Efficient DNA fluorescence labeling via base excision trapping | Panda Anku

Labeling strategy

A previous study described the synthesis of rotor dyes conjugated to aminooxy groups (UBER) and their application in assays for DNA base excision repair activity9. The best-performing probe was CCVJ-1, which contains a cyanovinyl julolidine dye and a three-atom linker to an aminooxy group (Supplementary Fig. 1). Two unusual features of the reaction of CCVJ-1 with AP sites in DNA are (a) an extremely rapid reaction rate, with rate constants up to 400 M−1 sec−1 at neutral pH, and (b) a high degree of light-up of the rotor dye, which is dark in solution but gives up to 500-fold enhancement upon covalent binding with AP sites in DNA. These features enabled its use in imaging DNA repair in cells and tissues; it was found that the dye in DNA is highly photostable and can be observed with common fluorescein filter sets, and can also be imaged by two-photon excitation10.

Given these potentially useful properties, we envisioned applying CCVJ-1 and related UBER probes as general fluorescent labels for oligodeoxynucleotides as well as dsDNA. The protocol would involve replacing deaminated DNA bases with such dyes in a single reaction, using inexpensive bacterial repair enzymes to excise deaminated bases (e.g., uracil and inosine) placed there during synthesis (Fig. 1b). We refer to this as Base Excision Trapping (BETr). While earlier studies have described the incorporation of fluorescent labels into AP sites for detecting products of DNA repair11,12,13, previous dyes were fluorescence quenched by DNA, giving low signal intensity and no light-up response8. For the DNA synthesis in the BETr method, we envisioned incorporating deaminated bases at one or more sites, either chemically via commercially available phosphoramidites, or enzymatically with DNA polymerase and available deaminated nucleoside triphosphates. Chemical DNA synthesis enables complete control over the location of deaminated bases, and can be performed commercially at reasonable cost unlike the chemical synthesis of DNA with fluorophores. On the other hand, polymerase-mediated synthesis could potentially enable the construction of long DNA strands containing many labeling sites, and might be performed in biological mixtures using target-specific primers. Finally, a DNA polymerase reaction, the enzymatic base excision, and the aminooxy labeling could potentially be performed in a single reaction, making the labeling process rapid and simple. Thus, we performed a series of experiments to test these various labeling possibilities.

Uracil-containing DNA and in situ covalent labeling

We examined the time required for BETr labeling by bacterial uracil-DNA glycosylase (UDG) excising deoxyuridine (dU) in DNA, and employing CCVJ-1 to form a covalent dye link with a dsDNA containing one dU (dsDNA_dU, Fig. 1c). The simultaneous excision and labeling process was complete in ca. 30 min, and fluorescence enhancement was highly reproducible and proportional to the amount of DNA present (Supplementary Fig. 2). Further experiments with an RNA complement showed that UDG excises a uracil base from both dsDNA and a DNA/RNA hybrid, and CCVJ-1 traps AP sites in both structures with fluorescence enhancement. This efficiency was somewhat surprising for the DNA/RNA case, as the native substrate of UDG is thought to be dsDNA14. The labeled DNAs could be isolated conveniently either by size exclusion spin column or by precipitation.

MALDI-TOF mass spectrometric analysis showed that UDG excised the deaminated base almost quantitatively (Supplementary Fig. 3). Notably, the UBER link stabilized the abasic site in DNA; in the absence of the dye, polyacrylamide gel (PAGE) gel analysis showed that 40–50% of the excised DNA was fragmented, likely through known elimination processes (Supplementary Fig. 4)14. However, in the presence of UBER to trap the AP site, the DNA was labeled almost quantitatively with UBER, leaving undetectable amounts of the native ODN, the excised ODN, and the truncated ODN on MALDI analysis. To evaluate multi-labeling efficiency, BETr labeling was performed with a dsDNA (4dU) containing four uracils (Supplementary Fig. 5). The results showed that all four uracils were replaced with CCVJ-1 in the ODN, leaving negligible amounts of under-labeled ODN, thus documenting sufficiently high efficiency for multi-labeling. A simple cost analysis of the process suggests that it is highly cost-efficient compared to a standard labeling approach (>100-fold less costly for single labeling in a 25-nt ODN, Supplementary Fig. 6), due to the low cost of enzyme and moderate cost of dU-containing ODNs as compared with chemically synthesized fluorescent ODNs.

