ANP32 Proteins Are Essential for Influenza Virus Replication in Human Cells

Influenza virus is the etiological agent behind some of the most devastating infectious disease pandemics to date, and influenza outbreaks still pose a major threat to public health. Influenza virus polymerase, the molecule that copies the viral RNA genome, hijacks cellular proteins to support its replication. Current anti-influenza drugs are aimed against viral proteins, including the polymerase, but RNA viruses like influenza tend to become resistant to such drugs very rapidly. An alternative strategy is to design therapeutics that target the host proteins that are necessary for virus propagation. Here, we show that the human proteins ANP32A and ANP32B are essential for influenza A and B virus replication, such that in their absence cells become impervious to the virus. We map the proviral activity of ANP32 proteins to one region in particular, which could inform future intervention.

Influenza virus requires the host cell machinery to support replication of its genome and production of new virions. The influenza genome is made up of eight segments of single-stranded negative-sense RNA (vRNA). Each segment is packaged in a double helical loop structure bound by nucleoprotein (NP) along its length, except for the pseudocomplementary 3= and 5= untranslated regions that comprise the promoter (3)(4)(5). These termini instead associate with an RNA-dependent RNA polymerase (RdRp) encoded by the virus (6). This key enzyme, a heterotrimer of polymerase basic protein 1 (PB1), polymerase basic protein 2 (PB2), and polymerase acidic protein (PA), functions as both a transcriptase and a replicase (reviewed by te Velthuis and Fodor [7]). Transcription of mRNA and replication through a positive-sense cRNA take place in the host cell nucleus (7,8). A viral complex containing RNA, NP, and RdRp is termed a ribonucleoprotein (RNP), which, depending on the sense of the RNA, is either a vRNP or a cRNP.
ANP32A and ANP32B are small acidic nuclear phosphoproteins (ANPs) that are attributed to a plethora of cellular functions (9), including chromatin remodeling (10,11), apoptosis (12,13), transcription regulation (14,15), and intracellular transport (16). ANP32 proteins are approximately 250 amino acids in length and contain an N-terminal leucine-rich repeat (LRR) region, a central domain, and an unstructured low-complexity acidic region (LCAR) at the C terminus. ANP32 proteins have been associated with influenza A polymerase function. A nuclear fraction containing ANP32A and ANP32B was shown to enhance the synthesis in vitro of vRNA from a short cRNA template (17). Knockdown of ANP32A or ANP32B in human cells reduced polymerase activity measured in minigenome reporter assays, as well as synthesis of viral RNA in infected cells (17,18). Direct interactions of ANP32 proteins with the RdRp or RNP have been documented but do not completely correlate with function (19)(20)(21)(22). The difference between avian and mammalian ANP32A proteins has been suggested to account for host range restriction of avian influenza strains in mammalian cells, and much of the work to date has focused on the avian orthologues, particularly those from chickens (23).
Here, we use CRISPR/Cas9 genome editing to render the Anp32A and/or Anp32B genes nonfunctional in low-ploidy human eHAP1 cells (24,25), thus obtaining a clean experimental platform in which to investigate the interplay between different influenza virus polymerases and mammalian ANP32 proteins. We find that although IAV and IBV polymerases can replicate in the absence of either ANP32A or ANP32B alone (i.e., in single-knockout cells), depletion of both proteins (double knockout) renders the cell impervious to RdRp activity. Furthermore, none of the IAV strains tested is capable of replication in the double knockout cells. Human ANP32A and ANP32B proteins are thus functionally redundant but essential for influenza virus replication. We further show that this redundancy is not present in the murine Anp32 orthologues. Only murine Anp32B (MusB) is able to recover IAV polymerase activity, although, surprisingly, murine Anp32A (MusA) can be coopted by IBV polymerase. Functionality mapped to leucine-rich repeat 5 of the LRR domain, thus assigning this domain of the host proteins as key for the support of influenza polymerase activity and a target for future interventions.

