key: cord-0426508-qso2s9a1
authors: Wells, Michael F.; Nemesh, James; Ghosh, Sulagna; Mitchell, Jana M.; Mello, Curtis J.; Meyer, Daniel; Raghunathan, Kavya; Tegtmeyer, Matthew; Hawes, Derek; Neumann, Anna; Worringer, Kathleen A.; Raymond, Joseph J.; Kommineni, Sravya; Chan, Karrie; Ho, Daniel; Peterson, Brant K.; Piccioni, Federica; Nehme, Ralda; Eggan, Kevin; McCarroll, Steven A.
title: Natural variation in gene expression and Zika virus susceptibility revealed by villages of neural progenitor cells
date: 2021-11-09
journal: bioRxiv
DOI: 10.1101/2021.11.08.467815
sha: 853f5c1395224262160c4a000b63a7a013101bdb
doc_id: 426508
cord_uid: qso2s9a1

Variation in the human genome contributes to abundant diversity in human traits and vulnerabilities, but the underlying molecular and cellular mechanisms are not yet known, and will need scalable approaches to accelerate their recognition. Here, we advanced and applied an experimental platform that analyzes genetic, molecular, and phenotypic heterogeneity across cells from very many human donors cultured in a single, shared in vitro environment, with algorithms (Dropulation and Census-seq) for assigning phenotypes to individual donors. We used natural genetic variation and synthetic (CRISPR-Cas9) genetic perturbations to analyze the vulnerability of neural progenitor cells to infection with Zika virus. These analyses identified a common variant in the antiviral IFITM3 gene that regulated IFITM3 expression and explained most inter-individual variation in NPCs’ susceptibility to Zika virus infectivity. These and other approaches could provide scalable ways to recognize the impact of genes and genetic variation on cellular phenotypes. HIGHLIGHTS Measuring cellular phenotypes in iPSCs and hPSC-derived NPCs from many donors Effects of donor sex, cell source, genetic and other variables on hPSC RNA expression Natural genetic variation and synthetic perturbation screens both identify IFITM3 in NPC susceptibility to Zika virus A common genetic variant in IFITM3 explains most inter-individual variation in NPC susceptibility to Zika virus

Variation in the human genome contributes to abundant diversity in human traits and 20

vulnerabilities, but the underlying molecular and cellular mechanisms are not yet known, and will 21 need scalable approaches to accelerate their recognition. Here, we advanced and applied an 22 experimental platform that analyzes genetic, molecular, and phenotypic heterogeneity across 23 cells from very many human donors cultured in a single, shared in vitro environment, with 24 algorithms (Dropulation and Census-seq) for assigning phenotypes to individual donors. We used 25 natural genetic variation and synthetic (CRISPR-Cas9) genetic perturbations to analyze the 26 vulnerability of neural progenitor cells to infection with Zika virus. These analyses identified a 27 common variant in the antiviral IFITM3 gene that regulated IFITM3 expression and explained 28 most inter-individual variation in NPCs' susceptibility to Zika virus infectivity. These and other 29 approaches could provide scalable ways to recognize the impact of genes and genetic variation 30 on cellular phenotypes. Humans and indeed all natural populations harbor immense diversity in biological traits, affecting 48 almost all organ systems and physiological processes. Reservoirs of natural variation allow 49 populations to adapt to existential crises and other strong selective pressures. Studies of 50 quantitative traits and disease risk in populations -whether of humans, animals, or plants -reveal 51 a large role for standing variation in shaping phenotypic diversity and enabling adaptation. And 52 yet, the mechanisms by which natural genetic variation acts through molecular, cellular and 53 developmental biology to shape trait variation are largely unknown. 54

Genetic variation in developmental processes has an important but still poorly understood role in 55

shaping phenotypes throughout life. The spatial and cell-proliferative events that establish the 56 cellular compositions of tissues -the cell types, their relative abundances, and their spatial 57 arrangements -almost certainly affect those organs' physiological properties years later. 58

