Arsenal of Nanobodies for Broad-Spectrum Countermeasures against Current and Future SARS-CoV-2 Variants of Concerns

Nanobodies offer several potential advantages over mAbs for the control of SARS-CoV-2. Their ability to access cryptic epitopes conserved across SARS-CoV-2 variants of concern (VoCs) and feasibility to engineer modular, multimeric designs, make these antibody fragments ideal candidates for developing broad-spectrum therapeutics against current and continually emerging SARS-CoV-2 VoCs. Here we describe a diverse collection of 37 anti-SARS-CoV-2 spike glycoprotein nanobodies extensively characterized as both monovalent and IgG Fc-fused bivalent modalities. The panel of nanobodies were shown to have high intrinsic affinity; high thermal, thermodynamic and aerosolization stability; broad subunit/domain specificity and cross-reactivity across many VoCs; wide-ranging epitopic and mechanistic diversity; high and broad in vitro neutralization potencies; and high neutralization efficacies in hamster models of SARS-CoV-2 infection, reducing viral burden by up to six orders of magnitude to below detectable levels. In vivo protection was demonstrated with anti-RBD and previously unreported anti-NTD and anti-S2 nanobodies. This collection of nanobodies provides a therapeutic toolbox from which various cocktails or multi-paratopic formats could be built to tackle current and future SARS-CoV-2 variants and SARS-related viruses. Furthermore, the high aerosol-ability of nanobodies provides the option for effective needle-free delivery through inhalation.


INTRODUCTION
Since we planned to test VHH-Fcs in hamsters for in vivo efficacy, we pre-emptively assessed their in 177 vivo stability and persistence. We chose 1d VHH-Fc as a representative and included VHH-72 VHH-Fc, 178 whose modified/enhanced version is currently in a phase 1 clinical trial, as a reference. Hamsters were 179 injected intraperitoneally (IP) with 1 mg of each antibody and serum antibody concentration was 180 monitored for up to four days by ELISA. Significant and comparable VHH-Fc concentrations were present 181 in the hamster sera for both 1d and VHH-72 VHH-Fcs on days 1 and 4 post injection (Fig. 3D) Table S6). This resulted in reduced % recoveries, 188 measures of VHH stability against aerosolization, and corresponding to the proportion of VHHs that 189 remained as soluble monomer following aerosolization ( Fig. 3E; Fig. S9; Table S6). The majority of VHHs 190 (18 out of 28 VHHs tested), however, were stable against aerosolization with high % recoveries (Fig. 3E). 191 Additionally, several VHHs still showed a high % recovery upon aerosolization (50 -70 %) despite the 192 formation of some visible aggregates. Comparison of ELISA-derived EC50s of select pre-aerosolized vs 193 post-aerosolized VHHs clearly demonstrated aerosolization did not compromise the binding activities of 194 VHHs ( Fig. 3E; Fig. S9). 195 196

Screening for neutralizing VHHs by surrogate virus neutralization assays (SVNAs) 197
A preliminary screen of 25 VHHs (14 RBD-specific, six NTD-specific and five S2-specfic) by  SPR-based SVNAs identified several potential neutralizers, predominantly from the RBD-binding cohort 199 ( Fig. S10; Table S7). A more relevant SVNA, which assessed the ability of antibodies to block binding of S 200 to Vero E6 cells displaying ACE2, was then used as a screen to identify neutralizing VHHs and VHH-Fcs. 201 Neutralizing VHHs displayed similar potencies (IC50: 5 -21 nM) and outperformed the benchmark VHH-202 72 (IC50: 59 nM) by as much as 12-fold ( Fig. S11; Table 2). Compared to VHHs, a larger number of VHH-Fcs 203 demonstrated neutralization capabilities ( Fig. 4A; Fig. S12; Table 2). While neutralizing monomer VHHs 204 did not benefit from reformatting (except for the VHH-72 benchmark), several non-neutralizing VHHs 205 (three RBD-specific and three NTD-specific) benefitted profoundly from reformatting and were 206 transformed into neutralizers that had potencies similar to other RBD-specific VHH-Fcs. All S2-specific 207 VHHs remained non-neutralizing as VHH-Fcs. 208 Extending our SVNAs to variants Alpha, Beta, Gamma, Delta and Kappa