Science and Nature

Purposeful single-cell genomics of human cytomegalovirus an infection

Summary

Conception how viral and host factors work collectively and the way in which perturbations impact an infection is the muse for designing antiviral interventions. Here we define the functional contribution of every and each viral and host factor interested by human cytomegalovirus an infection in necessary human fibroblasts by way of pooled CRISPR interference and nuclease screening. To resolve how genetic perturbation of essential host and viral factors alters the timing, route and development of an infection, we applied Perturb-seq to sage the transcriptomes of tens of thousands of CRISPR-modified single cells and came across that, on the general, most cells apply a stereotypical transcriptional trajectory. Perturbing essential host factors does no longer commerce the stereotypical transcriptional trajectory per se however can stall, prolong or skedaddle development alongside the trajectory, allowing one to pinpoint the stage of an infection at which host factors act. Conversely, perturbation of viral factors can put definite, abortive trajectories. Our outcomes give an explanation for the roles of host and viral factors and present a roadmap for the dissection of host–pathogen interactions.

Significant

The β-herpesvirus human cytomegalovirus (HCMV) is a pervasive pathogen that establishes lifelong an infection in the bulk of the human population. Activation of its lytic cycle triggers a attribute cascade of events, starting with stereotypical waves of viral gene expression, continuing with the replication of its ~235-kb double-strand DNA genome and culminating in the budding of viral progeny1. Quite a bit of systems-stage be taught hold described these phenomena at the stage of the transcriptome, the feature of translated messages and the proteome in time and residence2,3,4,5,6,7, generally learning an infection in fibroblasts. HCMV has a vast tropism in its human host, however fibroblasts are predominant in the pool of contaminated cells in vivo and had been broadly used for propagation in cell custom8. A core feature of viral genes predominant for replication used to be established by systematic mutagenesis9,10. These be taught highlighted the complexity of a entire bunch of viral genes cooperating to manipulate the host and undermine its protection machinery, raising the query of of what primarily the most efficient targets are for antiviral intervention.

CRISPR–Cas9 technology gives tools to systematically measure the functional contribution of every and each viral gene and host factor interested by productive infections11. On the assorted hand, it remains a area to translate a list of factors staunch into a systematic knowing of their particular person roles and group in pathways. We take care of this area with Perturb-seq, a scalable skill that mixes CRISPR-primarily based fully genetic perturbations with properly to place phenotypic profiling by single-cell transcriptomics. First, we performed systematic pooled CRISPR monitors for each and each host and viral factors affecting survival of necessary human fibroblasts upon HCMV an infection. Survival monitors are scalable to genome-vast libraries however employ fully a compressed image of the molecular events unfolding upon perturbation of host or viral factors. Furthermore, lytic an infection is inherently dynamic over time and heterogeneous from cell to cell12,13. We, this skill that truth, recorded the transcriptomes of tens of thousands of single cells and monitored how perturbation of a feature of essential factors—identified in the pooled monitors—alters the timing, route and development of an infection.

Our records paint a excessive-resolution image of the HCMV lytic cycle as a deterministic program that is distinctly inclined to host- and virus-directed interventions. We name host restriction factors as properly as dependency factors essential for viral entry and for development from early to dull stages of an infection. Conversely, we order that focusing on key viral factors derails the viral gene expression program in specific ways. Taken collectively, our findings give an explanation for a dichotomy between the roles of host and viral factors, with the feature of viral factors fully defining the trajectory of an infection and host factors growing the ambiance permitting the execution of that program.

Outcomes

High-resolution functional scanning of the HCMV genome

Cas9 is an efficient tool for making targeted disruptions in the cytomegalovirus genome. Focusing on particular person predominant herpesvirus genes by CRISPR–Cas9 used to be confirmed to disrupt their expression directly and their feature by way of errors presented by the host DNA restore machinery14. Cleavage of the viral DNA in non-predominant areas has minimal form on HCMV replication and host cell viability—seemingly due to DNA restore is snappy relative to replication—however can hold an tag on expression of genes proximal to the lower internet sites (sight ref. 14 and our records below). To enable excessive-resolution scanning of viral components for a complete functional annotation of the HCMV genome, we designed a CRISPR library that targets each and each protospacer-adjoining motif (PAM) for Streptococcus pyogenes Cas9 (NGG PAM sequence present roughly each and each 8 bp) alongside the genome of the scientific stress Merlin (Fig. 1a and Supplementary Desk 1). We delivered the library into necessary human fibroblasts engineered to specific Cas9, in inform that, upon an infection, each and each cell executes a lower at an outlined recount alongside the viral genome, collectively tiling its entirety.

Fig. 1: Virus-directed CRISPR nuclease show conceal conceal maps the phenotypic panorama of the HCMV genome.
figure1

a, Experimental execute for pooled, virus-directed CRISPR screening. Our HCMV tiling library comprises ~33,000 components and used to be lentivirally delivered into necessary human foreskin fibroblasts expressing the CRISPRn machinery, followed by an infection with HCMV. sgRNA cassettes had been quantified by deep sequencing in the initial (t0) population and the surviving population. b, Phenotypic panorama of the HCMV genome got by locally averaging the phenotypes of particular person sgRNAs alongside the genome and across two biological replicates. Strong adjustments in the magnitude of the phenotype coincide with gene–gene boundaries (inset).

We mapped the functional panorama of the HCMV genome by quantifying the abundance of particular person single records RNA (sgRNA) cassettes in a population earlier than/after an infection (Prolonged Data Fig. 1). We came across that reducing phenotypes are slightly constant inside of particular person genes, indicating that the determining factor is which gene is targeted, reasonably than the put Cas9 cuts at some level of the gene body. Cutting adjoining sets of genes steadily showed identical phenotypes. On the assorted hand, some gene boundaries had been marked by abrupt phenotype adjustments, arguing that order penalties of Cas9 cuts are restricted to their instantaneous vicinity (Fig. 1b, Prolonged Data Fig. 1 and Supplementary Desk 1).

At a elevated scale, adjustments in the direction and magnitude of the phenotypes outlined six necessary genomic modules. Cuts in each and each distal areas of the genome, which lack genes predominant for viral replication9,10, had minimal form on host cell survival. As expected, focusing on the areas overlaying UL48A–UL73 and UL96–UL150, each and each of which have predominant genes interested by viral DNA replication, packaging and nuclear egress10,15, strongly safe contaminated cells. In the two ultimate areas, we came across that disruption of genes required for viral replication did no longer primarily give protection to the host. Cuts at some level of the UL32–UL47 space, which comprises predominant genes, the truth is led to a strongly elevated skill of the virus to abolish cells. Essentially the most strongly sensitizing phenotypes mapped to the identified viral apoptosis inhibitors UL36, UL37 and UL38 (ref. 16). Despite the true fact that this behavior could well furthermore also be rationalized for virally encoded anti-apoptotic proteins, it extended to many other virus-predominant genes with out identified anti-apoptotic roles, in conjunction with the DNA polymerase processivity factor UL44. At final, and counterintuitively, cuts in the central space spanning UL75–UL88 triggered very gentle phenotypes. Many genes in this space encode predominant components of the viral envelope, tegument and capsid, but the outcomes on host survival after Cas9 reducing had been much like focusing on the US distal space, ensuing in even mildly enhanced cell death upon an infection.

Focusing on predominant viral genes, by definition, undermines the production of viral offspring. It does no longer primarily concomitantly give protection to the contaminated cell, underscoring that a success viral replication and death of the contaminated cell are definite phenomena. Evidently disrupting predominant genes interested by viral DNA replication largely protects the host. On the assorted hand, interfering with the later steps of assembling new virions could well perchance no longer fully be ineffective in preserving the host however could well perchance even recount an additional burden.

Genome-vast monitors for host factors of HCMV

Subsequent, we performed a pooled show conceal conceal for host factors of HCMV an infection by systematically repressing host genes by CRISPR interference (CRISPRi)17,18. Phenotypes had been outlined by enrichment or depletion of sgRNA cassettes earlier than/after an infection and a mock-contaminated hold a watch on population (Fig. 2a and Supplementary Desk 2).

Fig. 2: Host-directed CRISPR monitors name host dependency and restriction factors.
figure2

a, Experimental execute for pooled, host-directed CRISPRi and CRISPRn screening. Genome-vast sgRNA libraries focusing on human genes with more than one sgRNAs each and each had been lentivirally delivered into necessary human foreskin fibroblasts expressing the CRISPRi or CRISPRn machineries, followed by an infection with HCMV. sgRNA cassettes had been quantified by deep sequencing in the initial (t0) population, the surviving population and an uninfected hold a watch on population to tale for gene essentiality in the absence of an infection. b, Outcomes of the host-directed CRISPRi monitors displayed as a scatter recount of average gene essentiality (that is, an infection-fair phenotype; y axis) versus protection/sensitization to death upon HCMV an infection (that is, an infection-dependent phenotype; x axis) averaged across two biological replicates. Highlighted are hit genes representing selected pathways. HAT, histone acetyltransferase.

Our show conceal conceal printed a various feature of host factors (Fig. 2b). Genes inquisitive relating to the biosynthesis of heparan sulfate had been amongst the strongest protective hits. Heparan sulfate proteoglycans on the cell ground enable viral cell adhesion19,20. Furthermore, we came across vesicle trafficking factors: RAB6A and its GEFs RIC1/KIAA1432 and RGP1, the conserved oligomeric Golgi (COG) advanced, individuals of TRAPP advanced III and UNC50. These factors converge on the Golgi equipment and mediate retrograde and anterograde transport. Just a few of them (COG, TRAPP and UNC50) had been beforehand implicated in the internalization of various bacterial and plant toxins, suggesting that HCMV and toxins exploit identical pathways for cell entry17,21,22,23,24,25.

