DNA Modifications Enabling Proximity Biotinylation

Abstract

Advances in peroxidase and biotin ligase-mediated signal amplification have enabled high-resolution subcellular mapping of endogenous RNA localization and protein–protein interactions. Application of these technologies has been limited to RNA and proteins because of the reactive groups required for biotinylation in each context. Here we report several novel methods for proximity biotinylation of exogenous oligodeoxyribonucleotides by application of well-established and convenient enzymatic tools. We describe approaches using simple and efficient conjugation chemistries to modify deoxyribonucleotides with “antennae” that react with phenoxy radicals or biotinoyl-5′-adenylate. In addition, we report chemical details of a previously undescribed adduct between tryptophan and a phenoxy radical group. These developments have potential application in the selection of exogenous nucleic acids capable of unaided entry into living cells.

With the development of the biotin ligase BioID in 2012,¹ proximity biotinylation has become a powerful tool for the mapping of protein–protein interactions and protein localization in living cells. More recent advances have enabled the mapping of protein and RNA localization across entire subcellular compartments with the ascorbate peroxidase APEX or its optimized form, APEX2.²⁻⁴ Despite the widespread success of these approaches, an equivalent method for mapping DNA localization and DNA–protein interactions has yet to be achieved. BioID and TurboID Biotin ligases catalyze synthesis of a highly reactive biotinyl-5′-adenylate intermediate from biotin and ATP, allowing proximity biotinylation of primary amine nucleophiles as reactive targets.^1,5,6 While such targets exist in proteins on lysine residues and N-termini,⁷ none are found in nucleic acids. APEX and APEX2, on the other hand, catalyze production of short-lived reactive phenoxy radical intermediates from biotin tyramide (BT) in the presence of H₂O₂.^8,9 Though proximity biotinylation by peroxidases has the impressive advantage of dual protein and RNA radical biotinylation, this approach cannot be used directly for biotinylation of unmodified DNA.¹⁰

We have previously shown that peroxidase-based biotinylation of oligodeoxyribonucleotides with BT is enabled by the addition of 5′ fluorescein and we described structural details of BT-fluorescein adducts.¹⁰ We demonstrated that an in vitro streptavidin gel shift assay is capable of detecting biotinylation of both RNA and 5′-fluorescein-modified DNA oligonucleotides and, importantly, provided molecular insights into differences in RNA and DNA biotinylation. Here, we expand the toolbox of DNA proximity biotinylation methods with new approaches suitable for peroxidases and, for the first time, biotin ligases. We envision application of these approaches for selection of exogenous DNAs capable of unaided entry into living cells.

It is known that tyrosine residues are protein reactive sites biotinylated by APEX2-generated BT phenoxy radicals.^11,12 To determine if other biological functional groups are reactive with BT phenoxy radicals, we devised a competition assay for the assessment of all other natural amino acids. In this assay, a streptavidin gel shift is used to detect horseradish peroxidase (HRP) and H₂O₂-dependent radical biotinylation of 5′-fluorescein-modified oligodeoxyribonucleotide 6132.¹⁰ Increasing concentrations of individual competing amino acids are added to the reaction. Amino acid reaction with BT radicals is detected by quenching of 6132 biotinylation. This results in a quantifiable inhibition of 6132 gel shifts.

We first observed that excess fluorescein reduces biotinylation-dependent streptavidin gel shifts as expected, validating that this assay is sensitive to competition for radicals (Figure 1). We then tested all 20 natural amino acids in competition with fluorescein-modified 6132 at 100-, 500-, and 1000-fold excess to 6132 (Figures S1–S3). As expected, excess tyrosine competes for phenoxy radicals and reduces the gel shift observed under competitor-free conditions (Figure 1). Notably, our investigation revealed that tryptophan also quenches the reaction. Tryptophan sensitivity to phenoxy radicals has not been previously described. The remaining 18 amino acids, exemplified here by glycine, were unreactive (Figure 1; Figures S1–S3). These results suggest that both phenol and indole functional groups might serve as simple biotin tyramide-receptive “antennae” for proximity biotinylation of appropriately conjugated DNA oligonucleotides.

Figure 1.

