A Genome-Wide CRISPR/Cas9 Screen Reveals the Requirement of Host Sphingomyelin Synthase 1 for Infection with Pseudorabies Virus Mutant gD–Pass

Herpesviruses are large DNA viruses, which encode up to 300 different proteins including enzymes enabling efficient replication. Nevertheless, they depend on a multitude of host cell proteins for successful propagation. To uncover cellular host factors important for replication of pseudorabies virus (PrV), an alphaherpesvirus of swine, we performed an unbiased genome-wide CRISPR/Cas9 forward screen. To this end, a porcine CRISPR-knockout sgRNA library (SsCRISPRko.v1) targeting 20,598 genes was generated and used to transduce porcine kidney cells. Cells were then infected with either wildtype PrV (PrV-Ka) or a PrV mutant (PrV-gD–Pass) lacking the receptor-binding protein gD, which regained infectivity after serial passaging in cell culture. While no cells survived infection with PrV-Ka, resistant cell colonies were observed after infection with PrV-gD–Pass. In these cells, sphingomyelin synthase 1 (SMS1) was identified as the top hit candidate. Infection efficiency was reduced by up to 90% for PrV-gD–Pass in rabbit RK13-sgms1KO cells compared to wildtype cells accompanied by lower viral progeny titers. Exogenous expression of SMS1 partly reverted the entry defect of PrV-gD–Pass. In contrast, infectivity of PrV-Ka was reduced by 50% on the knockout cells, which could not be restored by exogenous expression of SMS1. These data suggest that SMS1 plays a pivotal role for PrV infection, when the gD-mediated entry pathway is blocked.


Introduction
Many alphaherpesviruses are characterized by a broad host range in vitro and in vivo. This is particularly true for the alphaherpesvirus pseudorabies virus (PrV; Suid alphaherpesvirus 1), causing the notifiable Aujeszky's disease. Despite the broad host range, pigs, especially domestic pigs and wild boar, are the only natural host and reservoir of PrV [1].
Here, we generated a porcine sgRNA library (SsCRISPRko.v1), targeting all annotated genes in the porcine genome (Sus (S.) Scrofa 10.2). This library was then used in a genomewide screen to study host cell factors, which play an important role for PrV infection. This approach resulted in the identification of sphingomyelin synthase 1 (SMS1) as an important entry factor when the canonical gD-mediated entry pathway is blocked.
Generation of a porcine CRISPR library. A library of sgRNAs for a genome-wide CRISPR knockout (KO) of protein-coding genes in the porcine genome was designed using custom scripts in Python 2.7. Designs were based on the genome assembly S. scrofa 10.2 [46], using RefSeq gene models from NCBI S. scrofa Annotation Release 105. Predicted genes with no published gene symbol or identified orthologues, with a coding sequence of less than 5000 base pairs were excluded. For each gene, the transcript variant conforming most closely to consensus, as defined by the highest average frequency of exon occurrence across all recognized transcript variants, was selected as the basis for the guide design. First, candidate guides were identified based on the presence of an "NGG" (N: any base; G: guanosine) PAM on either strand giving rise to a predicted Cas9 cut site within an exon, at least 2 nucleotides (nt) from the intron-exon junction. Candidates were rejected if (i) more than one-third of the 30 nt sequence (20 nt protospacer with 4 additional 5 nt and 6 additional 3 nt including the PAM sequence) consisted of repetitive or low-complexity elements, (ii) there is more than one ambiguous base, or (iii) the protospacer contained a BsmbI restriction site. Potential cutting efficiency ("on-target score") was estimated using the Doench Rule Set 2 model in Python package scikit-learn version 0.16.1. [34]. To screen for off-target matches, candidate guide sequences (with appended PAM sequence) were aligned to the genome using BLAT [47], and the likelihood of an off-target match was estimated using a previously published Cutting Frequency Determination (CFD) score [34]. Four guides were chosen for each gene, but if insufficient candidates were available, guides exact matches were permitted, for example to allow targeting of genes with multiple paralogues. Corresponding oligonucleotides were generated by solid phase dense array synthesis by the Broad Institute.
Prior to library synthesis, the suitability of the design was tested using sgRNAs targeting two host genes known to be involved in influenza virus replication [40,48,49]. The sialic acid transporter SLC35A1 was targeted by two sgRNAs, while in addition the transcription factor interferon regulatory factor 7 (IRF7) was targeted by one sgRNA (Table 1). The sgRNAs were cloned into the pLentiCRISPRv2 backbone (Addgene #52961, containing the Cas9 gene of S. pyogenes) using the BsmBI restriction overhangs and Lentiviral vectors were generated as described above. For validation of individual guide cutting efficiency, PK15 cells were transduced with each of the lentiviral vectors as described in the following section and selected with 1.25 µg/µL puromycin for 10 days. To confirm efficient genome editing, genomic DNA (gDNA) was extracted using the DNeasy kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. The region of gDNA, approximately 300 bp either side of the cut site, was then amplified by PCR using primers given in Table 1. The PCR product was sequenced using Sanger sequencing and the resulting chromatogram was analyzed for cutting efficiency using the TIDE website (http://shinyapps.datacurators.nl/tide/, 14 June 2017).
