A comprehensive Drosophila resource to identify key functional interactions between SARS-CoV-2 factors and host proteins

SUMMARY

Development of effective therapies against SARS-CoV-2 infections relies on mechanistic knowledge of virus-host interface. Abundant physical interactions between viral and host proteins have been identified, but few have been functionally characterized. Harnessing the power of fly genetics, we develop a comprehensive Drosophila COVID-19 resource (DCR) consisting of publicly available strains for conditional tissue-specific expression of all SARS-CoV-2 encoded proteins, UAS-human cDNA transgenic lines encoding established host-viral interacting factors, and GAL4 insertion lines disrupting fly homologs of SARS-CoV-2 human interacting proteins. We demonstrate the utility of the DCR to functionally assess SARS-CoV-2 genes and candidate human binding partners. We show that NSP8 engages in strong genetic interactions with several human candidates, most prominently with the ATE1 arginyltransferase to induce actin arginylation and cytoskeletal disorganization, and that two ATE1 inhibitors can reverse NSP8 phenotypes. The DCR enables parallel global-scale functional analysis of SARS-CoV-2 components in a prime genetic model system.

Graphical abstract

In brief

Guichard et al. develop Drosophila tools for conditional tissue-specific expression of individual SARS-CoV-2 factors and candidate human binding partners, as well as functional characterization of their fly orthologs. This resource is shared with the scientific community to promote analysis of in vivo interactions between viral and host factors.

INTRODUCTION

The current COVID-19 pandemic crisis has inflicted catastrophic global impacts on public health and the economy.^1–3 Moreover, recurrent waves of infection have produced mutations altering viral infectivity and/or pathogenicity.⁴ SARS-CoV-2 can also cause a myriad of persisting symptoms often referred to as long-COVID or post-acute sequelae of COVID (PASC), which may pose a continuing public health burden in the future. Gaining mechanistic knowledge of SARS-CoV-2 interaction with host processes at the molecular, cellular, and tissue/organ levels is important for developing new therapeutic strategies to meet these current and future challenges.

SARS-CoV-2 is an enveloped, positive-sense, single-stranded RNA virus, with a 29 kb genome encoding seven open reading frames (ORF1–7) and 29 proteins. The first event following cell entry is the direct translation of ORF1ab into a polyprotein, which is self-processed to liberate 16 separate non-structural proteins (NSP1–16). The early-acting NSP proteins (NSPs) encoded by Orf1a engage in a variety of interactions with host cellular processes as well as the viral replicative machinery encoded by ORF1b.^5,6 In addition, proteomic studies have identified a wealth of physical interactions between viral and host proteins.^7–12 The functional significance of most of these interactions, however, remains to be determined.

Here we report the development of a comprehensive Drosophila COVID-19 resource (DCR), an extensive toolkit to systematically probe the functional interactions between viral and host proteins using the fruit fly Drosophila melanogaster as an in vivo discovery system. The design of these reagents and the strategies employed parallel previous studies delineating viral protein functions in flies. For example, one study relating to the influenza virus¹³ and another to the Zika virus^14–16 led to insights into the molecular functions of viral factors. In the former study, we identified and characterized a novel Hedgehog (Hh) pathway modulating activity associated with the influenza protein NS1.¹³ In a second study, we documented that the Zika virus protein NS4A acts through the host ANKLE2 protein to disrupt neural development and induce microcephaly.^16–20 Drosophila has also been employed to study a variety of pathogenic factors^17–19 including proteins relevant to SARS-CoV-2 infection,^20–22 such as ACE2-related receptors,²³ ORF6a,²⁰ and NSP6.²⁴ These compelling Drosophila-based mechanistic discoveries prompted us to generate the DCR, a comprehensive collection of publicly available reagents to accelerate COVID research.

The DCR leverages the full power of advanced Drosophila genetics to enable parallel exploration of basic biological features associated with SARS-CoV-2 infection (Figure 1). We first created transgenic lines encoding each viral protein to produce characteristic phenotypes. In parallel, based on data from a previous proteomics study which identified 332 human proteins physically binding to viral factors,⁹ we created 234 transgenic strains carrying full-length human cDNAs. Co-expression of these human proteins with the corresponding binding viral proteins can reveal whether the human and viral proteins functionally interact. We also created a library of 313 mutant stocks of the fly homologs of the human genes/proteins interactors by inserting T2A-GAL4 or Kozak-GAL4 encoding cassettes.^25–27 This comprehensive set of GAL4 insertional strains can be employed to determine loss-of-function (L-o-F) phenotypes associated with the targeted genes and their expression patterns or to express human interacting factors in patterns mimicking those of their fly homologs for “humanization” assays. The flexible and integrated features of the DCR permit rapid identification of phenotypes associated with expression of SARS-CoV-2 proteins in specific cell types, tissues, or organs and provide the genetic leverage to perform in-depth analysis of fly phenotypes to establish the mechanistic basis for virus-host interactions (Figure 1). We provide several proof-of-concept examples for employing the DCR toolbox to uncover new functions of SARS-CoV-2 proteins in vivo, highlighting the discovery of a novel interaction between the viral NSP8 protein human ATE1 arginyltransferase, which modifies actin and alters the cellular cytoskeleton.

Figure 1. Functional analysis of SARS-CoV-2 factors in Drosophila.

Three types of reagents were created for expression and functional analysis of SARS-CoV-2 factors and dedicated host interactors in Drosophila (left boxes): UAS lines expressing viral factors (top), UAS lines expressing their human binding partners (middle), and GAL4 gene trap lines allowing “humanization” for the fly homologs of these human genes (bottom). These reagents are used to produce specific phenotypes pertaining to lethality, wing and eye development, and neural function (top right). Phenotypes produced by viral factors can be used to probe for genetic interactions with human candidate genes (middle right). Results obtained from phenotypic and genetic analysis can be validated through further investigation: testing activity of selected pathways and testing co-expression of viral factor with RNAi of corresponding fly homologs. T2A-GAL4 and Kozak-GAL4 gene trap lines can be used to test functional conservation between fly and human orthologs and examining expression patterns of these fly genes (bottom right).

RESULTS

We systematically evaluated the activities of SARS-CoV-2 factors and their interactions with human candidate partners in Drosophila (Figure 1) by creating three classes of reagents.

UAS-SARS-CoV-2 transgenic lines

The SARS-CoV-2 genome encodes 16 non-structural proteins (NSP1–16), four structural proteins (spike [S], envelope [E], membrane [M], and nucleocapsid [N]), and nine putative accessory factors (ORF3a, ORF3b, ORF6, ORF7a, ORF7b, ORF8, ORF9b, ORF9c, and ORF10)⁹ (Figure 2A). We obtained SARS-CoV-2 ORF plasmids,²⁸ encoding 28 of the 29 viral proteins. We used these constructs to generate a collection of transgenic lines carrying upstream activator sequence (UAS)-driven constructs that permit cell- or tissue-specific expression in the presence of the GAL4 trans-activator protein.²⁹ Since NSP11 was not included in the publicly available SARS-CoV-2 plasmid collection, we subcloned it separately. The sequence of the proteins is based on the “2019-nCoV” reference strain (GenBank: MN908947).³⁰ We initially generated two versions for each viral transgene, one followed by a stop codon (closed clones) and another lacking the stop codon (open clones), into pGW-HA.attB³¹ (Figure 2B). A closed clone will produce an untagged protein upon expression of GAL4, while an open clone will encode a protein with a C-terminal 33HA (hemagglutinin) tag. Non-tagged proteins are likely to behave more similarly to the endogenous viral proteins while tagged proteins would permit protein quantification, subcellular localization analysis, and identification of host interacting proteins via co-immuno-precipitation.³² The UAS-cDNA expression cassettes were inserted into the same genomic locus on the second chromosome (VK37 docking site) via φC31-mediated transgenesis³³ (Table S1). We also generated UAS-Kozak-NSP lines with an optimized Kozak sequence inserted before the viral cDNA for the first ten early-acting NSPs (NSP1–10). This modification leads to a significant increase in translational efficiency and concomitant potency of the viral factors (Figures 2B and S3).

Figure 2. Generation of UAS-SARS-CoV-2, UAS-human-cDNA, and T2A-/Kozak-GAL4 stocks.

(A) Genome organization of SARS-CoV-2. The viral genome encodes 29 proteins, including 16 non-structural proteins, four structural proteins, and nine accessory factors.

(B) Structure of UAS-SARS-CoV-2 transgenes. Each transgene was inserted into the same genomic site using φC31-mediated insertion. The total number of transgenes is indicated. “W+” indicates the mini-white genetic marker.