using all of the RBD-specific 209 and a subset of NTD-specific VHH-Fcs ( Fig. 4B; Table S8), several observations were made. First, for 210 cross-neutralizing VHHs the IC50s across variants did not change significantly. Second, while all Wuhan 211 neutralizers also remained Alpha neutralizers, some lost their capability to inhibit Beta, Gamma, Delta 212 and Kappa with variable cross-neutralizing patterns. In particular, with respect to the RBD-specific VHHs, 213 the cross-neutralization profiles for Beta vs Gamma and Delta vs Kappa were identical, similarly 214 reflective of the key escape mutations in these variants (K417N, E484K and N501Y for Beta vs K417T, 215 E484K and N501Y for Gamma; L452R and T478K for Delta vs L452R and E484Q for Kappa). Third, and 216 importantly, 12 out of 20 VHH-Fcs (10 RBD-specific, two NTD-specific) were Delta neutralizers, nine of 217 which (eight RBD-specific, one NTD-specific) neutralized across all variants. The majority of these nine 218 pan-neutralizers (six RBD-specific, one NTD-specific) also neutralized SARS-CoV.  Table 2). For RBD-specific VHHs, potency increases (IC50 decreases) of 2 -100-fold were 228 observed; only one VHH (18) was unaffected with reformatting (IC50 range: 2.3 -30.8 nM; median: 7.6 229 nM). NTD-specific VHH-Fcs demonstrated weaker potencies (IC50 range: 11.3 -86.9 nM; median: 18.5 230 nM; 4 of 9 non-neutralizing). However, bivalency also significantly improved (~9-fold) the potencies of 231 SR01 and SR03 and transformed a non-neutralizing VHH (SR16) into a potent neutralizing VHH-Fc. 232 Consistent with the aforementioned SVNA results and previous data 29 , the VHH-72 benchmark also 233 improved, elevated from a weak VHH (IC50: 490 nM) to a strong VHH-Fc (25 nM) neutralizer. S2-specifiic 234 VHHs remained non-neutralizing with reformatting. 235 236 All RBD-and NTD-specific VHH-Fcs that were neutralizing by PVNA were also neutralizing in a live 237 virus neutralization assay (LVNA) (Fig. 4A; Table 2; Fig. S14). However, compared to the former method, 238 the LVNA IC50 values were lower and more variable. For RBD-specific VHH-Fcs an IC50 (range: 0.0008 -76 239 nM; median: 2.8 nM) was observed. The most potent VHH-Fcs belonged to bin 2/3/4 (IC50 range: 0.0008 240 -3.1 nM; median: 1 nM), with 05 showing the greatest potency (IC50: 0.0008 nM) followed closely by 02 241 and MRed05 (IC50s: 0.12 and 0.17 nM, respectively). Bin 1 neutralizers, to which VHH-72 belonged and 242 displayed a similar IC50 (8.5 nM), exhibited intermediate potencies (range: 1.9 -11.2 nM; median: 6.3 243 nM), followed by bin 5/6 neutralizers (range: 9.9 -76 nM; median: 58 nM). Weaker neutralizing 244 potencies were observed with NTD-specific VHH-Fcs. Here, six of nine VHH-Fcs, representing at least one 245 epitope bin, were neutralizing. Interestingly, three new neutralizers emerged from the pool of S2-246 specific VHH-Fcs using the LVNA, with S2A3 the most potent. (~10-fold). Although from the most potent bin (2/3/4), 02, 04 and 05, consistent with the cross-reactivity 256 data ( Fig. 2A), were completely abrogated presumably by the Beta mutations in the RBD (K417N, E484K, 257 N501Y), several others including MRed05, 10 and 15 did retain their high neutralizing potencies against 258 both Alpha and Beta variants. A similar trend was observed for the NTD-specific neutralizing VHHs: 259 against the Alpha variant, potencies either remained essentially the same as those for the Wuhan 260 variant or improved, while against the Beta variant, potencies diminished. Nonetheless, SR01 and SR16 261 maintained respectable neutralization potencies against Beta. The potencies of S2-specific neutralizers 262 (S2A3, S2G3, S2G4) were also decreased with variants. However, the lead S2A3 still maintained 263 comparable potencies across all three variants (IC50 of 12.2 nM, 31 nM and 54 nM for Wuhan, Alpha and 264 Beta [ Table 2]). Collectively, the neutralization profiles across Wuhan, Alpha and Beta variants were 265 consistent with cross-reactivity profiles ( Fig. 2A). Based on the cross-reactivity ( Fig. 2A)  To identify the number of non-overlapping epitopes, VHHs were subjected to epitope binning 