Whole genome CRISPR screens identify LRRK2-regulated endocytosis as a major mechanism for extracellular tau uptake by human neurons

Pathological protein aggregation in Alzheimer’s disease and other dementias is proposed to spread through the nervous system by a process of intercellular transfer of pathogenic forms of tau protein. Defining the cellular mechanisms of tau entry to human neurons is essential for understanding dementia pathogenesis and the rational design of disease-modifying therapeutics. Using whole genome CRISPR knockout screens in human iPSC-derived excitatory neurons, the primary cell type affected in these diseases, we identified genes and pathways required specifically for uptake of monomeric and aggregated tau. Monomeric and aggregated tau are both taken up by human neurons by receptor-mediated endocytosis, with the low-density lipoprotein LRP1 a significant surface receptor for both forms of tau. Perturbations of the endolysosome and autophagy systems at many levels, and specifically endosome sorting and receptor recycling, greatly reduced tau uptake. Of particular therapeutic interest is that loss of function of the endocytosis and autophagy regulator LRRK2, as well as acute inhibition of its kinase activity, reduced neuronal uptake of monomeric and aggregated tau. Kinase-activating mutations in LRRK2 are a cause of Parkinson’s disease accompanied by neuronal tau aggregation, suggesting that LRRK2 mediates tau spreading in vivo and that LRRK2 inhibition has the potential to inhibit interneuronal spread of tau pathology, slowing disease progression. Overall, pathways for tau entry share significant similarity with those required for virus entry by receptor-mediated endocytosis, suggesting that tau spreading is a quasi-infectious process.


INTRODUCTION
Dementias that involve the microtubule-associated protein tau (tau, MAPT), such as Alzheimer's disease, frontotemporal dementia and progressive supranuclear palsy, are referred to collectively as tauopathies 1 . Characterised by the formation of intracellular neurofibrillary tangles, or aggregates of tau protein, these diseases also display stereotyped spatiotemporal progressions through the central nervous system, including the cerebral cortex 2,3 . Different forms of these diseases first present with pathology in defined brain regions, before progressing through the brain 3 . The spatial progression of tau aggregation mirrors the connectivity of the central nervous system 4 , and the involvement of different brain regions is reflected in neurological symptoms. A current working model for tauopathy progression is that pathogenic forms of tau, proposed to be oligomers or aggregates, are released from neurons containing protein aggregates and then taken up by synapticallyconnected neurons, in which they then seed further tau aggregation and neuronal dysfunction 2 .
There is considerable interest in identifying the molecular agents that mediate intercellular disease spreading and the biological mechanisms involved, including cell surface receptors.
In addition to understanding disease pathogenesis, such insights are essential to the rational design of therapies to slow disease progression 5 . While the identity of the pathogenic tau species responsible for disease propagation remain unknown, progress has been made in understanding the mechanisms of cellular uptake of tau 2 . Cell surface heparan sulfate proteoglycans have been found to be necessary for cellular uptake of tau aggregates 6 , and we and others have found that human neurons efficiently take up both monomeric and aggregated tau by overlapping but distinct mechanisms, consistent with dynamin-dependent receptor-mediated endocytosis 7,8 . Recent work has identified the low density lipoprotein receptor LRP1 as a major neuronal receptor for tau uptake via receptor-mediated endocytosis 9 .
To define the cell biology of neuronal uptake of extracellular tau, we used whole-genome CRISPR knockout screens in human iPSC-derived neurons to identify genes and pathways required for neuronal uptake of tau protein. It is currently not known which forms of tau mediate pathogenic spreading of tauopathy between neurons 10 , although oligomers or aggregates of either full length tau or fragments that include the microtubule-binding region are considered strong candidates 11 . Therefore, we designed whole genome CRISPR knockout screens 12 for genes required for neuronal uptake of full-length monomer and aggregates of tau 7 , using human iPSC-derived excitatory neurons 13 , the primary cell type affected by tau aggregation into neurofibrillary tangles in vivo 1 .