(C) Structure of transgenes expressing human interactors. UAS-human cDNAs were generated either with a native stop codon or with a C-terminal 33HA tag. “W+” indicates the mini-white marker.

(D) Strategies to generate T2A-GAL4 and Kozak-GAL4 alleles using CRISPR-mediated homologous recombination. The T2A-GAL4 cassette is inserted into a suitable coding intron (intron flanked by two coding exons) of the target locus. The Kozak-GAL4 cassette replaces the coding sequence when no suitable introns are present. P, attP; F, FRT; SA, splice acceptor.

UAS-human cDNA lines

A set of 332 human proteins physically interacting with individual SARS-CoV-2 proteins has been previously identified.⁹ We obtained full-length human cDNAs from public or commercial sources encoding 261 of these proteins, subcloned them into the pGW-HA.attB or pUASg-HA.attB UAS plasmids,³¹ and generated fly lines carrying UAS-human cDNA constructs via φC31-mediated transgenesis³³ (Figure 2C and Table S2). These human SARS-CoV-2-interacting proteins have been implicated in diverse biological processes that are potentially impacted by SARS-CoV-2 infection, such as endoplasmic reticulum (ER) homeostasis and stress response, protein glycosylation, mitochondrial function, centrosome and cytoskeletal dynamics, vesicular trafficking, and exosome formation.⁹ Because components of such core cellular pathways are highly conserved, they are readily amenable to functional studies in Drosophila. Indeed, 90% of the identified human proteins that interact with SARS-CoV-2 viral proteins have ortholog candidates in flies.²¹ To date we have created 242 UAS-human cDNA stocks, and 19 more lines are currently under construction (Table S2).

T2A-GAL4 and Kozak-GAL4 insertions in fly ortholog candidates of human interactors

An important aspect of our assessment of interactions between SARS-CoV-2 factors and candidate human proteins is to establish the function of their Drosophila ortholog candidates. The 332 human genes are predicted to be homologous to 407fly genes (using the ortholog predictor DIOPT, with scores ≥3).³⁴ There are more fly ortholog candidates than human genes because some human genes have several fly orthologs. We inserted an artificial exon composed of attP-FRT-SpliceAcceptor(SA)-T2A-GAL4-polyA- 3XP3EGFP-polyA-FRT-attP into coding introns (introns that separate two coding exons) for 177 out of 186 genes that have a sufficiently large coding intron using CRISPR-mediated homology-directed repair (HDR) or by recombinase-mediated cassette exchange (RMCE) of existing MiMIC lines^{25,27,35–37} (Figure 2D). These inserts typically create strong L-o-F alleles of the targeted gene. Additionally, such alleles can be employed to determine the expression patterns of targeted genes and establish whether the human and fly genes have conserved molecular functions through rescue experiments.^38–47 We targeted another 136 out of 210 genes that do not have a suitable coding intron employing a knockout/knockin approach using Kozak-GAL4-FRT-3XP3EGFP-FRT cassette to replace the coding sequence of the gene via HDR.²⁶ Additionally, in three cases where the only available intron is in the 5′ UTR, we opted to replace the coding region with a SA-KozakGAL4 cassette²⁶ (Figure 2D). These alleles permit rescue strategies similar to those employed with the T2A-GAL4 insertions. In summary, we generated GAL4 alleles for 313 fly gene ortholog candidates corresponding to 216 of the 332 human genes that encode for SARS-CoV-2 interactors (Table S3).

Thirteen out of 29 SARS-CoV-2 factors cause lethality when ubiquitously expressed

As a first step in assessing the functions of individual SARS-CoV-2 proteins, we determined whether ubiquitous expression of these viral factors causes lethality (Figure 1). We selected three ubiquitous drivers: Actin(Act)-GAL4, Tubulin(Tub)-GAL4 (strong drivers), and daughterless(da)-GAL4 (milder driver). Among the 29 SARS-CoV-2 proteins tested expressed by the non-Kozak insertions, 13 (NSP1, NSP3, NSP5, NSP6, NSP13, NSP14, NSP15, E, M, ORF3a, ORF6, ORF7a, and ORF7b) caused lethality using at least one driver and seven proteins caused lethality with all three (Figures 3A and 3B), suggesting disruption of an essential activity. The stages of lethality produced by all transgenes with da-GAL4 are listed in Table S4. We also systematically compared the severity of phenotypes caused by overexpression of the native versus HA-tagged transgenes (Figure S3). These experiments revealed that 16 genes (55%) caused comparable lethality whether tagged or not, while 11 genes (38%) generated more severe phenotypes when HA tagged (Figure 3C). Since previous studies have demonstrated that HA tags do not typically alter protein levels, distribution, or activity,³¹ the basis for these differences will need further examination.

Figure 3. Functional evaluation of SARS-CoV-2 cDNAs through ubiquitous expression.

(A) Lethality phenotypes resulting from overexpression of UAS-SARS-CoV-2 and UAS-SARS-CoV-2-HA transgenes driven by ubiquitous GAL4s (Tub-GAL4, Act-GAL4 and da-GAL4).

(B) Venn diagram summarizing lethality based on ubiquitous overexpression of SARS-CoV-2 transgenes.

(C) Severity of lethal phenotypes caused by UAS-SARS-CoV-2 and UAS-SARS-CoV-2-HA ubiquitous overexpression. Adding the HA tag tends to cause more severe phenotypes.

For (A) to (C), the flies were raised at 29°C.

Photoreceptor-specific expression of SARS-CoV-2 proteins alter neural function

Numerous reports have documented entry of the SARS-CoV-2 virus into brains of infected patients,⁴⁸ and approximately 70% of hospitalized COVID-19 patients present neurological symptoms.⁴⁹ We therefore characterized the effects of a subset of SARS-CoV-2 proteins in the nervous system by overexpressing individual viral proteins in photoreceptor neurons (using Rhodopsin1(Rh1)-GAL4) and performed electroretinogram (ERG) recordings on adult flies (Figure S1A). Of the seven viral transgenes that cause lethality with all three GAL4 lines, six caused ERG defects (NSP1, NSP5, NSP13, NSP15, ORF3a, and ORF6; Figure S1B). Among them, NSP5 severely altered ERGs, including a marked reduction in on-transient, off-transient (both indicating defects in neuronal connection or synaptic transmission), and depolarization (reflecting defects in phototransduction), at 10 days post eclosion (Figure S1C). The other five proteins displayed more modest and age-dependent ERG defects. In these lines, reduction of on-transient, off-transient, and/or depolarization defects became obvious only by 20 days post eclosion (Figures S1B and S1D–S1F). These data suggest that SARS-CoV-2 proteins disrupt multiple neurological processes, several of which worsen with age.

An allelic variant in ORF3a reduces protein activity

As is typical for RNA viruses, the genome of SARS-CoV-2 has evolved since its original documentation.⁵⁰ Determining which of the emerging mutation(s) alter pathogenicity represents a significant challenge. For example, the accessory protein ORF3a encodes a viroporin that promotes apoptosis and viral release.^51–55 The p.Q57H variant in ORF3a was initially proposed to be associated with elevated infectivity and mortality.^56,57 However, another study reported that the activity of ORF3a^p.Q57H is comparable to that of the reference ORF3a allele.⁵⁸ We compared the relative activities of these two alleles by generating a strain of flies carrying the UAS-ORF3a^H57 allele inserted into the same genomic site as the reference UAS-ORF3a^Q57 allele. Expression of reference ORF3a using a variety of GAL4 drivers caused lethality at the embryonic stage (Figures S2A and S2B). Expression of ORF3a^p.Q57H in the developing wing (nubbin-GAL4), pan-neuronally (elav-GAL4), or ubiquitously (da-GAL4) also caused lethality, but at later developmental stages. When expressed in developing wings, ORF3a^p.Q57H causes significantly milder phenotype than that produced by the reference ORF3a (Figures S2C and S2D). Consistently, expression of ORF3a^p.Q57H in the developing eye and head using the ey-GAL4 driver caused milder morphological defects than the reference protein (Figure S2E). This difference was not associated with reduced protein levels (Figure S2F), indicating that the p.Q57H variant is less toxic than the reference ORF3a protein.