271 experiments by SPR and sandwich ELISA. SPR assays were performed by injecting paired combinations of 272 eight RBD-specific VHHs, six NTD-specific VHHs and ten S2-specific VHHs over a SARS-CoV-2 spike 273 glycoprotein surface (Fig. S16A). A conceptually similar assay to SPR was also performed by sandwich 274 ELISA (Fig. S16B). From the 33 VHHs tested, 14 unique epitope bins were identified: six for RBD-specific 275 VHHs, three for NTD-specific VHHs and five for S2-specific VHHs ( Fig. 5A; Table 1). The benchmark VHH-72 276 binned with RBD-specific VHHs 1d, 07, 12, 18, 20 and MRed04. With the exception of VHH 04, all 277 remaining bin 1, 2, 3 and 4 VHHs (13 in total), as well as VHH-72, binned with ACE2, consistent with them 278 being potent neutralizers. The results of sandwich ELISAs extended to include variants Alpha, Beta, 279 Gamma, Delta and Kappa (Fig. S16C), using lead VHHs from various bins as capture antibodies, were 280 consistent with epitope binning as well as cross-reactivity ( Fig. 2A) data. 281 282 Next, we investigated the conformational nature of the epitope bins at peptide-level resolution with 283 hydrogen-exchange mass spectrometry (HDX-MS) ( Table S9) HDX-MS profiles and previously described epitope bins and subunit/domain specificity. A common 286 binding mode adjacent to the RBM and distant from known VoC mutations was observed for bin 1 (Fig.  287   5B). This overlaps with core binding contacts of VHH-72 on SARS-CoV 29 where neutralization is achieved 288 by steric blocking of ACE2. The binding profiles for the strongest neutralizers in bins 2/3/4 overlap the 289 ACE2 binding site 10 and known conformational hotspots 60 . It was not possible to further resolve 290 epitope diversity within the context of this dataset, however it is evident that a range of binding 291 patterns exists 61, 62 , and there is a correlation between stabilizations spanning mutations in VoCs and 292 the loss/attenuation of neutralization (Fig. 5B). Such granularity assists in understanding and predicting 293 neutralization potency as novel variants emerge. 294 295 Epitopes for bin 6 (VHH 03 and 11) span the C-terminus of the RBD and SD1 63 (Fig. 5B), explaining 296 why binding is limited to constructs containing SD1 ( Table S2). Stabilization of the SD1 hinge responsible 297 for RBD motion highlights a potential inhibitory mechanism for VHH 11 64 . While it is challenging to 298 delineate between an epitope and conformational effects based on HDX profiles alone, distinct binding 299 responses with common conformational hotspots were observed for the NTD binders ( Fig. 5C; Fig. S17). 300 Interestingly, none of the NTD-supersite loops 14-17, 19, 65-67 covered here displayed significant HDX shifts, 301 except for N4 stabilized by SR02, suggesting a range of binding modes beyond the NTD-supersite. 302 Further, stabilizations partially overlap a previously described conformationally active epitope with low 303 variability and neutralization vulnerability 15 . Supersite binders appear to be vulnerable to escape 304 mutants 7, 14, 68 , highlighting the importance of targeting and characterizing alternative NTD epitopes. 305 An epitope for S2A3 spanning the linker/CD/HR2 motifs 21 is described in Fig. 5D. This region is 306 upstream of known S2 epitopes and is crucial for the structural transition required for virus-cell fusion 69 . 307 We cannot rule out the involvement of other residues within CD/HR2 regions due to gaps in coverage. 308 Given that none of the mutations within the six SARS-CoV-2 variants overlap the epitope, the cross-309 reactivity against the six variants ( Fig. 2A) and cross-neutralization against Alpha and Beta ( were recognizing linear epitopes, with the majority (9 out of 12) being S2-specific (Table 1; Fig S18). 314 315

In vivo therapeutic efficacy of VHH-Fcs 316