Various protective hits included individuals of the LAMTOR/Ragulator advanced, folliculin (FLCN) and the lyspersin (C17orf59) subunit of the BORC advanced, all linked to lysosome positioning and nutrient sensing26,27,28. This supports the contemporary commentary that HCMV an infection adjustments lysosome dynamics7. Furthermore, host cell death used to be diminished by knockdown of cullin-RING E3 ligases (RBX1 and CUL3), their adaptor subunits (DDA1 and TCEB2/ELOB), substrate receptors (DCAF4) and the connected neddylation (NEDD8, NAE1 and RBX1) and deneddylation (COP9 signalosome) machineries. Many viruses hijack this pathway to degrade host restriction factors, that will furthermore be steer clear off by broadly acting Nedd8-activating enzyme inhibitors29,30. At final, we identified genes interested by tail-anchored protein insertion into the endoplasmic reticulum (ER) as properly as ER-connected degradation: AMFR, an E3 ligase, and the TRC40/GET pathway individuals BAG6 and ASNA1, which had been confirmed to be required for insertion of membrane proteins of herpes simplex virus 1 (HSV-1)31.

Our monitors furthermore identified genes whose knockdown sensitizes cells to death upon an infection—as an instance, identified restriction factors similar to PML and DAXX and individuals of the interferon (IFN) form I pathway. Despite the true fact that restriction factors are strictly outlined by their form on battling viral replication, our findings indicate that screening for genes that sensitize the host to cell death when knocked down is a viable skill for their identification. Sensitizing hits included subunits of the NuA4 histone acetyltransferase advanced, which used to be confirmed to counteract hepatitis B virus replication by repressing transcription from chromatinized viral DNA32 and used to be described as an interactor of the HIV-1 TAT protein33. Furthermore, we identified individuals of the signal recognition particle, the translocon and connected factors as properly as genes interested by ER stress34,35. At final, we came across genes with anti-apoptotic feature, in conjunction with various caspases, whose knockdown seemingly increases the sensitivity to apoptosis triggered by HCMV an infection.

A contemporary peek reported PDGFRA because the dominant hit in a CRISPR nuclease (CRISPRn) knockout show conceal conceal designed to name host factors required for HCMV entry36, underscoring its reported role because the receptor on the bottom of fibroblasts exploited by HCMV lines expressing the trimeric virion glycoprotein advanced37,38,39. In our CRISPRi show conceal conceal, PDGFRA knockdown conferred gentle protection from cell death upon an infection (Fig. 2b).

To validate and lengthen the host factors identified by CRISPRi, we performed a knockout show conceal conceal under the same prerequisites, the usage of a longtime CRISPR reducing library40 (Prolonged Data Fig. 2a). CRISPRi and CRISPRn screening identified identical numbers of protective hits, of which around 20% scored as necessary in each and each screening modes (Supplementary Desk 2). We came across elegant settlement at the pathway stage although gene-stage phenotypes various (Prolonged Data Fig. 2b). Of show conceal, protective hits in the knockout show conceal conceal had been dominated by genes interested by virus entry, similar to PDGFRA and heparan sulfate biosynthesis genes. This underscores that genes with very valid protective phenotypes are more readily identified in knockout monitors, seemingly due to chance rigidity can act more strongly on cells with factual null alleles. Overall, our findings emphasize the advantages of combining orthogonal modes of genetic screening41.

The lytic cascade resolved by single-cell transcriptomics

Our pooled monitors present a genome-scale image of the factors interested by lytic HCMV an infection, however inserting them into biological pathways and linking them to a stage of the viral lifestyles cycle requires prior records or dedicated apply-up experiments. To study the roles of essential host and viral factors systematically in more depth, we used Perturb-seq, which mixes CRISPR-primarily based fully genetic perturbations with a properly to place single-cell transcriptional readout34,42,43,44. Measuring tens of thousands of single-cell transcriptomes from a population with a library of genetic perturbations gives a hugely parallel way of assessing the pause result of these perturbations under uniform prerequisites. The single-cell nature of this implies makes it particularly acceptable for learning viral an infection, a route of with immense inherent variability from cell to cell12,13,45,46,47,48,49.

Because the groundwork for the Perturb-seq prognosis, we first explored the development of HCMV an infection by recording single-cell transcriptomes of 12,919 genetically unperturbed cells, sampled from eight time aspects with two multiplicities of an infection (MOIs) each and each (Fig. 3a,b). In preference to relying totally on synchronizing cells experimentally, which has inherent limits as a result of intrinsic heterogeneity in the timing of initial an infection and its development, we staged cells computationally by their transcriptional signatures. The ultimate sources of variability between cells had been the extent of IFN signaling and the allotment of viral RNA per cell (‘viral load’), which reached phases of around 75% (Fig. 3c,d and Supplementary Desk 3). Cells with excessive viral loads showed a marked make bigger in their complete noticed mRNA molecules (that is, strange molecular identifier (UMI) counts) per cell. This indicates that the eponymous make bigger in cell dimension at some level of an infection (cyto megalo—neat cell) is reflected in a elevated cell RNA converse material (Prolonged Data Fig. 3a,b). Collectively, these properties define three predominant subpopulations of cells: a naive population (uninfected, IFN-detrimental); a bystander population (no longer expressing viral genes, IFN-sure); and an internally heterogeneous contaminated population with a variety of portions of viral transcripts, which we divided into more than one subclusters (Fig. 3e and Prolonged Data Fig. 3b). The distribution of cell cycle phases used to be altered between clusters (Prolonged Data Fig. 3c), primarily based fully on the properly-established G1 arrest early at some level of an infection50 and a subsequent shift toward a recount such as S-allotment in cells with very excessive viral loads.

Fig. 3: Single-cell an infection time route defines the lytic cascade of expression events as a trajectory in gene expression residence.
figure3

a, Cells had been contaminated with a low (0.5) or excessive (5.0) MOI of HCMV, harvested once in a while ranging from 6 to 120 h.p.i., pooled and subjected to emulsion-primarily based fully single-cell RNA sequencing (scRNA-seq). b–e, UMAP projections of the transcriptomes of 12,919 single cells (each and each host and viral transcripts), color-coded by experimental time after an infection (b), proportion of viral transcripts per cell (c), IFN salvage, calculated from the normalized expression of ISGs (d), and cluster membership (e). f, Share of cells in each and each cluster as a feature of time after an infection and MOI. g, Expression dot recount of purchase viral and host genes in numerous clusters. Sizes of dots checklist the expression normalized to the cluster with the perfect expression for each and each gene. Colors checklist absolute expression phases (scaled UMIs per gene per cell, averaged for all cells of a cluster). lncRNA, long noncoding RNA.

Each and each cluster contained cells from each and each the low- and excessive-MOI samples, and the gene expression patterns between these groups of cells had been extraordinarily extremely correlated (Prolonged Data Fig. 3d). This highlights each and each the supreme technical reproducibility of our transcriptomics workflow and that the MOI determines the population-stage response (that is, the allotment of cells at a given stage of an infection) reasonably than the character of the transcriptional responses in particular person cells.

Apparently, viral gene expression and the expression of interferon-stimulated genes (ISGs) had been nearly entirely mutually abnormal, a phenomenon that has been noticed for HSV-1 (ref. 46). Cells with excessive viral loads showed entirely suppressed IFN signaling. Most efficient cells in cluster ‘contaminated 1’ (Prolonged Data Fig. 3b) expressed each and each ISGs and low portions of viral RNA, mainly classic instantaneous-early genes similar to UL123 (IE1) (Fig. 3g), indicating that these are cells in the earliest stage of an infection. Collectively, this underscores the rapidity with which the virus effectively suppresses IFN signaling and highlights the vitality of single-cell analyses in resolving this phenomenon, which could well seem as concurrent expression of viral genes with ISGs in bulk measurements.

We subsequent tracked how the cell population is disbursed to different clusters as a feature of time and MOI (Fig. 3f and Supplementary Desk 3). All cells are at the origin in the naive cluster. At 6 hours put up an infection (h.p.i.), cells originate to transition to either the bystander or contaminated clusters, and, at 20 h.p.i., nearly no naive cells stay. The ratio of contaminated to bystander cells relies on the initial MOI and stays slightly constant between 20 and 48 h.p.i., with contaminated cells progressing to clusters with elevated viral load. After 48 h.p.i., we detected but some other marked make bigger in the series of contaminated cells and a concomitant decrease in bystander cells, much like a second wave of an infection, seemingly triggered by virions launched from cells that had been contaminated early. By 96 h.p.i., even the population with low initial MOI is nearly fully contaminated.

Among the many contaminated clusters, most cells apply a dominant trajectory with increasing viral load (clusters ‘contaminated 1–6’), and viral marker genes for specific phases height in outlined clusters (Fig. 3g and Supplementary Desk 3). To define the viral gene expression patterns alongside this trajectory at excessive resolution, we grouped cells from these clusters into slim viral load bins and determined the profiles of all robustly quantified viral genes (Prolonged Data Fig. 4a). Many genes uncover stereotypic kinetics, similar to instantaneous-early (UL123, US10 and US2) or factual-dull (UL99 and UL100, amongst others) genes. On the assorted hand, our excessive-resolution skill printed that the (pseudo)temporal patterns of many genes had been subtly definite from one but some other, and plenty did no longer align with the canonical temporal patterns. For instance, US6, UL78, US26, UL42 and US34 showed bimodal kinetics, and expression of UL4 and UL48A peaked mid-an infection.

A exiguous however prominent feature of host transcripts had been upregulated with increasing viral load, such as ‘leaky-dull’ or ‘factual-dull’ viral gene kinetics (Fig. 3d): APOE, CD55, TSPAN13, HES6, SPINT2 and PCSK5. Upregulation of CD55 has furthermore been noticed at the protein stage5, and it used to be confirmed to be integrated into budding virions to counteract the complement arrangement51.

A exiguous subpopulation, ~1% of contaminated cells, did no longer apply the dominant trajectory however diverted from contaminated cluster 2, following an off-trajectory the put cells attain excessive viral loads in a definite space of gene expression residence (Fig. 3c). Cells in that cluster had been outlined by lower UMI counts, suggesting no make bigger in cell dimension (Prolonged Data Fig. 3b). Their sample of viral gene expression used to be markedly different (Prolonged Data Fig. 4b and Supplementary Desk 3). Rapid-early and delayed-early genes had been strongly overexpressed, whereas factual-dull genes and all long noncoding RNAs had been strongly depleted. As factual-dull gene expression relies on genome replication52, we form that this trajectory is abortive. Attempting at host transcripts, cells on the abortive trajectory had been characterized by lack of suppression of the IFN response and had been the fully ones expressing interferon-β (IFNB1) itself, alongside with other cytokines and plenty stress response genes, prominently from the NF-κB pathway (NFKBIA and RELB) as properly as JUN. This sample used to be paying homage to the discovering that HSV-1 initiates pathways upstream of IFN production in a uncommon subset of abortively contaminated cells46. Collectively, our findings indicate that, equally, a exiguous subpopulation of cells on an abortive viral trajectory actively secrete interferon-β whereas escaping suppression of the downstream response to interferon, perchance though-provoking autocrine solutions loops as properly as to paracrine signaling.