Next, we tested a simultaneous in situ DNA synthesis and labeling process. For an ODN template containing nine adenine bases (9dA), a complementary primer was designed to bind at the 3′-end for initiating polymerase synthesis with deoxyuridine triphosphate (dUTP) replacing dTTP among the dNTPs15. Monitoring fluorescence intensity (538 nm) revealed a marked real-time fluorescence increase as the three steps proceeded simultaneously (Fig. 1d). Controls established that all three steps are needed for successful in situ synthesis and labeling. Moreover, this in situ DNA synthesis and BETr labeling was also applicable on RNA templates using reverse transcriptase (RTase) and specific primers (Fig. 1e). We know of no prior method that achieves DNA synthesis and covalent labeling in situ with light-up signaling on both DNA and RNA templates. Polymerase screening results showed that a lack of 3′ to 5′-exonuclease activity is beneficial for this labeling process, as is avoidance of high monovalent salt concentrations that interfere with UDG (Supplementary Fig. 7)16. The UBER-labeled ODN exhibited brighter fluorescence than achieved with Fl-dUTP (a common fluorophore for the labeling of ODNs) (Supplementary Fig. 8a–c).

We further envisioned the potential application of in situ synthesis and BETr labeling method to dsDNA obtained from cells, which can be useful for the preparation of FISH probes and blotting techniques17, 18. Genomic DNA (gDNA) extracted from HEK293T cells was subjected to nick translation and in situ BETr labeling (Fig. 1f). During the nick translation process, DNase I produces single-stranded nicks and polymerase I elongates the 3′ hydroxyl terminus, incorporating dU into the target and removing nucleotides by 5′ to 3′ exonuclease activity, then UBER labels the incorporated dU sites through BETr. We observed that all the reactions proceeded simultaneously in a single tube (Fig. 1g), and yielded a real-time fluorescence increase, reporting on reaction progress. The use of 100% dUTP in place of dTTP afforded the highest fluorescence enhancement, implying that all the sites opposite dA in gDNA were labeled with UBER (Fig. 1h). This labeling density is far higher than can be achieved with standard nick translation (Supplementary Fig. 8d). Moreover, control experiments with no dUTP showed negligible fluorescence enhancement, establishing that the levels of endogenous AP sites and dU in gDNA are low enough to yield very little background signal.

Mechanistic study of in situ DNA synthesis and BETr labeling

We examined the effects of the relative numbers and positioning of labeling sites in DNA regarding the extent of fluorescence enhancement and labeling efficiency, employing template DNAs with varied numbers of adenines. In a 60-mer template context, 2–11 adenines were introduced, then subjected to the combined polymerase/BETr procedure (Fig. 2a, Supplementary Fig. 9). While more adenines generally resulted in higher fluorescence enhancement up to as many as nine adenines (9dA) among 42 bases, the sequence with 11 adenines (11dA) showed a slightly diminished fluorescence intensity. Considering that the labeling sites reside at every three or four nucleotides in 11dA (Supplementary Table 2), there were three possible reasons for the diminished fluorescence intensity in the ODN with more fluorophores: (i) inhibition of UDG by a nearby label, (ii) a self-quenching effect between adjacent dyes, and (iii) destabilization of the duplex by the fluorophores introduced at high density. First, possible inhibition of UDG by a nearby label was tested using an ODN containing three consecutive dU nucleotides (Supplementary Fig. 10). The MALDI analysis results showed that all three consecutive sites were labeled with CCVJ-1/UDG without leaving any detectable under-labeled ODN, confirming no inhibitory effect of UDG by a nearby label already installed. To evaluate the quenching effect of the dye, dsDNAs with 1–3 consecutive dU or two dU at varied distances were prepared (Fig. 2b) and subjected to labeling. When fluorophores were introduced consecutively, the higher number of fluorophores exhibited lower fluorescence intensities, implying self-quenching at close distances (red arrow in Fig. 2b). Testing the effect of distance between two dye sites revealed that separation by two intervening base pairs gave little or no self-quenching (blue arrow in Fig. 2b). For closer substitutions, the introduction of UBER (which is larger than canonical bases) causes local destabilization of the duplex (Supplementary Fig. 11), which may place the dye in a partially single-stranded context where fluorescence intensity is reduced somewhat relative to fully doubled-stranded DNA9. Interestingly, this destabilization effect can be harnessed in turnover labeling (see below). Overall, we conclude that high UBER density and brightness can be achieved with a spacing of 2–3 nucleotides between sites, where the labeled DNA appears to be stable and self-quenching is minimal.

