Protein Labeling and Crosslinking by Covalent Aptamers

Abstract

We developed a new approach to selectively modify native proteins in their biological environment using electrophilic covalent aptamers. These aptamers are generated through introduction of a proximity-driven electrophile at specific nucleotide sites. Using thrombin as a proof-of-concept, we demonstrate that covalent aptamers can selectively transfer a variety of functional handles and/or irreversibly crosslink to the target protein. This approach offers broad programmability and high target specificity. Furthermore, it addresses issues common to aptamers such as instability towards endogenous nucleases and residence times during target engagement. Covalent aptamers are new tools that enable specific protein modification and sensitive protein detection. Moreover, they provide prolonged, nuclease-resistant enzyme inhibition.

Graphical Abstract

Nucleic acid aptamers were developed that are capable of covalently labeling their protein target with a functional molecule. Additionally, with little modification, the aptamer covalently crosslinked to its target instead, thereby inhibiting its function. Covalent aptamers offer a new tool for selective protein modification, detection, and inhibition.

Covalent modification of proteins via bioorthogonal reactions is essential for many biological investigations and the development of protein-based therapeutics. Common approaches to generate protein conjugates include metabolic labeling,^[1] activity-based labeling,^[2] fusion of the protein of interest to an enzyme or peptide tag,^[3] and unnatural amino acid incorporation.^[4] Specific covalent modification of native proteins in their biological environment has been challenging. One approach involves the modification of small molecule inhibitors and ligands with electrophilic “warheads”. In contrast to traditional, reversible protein-ligand interactions, a covalent ligand is designed such that the reversible association is followed by proximity-driven irreversible crosslinking between a strategically placed electrophile on the ligand and a nucleophilic residue on the target protein.^[5] Small molecule covalent ligands offer several advantages over their non-covalent counterparts, including the ability to outcompete endogenous ligands with similar affinities, dosage at lower concentration, and a duration of action that matches the protein turnover rate. Prominent examples include ibrutinib and afatinib, FDA-approved inhibitors that covalently crosslink to Bruton’s tyrosine kinase and epidermal growth factor receptor, respectively.^[6] In a complementary approach, chemoselective and site-selective proximity-driven electrophiles were utilized to transfer a functional label to a target protein.^[7] Electrophiles were also escorted to a nucleophilic residue by a protein binding ligand – similar to covalent inhibitors, but with ligand release after label transfer.^[8] This approach enables specific modification of native proteins and was utilized, e.g., to covalently modify carbonic anhydrase with a ¹⁹F NMR probe to measure inhibitor binding in live cells.^[9]

In order to expand the scope of this exciting methodology beyond small molecule ligands and their corresponding targets,^[10] we selected aptamers as the protein-binding modality. Aptamers are short, single-stranded nucleic acid molecules that are generated through powerful selection strategies and bind targets with exquisite affinity and specificity.^[11] They are versatile nucleic acid tools with numerous applications as biological probes, as well as therapeutic, diagnostic, and drug delivery agents.^[12] However, aptamers are not catalytically active, in contrast to nucleic acid enzymes that have been shown to make and break bonds.^[13]

Aptamers’ broad scope in protein binding complements the functionality provided by transferable synthetic motifs in a powerful fashion. Aptamers interact with their protein targets over a large surface area; therefore, chemical modification of a single nucleotide is unlikely to abolish their binding affinity. Whereas small molecule inhibitors are often unamenable to such change. Additionally, oligonucleotides are generated via automated chemical synthesis and thus are easily (commercially) accessible.

Here, we report the first examples of aptamers modified with cleavable electrophiles that provide transfer of chemical motifs to a protein (Figure 1). The aptamer serves as an affinity ligand for a protein target, bringing its conjugated electrophile in the vicinity of a nucleophilic residue, followed by selective label transfer to the protein through covalent bond formation. The aptamer is cleaved from the electrophile at the same time as the covalent crosslink is formed and can then dissociate from its target. In this first study demonstrating the overall viability of this concept, we investigated the installation of electrophilic warheads into the thrombin binding aptamer, TBA, a 15-nucleotide DNA molecule.^[14] Thrombin, a serine protease, converts fibrinogen to insoluble fibrin, thereby contributing to clotting. Consequently, thrombin misregulation can lead to myocardial infraction, pulmonary embolism, stroke, and venous thrombosis.^[15] We selected TBA as our initial target due to its facile synthesis, robust protein binding, and therapeutic potential as an anticoagulant by inhibiting thrombin’s fibrinogen binding site with nanomolar affinity.^[16]

Figure 1.

