key: cord-0302058-pps3qmlg
authors: Springstein, Benjamin L.; Deighan, Padraig; Grabe, Grzegorz; Hochschild, Ann
title: A bacteria-based assay to study SARS-CoV-2 protein-protein interactions
date: 2021-10-08
journal: bioRxiv
DOI: 10.1101/2021.10.07.463611
sha: 6e9549ff7f4a6f2aff3621d76bfcc7cea935a389
doc_id: 302058
cord_uid: pps3qmlg

Methods for detecting and dissecting the interactions of virally encoded proteins are essential for probing basic viral biology and providing a foundation for therapeutic advances. The dearth of targeted therapeutics for the treatment of COVID-19, an ongoing global health crisis, underscores the importance of gaining a deeper understanding of the interactions of SARS-CoV-2-encoded proteins. Here we describe the use of a convenient bacteria-based two-hybrid (B2H) system to analyze the SARS-CoV-2 proteome. We identify sixteen distinct intraviral protein-protein interactions (PPIs), involving sixteen proteins. We find that many of the identified proteins interact with more than one partner. We further show how our system facilitates the genetic dissection of these interactions, enabling the identification of selectively disruptive mutations. We also describe a modified B2H system that permits the detection of disulfide bond-dependent PPIs in the normally reducing Escherichia coli cytoplasm and we use this system to detect the interaction of the SARS-CoV-2 spike protein receptor-binding domain (RBD) with its cognate cell surface receptor ACE2. We then examine how the RBD-ACE2 interaction is perturbed by several RBD amino acid substitutions found in currently circulating SARS-CoV-2 variants. Our findings illustrate the utility of a genetically tractable bacterial system for probing the interactions of viral proteins and investigating the effects of emerging mutations. In principle, the system could also facilitate the identification of potential therapeutics that disrupt specific interactions of virally encoded proteins. More generally, our findings establish the feasibility of using a B2H system to detect and dissect disulfide bond-dependent interactions of eukaryotic proteins. Importance Understanding how virally encoded proteins interact with one another is essential in elucidating basic viral biology, providing a foundation for therapeutic discovery. Here we describe the use of a versatile bacteria-based system to investigate the interactions of the protein set encoded by SARS-CoV-2, the virus responsible for the current pandemic. We identify sixteen distinct intraviral protein-protein interactions, involving sixteen proteins, many of which interact with more than one partner. Our system facilitates the genetic dissection of these interactions, enabling the identification of selectively disruptive mutations. We also describe a modified version of our bacteria-based system that permits detection of the interaction between the SARS-CoV-2 spike protein (specifically its receptor binding domain) and its cognate human cell surface receptor ACE2 and we investigate the effects of spike mutations found in currently circulating SARS-CoV-2 variants. Our findings illustrate the general utility of our system for probing the interactions of virally encoded proteins.

mutations. We also describe a modified B2H system that permits the detection of disulfide bond-23 dependent PPIs in the normally reducing Escherichia coli cytoplasm and we use this system to 24 detect the interaction of the SARS-CoV-2 spike protein receptor-binding domain (RBD) with its 25 cognate cell surface receptor ACE2. We then examine how the RBD-ACE2 interaction is 26 perturbed by several RBD amino acid substitutions found in currently circulating SARS-CoV-2 27 variants. Our findings illustrate the utility of a genetically tractable bacterial system for probing 28 the interactions of viral proteins and investigating the effects of emerging mutations. In principle, 29 the system could also facilitate the identification of potential therapeutics that disrupt specific 30 interactions of virally encoded proteins. More generally, our findings establish the feasibility of 31 using a B2H system to detect and dissect disulfide bond-dependent interactions of eukaryotic 32 proteins.

3 Importance 34 Understanding how virally encoded proteins interact with one another is essential in elucidating 35 basic viral biology, providing a foundation for therapeutic discovery. Here we describe the use of 36 a versatile bacteria-based system to investigate the interactions of the protein set encoded by 37 SARS-CoV-2, the virus responsible for the current pandemic. We identify sixteen distinct 38 intraviral protein-protein interactions, involving sixteen proteins, many of which interact with 39 more than one partner. Our system facilitates the genetic dissection of these interactions, 40 enabling the identification of selectively disruptive mutations. We also describe a modified 41 version of our bacteria-based system that permits detection of the interaction between the 6 RBD, motivating efforts to gain a systematic understanding of the effects of RBD amino acid 96 substitutions on ACE2 binding [23] .