After the successful preliminary test, the library consisting of 83,381 specific sgRNA and in addition 1000 non-targeting controls was synthesized. Following synthesis, the oligonucleotides were pooled and amplified using non-variable primers that introduce BsmBI restriction sites. The amplified products were then size fractionated by electrophoresis, purified and shotgun cloned into plentiCRISPRv2 (Addgene #52961). To maintain sgRNA representation, multiple repeated cloning reactions were performed and electroporated into highly competent bacteria, which were plated on 25 cm 2 agar plates. Colonies were harvested by scraping and DNA was extracted by Maxi Kit (Qiagen, Hilden, Germany). Relative coverage and representation of sgRNAs were confirmed by deep sequencing.
The remaining cells were infected with PrV (PrV-Ka or PrV-gD -Pass) at a multiplicity of infection (MOI) of 0.5. The medium was changed four days after infection and then every week, containing 20% conditioned medium from uninfected and untreated PK15 Viruses 2021, 13, 1574 5 of 20 cells. After approx. ten to 16 days, cell colonies became detectable. These cell colonies were harvested and pooled. Half of the sample was stored at −20 • C and the other half was seeded into 20 cm dishes and re-infected as described above. The survivors of the second infection were harvested approx. 28 days after the first infection, pelletized and stored at −20 • C until further analysis.
The sequencing library was generated by PCR using the different gDNAs as template. For each sample four reactions with each 50 µL PCR reaction mix containing ExTaq DNA polymerase (Clontech, Saint-Germain-en-Laye, France), ExTaq buffer, Mg 2+ , dNTPs, P5barcode and P7 reverse primers ( Table 2) and 5 µg of gDNA were prepared according to manufacturer's specifications. For each sample, a different P5 barcode primer was used. The PCR was performed as follows: 95 • C for 1 min, 28 cycles of 95 • C for 30 s, 53 • C for 30 s, 72 • C for 1 min with a final amplification at 72 • C for 10 min. After amplification, the four PCR reactions per sample were pooled, the 354 bp fragment was gel purified using a gel extraction kit (Zymo Research, Freiburg, Germany) and eluted in nuclease-free water. Isolation of gDNA, PCR, as well as gel extraction were performed in parallel with the corresponding controls for each set of samples to minimize any bias. The screening procedure is illustrated in Figure 1. with puromycin and expanded for ten to 14 days to achieve at least 500-fold coverage for each sgRNA as recommended [35]. From this cell pool, 2 × 10 7 cells were harvested as control and stored for further analysis. The remaining cell pool was then infected with either PrV-Ka or PrV-gD -Pass at an MOI of 0.5. Interestingly, no cells survived the infection with PrV-Ka in repeated assays, even after a prolonged incubation of up to three months indicating that PrV can use a wide panel of different host cell factors for successful infection. Because of this, we tested a mutant, PrV-gD -Pass, which lacks the receptor-binding protein gD but regained infectivity after repeated passages in cell culture [17]. Mutations in gB and gH have been shown to be involved in the gD-independent entry pathway but the exact mechanism is yet unknown [21]. Our genome-wide CRISPR/Cas9 approach should shed more light on this mode of infectious entry.
After infection with PrV-gD -Pass single cell colonies were visible after 10 days post infection (p.i.). These cell colonies were harvested 14 to 16 days p.i., pooled and one half of the cell pool was saved as 'survivors #1' for gDNA isolation and sequencing, while the other half was reinfected with PrV-gD -Pass at an MOI of 0.5. 'Survivors #2′ were harvested 12 to 14 days after the second round of infection. The screening was done in three Samples were sequenced on an IonTorrent Ion S5 TM XL System (Invitrogen, Dreieich, Germany). The sequencing data were analyzed on the Galaxy web platform (usegalaxy.eu) [48]. The sequencing reads were trimmed to 25 nt sequences with Cutadapt (Galaxy Version 1.16.5) using the 5 front adapter sequence (TTGTGGAAAGGACGAAACACCG) [49]. The trimmed reads were mapped to the original library file (Supplementary Table S1) analyzing the quality of the sequencing run using the MAGeCK count tool with default settings (Galaxy version 0.5.8.4) [51]. To compare the sequencing results of two different conditions ("Control" vs. "Survivor") we used the robust rank aggregation (RRA) method of the MAGeCK test tool (Galaxy version 0.5.8.1) [51].