General scheme of handle transferring nucleic acid ligands. An aptamer directs its conjugated functional label and cleavable electrophile to a target protein. A nucleophilic residue quickly reacts with the aptamer-conjugated electrophile resulting in an irreversible transfer of the handle to the protein.

Electrophilic aptamers were generated by incorporating alkyne-modified phosphoramidites into TBA via solid-phase oligonucleotide synthesis and conjugated with the electrophilic warheads. Electrophiles could not be directly incorporated due to their sensitivity to basic conditions required to remove nucleobase protecting groups post-oligomerization. The modified nucleotide 5-ethynyl-2’-deoxyuridine (EdU) phosphoramidite was synthesized in four steps from 5-iodo-2’-deoxyuridine (Supporting Figure S1).^[17] By individually synthesizing each modified aptamer, the six thymidines on TBA (5’-GGTTGGTGTGGTTGG-3’) were systematically replaced with EdU to screen for positions best suited for selective thrombin modification without disrupting the aptamer structure. Oligonucleotides were purified by PAGE and characterized by MALDI. The azido-modified tosyl electrophile 1 capable of transferring a biotin handle was synthesized in four steps (Supporting Figure S2). Biotin was chosen as the first transferrable handle since it enables both sensitive detection and facile enrichment of the targeted protein. Each alkyne-bearing aptamer was then individually conjugated to the cleavable electrophile and handle through a copper-catalyzed [3+2] cycloaddition that yielded ~60% conversion and purified by HPLC. This conjugation strategy was chosen over other common methods (e.g., amine-NHS ester coupling) since it is compatible with the electrophile intended for protein conjugation (e.g., tosylate). Conjugation yields were modest due to partial hydration of the ethynyl group to a methyl ketone during oligonucleotide synthesis and basic deprotection.^[18] We addressed this problem by employing a known triisopropylsilyl (TIPS) protected EdU phosphoramidite (Supporting Figure S3),^[18a] followed by a brief TBAF exposure of the synthesized oligonucleotide after nucleobase deprotection and resin cleavage. Subsequent bioconjugation via [3+2] cycloaddition then resulted in >95% conversion.

Aptamers modified with the tosyl electrophile 1, a warhead that has broad reactivity toward His, Tyr, Glu, Asp, and Cys, and has been used for protein labeling in live cells,^[19] were individually incubated with recombinant thrombin. A 300 nM thrombin concentration was chosen in order to mimic its native setting during coagulation,^[20] and the reaction mixtures were incubated for four hours in PBS (pH 7.4) at 37 °C. The handle transfer efficiency of each aptamer was evaluated by western blot using a streptavidin-horse radish peroxidase (HRP) fusion protein and a luminol-peroxide solution for visualization (Figure 2A, Supporting Figure S4A). Importantly, a distinct structure-activity relationship was observed, as the electrophile position on the aptamer had a significant impact on its ability to transfer biotin onto thrombin, with position 13 (numbered from 5’ to 3’) appearing to be most efficient. However, when the same experiment was performed with a shortened, 1 h incubation period, no labeling was detected, suggesting that the warhead may exhibit slow handle-transfer kinetics. Nuclease-mediated oligodeoxynucleotide degradation – a limitation to traditional aptamers – occurs on relatively fast (t_1/2 ~ 1 h) timescales;^[21] we reasoned this issue could be alleviated using our electrophile conjugates if labeling occurred at timescales faster than degradation. Moreover, the overall rate of covalent bond formation is defined by the ratio of the labelling rate constant to the dissociation constant;^[22] therefore, for efficient covalent protein labeling it is key to utilize electrophilic warheads with fast reaction rates.

Figure 2.

Individual aptamers with specific thymidines modified with electrophile A) 1 or B) 2 were incubated with thrombin for 4 h and 1 h, respectively, and biotinylation efficiency was analyzed by western blot. The T13 and T3-modified aptamers showed most efficient label transfer to the protein, respectively. General scheme of the labeling reaction is displayed at the bottom.

To address this, the higher reactivity N-acyl sulfonamide electrophile 2, which selectively labels lysines,^[22b] was synthesized in four steps (Supporting Figure S5) and conjugated to TBA. Aptamers modified with 2 were incubated with thrombin for one hour under the same conditions as the previous labeling reaction (Figure 2B, Supporting Figure S4B). Like electrophile 1, this warhead showed distinct, position-dependent reactivity, with positions 3 and 7 being the most efficient, while 12 and 13 also showed some labeling. When comparing labeling kinetics of the aptamers TBA(13)-1 and TBA(3)-2, the N-acyl sulfonamide 2 showed significantly faster kinetics than the tosyl 1 and was employed in all further experiments (Supporting Figure S6).

