Targeting dysregulated RNA splicing in neurodegenerative diseases

National Institute Of Neurological Disorders And Stroke

Investigators

Abstract

Aim 1: Define cell-type specific splicing alterations associated with disruption of RNA binding proteins in isogenic iPSC-derived cells In this aim we proposed to use cell-type specific iPSC differentiation and long-read RNA sequencing (MAS-ISO- Seq) to identify splicing alterations associated with loss or mutation of RNA binding proteins involved in ALS/FTD. We have engineered forty-five iPSC lines with FTD/ALS-associated mutations including in TARDBP, FUS, MATR3, HNRNPA1, TIA1 and ANXA11, and revertant lines as control, to express cell type-specific transcription factors (NGN2 or hNIL). These lines were obtained from the iPSC Neurodegenerative Disease Initiative (iNDI) and further engineered using modified piggybac transposase/transposon delivery. We have also optimized piggybac constructs and engineered KOLF2.1J lines to express a puromycin resistance marker and tet-inducible transcription factors for differentiation into astrocytes (NFIA.1-SOX9), inhibitory neurons (ASCL1-DLX2), skeletal muscles (MYOD1-shOCT4), glutamatergic neurons with optimized NGN2 (oNGN2), motor neurons (hNIL) or sensory neurons (BRN3). An iPSC line expressing CLYBL for differentiation into microglia was obtained from iNDI. Each of these lines were further engineered to express dCas9. We generated a new vector containing Zim3 dCas9 with 2 mycNLS (to improve its nuclear localization in iPSC models) and a Halotag with blasticidin resistance. Integration with the same lines of constructs expressing cell-type specific transcription factors and dCas9 were verified using BFP and Halotag, respectively. Finally, we cloned new vectors for expression of dual guide RNAs with a Hygromycin resistance and a SNAP- Tag. Guide RNAs were selected and vectors generated to target TDP-43, FUS, hnRNPA1, hnRNPA2B1, hnRNPA3, MATR3, ANXA11, and ATXN2. Conditions were optimized and robust knock-down for these different RNA binding proteins was validated using qRT-PCR, western blot and immunofluorescence in glutamatergic neurons at different time points of differentiation. A large-scale experiment was undertaken to generate samples from glutamatergic neurons with CRISPRi-mediated depletion of each RBP and non-targeting gRNAs as controls (n=6 replicates per condition; 10 conditions; samples for RNA extraction and for immunofluorescence). The RNA samples will be submitted to the Broad sequencing core for MAS-ISO-seq in the coming weeks. Parallel efforts have been undertaken to optimize differentiation protocols and CRISPRi in motor neurons, astrocytes and microglia. We will continue validating the efficiency of differentiation and establish conditions for CRISPRi downregulation of RBPs in different cell types. While we are systematically downregulating different FTD/ALS-associated RBP, we have leveraged already available RNA samples with TDP-43 downregulation in NGN2 neurons to develop an analysis pipeline from MAS-ISO-seq and compare with short read sequencing from the same samples. In collaboration with Drs. Haas and Al'Khafaji at the Broad Institute, we generated long-read sequencing via PacBio MAS-Iso-Seq (bulk Kinnex) from NGN2-neurons with CRISPRi against TDP-43 or non-targeting gRNAs. Approximately 6 million, mostly multi-exonic, reads per samples were obtained with a median size of 1.2 kb. A new workflow was developed for isoform discovery and quantification of differentially expressed features (Haas et al. in preparation). Differentially expressed isoforms were identified with both known and novel cryptic exons observed in TDP-43 targets such as STMN2, G3BP1, FEZ1, ETV5 and RANBP1. Aim 2: Identify splicing profiles in postmortem tissues from patients with distinct forms of ALS and FTD In this aim, we proposed to perform FACS-based isolation of neurons, astrocytes, and microglia dissociated from FTD/ALS patient post-mortem cortex with TDP-43 or FUS mislocalization, and to characterize mis-splicing of RBP-depleted cells using long read sequencing (MAS-ISO-seq). Since our submission, the sequencing technology underlying MAS-ISO-seq (originally developed at the Broad) advanced with the release of the PacBio Revio platform and refinement of the commercial MAS-ISO-seq kits. The combined result is that MAS-ISO-seq PHS 398 Page 1 1RM1 NS133601-01 Lagier-Tourenne, Blainey and Ward, multi-PIs now enables quite cost-effective long-read sequencing of single cells obtained using traditional 10x library prep kits. One can use the transcriptomic