RESULTS
Generation of eHAP1 knockout cells. eHAP1 cells lacking ANP32A (AKO), ANP32B (BKO), or both proteins (dKO) were generated by CRISPR/Cas9 genome editing using a double-nickase approach for enhanced specificity and minimal off-target DNA cleavage (26,27) (Fig. 1a). Control cells (control) were treated in an identical manner with nontargeting guide RNAs (28). Two independent clones with diallelic disruption of the Anp32A or Anp32B locus were verified by next-generation sequencing (NGS) and Sanger sequencing of individual alleles, and loss of protein expression was confirmed by Western blotting against ANP32A or ANP32B, respectively ( Fig. 1b and data not shown). Double-knockout cells were generated by tandem CRISPR from a BKO clone, using the guides against the Anp32A locus. Three independent dKO clones were verified by Sanger sequencing, and loss of expression was confirmed by Western blot analysis ( Fig.  1b and  ) obtained by one-way analysis of variance (ANOVA) from one representative repeat (n Ͼ 3). Accompanying Western blots show expression of respective vRNP components in each cell type (representative of one minigenome assay). ns, not significant; *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ****, P Ͻ 0.0001.

2006)
. Surprisingly, absence of ANP32A in human cells did not result in loss of influenza polymerase activity; in fact, activity increased for some polymerase constellations. (Fig.  1c to f). Polymerase activity in BKO clones was either unaffected or decreased but not abrogated ( Fig. 1c to f). A similar pattern was observed in A549 cells lacking ANP32A or ANP32B (data not shown). Strikingly, however, none of the polymerases showed any activity in three independent double-knockout lines (Fig. 1c to f and data not shown), despite robust expression of vRNP components ( Fig. 1c to f). These data suggest functionally redundant roles for ANP32A and ANP32B in supporting influenza virus polymerase activity in human cells.
In order to confirm redundancy, polymerases were coexpressed in dKO cells with plasmids encoding exogenous ANP32A, ANP32B, or equal amounts of both. All polymerases tested regained activity in the presence of either ANP32 protein ( Fig. 2a to d). Provision of both proteins at once did not further enhance rescue. These results were corroborated at the single-cell level: ANP32A or ANP32B proteins fused to the red fluorescent protein mCherry were coexpressed in dKO cells with H5N1 (PB2 627K) 50-92 vRNP components PB1, PB2-627K, PA-green fluorescent protein (GFP), and NP, with an influenza minigenome encoding blue fluorescent protein (BFP) as a reporter. Blue fluorescence resulting from active polymerase was observed only in cells that expressed ANP32 proteins, and either paralogue was able to rescue activity (Fig. 2e).
These data demonstrate that ANP32A and ANP32B proteins are essential but redundant for influenza A and B polymerase activity in human cells. Intriguingly, we observed that recovery of polymerase activity in dKO cells was more efficiently achieved by expression of ANP32B than ANP32A for specific polymerase constellations, namely avian H5N1 50-92 (PB2 627K) and IBV Florida 06 ( Fig. 2b and d). These two polymerase constellations were also more affected by loss of ANP32B expression ( Fig.  1d and f), suggesting ANP32B is the preferred host factor for these polymerases in human cells.
IAV replication is abrogated in cells lacking ANP32A and ANP32B. To investigate the consequence of absence of ANP32A or ANP32B proteins on infectious IAV replication in human cells, control, single-knockout, and double-knockout cells were infected at a multiplicity of infection (MOI) of 0.005 with three different viruses whose genetic content corresponded to the polymerase constellations tested in Fig. 1 and 2, i.e., from Vic/75, Tky/50-92 (E627K), or Eng195. While virus replicated to high titers in control and single-knockout (KO) cells, replication in dKO cells was completely abrogated ( Fig. 3a to c). This suggests that viral proteins such as NEP and NS1 (expressed during viral infection but not provided in the minigenome assay) cannot overcome the block in replication imposed by the absence of ANP32 proteins. Replication of the H1N1 laboratory-adapted strain A/PR/8/34 was also abrogated in dKO cells (data not shown). Reconstitution of dKO cells with both ANP32A and ANP32B proteins by transient transfection prior to infection restored PR8 virus replication (Fig. 3d).