FACS-based CRISPR knockout screens in human neurons to identify genes and pathways required for uptake of extracellular tau
We have previously found that aggregated and monomeric tau are taken up by human neurons by overlapping, but distinct pathways 7 . To identify and compare the cellular pathways by which monomeric and aggregated tau enter neurons, FACS-based assays were optimised to measure uptake of both forms of tau by human iPSC-derived cortical neurons ( Figure 1; Supplementary Figure 1). To focus the screens on genes required specifically for tau uptake, and not general mechanisms of receptor-mediated endocytosis, assays were designed to measure uptake of transferrin and tau by the same neurons ( Figure   1). Importantly, transferrin endocytosis does not alter tau uptake at a range of concentrations For CRISPR knockout screens, cerebral cortex progenitor cells were generated from KOLF2 human iPSCs, constitutively expressing Cas9 from the AAVS1 locus, using our previously described methods (Supplementary Figure 1) 13,14 . A lentivirus library composed of 100,090 gRNAs, targeting 18,025 genes 12 , was introduced by infecting cortical progenitor cells at an MOI of 0.3, and the progenitor cells subsequently differentiated to cortical excitatory neurons for 30 days (Figure 1).
To identify lentivirus-infected, guide RNA-expressing neurons that failed to take up tau but remained competent to endocytose transferrin, neurons were exposed to extracellular tau (labelled with Dylight-488) and transferrin (conjugated with Alexa-633) for either 4 or 5 hours, before being dissociated and fixed. Lentivirus-infected, gRNA-expressing neurons were gated and the populations of tau+ /transferrin+ and tau-/transferrin+ cells collected. Libraries of gRNAs were prepared and sequenced from each population, and replicate screens analysed to identify gRNAs enriched in neurons that failed to take up tau using the MAGeCK algorithm, reflecting genes whose loss of function resulted in reduced tau uptake 15 .

Receptors and pathways required for neuronal uptake of monomeric tau
Duplicate screens for monomeric tau uptake were analysed using MAGeCK 15 to identify genes whose loss of function results in reduced neuronal tau uptake (Figure 2). Applying a significance cutoff of p<0.01 identified 214 genes required for tau uptake (Figure 2). Notably the low density lipoprotein receptor LRP1, which has been shown to act as a tau receptor 9 , was one of the two highest ranked genes identified in the monomeric tau uptake screens, confirming the validity of the screens. The most highly ranked gene identified was LRRK2, a large multifunctional protein that regulates diverse intracellular vesicle trafficking processes, mutations in which are a cause of autosomal dominant Parkinson's disease 16 .
In addition to LRRK2, there was a notable number of genes encoding regulators of endocytosis required for monomeric tau uptake, including several AP2 subunits (AP2M1, AP2S1), dynamin-2 (DNM2), clathrin heavy chain (CLTC), RAB7A, HGS and PI3 kinase subunits C3 and R4. Several genes involved in endosomal sorting, such as SNX16, and specifically recycling of the LRP1 receptor were also required for monomeric tau uptake, including SNX17 17 and the LDL receptor chaperone MESD/MESDC2 18 . In addition, genes involved in lysosome and autophagosome biogenesis and acidification were prominent, including the vATPase subunit ATP6V1D, the vATPase assembly protein VMA21 19 , and the autophagy regulating proteins UVRAG and ATG4A 20 .
In addition to individual genes, gene ontology and pathway analyses of the set of genes required for tau uptake provides a useful overview of the cellular mechanisms required for tau uptake. Gene ontology and functional analysis of the set of genes above the empirical significance threshold of p<0.01 found significant enrichment in several categories related to receptor-mediated endocytosis, including lipoprotein particle receptor binding, the endocytic vesicle membrane and the CCC complex ( Figure 2B). In terms of cellular localisation, genes required for monomeric tau uptake were significantly enriched in proteins that are localised to clathrin-coated vesicles, endosomes, including late endosomes, COPI vesicles and ER-Golgi transport ( Figure 2C).

