Functional Analysis of SARS-CoV-2 Proteins in Drosophila Identies Orf6-induced Pathogenicity Attenuated by Selinexor

Background: SARS-CoV-2 causes COVID-19 with a widely diverse disease prole that affect many different tissues. The mechanisms underlying its pathogenicity in host organisms remain unclear. Animal models for study the pathogenicity of SARS-CoV-2 proteins are lacking. Methods: Using bioinformatic analysis, we showed that the majority of the virus-host interacting proteins are conserved in Drosophila. Therefore, we generated a series of transgenic lines for individual SARS-CoV-2 genes and used the Gal4-UAS system to express them in various tissues to study their pathogenicity. Results: We found that the Nsp6, Orf6 and Orf7a transgenic ies displayed reduced trachea branching and muscle decits resulting in “held-up” wing phenotype and poor climbing ability. Furthermore, muscle tissue in these ies showed dramatically reduced mitochondria. Since Orf6 was found to bind nucleopore proteins XPO1, we tested Selinexor, a drug that inhibits XPO1, and found that it could attenuated the Orf6-induced lethality and tissue-specic phenotypes in ies. Conclusions: Our studies here established new Drosophila models for studying the function of SARS-CoV2 genes, identied Orf6 as a highly pathogenic protein in various tissues, and demonstrated the effects of Selinexor for inhibiting Orf6 toxicity with an in vivo model system.