Here we employ a bacterial two-hybrid (B2H) system [24, 25] to study the PPIs of SARS-

CoV-2 in a heterologous non-eukaryotic system. Using this system, we describe a bacteria-99 based intraviral interactome. We further demonstrate the utility of the bacterial system for 100 genetically dissecting the SARS-CoV-2 PPIs by identifying mutations that selectively affect one 101 or another interaction. In addition, we describe a modified B2H system that allows us to detect 102 disulfide bond-dependent PPIs in the otherwise reducing Escherichia coli cytoplasm. We use 103 this system to detect the spike RBD-ACE2 interaction and to investigate the effects of mutations 104 found in VOCs. Our findings set the stage for further investigations of viral PPIs in a convenient 105 and genetically tractable bacterial system, as well as establishing the feasibility of using our 

Bacterial two-hybrid system to detect interactions of SARS-CoV-2 proteome 111 Previous studies have used yeast two-hybrid (Y2H) systems, a mammalian two-hybrid system 112 and co-immunoprecipitation experiments (co-IPs) to investigate the SARS-CoV-1 and SARS-

CoV-2 protein interactomes, identifying overlapping but also distinct interactions depending on 114 the employed system [6, 7, 26, 27] . Compared with bacteria, yeast have a relatively slow growth 115 rate and are more difficult to culture and transform for labs that do not routinely work with yeast.

To provide a more accessible alternative to Y2H systems as well as the less commonly used 117 mammalian two-hybrid system, we here describe the successful use of a B2H system 118 developed in our lab (Fig. 1A) [24, 25] to test for viral PPIs. We fused all NCBI-predicted E. coli 7 codon-optimized SARS-CoV-2 open reading frames (ORFs; listed in Fig. 1B , see also NCBI 120 accession #: NC_045512.2) to the DNA binding protein CI of bacteriophage  (CI) and to the 121 N-terminal domain of the  subunit (NTD) of RNA polymerase (RNAP). We then tested each 122 SARS-CoV-2 ORF for interaction with the other SARS-CoV-2 ORFs and itself. Interaction 123 between two given ORFs (X and Y), fused to NTD and CI, respectively, stabilizes the binding 124 of RNAP to the test promoter such that the magnitude of the lacZ reporter gene expression 125 correlates with the strength of the PPI (Fig. 1A ).

Identification of the SARS-CoV-2 interactome using a B2H system 127 Using our B2H system, we initially tested each SARS-CoV-2 ORF against each other SARS-

CoV-2 ORF and itself in biological duplicate. Protein pairs with at least a 2-fold activation of lacZ 129 over background in one of the replicates were selected for further analysis. The list of interacting 130 proteins was further refined by performing repeat experiments with three biological replicates for 131 each initially identified potential PPI pair. This resulted in a final list of sixteen interacting SARS-

CoV-2 protein pairs, including four self-interactions (Fig. 2) . Some of these interactions were 133 identified only with a specific fusion partner combination (i.e., protein X fused to NTDand 134 protein Y fused to CI, or the other way around), while others were fusion partner-independent 135 (i.e., interaction between proteins X and Y regardless of their fusion to NTD or CI). Self-136 interacting proteins (Nsp7, Nsp9, ORF6 and ORF10) were by definition fusion partner-137 insensitive; however, four other pairs of proteins (Nsp7+Nsp8, Nsp10+Nsp14, Nsp10+Nsp16,

Nsp3+N, and Nsp8+ORF6) also interacted detectably regardless of the fusion partner 139 ( Supplementary Fig. 1 ).

Among the identified interacting pairs, several particularly strong PPIs were observed,

including the Nsp7 self-interaction, Nsp7+Nsp8, Nsp10+Nsp16, N+Nsp3 and Nsp9+Nsp11 (Fig.   142 3). In fact, the Nsp10+Nsp16 pair interacted significantly more strongly than our positive control, 8 representing one of the strongest interactions we have ever measured with our B2H assay. For 144 our B2H assays, we routinely consider an interaction to be reliable when we detect at least a 145 two-fold increase in lacZ reporter gene expression (measured as -galactosidase activity) over 146 the background (obtained with the negative controls). Applying this cut-off to our experimental 147 data, we identified several medium-to-weak interactions (2-to 5-fold increase over the negative 148 controls; Supplementary Fig. 2 ). The interactions of Nsp8+ORF7b and ORF10+ORF10 closely 149 missed the 2-fold cutoff but were nonetheless included in the list because a previous SARS-

CoV-2 interactome study also identified those interactions (based on co-IP data) [6] .