CRISPR screen identifies genes and pathways required for neuronal uptake of aggregated tau
Screens for human neuronal uptake of aggregated tau were carried out in triplicate and significantly enriched genes identified by MAGeCK analysis of the three screens in combination 15 . Using the same significance cutoff of p<0.01 as for analysis of the monomeric tau screen identified 228 genes required for aggregated tau uptake ( Figure 3). LRP1 did not reach that threshold (p=0.024), although highly ranked in the screen output (ranked gene 522/18019).
As for monomeric tau, genes required for neuronal uptake of aggregated tau include regulators of endocytosis, such as EEA1, HGS and the PI3 kinase subunits C3 and R4.
Similarly, genes encoding vATPase subunits were required for aggregated tau uptake (ATP6V0E2, ATP6V0D1, ATP6V0A1), as were autophagy genes such as GABARAPL1 and UVRAG. Several genes encoding proteins involved in intracellular vesicular trafficking were also required for aggregated tau uptake, including the AP-1 and AP-3 complexes (AP1G1, AP3S1), and the COG complex that regulates retrograde trafficking within the Golgi 21 (four of the eight COG subunits: 1, 4, 7 & 8).
Gene ontology and functional analysis identified categories enriched in the set of genes required for aggregated tau uptake, including vacuolar acidification, ER to Golgi vesiclemediated transport, COPI and II vesicle and the COG complex. In terms of cellular compartments, genes required for aggregated tau uptake code for proteins that are significantly enriched in the endosome, phagophore, late endosome, Golgi, ER and ER-Golgi transport ( Figure 3C).

Monomeric and aggregated tau access neurons by receptor-mediated endocytosis dependent on LRP1 and LRRK2
The CRISPR screens used here to identify genes required for neuronal tau uptake are based on a chronic loss-of-function strategy, with all genes targeted in neural progenitor cells, and tau uptake assayed in terminally differentiated neurons approximately 30 days later. An advantage of this approach is that it identifies genes that are not essential for neuronal differentiation and viability, and thus pathways that could potentially be manipulated therapeutically in vivo to slow disease progression. To test whether acute inhibition of key proteins would also alter tau uptake, we used a live-imaging assay of neuronal tau uptake, using tau labelled with the pH-sensitive dye (pHrodo) that fluoresces in the low pH environments of the late endosome and lysosome (see Experimental Procedures for details).
LRP1 has recently been identified as a major tau receptor for both monomeric and oligomeric tau 9 , and was one of the two most significant genes identified as required for monomeric tau entry. However, LRP1 was ranked lower for aggregated tau entry (ranked 522 of 18,019 genes; p=0.024) in the FACS-based screen. To test the dependency of monomeric and aggregated tau entry on interaction with LRP1, we used two different acute approaches to inhibiting tau-LRP1 interactions, combined with live-imaging of neuronal uptake of pHrodo-labelled tau.
Blockade of all LDL receptors by addition of the pan-LDL receptor chaperone RAP 22 reduced uptake of both monomeric and aggregated tau by approximately 50% (Figure 4). In separate experiments we simultaneously added pHrodo-tau and recombinant domain IV of LRP1, the region defined as binding tau 9 , to the culture medium, before imaging tau uptake ( Figure 4).
As with RAP, the addition of the LRP1 fragment reduced uptake of both monomeric and aggregated tau by approximately 50% (Figure 4), confirming LRP1as a major receptor for neuronal uptake of both monomeric and aggregated tau.
The large, multidomain protein encoded by LRRK2 has multiple roles in endocytosis and vesicle trafficking 16 . CRISPR knockout of LRRK2 in progenitor cells removes all LRRK2 function from that stage, through neuronal differentiation, and would be expected to cause considerable disruption of the endolysosome-autophagy system. Mutations in LRRK2 that are causal for PD with tau aggregation 23 are concentrated in the kinase domain (e.g., G2019S) and have been shown to be activating or gain of kinase function mutations 24 .
Therefore, to test the dependency of tau uptake on kinase activity, and ask whether acute inhibition would alter tau uptake, iPSC-derived neurons were treated with small molecule LRRK2 inhibitors for 3 hours before exposure to extracellular monomeric or aggregated tau ( Figure 4C, D). Acute inhibition of LRRK2 kinase activity with two different small molecule inhibitors (GSK2578215A and MLi-2) reduced uptake of monomeric tau by between 50% and 70%. Inhibition of LRRK2 also reduced uptake of aggregated tau, but to a lesser degree of between 40 and 50%, which may contribute to why LRRK2 was not identified in the CRISPR screen for aggregated tau uptake.
Two subunits of PI3-kinase, PIK3C3 and PIK3R4, were identified as required for uptake of both monomeric and aggregated tau. PIK3C3 (VPS34) is a catalytic subunit of several PI3K complexes involved in regulating multiple aspects of the endolysosome-autophagy system, including endosome trafficking 25 . PIK3R4 forms a complex with PIK3C3 and another protein required for tau uptake, UVRAG, that regulates endocytosis, endosomal trafficking and autophagy (together with Beclin-1referred to as class III PI3 kinase complex II 25 ). Loss of three of the subunits of complex II of class III PI3K all leads to reduced tau uptake, suggesting that class III PI3K activity is required for tau uptake.
To test the role of PI3K in tau uptake, all PI3K activity in human neurons was acutely inhibited by wortmannin 26 administration before addition of extracellular tau ( Figure 4C, D).
Live imaging of pHrodo-tau uptake demonstrated that wortmannin reduced uptake of extracellular monomeric and aggregated tau in a dose-dependent manner (Figure 4), demonstrating the dependency of tau uptake on PI3K activity.