It has been suggested that ANP32A and ANP32B specifically support the synthesis of negative-sense vRNA from a positive-sense intermediate template (cRNA) (17). This is believed to occur after primary transcription in cis of the vRNA by the incumbent RdRp and requires newly synthesized RdRp molecules to stabilize the cRNA in trans (29,30). Therefore, without replication, secondary transcription and accumulation of viral proteins will not occur. We used immunofluorescence microscopy to visualize accumulation of viral nucleoprotein (NP) in control and dKO cells 5 h postinfection with H1N1 PR8 virus. NP accumulation exceeded background level only in cells containing ANP32 proteins (Fig. 4a), but this approach was not sufficiently sensitive to image NP protein products of primary transcription. In order to assess which viral RNAs were synthesized in cells that lack expression of ANP32 proteins, we preexpressed an inactive influenza polymerase complex (to stabilize any cRNA generated) for 20 h before infecting with high-MOI virus in presence or absence of cycloheximide (CHX). Five hours later, levels of vRNA, cRNA, and mRNA generated from segment 6 of the incoming virus were assayed by reverse transcription-quantitative PCR (qRT-PCR) detected in the presence of CHX in control cells. Thus, our data support the block to replication occurring at the step of copying cRNA back to vRNA in the absence of ANP32 proteins. Murine Anp32B supports IAV polymerase. We used the complementation assay in dKO cells to ask whether Anp32 proteins from nonhuman species relevant to influenza virology were capable of supporting polymerase function. We carried out minigenome reporter assays coexpressing Anp32A proteins from pig (SusA), mouse (MusA), duck (AnasA), and chicken (GallusA) with IAV polymerase (H3N2 Victoria/75). While the avian and porcine orthologues could support IAV polymerase, MusA could not (Fig. 5a). Bearing in mind that our results suggest that human ANP32B might be the more potent host factor for some polymerase constellations, we hypothesized that in mice, influenza virus might rely solely on MusB to support its replication. Indeed, expression of MusB could recover Vic/75 polymerase activity in dKO cells (Fig. 5b), despite equal levels of expression of both murine Anp32 proteins and their localization to the cell nucleus ( Fig. 5b and c).
An alignment of murine and human ANP32 proteins showed several unique features in the MusA sequence, mapping largely to surface-exposed residues within LRR 5 (Fig.  5d). In order to determine whether these differences were responsible for the lack of functionality of MusA, we generated a chimera of murine Anp32A and B (MusA 128-153 ) by substituting a 26-amino acid (aa) segment (aa 128 to 153) of MusB into MusA (Fig.  5d), and then tested if this conferred gain of function on MusA to support IAV polymerase. The chimera was indeed capable of recovering activity of Vic/75 polymerase in dKO cells, although not to the level shown by MusB (Fig. 5b). Western blot and immunofluorescence analysis of the FLAG-tagged chimeric construct demonstrated expression and nuclear localization (Fig. 5 b and c).
We identified a single amino acid in LRR 5 at position 130 that was the same in human ANP32A or ANP32B and MusB (aspartic acid, D) but which differed in MusA (alanine, A) ( Fig. 5e and f). Introduction of a D130A single-point mutation in human ANP32A significantly reduced its ability to support Vic75 polymerase activity, and conversely, introduction of A130D to MusA produced a small but significant increase in its ability to support viral polymerase (Fig. 5g).
Finally, we explored whether MusA or MusB could support activity of polymerases derived from other IAV strains or from IBV. As seen for Vic/75 polymerase, MusA was nonfunctional for IAV polymerases from A/Tky/50-92/91 and A/Eng/195 ( Fig. 6a and b); however, IBV Florida 06 polymerase recovered some activity in dKO cells in the presence of MusA (Fig. 6c). MusB, however, was the more potent factor for support of IBV polymerase. Intriguingly, IBV Florida 06 polymerase activity was even greater in the presence of the MusA/MusB chimera than in that of MusB alone (Fig. 6c).

DISCUSSION
Here, we show that human ANP32A and ANP32B are functionally redundant in their support for influenza virus polymerase in human cells and that the RdRp does not carry out RNA replication in the absence of both family members. Our findings corroborate those of Zhang et al. (2019) who used a similar CRISPR approach (31) We further show that IBV polymerase is also dependent on human ANP32 proteins and can also utilize either orthologue to support activity. Our demonstration of redundancy in use of these essential host factors illuminates a deficiency in RNA interference (RNAi) or CRISPR screens where host genes are knocked down one at a time. Functionally redundant pairs or larger groups of host factors that can be used by a virus will escape detection, as have ANP32A and ANP32B individually in previous screens (32,33).