Comparison of our SARS-CoV-2 B2H data with the previously reported SARS-CoV-2 152 Y2H and co-IP data [6] revealed four PPIs that were shared among the three assay systems,

providing strong support for their biological relevance ( Supplementary Fig. 3 ). These included 154 Nsp7+Nsp8, Nsp8+ORF10, Nsp10+Nsp14 and ORF6+ORF6. Others were identified either in 155 only one of the assay systems (i.e., B2H, Y2H or co-IP) or in two assay systems (B2H and Y2H,

B2H and co-IP, or Y2H and co-IP) ( Supplementary Fig. 3 

Nonetheless, most of the interaction partners we identified for Nsp8 in SARS-CoV-2 are 9 different than those identified previously for 26, 27] (Supplementary Fig. 4 ).

Overall, only six PPIs were identified in our SARS-CoV-2 B2H analysis and at least one of three 170 independent SARS-CoV-1 Y2H studies, including two involving Nsp8 ( Supplementary Fig. 4 ).

Notably, there are considerable differences between the results of the three previous Y2H 172 studies [7, 26, 27] and only three PPIs (Nsp8+Nsp7, Nsp10+Nsp14, and Nsp10+Nsp16) were 173 independently identified in two SARS-CoV-1 two-hybrid assays and our SARS-CoV-2 B2H 174 assay ( Supplementary Fig. 4 ). This could reflect significant differences between the PPI 175 networks in SARS-CoV-1 and SARS-CoV-2 and/or differences in the assays themselves 176 (procedures and background organism).

Targeted mutational screens identify interaction partner-specific sites of protein-protein 178 interaction in CoV-2 proteins with more than one interaction partner 179 As a genetic assay, the B2H system facilitates the dissection of specific PPIs through both 180 targeted and random mutagenesis. Having established the utility of the B2H assay in testing for 181 viral PPIs, we next sought to use this assay to dissect the interactions of selected viral proteins 182 through targeted mutational analysis. Specifically, we chose proteins that interacted with more 183 than one partner and sought to disrupt the interaction of such a protein with one of its partners 184 while preserving its interaction with another. We initially selected Nsp10 with two known 185 interaction partners, Nsp14 and Nsp16, and attempted to disrupt only its interaction with Nsp14.

To identify suitable targets for mutagenesis, we analyzed the crystal structures of Nsp10-Nsp14 

4B). We note that the close approach of amino acid side chains at a protein-protein interface as 196 revealed by X-ray crystallography does not necessarily indicate that they participate in a 197 functionally important interaction. However, the loss of a detectable interaction between each of 198 the three Nsp10 mutants and Nsp14 in our B2H assay suggests that at least a subset of the 199 selected residues make stabilizing contacts. Furthermore, although the Nsp10-Nsp16 200 interaction serves as a control, we also confirmed that the introduced amino acid substitutions 201 were not generally destabilizing ( Supplementary Fig. 5 ).

We then focused on Nsp16 with two interaction partners, Nsp10 and Nsp15, targeting 203 the Nsp16-Nsp10 pair, which displayed a significantly higher B2H signal than that of the Nsp16-

Nsp15 pair. Here we also utilized the available crystal structure for the Nsp16-Nsp10 complex; 205 however, as there is no structure for the Nsp16-Nsp15 complex, the substitutions introduced 206 into Nsp16 were based solely on their predicted effects on its interaction with Nsp10.

Endeavoring to disrupt the Nsp16-Nsp10 interaction, we created two Nsp16 triple substitution 208 mutants, targeting hydrophobic (I40A/M41A/V44A) or hydrophilic (K76A/Q87A/D106A) contacts, 209 and a mutant with the six substitutions combined (Fig. 4C ). The data reveal drastic effects of 210 these substitutions on the binding of Nsp16 to Nsp10, resulting in near background or 211 background levels of reporter gene expression for each of the mutants (Fig. 4D ). The effects of 212 the same substitutions on the binding of Nsp16 to Nsp15 were modest and not statistically 213 significant. Notably, even though Nsp16 interacts much more weakly with Nsp15 than with 214 Nsp10, reporter gene expression was lower for each of the Nsp16 mutants in combination with 215 11 illustrate a proof-of-principle approach that can be used to obtain functionally informative 218 mutants within a PPI network.

