti-H4A3 antibody. Scale bar is 20 µm. (E) Quantification of TDP-43 immunofluorescence on BORCS7 knockout neurons. Both BORCS7 guides significantly reduced the amount of TDP-43 IF signal compared to an NT guide, light blue dots. N=10 wells, 9 images per well (light gray dots). Significant p-values indicated on graph.

Figure 6.. TDP-43 protein levels require active and neuritically localized lysosomes.

(A) Drugs targeting different aspects of lysosome function decrease Halo-TDP-43 levels. All drugs tested were significant compared to a DMSO treatment, indicated by red dots. N=12 wells, 9 images per well (small gray dots), mean of each well is indicated by large dot. (B) Representative images of drug treatments that decrease Halo-TDP-43 levels. Scale bar is 20 µm. (C) Representative 30 s kymographs of Halo-TDP-43, lysosomes (LAMP1-mApple), mitochondria (mito-mEmerald), showing co-transport of Halo-TDP-43 with organelles. Green arrowheads indicate co-localization with static mitochondria; blue arrows indicate co-localization with motile lysosomes. Scale bar is 10 µm. (D) Quantification of Halo-TDP-43 signal with organelles. Motile Halo-TDP-43 in neurites is primarily co-transported with lysosomes or lysosomes and mitochondria together. Stationary Halo-TDP-43 is more likely to be associated with mitochondria. (E) Upon BORCS7 KD, TDP-43 is less mobile compared to a non-targeting (NT) guide, likely due to restriction of lysosomes to the soma. N=4 wells (mean per well indicated by large dot), 8 images per well (small dots).

Figure 6—figure supplement 1.. Lysosome inhibitors alter Halo-TDP-43 levels.

(A) Additional drug treatments from experiment shown in Figure 5B that do not change Halo-TDP-43 levels. An untagged control line (‘no Halo tag’) shows significantly less Halo signal than the DMSO control (repeated from Figure 5B). (B) Replication of drug treatment experiment shows the same trend as previous experiment. (C) Apillimod does not alter levels of Halo-TDP-43. (D) Quantification of Halo-TDP-43-organelle association after BORCS7 KD. Halo-TDP-43 is significantly reduced on lysosomes, with a small increase of Halo-TDP-43 on mitochondria. More Halo-TDP-43 is transported not on either lysosomes or mitochondria in BORCS7 KD neurons.

Figure 7.. BORC KD does not affect TDP-43 RNA levels.

(A) qPCR of non-targeting (NT), BORCS6 KD, and TDP-43 KD neurons showing decreased TDP-43 mRNA levels in TDP-43 KD, but not BORCS6 KD. Normalized to PPIA. NT n=3, BORCS6 KD n=4, and TDP-43 KD n=4. (B) Normalized transcript counts of NT and BORCS6 KD neurons showing about half as much BORCS6 mRNA in BORCS6 KD neurons. NT n=4, BORCS6 KD n=3. (C) Normalized transcript counts of NT and BORCS6 KD neurons showing no change in TARDBP (TDP-43) mRNA levels in BORCS6 KD neurons. NT n=4, BORCS6 KD n=3. (D) Volcano plot comparing BORCS6 KD with NT neurons. 98 genes increase expression upon BORCS6 KD (right of graph), while 139 genes decrease expression (left of graph). NT n=4, BORCS6 KD n=3. (E) GO Biological Process (BP) enrichment analysis from Enrichr of genes significantly altered in BORCS6 KD neurons compared to NT neurons shows several neuron and axon-related terms.

Figure 7—figure supplement 1.. TDP-43 mRNA levels are unchanged after BORCS6 KD.

(A) qPCR of non-targeting (NT), BORCS6 KD, and TDP-43 KD neurons showing decreased TDP-43 mRNA levels in TDP-43 KD and slight, but significant increase in TDP-43 mRNA in BORCS6 KD. Normalized to PGK1. NT n=3, BORCS6 KD n=4, and TDP-43 KD n=4. (B) Principal component analysis plot of RNA sequencing data shows most of the variance between the samples (90%) can be explained by a single principal component and that the samples separate based on knockdown status. (C) REPAC results comparing BORCS6 KD to NT show changes in polyadenylation across the transcriptome, with increased usage of an alternative polyA site compared to the reference site on the right of the volcano, and decreased use of the polyA site compared to the reference site on the left side. TDP-43 (TARDBP) polyA is not changed in BORC KD neurons. (D) For alternative polyA sites that result in a lengthening event (right side of volcano in C), GO Biological Process (BP) shows enrichment in pathways including splicing, intracellular transport, and posttranslational modifications. (E) For alternative polyA sites that result in a shortening event (left side of volcano in C), GO BP shows enrichment in terms related to catabolism, metabolism, and biosynthesis.

Figure 8.. BORC KD alters the abundance and half-lives of TDP-43 and other proteins.

(A) TDP-43 protein has decreased total abundance in day 7 BORCS6 neurons. (B) Volcano plot of total protein abundance comparing BORCS6 KD to a non-targeting guide. Some proteins (purple) have increased abundance, while a larger number (teal) have decreased abundance. Nonsignificant or low log₂(fold change) proteins are indicated in gray. Vertical lines at log₂(fold change) of 1 and –1, horizontal line at p-value = 0.05. (C) TDP-43 has longer half-life in day 7 BORCS6 KD neurons. (D) Volcano plot of half-life changes in BORCS6 KD neurons as compared to neurons expressing a non-targeting guide. Nonsignificant or low log₂(fold change) protein turnovers are indicated in gray. Horizontal line at p=0.05, vertical lines at log₂(fold change) of 0.5 or –0.5. (E) Density plot of global protein turnover shows longer protein half-lives upon BORC KD. (F) Gene ontology enrichment analysis using proteins with longer half-lives in BORCS6 KD neurons shows KEGG pathways related to protein export, proteasome, and metabolism. n=6.

Figure 8—figure supplement 1.. Protein turnover is increased at d15 in BORC KD neurons.

(A) TDP-43 translation is not altered upon BORCS7 KD, as measured by fraction of RNA puncta with SunTag puncta by microscopy. N=4 wells, 8 images quantified per well. (B) Density plot of total protein half-life shows longer protein half-lives in BORCS6 KD neurons compared to non-targeting (NT) neurons at d15. (C) Overlay of protein half-life density plots at d7 (red) and d15 (teal) shows both days have longer protein half-lives in BORCS6 KD neurons. (D) Volcano plot of total protein abundance comparing BORCS6 KD to an NT guide at d15. Some proteins (purple) have increased abundance, while a larger number (teal) have decreased abundance. Nonsignificant or low log₂(fold change) proteins are indicated in gray. Vertical lines at log₂(fold change) of 1 and –1, horizontal line at p-value = 0.05. (E) Volcano plot of total protein half-life at d15 shows many proteins have longer half-lives (log₂(fold change)>0) in BORCS6 KD neurons as compared to neurons expressing an NT guide. Nonsignificant or low log₂(fold change) protein turnovers are indicated in gray. Horizontal line at p=0.05, vertical lines at log₂(fold change) of 0.5 or –0.5. (F) Gene ontology enrichment analysis using proteins with longer half-lives in BORCS6 KD neurons shows REACTOME pathways related to ribosomes, translation, and mRNA binding.

Author response image 1.