Expression of NSP1–10 in developing wings produces distinct phenotypes

The Drosophila wing can be employed as a sensitive in vivo system for assessing biological activities of candidate virulence factors based on overexpression phenotypes and genetic epistasis analyses.^13,59–61 The wing consists of five major longitudinal veins (L1–L5) (Figure 4A). Defects in major developmental signaling pathways result in characteristic wing phenotypes.⁵⁹ For example, loss of epidermal growth factor receptor/ mitogen-activated protein kinase (EGFR/MAPK) signaling causes vein truncations and small elongated wings,^62,63 reduced bone morphogenetic protein (BMP) signaling decreases L2-L3 and L4-L5 vein spacing,^64,65 loss of Hh signaling reduces L3-L4 vein spacing,^66,67 and reduced Notch signaling leads to thickened veins and gaps along the wing margin⁶⁸ (Figure 4A). Hence, wing phenotypes induced by viral factors can provide clues about which signaling pathways they impact. For this analysis, we employed the UAS-Kozak-NSP lines to produce the NSP1–10 proteins at higher expression levels (Figures S3A–S3C).

Figure 4. NSP1–10 cause specific wing phenotypes.

(A) A wild-type wing with key signaling pathways and their roles in patterning during development. Hedgehog signaling (Hh, light blue) defines the distance between central veins L3 and L4. Dpp signaling (purple) defines the positioning of outer veins L2–L5. Notch signaling (green) defines vein thickness and margin fate. EGFR/MAPK signaling (black) controls cell proliferation/survival and vein fate. Scale bar represents 500 μm.

(B) A wild-type wing disc (the larval wing primordium) expressing mCherryNLS (magenta) driven by a strong dorsal-wing-specific GAL4 (MS1096 referred to as wingGAL4) and stained with an anti-Delta antibody (white). The Notch ligand Delta marks the position of the margin and longitudinal veins L3–L5. Scale bar represents 100 μm.

(C) Specific wing phenotypes caused by UAS-Kozak-NSP1–10 expressed under the control of wingGAL4 driver. Four expression levels were defined by different conditions to obtain viable and visible phenotypes: low expression (+, tubulinGAL80ts lowers expression to obtain viable males at 25°C), moderate expression (++, females raised at 18°C), high expression (+++, males raised at 25°C), and very high expression (++++, two copies of the UAS transgene in individuals raised at room temperature). Colored arrows point to specific phenotype features in relation to the signaling pathways depicted in (A). Indigo arrows indicate blisters reflective of cell adhesion defects; gray arrows point to curved wings indicating cell proliferation/survival defects. Scale bar represents 500 μm.

(D) Imaginal discs stained with an anti-activated MAPK antibody (pMAPK). Elevated pMAPK marks the margin primordium and veins L3–L5 in a wild-type disc (left). In an NSP3-expressing disc (right), pMAPK staining is reduced, as predicted by the loss-of-vein phenotype in (C). Lower panel shows a pMAPK intensity profile for four discs of each genotype: NSP3-expressing discs show consistently lower pMAPK levels.

(E) Imaginal discs stained with an anti-phosphorylated Mad antibody (pMad), which reflects activation of Dpp signaling. pMad signal is elevated in a central zone between L2 and L5 vein primordia. In an NSP8-expressing disc (right), pMad staining is greatly reduced. Lower panel shows a pMad intensity profile for four discs of each genotype: NSP8-expressing discs show consistently lower pMad levels, as predicted by the adult wing phenotype in (C). Scale bars in (B), (D), and (E) represent 100 μm.

We exploited a wing-specific GAL4 driver (MS1096,⁶⁹ referred to as wingGAL4 hereafter) to express individual NSP proteins throughout the developing wing pouch, with higher levels in the dorsal surface (Figure 4B). Expression of all but one of the NSPs tested generated reproducible, penetrant, and distinctive phenotypes that depended on transgene expression levels (Figure 4C). These phenotypes include small misshapen wings with blisters suggesting loss of cell adhesion (NSP1 and NSP5), small curled wings suggesting reduced growth (NSP2, NSP3, and NSP6), truncated veins reflective of reduced EGFR/MAPK signaling (NSP3 and NSP6), thickened L3 and L5 veins typical of impaired Notch signaling (NSP4, NSP6, and NSP10), reduced L3-L4 spacing reminiscent of reduced Hh signaling (NSP5), approximation of the L2 and L3 veins suggesting decreased BMP signaling (NSP6 and NSP8), and ectopic veins (NSP9) as seen in net⁷⁰ or plexus⁷¹ mutants. We find that NSP1 and NSP3–5 have the strongest effects (expression of these factors needs to be reduced to obtain viable flies), NSP6 and NSP8 have intermediate strength, while NSP2, NSP9, and NSP10 have weak activity. Only NSP7 failed to produce any phenotype. These relative phenotype severities in the wing parallel observations from the lethality analyses (Figures 3 and Figure S3A).

Because phenotypes from overexpressing NSP3 and NSP6/8 resemble those caused by compromised EGFR/MAPK and BMP signaling, respectively, we assessed the activities of these pathways in larval wing imaginal discs using antibodies specific for the activated forms of the MAPK (pMAPK)⁷² or Mad (pMad) pathway effectors.⁷³ Paralleling the observed adult phenotypes, wing discs expressing NSP3 displayed greatly reduced pMAPK staining (Figure 4D), as observed previously with anthrax lethal factor.⁷⁴ Similarly, wing discs expressing NSP6 or NSP8 exhibited reduced levels of pMad staining (Figures 4E and S4). These congruent observations, linking reduced activities of specific signaling pathways in larvae with adult phenotypes, indicate that the viral factors are likely to disrupt these canonical signaling pathways.

NSP3 and NSP5 are proteases (PLpro and 3CLpro, respectively) that process the primary SARS-CoV-2 pp1a and pp1ab polyproteins. We wondered whether the catalytic activities of these enzymes are required for producing wing phenotypes. For NSP5, the p.C145A mutation of a critical catalytic residue greatly diminishes phenotype severity (Figure S5A), suggesting that NSP5 cleaves essential host proteins. In contrast, the p.C857A mutation of NSP3 did not alter its activity, revealing that NSP3 protease activity is not required to produce its wing phenotype (Figure S5B). These different outcomes highlight the utility of the DCR and sensitive wing assays to parse out contributions of specific domains or the relevance of catalytic activities of viral proteins.

NSPs exhibit potent synergistic interactions in the wing

Some SARS-CoV-2 proteins have been shown to physically bind to each other,^10,75–78 although the functional consequences of these interactions on host cells remain to be determined. An informative approach for identifying viral factors working in concert is to co-express them and test for potential synergistic or inhibitory interactions.

We expressed NSPs in pairwise combinations in the developing wing, focusing our analysis on NSPs producing viable offspring with the wingGAL4 driver (NSP2, NSP6, and NSP7–10). We observed several consistent cooperative interactions among these viral factors resulting in much more severe phenotypes than those produced by individual NSPs. While weak to moderate-strength genetic interactions were observed in some pairwise combinations (NSP2–NSP9, NSP2–NSP7, and NSP2–NSP10, Figure 5E), NSP8 stood out by exhibiting strong interactions with several other NSPs, most notably synergizing with NSP6, NSP7, NSP9, and NSP10 (Figure 5). When expressed at submaximal levels (+++), NSP7, NSP9, and NSP10 generated no phenotypes on their own. However, when combined with NSP8, they each strongly enhanced its wing phenotype (Figures 5A and 5B). Similarly, under weaker expression conditions where NSP6 by itself caused a moderate phenotype (Figure 5C) and NSP8 alone caused no phenotype, co-expression of NSP6–NSP8 resulted in a strong enhancement (Figures 5C and 5D). Since NSP6 and NSP8 each produce BMP-like phenotypes (Figures 4C, 4E, and S4), their synergistic interaction presumably reflects converging effects on this pathway.

Figure 5. NSP8 interacts with NSP6, NSP7, NSP9, and NSP10.

(A) Wing phenotypes from male flies of the indicated genotypes raised at 25°C. Under these conditions, NSP7, NSP9, and NSP10 do not produce any phenotype (wingGAL4>NSP in a CyO background, left panels). For flies co-expressing NSP8 + NSP7, NSP8 + NSP9, and NSP8 + NSP10, clear synergistic phenotypes are shown on right panels.

(B) Graph showing wing area quantifications of the genotypes shown in (A), with significant reduction in wing size when factors are expressed in pairwise combinations. Number of biological replicates is indicated. ****p < 0.0001.

(C) Wing phenotypes from wingGAL4>NSP8, wingGAL4>NSP6, and wingGAL4>NSP6 + NSP8 females raised at 18°C, showing synergism between NSP6 and NSP8.

(D) Graph showing wing area quantifications of genotypes shown in (C), with significant reduction in wing size when both factors are expressed. Number of biological replicates is indicated. ****p < 0.0001; ns, not significant.