Host perturbations can block, slack or skedaddle an infection

We subsequent performed a series of Perturb-seq experiments exploring the form of focusing on host and viral factors on the viral lifestyles cycle. In disagreement to the pooled show conceal conceal, the put phenotypes emerge by enrichment or depletion of cells over more than one days, Perturb-seq gives a excessive-resolution peek of the form of focusing on an predominant gene over the first 72 h.p.i., overlaying roughly one viral replication cycle.

We first selected 52 host genes identified as hits in the pooled monitors, cloned them staunch into a targeted library, alongside with non-focusing on hold a watch on sgRNAs, and delivered the library staunch into a population of fibroblasts expressing the CRISPRi machinery (Fig. 4a,b and Supplementary Desk 4). We challenged that population with an MOI of HCMV of 0.5 for 1 h, followed by a medium commerce, and monitored the implications of the genetic perturbations in a median of 165 ± 50 cells per goal per time level (Prolonged Data Fig. 5a). CRISPRi diminished expression of the host targets by a median of 87% (Prolonged Data Fig. 5b) and triggered goal-specific transcriptional responses (Prolonged Data Fig. 5c). In uninfected cells, we noticed the strongest transcriptional responses after knockdown of IFN pathway individuals, LAMTOR/Ragulator subunits and the cullin-RING/neddylation machinery, as properly as gentle responses to the knockdown of vesicle trafficking factors. The patterns of the transcriptional responses to knockdown organized host factors by pathway in a principled kind (Prolonged Data Fig. 5c), providing a layer of records that continues to be unresolved by pooled screening.

Fig. 4: Perturbing host factors can alter the propensity of a cell to be contaminated.
figure4

a, Host dependency and restriction factors had been selected from the pooled show conceal conceal, cloned staunch into a Perturb-seq library and delivered into dCas9-expressing fibroblasts, which had been challenged with an MOI of 0.5 of HCMV for 24–72 h. b, Chosen host factors hold a vast series of sensitizing to protective phenotypes, hold a variety of degrees of essentiality and quilt different pathways. c, UMAP projection of the transcriptomes of 35,601 cells with confidently identified sgRNAs presentations the same naive, bystander and productively and abortively contaminated clusters came across in the unperturbed an infection time route (Fig. 3). d, Cluster membership as a feature of time after an infection for cells expressing non-focusing on hold a watch on sgRNAs (sgControl), as properly as sgRNAs focusing on UBA3, EXT2 and IFNAR2, as representatives for different sorts of responses. For a entire feature of cluster membership graphs, sight Prolonged Data Fig. 5h. e, UMAP illustration of the cluster membership records (Prolonged Data Fig. 5h) organizes host factors by their phenotypes of altered development of an infection in single cells, spanning cell entry phenotypes, partial protection from an infection, shut to-normal development and accelerated development of an infection.

We noticed the same split of the population staunch into a naive, a bystander and an contaminated cluster, branching staunch into a productive and an abortive trajectory (Fig. 4c and Prolonged Data Fig. 5d–f). Cells from the uninfected (0 h) sample had been overwhelmingly in the naive cluster and transitioned into the bystander and contaminated clusters, starting at 24 h.p.i. (Fig. 4d, Prolonged Data Fig. 5h and Supplementary Desk 4). The kinetics of transition of cells between the clusters used to be markedly different in cells with definite host factor knockdowns when when in contrast with hold a watch on cells. Focusing on individuals of the heparan sulfate biosynthesis pathway, similar to EXT2, COG advanced individuals, the KIAA1432/RIC1–RGP1 advanced and other trafficking factors similar to UNC50 efficiently steer clear off an infection. Cells lacking the NEDD8-activating enzyme subunits UBA3 and NAE1, as properly as LAMTOR advanced individuals, was contaminated however in lowered numbers. Conversely, focusing on SEC61B, a non-predominant subunit of the translocon, elevated the numbers of contaminated cells at 24 h. In a similar kind, focusing on the interferon receptor subunit IFNAR2, or its downstream effectors STAT2 and IRF9, elevated an infection rates early. Furthermore, cells with these perturbations failed to mount the interferon-driven transcriptional response attribute of bystander cells and remained transcriptionally naive so long as they stayed uninfected. These cells saved getting contaminated at elevated rates, especially at later time aspects, when most other cells showed a sturdy interferon response. For a systematic classification of host targets by their development phenotypes, we performed dimensionality reduction of the temporal cluster membership records (Prolonged Data Fig. 5h), organizing different host factors by phenotype on a spectrum ranging from cell entry defects to accelerated development (Fig. 4e).

Subsequent, we extended our Perturb-seq prognosis to viral factors, the usage of CRISPRn because the mode of genetic perturbation. We reasoned that, when focusing on a viral factor, fully contaminated cells would elevate meaningful records. We, this skill that truth, challenged the cells with a excessive MOI of 5.0 of HCMV, with out removing the inoculum, to maximise the proportion of contaminated cells (Fig. 5a). We selected 31 viral gene targets in step with their valid phenotypes in the pooled, virus-directed monitors. Furthermore, we added knockout guides focusing on a consultant feature of 21 host factors as properly as safe-focusing on guides focusing on non-predominant areas of the human and HCMV genomes (Fig. 5b and Supplementary Desk 5). We recovered 188 ± 77 cells per goal per time level (Prolonged Data Fig. 6a). Our experimental prerequisites resulted in more than 50% contaminated cells at 24 h.p.i. and elegant illustration of cells in numerous contaminated subclusters (Fig. 5c).

Fig. 5: Host- and virus-directed perturbations stall or skedaddle development or shift the patterns of viral gene expression.
figure5

a, Host and viral factors had been selected from the pooled monitors, cloned staunch into a Perturb-seq library and delivered into Cas9-expressing fibroblasts, which had been challenged with an MOI of 5.0 of HCMV for 24–72 h. b, Chosen factors organized by their respective phenotypes in the pooled monitors, essentiality for the host (determined by the uninfected arm of our pooled show conceal conceal) and the virus10, and pathway membership. IE, instantaneous early. c, UMAP projection of the transcriptomes of 39,297 cells with confidently identified sgRNAs presentations the same naive, bystander and productively and abortively contaminated clusters came across in the unperturbed an infection time route (Fig. 3) and the host-directed Perturb-seq dataset (Fig. 4). d, Cluster membership as a feature of time after an infection for cells expressing sgRNAs focusing on two host factors (UNC50 and FLCN), a valid-goal space of the viral genome (sgControl) and three viral factors (UL105, UL42 and UL69), as representatives for different sorts of responses. For a entire feature of cluster membership graphs, sight Prolonged Data Fig. 6c,d. e, UMAP illustration of the cluster membership records (Prolonged Data Fig. 6c,d) organizes host and viral factors by their phenotypes of altered development of an infection in single cells, spanning cell entry phenotypes, partial protection from an infection, shut to-normal development and accelerated development of an infection as properly as patterns specific to viral targets.

The event of an infection all over again various widely looking on the targeted gene, visualized by the distributions of viral loads for each and each goal (Prolonged Data Fig. 6b) and cluster membership of cells with a given goal (Fig. 5c,d, Prolonged Data Fig. 6c,d and Supplementary Desk 5). Host-directed knockouts confirmed our findings from the host-directed CRISPRi experiment. The excessive percentages of contaminated cells improved the resolution of some protective phenotypes, distinguishing two scenarios: a diminished propensity of a cell to be contaminated versus delays in development from the early to later stages of an infection (Prolonged Data Fig. 6c). For instance, knockout of PDGFRA (the proposed viral receptor on fibroblasts38) or HS6ST1 (interested by heparan sulfate biosynthesis) nearly entirely steer clear off an infection, even with excessive MOI. Equal phases of protection had been noticed in cells the put COG8 or UNC50 used to be knocked out, implicating these factors in viral entry. Conversely, perturbation of FLCN, LAMTOR2/3, KIAA1432/RIC1 and RGP1 permitted an infection (albeit at diminished phases) however severely slowed the development of an infection to the dull stage, indicating that these factors are predominant in early-stage an infection, acting downstream of viral uncoating however earlier than genome replication. Handbook examples for each and each development phenotype are confirmed in Fig. 5d.

Virus-directed perturbations alter the an infection trajectory

When put next with host genes, focusing on viral genes led to qualitatively different outcomes (Fig. 5d,e and Prolonged Data Fig. 6d). Expectedly, cells with virus-focusing on sgRNAs in total had equal propensities to alter into contaminated (Prolonged Data Fig. 6d), with the exception of cells focusing on UL122 and UL123, which gave the influence to hold slightly diminished propensities. This discovering confirms the identified roles of these two genes in initiating instantaneous-early viral gene expression, which, when suppressed, can make an contaminated cell present as uninfected in gene expression residence.

As soon as contaminated, cells with virus-focusing on sgRNAs progressed in ways specific to the goal gene, evident by more advanced patterns of viral load distributions (Prolonged Data Fig. 6b) and, consequently, of their development by way of different clusters (Prolonged Data Fig. 6d and Fig. 5d). This commentary printed that focusing on a viral factor does no longer good stall an infection at the stage the put that factor turns into predominant, as is the case with host factors. Moderately, cells with virus-focusing on guides remove definite areas in transcriptional residence however fully in contaminated cells (Prolonged Data Fig. 6e–g). Collectively, these findings brought on us to peek in more factor the patterns of viral gene expression in contaminated cells.