Amino acids tyrosine and tryptophan compete for phenoxy radical biotin tyramide intermediates to quench biotinylation. (A) Excess amino acids or fluorescein are added to in vitro HRP-mediated biotinylation reaction and to monitor competition for radical reactivity. Covalently biotinylated oligonucleotides are detected by streptavidin gel shift, which is quenched in the presence of reactive competitors. (B) Quantification of gel shifted material in the presence of various amino acids or fluorescein. Fraction shifted is normalized to competitor-free conditions in each gel, typically approximately 0.45.

Having screened all natural amino acids for phenoxy radical reactivity, we set out to test tyrosine and tryptophan as DNA conjugates in place of fluorescein in our established gel shift assay of phenoxy radical biotinylation. Tyrosine hydrazide was conjugated to 5′ aromatic aldehyde-modified oligodeoxyribonucleotide 6542 to generate 6542–5′Tyr (Figure 2A). Tyrosine-conjugated oligonucleotides were purified from denaturing polyacrylamide gels (Figure S4A). Unexpectedly, we first found that the 5′ aromatic aldehyde modification of 6542 was itself sensitive to phenoxy BT radicals, though only approximately 15–20% as reactive as 5′ fluorescein (Figure 2B, lane 7; Figure S4B, lane 7; Figure S4C). 6542–5′Tyr conjugate was also slightly less reactive than the 5′ fluorescein of 6132, though significantly more reactive than 6542 (Figure 2C, lane 7; Figure S4B,C). Biotinylation of both the 5′ aromatic aldehyde of 6542 and 6542–5′Tyr requires the simultaneous presence of BT, H₂O₂, and HRP, confirming that the catalyzed reaction is similar to that described in peroxidase proximity labeling of proteins.¹¹ Thus, despite being approximately 25% less reactive than 5′ fluorescein, 5′ tyrosine modifications present a useful alternative for DNA oligonucleotide biotinylation in cases where use of a fluorescent substrate is incompatible with the required application. This may be helpful in fluorescence microscopy applications when other fluorescein-labeled reagents are required.

Figure 2.

Tyrosine DNA oligonucleotide conjugates react with phenoxy biotin tyramide radicals. (A) Reaction scheme for covalent conjugation of tyrosine hydrazide and a 5′ aromatic aldehyde-modified oligonucleotide. (B) Biotinylation and streptavidin gel shift of 5′ aldehyde modified oligonucleotide 6542 (black dot; ∼5% yield under these conditions) occurs only in the presence of H₂O₂, HRP, and BT. (C) Biotinylation and streptavidin gel shift of 5′ tyrosine-conjugated oligonucleotide 6542 (black dot) occurs only in the presence of H₂O₂, HRP, and BT. DNA is visualized by SYBR Gold.

We next tested a 5′ tryptophan conjugate as an alternative to fluorescein to facilitate DNA oligonucleotide biotinylation. A Nα-Boc-l-tryptophan N-hydroxysuccinimide (NHS) ester (Trp-NHS) was coupled to 5′ primary amine modification of oligodeoxyribonucleotide 6543 (Figure 3A). The conjugated product, 6543–5′Trp, was purified by denaturing polyacrylamide gel electrophoresis (Figure S5A). 5′-conjugation of Trp-NHS was further confirmed by high resolution LC-Orbitrap-MS to identify the predicted product (Figure S5A–D). 6543–5′Trp was amenable to phenoxy radical biotinylation and streptavidin gel shift approximately 50% reactivity compared to the fluorescein conjugate (Figure 3B, lane 7; Figure 3C). Remarkably, we found that biotinylation of 6543–5′Trp occurred even in the absence of HRP. requiring only BT and H₂O₂ (Figure 3B, lane 6; Figure 3C). This reaction, enhanced by the addition of peroxidase, could serve as a tool for nonenzymatic biotinylation of Trp-containing DNA conjugates. These results also suggest that proximity biotinylation by APEX2 might be subject to nonenzymatic background biotinylation upon prolonged cell incubation with biotin tyramide and H₂O₂.

Figure 3.