To further evaluate the SsCRISPRko.v1 library performance, an essential gene analysis was performed on the sequencing results from the control cells, representing the modified cell pool after selection and expansion. First, enrichment or depletion of sgRNAs in the control cells relative to the sequenced input plasmid pool was assessed using MAGeCK as above. Target genes were subsequently ranked according to fold change sgRNA enrichment in the cells versus the plasmid stock. Enrichment of essential genes, as defined using previously published reference lists of essential genes [52,53], in gene rankings was evaluated by gene set enrichment analysis in the MAGeCK implementation of GSEA (Galaxy version 0.5.8 [51]), using 10 6 permutations for p-value estimation. The lists for this analysis consisted of: (i) a longer list of 1580 essential genes identified by CRISPR/Cas9 fitness screening in five human cell lines [52], of which 1308 could be matched to unique porcine genes used in the porcine screen or (ii) a refined list of "core" essential genes [53], derived from the previously mentioned long list by network analysis. This included 47 genes matched to unique porcine gene names. We visualized the correlation between the gene expression values (LogFC) as a violin plot. Violin plots were generated in Python 3.7, using package matplotlib v3.1.0.
Generation of RK13-sgms1 KO cells. For generation of RK13 knockout (KO) cells four sgRNAs were designed targeting the third exon of sphingomyelin synthase 1 (SMS1; sgms1, ENSOCUG00000010965.4) as predicted in the rabbit genome OryCun2.0 (Oryctolagus cuniculus, ensemble.org [54]) with the help of the corresponding tool in Geneious software 11.1.5 (https://www.geneious.com, 1 Decmeber 2020). The four gRNAs with the highest score and the lowest probability for off-target effects were selected (Table 3). sgRNAs were ordered as unmodified DNA oligonucleotides (MWG Eurofins, Ebersberg, Germany) with BbsI restriction overhang, hybridized and inserted into the BbsI-digested vector pX330-NeoR as previously described [55,56]. Correct cloning was verified by sequencing with primer HU6-F (ATAATTTCTTGGGTAGTTTGCAG). RK13-sgms1 KO cell lines were generated by co-transfection of all four sgRNA-containing pX330-NeoR plasmids (1.5 µg per plasmid) into cells using calcium phosphate-co-precipitation [57]. Two days after transfection in 6 well dishes, cells were transferred to 10 cm plates (Corning, Sigma-Aldrich, Darmstadt, Germany) and selected in medium containing 500 µg/mL G418 (Invitrogen, Karlsruhe, Germany). Ten to 14 days after transfection cell colonies were picked by aspiration and tested for a biallelic KO by sequencing after PCR amplification of the targeted gene sequence using the control primers (Table 3) as described previously [56]. One cell clone, designated as RK13-sgms1 KO , which contained two deletions in exon 3, was further analyzed.
Cell Viability Assay. Cell viability was analyzed using the PrestoBlue™ Reagent (Thermo Scientific, Dreieich, Germany), a resazurin-based metabolic assay. For this, 1 × 10 4 cells were seeded in 90 µL volume in a black 96 well plate with a flat and clear bottom (Corning, Sigma-Aldrich, Darmstadt, Germany). At 24, 48 and 72 h after seeding 10 µL PrestoBlue TM Reagent was added to the cell culture supernatant and mixed. The samples were incubated for 30 min at 37 • C. For each time point, six replicates were used including six wells containing medium for background estimation. Before bottom-read measuring of fluorescence in a Tecan Reader at Ex 560 nm /Em 590 nm , the plate was agitated for 5 sec. Multiple reads per well (3 × 3) were performed using the i-control™ microtiter reader software. Blank-reduced raw data (fluorescence intensities) are given, while the statistics were done with standardized values. For standardization, the following formula was used: (measured value-mean)/standard deviation.
Oil Red O Staining. The intracellular lipid distribution was visualized by Oil Red O staining. For this, 5 × 10 5 cells were seeded on coverslips in a 6 well plate, incubated for 24 h and then fixed with 4% paraformaldehyde in 1× PBS for 15 min. Fixed cells were washed three times with 50 mM NH 4 Cl in 1× PBS and then incubated for 30 min in the same buffer to quench the free aldehyde groups. A 0.3% Oil Red O (Serva, Heidelberg, Germany) stock solution was prepared in 100% 2-propanol. The Oil Red O stock solution was diluted in A. dest (60% Oil Red O Stock) and filtered through a 0.45 µm syringe filter unit. After two washing steps with A. dest, the cells were incubated for 30 min in diluted Oil Red O staining solution. Subsequently, cells were washed five times in A. dest, nuclei were counterstained for 10 min with Hematoxylin solution (Medite, Burgdorf, Germany) and cells were mounted in a drop of aqueous mounting medium (Aquatex, Merck, Darmstadt, Germany). The cells were imaged using the Aperio ScanScope CS system (Leica, Wetzlar, Germany). Scale bars indicate 50 µm.
Membrane staining. For membrane staining, 5 × 10 5 cells were seeded on coverslips in a 6 well plate and treated for 24 h with CellBrite TM (Neuro DiO-based, 1:600 in cell culture medium, Hölzel Diagnostika, Köln, Germany) before fixation with 4% paraformaldehyde in 1× PBS for 15 min. Fixed cells were washed three times with 50 mM NH 4 Cl in 1× PBS and then incubated for 30 min in the same buffer. Membranes were directly visualized via the autofluorescence of the staining dye. The nuclei were counterstained with 300 mM DAPI for 5 min and cells were mounted in a drop of Kaiser's glycerol gelatin (Merck, Darmstadt, Germany). Samples were analyzed using a confocal laser scanning microscope (Leica DMI 6000 TCS SP5, 63× oil-immersion objective, NA = 1.4; Leica, Wetzlar, Germany). The Viruses 2021, 13, 1574 9 of 20 different cell lines were processed in parallel and all images were acquired with identical parameters (laser excitation, detector settings, and zoom factor). Representative single images were processed using the Fiji software version 1.52 [58,59]. Scale bars indicate 10 µm.