The in vivo therapeutic efficacy of VHH-Fcs which were neutralizing by LVNA were assessed in a 317 hamster model of SARS-CoV-2 infection. Five VHH-Fcs were selected to cover a wide range of important 318 attributes including in vitro neutralization potencies and breadth, epitope bin, subunit/domain 319 specificity and cross-reactivity pattern. These included three RBD-specific (1d, 05, MRed05), one NTD-320 specific (SR01) and one S2-specific (S2A3) VHH-Fcs. Cocktails of two VHH-Fcs were also included to 321 explore synergy between the antibody pairs recognizing distinct epitopes within the RBD (1d/MRed05) 322 or RBD and NTD (1d/SR01). 323

324
Hamsters were administered IP with 1 mg of VHH-Fcs 24 h prior to intranasal challenge with SARS-325 CoV-2 Wuhan isolate. Daily weight change and clinical symptoms were monitored. At 5 dpi, lungs were 326 collected to determine viral titers. Viral titer decrease and reversal of weight loss in antibody treated 327 versus control animals were taken as measures of antibody efficacy. Animals treated with RBD binders 328 1d, 05, and MRed05 showed reduced lung viral burden by three, five and six orders of magnitude, 329 respectively, relative to PBS or VHH-Fc isotype controls, with 05 and MRed05 reducing viral burden to 330 below detectable levels (Fig. 6A). The RBD-specific VHH-72 benchmark caused a mean viral decrease of 331 four orders of magnitude. The NTD binder SR01, and interestingly, the S2 binder S2A3, were also 332 effective neutralizers, decreasing mean viral titers by four and three orders of magnitude, respectively. 333 Both 1d/SR01 and 1d/MRed05 cocktails decreased viral titers by 6 orders of magnitude to undetectable 334 levels of virus infection. While it was not possible to unravel potential synergies for 1d/MRed05, as 335 MRed05 alone displayed essentially the same efficacy as the 1d/MRed05 combination, it was apparent 336 that the 1d/SR01 combination benefited from synergy, decreasing viral titers by a further 2 -3 orders of 337 magnitude to undetectable levels, relative to 1d or SR01 alone. Moreover, in accordance with the viral 338 titer decreases, a gradual reversal of weight loss in infected animals was observed with antibody 339 treatment starting on 2 dpi (Fig. 6B,C). A strong negative correlation (r = -0.9436; p <0.0001) was 340 observed between weight change and viral titer at 5 dpi (Fig. 6D). macrophages and T lymphocytes, in the lung parenchyma 70 . As expected, this was the case for the non-350 treated PBS and isotype control groups. In contrast, we observed a substantial reduction of 351 macrophages and T lymphocytes infiltrate in lung parenchyma with antibody treatment (Fig. S19, S20). 352 The most dramatic decreases in the number of macrophages and T lymphocytes were seen with 05, 353 MRed05, 1d/MRed05 and 1d/SR01 treatments. Interestingly, a reduction in inflammatory responses was 354 also associated with a decrease in the number of apoptotic cells in antibody-treated animals (Fig. S21). 355 Altogether, the viral titer, weight change and immunohistochemistry results consistently demonstrate 356 that a single dose of several of our VHH-Fcs reduced viral burden, immune cell infiltration and apoptosis 357 in the lungs of infected hamsters. 358 359

DISCUSSION 360
The efficacy of COVID-19 therapeutic antibodies and vaccines is persistently being threatened by 361 emergence of new VoC escape mutants, presenting a pandemic state with no end in sight. Developing 362 broad-spectrum antibodies that can neutralize current and emerging VoCs represents an effective 363 countermeasure against the current SARS-CoV-2 pandemic and future ones. 364

365
Two key structural features of nanobodies make them promising candidates as broad-spectrum 366 therapeutics. First is their ability to access epitopes on the surface of the spike glycoprotein that are 367 hidden from mAbs and conserved across VoCs 34, 37, 45 . Second is their high modularity, allowing for their 368 rapid conversion into multimeric/multi-paratopic constructs with favorable manufacturability profiles. 369 Multimerization can lead to drastic increases in the efficacy of anti-COVID-19 nanobodies and broaden 370 their cross-reactivity across variants 29, 34, 36, 39 . Significantly, the chance of developing multimeric/multi-371 paratopic constructs with the desired cross-neutralization breadth improves with the diversity of 372 nanobody building blocks available. 373

374