Next, to ensure that the installed electrophile did not reduce the aptamer’s binding affinity – thereby impeding overall potency – and that the positional preferences observed in Figure 2B did not stem from steric interference caused by the warhead in certain positions, the binding constants of TBA modified at position three (TBA(3)) and nine (TBA(9)), which exhibited high and low handle-transferring efficiencies, respectively, were measured. Both aptamers were synthesized to contain the inactivated N-acyl sulfonamide warhead 2b, which lacks the N-cyanomethylene electron withdrawing group, (Supporting Figure S5) as well as a 3’ fluorescein. A fluorescence polarization assay (Supporting Figure S7) revealed very similar binding constants for TBA(3)-2b (395 nM) and TBA(9)-2b (520 nM), which also matched the unmodified aptamer (220 nM). This confirms that the differences in labeling efficiencies were not due to interference with the aptamer-protein interaction (e.g., due to steric hinderance), but are likely due to distinct positioning of the warhead in relation of accessible Lys residues on the protein surface. This is further supported by the fact that TBA(3)-2 only biotinylated 5 out of 19 lysines when incubated with thrombin, as determined by mass spec sequencing. Not surprisingly, these lysines are located at the aptamer-protein binding interface, based on a TBA-thrombin crystal structure (Figure 3A). Quantification revealed that K149, K109–110, and K36 were 91%, 74%, and 4% biotinylated, respectively, thus demonstrating site-selectivity of the handle-transferring aptamers (Figure 3B).

Figure 3.

A) Crystal structure of TBA (blue) complexed with thrombin (pink). A zoomed-in image shows that T3 is proximal to labeled lysines (green) and distant from non-labeled ones (yellow), based on mass spec sequencing. One of the biotinylated lysines, K149, is missing from the crystal structure. PDB 4DII. B) Diagram highlighting labeling yields of biotinylated lysines (green) and unmodified lysines (red). Peptides not covered by mass spec are shown in grey. Biotinylation percent represents the fraction of thrombin molecules biotinylated at the indicated Lys residues.

The peptide containing K81 was detected only when biotinylated and, therefore, labeling efficiency could not be quantified. K109 and K110 are grouped because they are located on the same peptide after tryptic digestion and hence cannot be differentiated. There is also a small part of the protein that we were unable to cover by MS sequencing, despite repeated attempts even in the case of unmodified thrombin.

Using the newly established TBA(3)-2, we determined the maximal protein labeling that can be achieved. Thrombin was treated with increasing concentrations of TBA(3)-2 and biotinylation was first observed at 40 nM and reached a maximum at 1 μM (Supporting Figure S8A). Enhancement of labeling kinetics through aptamer-based electrophile presentation to proximal nucleophilic residues was then confirmed by comparing labeling at 1 μM of TBA(3)-2 with that of the unconjugated small molecule electrophile 2 at increasing concentrations. The small molecule alone required a 50-fold higher concentration than the aptamer for similar labeling yields (Supporting Figure S8B). Thus, the proximity effect imparted by the aptamer provided significant enhancements in labeling kinetics, efficiency, and residue specificity. Next, we investigated protein target selectivity of the covalent aptamers and found that thrombin (300 nM) was exclusively biotinylated in the presence of an excess BSA (1.5 μM) (Figure 4A), a protein that has three times the number of lysines (60). Additionally, to assess the potential use of our handle-transferring aptamers in thrombin’s native environment, thrombin and TBA(3)-2 were incubated in human plasma (~1.5 mM total protein concentration), and again highly efficient target labeling was observed while no other proteins were biotinylated at aptamer concentrations of up to 500 nM (Figure 4B). A time-course experiment then revealed that covalent modification of thrombin is not only selective but also fast, with a t_1/2 of 9.6 min and 34 min at 1 and 0.5 μM of TBA(3)-2, respectively (Figure 4C, Supporting Figure S9). Mass spectrometry analysis showed that 81% of thrombin was modified within 1 h at 37 °C with an average of 1.4 biotins per protein in the presence of 1 μM of TBA(3)-2 (Supporting Figure S10); and that the TBA(3)-2 concentration was strongly correlated to the number of biotinylation events.

Figure 4.

A) Thrombin (300 nM) selectivity experiment in the presence of excess BSA (1.5 μM). B) Thrombin (300 nM) selectivity experiment in human plasma. The concentrations of TBA(3)-2 used were 0, 0.06, 0.12, 0.25, 0.5, and 1 μM. C) Thrombin biotinylation time-course experiment. The concentration of thrombin was maintained at 300 nM. Data represent averages and error bars are standard deviations from at least two independent experiments.