Monomeric and aggregated tau use similar mechanisms for neuronal entry
Given the clear overlap between the genes identified as required for monomeric or aggregated tau entry, and the finding that proteins identified as required for monomeric tau entry are also required for aggregated tau entry, the degree of similarity between the mechanisms of neuronal uptake of both forms of tau was analysed at the levels of genes and protein complexes ( Figure 5). To do so, we compared the gene sets identified above from the monomeric and aggregated tau entry, defined using the empirical threshold significance of p<0.01.
There was a significant overlap in individual genes identified by the screens for monomeric and aggregated tau uptake (hypergeometric test p <.001; Figure 5A), confirming the overall similarity of the mechanisms for neuronal uptake of each form of tau. Genes common to both screens included the COG complex member COG4, the three class III PI3K complex members discussed above (PIK3C3, PIK3R4 and UVRAG), and the tyrosine kinase HGS, which is a known downstream effector of PI3 kinase, ESCRT component and regulator of endosome trafficking 27 .
In addition to individual genes, genes that contribute to four protein complexes were significantly enriched (FDR <0.05) among combined hits of the monomeric and aggregated tau uptake screens ( Figure 5B). These four were the Conserved Oligomeric Golgi (COG) complex, involved in vesicular trafficking within the Golgi, the vesicular ATPase that acidifies the lysosome and autophagosome, the PIK3C complex discussed above, and the CCC complex, which regulates endosomal recycling, including recycling of LDL receptors 28 .
Together, these emphasise the role of the endolysosome in uptake of both forms of tau, including receptor recycling, which was confirmed by the enrichment of categories including endocytic vesicle membrane and pH reduction in the ontology analysis ( Figure 5C) and the localisation of the proteins encoded by genes common to both screens in endosomes and lysosomes ( Figure 5D).