With the goal of developing broad-spectrum therapeutics, we employed multiple immunization, 375 phage display library construction and panning strategies to identify a diverse collection of nanobodies. Our study also provides valuable insights into how various in vitro neutralization assays predict 389 antibody efficacy. The flow cytometry-based SVNA was shown to effectively identify neutralizing 390 antibodies that were RBD-specific, providing a viable alternative to the PVNA or LVNA, which are labor-391 intensive, difficult to standardize, inconvenient and not readily accessible as they require operating in 392 biosafety level 2 or 3 labs. However, the SVNA occasionally missed NTD-specific neutralizing antibodies, 393 and, similar to the PVNA, failed altogether to identify neutralizing antibodies that were S2-specific. 394 Furthermore, the SVNA did not have the sensitivity of the LVNA to identify VHHs with weaker potencies 395 or, similar to the PVNA, to fine-rank neutralizing antibodies. 396

397
For several reasons, the number of epitope bins (14) identified in the current study likely under-398 estimates the number of actual distinct epitopes. First, VHHs that recognize (i) partially overlapping 399 epitopes, (ii) fully overlapping epitopes of significantly different nature, or (iii) non-overlapping epitopes, 400 but manifest exclusive binding as a consequence of conformational competition or steric clashes 401 between the VHH pairs, would fall under the same epitope bins. Second, indicators of distinct epitopes 402 such as differential HDX-MS footprints, epitope types, cross-reactivity profiles, neutralization potencies, 403 and cross-neutralization profiles not accounted for by affinity alone, are seen amongst VHHs within the 404 same bin. Thus, the repertoire of structurally and functionally distinct epitopes are more diverse than 405 what can be gleaned from epitope binning analysis alone. 406

407
In vitro neutralization assays with the Wuhan SARS-CoV-2 variant showed that the majority of the 408 nanobodies were RBD-specific. Importantly, and to our knowledge for the first time, we demonstrated 409 that several NTD-and S2-specific VHHs were also potent and efficacious neutralizers in vitro and in vivo. 410 Significantly, neutralizing nanobodies showed high epitopic diversity, originating from at least eight 411 different epitope clusters. The vast majority of these VHHs -including NTD and S2 VHHs -remained 412 potent in vitro neutralizers against the Alpha, Beta, Gamma, Delta and Kappa variants as well. A sample 413 of in vitro neutralizing nanobodies, representing RBD-, NTD-and S2-specific VHHs, were also shown to be 414 efficacious in vivo neutralizers as single VHH-Fcs or as paired combinations of VHH-Fcs that targeted RBD 415 and NTD, with some capable of complete viral clearance from hamster lungs. These results also confirm 416 the strong positive correlation between the in vitro and in vivo neutralization data and indicate that the 417 remaining in vitro neutralizers should also be in vivo neutralizers. Moreover, given the broad cross-418 reactivity and cross-neutralization profiles of many of these nanobodies, it is reasonable to expect these 419 pan-reactive antibodies will also effectively neutralize currently untested and emerging VoCs, e.g., 420 Omicron, and SARS-related viruses. 421

422
The VHH-72 benchmark nanobody, a modified, enhanced VHH-Fc version of which is currently being 423 developed for COVID-19 therapy, has the advantage of being broadly neutralizing and binding to a highly 424 conserved cryptic epitope in the RBD region which is difficult to access with conventional mAbs 29, 39 . 425 Here we have identified five VHH-72-like nanobodies (1d, 07, 12, 20, MRed04) that map to the same 426 epitope as VHH-72 and demonstrate similar cross-reactivity and cross-neutralization profiles in the VHH-427 Fc format. In monomeric formats, however, the VHHs significantly outperform VHH-72, but whether they 428 would do the same against the enhanced version of VHH-72 remains to be seen. A similar broad cross-429 reactivity profile to VHH-72 was seen with two other neutralizing nanobodies (11, SR01) that by HDX/MS 430 experiments map to different but well-conserved epitopes, indicating that these nanobodies may 431 similarly bind to cryptic epitopes conserved across variants. The doubly pan-specific SR01/1d cocktail 432 completely cleared hamsters of viral burden, making for a promising broad-spectrum combination 433 therapeutic. 434

