Porcine Complement Regulatory Protein CD46 Is a Major Receptor for Atypical Porcine Pestivirus but Not for Classical Swine Fever Virus

Pestiviruses comprise animal pathogens such as classical swine fever virus (CSFV) and bovine viral diarrhea virus (BVDV) that cause notifiable diseases with great economic impact. Several additional pestivirus species affecting animal health were recently identified, including atypical porcine pestivirus (APPV).

T he genus Pestivirus belongs to the family Flaviviridae and currently comprises 11 different species termed Pestivirus A to K and a growing number of putative new species (1). Pestiviruses such as bovine viral diarrhea virus (BVDV) and classical swine fever virus (CSFV) are RNA viruses of outstanding veterinary and economic relevance, being causative agents of notifiable diseases (2,3). Porcine host species are significantly implicated in pestivirus biology. CSFV (Pestivirus C) is the only established pestivirus selectively infecting porcine hosts. Moreover, pigs are susceptible for several pestivirus species, including ruminant pestiviruses like BVDV type 1 and 2 (Pestivirus A and B) and Border disease virus (BDV, Pestivirus D). In addition to CSFV, three additional porcine pestivirus species were discovered, comprising atypical porcine pestivirus (APPV, Pestivirus K), Bungowannah virus (BuPV, Pestivirus F), and LINDA virus (LindaV, species not yet approved). These three pestivirus species can cause severe diseases in young piglets (4)(5)(6). It appears likely that infections of pigs with LindaV and BuPV were results of rare or even unique spillover infections from so far unknown reservoir hosts since subsequent attempts to detect these pathogens in domestic pig and wild boar populations failed (5)(6)(7)(8). In contrast to BuPV and LindaV, APPV was found to be highly abundant in domestic pigs from many countries worldwide, and wild boar must be considered a natural reservoir host (9,10). Pestivirus genomes that are most closely related to APPV were detected in bats, which together with rodents represent the first non-ungulate hosts harboring novel pestivirus species (11)(12)(13). So far, it is not known whether pigs are the only host for APPV, and investigation of in vitro cell tropism was restricted due to lack of an APPV cell culture isolate. In contrast to that of BuPV and LindaV, APPV isolation and propagation is highly inefficient on established porcine cell lines, and virus isolation was only reported on embryonic porcine kidney epithelial cells (SPEV cells) with very low infectious titers (14,15).
Although it is a crucial step in the viral replication cycle, the entry process of pestiviruses is still poorly understood. Pestiviruses enter the host cell by receptor-mediated endocytosis, likely via a multistep process using different host cell factors (16,17). Historically, bovine complement control protein CD46 (CD46 bov ) was identified as a receptor for BVDV by a set of neutralizing monoclonal antibodies (mabs) that were directed not against the viral antigen but against the cellular protein (18,19). CD46 was previously named a "pathogen's magnet," as it is used as a receptor by diverse bacterial and viral pathogens, including measles virus (MeV) vaccine strains, certain human adenoviruses, and herpesviruses such as human cytomegalovirus (20,21). Expression of CD46 bov on porcine cells increases the susceptibility to BVDV (18). Nevertheless, BVDV was not able to infect nonpermissive cells substituted with CD46 bov . As CD46 bov is an almost ubiquitously expressed molecule, it remained unclear what determines tropism of BVDV for certain tissues. One possible explanation would be the use of coreceptors or different CD46 bov variants by the virus (20,22). Evidence for an additional BVDV receptor was provided by a study which demonstrated cellular entry independent from binding of E2 to CD46 bov (23). Subsequently, porcine CD46 (CD46 pig ) was suggested to play a role in the entry process of CSFV, as CD46 pig -specific monoclonal antibodies (mabs) diminished infectivity to some degree (24).
In the present study, we systematically investigated the role of CD46 pig for entry of the porcine pestivirus species APPV, CSFV, and BuPV. The results of our study show that CD46 pig is essential for efficient entry of APPV, while CSFV and BuPV rely on different host cell factors for cellular entry.

RESULTS
Adaptation of atypical porcine pestivirus (APPV) to cell culture conditions. Determination of the complete polyprotein coding sequence of APPV from the 100th passage (APPV P100 ) using next-generation sequencing revealed a total of 23 nucleotide exchanges compared to the original sequence obtained directly from the sample material (GenBank MF167291). These genome alterations consist of 7 synonymous and 16 nonsynonymous mutations (Fig. 1B). Three synonymous and three nonsynonymous substitutions, namely, H330Q (E rns ), N751K (E2), and D752N (E2), were located within the predicted glycoprotein coding regions (Fig. 1B). Sanger sequencing of viral genomes obtained from different APPV passages confirmed that these mutations were acquired during passaging, between the 45th and 100th passages. APPV stock from the 17th passage (APPV P17 ) did not show any mutations within the predicted glycoprotein coding regions. In five independent repetitions of passaging, the same synonymous and nonsynonymous mutations occurred within the predicted glycoprotein coding regions and correlated with an increased number of infected SPEV cells, suggesting their importance for the adaptation of APPV to cell culture conditions.