The B2H system as a tool to study circulating spike variants and their binding to ACE2

To further assess whether our B2H system can facilitate the study of emerging mutational 221 changes in viral populations, we next asked whether we could use our system to study the 222 interaction between the SARS-CoV-2 spike protein and ACE2. For this, we obtained an E. coli 

Having adapted our B2H system for the study of disulfide bond-dependent PPIs, we 247 sought to test different spike (RBD) circulating variants for their abilities to bind ACE2. The RBD 248 amino acid substitutions included in our study are found in several SARS-CoV-2 variants that 249 were previously designated VOCs by the Centers for Disease Control and Prevention (USA; 

[32]), which was not identified by either Y2H or co-IP analyses [6] . Of the six interactions we 286 detected that were not previously described in the context of SARS-CoV-2, three were 287 previously detected by Y2H analyses in the context of SARS CoV-1 ( Supplementary Fig. 4 ). Of 288 the remaining three interactions, not previously described, two (Nsp9+Nsp11 and Nsp3+N) were 289 particularly strong as assessed in our B2H assay (Fig. 3 ).

14

Although the different assays that have been used to characterize the SARS-CoV-2 Fig. 3 ). We note, however, that a comprehensive phosphoproteomics analysis 305 of SARS-CoV-2-infected cells [57] suggests that other than interactions involving the N protein,

which was found to be phosphorylated at multiple sites, most of the viral PPIs that were 307 detected by Y2H analysis but not in our B2H system involve proteins that were not detectably 308 phosphorylated.

A benefit of two-hybrid approaches for studying PPIs is that detected interactions can be 311 readily dissected genetically, something that is particularly straightforward to do with our B2H 312 system. As a proof-of-principle, we used a structure-based approach to investigate the effects of 313 targeted mutations on specific SARS-CoV-2 PPIs, identifying substitutions that disrupt one 314 interaction but not another. In addition to facilitating the evaluation of specific circulating or targeted mutations, our B2H system can readily be adapted to screen for randomly generated 

Nsp2 open reading frame (ORF) was cut from pS85 by NotI+BamHI and inserted into 50 ng

The Global Phosphorylation Landscape of SARS-CoV-2 605 Infection

CbtA toxin of Escherichia coli inhibits cell division 607 and cell elongation via direct and independent interactions with FtsZ and MreB

A Bacterial Small-Molecule Three-Hybrid System

Woods RJ & Wells L 613 (2020) Virus-Receptor Interactions of Glycosylated SARS-CoV-2 Spike and Human ACE2

ACE2 glycans 616 preferentially interact with SARS-CoV-2 over SARS-CoV

Dual nature of human ACE2 glycosylation in binding to 618 SARS-CoV-2 spike

DeFrees S & Wakarchuk W 620 (2019) A Bacterial Expression Platform for Production of Therapeutic Proteins Containing 621

Human-like O-Linked Glycans

Reduced sensitivity of SARS-CoV-2 variant Delta to antibody neutralization

Structural 629 basis of receptor recognition by SARS-CoV-2

Enhanced binding of the N501Y-mutated SARS-CoV-2 631 spike protein to the human ACE2 receptor: insights from molecular dynamics simulations

The Fc-mediated effector functions of a potent SARS-CoV-2 638 neutralizing antibody, SC31, isolated from an early convalescent COVID-19 patient, are 639 essential for the optimal therapeutic efficacy of the antibody

SARS-CoV-2 B.1.1.7 and

.351 spike variants bind human ACE2 with increased affinity

The new SARS-CoV-2 strain shows a stronger binding 644 affinity to ACE2 due to N501Y mutant

silico investigation of the new

South African (501y.v2) SARS-CoV-2 variants with a focus at the ace2-647 spike rbd interface

Mutations in 649 the B.1.1.7 SARS-CoV-2 Spike Protein Reduce Receptor-Binding Affinity and Induce a 650 Flexible Link to the Fusion Peptide

SARS-CoV-2 entry related viral and host genetic variations: 652 Implications on covid-19 severity, immune escape, and infectivity

Comprehensive mapping of mutations in the SARS-CoV-2 receptor-binding domain that 655 affect recognition by polyclonal human plasma antibodies

Complete Mapping of 660

Mutations to the SARS-CoV-2 Spike Receptor-Binding Domain that Escape Antibody 661 Recognition

Enhanced Receptor Binding by Rapidly Spreading SARS-CoV-2 Variants

Key residues of the receptor binding motif in the spike protein of 667

SARS-CoV-2 that interact with ACE2 and neutralizing antibodies

Transmission, infectivity, and neutralization 676 of a spike L452R SARS-CoV-2 variant