To demonstrate versatility in the delivery of different handles and to simplify detection of labeled proteins, a diethylamino coumarin (DEACM) transferring electrophile 3 was synthesized and then conjugated to T3 of TBA (Figure 5A, Supporting Figure S11). The fluorescent conjugate (1 μM) was then used in a thrombin (300 nM) labeling reaction (Figure 5B), and DEACM was selectively transferred to thrombin, even in the presence of excess BSA (1.5 μM). Furthermore, the covalent aptamer was again selective for the target protein when tested in human plasma and the labeled thrombin was readily visualized through gel-fluorescence imaging (Figure 5C).

Figure 5.

Fluorescent handle transfer analyzed on a denaturing gel. A) DEACM N-acylsulfonamide electrophile structure. B) Selective handle transfer to 300 nM thrombin in the presence of excess BSA (1.5 μM) using the conjugate compared to non-selective labeling using the unconjugated electrophile. C) Selective handle transfer to thrombin (300 nM) in human plasma. The concentrations of TBA(3)-3 used are 0.03, 0.06, 0.12, 0.25, 0.5, and 1 μM.

While all covalent labeling reactions so far resulted in release of the aptamer ligand from the protein target, we realized that “inverting” the warhead would result in a covalent crosslink between the targeted protein and the aptamer, rather than a transfer label (Figure 6A). This was achieved by conjugating the nucleic acid with the electrophile via its amide motif, rather than the aryl sulfonamide as before. Covalent protein-oligonucleotide conjugates have found diverse applications including protein immobilization, assay development,^[5h] and drug delivery.^[23] Furthermore, owing to fast labeling timescales, we reasoned that the protein may shield the crosslinked aptamer from nuclease-mediated degradation. Few examples exist in which nucleic acids were engineered to covalently bind proteins,^[24] including DNA decoys, nucleic acids that the consensus binding site of transcription factors, and aptamers crosslinking to purified and cell-surface proteins.^[25]

Figure 6.

A) Aptamer-protein crosslinking using the inverted electrophile B) Concentration-dependent aptamer-protein crosslinking (300 nM of thrombin). C) Selectivity experiment in human plasma. D) Time-course biotinylation experiment (300 nM of thrombin). E) Position-dependent crosslinking of TBA-4 (1 μM) with thrombin (300 nM). F) Fibrinogen clotting assay using 500 nM of thrombin. Data represent averages of at least two experiments and error bars are standard deviations from at least two independent experiments.

The inverted electrophile 4 was synthesized in 3 steps (Supporting Figure S12) and conjugated to TBA to form TBA(3)-4. It was then incubated at increasing concentrations with 300 nM of thrombin in PBS at 37 °C for 1 h and analyzed via SDS-PAGE. As expected, the aptamer crosslinked to its target, as indicated by a band shift, and the bands were integrated to quantify conjugation yield (Figure 6B, Supporting Figure S13). Interestingly, only a single aptamer crosslinked per protein, in contrast to the multiple biotin transfers seen before, likely due to the crosslinked aptamer blocking further target interaction with other aptamers. To examine its selectivity, a 5’ fluorescein-modified TBA(3)-4 was generated and incubated in human plasma spiked with thrombin. A denaturing gel (Figure 6C) again revealed selective on-target crosslinking at aptamer concentrations up to 500 nM (1 h at 37 °C), matching our previous results with the biotin-delivery aptamer (Figure 4B). A time-course experiment revealed that thrombin crosslinking was nearly quantitative within 1 h (Figure 6D, Supporting Figure S14). Because thrombin crosslinking with TBA(3)-4 showed similar reaction kinetics and target selectivity to biotinylation using TBA(3)-2, we explored the crosslinking reactivities of the remaining positions on TBA. Crosslinking electrophile 4 was conjugated to the aptamers and they were then incubated at 1 μM with 300 nM of thrombin (1 h at 37 °C in PBS). Analysis on a denaturing gel (Figure 6E) showed that crosslinking position-dependence was similar to that of biotinylation (Figure 2B), thus demonstrating that warheads are readily interchangeable from label-transfer to crosslinking.

Once fast and selective crosslinking to thrombin was confirmed, we sought to determine whether TBA-thrombin complex stabilization via the generated covalent bond would enhance aptamer-mediated inhibition of thrombin’s function. Thrombin is a key enzyme in maintaining hemostasis and hyperactivation leads to thrombosis – one of the most common causes of death.^[26] To assess enzymatic activity, we capitalized on thrombin’s cleavage of fibrinogen into insoluble fibrin, which can be quantified via absorbance at 350 nm.^[27] Thrombin was incubated with TBA for 1 h in PBS (pH 7.4) at 37 °C. Fibrinogen was added and absorbance was recorded over 30 min (Supporting Figure S15), followed by extrapolation of clotting times (Figure 6F). Compared to unmodified TBA, TBA(3)-4 lead to a 2-fold prolonged clotting time, demonstrating that the potency of an aptamer can be enhanced through covalent crosslinking.

