Broadly neutralizing aptamers to SARS-CoV-2: A diverse panel of modified DNA antiviral agents

Abstract

Since its discovery, COVID-19 has rapidly spread across the globe and has had a massive toll on human health, with infection mortality rates as high as 10%, and a crippling impact on the world economy. Despite numerous advances, there remains an urgent need for accurate and rapid point-of-care diagnostic tests and better therapeutic treatment options. To contribute chemically distinct, non-protein-based affinity reagents, we report here the identification of modified DNA-based aptamers that selectively bind to the S1, S2, or receptor-binding domain of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) spike protein. Several aptamers inhibit the binding of the spike protein to its cell-surface receptor angiotensin-converting enzyme 2 (ACE2) and neutralize authentic SARS-CoV-2 virus in vitro, including all variants of concern. With a high degree of nuclease resistance imparted by the base modifications, these reagents represent a new class of molecules with potential for further development as diagnostics or therapeutics.

Graphical abstract

We report a generally applicable method for developing modified DNA aptamer antagonists to viral targets, in this case, the SARS-CoV-2 spike protein. We identified an antiviral agent with potent activity against all variants of concern of SARS-CoV-2. Broad neutralization activity is due to occupancy of a conserved epitope overlapping the receptor-binding motif.

Introduction

Nearly 3 years into the COVID-19 pandemic, the world is still struggling with surges, lockdowns, loss of human life, and economic stress despite the availability of effective vaccines since early 2021.¹ Furthermore, the emergence of increasingly transmissible variants of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emphasizes the need for accurate and rapid point-of-care diagnostic tests that do not require complex instrumentation, and better therapeutic treatment options.

SARS-CoV-2 is an enveloped positive-sense single-stranded RNA virus that typically infects respiratory and gastrointestinal tracts in a wide range of mammalian hosts.² The SARS-CoV-2 genome encodes only four structural proteins that function in viral particle assembly and infection.³ The trimeric spike (S) protein coats the surface of the virus and is responsible for gaining host cell entry through the angiotensin-converting enzyme 2 (ACE2) protein.^4,5 Spike is a heavily glycosylated type I transmembrane protein that is produced as a precursor that is cleaved by furin-like protease into the non-covalently associated S1 fragment containing the ACE2 receptor-binding domain (RBD), and the membrane fusion fragment S2.^6,7,8 The RBD is dynamic and exists in either the ACE2 binding competent “up” conformation or the inaccessible “down” conformation.⁹ ACE2 receptor binding by spike RBD triggers a second cleavage event in the S2 fragment (S2′) by cellular proteases such as TMPRSS2 and cathepsin L, causing the dissociation of S1 and the irreversible conformational change of S2 to form the post-fusion construct, which ultimately leads to membrane fusion.^10,11,12 Seventeen residues from the spike RBD comprise the binding interface with ACE2, of which eight are conserved between SARS-CoV and SARS-CoV-2¹³ (overall amino acid sequence identity for the full-length spike protein is 88%).

One concern throughout the pandemic has been the emergence of genetic variants of SARS-CoV-2. Isolates identified in multiple infection clusters or countries that possess substitutions likely to result in a phenotypic change are designated variants of interest (VOI) by the World Health Organization (WHO). Those isolates that are associated with an increase in transmissibility or virulence or decrease the effectiveness of existing vaccines, antiviral drugs, or diagnostics, are classified as a variants of concern (VOC). Both VOIs and VOCs are now assigned letters from the Greek alphabet for easy reference. Among VOCs, the once dominant Alpha variant was swiftly overtaken in the summer of 2021 by the Delta variant, which originated in India. The Delta variant is more than twice as contagious as previous variants and became the dominant circulating variant during 2021. The Omicron variant, first identified in South Africa in November 2021, carries multiple spike substitutions, of which 12 in the RBD alone had not been seen in previous VOCs, rapidly replaced other VOCs as the globally dominant lineage by January 2022.^14,15 Since the spike protein is a primary target for vaccine-elicited antibodies and is a suitable target for rapid detection tests and therapeutic interventions, the ongoing appearance of such mutations could allow the virus to evade current response strategies.

DNA aptamers are non-protein-based molecular recognition reagents with potential application in the COVID-19 diagnostic and therapeutic arenas. These relatively small (8–15 kDa) synthetic compounds are frequently compared with antibodies in terms of target binding affinity and specificity.¹⁶ Unlike antibodies, traditional DNA aptamers lack the chemical diversity inherent in the 20 amino acids due to the similarity of physico-chemical properties of the four nucleobases. We have therefore developed a new class of DNA-based aptamers called slow off-rate modified aptamers (SOMAmers) containing versatile chemical moieties that endow nucleotides with protein-like functional groups uniformly substituted at the 5-position of deoxyuridine (dU).^17,18 These functional groups do not interfere with base pairing and are therefore compatible with the polymerases used in the amplification steps of the discovery process, called systematic evolution of ligands by exponential enrichment (SELEX), with the greatest success achieved using moieties with hydrophobic aromatic character resembling the side chains of similar amino acids. We have more recently demonstrated the ability to introduce multiple protein-like modifications into random nucleic acid libraries simultaneously on two of four nucleobases, dU and deoxycytidine (dC), which further expands their chemical heterogeneity and allows the selection of ligands with improved properties including higher affinity and specificity and enhanced ligand efficiency, metabolic stability, and epitope coverage.¹⁹ It is useful to note that the sequence solutions to high-affinity binding, that is, sequences of aptamers derived from affinity-enriched pools, are highly dependent on the composition of the starting randomized libraries. For example, all RNA and DNA aptamers to the same targets have different sequences, but so do aptamers derived from libraries with different 2′ modifications (2′-amino, 2-fluoro, or 2′-O-methyl modifications), as well as different 5-position modifications. Even libraries with closely related 5-position modifications, for example, 1-naphthylmethyl and 2-naphthylmethyl modifications, have to date, over the course of more than 7,000 selections, always yielded aptamers with different sequences. This is not the result of stochastic selection of rare sequences since the SELEX process with the same starting libraries generally leads to the identification of the same winning sequences in a reproducible manner. Different sequences, in turn, lead to different folding and three-dimensional structures, which then recognize different epitopes on proteins through exquisite shape complementarity.¹⁶

We were the first to demonstrate the utility of aptamers in neutralizing human viruses,²⁰ and we found that such ligands could target viral epitopes that were unavailable to antibodies and thereby generate broadly neutralizing ligands.^21,22,23 Here, we report the discovery of modified DNA-based aptamers targeting the SARS-CoV-2 spike protein. Using chemically diverse sets of both single (dU) and double-modified (dU/dC) SELEX libraries, we identified high-affinity binding reagents to the various domains of SARS-CoV-2 spike protein including RBD, S1, and S2. Many of the SOMAmers that bind the SARS-CoV-2 spike RBD inhibit binding to the ACE2 receptor and display antiviral activity against authentic virus in a focus-reduction neutralization assay. Remarkably, nearly all top candidates retain high-affinity binding to recombinant spike protein mutants, while our most potent antiviral agent maintains neutralization activity against all clinically relevant VOCs, including Omicron. These findings, combined with the enhanced nuclease resistance resulting from the base modifications, establish these reagents as candidates for further diagnostic and therapeutic development.