Fig. 2: Optimization of labeling frequency.
figure 2

a Fluorescence intensity depending on the number of labeling sites in ODNs after 1000 min of reaction. b Fluorescence intensities of different numbers of CCVJ-1 in varied distances. c Fluorescence enhancement during the in situ synthesis and BETr labeling with the template ODNs with consecutive dA. d Fluorescence enhancement measured with DNAs containing one dU at varied positions from the end of the strand. [CCVJ-1] = 20 µM, [UDG] = 10 U/mL, [Templates] = 1 µM, [Primers] = 5 µM, [dATP, dCTP, dGTP] = 50 µM, [dUTP] = 100 µM, [Polymerase] = 2 U/mL. Data acquisition after 2.5 h incubation, measured in a Fluoroskan Ascent microplate reader (485/538 nm).

We evaluated the applicability of in situ DNA synthesis and BETr labeling for yet more dense fluorescence labeling on ODNs (Fig. 2c), testing ODNs encoding 1–5 consecutive labeling sites next to the primer binding site (1A–5A). We found that consecutive adenines in a template do not prevent success in labeling with the polymerase/BETr procedure. However, the intensity slightly decreases with added dye numbers, likely due to self-quenching and local DNA destabilization as noted above. We also tested BETr-labeling efficiency at the end of DNA strands, as the number of flanking base pairs alongside a lesion significantly affects the activity of glycosylases (Fig. 2d)19, 20. Labeling with UDG and CCVJ-1 gave fluorescence enhancement when lesions were located more than 2 nt away from the end, implying UDG requires at least two flanking bases alongside the lesion.

Site-specific two-color labeling

Most DNA repair glycosylases are specialized to specific forms of damage, and recognize only a small number of related base lesions21. This fact raises the interesting possibility of a strategy for multiple orthogonal DNA repair reactions in one DNA for selective multi-color labeling, by use of two different lesions and two different repair glycosylases. As a test of this possibility, we developed an additional aminooxy-substituted rotor dye (UBER Red) (see details in the Supporting file, Supplementary Fig. 12). This dye also reacts with an AP site, affording 15-fold fluorescence enhancement at a red-shifted wavelength (ca. 620 nm) distinct from that of CCVJ-1 (Fig. 3a). For two-color labeling, we prepared a hairpin ODN (dU/dI) with two types of deaminated DNA lesions in the stem: (i) dU (a substrate of UDG), and (ii) deoxyinosine (dI) which is a substrate of N-methylpurine DNA glycosylase (MPG) (Fig. 3b). Adding MPG to the hairpin ODN successfully labeled it with UBER Green (CCVJ-1), and control experiments with a hairpin structure containing only dU (dU_sample) confirmed the dI excision specificity (Fig. 3c). The labeling capacity of UBER Red through BETr was also confirmed with UDG, showing red fluorescence with a negligible overlap in the green channel. We then doubly labeled the ODN in a site-specifically by consecutive treatment with each glycosylase/dye combination (Fig. 3c). The fluorescence intensities and the colors of labeled ODN documented that each dye was incorporated exclusively in the designated positions in a site-specific manner (Fig. 3c). Thus, the data show that dual labeling can be achieved by the use of two deaminated bases and enzymes selective for each.