Because TBA(3)-4 was capable of crosslinking to thrombin in spite of nucleases present in human plasma, we next explored the aptamer’s stability once crosslinked. Although nuclease-mediated aptamer degradation has been partially addressed by modification of the 2’ position on the individual nucleotides, tedious structure-activity relationship studies are required to identify positions that can be modified without significantly reducing binding affinity.^[21] Fluorescein-modified TBA(3)-4 (1 μM) was incubated with thrombin (300 nM) for 1 h in PBS (pH 7.4) at 37 °C. The crosslinking reaction was quenched through addition of glycine (10 mM) followed by a 1 h incubation period. The conjugate was then diluted 10-fold into human plasma and continued to incubate at 37 °C. When analyzed by SDS-PAGE, no band shift and no reduction in fluorescence was observed after 24 h, indicating that the crosslinked aptamer is fully stable in human plasma (Supporting Figure S16).

With aptamers conjugated to electrophiles 2, 3, and 4, delivering biotin, coumarin, and a nucleic acid molecule to thrombin established, we aimed to evaluate the diagnostic potential of covalent aptamers by measuring the protein detection limit of the target protein using three approaches. Sensitive thrombin detection has therapeutic implications as concentrations are highly varied between individuals, suggesting that thrombotic disorder treatments can be personalized based off these concentrations.^[28] Thus, we incubated the biotinylating aptamer TBA(3)-2 with decreasing thrombin concentrations in PBS (pH 7.4) at 37 °C for 1 h. The detection limit was 4 nM when measured via western blot (Figure 7, Supporting Figure S17). Compared to traditional western blots, this approach is more economical since oligonucleotides, as opposed to antibodies, can be quickly generated via chemical synthesis, are homogenous, have little batch-to-batch variability, and exhibit long shelf-lives. Additionally, membrane incubation periods are reduced to 1 h due to the high affinity biotin-streptavidin interaction, enabling a faster workflow. To simplify this assay, the detection limit was measured using fluorophore transferring aptamer TBA(3)-3. The detection limit was 33 nM; however, the same assay will likely yield lower limits if used with an electrophile that transfers a brighter fluorophore.^[29] Transfer of different fluorophores also lends itself easily to multiplexing, as this approach offers a fast (~2 h) and convenient method for native protein detection. In an attempt to decrease the detection limit, the crosslinking aptamer TBA(3)-4 was phosphorylated at the 5’ end using a T4 polynucleotide kinase in the presence of γ−³²P ATP. The radiolabeled conjugate was used at a greatly reduced concentration (20 nM) and was incubated with thrombin at decreasing concentrations under the same conditions as before. Radiolabeling of thrombin was analyzed on a denaturing gel that was visualized with a phosphorimager. Using this approach, the detection limit was reduced to 400 pM, thus demonstrating that covalent aptamers provide a platform to detect proteins of low abundance.

Figure 7.

A) Comparison of three covalent aptamer-mediated approaches for thrombin detection using 0.5 μM of aptamer for fluorescence and luminescence-based detection or 20 nM for radiography. B) A zoomed-in representation of the data (0–15 nM) from Figure 7A. Data represent averages and error bars are standard deviations from at least two independent experiments.

In summary, an aptamer capable of fast and selective protein covalent labeling was developed; it transferred functional motifs such as biotins and fluorophores to native thrombin in buffer as well as in human plasma. Labeling occurs with complete protein specificity and complete lysine chemoselectively. Partial site-specificity was observed, with only lysines at the aptamer binding interface being modified. Inversion of the crosslinking electrophilic warhead enabled rapid protein-oligonucleotide conjugate formation, with only a single aptamer transferred per protein. Our covalent aptamer inhibited thrombin’s enzymatic function with greater efficacy than its noncovalent counterpart. Additionally, labeling occurs on timescales that outcompete degradation by endogenous nucleases, and crosslinked aptamers remain immune to nuclease-mediated degradation for >24 h. Each of the three crosslinked chemical motifs enabled a unique approach to protein detection, at concentrations in the pico- to nanomolar range, using very simple western blot, fluorescence, and radiography protocols. We expect that the approach presented here is broadly applicable to existing and future aptamers and that it enables new applications of this important class of nucleic acid-based probes in the modification, detection, and inhibition of proteins.