Permissivity of different cell lines to atypical porcine pestivirus (APPV). We investigated the permissivity of different porcine and nonporcine cell lines to APPV infection using the newly established cell culture-adapted APPV strain (APPV P100 ). All porcine cell lines except 38A 1 D (lymphoma) were susceptible to APPV P100 infections, however, to various extent (Fig. 1C). This is in line with the situation in vivo, as APPV can be detected in cells of diverse tissues and different cell types of infected piglets (25). APPV P100 replicated most efficiently in SPEV (kidney) cells, resulting in an almost completely infected monolayer of SPEV cells at 72 h postinfection (p.i.) after infection with an MOI of 1. In contrast, on PK15 (kidney) cells, which are routinely used for propagation of CSFV, only few APPV P100 -infected foci were observed under equal conditions. Similar to that on PK15 cells, inefficient and limited viral replication of APPV P100 was observed on porcine SK6 (kidney), NPTr (neonatal trachea), IPECJ2 (jejunum), and ST (testis) cells. These results demonstrate so far unknown differences in the permissivity of different porcine cell lines to APPV infection (Fig. 1C). (B) Schematic representation of the 100th passage of APPV isolate L277 (APPV P100 ) genome including locations of synonymous (blue) and nonsynonymous mutations (red). (C) The cell culture-adapted APPV P100 was used to infect different porcine cell lines with an MOI of 1. All porcine cell lines except 38A 1 D were permissive to APPV, showing either nearly complete APPV-infected monolayer (SPEV) or only infected foci (PK15, SK6, NPTr, IPECJ2, and ST). The respective noninfected cells (NIC) served as a control. Immunofluorescence staining was performed at 72 h p.i. using porcine APPV-specific antiserum (red) and DAPI (blue).
Several nonporcine cells (MDBK, CRFK, RK13, Vero76, Lovo, HypNi/1.1, EidNi/41; Table  1) were also tested for their permissivity to APPV, since BuPV was shown to have a broad cell tropism in vitro and sequences of pestiviruses closely related to APPV were detected in diverse rodent and bat species (11)(12)(13). None of the investigated nonporcine cell lines was permissive to APPV P100 (data not shown). Thus, the data provide no evidence that APPV can efficiently infect and replicate in nonporcine cell lines.
Characterization of CD46 pig expression patterns in porcine cell lines. The observed differences in permissivity of porcine cell lines to APPV infection prompted us to characterize the CD46 pig surface expression level of these cells, as CD46 bov was previously shown to represent a major receptor for the related ruminant pestivirus BVDV (18). In addition, CD46 pig was suggested to be a major entry factor of CSFV (24). With the exception of two cell lines, immunofluorescence staining of CD46 pig using commercially available CD46 pig -specific mabs revealed that most of the porcine cell lines, including SPEV and PK15, showed similar surface expression of CD46 pig ( Fig. 2A). The only porcine cell line (38A 1 D) nonpermissive to APPV P100 was also the only one not expressing any CD46 pig . A neonatal porcine tracheal cell line NPTr revealed expression of CD46 pig only by some clusters of cells ( Fig. 2A). CD46 pig -positive NPTr cells were particularly located in the upper layers of cells that grew in multilayer clusters. Costaining of APPV P100 and CD46 pig showed that APPV P100 -positive cells were located only on the CD46 pig -positive cell clusters (Fig. 2B).
Sequence analysis of CD46 pig obtained from SPEV and PK15 cells revealed no differences in the deduced amino acid sequences of the four ccp domains. In 38A 1 D cells, no mRNAs coding for the individual ccp domains (ccp1 to ccp4) were detectable by different RT-PCRs (Fig. 2C), demonstrating the absence of CD46 pig coding transcripts. This confirms the negative immunofluorescence results when staining for CD46 pig . Taken together, these data provide strong evidence that 38A 1 D cells are deficient for CD46 pig expression and highlight the importance of CD46 pig for the APPV entry process.