Consistent with this idea, many genetic discoveries in autism and schizophrenia point to genes 59 that are most highly expressed during fetal brain development genes -in every adult organ analyzed. The relationships of these and other haplotypes to gene 73 expression during development is largely unknown. 74

Relationships between common genetic variants and gene expression can also be revealed using 75

in vitro stem cell-derived models, which have become increasingly adopted with advancements 76

in differentiation protocols and the demonstrated similarities of these cells to corresponding cell 77 types in tissues. Fast, reliable induction techniques for generating NPCs, neurons, glia, and 78 oligodendrocytes from human embryonic stem cells (hESCs) and induced pluripotent stem cells 79

(iPSCs) now enable investigations into gene expression variation. However, maintaining cells 80 from many human donors in separate culture environments results in technical variation that can 81 obscure biologically-relevant effects, and requires substantial resources and effort. 82

Though thousands of eQTLs have been found in every tissue, in iPSCs, and in iPSC-derived cells  83 of various types, we know little today about the ways in which these effects percolate through 84 cells to influence their functional phenotypes. Which of these genetic variants actually change a 85 cell's phenotype in a meaningful way? Biological pathways are built to be robust to many kinds of 86 perturbation and may contain the effects of many genetic variants, even those that regulate a 87 gene's expression . The precise expression dosage of most genes is of unknown biological  88  significance. Thus, it is essential to understand the relationships among genetic variants, gene  89 expression, and the physiological phenotypes of cells that shape disease processes. 90

Here we describe the coordinated application of three approaches for dissecting these aspects of 91 a cell's life and their relationships to one another. One is genetic multiplexing, in which tens of 92 thousands of cells from scores of donors are analyzed simultaneously by single-cell RNA-seq, 93

revealing effects of genetic variants, cell-donor and cell-source properties on RNA expression at 94 population scale. Another is "Census-seq", a rapid, inexpensive method for relating cellular 95 phenotypes to natural genetic variation by sequencing the genomic DNA from such "cell villages" 96 (Mitchell et al., 2020) . A third is functional CRISPR-Cas9 screens, which explore thousands of 97 artificial genetic perturbations simultaneously (Joung et al., 2017; Ophir et al., 2014) . 98

By applying these experimental platforms in a coordinated way, we find high levels of inter-99

individual variation in susceptibility of NPCs to the neurotropic Zika virus (ZIKV), then uncover a 100 common, large-effect SNP in the antiviral IFITM3 gene that explains more than 50% of this 101

physiological variation and appears to act by regulating expression levels of IFITM3 in NPCs. 102

Analyses of the high-volume data from these same experiments helped us understand genetic, 103 biological and technical sources of variation on gene expression in human pluripotent stem cells 104

(hPSCs) and NPCs. The population-scale nature of these experiments made it possible to 105 describe and quantify the effects of donor sex, cell source, and cell reprogramming methods. We 106 also identify hundreds of NPC eQTLs, some of which affect expression levels of critical 107 neurodevelopmental genes. 108

We hope that these results enable better understanding of sources of variation in hPSC-derived 109 systems while suggesting possibilities for integrated experimental approaches that use natural 110 and synthetic genetic perturbations to understand disease-relevant functional variants and their 111 effects on specific cell types. 112 113 RESULTS 114

Analysis of cellular variation in "Dropulations" 115

To understand how natural genetic variation and donor-specific biological effects shape cellular 116 phenotypes, we sought to eliminate other sources of variation by culturing cells from very many 117 donors in a common environment and analyzing them all together. Transcribed SNPs contain 118 abundant information that can be used to identify the donor of a biological sample using the alleles 119 that are present in RNA transcripts. The combination of hundreds of transcribed SNPs can act as 120 a cell-intrinsic barcode that uniquely identifies the donor of an individual cell (Kang et al., 2018) . 121 We further developed genetic multiplexing analysis: (i) to address challenges inherent to scRNA-122 seq experiments, such as ambient RNA; (ii) to utilize unique molecular indicators (rather than 123 reads) as the basis for demultiplexing; and (iii) to allow scalability up to hundreds of potential 124 donors (Methods). We provide the resulting software ("Dropulation", for droplet-based sequencing 125 of populations) in an open-source format (https://github.com/broadinstitute/Drop-seq). 126