Introduction SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2), the cause of coronavirus disease 2019 , is the latest in a string of outbreaks in the human population caused by highly pathogenic coronaviruses. Its high transmission rate and virulence have culminated in a pandemic to which no end is insight at the time of this writing. At the time of this writing, SARS-CoV-2 has infected nearly 80 million people globally, causing more than 1,7 million deaths (source: Johns Hopkins University). While SARS-CoV-2 shares many similarities with its predecessors, SARS-CoV (2002)(2003)(2004) and MERS-CoV (Middle East respiratory syndrome-CoV; 2012), investigations have already revealed several unique features. For example, the SARS-CoV-2 spike (S) protein structure includes a distinct loop with exible glycyl residues in place of the SARS-CoV rigid prolyl residues. Molecular modeling has indicated the SARS-CoV-2 receptor binding domain has a higher a nity for ACE2 compared to its SARS-CoV counterpart (Y. Chen, Guo, Pan, & Zhao, 2020), which might contribute to its high virulence. These ndings highlight the important roles speci c viral genes can play, yet very little is known of the functions of individual SARS-CoV-2 proteins and the host systems they affect.
The SARS-CoV-2 genome encodes 28 con rmed proteins ( Figure 1A). These include the polyproteins Orf1ab and Orf1a that are further cleaved into 165 non-structural proteins (Nsp1-16) that form the viral transcription/replication complex, as well as the 4 structural proteins (spike (S), envelope (E), membrane (M) and nucleocapsid (N) proteins) and 8 accessory proteins (Orf3a, Orf3b, Orf6, Orf7b, Orf8, Orf9b and Orf 10) (Gordon et al., 2020;Yoshimoto, 2020). The functional signi cance of these accessory proteins remains largely unresolved, since they lack well-de ned domain structures. Interestingly, they are also the least conserved between the SARS-CoV-2 and SARS-CoV viruses (Gordon et al., 2020). In fact, speci c host interaction networks have been shown for each of the SARS-CoV-2 proteins (Gordon et al., 2020). In order to identify targets and develop effective therapeutics for COVID-19, it will be vital to understand how SARS-CoV-2 hijacks host pathways and which of its proteins are key for these interactions and ultimately are the primary determinants of pathogenicity.
Human models are limited to in vitro studies which lack a whole body context, while conventional animal models are not conducive to screening multiple individual proteins and many, including mice, have been hampered by lacking natural susceptibility to SARS-CoV-2 infection (Takayama, 2020). Therefore, we sought to establish a new animal model, namely the fruit y (Drosophila). Notably, we found that the vast majority of the human host protein interactome identi ed for SARS-CoV-2 proteins have y orthologs, and their related biological processes are conserved in ies (Figure 1B and C; Supplemental Table S1). Drosophila's ease-of-use, while offering complexity with all major organs present (including heart, lung, kidney, muscle, blood and brain) and encoded by a compact genome with little redundancy, combined with the abundance of available genetic resources, has already been effectively applied to investigate several human relevant viruses in recent years (Hughes et al., 2012), including HIV-1 and Zika. For example, the HIV-1 Nef gene was studied in Drosophila using the UAS-Gal4 system (Lee, Park, Jung, & Chung, 2005). The study discovered the interaction between the HIV-1 Nef protein and the JNK and NF-kappaB signaling at the plasma membrane of y wing disc cells, both are evolutionary highly conserved signaling pathways. The HIV-1 Tat protein was similarly studied in ies and found to disrupt microtubule polymerization and kinetochore dynamics via a direct interaction with tubulin (Battaglia, Zito, Macchini, & Gigliani, 2001). This nding led to the discovery of interaction between Tat and human tubulin proteins and the important role of Tat in inducing apoptosis of human cells through targeting the microtubule network in human 293T cells (D. Chen, Wang, Zhou, & Zhou, 2002). Drosophila studies, using the UAS-Gal4 system to express Zika virus (ZIKV) proteins in speci c tissues, have also provided signi cant contributions in the study of ZIKV infection and related microcephaly. One such study found that neurons employ NF-kappaB-dependent in ammatory signaling in response to ZIKV infection (Liu et al., 2018). This, in turn, induces expression of Drosophila stimulator of interferon genes (dSTING) speci cally in the brain of ZIKV infected ies. Activation of dSTING leads to antiviral autophagy as an innate defense to control the viral infection thereby restricting ZIKV to the brain. Another high-pro le study used y models to demonstrate that genetic variants in ANKLE2, a protein linked to hereditary microcephaly, and ZIKV protein NS4A which interacts with and inhibits Ankle2 protein function, converge on the same pathological pathways (Link et al., 2019). They elegantly demonstrate that Ankle2 interaction with Ballchem/VRK1 regulates asymmetric protein localization during neuroblast division, and that disruption of this pathway leads to microcephaly in both human patients due to genetic causes and ies induced by ZIKV infection. Further, a study into host transcriptomics changes following ZIKV infection in Drosophila adult ies, revealed the importance of the JAK/STAT signaling pathway in viral pathogenesis (Harsh, Fu, Kenney, Han, & Eleftherianos, 2020). Interestingly, further exploration of the interaction between JAK/STAT signaling and the individual ZIKV non-structural proteins revealed tissue-speci c regulation of viral infection via highly conserved host signaling pathways. Together, these studies demonstrate that ies provide a powerful model to identify the prime mechanistic tissue-speci c pathways used by speci c viral pathogenic proteins, and that these pathways are conserved from y to human. However, as of yet, no literature reports the use of Drosophila to study SARS-CoV-2.
We used Drosophila to identify the foremost pathogenic SARS-CoV-2 genes, and then examined their effects on host function in a whole organism. Data revealed SARS-CoV-2 Orf6, Orf7a and Nsp6 proteins display pathogenicity in vivo. Each of these proteins individually can cause developmental lethality, reduced lifespan, defect in trachea morphology (reduced branching in y equivalent of lung), aberrant muscle function without signi cant morphological changes, and reduced mitochondria in muscle tissue. Interestingly, while all three proteins induced similar phenotypic defects, their underlying pathomechanism appears to be different. Selinexor, an FDA-approved selective inhibitor of nuclear export and predicted to disrupt SARS-CoV-2 Orf6 interaction with the host nuclear pore system, was able to counteract virus protein induced developmental lethality in Orf6 transgenic ies, but not in ies with Orf7a or Nsp6 overexpression. Indeed, Selinexor attenuated each of the observed phenotypes in SARS-CoV-2 Orf6 transgenic ies. Taken together, these data demonstrate Drosophila provides a valuable resource for studying SARS-CoV-2 mechanism-of-action and pharmacological interventions. Furthermore, these ndings make a strong case for the importance of studying individual SARS-CoV-2 proteins and illustrate their potential as therapeutic targets to counteract the pathogenic mechanisms culminating in COVID-19 symptomatology, as a complement to viral inhibition.