Generation and characterization of CD46 pig knockout cell lines. To investigate the CD46 pig dependency in the entry process of different porcine pestiviruses, CD46 pigdeficient cell lines were generated by CRISPR/CAS9 technology and lentiviral transduction of CD46 pig -specific guide RNAs. Two different guide RNAs (gRNAs CD46-2 and -7) were designed to avoid misinterpretation due to off-target effects (Fig. 2C, Fig. 3B, and Table 2). Both oligonucleotides were located very close to each other targeting the genomic region encoding the ccp1 domain of CD46 pig , the receptor-binding site of CD46 bov for the ruminant pestivirus BVDV (26). Two different CD46 pig knockout SPEV cell lines (SPEVDCD46) were obtained with gRNA CD46-2 (SPEVDCD46 clone 2) and gRNA CD46-7 (SPEVDCD46 clone 7), respectively (Fig. 3). Additionally, one CD46 pig knockout PK15 cell line was generated with gRNA CD46-2 (PK15DCD46 clone 2, Fig. 3) prompted by the observed differences in permissivity of wild-type (WT) SPEV and PK15 cells. All CD46 pig knockout cell lines were cloned biologically to obtain clonal cell lines. Immunofluorescence staining using mabs against CD46 pig confirmed that all generated knockout cell lines were phenotypically CD46 pig -deficient (Fig. 3A). Sequence analysis of plasmids containing cloned PCR products of the N-terminal region of CD46 pig ccp1 domain confirmed different genetic alterations in this region, due to random repair mechanisms of the cell. Genetic characterization of at least 10 independent CD46 pig PCR plasmid clones demonstrated the presence of altered CD46 pig coding sequences in a diploid set of chromosomes for each cell clone. Each one of the altered alleles revealed different frameshifts within the N-terminal region of the ccp1 domain (Fig. 3B). The SPEVDCD46 clone 2 revealed to have a small deletion of five nucleotides on one allele resulting in a stop codon and codes for only three unchanged amino acids at the N terminus of ccp1. In the other allele, an extended large deletion of 289 nucleotides was found, affecting the parts coding for the SP and the complete ccp1 sequence (195 nucleotides) as well as the N-terminal half of ccp2 (94 of 180 nucleotides, Fig. 3B). Both alleles of SPEVDCD46 clone 7 revealed to have deletions of seven and eight nucleotides, respectively, resulting in frameshifts and stop codons (Fig. 3B). In consequence, only the first six amino acids following the signal peptide (SP) of CD46 pig remained unchanged in SPEVDCD46 cell clone 7 (Fig. 3B). In the PK15DCD46 clone 2, in one allele, a single nucleotide insertion results in an immediate stop codon. The other allele contains the same extended deletion as observed in SPEVDCD46 clone 2 affecting the SP, ccp1, and ccp2 sequences. In consequence, in addition to the 38A 1 D cell line that was identified to be naturally deficient for CD46 pig , three genetically engineered CD46 pig knockout porcine cell lines were successfully generated with the aim to determine the relevance of CD46 pig in the entry of porcine pestiviruses.
Relevance of CD46 pig for the entry of porcine pestiviruses. A comparative analysis regarding the permissivity of WT and CD46 pig knockout cell lines was performed using APPV, a set of CSFV strains, and BuPV. Early (non-cell-culture-adapted APPV P17 ) and late (cell culture-adapted APPV P100 ) passages of the APPV isolate were used. Different CSFV isolates that are representatives of genotypes 1 and 2 with different virulence properties were selected, including the attenuated vaccine strain Riems (gt. 1.1), the highly virulent strain Koslov (gt. 1.1), the strains Paderborn (gt. 2.1), Diepholz (gt. 2.3), and Alfort-Tübingen (AlfT) (gt. 2.3), which was rescued from a reverse genetic system (27,28). Sequence analysis of glycoprotein E rns coding regions revealed that selected CSFV isolates do not possess a point mutation (Ser 476 to Arg 476 ) which was previously reported to be responsible for cell culture adaptation (29). CSFV isolate Paderborn naturally possesses Arg 476 as also described earlier (GenBank ADI58615; AAL68894).
Infection of WT SPEV and SPEVDCD46 cells revealed a strong impact of CD46 pig expression on permissivity to APPV P100 at 16 or 72 h p.i. (Fig. 4 and 5 and Fig. 6B). This effect was observed on both SPEVDCD46 cell lines and confirmed the results observed with the naturally CD46 pig -deficient 38A 1 D lymphoma cell line. Nevertheless, single small foci of APPV P100 -infected cells were evident on both SPEVDCD46 cell lines. Viral titers were determined 72 h p.i. (Fig. 5A). High titers were detected after de novo infections with APPV P100 reaching approximately 1 Â 10 6 50% tissue culture infective dose (TCID 50 )/ml in the supernatant of WT SPEV cell line. In contrast, no infectious titers could be determined in the supernatants of the two different CD46 pig knockout cell lines. Determination of APPV P100 genome equivalents revealed 11 to 22 times less viral RNA in the SPEVDCD46 cells than in WT SPEV at 72 h p.i. (Fig. 5B). In the case of PK15 cells, the dependency on CD46 pig for APPV P100 permissivity was less pronounced since the WT PK15 cell line already displayed a low permissivity comparable to that of SPEVDCD46 cells (Fig. 4 and 5). Quantification of APPV P100 genome equivalents in three independent experiments revealed on average 90 times lower genome equivalents in PK15 cells than in SPEV cells (Fig. 5B). Accordingly, no infectious titers could be determined in the supernatants of the PK15 WT and knockout cell lines due to the low viral load. Interestingly, WT PK15 cells contained an approximately 10 times smaller amount of APPV P100 genomes than did the two SPEVDCD46 cells, suggesting a less efficient RNA replication in PK15 cells.