We performed a variety of in silico and experimental analyses to confirm the accuracy with which 127 cells were correctly assigned to donors by this approach (Methods, Fig. 1a , Supplementary Note, 128 Fig. S1 ). 129

The ability to measure mRNA expression in a common culture environment made it possible to 130 quantify effects that have long been of great concern (but uncertain practical impact) in stem cell 131 research, including the effects of donor sex and cell source. To characterize such effects, we 132 established a 104-donor human iPSC "cell village" (Figure 1a ) that included iPSCs derived from 133 multiple cell sources (skin and blood cells) and from male and female donors (Figure 1b , Figure  134 S2a). We profiled this cell village by scRNA-seq (10X Genomics platform) five days after pooling 135 the 104 iPSC lines. We analyzed 86,185 cells by scRNA-seq, sampling their RNAs to an average 136 depth of 104,160 UMIs per cell and assigning each cell to its donor-of-origin using Dropulation 137

( Figure S2b -c). 138

When cultured together, the iPSCs from the 104 donors exhibited highly similar RNA-expression 139 patterns, with cells from the 104 donors distributing similarly across a two-dimensional projection 140 of the single cells' RNA-expression patterns ( Figure 1c ). The primary source of variation in the 141 individual cells' expression profiles appeared to be their progress through the cell cycle ( Figure  142 1d). Heterogeneity in cell state or identity can also impact gene expression ( Figure S2d ). As 143 expected, the great majority of the hiPSCs expressed pluripotency markers NANOG and OCT4 144 at high levels ( Figure 1e ); an exception (cluster 10, 4.3% of all cells) involved cells with higher-145 than expected expression of neural progenitor markers RIPPLY3 and SIX6, suggesting that they 146 have undergone spontaneous neural differentiation ( Figure S2e ). Another set of cells (clusters 6 147 and 7, 3.1% of all cells) exhibited high expression of UTF1 and MTG1, suggesting poor 148 differentiation potential; notably, these cells came from only a handful of cell lines ( Figure S2f) . A 149

subset of single-cell expression profiles exhibited lower UMI counts and higher percentages of 150 nascent transcripts with intronic reads ( Figure S2g -h); we interpret these profiles to have arisen 151 from nuclei (likely generated by unintentional cell lysis during handling) rather than intact cells. 152 iPSC lines are routinely created from skin (fibroblasts) or from blood (PBMCs); fundamental 153 methodological questions involve the comparability of iPSCs created from different tissue 154 sources. While it is reasonable to hypothesize that tissue source for iPSC reprogramming 155 influences gene expression, our data suggests that such effects were small: Only four RNAs (all 156 noncoding) exhibited genome-wide-significant differences in expression between the 56 skin-157 derived and 45 PBMC-derived iPSC lines (Figure 1fg ; Data File 1). A more-sensitive enrichment 158 analysis of sub-significant differential expression measurements indicated a modest enrichment 159 associated with the somatic cell of origin (Data File 1): genes preferentially expressed in blood-160 derived iPSCs were nominally enriched in a myeloma related pathway, while genes preferentially 161 expressed in skin-derived iPSCs were nominally enriched in a gene set associated with 162 melanoma relapse ( Figure S2i ). This finding suggests that there is modest retention of epigenetic 163 memory inherited from the parental cell source of origin in the iPS cell lines, but that few protein-164

coding genes are strongly affected by this memory. 165

Many human phenotypes show sex-biased differences, and RNA-expression levels of many 166 genes differ on average between males and females in various tissues (Oliva et al., 2020) . The extent to which cell-autonomous biology -as opposed to, for example, circulating hormones -168 contributes to such differences is unknown. The experimental design made it possible to isolate 169 and quantify cell-autonomous effects, since (i) environmental and non-cell-autonomous effects 170

were controlled across cells growing in the same culture, and (ii) very many male and female 171 donors were sampled. The RNA-expression profiles of the individual cells initially appeared to be 172 strongly distinguished by donor sex (Figure 1h ), though this difference largely disappeared when 173 we limited analysis to autosomal genes ( Figure 1i -j), suggesting that it arose largely from Y-linked 174 genes and genes that escape X chromosome inactivation. 175