Results
SARS-CoV-2 Nsp6, Orf6 and Orf7a transgene expression causes developmental lethality and reduced longevity in ies The SARS-CoV-2 genome encodes 28 con rmed proteins (Orf9c genetic code does not lie within the veri ed SARS-CoV-2 open reading frame) ( Figure 1A). To date, few studies have looked at the individual viral proteins, those that did have been limited to in vitro models. Here we use the fruit y, Drosophila melanogaster, as a whole organism model to carry out a detailed study of pathogenic effect and function of individual SARS-CoV-2 proteins. We integrated data on predicted topology and subcellular localization (based on topological motifs such as transmembrane domain, signal peptide, membrane-embedded alpha-helix and nuclear localization sequence), with data from UniProt Knowledgebase (UniProt, 2019) and the scienti c literature to prioritize viral genes based on likelihood to instigate pathogenic host interactions. Viral proteins with main purported functions in viral entry, replication or packaging were initially not favored as they remain a focus of much of the latest literature. Twelve SARS-CoV-2 proteins (Nsp1, Nsp2, Nsp3, Nsp6, Orf3a, Orf3b, Orf6, Orf7a, Orf7b, Orf8, Orf9b and Orf10) were thus prioritized for analysis. In order to investigate the function of SARS-CoV-2 genes in vivo, we produced transgenic Drosophila lines carrying each of these individual SARS-CoV-2 genes, in the UAS-Gal4 Tub system. The ubiquitous enhancer Tubulin (Tub) is active in all tissues and throughout development from embryonic stages, therefore we reasoned that Tub enhancer driven expression of SARS-CoV-2 genes that are primary determinants of virus pathogenicity would result in lethality.
For this assay, male and female ies of the designated genotypes were crossed to produce progeny carrying the UAS-SARS-CoV-2 gene construct, with ubiquitous expression of the individual virus gene driven by Tub-Gal4 (red eyes, long hair), or the balancer (orange eyes, short hair) ( Figure 2A). Ratios of these two variants among the progeny showed high mortality among ies expressing SARS-CoV-2 Nsp6, Orf6 or Orf7a, and a moderate increase in mortality among Nsp3 transgenic ies ( Figure 2B and C). In addition, we monitored y survival rates. Under typical maintenance conditions, the majority of wild type ies survive 50-60 days. Orf6 and Orf7a expression ies displayed reduced lifespan such that all ies had died by 20 and 18 days, respectively ( Figure 2D). Nsp6 expression was associated with a severe lifespan reduction such that ies survived 12 days at most ( Figure 2D). Nsp3 which induced more moderate developmental lethality, also considerably reduced y lifespan to 26 days. Interestingly, though Orf3a expression did not cause developmental lethality in our mortality index, it did shorten adult y lifespan (34 days maximum), which suggests it may induce more moderate levels of toxicity or affect adult y pathways. Together, these results suggest that Nsp6, Orf6 and Orf7a might be primary determinants of SARS-CoV-2 pathogenicity.
SARS-CoV-2 Nsp6, Orf6 and Orf7a transgene expression leads to reduced branching of Drosophila trachea COVID-19 is an acute respiratory disease which is characterized by pneumonia when disease progresses, as well as additional complications. Due to its primary effect on the lungs, we tested whether expression of SARS-CoV-2 Nsp6, Orf6 and Orf7a-most pathogenic viral proteins based on our developmental lethality and survival assays-by themselves would lead to change in long morphology. The Drosophila tracheal system is similar to human lung and shares a striking resemblance in branching morphology ( Figure 3A). In wild type Drosophila larvae, the tracheal system contains a central branch and multiple classes of terminal branches ( Figure 3B). SARS-CoV-2 Nsp6, Orf6 and Orf7a expression led to dramatically reduced numbers of class II terminal branches, while the central branch and class I terminal branches appeared unchanged ( Figure 3C and D). The changed morphology could result in a functional defect of the y tracheal system and possibly contribute to the reduced developmental viability we observed in these ies.
Interestingly, we observed ies with an "held-up" wing position phenotype among those expressing SARS-CoV-2 Nsp6, Orf6 or Orf7a ( Figure 4B and C). The "held-up" wing defect is typically due to a defect of the indirect ight muscle. Hence, we took a closer look at the indirect ight muscle morphology in the Nsp6, Orf6 and Orf7a transgenic ies, but did not observe any signi cant morphological changes in muscle ber ( Figure 4D). However, when studying the mitochondria in the indirect ight muscle of these ies, their numbers were dramatically reduced ( Figure 4D and E). Mitochondria are essential for muscle function and their signi cant reduction could explain both the "held-up" wing phenotype and the reduced climbing ability of the adult ies. These data suggested SARS-CoV-2 Nsp6, Orf6 and Orf7a may affect muscle function by reducing the available function mitochondria. Additional studies are needed to understand the full mechanism underlying these ndings, and their implications.
Selinexor attenuates SARS-CoV-2 Orf6 induced developmental lethality, aberrant tracheal branching, locomotion defect and reduced mitochondria Selinexor, an FDA-approved selective inhibitor of nuclear export (Kashyap et al., 2016), has been proposed as potential treatment for COVID-19, based on its predicted ability to disrupt SARS-CoV-2 Orf6 interaction with the mRNA nuclear export complex (RAE1 and NUP98) (Gordon et al., 2020). These Orf6 interaction proteins are conserved between human and y ( Figure 5A). Therefore, we tested whether Selinexor could Orf7a. The SARS-CoV protein was shown to interact with Bcl-X L and other pro-survival proteins (Bcl-2, Bcl-2, Mcl-1, and A1), but not with pro-apoptotic proteins (Bax, Bak, Bad, and Bid) (Tan et al., 2007). The ndings suggested Orf7a might trigger apoptosis by interfering directly with the pro-survival function of Bcl-X L , indeed both proteins were shown to co-localize at the endoplasmic reticulum and mitochondria (Tan et al., 2007). Together, these ndings indicate that even though the SARS-CoV-2 Orf6, Nsp6 and Orf7a proteins lead to similar viability defects and tissue damage, the underlying pathomechanisms are likely different. . We similarly report SARS-CoV-2 Orf6 protein-protein interaction with many key members of the nuclear pore machinery, including nucleoporins and karyopherins (both importins and exportins) (Lee et al., this issue). Selinexor, an FDA-approved selective inhibitor of nuclear export (Syed, 2019), has been predicted to disrupt this interaction of SARS-CoV-2 Orf6 with the host nuclear pore complex (Gordon et al., 2020). We report it reduced SARS-CoV-2 Orf6 cytotoxic effects in a human in vitro culture model (Lee et al., this issue). Notably, the nuclear pore proteins in the Orf6 interaction network are extremely conserved between human and y ( Figure 1) (Lee et al., this issue), therefore we similarly treated the ies with this nuclear export inhibitor. Indeed, Selinexor treatment greatly attenuated the Orf6-induced muscle and trachea (lung) defects observed in y ( Figures   5 and 6). However, even though we found phenotypic overlap between Orf6, Orf7a and Nsp6 in our assays, Selinexor was unable to counteract Orf7a and Nsp6 effects, supporting the notion that the viral proteins act through different pathomechanisms. These ndings further underpin our hypothesis that inhibition of viral entry/replication (the strategy of COVID-19 interventions like Remdesivir) might proof insu cient by itself as that approach might stop the virus in its tracks but does not remedy the multitissue effects of virus already present in the host. Therefore, we propose to use our Orf6 ndings as a starting point to identify additional virus-host pathogenic interactions, then to develop drugs to target and disrupt these interactions. Therapeutic strategies, combining inhibition of virus entry/replication and pathogenic host interactions, would both stop SARS-CoV-2 spread and ameliorate the devastating progression of COVID-19 symptomatology.