(E) Summary diagram of genetic interactions found between NSP factors. NSP2 and NSP6–10 were tested in these co-expression experiments, but not NSP1 and NSP3–5, which require tubGAL80ts to obtain viable phenotypes. NSP8 displays strong or moderate positive interactions with NSP6, NSP7, NSP9, and NSP10, while other factors show no or weak interactions.

The interaction of NSP8 with NSP7 is also informative because NSP7 does not generate any phenotype even when expressed at the highest levels. Thus, its synergism with NSP8 reveals its activity in wing cells. This genetic interaction is consistent with prior studies indicating its direct binding to NSP8 to form a hexadecamer complex.^77,78 Thus, with its multiple synergies revealed in the wing, NSP8 appears to act as an interaction hub, coordinating some of the activities of NSP6, NSP7, NSP9, and NSP10 in vivo. We conclude that the Drosophila wing is a sensitive system for detecting specific functional connections among NSPs.

NSPs functionally interact with some human host factors identified by mass spectrometry

Comprehensive proteomic studies have identified a multitude of human proteins that may engage in virus-host interactions.^7–10,12 These proteome-wide studies call for further analysis to establish which of these many physical interactions have functional relevance in vivo and how they may contribute to viral infection. Because NSP8 emerged as a central hub interacting with four other NSPs (Figure 5), we examined potential functional connections between NSP8 and its human binding candidates.^8,9 Twenty-four host proteins were reported to bind to NSP8,⁹ including constituents of the mitochondrial ribosome, exosome components, signal recognition particle (SRP) subunits, and other activities (Figure 6A). We expressed each of the 17 available individual human NSP8-interacting proteins alone or with NSP8 in the wing. Nine of these factors did not modify the baseline NSP8 wing phenotype, nor did they cause any phenotype alone (Figure S6A). In contrast, seven factors (ATE1, NSD2, EXOSC2, EXOSC5, NARS2, DDX10, and MRPS27) enhanced the NSP8 phenotype to varying degrees (Figures 6A–6C, S6B, and S6C), and only SRP19 produced a weak suppression of the NSP8 phenotype. When expressed alone, most of these genes had no effect except for NSD2 and EXOSC2, which produced mild and moderate small wing phenotypes, respectively. Consistent with EXOSC2 and NSP8 interacting strongly, the phenotypes of EXOSC2 were significantly exacerbated by co-expression with NSP8 even under conditions of lower expression where NSP8 had little effect alone (Figure S6C, right panel). Among the most notable enhancers of the NSP8 phenotype were EXOSC5 and EXOSC2, two components of the RNA exosome complex,⁷⁹ and ATE1, an arginyltransferase that modifies specific proteins to alter their activity or targets them for degradation.⁸⁰ Most dramatically, co-expression of ATE1 with NSP8 reduced wings to a mere rudiment (Figures 6B and 6C). Consistent with this synergism, RNAi knockdown of the endogenous Drosophila ortholog of ATE1 (Ate1) with either of two independent lines resulted in robust suppression of the NSP8 phenotype, indicating that this protein is an important mediator of basal NSP8 activity in the wing (Figures 6B and 6C). We find that theT2A-GAL4 insertion into the Ate1 locus is homozygous lethal, suggesting that Ate1 is an essential Drosophila gene. Expression of ATE1 driven by the T2A-GAL4 in an Ate1 mutant background partially rescued this lethality (Figure S7), demonstrating that fly Ate1 and human ATE1 are functionally conserved.

Figure 6. NSP8 interacts functionally with a set of human candidate genes.

(A) Diagram summarizing the results from co-expression experiments involving NSP8 and individual human candidate cDNAs. The activity of the human cDNA alone was also assessed (in wing-GAL4>h-cDNA; CyO flies). Strong genetic interactions are indicated with a dark-pink circle, moderate interactions with a pink circle, and weak interactions with a light-pink circle. Suppression of the NSP8 phenotype is indicated with a blue circle; no interaction is indicated with a gray circle.

(B) Examples of strong and moderate interaction phenotypes between NSP8 and human candidate genes ATE1 and EXOSC5 (right panels), which alone do not produce any phenotype (middle panels). The NSP8 phenotype is strongly suppressed by co-expression of an h-ATE1 RNAi, indicating that the NSP8 phenotype is mediated by endogenous d-Ate1 (left bottom panel).

(C) Graph showing wing area quantifications of the genotypes shown in (B), with significant reduction in wing size when NSP8 is co-expressed with human candidate gene ATE1 or EXOSC5. d-ATE1 RNAi alone does not produce any phenotype but significantly suppresses the NSP8 phenotype. Number of biological replicates is indicated. ****p < 0.0001; ns, not significant.

NSP8 synergizes with ATE1 to induce actin arginylation and cytoskeletal alterations

Our results indicate that NSP8-induced wing phenotypes are in part mediated by Ate1. Protein arginylation plays various roles including regulation of cytoskeletal dynamics, particularly the F-actin network, during embryogenesis, cardiovascular development, and angiogenesis.^81,82 Furthermore, the actin cytoskeleton is disrupted during infection with SARS-CoV-2, as illustrated by characteristic spherical inclusions in SARS-CoV-2-infected pulmonary cells.⁸³ We therefore evaluated the impact of NSP8 and ATE1 on the actin cytoskeleton. In wild-type wing imaginal discs, polymerized F-actin was primarily restricted to an apical band underlying intercellular junctions as revealed by phalloidin staining. Fainter F-actin lines along the plasma membrane extending along the apico-basal axis were also observed (Figure 7A, left panel). This pattern was mildly altered in wing discs expressing either NSP8 or ATE1. NSP8 induced formation of a few intracellular F-actin foci (Figure 7A, second panel), whereas ATE1 produced irregular vertical F-actin labeling. These mild cytoskeletal phenotypes were notably enhanced by co-expression of ATE1 and NSP8, which profoundly disrupted the apical F-actin network and caused the appearance of many scattered F-actin-rich inclusions (Figure 7A, rightmost panel).

Figure 7. NSP8 cooperates with human ATE1 to deregulate actin dynamics and arginylation.

(A) F-actin distribution in control, NSP8, ATE1, and NSP8 + ATE1-expressing discs (using the wingGAL4 driver). The F-actin network is mildly deregulated in NSP8 (few F-actin accumulations, arrowheads) and ATE1 discs, but severely disorganized in discs co-expressing NSP8 and ATE1 (many F-actin accumulations, arrowheads). Scale bars represent 10 μm.

(B) In salivary glands, expression of NSP8 or ATE1 alone or in combination results in formation of abnormal F-actin structures (arrowheads). The wingGAL4 driver used here is also expressed in salivary glands. Scale bars represent 100 μm.

(C) Western blot analysis of actin arginylation, using an anti-arginylated β-actin (R-actin) antibody. Control (hsGAL4/+), hsGAL4>NSP8, hsGAL4>ATE1, and hsGAL4>NSP8 + ATE1 adult flies were heat shocked for 90 min and proteins were extracted after a 3 h 30 min at room temperature. Actin arginylation is increased by NSP8 and ATE1 alone and cooperatively by both factors. Total levels of β-actin are not altered by expression of either factor (bottom). Actin arginylation following NSP8 and ATE1 co-expression increases over time after heat shock (right).

(D) In situ detection of actin arginylation in salivary glands. In control tissues (wingGAL4/+, left panels), R-actin (red) appears as a dotted ubiquitous stain mildly elevated at cell-cell junctions, which correlates with local interruptions (arrowheads) of the F-actin (green) network. Upon co-expression of NSP8 and ATE1 (wingGAL4>NSP8 + h-ATE1, bottom panels), R-actin co-localizes with F-actin at points of cortical accumulations (arrows) and some weaker staining F-actin rings (arrowheads). Scale bar represents 10 μm.

(E) Wing size quantifications of control (w1118; WT) and wingGAL4>NSP8 males (from wingGAL4; NSP8/CyO females crossed to w1118 males at 25°C) on different concentrations of suramin (Sur), with or without 30 μM merbromin (Mer). In absence of any drug, NSP8 produces a small wing phenotype in males. This phenotype is suppressed significantly in the presence of 100 μM suramin or 25 μM suramin + 30 μM merbromin. Addition of 30 μM merbromin also significantly ameliorates the NSP8 phenotypes obtained with 25 μM or 100 μM suramin. This suppressive effect is specific to NSP8-expressing wings, as the highest drug concentration (100 μM suramin + 30 μM merbromin) does not increase wing size in control w1118 animals but rather produces a small reduction in wing size. ***p < 0.001, ****p < 0.0001.