Cellular mechanisms for tau uptake and processing
To generate a coherent understanding of the pathways of tau entry and processing with neurons, the non-redundant set of genes required for uptake of either monomeric or aggregated tau was analysed for known direct protein-protein interactions ( Figure 6A). The combined set of 431 genes/proteins contained many known protein-protein interactions, which formed several discrete subnetworks ( Figure 6A).
Within the network is a large subnetwork of endosomal and endocytosis proteins ( Figure   6A), which includes clathrin heavy chain (CLTC), dynamin-2 (DNM2), EEA1 and AP2 subunits. This subnetwork is part of a larger set of interactions that includes RAB7A, LRRK2, PIK3C3 and PIK3R4, all of which have roles in regulating endocytosis and endosome trafficking 16,25 . Two other notable protein interaction networks/complexes in the dataset regulate vesicle trafficking and endosome sorting. Multiple components of the CCC complex that regulates endosomal sorting and receptor recycling are required for tau uptake, including CCDC22, CCDC93 and three COMMD proteins (2, 3 and 10). There is also a subnetwork centred on the Golgi vesicle trafficking COG complex (5 of the 8 core COG proteins), which is involved in Golgi vesicle trafficking. Four subunits of the vacuolar ATPase (ATP6V1D, ATPV0A1, ATPV0D1, ATPV0E2), together with three accessory or assembly subunits of the complex, VMA21, TMED119 and CCDC115, are all required for tau uptake.
Given that acidification of late endosomes and lysosomes is a key requirement for receptor recycling, disruption of that process is likely to have considerable impact on surface receptor number and composition 29 .
LRRK2 is the most densely connected protein within the protein-protein interaction network, with RAB7A and clathrin heavy chain (CLTC) also in the top five most connected proteins ( Figure 6B-D). LRRK2 interacts directly with both RAB7A and CLTC, as well as interacting indirectly via mutually-shared interactors. These include the adaptor complex subunit AP2M1, which interacts with each of LRRK2, RAB7A and CLTC ( Figure 6B-D). Overall, the number and density of protein-protein interactions in the tau uptake network that mediate endocytosis, endolysosome function and vesicular traffic underscore the importance of receptor-mediated endocytosis for neuronal uptake of extracellular tau.

Neuronal uptake of extracellular tau shares functional similarities with receptor-mediated viral entry
Current models for tau-mediated propagation of pathology between neurons proposes that pathogenic forms of tau seed aggregation in receiving cells in the cytoplasm. Thus, following endocytosis, extracellular tau is likely to exit the endolysosome as part of its pathogenic mechanism. This echoes the life-cycle of many viruses, which use a variety of mechanisms to enter human cells, including receptor-mediated endocytosis and micropinocytosis 30 , and thence to deliver genetic material to the cytoplasm or nucleus for replication. The host factors or genes that regulate the entry and life cycle of several viruses have been comprehensively studied using RNAi and CRISPR screens. These include the neurotropic Zika virus 31 and Influenza A 32 , which enter cells via receptor-mediated endocytosis, and Ebola 33 , which uses micropinocytosis.
To compare the biology of tau entry and processing with that of genes required for different viruses entry, the sets of genes identified in the screens here were compared with those reported for Ebola, Zika and Influenza A infection (Figure 7). Significant overlaps in the number of genes common to both the monomeric and aggregated tau screens were found with those reported for Zika and Influenza A infection ( Figure 7A), but not with Ebola infection ( Figure 7A). Genes required for both tau entry and either Zika or Influenza A infection and replication are significantly enriched in genes encoding the vATPase or its assembly, the COG complex and intracellular vesicular trafficking, as reflected in the gene ontology enrichments ( Figure 7B, C), with their encoded proteins enriched in proteins localised to the endosome, lysosome and Golgi network ( Figure 7D, E). The overlap in routes of cell entry by viruses and tau protein, at the levels of both individual genes and cellular pathways, suggests that tau entry in disease spreading is a quasi-infectious process.