SARS-CoV
In the current study, we have identi ed SARS-CoV-2 primary determinant pathogenic proteins (Orf6, Nsp6 and Orf7a). Capitalizing on the whole y, enabled us to study the pathogenic effects of the individual viral proteins on muscle and trachea (lung). We expanded our understanding of the pathomechanism invoked by Orf6 and showed that Selinexor was able to attenuate Orf6 pathogenicity speci cally in vivo. Together with the high conservation across the SARS-CoV-2 virus-host protein interaction network between human and y, these data strongly support the y as an effective model system for studying individual SARS-CoV-2 proteins.

Conservation SARS-CoV-2 and human host interactome in Drosophila
We used the DRSC Integrative Ortholog Prediction Tool (DIOPT) version 8  for the identi cation of y orthologs for the SARS-CoV-2 binding human proteins. DIOPT summarizes heterogeneous sources of conservation study tools and databases, providing integrative ortholog predictions, where score 0 indicates no orthologs and score 16 indicates all sources predicted the humany ortholog pair. Cytoscape version 3.8.0 (Shannon et al., 2003) was used to map the y ortholog information (i.e. DIOPT scores) on to the SARS-CoV-2 human host interactome (Gordon et al., 2020).

Mortality at eclosion
In order to assay the effect of viral genes on viability, a balancer system was used. UAS-SARS-CoV-2 gene OE ies were crossed with a Tub-Gal4/TM3, Sb line. Offspring either carry UAS-SARS-CoV-2 gene OE/TM3, Sb which carry the balancer chromosome resulting in orange eye color, shortened (stubbly; Sb) hairs on the back and no transgene expression, or they carry UAS-SARS-CoV-2 gene OE/Tub-Gal4, resulting in expression of the SARS-CoV-2 gene driven by Tub-Gal4 and transgenic ies with the typical red eyes and long hair (Figure 2A). Embryo progeny were collected and allowed to develop under standard conditions. Mortality at eclosion (adult emergence from pupa stage) was based on the percentage of ies with SARS-CoV-2 gene expression (red, long) that failed to emerge as adults, relative to siblings that did not express the SARS-CoV-2 gene construct (orange, short). The result was presented as a Mortality Index calculated as: (long hair -short hair) / short hair X 100.