435
With an abundance of neutralizing nanobodies on hand, many possibilities exist for designing 436 optimized multimeric/multi-paratopic therapeutic agents. This is implicitly evident from our epitope 437 binning experiments performed against several SARS-CoV-2 variants (Fig. S16C)

Recombinant antigens and ACE2 465
Purified recombinant spike and ACE2 proteins used in the current study are described in Table S1. They 466 were either purchased or produced in-house as described (Table S1) ( 76-80 ). Proteins were purified using 467 standard immobilized metal-ion affinity chromatography or protein A affinity chromatography. 468  presence of antibodies that block the binding of S to ACE2 (surrogate for neutralization) in the immune 527 sera of llamas, 400 ng of chemically biotinylated SARS-CoV-2 S was mixed with 1 ´ 10 5 Vero E6 cells in 528 the presence of 2-fold dilutions of sera (pre immune, day 21 and day 28 sera) in a final volume of 150 µL. 529 Following 1 h of incubation on ice, cells were washed twice with PBSB by centrifugation for 5 min at 530 1200 rpm and then incubated for an additional hour with 50 µL of Streptavidin, R-Phycoerythrin 531 Conjugate (SAPE, Thermo Fisher, Cat#S866) at 250 ng/mL diluted in PBSB. After a final wash, cells were 532 resuspended in 100 µL PBSB and data were acquired on a CytoFLEX S flow cytometer (Beckman Coulter, 533 Brea, CA) and analyzed by FlowJo software (FlowJo LLC, v10.6.2, Ashland, OR). Percent inhibition 534 (neutralization) was calculated according to the following formula: 535 Where, 538 Fn is the measured fluorescence at any given competitor serum dilution 539 Laboratories. Two independent phage-displayed VH/VHH libraries were constructed from ∼5 × 10 7 546 PBMCs as described previously 81, 82, 84 . Briefly, total RNA was extracted from PBMCs using TRIzol™ Plus 547 RNA Purification Kit (Thermo Fisher, Cat#12183555) following the manufacturer's instructions and used 548 to reverse transcribe cDNA with SuperScript™ IV VILO™ Master Mix supplemented with random 549 hexamer (Thermo Fisher, Cat#SO142) and oligo (dT) (Thermo Fisher, Cat#AM5730G) primers. VH/VHH 550 genes were amplified using semi-nested PCR and cloned into the phagemid vector pMED1, followed by 551 transformation of E. coli TG1 (Lucigen, Middleton, WI, Cat#60502-02) to construct two libraries with 552 sizes of 1 × 10 7 and 2 × 10 7 independent transformants for Green and Red, respectively. Both libraries 553 showed an insert rate of ∼95% as verified by DNA sequencing. Phage particles displaying the VHHs were 554 rescued from E. coli cell libraries using M13K07 helper phage (New England Biolabs, Whitby, Canada, 555 Cat#N0315S) as described in 81 and used for selection experiments described below. 556 557 b) Library selection and screening. Library panning and screening were performed essentially as 558 described 81, 82, 85 , using SARS-CoV-2 Wuhan spike glycoprotein fragments as target antigens. Library 559 selections were performed on microtiter wells under six different phage binding/elution conditions 560 designated P1 -P6. Briefly, for the phage binding step, library phages were diluted at 1 ´ 10 11 colony-561 forming units (CFU)/mL in PBSBT [PBS supplemented with 1% BSA and 0.05% Tween 20] and incubated 562 in antigen-coated microtiter wells for 2 h at 4°C. For P1 -P4, phages were added to wells with passively-563 adsorbed S (10 µg/well; P1), passively-adsorbed S2 (10 µg/well; P2), streptavidin-captured biotinylated 564 S1 (0.5 µg/well; P3) and streptavidin-captured biotinylated RBD (0.5 µg/well; P4). For P5, phages were 565 pre-absorbed on passively-adsorbed RBD wells (10 µg/well) for 1 h at 4°C and then the unbound phage 566 in the solution was transferred to wells with streptavidin-captured biotinylated S1 (0.5 µg/well) in the 567 presence of non-biotinylated RBD competitor in solution (10 µg/well). Following the binding stage (P1 -568 P5), wells were washed 10 times with PBST and bound phages were eluted by treatment with 100 mM 569 glycine, pH 2.2, for 10 min at room temperature, followed by immediate neutralization of phages with 2 570 M Tris. Similar to P4, in P6, phages were bound on streptavidin-captured biotinylated RBD but elution of 571 bound phages were carried out competitively with 50 nM human ACE2-Fc following the washing step. 572 For all pannings, a small aliquot of eluted phage was used to determine the titer on LB-agar/ampicillin 573 plates and the remaining phage were used for subsequent amplification in E.coli TG1 strain 81 . The 574 amplified phages were used as input for the next round of selection as described above. 575 After two rounds of selection, 16 (Green) or 12 (Red) colonies from each of the P1 -P6 selections 576 were screened for antigen binding by monoclonal phage ELISA against S, S1, S2 and RBD. Briefly, 577 individual colonies from eluted-phage titer plates were grown in 96 deep well plates in 0.5 mL 2YT 578 media/100 µg/mL-carbenicillin/1% (w/v) glucose at 37°C and 250 rpm to an OD600 of 0.5. Then, 10 10 CFU 579 M13K07 helper phage was added to each well and incubation continued for another 30 min under the 580 same conditions. Cells were subsequently pelleted by centrifugation, the supernatant was discarded and 581 the bacterial pellets were resuspended in 500 µL 2YT/100 µg/mL carbenicillin/50 µg/mL kanamycin and 582 incubated overnight at 28°C. Next day, phage supernatants were recovered by centrifugation, diluted 3-583 fold in PBSTC and used in subsequent screening assays by ELISA. To this end, antigens were coated onto 584 microtiter wells at 50 ng/well overnight at 4°C. Next day, plates were blocked with PBSC, washed five 585 times with PBSTC, and 100 µL of phage supernatants prepared above were added to wells, followed by 586 incubation for 1 h at room temperature in an orbital shaking platform. After 10 washes, binding of and Beta variants were also tested by SPR and amine coupled using the conditions described above. All 658 affinities were calculated by fitting reference flow cell-subtracted data to a 1:1 interaction model using 659 BIAevaluation Software v3.0 (Cytiva). 660 For VHH 12 and MRed05, VHH-Fc formats were used in SPR experiments. Approximately 200 RUs of 661 VHH-Fcs (2 µg/mL) were captured on goat anti-human IgG surfaces (4000 RUs, Jackson ImmunoResearch, 662 Cat#109-005-098) at a flow rate of 10 µL/min for 30 s. A range of SEC-purified RBD fragments (Table S1; 663 Superdex 75™ 10/300 GL column (Cytiva) and PBS as running buffer, as described above. Protein 734 fractions corresponding to the chromatogram's monomeric peak were pooled, quantified and its 735 concentration adjusted to 0.5 mg/mL. One mL of each VHH was subsequently aerosolized at room 736 temperature with a portable mesh nebulizer (AeroNeb Solo, Aerogen, Galway, Ireland), which produces 737 3.4-μm particles. Aerosolized VHHs were collected into 15 mL round-bottom polypropylene test tubes 738 (VWR, Cat#C352059) for 5 min to allow condensation and were subsequently quantified and kept at 4°C 739 until use. Then 200 µL aliquots of pre-and post-aerosolized VHHs were subjected to SEC to obtain 740 chromatogram profiles. Additionally, condensed VHHs were closely monitored for the formation of any 741 visible aggregates, and in cases where aggregate formation were observed, they were removed by 742 centrifugation prior to concentration determination, SEC analysis and ELISA. % soluble aggregate was 743 determined as the proportion of a VHH that gave elution volumes (Ves) smaller than that of the 744 monomeric VHH fraction. % recovery was determined as the proportion of a VHH that remained 745 monomer following aerosolization. 746

747