How strong is sex-biased expression in iPSCs relative to the routine effects of inter-individual 176 variation? We compared pairs of donors within and between sex to generate pairwise distributions 177 of differentially expressed genes. We found similar numbers of differentially expressed autosomal 178 genes in same-sex comparisons (XX vs XX; XY vs XY) as we did in across-group (XX vs. XY) 179

comparisons, indicating that on average iPSCs from XX and XY individuals were roughly as 180 different from each other as same-sex individual pairs regardless of cell source. These results 181 suggest that differences in gene expression generated by donor sex and source cell-type are 182 small compared to the routine effects of inter-individual variation, and that differences in 183 expression of sex-chromosome genes do not lead to broader effects on cells' biology in this 184

context. 185

To further explore these modest, quantitative effects that become apparent in a population-scale 186 analysis, we computed differential gene expression across all 104 donors using the limma/voom 187 software package ( Signatures Database (MSigDB; (Subramanian et al., 2005) . The most significant gene sets 194 associated with differences across male and female donor lines mapped to pathways related to 195 X chromosome inactivation and imprinting ( Figure S2j ). This signal disappeared when the 196 analysis was limited to autosomal genes, indicating that differences in gene expression due to 197 donor sex was largely driven by genes on the sex chromosomes (Data File 1). 198

The small effects of donor sex and cell source, relative to the routine effects of inter-individual 199 variation, suggest that population-scale iPSC experiments can successfully utilize blood-and 200

fibroblast-derived iPSCs together, though will ideally use designs in which these factors do not 201 confound the effects of other variables of interest (such as genotype or case/control status). 202 203

Effects of common genetic variation on RNA expression in hiPSCs 204

To map expression QTLs, we tested for associations between donors' genotypes and their gene 205 expression (summed across their individual iPSCs) using a linear regression model that included 206 independent variables for biological and technical covariates, utilizing the MatrixEQTL analytical 207 pipeline (Shabalin, 2012) ( Figure 2a ). Some 6,247 genes exhibited significant association to one 208 or more SNPs ("eGenes", at a q-value < 0.05) across the 104 hiPSC lines (Figure 2b, Data File  209 2). 210

To critically evaluate the utility of the genetically multiplexed format for eQTL discovery, we 211 compared rates of eQTL detection across studies of diverse tissue and cell types, including GTEX 212 and a report discovering eQTLs in human stem cells that were cultured and analyzed in an 213 arrayed format (DeBoever et al., 2017). Across these studies, power to discover eQTLs increased 214 almost linearly in relation to the number of donors sampled (with a slope of 29.2 eQTLs per donor). 215

Analysis of cells in a "cell village" format identified 60 eQTLs per donor (Figure 2b ), suggesting 216 that the multiplexed approach helped increase the sensitivity of eQTL detection, likely by 217

controlling for technical and environmental factors that introduce noise into most eQTL discovery 218 efforts. 219

To assess the relationship of these iPSC eQTLs to eQTLs discovered by large-scale analysis of 220 human tissues, we performed a sign test, asking what fraction of these SNPs exhibited the same 221 direction of association to the same gene-expression phenotypes in other tissues and 222 experiments. We found an 88% concordance (by sign test) between the eQTLs we found, and 223 eQTLs found by the arrayed analysis of many individual iPSC lines. In addition, we observed 224 similar sign test results when comparing to GTEx tissue and cell type data, with the strongest 225

concordance coming from tissues with many mitotic cells ( Figure S3a ). 226

Finally, we explored the significance of eQTL search parameters, such as the search window 227