signature of individual cells as surrogates of the cell types we proposed to study (neurons, astrocytes, microglia), as well as numerous subtypes of each of these and additional cell types that we didn’t originally propose, all with the benefit of long read driven discovery and identification of splice isoforms. Therefore, use of single cell MAS-ISO-seq has the potential to be superior to FACS-based methods. We performed a pilot study to see if replacing FACS-based sorting with MAS-ISO-seq of single cells or single nuclei from post-mortem brain would enable identification of cell-specific cryptic splicing. We obtained 10x single cell libraries from two rapid-autopsy ALS cases and one control from Dr. Christine Vande Velde at the University of Montreal. Sequencing of the rapid autopsy single cells libraries revealed that most of the cells that survived dissociation were glia, with few neurons that were sequenced, likely an artifact of the dissociation procedure. Nonetheless, approximately 50-120 million reads from ~4,000 cells (mostly microglia) were obtained per samples. Median read length was ~600 bp with mostly multi-exonic reads. We identified cryptic exons that were nominated from our iPSC studies in glia from ALS cases, which appeared to be present in more cells than in the control case. Indeed, while the neuronal STMN2 cryptic event was not detected in this dataset, cryptic exons were found to be enriched in glial cells including in RANBP1, FEZ1 and DNAJB6. In addition, scRNAseq libraries were successfully generated from nuclei from 8 healthy controls and 12 FTD/ALS cases [4 sporadic FTD patients with TDP-43 pathology (FTLD-TDP), 4 FTD patients with repeat expansions in the C9ORF72 gene (c9FTLD-TDP) and 4 FTD patients mutations in the progranulin gene (GRN-FTLD-TDP)] obtained from the Mayo clinic brain bank in collaboration with Dr. Len Petrucelli. These postmortem tissues (obtained from both males and females) were selected based on extensive pathological characterization of their TDP-43 burden. This approach yielded a robust representation of nuclei from different cell types including neurons, astrocytes and oligodendrocytes and identified cryptic splicing in targets such as STMN2 (in neurons), ETV5, FEZ1 and RANBP1 (both in neurons and glia). However, the single nuclei Kinnex approach captured mostly intronic reads (from pre-mRNAs) with ~100-fold fewer multi-exonic than the bulk and single cell kinnex approach. We will continue analyzing these results and integrating with data generated from iPSC-derived cells (Aim 1). Our goal is to nominate additional cell type specific splicing events for genomic screens in Aim 3. Finally, in collaboration with industry colleagues at insitro, we performed a pilot study to test the efficiency of a novel probe-based scRNAseq technology (10x FLEX) that enables direct sequencing of mRNAs in fixed single cells or nuclei. This technology does not suffer from 5’ or 3’ bias, has higher read depths per cell, and does not require long read sequencing. We designed a set of custom probes targeting cryptic exons identified in our multi- CNS cell iPSC TDP-43 knockdown studies, and with Insitro validated that they could detect TDP-43 induced cryptic exons in iNeurons. We then applied this panel of cryptic exon probes, plus the standard set of 10x FLEX probes for transcriptional profiling, to single nuclei isolated from 8 control and 8 FTLD-TDP postmortem brain (from Mayo Brain Bank). Sequencing of subsequent libraries was successful, and we are currently in the process of analyzing these data and comparing to the above alternative single cell sequencing approaches. Aim 3: Apply FACS-based and optical pooled CRISPR screens to identify cell-type specific genetic factors that correct selected splicing alterations associated with neurodegeneration Aim 3a of our proposal involves conducting FACS-based pooled CRISPRi screens using iPSC lines equipped with Zim3-dCas9 and dox-inducible TF differentiation cassettes. Twelve screens in total across three CNS cell types are planned. The first screen aims to identify genetic modifiers that affect STMN2 CE expression in TDP- 43 depleted neurons and other cell types, focusing on increasing or rescuing CE expression. We have now fully optimized conditions for performing FACS-based genome-wide HCR-FISH screens in iNeurons. Substantial alterations to previous FACS-based iNeuron screens were required, including altering fixation conditions from standard PFA to a crosslinker that is reversible in reducing conditions (DSP) plus methanol, due to low yields of reads from