DISCUSSION
We report here the characterisation by whole genome CRISPR knockout screens of the pathways for extracellular tau uptake by human cortical excitatory neurons, the primary cell type affected in tauopathies. Independent screens for full length monomeric and heparinaggregated tau avoided making assumptions about the forms and fragments of tau that are present in interstitial fluid in the CNS and of the forms that are competent to enter neurons, and provided a broad view of the mechanisms for tau entry to human neurons. Consistent with our previous findings 7 , whole genome functional screens found that monomeric and aggregated tau are taken up by neurons by overlapping pathways, with some significant differences, most notably the dependency of aggregated tau uptake on protein glycosylation.
The major mode of entry identified in this study for both monomeric and aggregated tau is LRRK2-regulated receptor-mediated endocytosis, which is perturbed by a variety of disruptions of the integrity of the endolysosome-autophagy system and intracellular vesicular transport between the Golgi and other cellular compartments.
This study confirmed that the low density lipoprotein receptor LRP1 is a primary receptor for both monomeric and aggregated tau in human neurons, as recently described 9 . The identification of LRP1 as one of the top two ranked genes required for monomeric tau uptake is an important validation of the functional genetic approach taken here to understanding the cell biology of neuronal tau uptake. It is noteworthy that two different strategies for acutely inhibiting tau-LRP1 interactions, using the RAP chaperone and domain IV of LRP1, found that LRP1 mediates at least half of tau uptake in neurons. However, neither strategy completely blocked tau entry, suggesting the LRP1 may not be the sole receptor for tau entry.
CRISPR screens for uptake of monomeric and aggregated tau identified similar genes and pathways required for uptake of each form of the protein, but with some differences.
Although the specific genes identified in each screen were not identical, there was a high degree of convergence of hits in protein complexes, organelles and cellular pathways, particularly in genes encoding proteins regulating later stages of intracellular trafficking and processing, beyond the early endosome, and more general intracellular vesicular trafficking.
For each form of tau, key regulators of receptor-mediated endocytosis were identified at the level of endocytosis for monomeric tau (clathrin heavy chain and dynamin-2), and within early endosomes for aggregated tau (EEA1). The differences here in specific genes identified is unlikely to reflect alternative routes for cellular entry of the forms of tau, as we and others have previously shown that aggregated tau is also dependent on dynamin-2 for neuronal uptake 7,8 , although it was not identified in CRISPR knockout screens here, confirming that both forms are internalised via dynamin-dependent endocytosis.
Beyond the early endosome, tau uptake is highly dependent on class III PI3-kinase, which was confirmed by acute small molecule inhibition. Class III PI3K is a major regulator of both endosome recycling to the plasma membrane, as well as autophagy and phagocytosis 25 , and its involvement highlights that perturbations of receptor recycling and of wider vesicular trafficking, including autophagy, results in a loss of neuronal capacity for tau uptake.
Alterations in receptor recycling is a well-characterised consequence of loss of function of proteins involved in endosome sorting and recycling 29 . We report here that loss of SNX17 and all of the individual components of the CCDC22/CCDC93/COMMD complex (CCC complex) reduces neuronal uptake of tau. Given that a key cargo for SNX17-mediated recycling is LRP1 17 , and that SNX17 interacts with the WASH/CCC complex 17 , it is likely that one consequence of loss of each of these proteins is a reduction in plasma membrane levels of LRP1. Similarly, late endosome and lysosome acidification is essential for receptor recycling during the sorting process 34 , and we find that loss of multiple vATPase subunits, as well as accessory subunits and assembly proteins, all reduce tau uptake. In contrast, whereas tau is involved in Alzheimer's disease pathogenesis, and mutations in MAPT/tau are a cause of frontotemporal dementia 1 , LRRK2 mutations or variants are not associated with the development of tauopathies. However, recent GWAS for rate of progression of a tauopathy, progressive supranuclear palsy, identified variants in a putative enhancer for LRRK2 that increase rate of progression 37 . We report here that LRRK2 is required for uptake of monomeric and aggregated tau by human neurons, indicating that LRRK2 mediates spreading of tau pathology in tauopathies, such as Alzheimer's disease.
Furthermore, acute inhibition of the kinase activity of LRRK2 is sufficient to block neuronal tau uptake, demonstrating that LRRK2 is actively involved in mediated receptor-mediated endocytosis of extracellular tau. These findings suggest that small molecule inhibition of LRRK2 may have a role in slowing disease progression in neurodegenerative diseases involving spread of tau aggregation, including Alzheimer's disease.