Tryptophan DNA oligonucleotide conjugates react with phenoxy biotin tyramide radicals. (A) Reaction scheme of conjugation of tryptophan NHS ester to 5′ primary amine modified oligonucleotides. (B) Phenoxy radical biotinylation of 6543–5′Trp occurs in the presence of H₂O₂ and BT alone (lane 6) but is enhanced by the addition of HRP (lane 7, ∼5% yield). (C) Quantification of gel shifts of 6542–5′Trp in with or without acceleration by HRP versus 5′ fluorescein modified 6132. (D) Proposed reaction scheme of unexpected HRP-free biotinylation of 5′-tryptophan-modified oligonucleotides.

High resolution LC-Orbitrap-MS analysis of the HRP-free biotinylation product further confirmed the unexpected uncatalyzed biotinylation of 6543–5′Trp. We identified a previously unreported 5′Trp-BT adduct (Figure S5E–H) and propose a novel reaction between tryptophan and BT that can be accelerated by peroxidase (Figure 3D, Figure S5G). This reaction can occur on 5′Trp-conjugated oligonucleotides or with free tryptophan, as evidenced by the ability of free tryptophan to quench biotinylation in a competition assay (Figure S6A). The reaction therefore does not depend on the Nα-Boc group present in the 5′Trp conjugate that we generated here. Tryptophan biotinylation does, however, require the tyramide group of BT as shown by the absence of a gel shifted species if the reaction is carried out with biotin substituted rather than BT (Figure S6B).

We next studied the possibility of expanding the suite of DNA proximity biotinylation tools from those compatible only with peroxidases to approaches suitable for biotin ligases. Proximity biotinylation by stably expressed biotin ligases relies only on exogenous addition of biotin to cells, unlike biotinylation by APEX2, which requires the addition of toxic H₂O₂. In the presence of endogenous ATP, biotin ligases generate highly reactive biotinyl-5′-adenylate intermediates that then react with proximal nucleophiles such as primary amines. Recently, directed evolution was used to develop the highly promiscuous biotin ligase TurboID which significantly outperforms previously established enzymes such as BioID.¹³ We therefore expressed and purified TurboID for in vitro reactions to model oligonucleotide biotinylation by biotin ligase (Figure S7).¹⁴ To first ensure that TurboID was enzymatically active and capable of labeling proteins, we combined TurboID, ATP, and biotin in vitro with HRP now serving as a convenient biotinylation target. Western blot analysis of the products using AlexaFluor647-streptavidin to mark labeled proteins revealed both promiscuous intramolecular biotinylation of TurboID and intermolecular biotinylation of HRP, confirming TurboID activity (Figure S8).

The parent enzyme of TurboID, BirA, transfers biotin to a target lysine through a reactive biotinyl-5′-adenylate intermediate.^15,16 Because DNA lacks aliphatic primary amines, we reasoned that the addition of convenient and commercially available amino modifications would allow oligonucleotide biotinylation by TurboID (Figure 4A).¹⁷ We first demonstrated this using streptavidin gel shift assays to detect biotinylation of a 5′C₆-amino modified oligonucleotide in the presence of TurboID, ATP, and biotin (Figure S9A,B). Oligonucleotides lacking the primary amine modification are not detectably biotinylated. Multiple amino-containing oligonucleotide modifications are commercially available. We compared biotinylation of oligonucleotides containing primary amines on spacers (C₆ and C₁₂) and those with internal amino-modified deoxythymidine bases (Figure 4C, Figure S9C–D). All variants carrying primary amines were similarly biotinylated (Figure S9E). To further demonstrate biotinyl-5′-adenylate specificity for primary amines, a competition assay was performed by adding excess glycine to the reaction mixture. Competition between amino-modified DNA and glycine for reactive intermediates resulted in a diminished gel shift (Figure S10A). We also demonstrated that observed gel shifts are dependent on the predicted streptavidin–biotin interaction by blocking streptavidin with excess biotin prior to adding the oligonucleotide, resulting in a reduced gel shifts (Figure S10B).

Figure 4.

DNA carrying a primary amine is a substrate for biotinylation by biotin ligase. (A) Schematic of reaction between ATP and biotin catalyzed by TurboID yielding biotinyl-5′-adenylate, which subsequently reacts with a primary amine modified oligonucleotide. (B) Comparison of streptavidin gel shifts (black dot) of an oligonucleotide containing no primary amine (6836), an oligonucleotide containing a 5′ primary amine with a C₁₂ spacer (6835), an oligonucleotide containing a 5′ primary amine with a C₆ spacer (6834), and an oligonucleotide containing a 5′ primary amine with a C₆ spacer and two internal primary amine modified deoxythymidine nucleotides (6870). Yields are ∼5% under these experimental conditions.