On a dimensionality-diminished projection of the viral transcriptomes in contaminated cells, the route of an infection could well furthermore also be visualized as a trajectory by a rolling average of the positions of cells with increasing viral load (Fig. 6a). Cells with host-focusing on sgRNAs all followed trajectories that are merely about congruent with the default trajectory (outlined by cells with hold a watch on sgRNAs). On the assorted hand, some host factor trajectories had been primarily shorter due to some perturbations preclude cells from reaching dull-stage an infection (Fig. 6b,c). In marked disagreement, cells with virus-focusing on sgRNAs followed trajectories that diverged from the default (Fig. 6d,e). These differences had been driven by viral expression patterns and no longer by concomitant diversifications in host gene expression.

Fig. 6: Virus-directed perturbations put more than just a few trajectories in viral gene expression residence.
figure6

a, UMAP projection of the viral components of the transcriptomes of 17,511 cells with >2.5% viral RNA, color-coded by the allotment of viral RNA per cell. b, Subsets of cells with host-directed sgRNAs, color-coded by records id. c, Trajectories of an infection for host-directed perturbations, determined by averaging the geometric recount of cells with a given sgRNA goal, ranked by viral load. d, Subsets of cells with virus-directed sgRNAs, color-coded by records id. e, Trajectories of an infection for virus-directed perturbations. f, UMAP illustration of different trajectories for each and each host- and virus-directed sgRNA goal, calculated by dimensionality reduction of the viral gene expression matrices for all virus- and host-directed perturbations confirmed in Prolonged Data Fig. 7a. All host-directed perturbations (gloomy plight) discontinue in trajectories that are most much just like the hold a watch on trajectories outlined from each and each the host and virus safe-focusing on controls. g, Pairwise correlation matrix of the relative viral gene expression matrices for different trajectories highlights the three predominant bundles of trajectories generated by virus-directed perturbations.

To peek the character of these alternate trajectories, we quantified the expression of all viral genes alongside each and each trajectory relative to the default trajectory. Prolonged Data Fig. 7a outlines this analytical workflow. Cells with sgRNAs focusing on non-predominant areas of the viral genome followed a trajectory nearly unchanged from these with host-focusing on controls, exhibiting fully gentle transcriptional outcomes on genes in the instantaneous vicinity of the lower internet sites, inside of ~10 kb (Prolonged Data Fig. 7b and Supplementary Desk 5). Diminished expression no longer good of the goal gene itself however furthermore of genes situated shut to the lower internet sites used to be a total feature for all virus-focusing on sgRNAs. Moreover to, for all virus-focusing on sgRNAs as antagonistic to our safe-focusing on controls, we noticed in kind upregulation or downregulation of genes in trans—that is, genes encoded removed from the sgRNA goal—indicating oblique outcomes on viral gene regulation triggered by disruption of the goal gene.

When evaluating the similarities of these trajectories, we noticed that they came in ‘bundles’ (Fig. 6e) and, particularly, that focusing on genes in the same space of the viral genome tended to discontinue in identical patterns of deregulated viral gene expression (Prolonged Data Fig. 7b,c). This applies each and each to the implications on genes in cis, which used to be expected, and for the expression adjustments of genes in trans, which indicates that genes are organized in the viral genome in functional modules. To hold the connection of the trajectories triggered by focusing on different genes, we projected the deviations in gene expression into two dimensions by uniform manifold approximation and projection (UMAP) (Fig. 6f) and quantified how correlated viral gene expression adjustments had been for different trajectories (Fig. 6g). Despite the true fact that all trajectories generated by focusing on host factors had been quantitatively much just like the default trajectory, viral trajectories came in three predominant classes. The predominant class of trajectories resulted from focusing on genes in the sensitizing space (UL35–UL43). These perturbations triggered diminished expression of RNA1.2 and RNA2.7, blended with overexpression to a variety of degrees of nearly all viral genes encoded downstream of UL48, with US3 and UL54 being amongst primarily the most strongly overexpressed. Cells following these trajectories once in a while reached very excessive viral loads, and illustration of these targets dropped severely in the later time aspects (Prolonged Data Fig. 6a), underscoring that these viral trajectories coincide with the cells present process apoptosis.

A second trajectory class used to be linked to perturbations of genes in one among the protective genomic modules (UL52–ORFL152C) as properly as to UL102 and UL105, that could be found around 50 kb downstream. These genes are all inquisitive relating to the viral replication machinery. Consequently, cells following these trajectories furthermore failed to achieve excessive viral loads and underexpressed dull viral genes, indicating that viral genome replication fails upon focusing on these factors.

A third trajectory class corresponded to perturbations of genes situated at some level of the UL115–UL148 space. Here, viral gene expression patterns had been most much just like the unperturbed trajectory. Of show conceal, the kinetics of development various amongst the targeted genes in this workforce (Fig. 5c, accurate panel), with perturbation of the necessary instantaneous-early transactivator genes UL122 and UL123 causing the strongest delays.

Notably, two viral gene targets led to trajectories of an infection that had been definite from one but some other and from viral targets in their instantaneous genomic vicinity: UL69 and UL112 (ORFL253W) (Prolonged Data Fig. 7b). Each and each genes showed slightly feeble protective phenotypes when targeted in the pooled tiling show conceal conceal (Prolonged Data Fig. 1b). Furthermore, each and each the UL112 and UL69 loci encode microRNAs, that are, nonetheless, thought to goal host reasonably than viral genes53.

UL69 has been described as a transactivator of gene expression54 and as interested by selling nuclear export of unspliced RNA55. Focusing on UL69 triggered a definite viral gene expression sample, in conjunction with downregulation of RL12, RL13, UL144 and UL155 and exiguous overexpression of the noncoding RNA1.2.

The UL112 locus offers upward push to more than one gene merchandise by more than just a few splicing, all of that are seemingly disrupted with Cas9. Some splice variants are interested by recruitment of the viral polymerase activator UL44 to nuclear replication internet sites56. Focusing on the UL112 locus triggered a viral gene expression sample that entails overexpression of US3 as properly as upregulation of genes in the 5′ space of the genome, similar to of UL22A and UL38. This sample bears some resemblance to the sample triggered by focusing on genes in the UL35–UL43 module, which sensitizes cells to cell death, whereas focusing on UL112 is protective.

Taken collectively, our outcomes order that the trajectory of an infection could well furthermore also be derailed in outlined ways by focusing on viral genes, whereas focusing on host factors determines the tempo of development alongside the default trajectory. This implies that host factors put a permissive ambiance, whereas viral genes fully orchestrate and hold a watch on the viral gene expression program.

Dialogue

The waves of viral gene expression at some level of lytic an infection are a key signature of herpesvirus biology1. Our peek redefines the lytic cascade at the one-cell stage as a continuum of cell states. We came across that the neat majority of cells apply this stereotypical trajectory in gene expression residence, whereas a exiguous however prominent subpopulation grasp an more than just a few, abortive trajectory.

Technologically, our peek breaks ground on various phases, in conjunction with the usage of saturating functional monitors of a neat viral genome, the total discovery of essential sets of genes on each and each side of a host–pathogen arrangement and the usage of single-cell analyses to achieve the functional penalties of focusing on these factors. Our work establishes Perturb-seq as a robust skill for functional genomics in a extremely dynamic virus–host arrangement. The single-cell skill is paramount to each and each taking pictures the inherent cell-to-cell heterogeneity of an infection and enabling a extremely multiplexed, functional prognosis of genetic perturbations.

In accordance with our excessive-dimensional dataset, we organized host factors each and each by the transcriptional responses in cells the put these factors are targeted and by how an infection progresses in these cells. This gives a systematic classification of host factors by functional category that displays a ramification of factors acting in viral entry, in early-stage an infection and as restriction factors. Finding out out genetic perturbation phenotypes as properly to place transcriptional signatures furthermore printed that, by focusing on viral factors, contaminated cells progress alongside trajectories in expression residence that are each and each non-productive and no longer encountered on the general.

HCMV is entirely relying on the transcriptional and translational machinery of its host. At the same time, our findings order that, in fibroblasts, the lytic cascade, once in motion, is a deterministic program that is hard-wired into the viral genome reasonably than reactive to the recount of the cell. HCMV aspects a vast cell tropism in its human host, and it harnesses different receptors and entry mechanisms1,39, underscoring that the feature of host factors and perchance essential viral factors will fluctuate across permissive cells. This is also attention-grabbing to study whether or no longer the dichotomy of host-directed versus virus-directed perturbations, which we noticed, is a total feature across cell sorts in the case of HCMV and of different virus–host systems more in total.

Our work gives a roadmap for the execute of antiviral aggregate therapies by deciding on sets of targets that drive the virus into definite non-productive pathways. In a similar kind, our records can repeat the execute of attenuated viral lines for vaccine pattern purposes. More in total, we envision that our skill of single-cell functional genomics can reduction as a blueprint for learning other viruses and could well perchance define their vulnerabilities to genetic or pharmacological interventions.

Programs

Cell and virus custom

Human foreskin fibroblasts (HFFs; CRL-1634) and HCMV (stress Merlin; VR-­1590) had been bought from the American Sort Culture Series. HFFs had been cultured in DMEM, supplemented with 10% FBS and penicillin/streptomycin. HCMV shares had been expanded by two rounds of propagation on HFFs and titered by serial dilution.

For valid expression of the CRISPRi/n machineries in HFFs, we modified established lentiviral (d)Cas9 expression vectors17 by inserting a minimal ubiquitous chromatin opening factor (UCOE)57 upstream of the SFFV promoter, ensuing in pMH0001 (UCOE-SFFV-dCas9-BFP-KRAB; Addgene no. 85969) and pMH0004 (UCOE-SFFV-Cas9-BFP; Addgene no. 174087). The UCOE steer clear off epigenetic silencing that affected the distinctive constructs.

Pooled CRISPR screening

The HCMV tiling library used to be designed to have sgRNAs focusing on each and each single one among the 33,465 PAMs in the HCMV Merlin genome (NC_006273.2) as properly as 533 non-focusing on controls (Supplementary Desk 1). It used to be synthesized and cloned staunch into a lentiviral vector (Addgene no. 84832) as beforehand described17,18. For focusing on host genes, we used the human CRISPRi v2 library (Addgene no. 83969)18 and the Yusa et al. human knockout CRISPR v1 library (Addgene no. 67989)40.