Due to the lack of highly specific APPV mabs and small amounts of viral antigen early after infection, immunofluorescence staining of APPV was not possible as early as 16 h p.i. In order to perceive the impact of CD46 pig at an early time point of the infection, a more sensitive fluorescence in situ hybridization (FISH) assay was performed on APPV P100 -infected WT and CD46 pig knockout cell lines (SPEVDCD46 clone 2 and PK15DCD46 clone 2) at 16 h p.i. (Fig. 6B). Results of the FISH assay were in line with the immunofluorescence staining of infected cells at later time points (72 h p.i.). APPV P100 genomes were detected only in very few CD46 pig knockout cells (SPEVDCD46 clone 2 and PK15DCD46 clone 2), whereas most of the CD46 pig -positive WT SPEV and PK15 cells were highly positive for APPV P100 genome (Fig. 6B). Quantitative analyses of the images revealed equal amounts of positive signal in both CD46 pig knockout cell lines, on average (SPEVDCD46 clone 2 and PK15DCD46 clone 2). Positive signals detected in WT SPEV cells were 1.9-fold higher than those in WT PK15 cells and 5.9-fold higher FIG 4 Relevance of CD46 pig for the entry of porcine pestiviruses. Wild-type (WT) SPEV and PK15 as well as CD46 pig knockout cell lines (SPEVDCD46 clones 2 and 7 and PK15DCD46 clone 2) were infected with APPV P17 , APPV P100 , BuPV, and CSFV strains Alfort-Tübingen (AlfT), Diepholz, Riems, Koslov, and Paderborn at an MOI of 1, respectively. Immunofluorescence staining was performed at 72 h p.i. using porcine APPV-specific antiserum, a porcine BuPV-specific antiserum, and a mab against CSFV, respectively. A strong reduction of APPV infection is evident on all SPEVDCD46 cell lines in comparison to that on SPEV cells. PK15 cells display significantly lower permissivity to APPV P100 compared to that of SPEV cells. Non-culture-adapted APPV P17 obtained from early passage revealed the same CD46 pig dependency as the culture-adapted variant (APPV P100 ). With regard to infections with CSFV and BuPV, there are no differences in permissivity between the WT and the CD46 pig knockout cell lines. than those in both CD46 pig knockout cell lines. The difference between the WT PK15 cells and the PK15DCD46 2 cells was 3.1-fold.
APPV harvested from early passage 17 (APPV P17 ) showed only inefficient replication on WT SPEV and PK15 cells. Nevertheless, the impact of CD46 pig expression on the permissivity to APPV P17 was obvious when WT and CD46 pig knockout cells were compared (Fig. 4). Multiple infected foci were detected on WT SPEV and PK15 cells, whereas only few single infected cells were present on CD46 pig knockout cell lines at 72 h p.i. (Fig. 4). Due to inefficient replication of APPV, infectious titers could not be determined in the supernatants of either WT or CD46 pig knockout cell lines.
In contrast to APPV, no visible differences between WT and CD46 pig knockout cell lines were observed with regard to the permissivity to CSFV and BuPV at 16 h and 72 h p.i. (Fig. 4 and Fig. 6A). Viral titers of CSFV AlfT determined at 72 h p.i. from cell culture supernatants of infected PK15 WT and CD46 pig knockout cell lines were comparable (Fig. 5). Titers obtained from the infected CD46 pig knockout cell lines ranged between Wild-type (WT) SPEV and PK15, as well as CD46 pig knockout cell lines (SPEVDCD46 clones 2 and 7 and PK15DCD46 clone 2), were infected with APPV P100 , CSFV Alfort-Tübingen (AlfT), and BuPV at an MOI of 1, respectively. (A) Supernatants were harvested 72 h p.i. to determine virus titers by using endpoint dilution assays in quadruplicates and in three repetitions. (B) Cells were collected at 72 h p.i. for RNA preparation and subsequent RT-PCR analysis. TaqMan based qRT-PCR assays were used for detection of CSFV and APPV genomes, whereas a SYBR green-based real-time RT-PCR was performed for detection of BuPV genomes. 30 ng total RNA was used per reaction. Samples collected from three individual experiments were tested in duplicates. Mean values with standard deviations are shown. APPV genome copy numbers obtained from WT cells are significantly higher compared to those from CD46 pig knockout cells (***, P , 0.0001, highly significant; *, P , 0.01, significant). CSFV and BuPV genome levels obtained from WT cells did not show significant differences compared to genome loads detected in infected knockout cells.