Fig. 3: Two-color labeling, and selectivity on a target DNA.
figure 3

a Chemical structures and fluorescence spectrum of UBER Green and UBER Red showing enhancement in AP site. b Illustrational description of two-color labeling of a hairpin ODN containing two different DNA lesions with corresponding glycosylases. c A picture of UBER labeled ODNs purified by DNA precipitation taken on a gel illuminator with 365 nm excitation, and fluorescence intensities at 538 and 590 nm while excited at 485 nm. d Illustration of selective labeling of target DNA in the presence of spectator DNAs in the mix. e Fluorescence changes in the in situ synthesis and labeling of DNA in the presence of spectator DNAs with and without target DNA. f Agarose gel (1.5%) electrophoresis results showing selective labeling of the target DNA in the presence of spectator DNAs. [UBERs] = 10 µM, [hairpin oligo] = 1 µM, λex = 480 nm (UBER Green), 515 nm (UBER Red), [UDG] = 20 U/mL, [MPG] = 50 U/mL, [Target DNA] = 1 µM, [Spectator DNAs] = 1 µM each, [Target-specific primer] = 5 µM, [Klenow] = 2 U/mL, [dNTP (A,C,G)] = 50 µM, [dUTP] = 100 µM. The fluorescence spectra in a were measured on a Jobin Yvon-Spex Fluorolog 3. The fluorescence intensities were measured on c, e a Fluoroskan Ascent microplate reader (485 nm/538 nm) at 37 °C.

Synthesis of labeled DNAs on a sequence-specific target nucleic acid

We envisioned that synthesis of fluorescent-labeled DNA selectively on a complementary target DNA or RNA might be applied to the detection of nucleic acids among mixtures as well as to in situ synthesis of FISH probes in biological context, while removing the need for washing steps (Fig. 3d). To test this, a mixture of five spectator DNAs (41–60 nt) was used along with a specific target DNA (60nt), then the polymerase/BETr in situ labeling was carried out with a target-specific primer (Fig. 3e, Supplementary Table 2). Fluorescence analysis showed that the spectator DNAs lacking primer complementarity resulted in negligible covalent labeling, while the addition of a target DNA to the pool exhibited significant in situ fluorescence enhancement over time, documenting sequence-specific synthesis of fluorescently labeled DNA. Agarose gel analysis further confirmed that the in situ covalent labeling of UBER proceeded only in the presence of target DNA among the spectator sequences (Fig. 3f). High specificity was also obtained with an RNA template in the presence of background tRNA (Supplementary Fig. 13). This suggests the possible future utility of polymerase/BETr labeling for detection of targets in complex mixtures.

Fluorescence signal reporting with isothermal amplification

We next considered strategies involving the use of BETr for signaling isothermal amplification. Isothermal rolling circle amplification (RCA) rapidly generates long concatemeric copies of circular DNAs, and is widely employed in biomolecular target detection and amplification22. Given that a short DNA primer (10–15 nt) can be extended by a polymerase into a very long DNA strand (>12,000 nt)23, we envisioned incorporation of many dU nucleotides with simultaneous BETr labeling, affording a long labeled DNA with high brightness. For this test, we employed M13mp18 circular ssDNA, ca. 7200 nt in length (Supplementary Fig. 14). Given the presence of consecutive adenine sites which would impede fluorescence enhancement in this circular DNA, we envisioned the use of mixtures of dUTP with dTTP to increase average spacing of labeling. When both dUTP and dTTP were provided in the reaction, the difference in ultimate fluorescence showed that dUTP and dTTP compete with each other, establishing that dUTP is close to dTTP in efficiency as a polymerase substrate (Fig. 4a). Interestingly, 50% dUTP in the mix resulted in 75% fluorescence enhancement, suggesting highly favorable substrate behavior with this enzyme. The results led us to test in situ RCA and BETr labeling with different ratio of dUTP to dTTP to optimize the ratio for efficient fluorescence labeling (Fig. 4b). In the beginning of the amplification (<100 min), a higher proportion of dUTP showed greater fluorescence enhancement, implying the faster incorporation and labeling. However, a low ratio (<2% dUTP) exhibited higher fluorescence enhancement at extended reaction times, plausibly due to the optimized frequency of fluorescence labeling (Supplementary Fig. 15).