Quantification of the cellular fluorescence was performed using Fiji software as previously described [60]. The integrated density (IntDen) of 25 cells was measured from five independent fields of view (5 cells per view, 80 × 80 µm) and in each view, two regions without fluorescence were measured for background reduction. The corrected total cell fluorescence (CTCF) was calculated using the following formula: CTCF = IntDen -(area of selected cell × mean fluorescence of background).
In vitro replication studies. To test for PrV propagation, RK13 and RK13-sgms1 KO cells were seeded with 2 × 10 5 cells per 24 well and subsequently infected with wildtypelike PrV-1112 or PrV-gD -Pass at an MOI of 5. Cells and supernatants were harvested at different time points after infection (0 h, 4 h, 8 h, 12 h, 24 h and 30 h p.i.) and stored at -80 • C. To determine the infectious virus titer, samples were thawed and cell debris was removed by centrifugation (2 min, 14,000× g). The supernatant was serially diluted (10 −1 to 10 −6 ) and used to infect RK13 cells in 48 well culture plates. After incubation for 1 h, the inoculum was replaced by a semi-solid medium allowing only direct cell-to cell spread of the virus. Cells were fixed after 2 days with formaldehyde and stained with crystal violet. Virus plaques were counted in at least two wells and mean values were calculated as plaque forming units per milliliter (pfu/mL). Shown are mean values of three independent experiments with corresponding standard deviations.
Determination of efficiency of plating. To check for efficiency of plating, cells were seeded at 3 × 10 5 cells per 12 well and subsequently infected with PrV-1112 or PrV-gD -Pass each with 200 pfu per well. After incubation for 1 h, the inoculum was replaced by a semi-solid medium. Cells were fixed after 2 days and stained with 5-bromo-4-chloro-3-indolyl-ß-D-galactopyranoside (X-Gal). The efficiency of plating was determined by counting the number of plaques in comparison to the RK13 control, which was normalized to 100%.
Statistics. For each assay at least three independent experiments were performed. The statistical significance of the data was determined by statistical tests using GraphPad Prism version 8.1.0. A p-value ≤ 0.05 was considered significant and is presented in the figures in form of asterisks (* ≤0.05, ** ≤0.01, **** ≤0.0001).

Results
Design of the porcine CRISPR library. A broadly applicable porcine sgRNA library for genome-wide knockout screens was designed based on the genome assembly of S. scrofa 10.2 [46]. For each gene, preferentially four specific 20 nt-sgRNA sequences were chosen that met the following criteria: (i) no exact off-target matches elsewhere in the genome, (ii) on-target score > 0.2, maximum CFD score for an off-target match <0.2, (iii) separated from a higher ranked guide by at least 5% of the coding sequence and (iv) located within the first 5-65% of the coding sequence. Where more than four candidate guides met these criteria, on-target scores (rounded to one decimal place) and off-target scores were used to rank guides further.
A total of 20,598 genes were targetable, with four guides each for 20,582 of them. The final library designated as SsCRISPRko.v1 contains 83,381 sgRNAs and 1000 nontargeting controls (Supplementary Table S1). The sgRNA design was validated by testing specific sgRNAs against host genes previously described to be important for influenza virus infection [40,48,49]. For this, PK15 cells were transduced with lentiviral vectors targeting SLC35A1 (two sgRNAs) or IRF7 (one sgRNA) ( Table 1). Following puromycin selection and gDNA isolation, the genomic region surrounding the cutting site was amplified from both transduced and non-transduced cells and sequences were analyzed using the web tool TIDE. This web tool compares the sequencing trace around the CRISPR cut site in pools of edited versus unedited DNA to estimate the spectrum and frequency of small edits. This is achieved by analyzing the shifts in sequence signal from small insertions and deletions (InDels) around the cut site [61].
From the cutting site onwards, the sequence from edited cells consisted of a mixture of signals derived from unedited DNA and from sequences shifted by a different number of nucleotides due to InDels [61]. Results shown in Supplementary Figure S1 illustrate the cutting efficiency with InDels generated from each guide thereby validating the sgRNA design and efficiency.
SsCRISPRko.v1 library performance. The performance of the porcine library was analyzed by sequencing the plasmid stock and the control cells after selection and expansion. In the porcine library plasmid stock > 99.95% of the sgRNAs were detected in three independent sequencing replicates with a mean zero count of 39. The sgRNA distribution of the library is persuasive, measured by a mean GiniIndex of 0.06343 (n = 3) assessed with MAGeCK Count. This shows that the synthesized library has a great overlap with the computationally designed library, without major losses during cloning and amplification.