Libraries had been packaged into lentiviruses and delivered into (d)Cas9-expressing HFFs at an MOI of 0.3–0.5, followed by puromycin chance. Pooled monitors had been performed at 500–1,000× coverage—that is, ~500–1,000 cells per library factor per sample taken.

A t0 sample used to be harvested, and the final word cells either had been passaged on the general or had been contaminated with HCMV at an MOI of 0.5–1.0 (for the HCMV tiling monitors) or 0.1 (for the host-directed monitors). Contaminated flasks had been washed with PBS and given new medium at days 3, 5 and 7 after an infection to grasp away ineffective cells and harvested at days 7–10. Genomic DNA used to be extracted and digested with MfeI (pCRISPRia v2-primarily based fully libraries) or HindIII (Yusa et al. library) to inaugurate a fragment containing the sgRNA cassette, followed by gel-primarily based fully extraction, PCR amplification and deep sequencing as described18.

Raw depend records had been normalized for read depth, and a exiguous constant used to be added to tale for lacking values. Phenotypes of particular person sgRNAs had been expressed as log2-remodeled ratios of adjusted read counts between samples (Supplementary Desk 2). We calculated the mean of all sgRNAs specific to each and each host gene and computed MAGeCK rankings58. For the HCMV tiling show conceal conceal, we calculated a rolling average in a 250-bp window, with the average of all non-focusing on sgRNAs defining the baseline.

Single-cell RNA sequencing

For the one-cell an infection time route, wild-form HFFs had been lentivirally transduced with barcoded Perturb-seq vectors to encode the experimental situation (pBA571, Addgene no. 85968; Supplementary Desk 3), followed by puromycin chance. Cells had been seeded at a density of 250,000 per properly of a 12-properly plate and contaminated with an MOI of 0.5 or 5.0 with no additional medium commerce earlier than harvest. An infection times had been staggered in inform that at any time when aspects for a given MOI had been harvested in parallel and pooled, aiming for roughly equal cell numbers for at any time when level, with a exiguous over-illustration of the 20- and 28-h time aspects (Prolonged Data Fig. 3a). For every and each MOI, pools of ~10,000 cells had been ready for single-cell transcriptomics the usage of 1 lane each and each of the Chromium Single Cell 3′ Gene Expression Answer v2 in accordance with the manufacturer’s instructions (10x Genomics) and sequenced on a NovaSeq platform (Illumina) at ~100,000 reads per cell. Barcodes encoding the experimental situation had been PCR amplified from the final word library and sequenced as a 5% spike-in as beforehand described34.

Perturb-seq

For the host-directed CRISPRi Perturb-seq experiment, we at the origin selected 53 candidate genes by their valid protective or sensitizing phenotypes in the pooled show conceal conceal (one gene used to be later eradicated at some level of prognosis; sight below). We manually picked the two easiest-performing sgRNAs for each and each candidate. Furthermore, we added six hold a watch on constructs focusing on green fluorescent protein (which is no longer present in our HFFs).

For the host- and virus-directed CRISPRn Perturb-seq experiment, we selected a feature of 21 host factors, of which 19 had been already amongst the targets of the CRISPRi Perturb-seq experiment and had no valid essentiality knockout phenotypes and identical protective or sensitizing phenotypes in each and each the pooled host-directed CRISPRi and CRISPRn monitors (Prolonged Data Fig. 2). We additional added PDGFRA and FLCN, each and each of which had been valid hits in the pooled CRISPRn show conceal conceal. For every and each host goal, we manually picked the two easiest-performing sgRNAs from the pooled show conceal conceal.

Moreover to, we selected 31 viral targets with valid protective or sensitizing phenotypes, much just like the three strongest modules identified in the HCMV tiling show conceal conceal (Fig. 1b and Prolonged Data Fig. 1b). From the tiling show conceal conceal, we selected the two perfect-ranking sgRNAs for each and each goal gene in step with the next scoring arrangement. From the pool of strange sgRNAs falling at some level of the gene boundaries and having a Doench salvage59 of more than 0.5, we calculated absolutely the average phenotype across replicates and subtracted a penalty outlined because the adaptation between replicates plus the average absolute essentiality phenotypes on a log2 scale.

We designed various safe-focusing on hold a watch on sgRNAs focusing on intergenic DNA in the US2–US12 space. This space used to be selected in step with its shut to-fair phenotypes in the tiling show conceal conceal (Prolonged Data Fig. 1b), its lack of predominant genes9,10 and its comparatively neat areas between consensus genes. Furthermore, in some bacterial artificial chromosome (BAC) constructs harboring HCMV genomes, this space used to get replaced by the BAC backbone, underlining its non-predominant nature at some level of an infection in tissue custom60. We picked five sgRNAs in step with their Doench rankings from a pool of strange sgRNAs focusing on the intergenic areas and having survival and essentiality phenotypes of <0.5 (log2 scale) in all replicates. Moreover to, we included four hold a watch on sgRNAs directed against safe-harbor loci in the host genome, which we repurposed from gene knock-in applications.

All sgRNAs had been synthesized as particular person oligonucleotide pairs (IDT) and cloned staunch into a barcode library-containing plasmid pool (pBA571, Addgene no. 85968), thereby linking each and each sgRNA to a strange records barcode contained at some level of the 3′ untranslated space of the puromycin resistance gene34. Barcodes had been validated to no longer have homo-oligomers or sequences such as transcription termination signals. All sgRNA and barcode sequences are listed in Supplementary Tables 4 and 5.

sgRNA vectors had been individually packaged into lentiviruses, titered one at a time and pooled to substantiate equal illustration. This workflow prevents scrambling of records sequences and connected barcodes by recombination, which is a area in pooled lentivirus preparations61. We delivered the pooled library into (d)Cas9-expressing HFFs at an MOI of 0.3 followed by puromycin chance. Cells had been seeded at 250,000 per properly of a 12-properly plate and contaminated with HCMV at an MOI of 0.5 for 1 h, followed by medium commerce (for the CRISPRi host-directed experiment), or an MOI of 5.0, leaving the inoculum on the cells, with the goal of maximizing the numbers of contaminated cells (for the CRISPRn host- and virus-directed experiment). Cells had been harvested in the uninfected recount (designated as 0 h) and at 24, 48 and 72 h.p.i. We geared toward a illustration of every and each library factor by around 100 cells per time level (for accurate cell numbers, sight Prolonged Data Figs. 5a and 6a).

Cells had been smooth and willing for scRNA-seq the usage of the 10x Chromium platform as described above for the one-cell an infection time route. Libraries had been sequenced on a HiSeq 4000 (Illumina) at ~40,000 reads per cell.

Single-cell records prognosis

Raw sequencing records had been submitted to cellranger v2.0.1 (10x Genomics) in accordance with the manufacturer’s instructions. We compiled a reference transcriptome from the hg19 human genome and a personalised assembly of HCMV coding transcripts in step with our earlier ribosome profiling dataset4 as disbursed as segment of the ‘Plastid’ Python library demo dataset62. We manually added four properly-established long noncoding RNA transcripts (RNA1.2, 2.7, 4.9 and 5.0). Interior open reading frames (iORFs) had been eradicated as they’d put ambiguous mappings, as had been ORFs overlapping with the aforementioned long noncoding RNAs.

Cells retained in the final word dataset needed to substandard the default cellranger quality thresholds as properly as hold one strange lentiviral barcode assigned with excessive confidence34.

In the course of records prognosis of the Perturb-seq experiments, three CRISPRn sgRNAs focusing on host genes had been eradicated computationally due to they had been came across to be lazy, as seen by lack of transcriptional responses and viral load patterns much like these in cells with hold a watch on sgRNAs. One host gene, RBBP5, used to be equally excluded from each and each the CRISPRi and CRISPRn datasets because it was apparent that its knockdown/knockout causes differentiation of cells and a valid transcriptional response reasonably than factual protection against an infection (Supplementary Tables 4 and 5).

Percentages of viral RNA (viral loads) had been calculated because the allotment of complete UMIs per cell mapping to viral genes.

Gene expression used to be normalized in each and each cell by a part scaling the total UMIs mapping to human transcripts to its average amount across all cells in a population. This accounts for the true fact that contaminated cells hold worthy elevated complete UMI counts, indicating that viral transcripts accelerate ‘on prime’ of human transcripts (Prolonged Data Fig. 3b).

From the unperturbed time route experiment, we outlined a feature of robustly detected genes as these with more than 10,000 UMIs complete across all cells in that population (3,588 genes in complete, of which 106 are viral genes). This feature of expressed genes used to be used constantly for the prognosis of all entire-transcriptome single-cell datasets in this peek.

For warmth scheme representations of gene expression as a feature of viral load, cells had been binned by viral load, and gene-stage expression values had been averaged in each and each bin. Bin widths of 2% or 10% had been selected looking on the obtainable series of cells. We visualized a slightly elevated feature of viral genes, particularly these expressed in more than 95% of cells in at the least one among these 2% viral load bins (114 genes in complete). Viral transcriptome-centric trajectory analyses (Fig. 6 and Prolonged Data Fig. 7) had been furthermore in step with this selection of viral genes.

Cell cycle phases had been scored in step with marker genes as beforehand described34. The employ of a identical skill, we calculated an IFN salvage by summing (in each and each cell) and this skill that truth z-scoring (across cells) the normalized expression values of the next feature of robustly quantified ISGs: PSMB8, PSMB9, PSME1, PSME2, ISG15, ISG20, IRF7, MX1, MX2, GBP1, GBP2, GBP3, IFI6, IFI44, IFI35, IFI16, IFI27, IFIH1, IFI44L, IFIT1, IFIT2, IFIT3, IFIT5, IFITM1, IFITM2, IFITM3, EIF2AK2, OAS1, OAS2, OAS3, CNP, PLSCR1, BST2, BTN3A2, XAF1, CASP1, CASP4, CASP7 and GSDMD.