We also examined salivary glands, for which high levels of F-actin mark the cell borders in control tissues. When expressed, NSP8 or ATE1 resulted in the appearance of ectopic cytoplasmic F-actin-rich inclusions (Figure 7B). Co-expression of NSP8 and ATE1 exacerbated these defects, resulting in a more irregular cortical F-actin network and in numerous atypical F-actin rings, similar to those reported in SARS-CoV-2-infected human cells,⁸³ as well as long F-actin filaments traversing the cytoplasm. We conclude that NSP8 acts in concert with ATE1 to disrupt F-actin organization in both wing discs and salivary glands.

We next examined whether NSP8/ATE1 induced arginylation of actin using an antibody specific for arginylated β-actin (R-actin)⁸⁴ for western blot and immunofluorescence analysis. In extracts from control adult flies (hs-GAL4/+), only a faint R-actin signal could be detected (42 kDa). Heat-shock-induced expression of NSP8 or ATE1 alone led to moderate increases in R-actin, while co-expression of NSP8 with ATE1 substantially elevated R-actin levels (Figure 7C, left panel), mirroring their strong positive phenotypic interactions observed in adult and larval wings. Actin arginylation increased sequentially following a single heat shock, indicating that this process is progressive (Figure 7C, right panel).

We also examined in situ actin arginylation in response to NSP8 and ATE1 co-expression in salivary glands. In control glands, we observed a unique subcellular distribution of R-actin relative to the F-actin network. R-actin in these cells displayed a punctate cytoplasmic staining pattern but also accumulated at intercellular junctions, particularly where the cortical F-actin stain was weaker, indicating that R-actin and F-actin have distinct but adjacent localization patterns (Figure 7D, left panels, arrows). In glands co-expressing NSP8 and ATE1, R-actin and F-actin co-localized in irregular cortical accumulations (Figure 7D, right panels, arrows). In addition, R-actin labeled a subset of ectopic F-actin cytoplasmic rings/spheres induced by NSP8/ATE1 co-expression (arrowheads), typically among those showing fainter F-actin labeling. Overall, the above experiments indicate that NSP8 activity is mediated to a great extent by ATE1. We tested this hypothesis further by feeding developing flies expressing NSP8 with suramin or merbromin, two chemical inhibitors of the ATE1 enzyme.⁸⁵ Both compounds, which have been also shown to reduce SARS-CoV-2 viral loads in cell-based infections,^86,87 ameliorated the NSP8 wing phenotype and did so in a cooperative fashion (Figure 7E). Cumulatively, these phenotypic, biochemical, and immunofluorescence data support a model in which ATE1 acts as a critical mediator of NSP8 activity, inducing actin arginylation, disrupting F-actin organization and producing ectopic F-actin-rich ring-like structures.

DISCUSSION

Human cells, organoids, and mammalian model organisms including mice, hamsters, minks, ferrets, shrews, and non-human primates have been used to study COVID-19.⁸⁸ While systemic infection of a mammalian model organism with SARS-CoV-2 can cause phenotypes reminiscent of disease symptoms in humans, it is often difficult to elucidate the precise molecular mechanisms underlying specific pathologies. In this study, we addressed these challenges by generating the DCR resource in Drosophila that enables multi-faceted parallel in vivo studies of virus-host interactions for SARS-CoV-2 to greatly accelerate the pace of research in this field.

Expression of individual viral genes can induce phenotypes that are sometimes very similar to those resulting from altered function of a fly gene,^13,14 pointing to potential biological processes or signaling pathways affected by the viral protein. In addition, if the human cDNA can rescue the loss of the fly gene, a compelling case can be made to study the function of the viral genes and their targets in vivo using Drosophila to guide further mechanistic analyses in vertebrates. The above reagents may also provide a robust platform to screen for drugs that suppress the phenotypes induced by UASSARS-CoV-2 genes or suppress the phenotypes associated with L-o-F mutations of the viral target gene. Similar approaches to suppress phenotypes associated with rare genetic disease have identified several Food and Drug Administration-approved drugs in flies.^89–92 Expression of seven SARS-CoV-2 encoded proteins using three different ubiquitous drivers caused lethality. Our study identifies the most toxic factors as NSP1, NSP3, NSP5, NSP13, NSP15, Orf3a, and Orf6, which induce lethality with three ubiquitous GAL4. These findings mirror to a large extent results from a fission yeast study in which these factors caused cell death and inhibited proliferation, suggesting that they act on broadly conserved targets.⁵⁸ Although it remains to be determined whether lethality in flies is caused by mechanisms that relate to mortality caused by COVID-19, lethality is a simple and robust phenotype to assess. We envision that screens to revert such lethality could be used to identify genes or small molecules that suppress the toxic effect of a viral protein.

Photoreceptor-specific overexpression experiments revealed that six out of seven SARS-CoV-2 proteins that cause lethality when ubiquitously expressed disrupt neuronal function in the visual system. Studies in monkeys have shown that SARS-CoV-2 affects cortical regions of the brain that receive direct olfactory input, suggesting that the virus may enter via the nose to infect neurons.⁹³ Such SARS-CoV-2 brain infections lead to neuroinflammation and neurodegeneration, a mechanism that may contribute to PASC in humans. PASC includes symptoms that relate to neuronal dysfunction including fatigue, headaches, memory impairment, and decreased concentration.^94,95 Since fly photoreceptors are amenable to many experimental assessments including histological and ultrastructural analysis,⁹⁶ this approach may provide molecular insights into how neurons are impacted in COVID-19 and PASC. Interestingly, six viral proteins cause age-dependent ERG defects, indicating that continuous long-term expression of these viral proteins causes progressive neuronal dysfunction.

Early-expressed NSPs generate distinct instructive wing phenotypes

Distinctive wing phenotypes induced by viral factors complement functional information obtained from lethality assays. Patterning defects can be diagnostic for disruption of known signaling pathways, providing clues for further functional analysis. For example, NSP3 adult phenotypes suggest impairment in EGFR/MAPK signaling, and NSP3 reduces downstream MAPK activation in wing vein primordia. Similarly, NSP6 and NSP8 generated vein-fusion phenotypes similar to those caused by reduced activity of the BMP (Dpp in flies) pathway^64,65 or expression of BMP antagonists.^65,97 Strengthening our observations, these two NSPs also reduced levels of pMAD, the primary BMP signaling transducer. Interestingly, SARS-CoV-2 infections trigger expression of transforming growth factor β (TGF-β) and BMP antagonists,^98–100 associated with damaging inflammation, apoptosis, and fibrosis.¹⁰¹ Thus, modulation of TGF-β/BMP signaling components by NSP6 and NSP8 during infection could be evaluated as targets for therapeutic intervention.¹⁰² Expression of other NSPs caused thickened veins typical of reduced Notch signaling (NSP4), blistering reflective of impaired cell adhesion (NSP1, NSP5), or ectopic veins (NSP9, NSP10), which could reflect ectopic MAPK or BMP pathway activation. Future molecular and genetic epistasis experiments should help define specific host components mediating the effects of individual NSPs to explore the connection between these activities and COVID-19.

Phenotypic analysis distinguishes activities of variant SARS-CoV-2 proteins

A defining feature of viruses is their ability to rapidly evolve, a characteristic which has proved to be a particular challenge in the case of SARS-CoV-2. Expression of SARS-CoV-2 proteins in Drosophila offers an avenue for assessing the functional effect of naturally occurring variants as well as for testing the role of specific amino acid residues using synthetically produced mutations. In this study, we provide two examples demonstrating the potential to resolve differing activities of SARS-CoV-2 protein variants in Drosophila.

First, we examined the phenotypic consequence of an ORF3a mutation (p.Q57H) that has been detected in multiple SARS-CoV-2 strains.¹⁰³ Our data indicate that the p.Q57H allele is less toxic to host cells than the reference allele in several settings. A possible interpretation for this result is that the weaker p.Q57H allele may have been selected because it promotes net viral infectivity. As has been demonstrated with other COVID strains, viral transmissibility and lethality can be tradeoffs: less-virulent strains frequently spread faster than those that are more lethal.¹⁰⁴ It should be informative to extend such functional analyses to variants arising in other SARS-CoV-2 encoded proteins reported in the CoV-GLUE-Viz database (http://vshiny1.cvr.gla.ac.uk/cov-glue-viz/). Similarly, we compared catalytically active and inactive forms of NSP3 and NSP5 proteases using the wing phenotype assay. We found that the NSP3 phenotype was independent of its protease activity. Further epistasis experiments using information from a recent NSP3 interactome study¹⁰⁵ should help define which host proteins may be functionally relevant for this phenotype. Of note, stress sensor ATF6, a protein that binds to NSP3¹⁰⁶ and regulates MAPK signaling,¹⁰⁷ is a prime candidate for further analysis. In contrast, the catalytically inactive NSP5^p.C145A caused a much milder wing phenotype compared to the reference NSP5. It will be informative to determine which host proteins are cleaved by NSP5 to mediate its phenotype. Candidates identified in a recent bioinformatics study should be considered for future functional screens.¹⁰⁶

NSPs interact synergistically with each other

NSPs form various complexes with each other to exert specific effector functions.^{10,75–78,108–110} In our initial survey of genetic interactions among NSPs in the wing, NSP8 stood out as a central “hub,” strongly interacting with several other NSPs. We note that both NSP7 and NSP9 are known to engage in direct protein-protein interactions with NSP8. Regarding NSP10, it has been reported to interact with the host protein CAND1, which also binds to NSP8¹⁰ and could potentially mediate the NSP8/NSP10 synergism. Alternatively, NSPs may impact distinct but intersecting pathways, potentially resulting in some of the observed genetic interactions.