CRISPR knockout of genes, which are essential for cell survival, should result in the depletion of the corresponding sgRNA sequences in the gDNA derived from the control cell pool after selection and expansion. To further evaluate the performance of the library, an essential gene drop-out analysis was done by comparing the read counts from the control cell pool and the input plasmid set. Genes were ranked by fold change (FC) in control cells vs. plasmid stock and compared in an enrichment analysis using two reference lists of essential genes [52,53]. sgRNAs targeting essential genes were significantly depleted in the transduced control cells compared to the input plasmid stock in all three replicates as shown by enrichment in higher-ranked genes, ranked in ascending order of fold change, in gene set enrichment analysis (GSEA, implemented in MAGeCK GSEA, version 0.5.8). A representative violin plot of one screen, as well as the statistical details of all three screens are given in Supplementary Figure S2. Depletion of sgRNAs targeting essential genes compared to the input pool confirms successful targeting by the porcine CRISPR/Cas9 library and negative selection in the initial phase of cell culture prior to infection, indicating efficient library performance.
Identification of SMS1 as entry factor for PrV. Before starting the genome-wide CRISPR screen, the knockout efficiency of Cas9 in PK15 cells was tested. Cells were transduced with lentiCRISPRv2 expressing a sgRNA targeting EGFP (Lenti-CRISPRv2 EGFP-sgRNA ) or with control (Lenti-CRISPRv2 empty ) and then selected with puromycin (1.25 µg/mL) until the non-transduced control cells were dead. After selection and expansion, cells were subsequently transduced with lentivirus expressing EGFP. Knockout efficiency of Cas9 was monitored by reduction of the EGFP signal using flow cytometry 3 days post transduction. We observed more than 97% reduction of EGFP positive cells compared to lentiCRISPRv2 empty transduced cells (Supplementary Figure S3). This indicated that PK15 cells are suitable tools for the genome-wide CRISPR/Cas9 screen.
The workflow for the genome-wide screen is outlined in Figure 1A. The porcine lentiviral library was used to transduce PK15 cells at an MOT of 0.3, to enhance the probability that every cell is transduced with only one sgRNA. The transduced cells were selected with puromycin and expanded for ten to 14 days to achieve at least 500-fold coverage for each sgRNA as recommended [35]. From this cell pool, 2 × 10 7 cells were harvested as control and stored for further analysis. The remaining cell pool was then infected with either PrV-Ka or PrV-gD -Pass at an MOI of 0.5.
Interestingly, no cells survived the infection with PrV-Ka in repeated assays, even after a prolonged incubation of up to three months indicating that PrV can use a wide panel of different host cell factors for successful infection. Because of this, we tested a mutant, PrV-gD -Pass, which lacks the receptor-binding protein gD but regained infectivity after repeated passages in cell culture [17]. Mutations in gB and gH have been shown to be involved in the gD-independent entry pathway but the exact mechanism is yet unknown [21]. Our genome-wide CRISPR/Cas9 approach should shed more light on this mode of infectious entry.
After infection with PrV-gD -Pass single cell colonies were visible after 10 days post infection (p.i.). These cell colonies were harvested 14 to 16 days p.i., pooled and one half of the cell pool was saved as 'survivors #1' for gDNA isolation and sequencing, while the other half was reinfected with PrV-gD -Pass at an MOI of 0.5. 'Survivors #2 were harvested 12 to 14 days after the second round of infection. The screening was done in three independent assays. For each assay, gDNA of all cell pools (control cells before infection, survivors #1 and #2) was isolated in parallel to avoid bias and used for PCR to amplify the integrated CRISPR regions. PCR products were purified and subjected to Ion Torrent sequencing.
Sequencing data were analyzed using the MAGeCK platform [51] which identified the significantly enriched sgRNAs present in the gDNAs of survivors compared to the non-infected control cells. Based on a robust rank aggregation (RRA) algorithm the absolute top hit candidate, with all four sgRNAs found enriched was identified as sphingomyelin synthase 1 (SMS1, sgms1) ( Figure 1B), a protein involved in sphingolipid metabolism. The other genes among the top 10 were found with a significantly higher and therefore less stringent RRA score and varied between the different independent assays, clearly separating them from the top hit SMS1.
Generation of a RK13-sgms1 KO cell line. Based on results from three independent screenings, we set out to validate the top hit by analyzing the role of SMS1 in our standard cell line, rabbit kidney cells (RK13). Although RK13 cells are not the natural host cells for PrV, they propagate the virus to high titers within a short time, are easy to transfect and no differences in the PrV replication cycle compared to cells of pig origin are evident.