Results

SOMAmer reagents with diverse compositions bind to multiple epitopes of the SARS-CoV-2 spike protein

To identify chemically distinct binding reagents to SARS-CoV-2 spike protein, we used the SELEX in vitro evolution method to discover SOMAmer reagents targeting four different monomeric constructs of the spike protein: the S1 domain, the S2 domain, the S1 and S2 ectodomain, and the RBD. We performed separate SELEX experiments with five chemically diverse random DNA libraries containing different 5-position modifications on pyrimidine bases: 1-naphthylmethyl (Nap), 3-indole-2-ethyl (Trp), 4-phenylbenzyl (PBn), or 3,3-diphenylpropyl (DPP) side chains as single modifications on all dU residues within a 40 nucleotide (40N) random region or double modifications comprising a 4-hydroxyphenyl-2-ethyl (Tyr) side chain on all dU residues and a Nap side chain on all dC residues within a 30 nucleotide (30N) random region (Figure 1). We used a shorter random region for the library with two modifications on the basis of our previous observation that high-affinity binding in such libraries is generally encoded in shorter sequences.¹⁹ To obtain reagents with slow off rates, we included a kinetic challenge step in the selection by adding a large molar excess of dextran sulfate following complex formation between the protein target and modified DNA library.¹⁷ Ten rounds of SELEX were completed for each target-library pair, with rounds 9 and 10 being conducted under conditions favoring fast on rates by decreasing the incubation time from 15 min (rounds 2–8) to 10 s (rounds 9–10), before the addition of dextran sulfate, while simultaneously lowering the protein concentrations. Sequencing of the final affinity-enriched pools indicated ample sequence convergence, and approximately 10 high-abundance sequences from each pool were selected for further evaluation. These reagents were synthesized as partially truncated sequence variants (50-mer for libraries with single modification or 40-mer for libraries with double modifications) in which the full random region was retained plus five nucleotides from each of the 5′ and 3′ primer regions. Many high-affinity binders (K_D < 10 nM) were identified from each of the 25 SELEX experiments (Figure 2). Of the 287 total SOMAmer reagents screened for binding, more than one-third of the reagents had K_D <1 nM to their target protein (39 to spike S1/S2 ectodomain, 48 to spike S1, and 17 to spike S2) (Figure 2E). Some of the sequences tested showed no binding activity to their intended target nor did the random library pools, indicating that these high-affinity sequences are specific and that the proteins are not general nucleic acid binders.

Figure 1.

Selection of SARS-CoV-2 spike protein

(A) Spike protein constructs and modified libraries used in SELEX. (B) Structures of modified nucleotides used in SELEX.

Figure 2.

Affinity distributions of partially truncated select SOMAmer reagents arising from selections with five spike protein variants and five modified DNA libraries

SOMAmers were screened for binding their respective SELEX targets: spike S1/S2 ECD (A), spike S1 domain (B), spike S2 domain (C), and spike receptor-binding domain (D) either as 50-mer for single modifications (40N random region plus five nucleotides from both the 5′ and 3′ primer regions) or as 40-mer for dual-modified sequences (30N random region plus five nucleotides from both the 5′ and 3′ primer regions). Equilibrium binding dissociation constants (K_D) were measured in a filter binding assay with radiolabeled SOMAmer. Line in boxplot represent the median of K_D values.

Next, we examined cross-reactivity to the various spike domains. For example, reagents selected against the full spike S1/S2 ectodomain were tested for binding the S1 domain, the RBD, and the S2 domain. In this manner, we gained a better understanding of the binding epitopes for many reagents or determined how the binding affinity of a reagent to a smaller domain (e.g., RBD) is altered in the context of the full S1/S2 ectodomain. Overall, we found that reagents targeting the isolated RBD had the highest affinities (48 below 1 nM, with 16 of those below 100 pM) but observed an approximate 10-fold reduction in affinity when measured against the full spike S1/S2 ectodomain (13 retained affinity below 1 nM). Nonetheless, we identified 87 SOMAmer reagents with affinity to the RBD below 10 nM to screen for potential inhibition of ACE2 receptor binding.

SOMAmer reagents to SARS-CoV-2 spike protein inhibit ACE2 receptor binding

The 87 SOMAmer reagents identified above were assessed for their ability to inhibit the binding of the spike RBD to immobilized ACE2 in a sandwich assay format. Given the high affinity of the SOMAmer reagents to the spike RBD protein, we used low concentrations of the RBD-Fc fusion protein (250 pM), allowing us to theoretically measure IC₅₀ values as low as 125 pM. Among these, 36 reagents displayed a concentration-dependent inhibition of receptor binding with IC₅₀ values of 0.09–20 nM. Many SOMAmer reagents inhibited the spike/ACE2 interaction with equal or greater potency compared with commercial anti-spike S1 monoclonal antibodies. Inhibitory activity of these 36 SOMAmers was also confirmed in a complementary assay where ACE2 binding to the immobilized spike RBD was tested to ensure that any loss of signal, interpreted as inhibition, was not due to SOMAmer interference with antibody detection. This assay gave similar results, thereby affirming our initial conclusions. The remaining 51 sequences showed modest to no inhibitory activity, thus serving as inactive controls. Representative inhibition curves are illustrated in Figure S1.

SOMAmer reagents exhibit potent inhibitory activity against the authentic virus in vitro

To assess the ability of a representative set of SOMAmer reagents to inhibit authentic SARS-CoV-2 virus from infecting Vero cells, we evaluated 92 SOMAmers in a focus-reduction neutralization assay against an isolate representing the first pandemic wave of SARS-CoV-2 (VIC01, Pango clade B). For this screen, we included the 36 reagents that effectively blocked the spike/ACE2 interaction in the sandwich assay, as well as 56 additional reagents that exhibited high-affinity binding to spike S2 and spike S1 outside the RBD or within the RBD but lacking direct ACE2 inhibition. These reagents were chosen to cover a broad array of epitopes on the SARS-CoV-2 spike protein since we aimed to explore all potential mechanisms of inhibition of viral entry including indirect occlusion of receptor binding or interference with the spike protein conformational changes necessary for infection. In addition, we included SOMAmer reagents with each type of chemical modification to increase the likelihood of occupying a broad range of epitopes on the spike protein.