Two observations imply preference of certain polymerase constellations for human ANP32B over ANP32A. First, specific polymerases were more efficiently enhanced by ANP32B when provided exogenously. Second, the absence of ANP32A in the cell enhances virus polymerase activity in some cases, where the opposite might be cotransfected with FLAG-tagged mouse Anp32A, Anp32B, or Anp32A 128-153 . Data show mean (SD) of firefly activity normalized to Renilla and analyzed by one-way ANOVA from one representative repeat (n ϭ 2 triplicate experiments). ns, not significant; *, P Ͻ 0.05; ****, P Ͻ 0.0001. Accompanying Western blot shows expression of vRNP component PB2 and coexpressed FLAG-tagged ANP32 constructs. expected and has previously been observed in knockdown experiments (17,18). The latter observation might be explained if ANP32B is held in heterodimers or larger protein complexes with ANP32A, thus absence of ANP32A might liberate the preferred ANP32B for recruitment by influenza virus RdRp. Alternatively, ANP32A and ANP32B, being very similar structurally, might compete for polymerase binding, and loss of ANP32A would then favor binding and more efficient activity mediated by ANP32B. Taken together, the observations point to ANP32B being functionally superior in humans to ANP32A for supporting influenza polymerase. However, these differences were not so readily apparent in the context of virus infection where replication continued largely unabated in single ANP32A or ANP32B knockout cells.
Interestingly, the redundancy observed in humans is not observed for murine Anp32 proteins. IAV polymerases cannot use MusA, but IBV polymerase can, albeit inefficiently. The ability of influenza A virus to replicate in mice is explained by the utility of murine Anp32B. Mapping this difference using a chimeric approach revealed LRR 5 as a key domain of ANP32 proteins for supporting IAV polymerase, and point mutation highlighted the role of a single amino acid in LRR 5 at position 130. It will be important to investigate whether this is a contact point in the interaction between the viral complex and the host protein, and this will be a crucial question for structural studies to address. The highlighted domain sits adjacent to a linker region between the structured LRR and highly flexible LCAR, and it may be important for defining overall structural arrangement and susceptibility to conformational change. It is interesting to note that in chicken cells, it is the Anp32A orthologue that is utilized by avian influenza polymerase, whereas chAnp32B does not support replication; this difference in functionality was also mapped to LRR 5 (31,34).
Current anti-influenza therapeutics, as well as drugs in development such as adamantanes (M2 ion channel inhibitors), neuraminidase inhibitors (NAIs) such as oseltamivir, and RdRp-targeting molecules (including the nucleoside analogue favipiravir and small molecules such as baloxavir), are all aimed at proteins encoded by the virus. A recurring issue with such drugs is the ease with which influenza virus evolves resistance to them, be it in a laboratory setting (35,45) or in the field (36)(37)(38). An alternative approach would be to target specific interactions of virus proteins with essential host factors, such that small-molecule inhibitors may temporarily block the interacting surface on the host protein without compromising its cellular functions. As influenza virus replication is completely abrogated in their absence, ANP32 proteins suggest themselves as potential candidates for such an approach.
Generation and screening of CRISPR clones. Pairs of guide RNAs against exon 2 of human Anp32A (GTCAGGTGAAAGAACTTGTCC and GAAGGCCCGACCGTGTGAGCG) and Anp32B (GAGCCTACATTTATTAA ACTG and GCAAGCTGCCTAAATTGAAAA) were designed with the aid of the CRISPR design tool at www.crispr.mit.edu (Feng Zhang Lab). The nontargeting guide RNA pair was GTATTACTGATATTGGTGGG and GAACTCAACCAGAGGGCCAA. The guides were cloned into plasmid pSpCas9n(BB)-2A-Puro (PX462) v2.0 (Feng Zhang Lab), obtained via Addgene, and equimolar amounts of plasmids were transfected using Lipofectamine 3000 (Thermo Fisher). Cells harboring at least one plasmid were enriched by selection with puromycin at 1.5 g · ml Ϫ1 for 3 to 5 days and single-cell sorted into 96-well plates containing growth medium, using a fluorescence-activated cell sorter (FACS) Aria IIIU (BD Biosciences) with an 85-m nozzle. Single cells were grown out into clonal populations over a period of 10 to 14 days. Genetic loci harboring insertion/deletion mutations (indels) were amplified by PCR using barcoded primers (AGTGACGGAGTGACTGACTG and GAGGTGAGGCCTACGTTGAT for Anp32A; TGTCTTGGACAATTG CAAATCAA and CCATGTGCTTTCTGCTACACT for Anp32B) (42). A total of 268 barcoded amplicons were then prepared for next-generation sequencing (NGS) using the NEBNext Ultra II kit (NEB) and sequenced using 150-bp paired-end reads on an Illumina MiSeq instrument. Reads were mapped using Burrows-Wheeler Aligner (BWA) v0.7.5. Indels occurring above a cutoff of 2.5% of reads were detected using an R script (https://github.com/Flu1/CRISPR). DNA sequences were analyzed in Geneious v6.