Sgms1 knockout RK13 cell lines were generated by cloning the four sgRNA specific oligonucleotides targeting the rabbit homologue of SMS1 into the Cas9 encoding vector pX330-NeoR [55]. Two of the four sgRNA sequences were identical to those present in the porcine library. All four plasmids were co-transfected into RK13 and cells were selected with 0.5 mg/mL G418. Single G418-resistant cell clones were picked approx. 14 days post transfection and the gDNA was isolated. The targeted sgms1 gene region was amplified by PCR with specific primers (Table 3), cloned and sequenced as described before [56]. The cell line, which was chosen for further analyses, carried two deletions comprising 17 nt and 364 nt in exon 3 of sgms1, removing amino acids 15 to 182 and introducing a frame shift (Figure 2A). Direct sequencing of the PCR product uncovered the modification with no sequence bias, already indicating a biallelic knockout. To further validate this, the PCR product was cloned into the vector pBluescript and the inserts of 10 randomly selected plasmid clones was determined, which is generally accepted for verifying CRISPRengineered modifications [62]. All cloned PCR products derived from RK13-sgms1 KO cells revealed the same deletions further showing that both alleles carry the identical modification ( Figure 2B).
Western blot analysis using an SMS1-specific antibody showed absence of the approx. 38 kDa signal in RK13-sgms1 KO compared to RK13 cell lysates ( Figure 2C) confirming the successful knockout of full-length SMS1 also on protein level. No smaller sized bands were detectable in RK13-sgms1 KO cell lysates, but the serum used was raised against the N-terminal 120 amino acids of SMS1. In case translation would start at a downstream codon, the truncated isoforms would not be detected by the specific antiserum. However, the corresponding protein product would lack the conserved sterile alpha motif (SAM) and at least the first of the predicted six transmembrane domains. The approx. 45 kDa signal detectable in all cell lysates is probably due to background staining. (Figure 2A). Direct sequencing of the PCR product uncovered the modification with no sequence bias, already indicating a biallelic knockout. To further validate this, the PCR product was cloned into the vector pBluescript and the inserts of 10 randomly selected plasmid clones was determined, which is generally accepted for verifying CRISPR-engineered modifications [62]. All cloned PCR products derived from RK13-sgms1KO cells revealed the same deletions further showing that both alleles carry the identical modification ( Figure 2B).  Characterization of RK13-sgms1 KO cells. The metabolic activity of RK13 and RK13-sgms1 KO cells was measured to test whether the SMS1 knockout affects cell proliferation and viability. For this, the resazurin-based PrestoBlue TM reagent was used. RK13 and RK13-sgms1 KO cells were seeded in 96 well plates, and cell viability was measured at 24 h, 48 h and 72 h, respectively. No significant differences in fluorescence intensity between the parental and the RK13-sgms1 KO cells were found ( Figure 3A) indicating that absence of SMS1 does not impair cell viability in presence of FCS.
Since SMS1 plays a crucial role in sphingolipid metabolism [63] and sgms1 deficiency was described to unbalance sphingolipid levels [64], we tested for an effect in the RK13-sgms1 KO cells. To detect the presence of neutral lipids, RK13 and RK13-sgms1 KO cells were fixed and stained with Oil Red O. In comparison to native RK13, RK13-sgms1 KO cells showed an accumulation of lipid dots in the cytoplasm ( Figure 3B) indicating that the lipid metabolism in RK13-sgms1 KO cells is altered.
To analyze whether SMS1 knockout results in differences in intracellular membranes we used CellBrite Dye, which labels preferentially cytoplasmic membranes after longer incubation times. Here, we stained the RK13 and RK13-sgms1 KO cells for 24 h before imaging by confocal microscopy. Imaging was performed with constant settings for both cell lines. As shown in Figure 3C,D, fluorescence intensity is significantly higher in the SMS1 knockout cells indicating an accumulation of cytoplasmic membranes. sgms1KO cells was measured to test whether the SMS1 knockout affects cell proliferation and viability. For this, the resazurin-based PrestoBlue TM reagent was used. RK13 and RK13-sgms1KO cells were seeded in 96 well plates, and cell viability was measured at 24 h, 48 h and 72 h, respectively. No significant differences in fluorescence intensity between the parental and the RK13-sgms1KO cells were found ( Figure 3A) indicating that absence of SMS1 does not impair cell viability in presence of FCS.  Nuclei were counterstained with DAPI, and fluorescence was detected with a confocal laser-scanning microscope with constant settings. Scale bars indicate 10 µm. (D) Fluorescence was quantified by calculation of the corrected total cell fluorescence. Significant differences were calculated by two-way ANOVA followed by Sidak's multiple comparison test (A, ns: not significant) or by an unpaired t-test (D, **** ≤0.0001).
SMS1 is required for efficient entry of PrV-gD -Pass. To analyze the importance of SMS1 on PrV replication, RK13 and RK13-sgms1 KO cells were infected with either the wildtype-like PrV-1112 (ßGal + ) or PrV-gD -Pass (gD -, ßGal + ) at an MOI of 5. Cells and supernatant were harvested at different time points p.i. and progeny virus titers were determined on parental RK13 cells. As depicted in Figure 4A, titers for PrV-1112 derived from RK13-sgms1 KO cells were approx. 10-fold lower, while titers of PrV-gD -Pass were more than 100-fold reduced compared to the parental RK13 cells.