(around each gene) used for cis-eQTL discovery, and the allele frequencies of the SNPs 228 considered; both involve a tradeoff between the increased sensitivity of a broader search, and the 229 reduction in sensitivity when multiple hypothesis testing is fully addressed in such a search ( Figure  230 2c). We hope that this information is useful in guiding other eQTL discovery studies. 231

hPSCs and hPSC-derived cell types offer a unique opportunity to analyze genetic influences upon 233 the physiological phenotypes of living cells. They also present an opportunity to learn from both 234 natural genetic variation and synthetic genetic perturbations. We sought to identify and 235 understand genetic effects upon a physiological phenotype, the vulnerability of human NPCs to 236 the neurotropic Zika virus. Many of these SNPs may affect the expression of human genes that play prominent roles in viral 242 entry or replication mechanisms, as well as those that mediate the innate immune response 243 (Kenney et al., 2017) , and in doing so influence the probability of productive infection and spread 244

to neighboring host cells. While numerous relationships between specific SNPs and viruses have 245 been identified across a range of cell types, our understanding of the interplay among human 246 genetic variation, viral susceptibility, and the brain is still in its infancy. 247

The Zika virus is a global health concern, with an outbreak that originated in South America and 248 spread to over 50 countries between 2015 and 2016 ( independently-cultured hESC-derived SNaP lines with ZIKV-Ug (MOI = 1) and measured 266 infectivity levels at 54 hpi ( Figure 3a ). We observed a surprisingly high degree of variability in viral 267 sensitivity, with mean infectivity rates ranging from 0.7% (RUES1 cell line) to 99.4% (Mel2; Figure  268 3b). 269

These data suggest that cell-intrinsic factors make a large contribution to the inter-individual 270 variation in vulnerability to ZIKV. We sought to understand this through both natural genetic 271 variation and synthetic genetic perturbations, which we hypothesized might lead us to the same 272 genes. We also detected 195 and 147 genes that were significantly depleted in the ZIKV-PR and ZIKV-318

Ug screens, respectively, suggesting that their ablation rendered cells more sensitive to ZIKV-319 mediated cell death (adjusted p-value < 0.05; Figure 3f -g). The 48 depleted genes that were 320 detected in both screens included the Type I interferon-responsive genes IFNAR1-2, STAT2, and 321

IFITM3. 322

The results of the primary survival screen were confirmed through CRISPR-and antibody-323 mediated secondary assays. We first generated a H1 hESC line that constitutively expresses 324

Cas9 (clone H1-36-23) and validated its genome-editing properties ( Figure S4c-f ). H1-Cas9 325 hESCs were then induced to SNaPs and transduced with lenti-gRNAs in an arrayed format for 326 infectivity assays at 54 hpi and cell viability assays at 120 hpi using ZIKV-Ug (MOI = 1). We 327 targeted a subset of genes that were significantly enriched in both SNaP survival screens and 328

represented the spectrum of protein complexes and biological processes identified by these 329 primary screens. As predicted, administering guides that targeted the EMC genes, SLC35B2, 330 OST complex components, and vATPase subunits, but not TAM receptors, resulted in 331 significantly improved cell viability and reduced infectivity compared to NT gRNA controls ( Figure  332 3h-i, S4g-i). Interestingly, disruption of the transmembrane-spanning integrin subunit ITGB5 333 improved cell viability and dramatically reduced levels of infection ( Figure S4j ). Similar protection 334 was observed when SNaPs were treated with anti-integrin aVb5 antibody prior to ZIKV-Ug 335 exposure ( Figure S4k treatments. Perhaps most importantly, these data nominate a small number of genes as potential 345 mediators of the differential viral susceptibility we observed across SNaP lines. 346 347

Having identified (in the CRISPR screens) genes that mediate ZIKV infectivity and death of 349

SNaPs, we sought to identify natural genetic variants that affect the expression of these genes in 350

SNaPs. We pooled 44 hESC-derived SNaP lines into a village (Village-44; Figure 4a ) and three 351 days later processed these cells for scRNA-seq and donor re-identification. As expected, cell 352 cycle stage and differentiation status contributed to variation in gene expression (Figure 4b-c) . 353 We used scRNA-seq data to assess the quality and identity of the cells in this village, finding that 354