genomic integrated sgRNAs following flow sorting after PFA crosslinking. These DSP fixation changes required further alterations to the HCR FISH protocol, which typically uses reducing agents in key steps, to prevent premature crosslinking that interfered with successful flow sorting. We included a second DSP fixation step immediately following the HCR-FISH portion of the protocol, thereby enabling us to successfully sort iNeurons without clumping. These conditions were used to perform a genome-wide CRISPRi screen with PHS 398 Page 2 1RM1 NS133601-01 Lagier-Tourenne, Blainey and Ward, multi-PIs technical duplicates in d7 iNeurons. Top gene knockdown hits that increased the STMN2 CE/total STMN2 ratio in iNeurons included TDP-43, as expected, as well as STMN2. We also identified sgRNAs targeting the RANBP1 gene as a top hit, which we had previously identified in a tagged STMN2-mScarlet iNeuron CRISPRi screen and have further validated using orthogonal knockdown methods and qRT-PCR. Follow up studies using CRISPR cutting nucleases revealed that it is likely that RANBP1 mRNA was not the biological driver of STMN2 loss. Interestingly, a snoRNA is embedded within the RANBP1 transcript. We are currently testing whether this snoRNA is driving the reduction in STMN2 expression. Numerous other hits were observed in interesting biological pathways, such as those involved in splicing and methylation. One caveat to these data that we discovered during interpretation of our results was that we used a ratio of STMN2 CE/total RNA. We selected this ratio as our primary sorting readout based on its optimal performance in separating TDP-43 KD iNeurons from control iNeurons. However, we found that a number of the hits in our screen were due to reductions in total STMN2 RNA, rather than alterations in STMN2 CE. Therefore, we decided to perform back-to-back screens of STMN2 CE and STMN2 total RNA in all future screens, to more specifically identify modulators of cryptic splicing vs total STMN2 expression. In addition, we are following up on the top 1000 hits from this previous screen using an arrayed CRISPR cutting library from Adriano Aguzzi’s lab, followed by multiplex immunostaining against a panel of TDP-43 relevant proteins (see below OPS phenotyping pipeline). This will enable us to identify hits that specifically induce TDP-43 loss of function, mislocalization, or STMN2 expression, in addition to clustering hits into convergent pathways based on image-based profiling. In addition, we proposed to perform CRISPR-based screens in TDP-43 depleted iNeurons, to identify modulators that may further exacerbate or rescue cryptic splicing. We cloned a piggybac based vector that constitutively co- expressed a shRNA targeting TDP-43 in the 3’ UTR of CAG-expressed miRFP670. We transduced iPSCs with this vector, performed serial dilutions, and picked iPSC clones with varied levels of miRFP670 expression, followed by HCR-FISH to identify iPSC clones with partial loss of TDP-43 expression. As expected, we found that STMN2 CE expression was inversely correlated with miRFP670 expression. iPSCs with the highest expression of miRFP (and lowest expression of TDP-43, per STMN2 CE surrogate measurements) also displayed poor growth kinetics. We thus selected a lead iPSC clone for further experiments that maintained good growth kinetics but also displayed partial loss of TDP-43 and modest increased expression of STMN2 CE. Unfortunately, following expansion of several of these clones with TDP-43 shRNAs to the large number of cells required for screens, we found that the TDP-43 knockdown was lost. This is likely due to selective pressure against TDP-43 loss during cell expansion (TDP-43 is required for cell division through unclear pathways). Thus, we re-designed our piggybac vector such that the polII/shRNA transgene was locked behind a loxP-stop cassette. Treatment of iPSCs with cre following expansion results in inducible TDP-43 knockdown, thus enabling us to expand iPSCs pre-differentiation to their required levels without loss of TDP-43 knockdown activity through negative selection pressure. We are currently in the process of conducting genome-wide CRISPRi and CRISPRa screens in iNeurons differentiated from iPSCs expressing this new transgene cassette. In Aim 3b we proposed to apply optical pooled screening (OPS) to identify modulators of RNA binding proteins’ function. Optical pooled screens have not yet been applied to iPSC-derived neurons and major efforts were developed from our three teams to optimize each step of the protocol for screens in neurons. Plating and