Initial screens were conducted at a single concentration of SOMAmer reagent (10 nM) and were performed in triplicate with a fixed dose of virus (100 foci). The number of focus-forming units in test wells was counted, and the effective neutralization threshold was set at a 20% reduction in plaque count compared with controls (Figure 3A). The single point tests led to the identification of 13 SOMAmer reagents that were clearly in the effective neutralization range, with one reagent, SL1111, having the greatest inhibitory potency (70% reduction in infectivity). Potency was determined by performing a 12-point titration of the 13 neutralizing SOMAmer reagents over a concentration range of 10⁻⁶ to 10⁻¹⁰ M for SL1111 or 10⁻⁵ to 10⁻¹¹ M for the remaining 12 reagents. Most of the SOMAmer reagents that bind the spike RBD were effective at reducing infectivity over the titration range and produced inhibition curves with Hill slopes of −1 (Figures 3B and S2). A subset of SOMAmer reagents had clear biphasic inhibition curves, indicating that there is a population of virus with conformationally available spike epitopes that our reagents can access for neutralization and a subset of less accessible epitopes, likely due to the diversity of receptor conformations on the surface. Reagents targeting spike S2 showed very little neutralization of infectivity. SL1111, which was selected to bind the spike RBD, was confirmed as the most potent inhibitor of viral infection in this assay, with NT₅₀ = 1.2 nM (Figure 3B).

Figure 3.

SOMAmer reagents in a foci neutralization assay

(A) Normalized infectivity data in the presence of 92 SOMAmer reagents tested at a single concentration of 10 nM. Line in the boxplot represents the median value of three replicates. Thirteen reagents having greater than 20% reduction in plaque count (within the green shading) were subject to a full titration. (B–D) Titration of the 13 reagents in the foci neutralization assay showed a range of effects with SL1115 (B) having no neutralizing effect, SL1111 (C) being the most potent, and SL1104 (D) having a biphasic profile. Assay was repeated three times in triplicate, and data were plotted as mean ± SEM in GraphPad Prism and fit to either a biphasic curve or a four-parameter dose-response curve.

High-affinity binding of SOMAmers is encoded in sequences of 28–44 nucleotides

With SL1111 emerging as the most robust inhibitor of viral infectivity, we sought to determine secondary leads that bind non-overlapping epitopes on the spike protein. To that end, we performed competition assays with SL1111 and the other 12 reagents to identify the SOMAmers that do not compete with SL1111 for binding spike. We found that six of the 12 reagents directly competed with SL1111 for binding spike S1/S2 (Figure S3A). All six of these SOMAmers also bind the spike RBD. The remaining six SOMAmers did not compete with SL1111 for binding spike S1/S2 protein (Figure S3B), and of these, we chose SL1107, SL1108, SL1114, and SL1115 to pursue for further optimization. These four reagents were selected based on their affinities, modifications, and epitopes on spike protein; SL1107 and SL1108 target the spike RBD outside the SL1111-binding epitope, while SL1114 and SL1115 target spike S2.

As a first means of optimization, we performed truncation studies with the five lead reagents to identify the shortest sequences required for high-affinity binding to spike S1/S2. For each reagent, a series of sequences were synthesized, sequentially removing nucleotides from the 5′ and 3′ ends, and the binding affinities for each were measured in a filter-binding assay. We found that the three sequences with dual modifications were highly amenable to truncation and resulted in final high-affinity binding reagents that were 28 (SL1114_18 and SL1115_18) and 29 nucleotides in length (SL1111_20). The sequences with single modifications could be truncated to somewhat longer lengths comprising 35 (SL1108_18) and 44 (SL1107_18) nucleotides (Table 1; Figure S4).

Table 1.

K_D values and affinity ratios for truncated SOMAmer reagents binding recombinant WT spike S1/S2 monomeric protein, spike S1/S2 stable trimer protein, and spike S1/S2 stable trimer variants

Reagent	SL1111_20	SL1107_18	SL1114_18	SL1108_18	SL1115_18
Domain targeted	RBD	RBD	S2	RBD	S2
Length (nt)	29	44	28	35	28
KD (nM) spike monomer	0.4	0.6	0.6	0.4	0.4
KD (nM) spike stable trimer	0.5	1	1.8	0.6	no binding
Variant	Ratio of binding: variant/WT monomer | variant/stable trimer
Alpha	3 | 3	2.5 | 1.5	7.5 | 2.5	3.5 | 2.5	no binding
Beta	0.3 | 0.5	0.4 | 0.3	1.3 | 0.8	2.0 | 0.9	no binding
Gamma	0.6 | 0.9	0.8 | 11	2.5 | 3.7	1.3 | 1.6	no binding
Delta	1.1 | 0.2	4.7 | 0.5	11 | 0.4	2.3 | 1.4	no binding
Omicron BA.1	0.1 | 0.1	2.8 | 1.7	2.0 | 0.7	1.8 | 1.2	no binding

nt, nucleotide; WT, wild type; K_D, equilibrium binding dissociation constants.

Truncated SOMAmer reagents bind to mutated spike protein and inhibit VOCs in vitro

We next sought to determine whether our five truncated lead reagents maintain high-affinity binding to recombinant spike protein monomer and stabilized trimer containing the mutations found in the five VOCs. The current list of five major VOCs and the mutations contained within the spike protein are shown in Table S1, and a schematic of the spike protein domains and variant mutations is shown in Figure 4. Remarkably, we observed no significant loss in binding affinity with four of our five SOMAmer reagents for binding all five VOC spike trimers compared with unmutated spike monomer and trimer proteins. These reagents include our lead, SL1111_20 and the other two RBD-binding reagents, SL1107_18 and SL1108_18, as well as the S2-targeting reagent SL1114_18. Conversely, the SL1115_18 reagent had no measurable binding affinity for recombinant spike trimer up to a 50 nM protein concentration. The observed lack of binding affinity could be due to occlusion of the binding epitope upon spike trimerization or the presence of multiple stabilizing mutations inserted into the recombinant trimer proteins, many of which are located in the S2 domain.²⁴

Figure 4.