SNaPs predominantly clustered with fetal NPCs when compared to an integrated reference 355 dataset composed of human fetal and adult brain cells measure SNaP similarities to in vivo cells, we found that 83.1% of the SNaPs were determined to 360 most closely resemble fetal NPCs, suggesting that the SNaPs have an expression profiles similar 361

to those of their in vivo counterparts. In addition to confirming the relevance of this cell type to its 362 in vivo counterpart, analysis confirmed the consistent production of high-quality SNaPs across all 363 donors ( Figure S5c-d) . 364

We identified 961 genes whose expression levels were associated with the genotypes of nearby 365

SNPs ("eGenes", Data File 4). Many of these genes have known roles in fetal brain development. 366

They included genes associated with neurodevelopmental disorders, including MAPK3, PEX6, 367

and DMPK as well as the fetal brain transcriptional activator CHURC1 and the Williams Syndrome 368 genes WBSCR16 and WBSCR27 ( Figure S5e , Data File 4). They also included many genes 369 associated with autism spectrum disorder, epilepsy, and microcephaly such as DPP10, HNRNPU, 370 and TSEN2. This analysis provides an inventory of genes regulated by common genetic variation 371 in human NPCs, which could be useful for future investigations of variant effects on a range of 372 neurodevelopmental processes and diseases. 373

We had hypothesized that screens of the human genome using synthetic (CRISPR) and natural 374 (genetic) perturbations might converge. We cross-referenced the eGenes from our eQTL analysis 375 to a list of 125 genes that were significantly enriched or depleted in both ZIKV survival screens. 376 We found an overlap of 8 genes (Figure 4d) IFITM3 at levels 4.82-fold higher than donors homozygous for the alternate allele (T) (p = 4.28 x 394 10 -6 , r 2 = 0.399; Figure 4e ). We hypothesized that by enhancing expression of this antiviral gene 395 the reference allele of rs34481144 confers protection from ZIKV infection, relative to the alternate 396 allele. 397

To formally test the hypothesis that the reference allele protects SNaPs from ZIKV infection, we 399 exposed SNaP Village-44 to ZIKV-Ug (MOI = 1) or mock media (Figure 5a ). At 54 hpi, we FACS-400 sorted the cells into four fractions based on ZIKV envelope protein 4G2 antibody-stained GFP 401 signal intensity (ZIKV-Negative, -Low, -Mid, -High) before harvesting pellets for DNA sequencing. 402 We then analyzed each cell fraction using Census-Seq (Mitchell et al., 2020) to estimate each 403 donor's cellular contribution to the different fractions. Donors with the rs34481144 TT genotype 404 were greatly over-represented relative to rs34481144 CC donors in the ZIKV-positive populations 405 relative to the ZIKV-negative pool (r 2 = 0.225, p = 3.04 x 10 -3 ; Figure 5b -e). No such relationship 406 was observed with other significant Village-44 eQTLs located in antiviral/host factor genes ( Figure  407 S6a). These data support the hypothesis that the rs34481144-T allele renders SNaPs more 408 vulnerable to ZIKV infection compared to cells harboring the rs34481144-C allele (Figure 5f ). 409

We next asked whether the apparently strong effect of the rs34481144 genotype on ZIKV 410 susceptibility is cell-intrinsic or arises in an unexpected way from the cell-village experimental 411 design. To do so, we analyzed data from SNaP infectivity assays of independently-cultured 412 individual SNaP lines (Figure 3b ) and again found a significant correlation between rs34481144 413 genotype and the percentage of ZIKV-infected SNaPs (at 54 hpi) in each culture, where the 414 rs34481144 TT cells were more prone to infection than rs34481144 CC SNaPs (r 2 = 0.345, p = 2.53 415

x 10 -3 ; Figure 6a -b). Plotting these results against the Census-seq data showed agreement of 416 results between the pooled-culture and arrayed-culture experimental designs (r 2 = 0.703, p = 2.63 417

x 10 -5 ; Figure S6b ). 418

To further replicate these results, we infected a separate set of 36 human iPSC-derived SNaP 419 lines in an arrayed format using ZIKV-Ug (MOI = 1) or ZIKV-PR (MOI = 10). In these experiments, 420

rs34481144 genotype explained 58.8% and 29.4% (respectively) of the variation in cell survival 421 (Figure 6c-e) . 422