DISCUSSION
Although pestiviruses are of outstanding importance in veterinary medicine, molecular determinants of pestiviral entry are still only poorly understood. The entry process of pestiviruses is likely to be a multistep process, as it is known to be for the related human hepatitis C virus (HCV), which uses at least four different cellular factors (30). The cellular uptake of BVDV particles is mediated by clathrin-dependent endocytosis, and it was shown to be dependent on interaction between viral E2 and cellular CD46 bov (14,16,26). For BVDV, CD46 bov was identified to function as the main receptor, but there are several lines of evidence indicating that an additional receptor may exist (23,26,31,32). An entry mechanism of genetically distinct pestiviruses such as APPV as well as rodent and bat pestiviruses so far could not be investigated due to lack of virus isolates and established cell culture systems. Generation of a cell culture-adapted APPV isolate (APPV P100 ) allowed us for the first time to characterize this novel pestivirus species biologically (7). In this study, reproducible genomic alterations within the predicted E rns and E2 regions were identified in the culture-adapted APPV P100 . Their appearance correlated with an increased number of infected SPEV cells (Fig. 1A). Additional alterations were also identified in the nonstructural proteins p7, NS2, NS3, NS5A, and NS5B (Fig. 1B). To clarify the relevance of observed mutations throughout the APPV genome for cell culture adaptation and to elucidate the underlying mechanisms of improved in vitro replication, further experiments, e.g., by reverse genetics or viral pseudotypes, will be necessary.
Recently established, genetically distinct pestivirus species, including bat and rodent viruses, are apparently not restricted to ungulate hosts. Pestiviruses identified in different bat species are most closely related to APPV (Fig. 7A). The BuPV, another pestivirus species identified in pigs, shows a broad host cell tropism and is able to efficiently replicate in cells of bat and even in cells of human origin (33). Previous work showed that in vivo APPV differs from CSFV in tissue tropism (25). So far, host cell tropism of APPV could not be investigated due to inefficient and limited viral replication in vitro. The presented data revealed that APPV P100 obviously has a narrow host tropism and (within the cell lines used in this study) selectively infects cells of porcine origin (Fig. 1C). Different nonporcine cell lines permissive to BuPV (bat and human cells) were not permissive to APPV P100 (data not shown). Investigated cell lines of porcine origin showed major differences in their permissivity to APPV P100 (Fig. 1C). The embryonic porcine kidney epithelial cell line SPEV allowed most efficient replication of APPV P100 , which may explain reports of successful virus isolation only on this cell line (14,15). Differences in permissivity to APPV P100 prompted us to investigate the expression of CD46 pig . SPEV and PK15 cells, displaying major differences in permissivity to APPV P100 , did not differ in their CD46 pig expression pattern. Moreover, sequence analysis did not identify differences in the regions coding for the ectodomain of CD46 pig . Thus, the major differences in PK15 and SPEV cell permissivity to APPV P100 cannot be explained by genetic differences in the ccp1 domain, which was previously reported to bind glycoprotein E2 of BVDV (26). The molecular basis for restricted growth of APPV P100 in established porcine cell lines and differences in permissivity of CD46 pig -expressing porcine cells should be in the focus of future studies with the aim to establish a fully permissive cell culture system. All porcine cell lines investigated in this study expressed CD46 pig in a uniformly distributed pattern with two exceptions (NPTr, 38A 1 D, Fig. 2A). Remarkably, neonatal porcine tracheal cell line (NPTr) revealed to express CD46 pig only in the upper layers of cell clusters, putatively in dependence on the differentiation status. Costaining of APPV P100 and CD46 pig in these cells indicated that CD46 pig expression is a prerequisite for APPV P100 infection (Fig. 2B). Along this line, it is an interesting finding that the porcine lymphoma cell line 38A 1 D is naturally deficient for CD46 pig . 38A 1 D was the only porcine cell line tested in this study that was nonpermissive to APPV P100 , and the absence of CD46 pig in 38A 1 D cells might be responsible for this phenotype (Fig. 1C). Oddly, previous results showed that a CSFV isolate could be grown on 38A 1 D cells to titers 10 times higher than it could on established PK15 cells (34). Thus, the finding of 38A 1 D cells to be deficient for CD46 pig is remarkable since CD46 pig was proposed to be a major receptor of CSFV entry (24).
To elucidate the role of CD46 pig in the entry of porcine pestiviruses, porcine knockout cell lines incapable of expressing CD46 pig were generated by CRISPR/Cas9, based on SPEV and PK15 cell lines. All cell lines were characterized genetically and phenotypically to be deficient for CD46 pig (Fig. 3). Differences in permissivity of WT and CD46 pig knockout cells to APPV P17 were observed at 72 h p.i. (Fig. 4). Moreover, with the use of culture-adapted APPV P100 , the high impact of CD46 pig expression on permissivity to APPV could be observed as early as 16 h p.i. and also at 72 h p.i. (Fig. 4 and Fig. 6B). These results demonstrate that entry of APPV is highly dependent on CD46 pig . A similar strategy based on culture adaptation was followed for other viruses that were difficult to isolate and to propagate in vitro, e.g., the related HCV and its culture-adapted JFH-1 clone (35). Adaptive mutations in glycoprotein regions can result in improved binding affinity of the viral ligands with attachment factors or with receptors and sometimes might even result in receptor switch. In the glycoprotein of MeV, only a single exchanged amino acid in the viral receptor ligand can be sufficient to result in a switch between usage of SLAM or nectin-4 receptors and entering the cell via human CD46 (e.g., as known for strain Edmonston). Detailed studies demonstrated that the interaction sites for the cellular receptors on the viral glycoprotein at least partially overlap and that alteration of a single amino acid can be sufficient to change receptor preference (36,37). However, the mutations observed in the APPV glycoproteins after cell culture adaptation apparently do not result in a receptor switch since both APPV harvested from early passage (APPV P17 ) and cell culture-adapted APPV P100 showed comparable dependency on CD46 pig . An increase in the affinity for CD46 pig in the cell culture-adapted virus is one plausible explanation. As known for heparan sulfate (HS) usage by BVDV and CSFV, improved attachment to the host cell might also help to bind to the receptor and thus result in more efficient entry. Consequently, the cell culture-adapted APPV P100 established in this study represents a precious tool to identify molecular determinants of entering the host cell.