Spike protein domain map and location of mutations found in the five variants of concern

NTD, N-terminal domain; RBD, receptor-binding domain; RBM, receptor-binding motif; FC, furin cleavage site; FP, fusion peptide; HR1 and 2, heptad repeat.

SL1111_20 binds to a conserved site of RBD distinct from antibody epitopes and potently neutralizes Omicron BA.1

As the substitutions found in the spike protein of Omicron result in a substantial escape from neutralizing antibodies generated by immune responses to earlier variants of SARS-CoV-2,²⁵ the ability of our lead neutralizing aptamer to bind to the RBD of Omicron BA.1 was very encouraging. To understand the basis for this cross-reactivity, we tested whether its binding site overlapped the known epitopes of a panel of structurally well-characterized antibodies using competition ELISA on both trimeric full-length spike (HexaPro spike) and the RBD. The results (Figures 5A and 5B) show that SL1111_20 competes for binding with antibodies that bind to various epitopes (based on Barnes classification²⁶ on the RBD including class 1 [FI-3A, C105], 1/4 [COVA1-16, C118, S2X-259], 3 [FD-11A, REGN10987], and 4 [CR3022, VHH72]) but no competition was seen with a class 2 antibody (C121). A similar pattern of blocking was observed on both the RBD and trimeric full-length spike.

Figure 5.

SL1111_20 epitope mapping with structurally characterized antibodies

(A and B) Competition ELISA results show that SL1111_20 substantially blocks binding to trimeric spike (A) and RBD (B) for a subset of several classes of antibodies whose binding sites have been mapped precisely by cryelectron microscopy (cryoEM) or X-ray crystallography. Data are presented as mean ± SEM. (C and D) Residues of binding sites determined for antibodies that are not competed by aptamer SL1111_20 are shown in green on a representative structure of RBD (PDB: 7T9L). Residues that are in bindings sites of antibodies that are competed by aptamer but not in the non-competing sites above are shown in red. Ambiguous residues showing partial competition are shown in amber. The van der Waals surface of ACE2 in complex with the RBD is shown in translucent cream. Antibody-binding residues on the RBD were determined using the PISA interface tool at the PDB in Europe from published structures and from those currently under peer review (HSW-1 and HSW-2) (Fan et al.³⁷). Public PDB structures used for determining RBD epitopes were PDB: VHH72 (6WAQ), CR3022 (7LOP), REGN10987 (6XDG), EY6A (6ZER), S309 (7TN0), C121 (7K8Y), REGN10933 (6XDG), FI-3A (7Q0G), FD-11A (7PQZ), FD-5D (7PR0), S2X-259 (7M7W), C105 (6XCM or 6XCN), COVA1-16 (7JMW), C118 (7RKS), and C022 (7RKU).

We identified the epitope for SL1111_20 by mapping the binding sites for antibodies that compete with the aptamer for RBD binding (red in Figures 5C and 5D), followed by subtracting all residues that overlap with binding sites of the non-competing antibodies (green in Figures 5C and 5D). Based on these analyses, SL1111_20 appears primarily to disrupt binding at a distinct region to the “rear” of the RBD, centered on Thr-500 (residues 499–505 plus 437–439, 404, 405, 407, 443, and 448, which are all clustered together in the folded RBD (Figures 5C and 5D). This region overlaps with part of the ACE2-binding site, potentially accounting for the neutralizing potency of SL1111_20. Residue 420, which lies exclusively in an epitope occupied by competing antibodies, is surrounded by epitopes occupied by non-competing antibodies, so it is unlikely to be in an aptamer-binding site. This notion is supported by the observation that competing antibodies, C105 and FI-3A, have additional contact sites that lie in the principal aptamer-binding region described above (Figure S5).

To complement these antibody competition experiments and gain additional insight regarding the potential binding epitope of SL1111_20 on the spike RBD, we performed hydrogen-deuterium exchange mass spectrometry (HDX-MS). A total of 53 peptides encompassing 77% of the primary sequence of the spike RBD were identified. Two regions could not be identified, amino acids 319–351 and 363–374, possibly due to glycosylation. When the spike RBD/SL1111_20 complex was analyzed, five peptides showed a significant change in deuteration compared with the free spike RBD (Figure S5), indicating they are either directly protected or consequentially affected by the binding of SL1111_20. Two of the five peptides contain the primary residues affected by antibody binding, as described above (residues 499–505, 443, and 448) (Figure S5).

Neutralization of VOCs by SL1111_20

To test whether this conservation of binding to RBD variants, albeit distinct from that of known antibodies, was reflected in conserved antiviral effectiveness, we undertook neutralization assays with live Gamma, Delta, and Omicron BA.1 VOCs. The results (Figure 6) showed that SL1111_20 neutralized all three VOCs with equal or better potency than prototype clade B VIC01. The epitope of SL1111_20 is strongly conserved between VOCs, with the exception of N501Y and Y505H, which are reported to enhance ACE2-binding affinity and clearly do not disrupt SL1111_20 RBD interaction.

Figure 6.

Neutralization of VOCs by SL1111_20

SL1111_20 potently neutralizes prototype authentic SARS-CoV-2 (B Victoria, black), Delta (blue), Gamma (purple), and Omicron (red) VOCs. SL1111_20 shows enhanced neutralization activity against Omicron, Gamma, and Delta variants compared with the prototype clade B VIC01. Assay performed in quadruplicate and data were plotted as mean ± SEM in GraphPad Prism and fit to a log(agonist) vs. response curve.

Inhibitory SOMAmer reagents exhibit a substantial degree of nuclease resistance in serum

The strong neutralization potency of SL1111_20 against all tested VOCs of SARS-CoV-2 makes it a highly attractive candidate for further development. To determine the metabolic stability of our nucleic acid-based reagents to nucleases present in human biological fluids, we measured the in vitro half-life of SL1111_20 and the other lead candidates in pooled human serum and compared them with unmodified DNA analogs in which NapdU is replaced with dT (SL1107_18 and SL1108_18) and TyrdU and NapdC are replaced with dT and dC, respectively (SL1111_20, SL1114_18, SL1115_18). Purified SOMAmers were incubated with 90% pooled human serum at 37°C, and samples were removed for analysis at various time points up to 96 h (Figure 7). All reagents tested contained 3′-inverted dT to impede 3′-5′ exonuclease activity. Samples were analyzed by denaturing PAGE, and the percentage of intact aptamer was plotted as a function of time and fit to a one-phase exponential decay model to determine half-life. We found that the two longest sequences, SL1108_18 (35-mer) and SL1107_18 (44-mer), which contain only NapdU modifications, had the shortest half-lives of 34 and 75 h, respectively. The half-lives for the sequences with two modifications were longer and more varied, ranging from 99 h for SL1115_18, to 161 h for SL1111_20, to 317 h for SL1114_18. The presence of distinct degradation bands with all lead reagents indicates that certain nucleotide positions are more sensitive to nuclease activity, suggesting that half-lives could be further improved with supplementary nuclease-stabilizing modifications such as 2′-O-methyl substitutions at deoxyribose and phosphorothioate backbone modifications. Not surprisingly, we found the unmodified analogs of our lead reagents were quickly degraded in human serum (half-life of 6–30 h). These results are consistent with prior observations that shorter, heavily modified sequences tend to have the longest half-lives, while unmodified DNA is rapidly metabolized.²⁷

Figure 7.