culture conditions to allow segmentation for image-based phenotyping and genotyping of individual neuron were optimized. Most importantly, we adopted a new sgRNA delivery vector (CROPseq-Multi) developed in the Blainey lab (Walton et al. BioRxiv) which we have shown to increase in situ sequencing spot detection in iPSC neurons. We adopted the new CROPseq-Multi vector for dual-targeting CRISPRi knockdown of target genes in NGN2 neurons. Combining the CROPseq-Multi vector with “Zombie” in vitro transcription-based pre-amplification of genomic guide/barcode sequences using a T7 pre-amplification step, we observed a dramatic increase in signal brightness, as well as restriction of the signal to the nucleus, which simplifies the assignment of sgRNA to cells in high-density cultures such as iPSC-derived neurons and other CNS cell types. With these technological advances, as well as major improvement in the speed of image acquisition, we have performed the first OPS screen that combine HCR-FISH and immunostaining. This approach allows us to determine in individual cells the cellular localization and level of expression of TDP-43 using immunostaining, and the function of TDP-43 by co-localization with HCR-FISH with probes for full length and truncated STMN2 PHS 398 Page 3 1RM1 NS133601-01 Lagier-Tourenne, Blainey and Ward, multi-PIs transcripts. We demonstrated that neurons with TDP-43 knockdown exhibit a marked decrease in full-length STMN2 transcript and a corresponding increase in mis-spliced STMN2, and that in situ sequencing of guide RNAs for OPS remains efficient after subjecting the sample to the multidimensional phenotyping protocol. We conducted a pilot screen targeting 330 genes using a custom library with 3 dual gRNAs constructs per gene, cloned in the newly developed CROPseq-Multi vector. using in vitro transcription of sgRNA barcodes under the T7 promoter prior to sequencing-by-synthesis, we achieved spot detection by in situ sequencing in >90% neurons. The analysis of hits from the screen validated that TDP-43 knockdown lowers nuclear TDP-43 levels and correspondingly increases STMN2 cryptic exon. Additional hits included KPNB1, CSE1L and NUP98, all involved in nucleocytoplasmic shuttling of TDP-43 itself, indicating that our platform can identify biologically relevant genes that modulate RBP cellular localization and splicing regulation. To capitalize on the robustness of the approach, we are currently generating libraries for genome-wide CRISPRi screens in neurons. Towards this, we are constructing a genome-wide dual-sgRNA library consisting of two constructs per gene that target the canonical transcription start site (TSS) of protein-coding genes. To account for cell type–specific alternative TSS usage, we are also generating a spike-in library of dual-sgRNA constructs that target the neuron-specific alternative TSSs. In parallel, we further characterized the impact of TDP-43 loss of function on organelle morphology in iNeurons. In previous studies in TDP-43 KO HeLa cells, it was reported that TDP-43 loss caused dramatic alterations in organelle morphology, in particular altering golgi morphology. Indeed, we found that iNeurons depleted with TDP-43 show severe golgi fragmentation. We further optimized a panel of antibodies against additional organelle markers for IF in iNeurons, including antibodies for mitochondria, early endosomes, lysosomes, peroxisomes, nuclear lamina, stress granules, TDP-43 itself, and nucleoli. Furthermore, we developed new antibodies that could robustly detect a cryptically-spliced protein specific to TDP-43 loss of function via immunostaining of TDP- 43 depleted iNeurons, HDGFL2-CE, which was recently discovered by the Ward lab. We optimized a multiplexed immunostaining protocol, IBEX, to allow us to stain 15 different antibodies across 5 serial rounds of imaging during the phenotyping stage of iNeurons in OPS. This antibody phenotyping panel will be used in a parallel OPS CRISPR-cutting screen of shRNA TDP-43-depleted iNeurons, to parallel the FACS-based screen. We anticipate that gene KOs that rescue TDP-43 LOF will additionally correct organelle dysmorphology and cryptic splicing of HDGFL2, and that the inclusion of a deep panel of antibody markers in the phenotyping stage of the screen will enable clustering of hits to identify pathways that rescue TDP-43 loss of function. Aim 4: Determine the functional impact of identified modulators for their capacity in rescuing splicing profiles and other FTD/ALS-associated phenotypes in iPSC-derived neurons This activity will be initiated in subsequent years.