To assess the effect of aerosolization on functionality of VHHs, the activities of post-aerosolized VHHs 748 were determined by ELISA and compared to those for pre-aerosolized VHHs. To perform ELISA, S1-Fc 749 (ACRO Biosystems, Cat#S1N-C5255) was diluted in PBS to 500 ng/mL, and 100 µL/well were coated 750 overnight at 4°C. Next day, plates were washed with PBST and blocked with 200 µL PBSC for 1 h at room 751 temperature. After five washes with PBST, serial dilutions of the pre-and post-aerosolized VHHs were 752 added to wells and incubated for 1 h at room temperature. Then plates were washed 10 times with 753 PBST and binding of VHHs to S1-Fc was detected with rabbit anti-6xHis Tag antibody HRP Conjugate 754  Table S9. Finally, significant changes in deuteration was assigned based on two cutoffs (3 x 817 SD and 1-p = 0.98) using MS Studio 94 . 818 819 4.9. Surrogate virus neutralization assays 820 a) ACE2 competition assay by ELISA. Wells of NUNC® Immulon 4 HBX microtiter plates (Thermo Fisher) 821 were coated overnight at 4°C with 50 ng/well of S in 100 µL PBS, pH 7.4. Next day, plates were blocked 822 with 250 µL PBSC for 1 h at room temperature. For ACE2/VHH competitive binding to SARS-CoV2 S 823 (Wuhan), 50 µL of human ACE2-Fc (ACROBiosystems, Cat#AC2-H5257) at 400 ng/mL was mixed with 50 824 µL of VHH at 1 µM, and then transferred to SARS-CoV-2 S coated microtiter plate wells. After 1 h 825 incubation at room temperature, plates were washed 10 times with PBST and the ACE2-Fc binding was 826 detected using 1 µg/mL goat anti-human IgG (Fc specific) HRP conjugate antibody (Sigma, Cat#A0170) in 827 100 µL PBSCT. After 10 washes with PBST, the peroxidase (HRP) activity was determined as described in orientation was also performed (VHH followed by VHH + ACE2). Surfaces were regenerated using a 12 s 840 pulse of 10 mM glycine, pH 1.5, at a flow rate of 100 µL/min. All pairwise combinations of VHHs and 841 ACE2 were analyzed. VHHs that competed with ACE2 for SARS-CoV-2 spike glycoprotein binding showed 842 no increase in binding response during the second injection. Conversely, a binding response was seen 843 during the second injection for VHHs that did not compete with ACE2. 844 845 c) ACE2 competition assay by flow cytometry. Experiments were performed as described in 4.3c, except 846 that biotinylated S/Vero E6 cells were mixed with VHHs or VHH-Fcs instead of sera. Additionally, assays 847 were performed against biotinylated SARS-CoV-2 Wuhan, Alpha, Beta, Gamma, Delta and Kappa as well 848 as SARS-CoV S. As internal reference of competition experiments, competition assay with recombinant 849 human ACE2-H6 in lieu of VHH was also included. A20.1, a C. difficile toxin A-specific VHH 82 was used as  Retrieval Solution 2 (ER2, EDTA buffer, pH 8.8) at 98°C for 20 min. Epitopes were exposed using ER1 for 927 Iba-1 and SARS-CoV-2 N and ER2 for CD3. After washing steps, non-specific endogenous peroxidases 928 summarizing VHH kinetic rate constants, kas and kds, determined by SPR. Diagonal lines represent 957 equilibrium dissociation constants, KDs (see also Table 1). Maps were constructed using the VHH binding 958 data ( Fig. S3; Table S2) against SARS-CoV-2 Wuhan S (all except 12 and MRed05) or RBD/SD1 (12 and 959 MRed05). VHH subunit/domain specificities were determined by SPR and ELISA ( Fig. S3; Table S2; Table  960 S3). Anti-SARS-CoV S VHH-72 which cross-reacts with SARS-CoV-2 RBD 29 and the monomeric ACE2 961 plots of % folded vs temperature ( Fig. S7; Table S5). C) Summary of VHH ΔG 0 data. ΔG 0 (as well as other 980 thermodynamic parameters, Cm and m values) are reported in Table S5. D) In vivo stability and 981 persistence of VHHs. Stability and persistence were determined by monitoring the concentration of a 982 representative VHH-Fc (1d) in hamster blood at various days post-injection by ELISA. VHH72-Fc was used 983 as the benchmark. E) Stability of VHHs against aerosolization. Summary of % recovery of all (i) and lead 984 (ii) VHHs are shown. Percent recovery represents the proportion of a VHH that remained soluble 985 monomer following aerosolization. Graphs were generated based on the data in Fig. S9A and Table S6. Wuhan, Alpha and Beta SARS-CoV-2 variants. See also Table 2 for IC50 values. Graphs were generated 995 based on the data in Fig. S12 -Fig. S15 and Table S8 S2A3  10  -nd  -12.2  31  54   S2A4  11  -------S2F3  12  -nd  ----S2G3  13  ----~200  ---4 -S2G4  14  ----~200  --MRed11  nd  -------MRed18  12  -----~400  -MRed19 12