Serum stability of SARS-CoV-2 spike reagents and unmodified analogs

(A) Modified lead SOMAmers and their corresponding dT (SL1108_28 and SL1107_26) and dT/dC (SL1114_36, SL1115_26, and SL1111_354) analogs were incubated in 90% pooled human serum, and sample was removed at various time points (3–96 h) and analyzed via denaturing PAGE. Sample time points from (A) were plotted as the percentage of intact reagent and fit to a one-phase exponential decay model to determine half-life for duel-modified sequences and their dT/dC analogs (B) and single-modified sequences and their dT analogs (C).

Discussion

Aptamers are a class of nucleic acid-based affinity reagents with several advantages over protein-based reagents including high thermal and chemical stability, the ability to rapidly refold after denaturation, compatibility with fully synthetic manufacturing, and low immunogenicity.²⁸ These properties make aptamers attractive candidates for development of antigen-based diagnostics and parenteral therapeutics. We have previously established that the incorporation of diversity-enhancing side chains at the 5 position of one or more pyrimidine bases substantially increases the efficiency with which high-affinity aptamers can be identified to a wide range of protein targets.¹⁷ For this reason, we used five different 5-position-modified starting libraries, paired with five different target protein constructs, to identify a group of modified aptamers with broad epitope coverage of the surface of the SARS-CoV-2 spike protein. All 25 selections yielded affinity-enriched pools from which we were able to identify high-affinity ligands with K_D values of <10 nM. This resulted in a diverse collection of sequences from which we could screen for molecules with desired functional properties. Although we have focused our study on antiviral agents, it is worth noting that the abundance of high-affinity reagents described here, including the identification of non-competing SOMAmer pairs, does identify candidate reagents that may be suitable for development of sandwich-based diagnostic tests for SARS-CoV-2.

The primary goal of this study was to identify sequences with inhibitory activity against SARS-CoV-2. Thirteen SOMAmer reagents exhibited antiviral activity, and through competition binding assays with the most potent inhibitor, SL1111, (Figure S2), we determined that six of these reagents could bind simultaneously to the spike protein, which establishes that they occupy separate epitopes (Figure S3). Of these six non-competitive inhibitors, we fully truncated four SOMAmers in addition to SL1111 and demonstrated that these reagents with 5-position modifications exhibit substantial resistance to nuclease degradation compared with unmodified DNA even in the absence of 2′-position modifications.

One of the most important findings of this work is that at least four out of five of the minimal SOMAmers retain high-affinity binding to the most clinically relevant VOCs: Alpha, Beta, Gamma, Delta, and Omicron BA.1. For the most potent inhibitor, SL1111_20, no loss in neutralization was observed with any of the five VOCs. These results are highly significant, as the evolutionary pressure of the virus to evade detection by the immune system will continue to result in the emergence of new variants. Such pressure does not exist for modified DNA reagents with inhibitory activity. Through epitope mapping and HDX-MS experiments, we identified a potential conserved binding surface for SL1111_20 on spike, overlapping with its receptor-binding motif (RBM), thus explaining the broad neutralization results. Taken together, these SOMAmer reagents represent promising candidates for the development of parenteral therapeutics for the treatment of COVID-19. The next steps in the development of therapeutics would require, inter alia, the development of an optimized and suitably formulated version of SL1111_20 and in vivo testing of such a candidate to determine pharmacokinetics and efficacy in models of SARS-CoV-2 infection following intravenous, intramuscular, subcutaneous, or inhalation delivery. Unique characteristics of SOMAmers (for example, relatively small size, low immunogenicity, compatibility with a wide range of excipients, fully synthetic manufacturing, etc.) could offer certain advantages over antibodies.

The approach described here might provide a generally applicable method for addressing future pandemics. Interaction surfaces engaged by aptamers can overlap with those engaged by antibodies, but because aptamers are not involved in the immune response to virus, there is no evolutionary pressure for the virus to mutate these binding sites. Furthermore, the development of SOMAmer antagonists of viruses with pandemic potential ahead of an outbreak could be rapidly deployed to help blunt the first phase of a future pandemic and allow time for traditional vaccine and antiviral approaches to develop.