In naturally infected pigs, APPV can be detected in many different tissues as reported earlier (25,38). This is in line with the finding of this study that APPV uses the ubiquitously expressed CD46 pig molecule for cell entry. Nevertheless, other cellular factors are apparently implicated in the entry process and will be required to complete the viral replication cycle successfully. Therefore, it is not surprising that APPV and BVDV show some differences in tissue tropism although both are highly dependent on CD46 during entry process. It remains obscure why BVDV preferentially infects certain cell types, e.g., epithelial and immune cells. One explanation might be interaction with additional (colocalized) surface proteins (20). Very recent data demonstrated that in polarized bovine respiratory epithelial cells, CD46 bov is a major BVDV receptor on the apical but not the basolateral cell membranes despite the observation that basolateral infection was more efficient than apical infection (32). In consequence, the entry process of BVDV is much more complex than previously recognized. Although CD46 pig represents a crucial factor in APPV entry, it remains unknown which additional factors are required for efficient replication. The future identification of such host factors may also explain differences in tissue tropism of BVDV and APPV. Nevertheless, the loss-of-function experiments presented in this study clearly demonstrate an important role of CD46 pig during cell entry of APPV. Diverse genetic and splice variations of CD46 bov have been reported to shift permissivity of bovine cells to BVDV (22). Future studies need to figure out whether CD46 pig complementation of CD46 knockout cells in trans can restore the permissive phenotype of wild-type cells. Successful complementation might depend on a specific CD46 pig variant or availability of specific cofactors and thus can provide novel insights into molecular interaction between CD46 pig and APPV during the entry process.
Several cellular factors were suggested to be involved in the entry process of CSFV, including CD46 pig , heparan sulfate (HS), laminin receptor (LamR/RPSA), the low-density lipoprotein receptor (LDLR), beta-actin, and vinculin (24,39,40). As antibody-mediated blocking of CD46 pig resulted in a reduction of CSFV infection, CD46 pig was proposed to be a main receptor of CSFV, along with the attachment factor HS and possibly other unknown receptors (24). However, the inhibitory effect of anti-CD46 pig antibodies may rely on steric side effects interfering with virus/receptor interaction. In a different study, blocking of the CD46 bov receptor with an antiserum provided no evidence for involvement in entry of CSFV, BDV, or Giraffe pestivirus (26). Importantly, amino acid composition of the BVDV E2 binding motif (E 66 QIV 69 and G 82 QVLAL 87 ) of CD46 bov and the analogous region in CD46 pig of porcine cells display major differences (26). The data obtained from infecting the well-defined porcine CD46 pig knockout cells clearly show that, in contrast to APPV, CD46 pig is not essential for cell entry of the porcine pestiviruses CSFV and BuPV ( Fig. 4 and 5 and Fig. 6A). Surprisingly, CSFV and BuPV seem to use an alternative route of cell entry. Cell-to-cell spread may result in overcoming the CD46 pig receptor usage via engagement of so far unknown alternate receptors on the target cell (41). Nevertheless, even as early as 16 h p.i., avoiding prominent effects of cell-to-cell transmission, no impact of CD46 pig on CSFV or BuPV infection became evident (Fig. 6A). A specific mutation in the E rns protein of CSFV (Ser 476 to Arg 476 ) was previously reported to result in cell culture adaptation via improved attachment to the cell by increasing the affinity to HS (29,42). CSFV isolates used in this study, including the recombinant CSFV strain AlfT as well as isolates Riems, Koslov, and Diepholz, do not possess this specific mutation and thus are not adapted to efficient HS attachment. All isolates were able to infect the CD46 pig -deficient knockout cells very efficiently. This strongly suggests that HS usage is not implicated in efficient entry of these CSFV strains ( Fig. 4 and Fig. 6A). In consequence, it is likely that a so far unknown receptor is used by CSFV to enter the cells. The representative selection of CSFV isolates (different genotypes and virulence properties) used in this study provide evidence that CD46 pigindependent entry of CSFV is not restricted to a particular CSFV isolate.