Production and characterization of human iPSC-derived cerebral cortex neurons
Human pluripotent stem cell lines used in this study were KOLF2-C1 (WTSIi018-B-1) and derivative constitutively expressing Cas9 protein (KOLF2-C1 Cas9). KOLF2-C1 Cas9 was generated by integration of a Cas9 transgene driven by a CAGG promoter at the AAVS1 locus. Cells were nucleofected (Lonza) with recombinant enhanced specificity Cas9 protein (eSpCas9_1.1), a synthetic crRNA/tracrRNA (target site ggggccactagggacaggat tgg) and a homology directed repair template (https://www.addgene.org/86698/) followed by selection in neomycin (50 ug/ml) and clonal isolation 38 . Integrated clones were identified using PCR across the homology arms and validated by Sanger sequencing of the entire transgene and measurements of Cas9 activity. Random integration was screened for using PCR within the antibiotic resistance gene of the template plasmid.
Directed differentiation of human iPSCs to cerebral cortex was carried out as described, with minor modifications 13,14 . To establish identity and quality of neuronal induction, gene expression profiling was performed on a custom gene expression panel 39  (A11010, all from Thermo Fisher Scientific). Samples were stained with DAPI (1:5000 in PBS). Images were acquired through Olympus Inverted FV1000 confocal microscope and processed using the Fiji software 41 .

Cas9 editing efficiency in human iPSC neurons
Knockout efficiency in KOLF2-C1 Cas9 neurons was assessed using a lentiviral reporter

Genome-wide CRISPR RNA guide lentiviral library
The Human CRISPR Library v.

Recombinant tau monomer and aggregate preparation
Recombinant tau protein was purified as previously described 7

Guide RNA sequencing
Genomic DNA was extracted from cell pellets using QuickExtract DNA Extraction Solution (Cambio; QE09050) as per the manufacturer's instructions. sgRNA amplification was performed using a two-step PCR strategy and purified using Agencourt AMPure XP (A63880; Beckman Coulter UK), Illumina sequencing (19-bp single-end sequencing with custom primers on the HiSeq2500 Rapid Run) and sgRNA counting were performed as described previously 12 .
CRISPR screen data analysis sgRNA count files from purified neuronal populations that had high levels of labelled transferrin and tau were compared to neuronal populations with high levels of labelled transferrin and low levels of labelled tau protein using MAGeCK (0.5.8) 15 . Replicate monomeric (n=2) and aggregated (n=3) tau uptake screens were analysed separately. sg-RNA rankings were summarised into gene-level statistics using the RRA algorithm. All subsequent analyses were performed in R 44 . For genes with two sets of sgRNAs, the higher ranking set was kept and the lower ranking set was discarded. An empirical threshold of unadjusted p-value < 0.01 was used to identify genes required for tau uptake in either screen, identifying 214 genes in the aggregated screen and 228 genes in the monomeric screen. Functional term enrichment analysis was performed using gProfiler 2 querying the GO, REAC, and CORUM databases 45 .
Compartment localisation enrichment was performed using localisation scores from the COMPARTMENTS database 46 . Localisation of each gene to a particular compartment is given a score up to 5 based on the strength of the supporting evidence. To calculate compartment localisation enrichment for the genes identified in the CRISPR screens, a gene set localisation score for each compartment was calculated by averaging the localisation score of all constitutive genes. The score ranged from 0 to 1, with 1 representing maximum localisation score for that compartment for all the genes in the set. Cellular compartments were removed from analysis if too undescriptive (gene-set localisation score >= 0.7), with low signal (gene-set localisation score <= 0.02), or with only weak evidence (no individual gene scoring 3.5/5 or above). Compartment localisation enrichment was calculated compared to the mean value obtained from 100,000 simulated random sets of 221 (aggregated, monomeric screens) or 431 (combined aggregated and monomeric), 15 (influenza A and combined aggregated and monomeric), or 8 (Zika and combined aggregated and monomeric) genes represented in the CRISPR library. Boot-strapping was used to calculate p-values and significance was adjusted for multiple comparison using false discovery rate. For significantly enriched compartments, the log2 fold-change of compartment score over random was mapped onto a diagram of cellular compartments using the sp package 47 .
Protein-protein interaction networks were constructed using the PSICQUIC package using the following databases: BioGrid, bhf-ucl, IntAct, MINT, UniProt, MBInfo, InnateDB 48 . All PPIs for the human proteins were maintained, removing reciprocal and self interactions.
Network graphs were drawn using igraph and visNetwork packages 49 .