Minigenome assay. In order to measure influenza virus polymerase activity, pCAGGS expression plasmids encoding PB1 (0.04 g), PB2 (0.04 g), PA (0.02 g), and NP (0.08 g) from each virus [H3N2 Victoria, H5N1 (PB2-627K) 50-92, pH1N1 England 195, or IBV Florida 06] were transfected into 200,000 eHAP1 or A549 cells using Lipofectamine 3000 (Thermo Fisher) at ratios of 2 l P3000 reagent per g plasmid DNA and 3 l Lipofectamine 3000 reagent per g plasmid DNA. As reporter constructs, we transfected 0.04 g PolI-luc, which encodes a minigenome containing a firefly reporter flanked by either influenza A or B promoter sequences, or, in Fig. 2e, PolI-BFP. pCAGGS-Renilla luciferase (0.04 g) was transfected as a transfection and toxicity control. For exogenous expression, 0.1 g pCAGGS plasmid encoding either the relevant FLAG-tagged Anp32 gene or Empty pCAGGS was coexpressed with the RNP components. The ratio of transfected plasmids was constant at all times, namely 2:2:1:4:2:2:5 PB1:PB2:PA:NP:PolI reporter:Renilla:ANP32/Empty (if present). At least 20 h posttransfection, cells were lysed in 100 l passive lysis buffer (Promega), and the dual-luciferase reporter assay kit (Promega) was used to measure bioluminescence on a FLUOstar Omega plate reader (BMG Labtech). In the case of a PolI-BFP reporter, please refer to the "Fluorescence microscopy" paragraph below. All minigenome assays were repeated in triplicate at least twice.
Fluorescence microscopy. At least 200,000 cells were cultured on glass coverslips in 24-well plates and transfected or infected as described. Cells transfected with plasmids encoding fluorescent proteins (BFP, GFP, or mCherry) were fixed in 4% paraformaldehyde and then visualized. Cells transfected with plasmids encoding (FLAG-tagged) nonfluorescent proteins were fixed and permeabilized in 0.2% Triton X-100. Primary antibodies used were rabbit anti-FLAG F7425 (Sigma) or mouse anti-IAV NP (Abcam 128193). Secondary antibodies were goat ␣-rabbit Alexa Fluor-594 (catalog number ab150080, Invitrogen) and goat ␣-mouse Alexa Fluor-568 (catalog number A11031; Invitrogen). Coverslips were mounted on glass slides using Vectashield mounting medium (H-1000-10; Vector Laboratories). Cells were imaged with a Zeiss Cell Observer widefield microscope with ZEN Blue software, using a Plan-Apochromat ϫ100 1.40-numerical aperture oil objective (Zeiss), an Orca-Flash 4.0 complementary metal-oxide semiconductor (CMOS) camera (frame, 2,048 ϫ 2,048 pixels; Hamamatsu), giving a pixel size of 65 nm, and a Colibri 7 light source (Zeiss). Channels acquired and filters for excitation and emission were 4=,6-diamidino-2phenylindole (DAPI) (excitation [ex], 365/12 nm, emission [em] 447/60 nm), GFP (ex 470/40 nm, em 525/50 nm), and TexasRed (ex 562/40 nm, em 624/40 nm). All images were analyzed and prepared with Fiji software (43). For images in Fig. 2e and Fig. S5b, the detection limit was adjusted individually for each channel (taking care to remain well above control background level), while in Fig. 3e, where we are comparing relative levels of NP expression, the lower detection limit in the TexasRed channel was set equal to that of the DAPI channel.
Influenza virus infection. Cells were infected with virus diluted in serum-free IMDM or DMEM at 37°C (MOI as indicated in the text or relevant figure legends) and replaced with serum-free cell culture medium supplemented with 1 g · ml Ϫ1 L-1-tosylamide-2-phenylethyl chloromethyl ketone (TPCK) trypsin