As has been discussed for reactions involving biological nucleophiles,¹⁸ reaction products may have different stabilities that obscure their formation. The DNA molecules in our studies possess one or more primary alcohol residues that could act as nucleophiles to produce esters upon reaction with biotinyl-5′-adenylate. Such esters may be unstable under these conditions, explaining the absence of detectable biotinylation of unmodified DNA. The aromatic amino groups of DNA bases are insufficiently nucleophilic to react.¹⁹

To demonstrate that amine-modified DNA oligonucleotide proximity biotinylation can be practically implemented beyond the in vitro reaction system, we created and characterized HEK293T cells stably expressing a nuclear localizing TurboID (TurboID-NLS) (Figure 5A).²⁰ We then tested whether a previously published, nuclear targeting aptamer can be selectively biotinylated in this cellular system when endowed with a 5′ primary amino group (7042).²¹ This nuclear targeting aptamer was selectively recovered from TurboID-NLS expressing cells vs HEK293T cells (Figure 5B), and recovery was significantly higher than a previously characterized negative control aptamer also endowed with a 5′ amino group (7041) (Figure 5B). Further, we demonstrated that a 5′ amino-modified DNA oligonucleotide (in this case 7041) can be selectively biotinylated, captured on streptavidin magnetic beads, and amplified by PCR even in the presence of a vast excess of unmodified DNA oligonucleotide (in this case 5589) (Figure 5C).

Figure 5.

Amine-modified aptamers can be specifically captured after TurboID biotinylation in cell nuclei and in solution. (A) HEK293T cells stably expressing TurboID-NLS were characterized by showing evidence of biotinylation only in nuclei after the addition of excess free biotin and staining with Alexa Fluor 647 labeled streptavidin counter-stained with DAPI. (B) A previously published karyophilic aptamer 7042 and negative control 7041 were each modified with a 5′ primary amine. 7042 was preferentially recovered using streptavidin magnetic beads as determined by qPCR compared with negative control 7041 (*p < 0.05) C) A solution of 1:100,000, 5′ amino modified DNA (7041):unmodified DNA (5589) was subjected to an in vitro biotinylation reaction with TurboID. 7041 and preferentially recovered with streptavidin magnetic beads as determined by qPCR compared to 5589 (*p < 0.05).

We have demonstrated several novel methods for biotinylation of exogenous DNAs. Methods for peroxidase-based biotinylation confirm our previous observation that fluorescein is readily biotinylated by BT. We now show the utility of tyrosine-DNA conjugates for biotinylation using BT. Tryptophan could also be biotinylated by BT in a peroxidase-catalyzed reaction. Remarkably, tryptophan was also detectably biotinylated by BT in the absence of peroxidase. Therefore, tryptophan may be less desirable as a conjugate “antenna” due to the possibility of background biotinylation unrelated to peroxidase activity. We then showed that biotin ligase-dependent biotinylation can be extended to amino-modified exogenous DNAs in the presence of biotin and ATP. The addition of biotin ligase-based tools allows proximity biotinylation of nucleic acids without the need for toxic H₂O₂. Based on our demonstration of selective biotinylation and recovery of amine-modified DNA, both in cells and in the presence of vast excess unmodified DNA, we envision a future application of these methods for selection of exogenous DNA oligonucleotides capable of homing to desired subcellular compartments.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.bioconjchem.2c00475.

• Materials and methods as well as supplemental figures (PDF)

Author Present Address

^# Department of Chemistry, Smith College, Northampton, Massachusetts, 01063 United States

Author Contributions

^¶ B.W. and K.P. contributed equally. Project conception: B.W., L.J.M., P.D. Manuscript preparation: B.W., K.P., and L.J.M. Experimental work: B.W., K.P., S.B., and L.B.B.

This work was supported by NIH grants GM128579 and GM143949 (L.J.M.), the Mayo Clinic Graduate School of Biomedical Sciences (B.W., K.P.), and an NSF graduate fellowship (B.W.).

The authors declare no competing financial interest.