To visualize single-cell datasets, we performed dimensionality reduction by UMAP63 (v0.1.1) in step with the matrix of scaled expression values of the feature of robustly detected genes (host + viral genes in Figs. 3 and 4 and Prolonged Data Figs. 3 and 4; viral genes in Fig. 6 and Prolonged Data Fig. 6). Clusters of cells had been outlined by Leiden clustering64 (leidenalg v0.8.0) or HDBSCAN65 (v0.8.26). To resolve trajectories, selected cells had been ranked by viral load, and the geometric recount of cells used to be averaged in a sliding window that used to be shifted in increments of 0.2 window sizes. Window sizes had been selected in step with the total series of accessible cells: 100 cells for each and each sgRNA goal; 500 cells for cells with hold a watch on sgRNAs.

UMAP used to be furthermore used for a dimensionality-diminished visualization of the similarities of the cluster membership records as a feature of time and sgRNA goal (Figs. 4e and 5e; underlying records in Supplementary Tables 4 and 5) and of the viral gene expression records alongside the trajectories outlined by cells with particular person sgRNA targets (Fig. 6f; underlying records in Supplementary Desk 5).

Reporting Summary

Additional records on be taught execute is provided in the Nature Analysis Reporting Summary linked to this article.

Data availability

Raw and processed sequencing records from all pooled monitors and single-cell experiments had been uploaded to the Gene Expression Omnibus (GSE165291).

Code availability

We used printed application for pooled show conceal conceal records processing (https://github.com/mhorlbeck/ScreenProcessing) and for the prognosis of Perturb-seq records (https://github.com/thomasmaxwellnorman/perturbseq_demo) with modifications.

References

  1. 1.

    Mocarski Jr., E. S., Schenk, T., Griffiths, P. D. & Trail, R. F. Cytomegaloviruses. in Fields Virology 1960–2014 (Lippincott Williams & Wilkins, 2013).

  2. 2.

    Gatherer, D. et al. High-resolution human cytomegalovirus transcriptome. Proc. Natl Acad. Sci. USA 108, 19755–19760 (2011).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  3. 3.

    Marcinowski, L. et al. Accurate-time transcriptional profiling of cell and viral gene expression at some level of lytic cytomegalovirus an infection. PLoS Pathog. 8, e1002908 (2012).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  4. 4.

    Stern-Ginossar, N. et al. Decoding human cytomegalovirus. Science 338, 1088–1093 (2012).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  5. 5.

    Weekes, M. P. et al. Quantitative temporal viromics: an skill to study host-pathogen interplay. Cell 157, 1460–1472 (2014).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  6. 6.

    Tirosh, O. et al. The transcription and translation landscapes at some level of human cytomegalovirus an infection give an explanation for new host-pathogen interactions. PLoS Pathog. 11, e1005288 (2015).

    PubMed 
    PubMed Central 

    Google Pupil
     

  7. 7.

    Jean Beltran, P. M., Mathias, R. A. & Cristea, I. M. A portrait of the human organelle proteome in residence and time at some level of cytomegalovirus an infection. Cell Syst. 3, 361–373 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  8. 8.

    Sinzger, C. et al. Fibroblasts, epithelial cells, endothelial cells and light muscle cells are necessary targets of human cytomegalovirus an infection in lung and gastrointestinal tissues. J. Gen. Virol. 76, 741–750 (1995).

    CAS 
    PubMed 

    Google Pupil
     

  9. 9.

    Yu, D., Silva, M. C. & Shenk, T. Purposeful scheme of human cytomegalovirus AD169 outlined by international mutational prognosis. Proc. Natl Acad. Sci. USA 100, 12396–12401 (2003).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  10. 10.

    Dunn, W. et al. Purposeful profiling of a human cytomegalovirus genome. Proc. Natl Acad. Sci. USA 100, 14223–14228 (2003).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  11. 11.

    Puschnik, A. S., Majzoub, Okay., Ooi, Y. S. & Carette, J. E. A CRISPR toolbox to peek virus-host interactions. Nat. Rev. Microbiol. 15, 351–364 (2017).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  12. 12.

    Russell, A. B., Trapnell, C. & Bloom, J. D. Wrong heterogeneity of influenza virus an infection in single cells. eLife 7, e32303 (2018).

  13. 13.

    Erhard, F. et al. scSLAM-seq displays core aspects of transcription dynamics in single cells. Nature 571, 419–423 (2019).

    CAS 
    PubMed 

    Google Pupil
     

  14. 14.

    van Diemen, F. R. et al. CRISPR/Cas9-mediated genome editing of herpesviruses limits productive and latent infections. PLoS Pathog. 12, e1005701 (2016).

    PubMed 
    PubMed Central 

    Google Pupil
     

  15. 15.

    Van Damme, E. & Van Loock, M. Purposeful annotation of human cytomegalovirus gene merchandise: an update. Front. Microbiol. 5, 218 (2014).

    PubMed 
    PubMed Central 

    Google Pupil
     

  16. 16.

    McCormick, A. L., Skaletskaya, A., Barry, P. A., Mocarski, E. S. & Goldmacher, V. S. Differential feature and expression of the viral inhibitor of caspase 8-induced apoptosis (vICA) and the viral mitochondria-localized inhibitor of apoptosis (vMIA) cell death suppressors conserved in primate and rodent cytomegaloviruses. Virology 316, 221–233 (2003).

    CAS 
    PubMed 

    Google Pupil
     

  17. 17.

    Gilbert, L. A. et al. Genome-scale CRISPR-mediated hold a watch on of gene repression and activation. Cell 159, 647–661 (2014).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  18. 18.

    Horlbeck, M. A. et al. Compact and extremely filled with life subsequent-technology libraries for CRISPR-mediated gene repression and activation. eLife 5, e19760 (2016).

    PubMed 
    PubMed Central 

    Google Pupil
     

  19. 19.

    Compton, T., Nowlin, D. M. & Cooper, N. R. Initiation of human cytomegalovirus an infection requires initial interplay with cell ground heparan sulfate. Virology 193, 834–841 (1993).

    CAS 
    PubMed 

    Google Pupil
     

  20. 20.

    Hetzenecker, S., Helenius, A. & Krzyzaniak, M. A. HCMV induces macropinocytosis for host cell entry in fibroblasts. Site visitors 17, 351–368 (2016).

    CAS 
    PubMed 

    Google Pupil
     

  21. 21.

    Pusapati, G. V., Luchetti, G. & Pfeffer, S. R. Ric1-Rgp1 advanced is a guanine nucleotide commerce factor for the dull Golgi Rab6A GTPase and an effector of the medial Golgi Rab33B GTPase. J. Biol. Chem. 287, 42129–42137 (2012).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  22. 22.

    Smith, R. D. et al. The COG advanced, Rab6 and COPI define a new Golgi retrograde trafficking pathway that is exploited by SubAB toxin. Site visitors 10, 1502–1517 (2009).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  23. 23.

    Liu, S., Dominska-Ngowe, M. & Dykxhoorn, D. M. Target silencing of components of the conserved oligomeric Golgi advanced impairs HIV-1 replication. Virus Res. 192, 92–102 (2014).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  24. 24.

    Bassik, M. C. et al. A systematic mammalian genetic interplay scheme displays pathways underlying ricin susceptibility. Cell 152, 909–922 (2013).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  25. 25.

    Selyunin, A. S., Iles, L. R., Bartholomeusz, G. & Mukhopadhyay, S. Genome-vast siRNA show conceal conceal identifies UNC50 as a regulator of Shiga toxin 2 trafficking. J. Cell Biol. 216, 3249–3262 (2017).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  26. 26.

    Starling, G. P. et al. Folliculin directs the formation of a Rab34-RILP advanced to hold a watch on the nutrient-dependent dynamic distribution of lysosomes. EMBO Fetch. 17, 823–841 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  27. 27.

    Filipek, P. A. et al. LAMTOR/Ragulator is a detrimental regulator of Arl8b- and BORC-dependent dull endosomal positioning. J. Cell Biol. 216, 4199–4215 (2017).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  28. 28.

    Pu, J., Keren-Kaplan, T. & Bonifacino, J. S. A Ragulator-BORC interplay controls lysosome positioning in accordance with amino acid availability. J. Cell Biol. 216, 4183–4197 (2017).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  29. 29.

    Becker, T., Le-Trilling, V. T. Okay. & Trilling, M. Cellular cullin RING ubiquitin ligases: druggable host dependency factors of cytomegaloviruses. Int. J. Mol. Sci. 20, 163 (2019).


    Google Pupil
     

  30. 30.

    Le-Trilling, V. T. Okay. et al. Substantial and potent antiviral boom of the NAE inhibitor MLN4924. Sci. Fetch. 6, 19977 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  31. 31.

    Ott, M., Marques, D., Funk, C. & Bailer, S. M. Asna1/TRC40 that mediates membrane insertion of tail-anchored proteins is required for ambiance pleasant inaugurate of herpes simplex virus 1 virions. Virol. J. 13, 175 (2016).

    PubMed 
    PubMed Central 

    Google Pupil
     

  32. 32.

    Nishitsuji, H., Ujino, S., Harada, Okay. & Shimotohno, Okay. TIP60 advanced inhibits hepatitis B virus transcription. J. Virol. 92, e01788-17 (2018).

    PubMed 
    PubMed Central 

    Google Pupil
     

  33. 33.

    Kamine, J., Elangovan, B., Subramanian, T., Coleman, D. & Chinnadurai, G. Identification of a cell protein that particularly interacts with the predominant cysteine space of the HIV-1 Tat transactivator. Virology 216, 357–366 (1996).

    CAS 
    PubMed 

    Google Pupil
     

  34. 34.

    Adamson, B. et al. A multiplexed single-cell CRISPR screening platform enables systematic dissection of the unfolded protein response. Cell 167, 1867–1882 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  35. 35.

    Cohen, Y. et al. The yeast p5 form ATPase, spf1, regulates manganese transport into the endoplasmic reticulum. PLoS ONE 8, e85519 (2013).

    PubMed 
    PubMed Central 

    Google Pupil
     

  36. 36.

    Wu, Okay., Oberstein, A., Wang, W. & Shenk, T. Draw of PDGF receptor-α at some level of human cytomegalovirus entry into fibroblasts. Proc. Natl Acad. Sci. USA 115, E9889–E9898 (2018).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  37. 37.

    Soroceanu, L., Akhavan, A. & Cobbs, C. S. Platelet-derived command factor-α receptor activation is required for human cytomegalovirus an infection. Nature 455, 391–395 (2008).