Adult survival assay
Following egg laying Drosophila larvae were kept at 25°C, standard conditions and an optimal temperature for UAS-transgene expression. Adult male ies were maintained in vials in groups of 20 or fewer. Number of life ies in each group was recorded every second day. Drosophila lifespan is typically 50-60 days for wild type ies. The assay was ended when no survivors were left for any of the transgenic lines. A 100 ies were assayed per genotype.

Climbing assay
Climbing male ies were monitored by analyzing their performance to climb 6 cm in a horizontal tube within 14 sec. A successful attempt was scored as 1, and failure to reach the top (6 cm line) as 0. Each y was assessed ve times to calculate the average climbing score. At least 30 ies per genotype were analyzed.

Drosophila wing observation
Wing position was observed using a ZEISS Stemi 305 Stereo Zoom microscope with 5:1 zoom. A total of 200 ies were counted (4 experimental replicates, 50 ies per group each time). Representative images were taken using a ZEISS SteREO Discovery.V12, modular stereo microscope with 12x zoom.

Drosophila indirect ight muscle imaging
Flies were dissected and xed for 30 min in 4% paraformaldehyde in phosphate-buffered saline (1x PBS). Indirect ight muscle bers from both wings were assayed. Alexa Fluor 647 phalloidin was obtained from Thermo Fisher. To label mitochondria, mouse anti-Atp5a antibody was used at 1:500, followed by Alexa Fluor 488-conjugated secondary antibodies. Confocal imaging of the Drosophila indirect y muscle was performed using a ZEISS LSM900 confocal microscope with a 63× Plan-Apochromat 0.8 N.A. oil objective. For quantitative comparisons of uorescence intensity, common settings were chosen to avoid oversaturation. ImageJ Software (version 1.52a) (Schneider, Rasband, & Eliceiri, 2012) was used to process images. Mitochondria were counted manually from the images, based on visual observation of rounded morphology separating individual segments (see Figure 6C, w 1118 for example).

Imaging Drosophila tracheal branching
For tracheal observations, 3 rd instar larvae were dissected as described previously (F. Chen, 2016). Six larvae were dissected for each group. Images of the tracheal branches for each were obtained using a using a ZEISS SteREO Discovery.V12, modular stereo microscope with 12x zoom. Tracheal branches in each segment were counted manually.

Selinexor treatment
Selinexor (Selleck, # KPT-330) was dissolved in dimethyl sulfoxide (DMSO; Sigma) and added to standard y food at indicated doses (0.1, 0.2, and 0.5 μM). For untreated, control (0 μM), DMSO alone was added to the food. Flies were treated from rst instar larval stage until the adult ies hatched.

Statistical analysis
Statistical tests were performed using PAST.exe software (http://folk.uio.no/ohammer/past/index.html) unless otherwise noted. Data were rst tested for normality by using the Shapiro-Wilk test (a=0.05). Normally distributed data were analyzed either by Student's t-test (two groups) and Bonferroni comparison to adjust the P value, or by a one-way analysis of variance followed by a Tukey-Kramer posttest for comparing multiple groups. Non-normal distributed data were analyzed by either a Mann-Whitney test (two groups) and Bonferroni comparison to adjust P value, or a Kruskal-Wallis H-test followed by a Dunn's test for comparisons between multiple groups. Statistical signi cance was de ned as P<0.05.

Declarations
Ethics approval and consent to participate Not applicable Consent for publication -The authors give our consent for publishing this work in Cell & Bioscience.
Availability of data and material -All data and materials generated in this study are available publicly upon request. Images of adult progeny emerging from pupa stage from cross in (A), distinguished by carrying the balancer (TM3, Sb; orange eyes and short hair on back; no viral transgene expression) or with expression of the SARS-CoV-2 gene driven by the ubiquitous Tubulin (Tub) enhancer (red eyes and long hair). w1118 is a wild type control. (C) Quanti cation of mortality rate prior to eclosion for the individually expressed SARS-CoV-2 genes from the cross in (A). Mortality calculated as: (long hair -short hair) / short hair X 100. (D) Graph displaying lifespan data for adult ies carrying SARS-CoV-2 Nsp6, Orf6, Orf7a, Nsp3 or Orf3a transgenes. w1118 is a wild type control. N=100 ies per group. Abbreviations: E, envelope protein; M, membrane protein; N, nucleocapsid protein; Nsp, non-structural protein; OE, overexpression; Orf, accessory protein; S, spike protein; Sb, stubble TM3, chromosome 3.