Live imaging of tau entry and processing by human iPSC neurons
Neurons grown (65 days after induction) on 96-well plates (CellCarrier96; Perkin Elmer Life Sciences; 6055308) were imaged for fluorescent pHrodo-labelled tau protein (excitation at 577nm and emission at 641nm in an acidic environment) using a 20× IncuCyte S3 Live-cell Analysis System (Sartorius). Bright-field and fluorescence emission images were collected at 45min or 1hr intervals for 16hrs at 37°C. Parameters for fluorescent objects were set (fluorescent intensity and contrast), quantified and intensity of the field was normalised to confluency using the Incucyte Analysis Software. Data were analysed using Prism Software

Statistical Analysis
The number of replica wells and experiments are indicated in the figure legends for each assay where appropriate. For live-cell imaging assays using pHrodo-labeled protein, intensity normalised to confluence of the field using the Incucyte Analysis Software.
Typically, 9 fields were recorded per well and an average generated from the Incucyte Analysis Software. Analyses were performed using the Prism version 8 software (GraphPad). One-way ANOVA with Dunnett's multiple testing or Student's t test was used where appropriate.
Hypergeometric tests were used to evaluate significance of overlap between gene sets.
Genes required for uptake of Ebola, influenza A, and Zika viruses were obtained from published CRISPR screens [31][32][33] . Since the Ebola data were also analysed with MAGeCK, we defined genes required for viral uptake using an unadjusted p-value threshold of 0.01; this information was not available for the influenza A and Zika studies, so we defined genes required for viral uptake as having an adjusted p-value < 0.05. Hypergeometric tests for overlap in screen hits were carried out using the set of genes analysed in both studies as background.     LRP-1 Cluster IV Fc chimera protein or vehicle control (PBS) for 3hrs prior to addition to extracellular media and live imaging of neuronal uptake of tau. Histograms report tau uptake at final timepoint, which was significantly reduced by both treatments, and for monomeric and aggregated tau. Minimum of 3 wells per treatment, and average of 9 fields of view per well. Error bars indicate SD. Significance was determined using one-way ANOVA (n = 3; * p < 0.05, * * p < 0.005, and * * * p < 0.0001, Dunnett's test for multiple comparisons). Significance was determined using one-way ANOVA (n = 3; * p < 0.05, * * p < 0.005, and * * * p < 0.0001, Dunnett's test for multiple comparisons).    A. The sets of genes required for aggregated tau uptake or monomeric tau uptake significantly overlap with the sets of genes required for cellular uptake of the Influenza A and Zika viruses, but not the Ebola virus (hypergeometric test, p-value < 0.01 (**), < 0.001 (***); ns: not significant).

B-C. Functional annotations enriched among 15 genes required both for uptake of Influenza
A virus and for either aggregated or monomeric tau uptake (B) or among 8 genes required both for uptake of Zika virus and for either aggregated or monomeric tau uptake (C) (representative selection).

D-E.
Genes required for uptake of Influenza A virus and for either aggregated or monomeric tau uptake (D) and genes required for uptake of Zika virus and for either aggregated or monomeric tau uptake (E) code for proteins with a significantly higher than random localisation score in particular cellular compartments in the COMPARTMENTS dataset (FDR