    CAS 
    PubMed 

    Google Pupil
     

  38. 38.

    Kabanova, A. et al. Platelet-derived command factor-α receptor is the cell receptor for human cytomegalovirus gHgLgO trimer. Nat. Microbiol. 1, 16082 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  39. 39.

    Martinez-Martin, N. et al. An fair show conceal conceal for human cytomegalovirus identifies neuropilin-2 as a central viral receptor. Cell 174, 1158–1171 (2018).

    CAS 
    PubMed 

    Google Pupil
     

  40. 40.

    Tzelepis, Okay. et al. A CRISPR dropout show conceal conceal identifies genetic vulnerabilities and therapeutic targets in acute myeloid leukemia. Cell Fetch. 17, 1193–1205 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  41. 41.

    Morgens, D. W., Deans, R. M., Li, A. & Bassik, M. C. Systematic comparability of CRISPR/Cas9 and RNAi monitors for predominant genes. Nat. Biotechnol. 34, 634–636 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  42. 42.

    Datlinger, P. et al. Pooled CRISPR screening with single-cell transcriptome readout. Nat. Programs 14, 297–301 (2017).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  43. 43.

    Dixit, A. et al. Perturb-seq: dissecting molecular circuits with scalable single-cell RNA profiling of pooled genetic monitors. Cell 167, 1853–1866 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  44. 44.

    Jaitin, D. A. et al. Dissecting immune circuits by linking CRISPR-pooled monitors with single-cell RNA-seq. Cell 167, 1883–1896 (2016).

    CAS 
    PubMed 

    Google Pupil
     

  45. 45.

    Zanini, F., Pu, S.-Y., Bekerman, E., Einav, S. & Quake, S. R. Single-cell transcriptional dynamics of flavivirus an infection. eLife 7, e32942 (2018).

    PubMed 
    PubMed Central 

    Google Pupil
     

  46. 46.

    Drayman, N., Patel, P., Vistain, L. & Tay, S. HSV-1 single-cell prognosis displays the activation of anti-viral and developmental programs in definite sub-populations. eLife 8, e46339 (2019).

  47. 47.

    Wyler, E. et al. Single-cell RNA-sequencing of herpes simplex virus 1-contaminated cells connects NRF2 activation to an antiviral program. Nat. Commun. 10, 4878 (2019).

    PubMed 
    PubMed Central 

    Google Pupil
     

  48. 48.

    Kotliar, D. et al. Single-cell profiling of Ebola virus disease in vivo displays viral and host dynamics. Cell 183, 1383–1401 (2020).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  49. 49.

    An, P. et al. Single cell transcriptomics displays a heterogeneous cell response to BKV an infection. J. Virol. 95, e02237 (2020).

  50. 50.

    Bogdanow, B. et al. Human cytomegalovirus tegument protein pp150 acts as a cyclin A2-CDK-dependent sensor of the host cell cycle and differentiation recount. Proc. Natl Acad. Sci. USA 110, 17510–17515 (2013).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  51. 51.

    Spear, G. T. et al. Host cell-derived complement hold a watch on proteins CD55 and CD59 are integrated into the virions of two unrelated enveloped viruses. Human T cell leukemia/lymphoma virus form I (HTLV-I) and human cytomegalovirus (HCMV). J. Immunol. 155, 4376–4381 (1995).

    CAS 
    PubMed 

    Google Pupil
     

  52. 52.

    Isomura, H. & Stinski, M. F. Coordination of dull gene transcription of human cytomegalovirus with viral DNA synthesis: recombinant viruses as attainable therapeutic vaccine candidates. Skilled Opin. Ther. Targets 17, 157–166 (2013).

    CAS 
    PubMed 

    Google Pupil
     

  53. 53.

    Zhang, L., Yu, J. & Liu, Z. MicroRNAs expressed by human cytomegalovirus. Virol. J. 17, 34 (2020).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  54. 54.

    Winkler, M., Rice, S. A. & Stamminger, T. UL69 of human cytomegalovirus, an open reading frame with homology to ICP27 of herpes simplex virus, encodes a transactivator of gene expression. J. Virol. 68, 3943–3954 (1994).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  55. 55.

    Lischka, P., Toth, Z., Thomas, M., Mueller, R. & Stamminger, T. The UL69 transactivator protein of human cytomegalovirus interacts with DEXD/H-Box RNA helicase UAP56 to promote cytoplasmic accumulation of unspliced RNA. Mol. Cell. Biol. 26, 1631–1643 (2006).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  56. 56.

    Schommartz, T., Tang, J., Brost, R. & Brune, W. Differential requirement of human cytomegalovirus UL112–113 protein isoforms for viral replication. J. Virol. 91, e00254-17 (2017).

    PubMed 
    PubMed Central 

    Google Pupil
     

  57. 57.

    Müller-Kuller, U. et al. A minimal ubiquitous chromatin opening factor (UCOE) effectively prevents silencing of juxtaposed heterologous promoters by epigenetic reworking in multipotent and pluripotent stem cells. Nucleic Acids Res. 43, 1577–1592 (2015).

    PubMed 
    PubMed Central 

    Google Pupil
     

  58. 58.

    Li, W. et al. MAGeCK enables sturdy identification of predominant genes from genome-scale CRISPR/Cas9 knockout monitors. Genome Biol. 15, 554 (2014).

    PubMed 
    PubMed Central 

    Google Pupil
     

  59. 59.

    Doench, J. G. et al. Optimized sgRNA execute to maximise boom and lower off-goal outcomes of CRISPR–Cas9. Nat. Biotechnol. 34, 184–191 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  60. 60.

    Borst, E. M., Hahn, G., Koszinowski, U. H. & Messerle, M. Cloning of the human cytomegalovirus (HCMV) genome as an infectious bacterial artificial chromosome in Escherichia coli: a new skill for construction of HCMV mutants. J. Virol. 73, 8320–8329 (1999).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  61. 61.

    Adamson, B., Norman, T. M., Jost, M. & Weissman, J. S. Approaches to maximise sgRNA-barcode coupling in Perturb-seq monitors. Preprint at bioRxiv https://doi.org/10.1101/298349 (2018).

  62. 62.

    Dunn, J. G. & Weissman, J. S. Plastid: nucleotide-resolution prognosis of subsequent-technology sequencing and genomics records. BMC Genomics 17, 958 (2016).

    PubMed 
    PubMed Central 

    Google Pupil
     

  63. 63.

    Becht, E. et al. Dimensionality reduction for visualizing single-cell records the usage of UMAP. Nat. Biotechnol. https://doi.org/10.1038/nbt.4314 (2018).

  64. 64.

    Traag, V. A., Waltman, L. & van Eck, N. J. From Louvain to Leiden: guaranteeing properly-linked communities. Sci. Fetch. 9, 5233 (2019).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  65. 65.

    Ding, J., Shah, S. & Condon, A. densityCut: an ambiance pleasant and versatile topological skill for computerized clustering of biological records. Bioinformatics 32, 2567–2576 (2016).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

  66. 66.

    Heer, J., Kong, N. & Agrawala, M. Sizing the horizon: the implications of chart dimension and layering on the graphical perception of time series visualizations. in Proc. 27th Int. Conf. Human Components in Computing Systems (ACM Press, 2009).

  67. 67.

    Murphy, E., Rigoutsos, I., Shibuya, T. & Shenk, T. E. Reevaluation of human cytomegalovirus coding attainable. Proc. Natl Acad. Sci. USA 100, 13585–13590 (2003).

    CAS 
    PubMed 
    PubMed Central 

    Google Pupil
     

Download references

Acknowledgements

We thank M. A. Horlbeck for designing the HCMV tiling library; L. A. Gilbert for relieve constructing pooled monitors; and T. M. Norman, M. A. Horlbeck, J. A. Hussmann and X. Qiu for relieve with records prognosis. A. Xu, J. A. Villalta and R. A. Pak provided technical assistance. The UCOE sequence used to be a reward from G. Sienski. We thank T. Magnificent for relieve with Perturb-seq experiments. We thank N. Stern-Ginossar, M. J. Shurtleff, M. Jost, R. A. Saunders, J. M. Replogle, X. Qiu and all individuals of the Weissman lab for insightful discussions. J. Winkler and A. S. Puschnik provided priceless comments on the manuscript. Special due to O. Wueseke for editorial relieve. J.S.W. is a Howard Hughes Medical Institute Investigator. M.Y.H. used to be supported by an EMBO long-time length postdoctoral fellowship (EMBO ALTF 1193-2015, co-funded by the European Commission FP7, Marie Curie Actions, LTFCOFUND2013, GA-2013-609409).

Author records

Affiliations

  1. Division of Cellular and Molecular Pharmacology, University of California, San Francisco, San Francisco, CA, USA

    Marco Y. Hein & Jonathan S. Weissman

  2. Howard Hughes Medical Institute, University of California, San Francisco, San Francisco, CA, USA

    Marco Y. Hein & Jonathan S. Weissman

  3. Chan Zuckerberg Biohub, San Francisco, CA, USA

    Marco Y. Hein

  4. Whitehead Institute for Biomedical Analysis, Massachusetts Institute of Technology, Cambridge, MA, USA

    Jonathan S. Weissman

Contributions

M.Y.H. and J.S.W. conceptualized the peek, interpreted the experiments and wrote the manuscript. M.Y.H. designed and performed the experiments and analyzed the records.

Corresponding authors

Correspondence to
Marco Y. Hein or Jonathan S. Weissman.

Ethics declarations

Competing interests

J.S.W. has filed patent applications connected to CRISPRi screening and Perturb-seq. J.S.W. consults for and holds fairness in KSQ Therapeutics and Maze Therapeutics and consults for 5AM Ventures.

Additional records

Admire overview records Nature Biotechnology thanks Antoine-Emmanuel Saliba and the assorted, anonymous, reviewer(s) for their contribution to the discover overview of this work.

Writer’s show conceal Springer Nature remains fair on the realm of jurisdictional claims in printed maps and institutional affiliations.