NSP8 interacts in vivo with a subset of human candidate proteins

The prime motivation for developing the DCR was to determine which human interactors identified by proteomic studies are functionally relevant. We note that among interactors identified in the first interactome study,⁹ 44% are found at least once again in independent comparable studies (Table S5; Chen et al.,^7,10–12 Li et al.,^7,10–12 Stukalov et al.,^7,10–12 Zhou et al.^7,10–12). The broad diversity of binding partners identified in multiple independent studies calls for additional functional tests, including functional interrogation screens, or co-expression experiments such as those described here. In the case of NSP8, 24 human NSP8-interacting proteins were identified in an initial systematic study.⁹ Taking advantage of the DCR toolbox, we observed seven positive interactions and one suppression out of the 17 UAS-h-cDNA lines available. Two of the strongest interactors (EXOSC2 and EXOSC5) belong to the RNA exosome, which mediates degradation of various types of RNA molecules.^111,112 Interestingly, low expression of EXOSC2 has also been recently identified through genome-wide association studies to be protective against SARS-CoV-2 infection.¹¹³ Other noteworthy NSP8 interactors include DDX10, a member of the of DEAD box family of RNA helicases, which contain regulators of SARS-CoV-2 infection,^114,115 and NSD2, a histone methyltransferase which dimethylates nucleosomal histone H3 to regulate chromatin integrity and gene expression.¹¹⁶ We note that among the seven interactors identified in our experiments, only MRPS27 has been validated in a functional interrogation screen.

NSP8 synergizes with ATE1 to arginylate actin and disrupt F-actin networks

The strongest NSP8 interactor we identified was ATE1, an arginyltransferase which, when co-expressed with NSP8, produced a strong synergistic rudimentary wing phenotype. Conversely, the NSP8 phenotype was potently suppressed by knockdown of fly Ate1 or by two known inhibitors of human ATE1,⁸⁵ suggesting that a key cellular activity of NSP8 depends on this host arginyltransferase. ATE1 is known to act on several cytoskeletal proteins, including actin, to regulate F-actin levels and dynamics.⁸² Here we observed abnormal ring-shaped F-actin structures formed in response to NSP8/ATE1 co-expression, reminiscent of those induced by SARS-CoV-2 infection in pulmonary cells,⁸³ presumably resulting from misregulated arginylation. These spherical structures may correspond to vesicles derived from the cell surface or from internal organelles. We note that the underexplored arginylation pathway plays roles in several processes relevant to SARS-CoV-2-induced pathogenesis, such as coagulation,¹¹⁷ cardiac inflammation and fibrosis,¹¹⁸ cellular respiration,¹¹⁹ ER stress,^120,121 and neural activity.^119,122 Consistent with ATE1 playing a role during infection, a recent study finds that the ATE1 protein levels are increased in cultured cells infected by SARS-CoV-2 and that ATE1 inhibitors (suramin, merbromin, and tannic acid), as well as ATE1 silencing, reduce viral loads.^86,87 Moreover, SARS-CoV-2 patients show alterations in their global arginylation landscape.¹²³ Altogether, these elements support the idea that the ATE1/NSP8 tandem plays a relevant role during SARS-CoV-2 infection and represents an attractive therapeutic target.

In conclusion, we provide the research community with the comprehensive DCR resource to enable functional analysis of all SARS-CoV-2 factors in Drosophila. The examples delineated above illustrate possible routes to explore the activities of these factors, although additional organs/tissues could be employed. A key aspect of these investigations lies in identifying which human proteins mediate viral activities and establishing the function of these host partners. The reagents described here should facilitate unraveling of the multiple impacts of SARS-CoV-2 factors on host proteins.

Limitations of the study

In this report we express viral factors in various Drosophila tissues, either alone or in combination with dedicated human partners, identified in a proteomics screen.⁹ Although a live organism offers many benefits as compared to cultured cells, we are aware that results obtained from an insect model require validation in other contexts, such as human cells or mammalian model organisms. Also, the synergistic phenotypes we show in this study may not necessarily reflect direct physical interactions: the interacting factor could be part of larger protein complexes or impinge on a common pathway. We also note that our study focused on a limited set of SARS-CoV-2 interacting proteins identified by Gordon et al.⁹ Other screens^7,10–12 have identified 1,693 additional interactors. Because of our past efforts to generate human cDNA transgenic resources, 551 additional lines are available, covering nearly one-third of all proteins identified in these screens (Table S5).

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by lead contact Hugo Bellen (hbellen@bcm.edu).

Materials availability

The list of the UAS-SARS-CoV-2 transgenic flies generated in this study can be found in Table S1. The available UAS-human cDNA stocks encoding SARS-CoV-2 protein interactors generated in this study are listed in Table S2. The available T2A-GAL4 or Kozak-GAL4 lines for fly ortholog candidates that encode for SARS-CoV-2 human interactors are listed in Table S3. All newly generated reagents have been deposited at BDSC or KDSC. Stock numbers are indicated in Tables S1–S3.

Data and code availability

The sequencing information and PCR verification of T2A-GAL4 and Kozak-GAL4 lines are available from the Drosophila GDP (Gene Disruption Project) Website (https://flypush.research.bcm.edu/pscreen/index.php). This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Tables S1–S5 are available at Mendeley Data: https://data.mendeley.com/datasets/4p3m97pkkx/1.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Fly stocks

All Drosophila melanogaster flies used in this study were generated in house or obtained from the Bloomington Drosophila Stock Center (BDSC), Vienna Drosophila Resource Center (VDRC) or Kyoto Drosophila Stock Center (KDRC) and reared on standard fly food. The names of each fly line that are not listed in Tables S1–S3 are listed below along with their genotypes and source information.

Fly line	Genotype	Source
Act-GAL4	y1 w*; P{Act5C-GAL4}17bFO1/TM6B, Tb1	BDSC #3954
Tub-GAL4	y1 w*; P{w[+mC]=tubP-GAL4} LL7/TM3, Sb1 Ser1	BDSC #5138
da-GAL4	w*; P{w[+mW.hs]=GAL4-da.G32}UH1, Sb1/TM6B, Tb1	BDSC #55851
MS1096-GAL4	w1118 P{w[+mW.hs]=GAL4}Bx[MS1096]	BDSC #8860
Hs-GAL4	w*; P{w[+mC]=GAL4-Hsp70.PB}89-2-1	BDSC #1799
UAS-Empty	y1 w*; PBac{UAS-empty}VK37/SM6a	Goodman et al., 2021
Rh1-GAL4	P{ry[+t7.2]=rh1-GAL4}3, ry[506]	BDSC #8691
Ate1 CRIMIC-TG 4	y1 w*; TI{CRIMIC.TG4.0}Ate1CR00893-TG4.0/SM6a	BDSC #81152
Ate1-Df	w 1118 ; Df(2R)Exel7162/CyO	BDSC #7896
Ate1-RNAi	y1 sc* v1 sev21; P{TRiP.GLC01821}attP2	BDSC #53867
Ate-RNAi	P{KK107889}VIE-260B	VDRC #104360

METHOD DETAILS

SARS-CoV-2 UAS-cDNA generation, related to Figure 2B

The UAS-cDNA lines expressing SARS-CoV-2 viral protein were generated using standard methodologies.⁴⁴ Briefly, the open and close versions of Gateway-compatible entry vectors, which were deposited by Fritz Roth Lab,²⁸ were obtained from Addgene. The cDNAs were transferred to the pGW-HA.attB destination vector using LR clonase II (ThermoFisher #11791020). All clones were confirmed with Sanger sequencings. Since NSP11 is too short to be compatible with the Gateway cloning system, the NSP11 cDNA (with or without a stop codon) was inserted into the pGW-HA.attB empty plasmid by Q5 site-directed mutagenesis (New England Biolabs) with primers listed below. The whole plasmid sequencing based on the Sanger method confirmed to be no second site mutations. These constructs were inserted into the VK37 docking site using φC31-mediated transgenesis³³ and balanced using SM6a. Final genotype:

y¹ w*; PBac{y[+mDint2] w[+mC]=UAS-SARS-CoV-2 cDNA}VK00037/(SM6a), or

y¹ w*; PBac{y[+mDint2] w[+mC]=UAS-SARS-CoV-2 cDNA-HA}VK00037/(SM6a).