Glycoprotein E2 of classical pestivirus species is the only known viral determinant possessing receptor-binding properties. However, simple transfer of BuPV E2 protein was not sufficient to expand host cell range of BVDV to human cells which are susceptible to BuPV (33). Interestingly, amino acid composition and predicted size of E2 glycoproteins of classical and recently discovered atypical pestiviruses display major differences (Fig. 7A). These differences can be expected to result in altered protein structure and different protein-protein interactions and may correlate with different mechanisms of cellular entry (43). There is strong evidence that a stretch of amino acids in domain II of pestivirus E2 functions as the receptor ligand (amino acids 834 to 863 in BVDV strain NADL, PDB 4JNT). This motif is exposed as a hairpin in the crystal structure of BVDV E2 (44). The distinct E2 amino acid composition of APPV did not allow for identification of a corresponding receptor-binding motif (Fig. 7B). Against this background, it is not surprising that the distantly related porcine pestiviruses APPV and CSFV use different strategies of cell entry. Remarkably, BVDV and APPV obviously share the same receptor, although they are only distantly related and infect different host species (ruminant and porcine, respectively).
Taken together, the presented results demonstrate that CD46 pig is a major cellular factor for efficient entry of APPV. With regard to its role in viral entry, the function of CD46 pig for APPV infection resembles the function of CD46 bov for BVDV entry. Additionally, our study demonstrates that (in contrast to BVDV and APPV) other porcine pestiviruses like CSFV and BuPV apparently use a CD46 pig -independent way to enter the host cell. This indicates that CD46 is not a general receptor of pestiviruses and that different pestivirus species apparently use diverse mechanisms for host cell entry.

MATERIALS AND METHODS
Cells and viruses. The porcine kidney cell lines SPEV and PK15 were maintained as a monolayer in Earle's minimal essential medium (EMEM) containing 5% fetal bovine serum (FBS) for the infection experiments and generation of the knockout cell lines. Other cell lines were maintained as a monolayer in Dulbecco's modified Eagle medium (DMEM) containing either 10% FBS or horse serum (MDBK and CRFK) ( Table 1).
Selected CSFV strains include representatives of the most relevant genotypes 1 and 2. CSFV strains Alfort-Tübingen (AlfT, CSF0904, genotype 2.3), Diepholz (CSF0104, genotype 2.3), Paderborn (CSF0277, genotype 2.1), Koslov (CSF0382, genotype 1.1), and Riems (CSF0913, vaccine strain, genotype 1.1) were obtained from the virus collection (CSF catalogue numbers are given in parentheses) of the Institute of Virology, University of Veterinary Medicine Hannover, Germany (45). These CSFV strains show differences in virulence properties, including the attenuated vaccine strain Riems, a variant of the live attenuated Cstrain, and the extraordinary virulent strain Koslov, which is often used in vaccination-challenge studies. Glycoprotein E rns coding regions were monitored for a point mutation (Ser 476 to Arg 476 ) which was previously reported to be responsible for cell culture adaptation via heparan sulfate (HS) usage (29). RT-PCR and subsequent Sanger sequencing were performed using primer pair 1539fw/2222rev (Table 2).
Increasing numbers of APPV-positive cells were observed during continuous passaging of persistently infected SPEV cells. For characterization of the cell culture adaptation, naive SPEV cells were infected with 1 ml APPV cell culture supernatant from the 25th, 45th, 55th, and 100th passage in six-well plates at the time point of seeding. Cells were heat fixed 72 h p.i. for immunofluorescence staining as described below. Viral titers were determined from the supernatants of 25th, 55th, and 100th passages of APPV. Genomic alterations within the predicted glycoprotein-encoding regions were monitored after the 17th, 25th, 35th, 45th, 55th, and 100th passages of APPV by RT-PCR and subsequent Sanger sequencing, using primer pairs 1114fw/1594rev, 1459fw/2002rev, 1721fw/2251rev, 2109fw/2554rev, 2353fw/2912rev, and 2749fw/3324rev (Table 2). Moreover, at the 100th passage (APPV P100 ), the complete polyprotein coding sequence was determined using next-generation sequencing as previously described (25).
Generation of CD46 pig knockout cell lines. Generation of knockout cells by CRISPR/Cas9 technology was performed as described previously (46). Briefly, oligonucleotides containing the guide sequences (primers gCD46-2 and gCD46-7) targeting the region encoding the N-terminal part of complement control protein 1 (ccp1) within CD46 pig were designed (Table 2) and cloned into the plentiCRISPR-v2 plasmid to generate plentiCRISPR-v2-CD46-2 and plentiCRISPR-v2-CD46-7 (47,48). For production of lentiviral particles, the recombinant plasmids were cotransfected with a packaging vector (pCMVDR8.91) and a plasmid encoding the glycoprotein of vesicular stomatitis virus (VSV-G-pMD.G) into HEK293T cells, using polyethylenimine transfection reagent (24765-1, Polysciences, Inc.). Harvested lentiviral particles were used for transduction of SPEV and PK15 cells, respectively, followed by a subsequent puromycin selection (P8833, Sigma-Aldrich) with 5 mg/ml puromycin-supplemented medium for 2 weeks. At least three rounds of biological cloning of SPEVDCD46 and PK15DCD46 cell lines were performed by single cell expansion. Purity of the obtained knockout cell lines was confirmed by immunofluorescence analysis and genetic characterization as described below.