Materials and methods

Selection of SOMAmer reagents

SOMAmer reagents targeting recombinant spike proteins, spike S1 and S2 extracellular domain (ECD) (SinoBiological, cat #40589-V08B1); spike S1 subunit SARS-CoV-1 (SinoBiological, cat #40150-V08B1); spike S2 ECD (SinoBiological cat #40590-V08B); spike S1 RBD (Creative Biomart, cat #Spike-190V); and spike S1 subunit SARS-CoV-2 (Creative Biomart, cat #Spike-191V) were discovered via the SELEX process^17,29 from libraries containing either a 40-nucleotide random region in which dT was substituted with TrpdU, NapdU, PBndU, or DPPdU or a 30-nucleotide random region in which dT was substituted for TyrdU and dC was substituted for NapdC. The 40-nucleotide random region was flanked by a 20-nucleotide forward primer (5′-GGTCGGGCACACTACGCATC-3') and a 21-nucleotide reverse primer (5′-GGGAAGAGAAAGGAGAAGAAG-3') while the 30-nucleotide library random region was flanked by a 20-nucleotide forward primer (5′-GGTCGGGCACACTACGCATC-3') and a 21-nucleotide reverse primer (5′-GGGAAGAGAAAGGAGAAGAAG-3'). The total length of each SELEX library was either 89 (40N library) or 79 (30N library) nucleotides including an eight-nucleotide poly-dA region on the 3′ end. SELEX was performed in 1XSB18T buffer (40 mM HEPES [pH 7.5], 102 mM NaCl, 5 mM KCl, 5 mM MgCl₂, 0.05% Tween 20). To preferentially select for modified aptamers with slow off rates, a kinetic challenge was included whereby protein/DNA complexes were incubated at 37°C in the presence of the polyanionic competitor dextran sulfate. The duration of the kinetic challenge was increased from 30 s in round 2 to 15 min in rounds 3–4, 30 min in rounds 5–6, and 45 min in rounds 7–8. Simultaneous to the polyanionic competitor challenge, the protein concentrations were lowered. SELEX was executed as follows. SELEX libraries were heated/cooled at 95°C for 5 min, and then cooled to 37°C at 0.1°C/s. Following heating/cooling, libraries were incubated at 37°C for 10 min with Protein Competitor Buffer (10 μM prothrombin, 10 μM casein, 0.01% human serum albumin) and 25 μg Hexa-His-bound His-tag Dyna beads (Invitrogen, cat #101-04D) for counter selection. After 10 min, the supernatant was removed and transferred to a clean plate. For round one of SELEX only, 50 pmoles of protein was immobilized 125 μg His-tag Dyna beads and transferred to the counter selected library and incubated at 37°C for 1 h with shaking. For rounds 2–8, proteins were in solution during library incubation at 37°C for 10 min with no shaking. Prior to capturing protein-DNA complexes with His-tag Dyna beads for rounds 2–8, the kinetic challenge was initiated with 5 mM dextran sulfate (final concentration). Beads were then washed five times in 1XSB18T buffer. Elution was achieved with 2 mM NaOH and neutralized with HCl and buffered to pH 7.5 with Tris (round one) or perchlorate elution buffer (40 mM PIPES [pH 6.8], 1 mM EDTA, 0.05% Triton) followed by qPCR for 25 cycles or until samples plateaued. After qPCR, DNA was captured on Dynabeads MyOne streptavidin beads, and the sense strand was eluted with 20 mM NaOH and discarded. The modified nucleotide sense strand was prepared with the appropriate nucleotide composition by primer extension from the immobilized antisense strand. After eight rounds of SELEX, the converged pools were sequenced. As some intended uses of spike protein SOMAmer reagents may require fast on rates, two additional rounds of SELEX were performed subsequent to round eight, as described above but with the following exceptions. The target proteins were incubated with the SELEX libraries for 10 s before the addition of polyanionic competitor dextran sulfate. Additionally, all protein concentrations were reduced 0.5 log from their round eight concentrations. Round 10 pools were sequenced and compared with the round eight sequencing results.

Modified aptamer synthesis

The modified deoxyuridine-5-carboxamide phosphoramidite reagents used for solid-phase synthesis were prepared by condensation of 5′-O-(4,4′-dimethoxytrityl)-5-trifluoroethoxycarbonyl-2′-deoxyuridine with the appropriate primary amine; 3′-O-phosphitylation with 2-cyanoethyl-N,N,N′,N′-tetraisopropylphosphene; and purification by flash chromatography on neutral silica gel.^30,31

The modified deoxycytidine-5-carboxamide phosphoramidite reagents used for solid-phase synthesis were prepared using a four-step synthetic strategy from 5′-OH-5-iodo-2′-deoxycytidine condensed with the appropriate primary amine; N-protection of the 3-amino position of cytidine; O-protection of the 5′ alcohol with 5,5′-dimethowxytrityl; 3′-O-phosphitylation with 2-cyanoethyl-N,N,N′,N′-tetraisopropylphosphene; and purification by flash chromatography on neutral silica gel.^30,32

Modified aptamers were produced by conventional solid phase oligonucleotide synthesis using the phosphoramidite method³³ The aptamers were synthesized at the 50-nmol scale in a plate-based system on a Mermade 192X DNA synthesizer with some adjustments to the protocol to account for the unique base modifications contained therein. Detritylation was accomplished with 10% dichloroacetic acid in toluene; coupling was achieved with 0.1 M phosphoramidites in straight acetonitrile or a mix of acetonitrile:dichloromethane activated by 5-benzylmercaptotetrazole and was allowed to react three times; capping and oxidation were performed according to instrument vendor recommendations. The resulting oligonucleotides were deprotected and cleaved from the controlled pore glass (CPG) by placing the plate containing synthesis columns in a reactor and incubating them with gaseous methylamine using an optimized time, pressure, and temperature. Deprotection by-products were removed by washing the columns containing the cleaved aptamers and spent CPG with high organic washes drawn through the bed via a vacuum manifold, then dried thoroughly. The desired product was next eluted from the CPG using high-purity water via the vacuum manifold and collected into Matrix tubes. The resulting products underwent no additional purification and were characterized by liquid chromatography MS (LCMS) using an Agilent Technologies Ultra Performance Liquid Chromatograph (1290 Infinity II) fitted with a single quadrupole mass spectrometer (Agilent Technologies 6130) and protein binding affinity in buffered aqueous solution.

Characterization of modified aptamers

Equilibrium dissociation constants (K_D) of synthetic 50-mer modified aptamers (40 nucleotides from the SELEX library modified region and five nucleotides from the 5′ and 3′ primer regions for 40N libraries) and 40-mer modified aptamers (30 nucleotides from the SELEX library modified region and five nucleotides from the 5′ and 3′ primer regions for 30N libraries) were determined by filter binding assay. K_D values of modified aptamers were measured in SB18T buffer. Modified aptamers were 5′ end labeled using T4 polynucleotide kinase (New England Biolabs) and γ-32PATP (PerkinElmer). Radiolabeled aptamers (∼20,000 CPM) were mixed with spike proteins at concentrations ranging from 10⁻⁷ to 10⁻¹² M and incubated at 37°C for 40 min. Following incubation, an equal volume of 10 mM dextran sulfate was added, and bound complexes were partitioned on His-tag Dynabeads at 37°C shaking at 1,850 RPM for 5 min and then captured on Durapore filter plates (EMD Millipore). The fraction of bound aptamer was quantified with a phosphorimager (Typhoon, GE Healthcare, Piscataway, NJ, USA), and data were analyzed in ImageQuant (GE Healthcare). To determine binding affinity, data were fit using the equation

$$\mathrm{y}=(\max -\min )\left(\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{t}\mathrm{e}\mathrm{i}\mathrm{n}\right)/({\mathrm{K}}_{\mathrm{D}}+\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{t}\mathrm{e}\mathrm{i}\mathrm{n})+\min$$

and plotted using GraphPad Prism v.7.00.