To test whether this effect is congruent with a defect in entry, cells were infected with approx. 200 pfu/well under plaque assay conditions, stained with XGal solution and plaques were counted two days p.i. As shown in Figure 4B, efficiency of plating was approx. 50% reduced for PrV-1112, while only 10-20% of plaques were observed for PrV-gD -Pass compared to the parental RK13 cells (set to 100%). This indicates that SMS1 plays a prominent role in entry related pathways of PrV infection.
To test whether this effect is congruent with a defect in entry, cells were infected with approx. 200 pfu/well under plaque assay conditions, stained with XGal solution and plaques were counted two days p.i. As shown in Figure 4B, efficiency of plating was approx. 50% reduced for PrV-1112, while only 10-20% of plaques were observed for PrV-gD -Pass compared to the parental RK13 cells (set to 100%). This indicates that SMS1 plays a prominent role in entry related pathways of PrV infection.  To test whether the observed effect is indeed due to the defect in SMS1 expression, we reconstituted RK13-sgms1 KO cells with SMS1. RNA of RK13 cells was isolated, reverse transcribed and the sgms1 open reading frame was amplified with specific primers ( Table 4). The 1.2 kb PCR product was cloned into pLV-X (pLV-sgms1 + ) and RK13-sgms1 KO cells were transduced either with the empty vector (pLV-X) or with pLV-sgms1 + . Transduced cells were selected in medium containing puromycin. In comparison to cells transduced with the empty control lentivirus, the 38 kDa protein was detectable in the restored RK13-sgms1 KO /sgms1 + , albeit with reduced intensity compared to parental RK13 cells ( Figure 2B). Since we used the original, unmodified sgms1 coding sequence, it might be speculated that Cas9 in combination with the sgRNAs is still partially effective and modifying the exogenous gene. Despite this reduced expression, efficiency of plating was increased for PrV-gD -Pass to approx. 90% in RK13-sgms1 KO /sgms1 + cells but remained at a 50% level for wildtype-like PrV-1112 ( Figure 4B) indicating that SMS1 plays a prominent role when the gD-mediated entry pathway is blocked.

Discussion
We generated a porcine-specific library, designated as SsCRISPRko.v1 to investigate host factors important for PrV infection. Although the function of most of the viral proteins in the PrV replication cycle is already understood to some detail [65,66] knowledge on host cell factors involved in efficient virus propagation is still lagging. The generated SsCRISPRko.v1 library contains 83,381 sgRNAs and 1000 control sgRNAs targeting a total of 20,598 genes. Sequencing, computational and experimental analyses using sgRNAs specific for host cell factors important for influenza virus infection and the depletion of sgRNAs targeting essential genes in the modified control cell pool confirmed a good performance of our newly designed library.
Recently, in parallel to our study, a genome-scale porcine CRISPR/Cas9 knockout library designated as PigGeCKO, with sgRNAs targeting 17,774 protein-coding genes, 11,053 long non-coding RNAs and 551 micro-RNA was used to identify host factors associated with Japanese encephalitis virus (JEV) replication [67]. While PigGeCKO requires Cas9-expressing cells, our library is cloned in a vector which also carries the genetic information for the Cas9 nuclease, which might render our library more broadly and easily applicable.
In our first screens, we used wildtype PrV-Ka for selection, but no cells survived. The same was true when we infected the cells with the highly attenuated vaccine strain Bartha [68]. This is in line with the perception that PrV is very flexible, can adapt to a number of different settings and can use multiple host cell proteins for successful propagation [4]. In contrast, infection of our porcine CRISPR library modified PK15 cell pool with a virus mutant, which is infectious in the absence of gD, PrV-gD -Pass, resulted in infection-resistant cells. Sequence analysis of surviving cells from three independent assays uncovered that all four sgRNAs directed against sgms1, which encodes the enzyme sphingomyelin synthase (SMS) 1, were found highly enriched and ranked on the top of the hit list. RRA values clearly separated the sgms1 hit from the other genes in the top 10 list, which in addition differed between the individual screens and were therefore not further analyzed.
SMSs synthesize sphingomyelin (SM) by transfer of phosphocholine from phosphatidylcholine to ceramide, yielding diacylglycerol as a side product [69]. Sphingolipids, such as SM and ceramides are essential and ubiquitous constituents of membranes and have distinct roles in a wide variety of cellular functions, e.g., in barrier function or the regulation of signaling pathways [63]. Proteins encoded by sgms1 (SMS1) and sgms2 (SMS2), which are distinct in their expression pattern and subcellular localization, have been identified with SM synthesis activity in mammalian cells [64]. While SMS1 is located in the lumen of the Golgi apparatus, SMS2 is present mainly in the outer leaflet of the plasma membrane [70]. Both contribute to SM production [71] and are required for SM homeostasis and cell growth. Since SMSs are active at a key point in sphingolipid synthesis, blocking the activity not only influences levels of SM and ceramide, but also of other related sphingolipids like glycosphingolipids (GSL), sphingosine and spingosin-1-phosphate [72,73].