For genetic characterization of CD46 pig expressed by parental SPEV and PK15 cells, RNA preparations were used for reverse transcription by SuperScript III reverse transcriptase (18080093, Invitrogen) and random hexamers (N8080127, Invitrogen). Amplification of a partial CD46 pig coding sequence was performed using ALLin HS Red Tag Mastermix (HSM0305, highQu) with two primer pairs (101fw/710rev and 604fw/1192rev, Table 2) for subsequent generation of a consensus sequence (1,092 bp). The following thermal profile was applied: 2 min at 95°C, 40 cycles; 15 s at 95°C, 15 s at 54°C, 15 s at 72°C, and final extension for 5 min at 72°C. Sanger sequencing was conducted by LGC genomics (Berlin, Germany). Attempts to amplify the CD46 pig coding sequence from 38A 1 D cells with these primers were not successful. Thus, four additional primer pairs were designed, which target genetic regions representing the individual domains ccp1 to ccp4 of CD46 pig (192fw/353rev, 387fw/543rev, 571fw/710rev, 766fw/937rev; Table 2).
To characterize the CRISPR/Cas9-induced genome alterations in the engineered CD46 pig knockout cells, primers (101fw/710rev) flanking the target site of the guide RNAs were applied. These primers were specific for regions coding for the signal peptide and the ccp3 domain, respectively ( Table 2). Gel purified amplicons (wild-type genome: 610 bp) were cloned into TOPO vector (450030, ThermoFisher Scientific) and propagated in TOP10 E. coli bacteria. Individual colonies were subjected to conventional PCR using vector specific primers T7fw and M13rev (Table 2). Subsequently, amplicons of at least 10 individual colonies were subjected to Sanger sequencing (LGC genomics, Berlin, Germany) to determine genetic variation within the individual cell clones. Obtained sequences were analyzed using the GENtle software (version 1.9.4.0) and compared to the CD46 pig coding sequences of SPEV and PK15 WT cells.
Infection of wild-type and CD46 pig knockout cell lines. Titers of virus stocks of APPV P100 , BuPV, and different CSFV strains were determined, and virus stocks were stored at -80°C until use for infection experiments. Titer of virus stock APPV P17 could not be determined due to inefficient viral replication. Titers were determined by three independent endpoint dilution assays each performed in quadruplicates. Visualization of viral infection was done by immunofluorescence staining.
One day prior to infection, 3 Â 10 4 SPEV, PK15, and CD46 pig knockout cells (SPEVDCD46 clones 2 and 7 and PK15DCD46 clone 2) were seeded into 24-well plates. Cells were infected with 500 ml of APPV P17 since the viral titer is unknown. Infections with APPV P100 , BuPV, and the different CSFV strains were performed at an MOI of 1 for 2 h at 37°C, respectively. Subsequently, inoculum was removed and infected cells were washed three times with PBS and incubated for 72 h at 37°C. Supernatant was harvested to determine the virus titer at 72 h postinfection. Cells were washed three times with PBS and subsequently analyzed by immunofluorescence staining or RT-PCR analysis. Additionally, infections of WT and CD46 pig knockout cell lines with CSFV and BuPV (MOI of 1, described above) as well as with APPV P100 (MOI of 0.5) were analyzed 16 h postinfection (p.i.) by either immunofluorescence staining (CSFV and BuPV) or fluorescence in situ hybridization (FISH) assay (APPV). For the FISH assay, 1 day prior to the APPV infection, 4 Â 10 5 SPEV, PK15, and knockout SPEVDCD46 clone 2 and PK15DCD46 clone 2 cells were seeded on Nunc Lab-Tek II CC2 chamber slides (S6690, Merck). All infection experiments were repeated three times.
As immunofluorescence staining was not suitable to detect APPV infection at an early time point of infection (16 h p.i.), FISH assay was performed. The assay was optimized for the cell lines used in this study based on the manufacturer's protocol (QVC0001, ThermoFisher Scientific). Formaldehyde fixation was applied for 30 min. Cy3 labeled APPV-specific probes were based on the sequence of APPV isolate L277 (GenBank MF167291) and located within the NS3 to NS4B region (3958-7236 base sequence, ViewRNA Type1 probe set, ThermoFisher Scientific). Fluorescein isothiocyanate (FITC)-labeled pig b-actin-specific probes were used as an internal reference (ViewRNA Type4 probe set, ThermoFisher Scientific).
Quantitative analyses of the images, which were generated from three independent FISH assays, were carried out using ImageJ software (version 1.51.0). The APPV-specific staining image was converted to a mask by implementing the Renyi Entropy thresholding method. APPV-specific signal is determined using the pixel count tool in histogram.