Competition binding

Modified aptamer SL1111 was 5′ end labeled using T4 polynucleotide kinase (New England Biolabs) and γ-32PATP (PerkinElmer). Competition assays were performed by incubating radiolabeled SL1111 (20 nM) and cold competitor SOMAmer reagents at concentrations ranging from 10⁻⁵ to 10⁻¹⁰ M with recombinant monomeric spike S1/S2 protein (Sino Biological) (10 nM) in SB18T buffer at 37°C for 60 min. Following incubation, an equal volume of 10 mM dextran sulfate was added, and bound complexes were partitioned on His-tag Dyna beads at 37°C shaking at 1,850 RPM for 5 min and then captured on Durapore filter plates (EMD Millipore). The fraction of SL1111 bound was quantified with a PhosphorImager (Typhoon, GE Healthcare), and data were analyzed in ImageQuant (GE Healthcare). Equilibrium dissociation constants (K_i) for the competitors were determined by non-linear regression analysis using the equation

$$\log {EC}_{50}=\log \left[{10}^{logKi}\ast \left(1+\frac{radioligand\left(nM\right)}{hotKd\left(nM\right)}\right)\right]$$

(GraphPad Prism).

ACE2:SARS-CoV-2 spike inhibition screening assay

Spike RBD-modified aptamers were tested in an ACE2:SARS-CoV-2 spike inhibition assay (BPS Bioscience, cat #79936) following the manufacturer’s protocol with a few exceptions. Briefly, for each modified aptamer to be tested, ACE2-His was diluted to 1 μg/mL, and 50 μL was added to 12 wells of the provided Ni-coated 96-well microplate. Modified aptamers were tested at protein concentrations ranging from 10⁻⁷ to 10⁻¹² M and added to the ACE2-containing wells prior to the addition of 0.25 nM spike RBD mFC tag protein. Secondary HRP-labeled antibody was diluted 1:1,000 in the provided blocking buffer and added to every well. The provided ELISA enhanced chemiluminescence (ECL) substrates were mixed 1:1 and added to every well just prior to reading the plate in chemiluminescence mode on a Spectramax (Molecular Devices). The buffer used in the assay was 1XSB18T 40 mM HEPES (pH 7.5), 102 mM NaCl, 5 mM KCl, and 0.05% Tween 20 with the addition of 2.5 mM MgCl₂. All incubations were done at room temperature for 1 h with slow shaking. After each incubation, the plate was washed 3× with the provided 1X Immuno Buffer and blocked for 10 min with the provided blocking buffer and then washed 3× again. All volumes of reagents corresponded to those recommended by the manufacturer. Data were analyzed as follows: the reading from the blank well (ACE2 only well) was subtracted from all data points. The percentage of activity for each modified aptamer tested was determined by calculating the fraction of the positive control signal (ACE2 plus spike RBD). Data were plotted in GraphPad Prism and fit to a four-parameter dose-response curve.

Virus isolates

Prototype isolate (Pango lineage B) was Victoria/01/2020, received at P3 from Public Health England (PHE) Porton Down (after being supplied by The Peter Doherty Institute for Infection and Immunity, University of Melbourne) in April 2020, passaged in VeroE6/TMPRSS2 cells, used here at P5, and confirmed identical to GenBank MT007544.1, B hCoV-19_Australia_VIC01_2020_ EPI_ ISL_ 406844_ 2020-01-25.

Beta (B.1.351, 20I/501.V2.HV001) isolate³⁶ was received at P3 from the Center for the AIDS Program of Research in South Africa (CAPRISA), Durban, in Oxford in January 2021, passaged in VeroE6/TMPRSS2 cells (NIBSC reference 100978), used here at P4.

Delta (B.1.617.2) isolate 83DJ-1 and Omicron BA.1 (B.1.1.529) isolate TBC were kindly provided by Piet Maes, Laboratory of Clinical and Epidemiological Virology (Rega Institute), KU Leuven, 3000 Leuven, Belgium.

Microneutralization assay (MNA)

The study was performed in the containment level three facility of the University of Oxford, operating under a license from the Health and Safety Authority, UK, on the basis of an agreed code of practice, risk assessments (under the Advisory Committee on Dangerous Pathogens guidance), and standard operating procedures. The MNA determines the concentration of aptamer that produces a 50% reduction in infectious focus-forming units (FFUs) of authentic SARS-CoV-2 in Vero CCL81 cells. Quadruplicate serial dilutions of serum, or aptamer (20 μL), were pre-incubated with 100–200 FFUs (20 μL) SARS-CoV-2 for 30 min at room temperature. After pre-incubation, 100 μL Vero CCL81 cells (4.5 × 10⁴) were added and incubated at 37°C, 5% CO₂. After 2 h, 100 μL of a 1.5% carboxymethyl cellulose-containing overlay was applied to prevent satellite focus formation. Eighteen hours post-infection, the monolayers were fixed with 4% paraformaldehyde, permeabilized with 2% Triton X-100, and stained for the nucleocapsid (N) antigen or spike antigen using monoclonal antibodies (mAbs) EY-2A and EY-6A, respectively.³⁴ After development with a peroxidase-conjugated antibody and TrueBlue peroxidase substrate, infectious foci were enumerated by ELISpot reader. Data were analyzed using four-parameter logistic regression (Hill equation) in GraphPad Prism 8.3.

Competition ELISA

The competition ELISA used to determine the binding epitope of the aptamer was done as previously described with a slight modification.^34,35 Briefly, 50 ng affinity-purified HexaPro Spike (Wuhan) protein or 25 ng RBD (displayed on the mi3 nanoparticle [RBD-mi3])^34,36 diluted in 1× phosphate-buffered saline (PBS) were coated on 96-well Thermo MaxiSorp plates for 2 h at room temperature (RT), washed with 1× PBS, and blocked with 300 μL 5% (w/v) dried skim milk in 2× PBS overnight at 4°C. The competing mAbs (diluted to 5 μg/mL in 1× PBS/0.1% BSA) in quadruplicates were mixed with SL1111_20 (200-fold molar excess of aptamer was used for HexaPro spike, 2,000-fold molar excess of aptamer was used for RBD-mi3) in a round-bottomed 96-well plate and transferred to the blocked and washed plates coated with HexaPro spike or RBD-mi3. Three wells containing the respective antibody only (no aptamer) were included to obtain the maximum binding of each of the mAbs. Wells containing aptamer was included to obtain the minimum binding (assay background). The PBD accession codes for the mAbs used in the assays are wwPDB: VHH72 (6WAQ), CR3022 (7LOP), REGN10987 (6XDG), EY6A (6ZER), S309 (7TN0), C121 (7K8Y), REGN10933(6XDG), FI-3A (7Q0G), FD-11A (7PQZ), FD-5D (7PR0), S2X-259 (7M7W), C105 (6XCM or 6XCN), COVA1-16 (7JMW), C118 (7RKS), and C022 (7RKU). The paper describing the structures of HSW-1 and HSW-2 is currently under peer review.³⁷ The plates were incubated for 1 h at RT. Plates were then washed, 50 μL secondary antibodies (rabbit anti-human, HRP conjugated, Dako P021402-2) diluted 1:1,600 in 1× PBS/0.1% BSA were added, and the plates were incubated for another hour at RT. Plates were then washed with 1× PBS and developed by adding POD substrate (11484281001, Roche) for 5 min before stopping the reaction with 1 M H₂SO₄. Absorbance OD₄₅₀ was measured using a Clariostar plate reader (BMG Labtech). Competition was measured using the formula below.