Prolonged records

Prolonged Data Fig. 1 High-resolution HCMV tiling show conceal conceal.

a, Data processing for the HCMV tiling show conceal conceal. We calculated log2 ratios of every and each particular person sgRNA in the surviving over the t0 populations, averaged across two biological replicates. Ratios had been averaged in a sliding 250 bp window. The in form of the ratios of the non-focusing on sgRNA population used to be feature because the baseline. The recount used to be then colored in step with the signal of the average phenotype and layered in bands of reducing lightness, one log2 unit vast. The detrimental residence used to be mirrored on the baseline, and bands had been stacked for the final word horizon recount illustration66. b, High-resolution horizon graph of the phenotypic panorama of the HCMV genome. Shades of blue denote sensitization to host cell death, shades of crimson denote protection from cell death upon HCMV genome cleavage. Predominant aspects of the HCMV genome are annotated. sgRNAs focusing on inside of and terminal repeat areas (hashed) generally hold more than one goal internet sites and sure discontinue in elevated-uncover fragmentation of the HCMV genome, exacerbating their respective phenotypes. Viral ORFs are classified by their essentiality for viral replication in step with ref. 10. ORFL150C, ORFL151C (at the birth named UL59, however thought to no longer be expressed as a protein67, causing it to be dropped from the consensus annotation), and ORFL152C had been the fully instant ORFs with valid phenotypes in areas of the genome devoid of consensus genes. UL48 used to be the fully gene that showed a the truth is perfect phenotype gradient inside of its gene body: Cutting the N-terminal space triggered gentle sensitization to death upon an infection, whereas reducing the C-terminus had the reverse form.

Prolonged Data Fig. 2 Host-directed CRISPRi and CRISPRn show conceal conceal.

a, Outcomes of the host-directed CRISPRn show conceal conceal displayed as a scatter recount of average gene essentiality (that is an infection-fair phenotype; y-axis) vs. protection/sensitization to death upon HCMV an infection (that is an infection-dependent phenotypes; x-axis), highlighting a manually selected feature of hit genes. Account for that as a result of experimental execute of the show conceal conceal, the obvious gene essentiality phenotypes are underestimating the actual essentiality due to t0 refers to the starting of HCMV an infection, no longer lentiviral provide of the sgRNA library. b, Scream comparability of CRISPRi and CRISPRn phenotypes, highlighting purchase host targets represented in each and each libraries. Hits interested by viral adhesion and entry are more pronounced in the CRISPRn show conceal conceal. Cullin/RING pathway individuals and some vesicle trafficking factors had been fully resolved in the CRISPRi show conceal conceal. Look Supplemental Desk 2 for a systematic comparability.

Prolonged Data Fig. 3 Single-cell an infection time-route.

a, b, Numbers of cells, as properly as distributions of UMIs per cell, proportion of viral transcripts per cell, and interferon salvage, broken down by cells for each and each MOI and each and each experimental time level (a), and for cells for each and each MOI and cluster membership (b). c, Distribution of cells across cell cycle phases in each and each of the clusters. d, Pearson’s correlation matrix of gene expression values (average logarithmized, scaled UMIs per gene per cell) for all clusters, broken down by low (L) and excessive (H) MOI prerequisites.

Prolonged Data Fig. 4 Viral gene expression profiles alongside the productive and abortive trajectory of an infection.

a, Gene expression profiles for robustly detected viral genes alongside the dominant trajectory (clusters ‘contaminated 1–6’). Cells had been grouped in bins spanning 2% of viral RNA and the gene expression (scaled UMIs per gene per cell) averaged for all cells in each and each bin. The heatmap presentations the expression relative to the perfect bin. Particular person viral genes are ordered by similarity of the profiles, and annotated by genome recount, phenotype of reducing at some level of the body of a gene in the pooled virus-directed CRISPR show conceal conceal (sight Fig. 1 and Supplementary Fig. 1), and by the temporal profile as determined in a bulk proteomics peek5. Account for the connection between a gene’s temporal class and its phenotype in the pooled show conceal conceal: Right-dull and leaky-dull genes predominantly showed protective phenotypes, whereas earlier classes furthermore contained sensitizing genes. b, Gene expression profiles of viral genes alongside the abortive trajectory (clusters ‘contaminated 1–2’ and ‘contaminated abortive’). Cells had been grouped in bins spanning 10% of viral RNA and the gene expression averaged for all cells in each and each bin. The heatmap presentations the expression relative to the expression in a similar bin of the dominant trajectory.

Prolonged Data Fig. 5 Host-directed CRISPRi Perturb-seq experiment.

a, Numbers of single cells for each and each sgRNA goal for each and each experimental time level in the host-directed CRISPRi Perturb-seq experiment. The in kind is 165 ± 50 (mean ± normal deviation) cells per sgRNA per time level. b, Knockdown phases for each and each sgRNA goal calculated from the expression of the goal gene in cells with a given sgRNA goal relative to cells with hold a watch on sgRNAs. No transcript at all used to be detected for VTCN1. Median knockdown stage used to be 87.1%. c, Hierarchical clustering of expression adjustments of primarily the most variable 100 genes (as antagonistic to the targeted factors) in accordance with host factor knockdown in naïve cells, relative to naïve cells with hold a watch on sgRNAs. a–g, UMAP projections of single-cell transcriptomes of cells from the host-directed Perturb-seq experiment (same as in Fig. 4c), color-coded by experimental time put up an infection (d), proportion of viral transcripts per cell (e), interferon salvage, calculated from the normalized expression of interferon stimulated genes (f), and by pathway of the targeted host accept as true with each and each cell (g). h, Cluster membership as a feature of sgRNA goal and time put up an infection.

Prolonged Data Fig. 6 Host- and virus directed CRISPRn Perturb-seq experiment.

a, Numbers of single cells for each and each sgRNA goal for each and each experimental time level in the host and virus-directed CRISPRn Perturb-seq experiment. The in kind is 188 ± 77 (mean ± normal deviation) cells per sgRNA per time level. Account for the over-proportional plunge in numbers in dull time aspects of cells with apoptosis-connected sgRNA targets. ‘Control’ denotes all safe-focusing on sgRNAs, that are 4 and 5 definite sgRNAs focusing on the host and virus, respectively. b, Violin plots of the distribution of viral RNA allotment per cell as a feature of time put up an infection and the sgRNA goal (crimson, protective phenotype; blue, sensitizing phenotype; gray, hold a watch on). Regions of the violin recount much like uninfected cells, as properly as early and dull stages of an infection are highlighted. Account for that uninfected cells hold non-zero background portions of viral RNA, and these background phases are elevated in later time aspects, indicating leaking of viral RNA from dying cells. c, d, Cluster membership as a feature of sgRNA goal and time put up an infection for cells with host-focusing on sgRNAS (c) and virus-focusing on sgRNAs (d). e–g, UMAP projections of single-cell transcriptomes of cells from the host and virus-directed Perturb-seq experiment (same as in Fig. 5c), color-coded by proportion of viral transcripts per cell (e), by pathway of the targeted host accept as true with each and each cell (f), and by viral goal in each and each cell (g) for cells with host and viral targets, respectively.

Prolonged Data Fig. 7 Trajectories in viral gene expression residence upon perturbation of viral factors.

a, Cartoon explaining the analytical workflow for evaluating viral trajectories across different sgRNA targets. b, Heatmaps of viral gene expression for all cells with virus-focusing on sgRNAs, much just like the center panel of the workflow comic strip. For every and each sgRNA goal, cells had been grouped in bins of 10% of viral RNA allotment, and the expression of viral genes plotted relative to a corresponding bin outlined by cells with host-directed, safe-focusing on sgRNAs (much like Supplementary Fig. 4b), representing the unperturbed, dominant trajectory. Each and each the columns (viral sgRNA targets) as properly because the rows (expressed viral genes) are ordered by genome recount. This facilitates the distinction of gene expression outcomes in cis, that is the instantaneous form of reducing on genes adjoining to the lower plight, as against in trans, that are reflecting an altered trajectory of an infection. Red bins order the sgRNA goal genes. c, Mapping the sgRNA targets onto the phenotypic panorama of the HCMV genome, indicating genome recount and phenotype in the CRISPRn tiling show conceal conceal.

Supplementary records

Supplementary Desk 1

sgRNA sequences of the HCMV tiling library, uncooked sequencing counts in the show conceal conceal and normalized records-stage phenotypes. Gene-stage phenotypes for consensus genes (in step with NCBI). Gene-stage phenotypes for all ORFs, in step with ref. 4.

Supplementary Desk 2

Raw sequencing counts for the human genome-vast CRISPRi monitors. Gene-stage phenotypes for the human genome-vast CRISPRi monitors, in conjunction with significance rankings calculated by MAGeCK58. Raw sequencing counts for the human genome-vast CRISPRn show conceal conceal. Gene-stage phenotypes for the human genome-vast CRISPRn show conceal conceal, in conjunction with significance rankings calculated by MAGeCK. Comparison of gene-stage phenotypes and MAGeCK rankings between the CRISPRi and CRISPRn show conceal conceal outcomes.

Supplementary Desk 3

Metadata annotations for all cells in the final word dataset. Desk of the expressed barcodes used to de-convolve the pooled cells into the experimental time aspects. Expression values of all robustly detected host and viral genes in the actual person clusters. Expression values of robustly detected viral genes alongside the default trajectory of an infection. Expression values of viral genes alongside the abortive trajectory of an infection.

Supplementary Desk 4

Metadata annotations for all cells in the final word dataset. sgRNA sequences, records barcodes and annotations for all components of the library. Desk of cell numbers in each and each cluster as a feature of experimental time. Expression values of all robustly detected host genes in the naive cluster, as a feature of sgRNA goal. Expression values of all robustly detected host genes in the bystander cluster, as a feature of sgRNA goal.

Supplementary Desk 5

Metadata annotations for all cells in the final word dataset. sgRNA sequences, records barcodes and annotations for all components of the library. Desk of cell numbers in each and each cluster as a feature of experimental time. Expression values of all robustly detected viral genes alongside the trajectories of an infection, as a feature of sgRNA goal.

About this article

Verify currency and authenticity via CrossMark

Cite this article

Hein, M.Y., Weissman, J.S. Purposeful single-cell genomics of human cytomegalovirus an infection.
Nat Biotechnol (2021). https://doi.org/10.1038/s41587-021-01059-3

Download citation

Related Articles

Back to top button
%d bloggers like this: