2021 May 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 In previous work, Hsieh and Higgins presented a novel structure of antibodies identified from malaria-exposed individuals, in which the extracellular immunoglobulin (Ig)-like domain of leukocyte-associated immunoglobulin-like receptor 1 (LAIR1) is presented on the third complementarity determining regions (CDR3) of the Ig heavy chain. Here we develop an Antibody Display technology based on this LAIR1-containing antibody, by grafting proteins of interest (POI) onto the heavy chain CDR3 while retaining the biological properties of the POI. As a proof of principle, we displayed the second extracellular domain of Tetraspanin12 (Tspan12EC2) and the receptor-binding domain (RBD) of SARS-CoV-2 spike protein on the heavy chain CDR3. Our data revealed that Antibody Display Tspan12EC2 bound to Norrie Disease Protein (Norrin) and Antibody Display SARS-CoV-2 RBD bound to angiotensinconverting enzyme 2 (ACE2) and neutralizing nanobodies. Collectively, Antibody Display technology offers the general strategy of designing novel antibodies by grafting POI onto the 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69

Antibodies are crucial reagents for biomedical research, diagnostics, and therapeutics. The conventional antibody, immunoglobulin G (IgG), has a highly conserved architecture that is composed of two heavy chains and two light chains (Owen et al., 2013). Each chain contains three hypervariable loops, known as complementarity-determining regions (CDRs), that form the antigen-binding site, which determines the binding specificity for antigens (Chothia et al., 1989) . Unexpectedly, a group of unusual antibodies were identified recently from malaria-endemic regions of Africa (Tan et al., 2016). These antibodies have a remarkable feature of the antigenbinding site in which an intact collagen-binding domain, adopting the immunoglobulin (Ig)-like fold, from leukocyte-associated immunoglobulin-like receptor 1 (LAIR1) was inserted into the CDR3 of the heavy chain (Figure 1A). The LAIR1 insert carrying multiple somatic mutations interacts with the repetitive interspersed families of polypeptides (RIFIN) that are expressed on the surface of Plasmodium falciparum-infected erythrocytes (Tan et al., 2016). A detailed structural analysis of this LAIR1-containing antibody (MGD21) has been reported (Hsieh and Higgins, 2017). Pieper et al., further found that malaria-exposed individuals had two additional types of LAIR1-containing antibodies in which the LAIR1 insert was located between the variable domain of the heavy chain (VH) and the constant domain of the heavy chain (CH1) or fused to the FC fragment (Pieper et al., 2017).

The Tetraspanin (Tspan) receptor family contains 33 members in humans and play central roles in diverse biological processes (cell-cell fusion, immune response, and vascularization) and diseases (pathogen infection and tumorigenesis) (Charrin et al., 2014; Hemler, 2014; Vences70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91

Catalan and Levy, 2018). Previous structural analyses revealed that Tspan receptors adopt a conserved architecture composed of four transmembrane helices (TM13TM4), intracellular Nand C-termini, a first and smaller extracellular domain (EC1; 13-30 residues) between TM1 and TM2, and a second and larger extracellular domain (EC2; 70-130 residues) between TM3 and TM4 (Kitadokoro et al., 2001; Oosterheert et al., 2020; Umeda et al., 2020; Yang et al., 2020b; Zimmerman et al., 2016) . Functional studies showed that EC2 of Tspan receptors is typically a critical domain for the interactions with partner proteins or pathogens on the cell surface (Green et al., 2011; Rajesh et al., 2012; Zhu et al., 2002). Tspan12 is important in central nervous system (CNS) blood vessel development and blood-brain/retina barrier formation (Junge et al., 2009; Lai et al., 2017; Wang et al., 2018; Zhang et al., 2018). Genetic studies revealed that (1) deficiency of Tspan12 results in familial exudative vitreoretinopathy (FEVR), a hereditary disorder charactered by abnormal development of the retinal vasculature, often leading to retinal detachment and vision loss (Junge et al., 2009; Nikopoulos et al., 2010; Poulter et al., 2010), and (2) Tspan12 is a co-activator of Norrie Disease Protein (also named Norrin) mediated ³-catenin signaling (Junge et al., 2009; Luhmann et al., 2005; Stenman et al., 2008; Xu et al., 2004). Norrin, a secreted cystine-knot growth factor, activates ³-catenin signaling by binding to the extracellular cystine-rich domain (CRD) of Frizzled4 (Fz4) and the N-terminal domains of lowdensity lipoprotein receptor-related protein 5/6 (Lrp5/6) (Chang et al., 2015; Ke et al., 2013; Smallwood et al., 2007; Xu et al., 2004) . Deficiencies of Fz4 and Lrp5 also result in FEVR (Robitaille et al., 2002; Toomes et al., 2004). Recently, Lai et al. suggested that the EC2 of Tspan12 (Tspan12EC2) is a critical region for Norrin binding (Lai et al., 2017). However, whether Tspan12EC2 interacts with Norrin directly remains obscure because of the technical challenge of 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 obtaining recombinant Tspan12EC2 without its conserved transmembrane helices (Junge et al., 2009; Lai et al., 2017).

The worldwide spread of coronavirus disease-2019 (COVID-19) causing by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has become a global pandemic (Zhou et al., 2020a; Zhu et al., 2020) . The surface of the SARS-CoV-2 particle is decorated with the heavily glycosylated trimeric spike (S) protein protruding from the virus membrane envelope (Fung and Liu, 2019; Shang et al., 2020; Walls et al., 2020; Wang et al., 2020b; Watanabe et al., 2020; Wrapp et al., 2020b; Yao et al., 2020) . Specifically, SARS-CoV-2 S protein contains the receptor-binding domain (RBD) which mediates the interaction with a virus entry receptor, angiotensin-converting enzyme 2 (ACE2). Therefore, SARS-CoV-2 S protein is an attractive target for therapeutic and vaccine development (Hoffmann et al., 2020; Walls et al., 2020; Zhou et al., 2020a; Zhou et al., 2020b) . SARS-CoV-2 S protein is immunogenic and the major target of neutralizing antibodies from COVID-19 convalescent individuals (Barnes et al., 2020; Baum et al., 2020; Brouwer et al., 2020; Chi et al., 2020; Ju et al., 2020; Seydoux et al., 2020; Walls et al., 2020; Wang et al., 2020a; Wu et al., 2020b) . Furthermore, Piccoli et al., showed that more than 90% of these neutralizing antibodies from COVID-19 convalescent plasma target the RBD of SARS-CoV-2 S protein for protective humoral responses (Piccoli et al., 2020) . However, recent studies showed that mutations in the RBD present in the current circulating SARS-CoV-2 variants decrease the potency of neutralizing antibodies (Greaney et al., 2021; Starr et al., 2021; Weisblum et al., 2020) . Therefore, understanding how the RBD of SARS-CoV-2 S protein responds to ACE2 and neutralizing antibodies is critical for ongoing vaccine development, 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 immunotherapy, and the assessment of the RBD mutations occurring in circulating SARS-CoV-2 stains.

In this report, we have developed an antibody-based display approach, termed Antibody Display, by grafting proteins of interest (POI) onto the heavy chain CDR3 to generate biological active and stably folded chimera antibodies. As a proof of principle, we used the structure-guided protein design approach based on the structure of LAIR1-containing antibody (Hsieh and Higgins, 2017) to display the EC2 of Tspan12 and the RBD of SARS-CoV-2 S protein as insertions at the tip of the heavy chain CDR3. We showed that Antibody Display Tspan12EC2 (Tspan12EC2-AD) bound to Norrin and Antibody Display SARS-CoV-2 RBD (RBD-AD) bound to ACE2 and two neutralizing nanobodies (VHH-72 and H11-D4). We also designed a humanized and engineering variable domain of the heavy chain alone to display the RBD of SARS-CoV-2 and confirmed its binding properties. 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159

The design of the Antibody Display was informed by structural studies of LAIR1-containing antibody (Figure 1A) (Hsieh and Higgins, 2017). We hypothesized that POI could graft onto the heavy chain CDR3 via two short peptide linkers to generate chimeric antibodies while retaining the biological features of the POI. To test this concept, we focused initially on Tspan12EC2, because (1) Tspan12EC2 protrudes from two transmembrane helices of Tspan12 suggesting that the N- and C-termini of Tspan12EC2 could be linked to b-strands G and F of the heavy chain in a manner that preserves their native spacing (Figure 1B) and (2) the biomedical importance of Tspan12 in retinal vascular diseases.

We began our design by searching for homologous structures and conducting a multiple sequence alignment of Tspan25 (CD53) (Yang et al., 2020b) , Tspan28 (CD81) (Kitadokoro et al., 2001; Zimmerman et al., 2016), and Tspan29 (CD9) (Oosterheert et al., 2020; Umeda et al., 2020) with Tspan12, as shown in Figure 13Figure supplement 1A. The obtained information led us to generate constructs having different boundaries of Tspan12EC2 for expression trials in HEK293T cells (Figure 13Figure supplement 1B). The constructs contained an N-terminal signal peptide and C-terminal monomeric Venus (mVenus) and 12xHis tags. To assess the level of secreted mVenus fusion protein expression, we used a time- and cost-efficient method, immobilized metal affinity chromatography (IMAC) followed by in-gel fluorescent imaging (Chang et al., 2020) . Variations in the location of the Tspan12EC2 N- and C-termini produced large effects on protein yield (Figure 13Figure supplement 1C). Construct 4 (Tspan12 residues 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 116-220) exhibited the highest yield, somewhat less than 0.1 mg/L, consistent with previous studies showing the technical difficulties in producing recombinant Tspan12EC2 (Junge et al., 2009; Lai et al., 2017). As the secretory pathway in mammalian cells has stringent protein quality-control machinery to ensure that secreted proteins are folded correctly (Trombetta and Parodi, 2003), the expression trials imply that residues 116-220 corresponds to a core folding unit of Tspan12EC2.

We next generated a three-dimensional model of Tspan12EC2 and designed Antibody Display Tspan12EC2 (Tspan12EC2-AD) model in a Fab format, Fab (Tspan12EC2), by computationally grafting the Tspan12EC2 onto the heavy chain CDR3 based on the structure of LAIR1-containing antibody (Figure 1C). We designed constructs based on these principles: (1) placing the core folding unit of Tspan12EC2 at the tip of the heavy chain CDR3 to maximally expose its binding regions for ligand binding, (2) preventing steric clashes between the Tspan12EC2 and the CDR loops of the heavy chain and light chain, and (3) testing variations in the length and flexibility of the connecting linkers between the Tspan12EC2 and the heavy chain. Five designs of these chimeric heavy chain constructs (Figure 13Figure supplement 2A) were co-transfected with light chain constructs in HEK293T cells to yield secreted Fab proteins and were assessed for expression level using IMAC followed by in-gel fluorescent imaging (Figure 13Figure supplement 2B). Interestingly, only two of these designs (construct 1 and 5) could be expressed as secreted Fab proteins by passing the protein quality-control machinery in mammalian cells (Figure 13Figure supplement 2B). Specifically, Tspan12EC2-AD (construct 5) with the highest yield contained Tspan12 residues 115-220 and had the shortest connecting linkers (Figure 13 Figure supplement 2B). This Tspan12EC2-AD was produced at 1 mg/L, a yield close to that of 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 the LAIR1-containing antibody (Hsieh and Higgins, 2017), and showed a 10-fold higher expression level than Tspan12EC2 (Figure 13Figure supplement 1C).

Next, to assess if Tspan12EC2-AD adopted the correct fold of Tspan12EC2, we tested the ability of recombinant Tspan12EC2 to bind to Norrin as measured by a colorimetric alkaline phosphatase (AP) reaction-based protein-protein interaction assay using an AP-Norrin fusion protein as a probe (Xu et al., 2004). Each bait protein had a biotinylated C-terminal Avi-tag and was captured on a streptavidin coated micro-well. Tspan12EC2 bound to AP-Norrin (Figure 1D and Figure 13 Figure supplement 3), as did the CRD of Fz4 (Fz4CRD) and the leucine-rich repeat domain of Leucine-rich repeat containing G-protein coupled receptor 4 (Lgr4LRR), extracellular domains that are known to bind to Norrin (Chang et al., 2015; Deng et al., 2013; Shen et al., 2015; Xu et al., 2004) . As expected, AP-Norrin did not bind to Fz5CRD (Figure 1D) (Chang et al., 2015; Smallwood et al., 2007) . Importantly, Tspan12EC2-AD bound to AP-Norrin as did Tspan12EC2 (Figure 1D). Taken together, these results show that (1) EC2 of Tspan12 mediates Norrin binding, in agreement with previous genetic studies using domain swaps within full-length Tspan12 together with co-immunoprecipitation (Lai et al., 2017), and (2) correctly folded Tspan12EC2 can be grafted onto the heavy chain CDR3 without losing Norrin binding.

To assess whether the design principle of the Antibody Display can apply to other protein folds, we tested the RBD of SARS-CoV-2 S protein. We began the design of the Antibody Display SARS-CoV-2 RBD (RBD-AD) by searching the three-dimensional RBD structures of SARS206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228

CoV-2 S protein (Barnes et al., 2020; Huo et al., 2020; Lan et al., 2020; Shang et al., 2020; Walls et al., 2020; Wang et al., 2020b; Wrapp et al., 2020b; Yuan et al., 2020; Zhou et al., 2020b) . As shown in Figure 2A, we computationally grafted the RBD structure of SARS-CoV-2 S protein onto the heavy chain CDR3 of LAIR1-containing antibody to generate a model in in a Fab format, Fab (RBD), by using the same design principles as Tspan12EC2-AD.

Furthermore, we hypothesized that a single VH domain could serve as a scaffold to display SARS-CoV-2 RBD in the absence of light chains. Indeed, human single VH domain mimicking nanobodies from camels has been developed as a minimum size of antigen recognition unit in the absence of light chains (Davies and Riechmann, 1995; Hamers-Casterman et al., 1993) , although specific hallmark mutations are required to maintain the solubility and the binding property of the single VH domain (Davies and Riechmann, 1995; Hamers-Casterman et al., 1993; Muyldermans, 2013) . More recently, Wu et al., identified a highly soluble and stable single VH domain from the human germline Ig heavy chain variable region 3-66*1 (IGHV3-66*1) allele (Wu et al., 2020a) . Therefore, we next designed an engineered single VH domain (eVH) based on the IGHV3-66*1 for the display of SARS-CoV-2 RBD, eVH(RBD), as shown in Figure 23

To verify whether RBD-AD retained binding to ACE2 and neutralizing antibodies, we generated AP fusion proteins with the extracellular domain of ACE2 (ACE2ECD-AP) and two anti-RBD nanobodies (NBVHH-72-AP and NBH11-D4-AP) (Huo et al., 2020; Wrapp et al., 2020a) (Figure 23 Figure supplement 2). We next prepared bait proteins of RBD-AD: (1) Fab (RBD) and 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 eVH(RBD) constructs had a C-terminal mVenus and biotinylated Avi-tag, immobilized on streptavidin-coated wells (Figure 2B and Figure 23Figure supplement 3), and (2) an eVH(RBD) construct contained the human IgG Fc fusion at the C-terminus, immobilized on a protein G-coated well (Figure 2C). As measured the binding by AP reaction-based assays, Fab (RBD) and eVH(RBD) bound robustly and specifically to ACE2ECD-AP, NBVHH-72-AP, and NBH11-D4-AP (Figure 2B-C and Figure 23Figure supplement 3). We conclude that RBD of SARS-CoV-2 S protein can be displayed on the heavy chain CDR3 in functional form. 252

We present a novel antibody engineering approach, named Antibody Display, by grafting POI onto the heavy chain CDR3, which were designed through a structure-guided approach built on the architecture of LAIR1-containing antibodies naturally occurred in malaria-exposed individuals (Hsieh and Higgins, 2017; Tan et al., 2016). Most importantly, each POI retained its biological properties by taking advantage of the structural conservation and stability of the antibody Ig fold. As a proof of principle, we showed that the design and production of Tspan12EC2-AD and RBD-AD were successful. Colorimetric AP reaction-based binding assays further revealed that Tspan12EC2-AD bound to Norrin and RBD-AD bound to ACE2 and NBs (VHH-72 and H11-D4). Our findings suggest a generic design principle of the Antibody Display, particularly for the POI adopting a non-Ig like fold: (1) a known structure or a computational model derived from homologous protein structures, (2) N and C termini in proximity to graft onto the b-strands G and F of the heavy chain, and (3) prevention of steric clashes with the antibody. At present, the limitations for selecting POI for Antibody Display such as protein shape and size remain unclear; further studies will be required to address these questions. Comparing with conventional Fc fusion proteins in which the POI is fused to the N terminus of Fc usually including the flexible hinge region, we reasoned that the Antibody Display, which grafts the POI onto the b-strands G and F of the heavy chain, provides several potential advantages: (1) design and production of a more conformationally constrained chimeric protein, (2) reduction of proteolysis with the POI inserted into a stable antibody Ig fold, and (3) extension 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 of the current antibody engineering toolkit for generating bispecific and chimeric antibodies. Further studies will be needed to explore these potential advantages.

An intriguing feature of using the Antibody Display for Tspan12EC2 is with a better yield of protein production than Tspan12EC2. Our functional assays demonstrated that Tspan12EC2 bound to Norrin directly, in agreement with previous genetic studies (Lai et al., 2017). As Tspan12 playing critical roles in CNS vascular development (Junge et al., 2009; Zhang et al., 2018) and tumorigenesis (Knoblich et al., 2014; Otomo et al., 2014), Tspan12EC2-AD may be a useful reagent for these studies. Moreover, Tspan proteins are well known for their important roles in regulating tumor migration, invasion and metastasis (Charrin et al., 2014; Hemler, 2014; VencesCatalan and Levy, 2018) . Antibodies targeting Tspan29 (CD9), particularly on its EC2 region, have been shown to inhibit the progression of colorectal and gastric cancers (Nakamoto et al., 2009; Ovalle et al., 2007). Furthermore, Tspan proteins are used by several pathogens for infection (Charrin et al., 2014) . For example, Hepatitis C Virus and P. falciparum use Tspan28 (CD81) for infection of hepatocytes (Pileri et al., 1998; Silvie et al., 2003). However, research on Tspan receptors has been hampered by the lack of an efficient approach to produce recombinant EC2 proteins. So far, only Tspan25 (CD53), Tspan28 (CD81), and Tspan29 (CD9) have been produced for biochemical and structural analyses among 33 Tspan receptors (Kitadokoro et al., 2001; Oosterheert et al., 2020; Umeda et al., 2020; Yang et al., 2020b; Zimmerman et al., 2016) . More studies will be required to investigate whether the Antibody Display can apply to other Tspan proteins. 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318

SARS-CoV-2 has resulted in a global pandemic and several types of COVID-19 vaccines such as nucleic acids, viral vectors, and protein subunits have been developed for vaccinations globally (Connors et al., 2021) . Ongoing studies have shown that the RBD of SARS-CoV-2 S protein is the major target of neutralizing antibodies from COVID-19 convalescent individuals (Barnes et al., 2020; Baum et al., 2020; Brouwer et al., 2020; Chi et al., 2020; Ju et al., 2020; Piccoli et al., 2020; Seydoux et al., 2020; Walls et al., 2020; Wang et al., 2020a; Wu et al., 2020b) , consistent with recent findings showing that the RBD, particularly decorated on nanoparticles, can induce a protective humoral immune response against SARS-CoV-2 infection (Cohen et al., 2021; Tan et al., 2021; Yang et al., 2020a) . In this report, we demonstrated that the Antibody Display can be used to display the RBD of SARS-CoV-2 S protein on the heavy chain CDR3 without losing its binding properties. It is not yet known whether RBD-AD can induce neutralizing antibodies, but it would be an interesting test case and it could be relevant for the future preparedness and response capacity against emerging SARS-CoV-2 variants. 319

The human complementary DNA clones of Fz4, Fz5, and Tspan12 were gifts from Jeremy Nathans and the human ACE2 clone (Shang et al., 2020) was obtained from Addgene (code 145033). Synthetic DNA fragments of engineered VH containing SARS-CoV-2 RBD (resi 333527; with a P527A mutation; UniProtKB code P0DTC2; Figure 23Figure supplement 2), NBVHH-72 (PDB code 6WAQ), and NBH11-D4 (PDB code 6YZ5) were obtained from Integrated DNA Technologies. Expression plasmids of MGD21 heavy chain pOPINVH and MGD21 light chain pOPINVL were described Hsieh and Higgins, 2017. The AP fusion Norrin construct was described in Xu et al., 2004.

The following five backbone constructs used for the Fc fusion or biotinylated baits and AP fusion probes were derived from the pHLsec vector (Aricescu et al., 2006) . (1) The pHLsec-3CFc (C103A)-8H vector contains a C-terminal Human Rhinovirus (HRV)-3C protease cleavage site followed by the human immunoglobulin heavy constant gamma 1 (resi 102-330; having a C103A mutation to remove an unpaired cysteine; UniProtKB code P01857) and an 8xHis tag. (2) The pHLsec-Fc-Avi-6H vector contains the human immunoglobulin heavy constant gamma 1 (resi 101-330; UniProtKB code P01857) followed by an Avi tag that can be biotinylated by BirA ligases and a 6xHis tag. (3) The pHLsec-3C-mVenus-Avi-8H vector derived from pHLsecmVenus-12H (Chang et al., 2015) was tagged with a C-terminally HRV-3C protease cleavage site followed by a mVenus fusion protein, an Avi tag and finally an 8xHis tag. (4) The pHL-NAP-Myc-8H vector was constructed N-terminally with the human alkaline phosphatase (AP; resi 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 1-506; NCBI code NP_001623) followed by a Myc tag and finally a C-terminal 8xHis tag. (5) The pHLsec-C-Myc-AP-8H vector contains a C-terminal Myc tag followed by the human AP (resi 18-506; NCBI code NP_001623) and finally an 8xHis tag.

Tspan12 coding segments for the EC2 were PCR-amplified and cloned into the pHLsecmVenus-12H (Chang et al., 2020; Chang et al., 2015) . Tspan12EC2 inserts were grafted into a modified MGD21 heavy chain vector containing a C-terminal HRV-3C protease cleavage site followed by a mVenus fusion protein and finally an 8xHis tag (pHLsec-mMGD21-3C-mVenusAvi-8H), by using three-fragments overlapping PCR experiments. The MGD21 light chain mutant (C91A and C93A) vector was generated by using a two-step overlapping PCR method. A DNA segment coding for CRD of Fz4 (resi 40-179) was constructed into the pHLsec-Fc-Avi-6H and pHLsec-3C-mVenus-Avi-8H. The CRD of Fz5 (resi 31-181) were cloned into the pHLsecFc-Avi-6H, pHLsec-3C-mVenus-Avi-8H, and pHL-N-AP-Myc-8H vectors. The Lgr4LRR was constructed into the pHLsec-3C-mVenus-Avi-8H vector. SARS-CoV-2 RBD insert was grafted into the pHLsec-mMGD21-3C-mVenus-Avi-8H vector. An engineered VH containing SARSCoV-2 RBD was constructed into the pHLsec-3C-mVenus-Avi-8H and pHLsec-3C-Fc (C103A)8H vectors. ACE2ECD (resi 19-615), NBVHH-72, and NBH11-D4 were cloned into the pHLsec-AP-8H vector. All constructs were confirmed by sequencing.

The HHpred server (Soding et al., 2005; Zimmermann et al., 2018) was used to search for homologs of Tspan12EC2 in the protein structure database of the Protein Data Bank (PDB). The structures of Tspan28 (CD81; PDB codes 1G8Q and 5TCX) were selected to generate an initial 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 computational model of Tspan12EC2 with Modeller (Eswar et al., 2006). Notably, the structures of Tspan25 (CD53; PDB code 6WVG), and Tspan29 (CD9, PDB codes 6RLR and 6K4J) were not available during the HHpred search. For the protein design of Antibody Display Tspan12EC2, LAIR1 insert was removed from the model of MGD21 Fab fragment (PDB code 5NST) and the Tspan12EC2 model was docked into the MGD21 VH manually by using COOT (Emsley et al., 2010) and PyMOL Molecular Graphic System (Schrödinger, LLC). For the structure-guided design of Antibody Display SARS-CoV-2 RBD, the crystal structure of SARS-CoV-2 RBD (PBD code 6YLA) was selected and docked into the VH region of the antibody manually by using COOT (Emsley et al., 2010).

HEK293T (ATCC CRL-11268) cells were maintained and transfected with the DNA using polyethylenimine (MilliporeSigma 408727) following the established procedures (Aricescu et al., 2006; Chang et al., 2015; Hsieh and Higgins, 2017; Hsieh et al., 2016) . For Antibody Display Tspan12EC2, DNA constructs expressing heavy chain and light chain were mixed into a 1 to 1 ratio and co-expressed in HEK293T cells. For the biotinylated bait preparations, bait constructs were mixed with a pHLsec-BirA-ER vector (Chang et al., 2015) into a 3 to 1 ratio and cotransfected in HEK293T cells in the presence of 0.1 mM Biotin (MilliporeSigma B4639). The biotinylated baits were purified from conditioned media by the immobilized metal affinity chromatography (IMAC) method as previously described in (Chang et al., 2020) . IMAC method was also used for the purification of His tagged Fc fusion and AP fusion proteins. To evaluate the expression level of secreted mVenus fusion proteins from conditioned media, IMAC 387 388 followed by an in-gel fluorescence imaging method was used as previously described in (Chang et al., 2020) .

The purified Fc fusion baits were captured on 96-well protein-G coated plates (Thermo Fisher Scientific 15131). The biotinylated baits were immobilized on 96-well streptavidin-coated plates (Thermo Fisher Scientific 15500). The wells were then washed three times with wash buffer (10 mM HEPES, pH 7.5, 0.15 M NaCl, 0.05% (w/v) Tween-20) and incubated with a 10-fold dilution of Blocker bovine serum albumin (BSA; Thermo Fisher Scientific 37525) in wash buffer for 1 h at 25 °C. The wells were washed with wash buffer and incubated with conditioned media containing AP probes at 4 °C overnight. The wells were subsequently washed three times with wash buffer and incubated with BluePhos phosphatase substrate solution (Kirkegaard and Perry Laboratories 50-88-00) to visualize the bound AP probes. 410 411 412 413 414 415 416 417 418 419 420

We thank Jeremy Nathans for access to the laboratory and valuable comments on the manuscript; and Matthew Higgins for critical reading the manuscript and valuable comments. F-L H was supported by the Howard Hughes Medical Institute. T-H C was supported by a Human Frontier Science Program long-term fellowship (LT000130/2017-L).

F-L H and T-H C conceived the project, conducted experiments, and wrote the manuscript.

F-L H and T-H C declare no competing interests.

Zhu, N., Zhang, D., Wang, W., Li, X., Yang, B., Song, J., Zhao, X., Huang, B., Shi, W., Lu, R., 382, 727-733. Blacklow, S.C. (2016). Crystal Structure of a Full-Length Human Tetraspanin Reveals a Cholesterol-Binding Pocket. Cell 167, 1041-1051 e1011.

Figure 13Figure supplement 3. Coomassie blue-stained SDS-PAGE and in-gel fluorescence imaging showed the biotinylated baits that were purified by IMAC from the conditioned media and used for the AP based binding assay (Figure 1D).

Figure 23Figure supplement 2. Coomassie blue-stained SDS-PAGE and the in-gel fluorescence results showed the biotinylated and Fc fusion baits and AP fusion probes that were purified by IMAC from the conditioned media. Constructs of ACE2ECD-AP, NBVHH-72-AP, and NBH11-D4-AP contain a C-terminal 8xHis tag. The AP based protein-protein interaction assays (Figure 2B, binding assay. (A) The in-gel fluorescence imaging shows the biotinylated Fab (RBD)-mVenusAvi-8H bait that was purified by IMAC from the conditioned media and used for the binding assay. (B) Biotinylated Fab (RBD) baits were immobilized on streptavidin-coated wells. Bound AP fusion probes were visualized by a colorimetric AP reaction (blue).