NSP11 constructs	Primers for mutagenesis
UAS-NSP11	Forward: ttttaaacgggtttgcggtgtaaAACCCAGCTTTCTTGTACAA
	Reverse: acgattgtgcatcagctgaCATGGTGCCTGCTTTTTTGTACA
UAS-NSP11-HA	Forward: ttttaaacgggtttgcggtgAACCCAGCTTTCTTGTACAA
	Reverse: acgattgtgcatcagctgaCATGGTGCCTGCTTTTTTGTACA

UAS-Kozak-NSP1-10 constructs were generated from the Addgene plasmid collection (#141255–141264). For each cDNA, a PCR fragment was generated with primers adding the CACC Kozak sequence right before the ATG start codon, and complementary overhangs for insertion into thepUASTattB vector (Addgene) downstream of the UAS(5X)/hs promoter sequences. The NEBuilder HiFi DNA Assembly kit (#E5520) was used to assemble PCR fragments, and the final sequence was verified.

MUTAGENESIS OF ORF3a, NSP3 AND NSP5

For ORF3a^Q57H mutagenesis, the ORF3a reference plasmid was obtained from Addgene (Addgene #149319). The variant was introduced using the Q5 site directed mutagenesis kit according to the manufacturer’s protocol (NEB# E0554). Briefly, gateway compatible constructs were transformed using high efficiency E. coli competent cells (NEB# C2987H) and successful mutagenesis was confirmed by Sanger sequencing the region of interest. The pDONR 207 entry vector was then transferred to the pGW-attB-3xHA.attB destination vector using LR clonase II (ThermoFisher #11791020), constructs were again transformed into high efficiency E. Coli competent cells, and the full sequences were confirmed by Sanger sequencing prior to injection. These constructs were inserted into the VK37 docking site using φC31-mediated transgenesis³³ and balanced using SM6a.

ORF3a constructs	Primers for mutagenesis
UAS-ORF3a p.Q57H-HA	Forward: CCGTGTTCCAtTCCGCCTCTAAGATCATC
	Reverse: CCAGCAGGGCCACACCCA

For mutant NSP3 (C857A), the UAS-Kozak-NSP3 plasmid was digested with Sbf1 and Stu1. The 1.6Kb fallout fragment was replaced with homologous DNA containing the 3nts C857A mutation and two complementary overhangs flanking the Sbf1 and Stu1 sites (30 nts on each side, gBlocks synthesized by IDT). Replacement was realized with NEBuilder kit, and sequence was verified. Similarly, for mutant NSP5 (C145A), the UAS-Kozak-NSP5 plasmid was digested with EcoRI and EcoNI, and the 0.6Kb fallout was replaced with a 0.6Kb DNA containing the C145A mutation and complementary overhangs.The constructs were inserted on the third chromosome docking site (attP2-3L68A4). Positive transformants were identified through the mini-white marker in the eyes and the inserted cassette was sequenced after PCR amplification of genomic sequences.

UAS-human cDNA generation

Gateway compatible full length human cDNA clones in pDONR221 or pDONR223 were obtained from a public Mammalian Gene Collection¹²⁴ or a commercial source (UltimateORF Clones, Thermo Fisher). The cDNAs were transferred to the pGW-HA.attB or pUASg-HA.attB destination vector using LR clonase II (ThermoFisher #11791020). All clones were validated by end sequencing. These constructs were inserted into the VK37, VK33³³ or ZH-86Fb¹²⁵ docking sites using φC31-mediated transgenesis³³ and balanced using SM6a or TM3 Sb Ser. In principle, we injected the transgenes on to VK37 (second chromosome docking site) if the best fly ortholog candidate of the human gene was located on the X, third or fourth chromosome. If the best fly ortholog candidate was located on the second chromosome, we injected them into VK33 or ZH-86Fb (third chromosome docking sites, latter being used during the early phase of the project). Two transgenes were inserted on attP40¹²⁶ or VK02³³ (both second chromosome docking sites), respectively, since they were generated during the pilot phase of this project. Second chromosome transgenes were balanced using CyO or SM6a, and third chromosome transgenes were balanced using TM3, Sb, Ser or TM6b, Hu, Tb, respectively.

T2A-/Kozak-GAL4 generation

T2A-GAL4 and Kozak-GAL4 CRISPR mediated homologous recombination alleles are generated using partially synthesized homology donor intermediates.^26,32 Briefly, 200 bps of homology arms are synthesized in pUC57_Kan_gw_OK vector by Genewiz/Azenta to generate homology donor intermediates. T2A-GAL4 or Kozak-GAL4 cassette is subcloned from pM37/pM37_kG4 vector respectively to the homology donor intermediate vector in BbsI or BsaI site Int200-T2AGAL4 and int200-KozakGAL4 constructs were injected at 250 ng/μL along with 100 ng/μL gene-specific gRNA(s) cloned in pCFD3 or pCFD5, respectively together with sgRNA1 (targeting GTAGTACGATCATAACAACGCGG) to linearize the homology donor construct.^127,128 To generate homology donor intermediates of int200_gRNA_T2A-GAL4 or int200_gRNA_Kozak-GAL4 lines, 200 bps homology arms and gene specific sgRNAs are commercially synthesized by Genewiz/Azenta. T2A-GAL4 or Kozak–GAL4 cassette is subcloned from pM37/pM37_kG4 vector respectively to the homology donor intermediate vector in BbsI or BsaI site int200-gRNA_T2AGAL4 and int200_gRNA_KozakGAL4 constructs are injected at 250 ng/ul final concentration. Injections of 400–600 embryos from y1w*; iso18; attP2(y+){nos-Cas9(v+)} for genes on the second or fourth chromosome and y1w* iso6 attP2(y+){nos-Cas9(v+)} for genes on the X chromosome and y1w*; attP40(y+){nos-Cas9(v+)}; iso5¹²⁹ for genes on the third chromosome were injected. Whole genome sequencing BAM files of isogenized lines can be found at https://zenodo.org/record/1341241. Resulting G0 males and females were crossed individually to y¹ w* flies as single fly crosses for 3XP3-EGFP detection. Positive lines were balanced, and stocks were established. Up to five independent lines were generated per construct per gene. The list of generated alleles can be found in Table S3. The sequences of homology arms and sgRNA(s) as well as the results of PCR validation and imaging on third instar larval brain are available at https://flypush.research.bcm.edu/pscreen/crimic/crimic.php. The stocks are deposited in the Bloomington Drosophila Stock Center (BDSC) on a regular basis. The stocks are available from the Bellen lab until they are deposited and established in the BDSC. Lines generated through RMCE of MiMIC alleles are generated through genetic crossing strategy.^27,37,130

Lethality assessment, related to Figures 3, S2, and S3

Crosses for assessment of lethality and morphological phenotypes were performed using the GAL4 drivers indicated in the text. Crosses were established using 5–10 virgin females crossed to a similar number of males. Parents were transferred to a new vial every 3–7 for multiple collections of progenies. Flies were collected after eclosion and genotypes were assessed based on the absence of balancers. When majority of the relevant progenies eclose and do not have obvious defect in movement and morphology, the combination is defined as viable. When some die before eclosion while there are also surviving adults, the lines are defined as semi-lethal. When there are no eclosions, the lines are defined as lethal. For lethal crosses, staging of lethality was performed based upon the oldest stage reached by the majority of flies. If no larvae were observed, the cross was recorded as embryonic lethal.

Wing phenotype assessment, related to Figures 3, 4, and 5

Wings were stored in isopropanol until mounting in Canada balsam (Sigma Aldrich Cat#101691). Pictures were taken with a 34x zoom factor using a Zeiss AXIO Zoom.V16 microscope, and further adjusted on Photoshop (21.2.4). Surface area was measured in FIJI (ImageJ 1.53t) using the “Ellipse Selection” followed by the command “Measure”. Data was exported to GraphPad Prism (version 9.4.1), for statistical analysis and graphs. Comparison between two conditions were tested with the Unpaired t test and between three or more conditions with Ordinary one-way ANOVA.