SARS-CoV-2 spike L452R variant evades cellular immunity and increases 681 infectivity

Acquisition of the L452R mutation in the ACE2-685 binding interface of Spike protein triggers recent massive expansion of SARS-Cov-2 686 variants

The Impact of Mutations in

SARS-CoV-2 Spike on Viral Infectivity and Antigenicity

Hochschild 691 A, Heyduk T & Severinov K (2002) A role for interaction of the RNA polymerase flap 692 domain with the σ subunit in promoter recognition

The bacteriophage λ Q 694 antiterminator protein contacts the β-flap domain of RNA polymerase

One-step preparation of competent Escherichia 697 coli: transformation and storage of bacterial cells in the same solution

Structural Basis for Bacterial Transcription-Coupled DNA Repair

High-throughput β-galactosidase assay for 703 bacterial cell-based reporter systems

After washing with TBST, 1068 proteins were then detected using a ChemiDoc MP system

SARS-CoV-2 Nsp16-Nsp10

Amino acids involved in hydrogen bond formation or substantially contributing to 1073 hydrophobic contacts in each complex were subjected to alanine mutagenesis and tested in 1074 B2H assays

The PyMOL Molecular Graphics System

Presentation of bacterial two-hybrid data and statistical analysis using one-way or two-way 1078

ANOVA with Tukey's or Dunnett's multiple comparison test was done using GraphPad Prism (v. 1079 9

The table lists the spike domains that were produced in E. coli (as B2H fusion proteins) and 838 tested for interaction with ACE2, their precise corresponding loci on the SARS-CoV-2 genome, 839 and the amino acids encoded by each test domain. 

The authors declare no competing interests.

All data generated during and/or analyzed during the current study are either provided within the 875 manuscript or are available from the corresponding authors upon reasonable request.NotI+BamHI-digested pBR or pACCI using T4 ligase (NEB) according to standard protocols, 949 creating pS179 and pS180, respectively. pS254 was generated by SDM using primers SARS_253+SARS_254 and plasmid pS215 as a 988 template. pS256 was generated by SDM using primers SARS_256+SARS_257 and plasmid 989 pS215 as a template. pS257 was generated by SDM using primers SARS_256+SARS_257 and 990 plasmid pS254 as a template. pS262 was generated by Gibson assembly using primers 991 SARS_268+SARS_269 and plasmid pS196 as a template. pS263 was generated by Gibson 992 assembly using primers SARS_270+SARS_271 and plasmid pS196 as a template. pS264 was 993 generated by Gibson assembly using primers SARS_272+SARS_273 and plasmid pS196 as a 994 38 template. pS267 was generated by Gibson assembly using primers SARS_280+SARS_281 and 995 plasmid pS65 as a template. pS271 was generated by Gibson assembly using primers 996 SARS_287+SARS_288 and plasmid pS65 as a template. pS272 was generated by Gibson 997 assembly using primers SARS_289+SARS_290 and plasmid pS65 as a template. pS273 was 998 generated by Gibson assembly using primers SARS_289+SARS_290 and plasmid pS271 as a 999 template. pS275 was generated by Gibson assembly using primers SARS_295+SARS_296 and 1000 plasmid pS65 as a template. pS276 was generated by Gibson assembly using primers 1001 SARS_291+SARS_292 and plasmid pS65 as a template. pS277 was generated by Gibson 1002 assembly using primers SARS_293+SARS_294 and plasmid pS65 as a template. pS278 was 1003 generated by Gibson assembly using primers SARS_291+SARS_292 and plasmid pS267 as a 1004 template. pS279 was generated by Gibson assembly using primers SARS_293+SARS_294 and 1005 plasmid pS267 as a template. pS280 was generated by Gibson assembly using primers 1006 SARS_293+SARS_294 and plasmid pS278 as a template.

-galactosidase assays 1008 -galactosidase assays to study the SARS-CoV-2 interactome were performed essentially as To verify the production of the respective fusion proteins, western blots of cell lysates from over-1045 night cultures were performed. For this, co-transformed cells were grown in the indicated IPTG 1046 concentration over-night in 550 µl total volume in 2 ml deep well plates at 30 or 37 °C, 800 rpm.

The next day, OD 600 values were recorded and 500 µl cells were pelleted by centrifugation (1 1048 min, 21,000 x g, room temperature (RT)) and either stored at -80 °C or directly processed. Cell 

680RD goat anti-mouse and IRDye® 800CW goat anti-rabbit (both 1:10,000 dil.; LI-COR