$$\%ofblocking=\frac{X-meanminimumbinding}{meanmaximumbinding-meanminumumbinding}$$

$$X=OD450ofthecompetingmAbinthepresenceofaptamers$$

HDX-MS

To identify peptic fragments of spike RDB protein, 154 pmol protein (2 μL) were mixed with 48 μL PBS (pH 7.2) and then mixed with 50 μL of a quenching buffer (6.67 M urea, 0.2 M TCEP, pH 3). The mixture (a total of 100 μL) was incubated at 6°C for 3 min and injected into a Waters HDX-LC box (Waters) in which protein was digested by an online pepsin column (Waters Enzymate BEH pepsin column, 5 μm, 2.1 × 30 mm), and the resulting peptic peptides were trapped and desalted on a Waters ACQUITY UPLC BEH C4 1.7 μm VanGuard Pre-column (2.1 × 5 mm) at 100 μL/min buffer A (0.1% formic acid in water) for 3 min. The digestion chamber was kept at 15°C, and the trap and analytical columns were at 0°C. The peptides were eluted with 3%–33% buffer B (0.1% formic acid in acetonitrile) between 0 and 6 min, 33%–40% B between 6 and 6.5 min, and 40%–85% B between 6.5 and 7 min. The eluted peptides were resolved a Waters ACQUITY UPLC Protein BEH C4 column (300 Å, 1.7 μm, 1 × 50 mm). Tandem MS (MS/MS) spectra were performed on a Thermo LTQ orbitrap Velos mass spectrometer. The peptides were ionized using electrospray ionization (ESI) with the source voltage = 4.5 kV and S-lens RF level = 60%. The capillary temperature was 275°C, and the source heater was at 80°C. The sheath gas flow was 10. Precursor ions were scanned between 350 and 1,800 m/z at 60,000 resolution with AGC 1 × 10⁶ (max ion fill time = 500 ms). From the precursor scan, the top 10 most intense ions were selected for MS/MS with 180-s dynamic exclusion (10-ppm exclusion window, repeat count = 1) and AGC 1 × 10⁴ (max ion fill time = 100 ms). Ions with unassigned charge states were rejected for MS/MS. The normalized collision energy was 35%, with activation Q = 0.25 for 10 ms.

MS/MS spectra were searched against a database consisting of the spike RBD protein sequence in a FASTA format using the MaxQuant/Andromeda program (v.1.6.3.4) developed by the Cox Lab at the Max Planck Institute of Biochemistry^38,39 The digestion mode was “unspecific,” and oxidation of methionine was set as a variable modification. The minimum peptide length for the unspecific search was 5. MaxQuant/Andromeda used 4.5 ppm for the main search peptide tolerance and 0.5 Da for MS/MS tolerance. The false discovery rate was 0.01. A list of peptides identified from the search was used as an exclusion list for the next round of LC-MS/MS, and a total of three LC-MS/MS rounds were performed for each protein.

HDX

500 μL PBS was dried using vacuum centrifugation and reconstituted with an equal volume of deuterium oxide (“D₂O buffer”). Spike protein (15 μL) was mixed with the same volume of either the compound (59 μM in water) or water and pre-incubated on ice for at least 1 h to ensure the complex formation. Ten minutes prior to the initiation of HDX reaction, 5 μL of the pre-incubated sample was transferred to a 0.5-mL tube and incubated at room temperature. HDX reaction was initiated by the addition of the D₂O buffer (45 μL) to the 5 μL sample (90% D₂O final). The reaction was quenched at 1 and 10 min after the initiation by the addition of the quench buffer (50 μL). The quenched sample was incubated at 6°C for 3 min, and the whole sample (100 μL) was injected into the Waters HDX-LC box and LC-MS was performed as described in the above section. For HDX samples, only MS1 spectra were recorded.

Data analysis

HDX-MS data were analyzed using Mass Spec Studio (v.2.4.0.3484) developed by the Schriemer lab at the University of Calgary^40,41 “Peptide.txt” and “evidence.txt” from the MaxQuant/Andromeda search results were used to generate the “identification” table for Mass Spec Studio using an in-house Python script. The raw files from the Thermo LTQ orbitrap Velos were converted to mzML files using ProteoWizard (v.3.0.20216, 64 bit). The default processing parameters were used except mass tolerance = 15 ppm, total retention time width = 0.15 min, XIC smoothing = Savitzky Golay Smoothing, and deconvolution method = centroid. All processed data were manually validated.

Serum stability assay

Serum stability studies were performed in 90% pooled human sera (10% PBS) using 500 nM aptamer, and samples were processed as described by Gupta et al.²⁷ Briefly, aliquots were extracted with phenol-chloroform and concentrated using a YM-10 molecular weight cutoff filter (EMD Millipore). Digestion products for all studies were separated from full-length aptamer by PAGE using a 15% polyacrylamide gel containing 8 M urea. Electrophoresis, using a Tris borate buffer system, was performed for 20 min at 200 V. To quantify bands, gels were stained with 2 μM SYBR Gold nucleic acid stain (Molecular Probes) for 10 min. Images of stained aptamers were obtained using a Typhoon 9500 (GE Healthcare) and quantified with the ImageQuant TL software (with background subtraction). The fraction of intact aptamer was plotted as a function of time and fit to a one-phase exponential decay model to determine half-life. All aptamers used in these studies were purified via HPLC or gel purification and had a 3′-3′-linked dT cap and a 5′-hydroxyl group.

Data availability

All data generated during this study are included in this published article and its supplemental information or are available upon request.