To validate the sgms1 hit in a different host cell background, we generated a sgms1knockout cell line of our standard rabbit kidney (RK13) cell line. The RK13-sgms1 KO cell clone isolated after targeted CRISPR/Cas9 mutagenesis carries two deletions, comprising 17 nucleotides and 364 nucleotides, respectively, affecting codons 15 to 182 of the rabbit homologue and additionally resulting in a frame shift in exon 3. Absence of a specific approx. 38 kDa protein could be demonstrated by western blot analysis using a monospecific antiserum raised against the N-terminal 120 amino acids of the human SMS1 homolog. SMS homologs are found throughout the animal kingdom and constitute multipass membrane proteins with six predicted transmembrane domains, with the N-and the C-terminal domain reaching into the cytosol [69]. Unfortunately, reagents detecting the C-terminal part were not available. However, if a truncated sgms1 gene product would be expressed in RK13-sgms1 KO cells it would lack at least one-third of the N-terminal domain including the sterile alpha motif (SAM) and the first transmembrane domain. Deletion of the SAM domain of human SMS1 resulted in a significant reduction of SM synthesis [74] indicating that even if a truncated SMS1 product is expressed it will most likely be functionally impaired.
Although SMSs play an important role for cell growth and propagation, RK13-sgms1 KO cells are viable and exhibited no impairment in metabolic activity and cell division. Viability was measured in the presence of fetal calf serum and SM levels present seem to be sufficient to compensate the defect in SMS activity [64]. To test for a functional effect in our RK13-sgms1 KO cell line, we analyzed whether lipid distribution is altered. Both, Oil Red O, which predominantly stains neutral lipids and CellBrite labeling of intracellular membranes, showed a significantly higher presence of lipids in RK13-sgms1 KO compared to the parental RK13 cells indicating that the knockout was successful. While we specifically detected all four sgRNAs directed against the sgms1 gene in our screening, none against SMS2 were found to be enriched after infection with PrV-gD -Pass.
Infection of the RK13-sgms1 KO cell line demonstrated the importance of SMS1 for entry of PrV-gD -Pass. Only approx. 10% to 20% of plaques were formed on RK13-sgms1 KO compared to RK13 cells. This defect could be restored by heterologous expression of wildtype SMS1 after lentivirus transduction validating that the observed defect is indeed due to inactivation of the cellular sgms1 gene. Although the number of plaques was also reduced for wildtype-like PrV-1112, restoration of SMS1 expression could not rescue this defect indicating that either the expression level of SMS1 was too low to compensate for the defect in PrV-1112 or that the reduction is not related to the impaired SMS1 function. This needs further investigation. Interestingly, PrV entry was inhibited by 54% in cells treated with sphingomyelinase [75]. Due to their distinct expression pattern and subcellular localization [64], we do not expect a SMS2 knockout or a SMS1/SMS2 double knockout to yield comparable or additive inhibitory effects on PrV entry pathways.
Sphingomyelin and other sphingolipids have previously been described to have an impact on attachment and entry of different viruses, e.g., JEV, HIV and Ebola virus [76][77][78]. Interestingly, an altered turnover of SM and GSL was observed in PrV and HSV infected cells [75,[79][80][81][82]. In addition, sgms1 was found in a CRISPR-based screen analyzing bovine cells in BoHV-1 infection ( [83], bioRxiv preprint). However, the study did not analyze SMS1 in more detail.
It has been proposed that lipid rafts, which are highly enriched in cholesterol and sphingolipids, are structured platforms that carry specific proteins and are preferentially used by viruses in various steps during replication [84][85][86]. SMS1 deficiency was shown to significantly decrease SM levels in plasma membranes [73,87]. Many viruses e.g., influenza virus or vesicular stomatitis virus [88,89] were described to induce and use lipid rafts as docking point for subsequent invagination from the plasma membrane in intracellular clathrin-coated vesicles [90,91]. Alphaherpesvirus receptor proteins like HVEM or nectin-1 are not associated with lipid rafts but HSV-1 gB was described to rapidly mobilize lipid rafts to serve as a platform for entry [92]. Although the pathway of PrV-gD -Pass entry has not yet been elucidated, the association of gB with lipid rafts on the cell surface might play a crucial role for gD-independent entry.
Although loss-of-function CRISPR/Cas9 screenings are powerful tools to identify cellular factors relevant for virus replication, this approach faces limitations, especially if factors are essential for host cell survival or if multiple factors exist-as this appears to be the case for PrV and probably also for other (alpha)herpesviruses. However, the data we present highlight the power of CRISPR screens using our newly generated library SsCRISPRko.v1 allowing further unraveling of pig-pathogen interactions, an important livestock species. Thus, our porcine library SsCRISPRko.v1 and the recently published PigGeCKO library [67] are valuable tools for CRISPR screens to identify cellular factors relevant for the replication of a multitude of porcine pathogens including swine influenza virus (SIV), porcine epidemic diarrhea virus (PEDV), porcine circoviruses (PCV) or the emerging African swine fever virus (ASFV).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study and not included in the supplement are available on request from the corresponding author.