ERG assessment, related to Figure S1

ERG recordings were performed using standard procedures.⁴⁵ Briefly, after eclosion, flies were maintained at 25°C with 12-h light (~3,500 Lux)/dark cycle before the recording. The flies were immobilized on a glass slide with glue. A recording electrode filled with 150 mM NaCl was placed on the surface of the eye while a reference electrode was placed in the thorax. Repeated light/ dark stimulations were given during the recording, and the response of photoreceptors were digitized using LabChart 8 software. We use the UAS-Empty as a control line, generated by injecting the pGW-HA.attB plasmid without an insertion into VK37.¹³¹

Western blotting, related to Figures 7, S2, and S3

For WB in Figure 7C, adult flies (1–2 days-old) were heat shocked at 37°C for 90 or 120 min. After a rest at RT of varying durations, groups of 10 flies were homogenized in 100 μL RIPA buffer with protease inhibitors. Cuticles and insoluble matters were eliminated by centrifugation at 4°C. For WB in Figure S3, L3 larvae were homogenized into RIPA Lysis and Extraction Buffer (Thermo Scientific) at 25 μL per larva. Proper amount of 4x Laemmli Sample Buffer (Bio-Rad) was added. 10 μl of lysate samples were loaded on SDS-PAGE, 4%–12% Mini-PROTEAN TGX Precast Gel (Bio-Rad) gels. After separation, proteins were transferred onto PVDF membranes (Millipore). Membranes were incubated with the corresponding primary and secondary antibodies and signal was obtained using Supersignal chemiluminescent substrate (Thermo Fisher # 34580). All antibodies used are documented in the key resources table. For protein quantifications in Figures S2 and S3, protein levels were analyzed by ImageJ.¹³² Briefly, the optical densities of bands were quantified and the density ratio of each immunoreactive band to the corresponding Actin band was calculated. The results were presented as relative fold changes.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Actin	Millipore	Cat # MAB1501; RRID:AB_2223041
Anti-NSP8	GeneTex	Cat #GTX632696; RRID:AB_2888329
Anti-NSP10	ABclonal	Cat # A20325
Anti pMAPK (197G2)	Cell Signaling	Cat # 4377; RRID: AB_331775
Anti pMad/pSmad	Cell Signaling	Cat #9516; RRID: AB_491015
Anti R-actin	Millipore	Cat #ABT264
Anti-β-actin	GeneTex	Cat #629630; RRID: AB_2728646
Alexa 647-Phalloidin	Thermo Fisher	Cat #222287
Mouse Anti HA	Millipore Sigma	Cat #11583816001; RRID:AB_514505
Goat anti-rabbit HRP	Jackson ImmunoResearch	Cat # 111-035-144; RRID: AB_2307391
Goat anti-mouse HRP	Jackson ImmunoResearch	Cat # 115-035-146; RRID: AB_2307392
Chemicals, peptides, and recombinant proteins
Merbromin	Millipore Sigma	Cat#129-16-8
Suramin	Millipore Sigma	Cat#S2671
Critical commercial assays
NEBuilder HiFi DNA Assembly	New England Biolabs	Cat#E5520
Experimental models: Organisms/strains
D. melanogaster: Act-GAL4	BDSC	BDSC #3954
D. m: Tub-GAL4	BDSC	BDSC #5138
D. m: da-GAL4	BDSC	BDSC #55851
D. m: UAS-Empty	Goodman et al., 2021	N/A
D. m: UAS-ATE1	This work	BDSC #93215
D. m: Rh1-GAL4	BDSC	BDSC #8691
D. m: Ate1CRMIC-TG4	This work	BDSC #81152
D. m: Ate1-Df	BDSC	BDSC #7896
D. m: Ate1-RNAi	BDSC	BDSC #53867
D. m: Ate1-RNAi	VDRC	VDRC #104360
D. m: MS1096-GAL4	BDSC	BDSC #8860
D. m: Heat Shock-GAL4	BDSC	BDSC #1799
D. m:vas-phiC31; VK33	BDSC and Kyoto Stock Center	BDSC #24871 KYOTO #130448
D. m:vas-phiC31; VK37	BDSC and Kyoto Stock Center	BDSC #24872 KYOTO #130449
Oligonucleotides
UAS-NSP11 Forward: TTTTAAACGGGTTTGCGGTGTAAAACCCAGCTTTCTTGTACAA	This paper, Sigma	N/A
UAS-NSP11 Reverse: ACGATTGTGCATCAGCTGACATGGTGCCTGCIIIIIIGTACA	This paper, Sigma	N/A
UAS-NSP11-HA Forward: TTTTAAACGGGTTTGCGGTGAACCCAGCTTTCTTGTACAA	This paper, Sigma	N/A
UAS-NSP11-HA Reverse: ACGATTGTGCATCAGCTGACATGGTGCCTGCI IIIIIGTACA	This paper, Sigma	N/A
UAS-ORF3ap.Q57H-HA Forward: CCGTGTTCCATTCCGCCTCTAAGATCATC	This paper, Sigma	N/A
UAS-ORF3a p.Q57H-HA Reverse: CCAGCAGGGCCACACCCA	This paper, Sigma	N/A
Recombinant DNA
pGW-HA.attB	Konrad Basler,(UZH)	GB: KC896838
pUASg-HA.attB	Konrad Basler,(UZH)	GB: KC896837
pUASg.attB	Konrad Basler,(UZH)	GB: EF362409.1
Software and algorithms
ImageJ 1.53c	https://imagej.nih.gov/	N/A
Graphpad Prism 8	https://www.graphpad.com/	N/A
Microsoft Excel	Microsoft	N/A
Microsoft Powerpoint	Microsoft	N/A
Adobe Acrobat Pro	Adobe	N/A
Adobe Illustrator 2020	Adobe	N/A

Immunostaining, related to Figures 4, 7, S4, and S7

For immunostainings, wing imaginal discs and salivary glands were dissected from L3 wandering larvae, fixed for 20 min in 4% formaldehyde in PBS, washed and blocked in 1% BSA PBS. Primary antibody was added at 1:100 dilution overnight at 4°C. Secondary incubations was at 1:400 for 2 h at RT and nuclear DAPI staining (Thermo Fisher #D1306) for 30 min at RT. The antibodies used were anti-pMad (Cell signaling, #9516), anti-pMAPK (Cell signaling, #4377), anti R-actin (Millipore #ABT264). For F-actin stain, 647-Phalloidin (Thermo Fisher #A22287, 1:200) was added with secondary antibodies (Thermo Fisher Alexa 488 anti Rabbit IgG, #A 21–200). Stained tissues were mounted in Slowfade Gold Antifade Mountant (Life Technologies #S36936), with a double-sided tape between slide and the coverslips for salivary glands to avoid crushing. Imaging was performed on a Leica TCS SP8 confocal microscope. For brain and wing disc imaging the larvae were dissected in PBS and fixed 20 minues in 4% formaldehyde in PBS. Samples are washed and blocked with PBS 0.3% Triton X-100 5% Normal Donkey Serum (NDS). Primary antibody (anti-mCherry, Genetex #GTX59788) was incubated at 1:200 dilution in PBS 0.3% Triton X-100 5% NDS. Multistack images were taken and exported for processing on FIJI (ImageJ 1.53t). Final panels were created using Adobe Photoshop software.

Drug treatments, related to Figure 7

Suramin and Merbromin were resuspended in H₂O to stocks of 5mM and 100mM, respectively. They were then diluted in 100% grape juice (Welch) to the appropriate concentrations. For each condition, 2mL of grape juice+drug was dispensed onto 1 gm of ground potato flakes placed in vials on top of regular brown fly food.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed using GraphPad software (GraphPad Prism v9.0; GraphPad Software, USA). Statistical analysis was performed using Student’s t-test. Data are presented using scattered plots with each dot representing measurement of an independent biological sample. Number n of biological replicates is indicated on each graph. Bars indicate mean with standard deviation. p values are indicated using standard symbolism: ns for p > 0.05; * for p < 0.05; ** for p < 0.01; *** for p < 0.001; and **** for p < 0.0001.

Highlights.

• A Drosophila resource enables functional analysis of all SARS-Cov2 factors

• This collection allows expression of viral factors and human binding partners

• It also includes GAL4 insertions mutating fly orthologs of host binding partners

• NSP8 interacts with other viral factors and with the ATE1 arginyltransferase

INCLUSION AND DIVERSITY

We support inclusive, diverse, and equitable conduct of research.