Our focus on the rs34481144 SNP had been driven by the implication of this gene in our CRISPR-423

Cas9 and eQTL screens. We wondered whether this SNP's effect on ZIKV infectivity was 424 sufficiently strong to have been identifiable in an unbiased genome-wide search for effects of 425 common variants on this cellular phenotype. We thus performed genome-wide association 426 analysis on the survival data from the 36 arrayed cell lines, measuring the association of this 427 phenotype with more than 1 million common SNPs. rs34481144 was the human genome's top-428

scoring SNP in the ZIKV-Ug dataset, reaching genome-wide significance even in this modest 36-429 donor sample (p = 9.0 x 10 -10 ; Figure 6f ). In an analogous analysis of infectivity with the PR strain 430

of ZIKV, rs34481144 also associated strongly, but just below genome-wide significance (p = 3.2 431

x 10 -7 ; Figure 6g) , and was the genome's second strongest common-variant association. 432

These results demonstrate that a genetically variable, cell-intrinsic property of human NPCs is the 433 major source of inter-individual variation in their susceptibility to a viral pathogen. findings, we believe that the rs34481144-T allele confers similar vulnerabilities to developing 446 human brain cells exposed to ZIKV. The authors declare no competing interests. 

A survey of human brain transcriptome diversity at 541 the single cell level

Large-Scale Profiling Reveals the Influence of 544

Genetic Variation on Gene Expression in Human Induced Pluripotent Stem Cells

Cellular Phenotypes in Human iPSC-Derived 548 Neurons from a Genetic Model of

Viruses are a dominant driver of 550 protein adaptation in mammals

Association 553 between genetic variants in nos2 and tnf genes with congenital zika syndrome and severe 554 microcephaly

Transcriptome of iPSC-derived 557 neuronal cells reveals a module of co-expressed genes consistently associated with autism 558 spectrum disorder

Pregnancy outcomes in interferon-beta-exposed patients with multiple 561 sclerosis: results from the European Interferon-beta Pregnancy Registry

Polymorphism of the Interleukin-10

Gene Is Associated with Susceptibility to Epstein-Barr Virus Infection

Distinct patterns of IFITM-mediated restriction 567 of filoviruses, SARS coronavirus, and influenza A virus

Population-scale single-cell RNA-seq 570 profiling across dopaminergic neuron differentiation

Genome-scale CRISPR-Cas9 knockout and 573 transcriptional activation screening

Multiplexed droplet single-cell RNA-576 sequencing using natural genetic variation

Human Genetic Determinants of 579 Viral Diseases

Ethnic variation in risk genotypes based on single nucleotide 581 polymorphisms (SNPs) of the interferon-inducible transmembrane 3 (IFITM3) gene, a 582 susceptibility factor for pandemic 2009 H1N1 influenza A virus

A probability-based approach for the analysis of large-scale 585 RNAi screens

Past and future spread of the arbovirus 588 vectors Aedes aegypti and Aedes albopictus

Genome-wide CRISPR screen for Zika virus 591 resistance in human neural cells

RACK1 controls IRES-mediated translation of 594 viruses

Disrupted in Schizophrenia 1 Regulates Neuronal 597 Progenitor Proliferation via Modulation of GSK3β/β-Catenin Signaling

Altered proliferation and 600 networks in neural cells derived from idiopathic autistic individuals

Polymorphisms of large effect 603 explain the majority of the host genetic contribution to variation of HIV-1 virus load

Mapping genetic effects on cellular phenotypes with cell villages

Zika Virus Associated with 609 Microcephaly

Delayed childhood 612 neurodevelopment and neurosensory alterations in the second year of life in a prospective 613 cohort of ZIKV-exposed children

Spatiotemporal 616 gene expression trajectories reveal developmental hierarchies of the human cortex

The impact of sex on gene 620 expression across human tissues

Infection-related microcephaly after the 2015 and 2016 Zika virus 623 outbreaks in Brazil: a surveillance-based analysis

Genome-Scale CRISPR-Cas9 Knockout Screening in Human Cells. Science

IFITM3 Rs12252-C Variant Increases Potential Risk for Severe Influenza Virus 629 Infection in Chinese Population

Opposing Brain 632 Differences in 16p11.2 Deletion and Duplication Carriers

Zika virus cell tropism in the 635 developing human brain and inhibition by azithromycin

limma 638 powers differential expression analyses for RNA-sequencing and microarray studies

Did Zika Virus Mutate to 641 Cause Severe Outbreaks?

Association between Zika Virus Microcephaly in Newborns with the rs3775291 Variant in Toll-645

Like Receptor 3 and rs1799964 Variant at Tumor Necrosis Factor-α Gene

Identification of Zika Virus and

Dengue Virus Dependency Factors using Functional Genomics

The IFITMs Inhibit Zika Virus Replication

Zika 653 virus disease in India -Update

Matrix eQTL: Ultra fast eQTL analysis via large matrix operations

Microcephaly epidemic related to the Zika virus and living conditions in Recife, 659 Northeast Brazil

IFITM3 directly engages and shuttles incoming virus particles to 662 lysosomes

A quantitative framework to 665 evaluate modeling of cortical development by neural stem cells

Comprehensive Integration of Single-Cell 668 Data

Gene set enrichment 671 analysis: A knowledge-based approach for interpreting genome-wide expression profiles

Zika Virus Infects Human Cortical 674 Neural Progenitors and Attenuates Their Growth. Stem Cell

Integrin αvβ5 Internalizes Zika Virus during Neural Stem Cells Infection and 677 Provides a Promising Target for Antiviral Therapy

Genetic Ablation of AXL Does Not Protect Human 680

Neural Progenitor Cells and Cerebral Organoids from Zika Virus Infection

Genome-wide screens in accelerated human 684 stem cell-derived neural progenitor cells identify Zika virus host factors and drivers of 685 proliferation

IFITM proteins inhibit entry driven by the 687 MERS-Coronavirus Spike protein: Evidence for Cholesterol-Independent Mechanisms

Camera: a competitive gene set test accounting for inter-gene 690 correlation

Identification of small-molecule inhibitors of Zika virus 693 infection and induced neural cell death via a drug repurposing screen

Interferon-induced transmembrane protein-3 genetic variant rs12252-C is associated 697 with disease severity in COVID-19

Interferon-induced transmembrane protein-3 genetic variant rs12252-700 C is associated with severe influenza in Chinese individuals

Interferon 702 induction of IFITM proteins promotes infection by human coronavirus OC43

High-Content Screening in hPSC-705

Neural Progenitors Identifies Drug Candidates that Inhibit Zika Virus Infection in Fetal-like 706 Organoids and Adult Brain

Zika Virus Targets Glioblastoma Stem Cells through 709 a SOX2-Integrin αvβ5 Axis

EVs Containing Host Restriction

Factor IFITM3 Inhibited ZIKV Infection of Fetuses in Pregnant Mice through Trans-placenta 712 Delivery

Non-targeting (NT) controls (pink, light blue dots) are also shown. C-F, Validation of H1-Cas9 stem cell line. (C) Diagram of AAVS1 targeting vector and site of genomic integration. (D) Immunostaining using an anti-FLAG antibody for clone H1-36-23. Scale bar = 200 µM. (E) PCR across the junctions for 4 clones, indicating proper targeting into one allele. (F) Fraction indels in neurons differentiated from H1-36-23 clone and infected with lenti-gRNAs, measured by nextgeneration sequencing. G-I, Confirmation of primary screen hits. H1-Cas9 SNaPs were transduced with individual gRNAs, expanded, and exposed to ZIKV-Ug (MOI = 1)