Fig. 4: Isothermal signal amplification and a bright DNA construct for tagging.
figure 4

Strategies for intense labeling via multiple incorporation of UBER dye into DNAs. a Fluorescence intensity changes of in situ DNA synthesis and BETr labeling with different ratio of dUTP and dTTP. [9dA] = 1 µM, [Primer] = 5 µM, [dUTP] = 100 µM. b Fluorescence changes in RCA with M13mp18 and in situ BETr labeling depending on the relative ratio between dTTP and dUTP. c Fluorescence intensities in real-time BETr labeling of ssDNA components versus the assembled TD. d Direct immunofluorescence images of HeLa cells (control and stressed) with anti-G3BP1 tethered with the labeled tetrahedral DNA, visualizing stressed granules. [M13mp18] = 2.5 µg/mL, [RCA Primer] = 10 µM, [dATP, dCTP, dGTP] = 1 mM, [dUTP+dTTP] = 1 mM, [Polymerase] = 10 U/mL, [UDG] = 20 U/mL, [CCVJ-1] = 50 µM (b), 20 µM (a, c). [ssDNAs] = 2 µM. Scale bar = 20 µm. Bar was plotted with the data at 5 h of incubation. The fluorescence intensities were measured on a Fluoroskan Ascent microplate reader (485 nm/538 nm) at 37 °C.

High-intensity multi-labeling of a DNA construct

DNA nanostructures have had a remarkable impact on nano-science and -technology24. In this regard, a method that enables direct labeling of a DNA nanostructure after its 3D assembly holds significant potential, as dU replacing dT provides fully canonical base pairs, avoiding interference with hybridization and DNA assembly. Taking advantage of the multi-labeling and spacing information generated above, we proceeded to label a compact DNA construct into a bright fluorescence tag. Multi-labeling is broadly useful for sensitive analyte detection and imaging. Due to the utility of multiple labels, DNAs have been studied as scaffolds to host multiple dyes; an example includes the use of dsDNA stained with DNA-intercalating dyes25, and self-assembly of multiple strands labeled with a single fluorophore at the end of the strands26. To test the labeling capability of BETr toward DNA assembly, tetrahedral DNA (TD) nanostructure was tested as a candidate for tagging. TDs are structurally compact and stable, are formed via simple self-assembly, and exhibit resistance to nuclease degradation27. A TD containing one deoxyuridine per 17 bp edge (six in total) was assembled from four ssDNAs, one of which was extended beyond the structure as a site for tethering via hybridization (Fig. 4c). BETr labeling of the assembled TD resulted in much greater fluorescence intensity than each ssDNA component (Fig. 4c), confirming the fluorescent advantage of a rigid duplex context for the UBER dye and labeling capability of BETr toward DNA assembly (Supplementary Fig. 16). Compared with fluorescein, a common and highly bright dye class28, the CCVJ-1 labeled TD showed three-fold greater fluorescence intensity than a fluorescein-labeled oligonucleotide (Supplementary Fig. 17). It is likely that yet greater brightness could be achieved with constructs containing greater numbers of CCVJ-1 dyes, while multiple labeling of DNA with fluorescein can result in strong self-quenching (Supplementary Fig. 8)8.

As a test of fluorescence tagging and imaging with the TD construct, we monitored the formation of stress granules in HeLa cells with a G3BP1 primary antibody conjugated with an oligonucleotide29. Stress granules are phase-separated aggregations in the cytosol composed of mRNAs and RNA-binding proteins30, and the protein G3BP1 is employed commonly as a target to image the granules31. The stress granules were induced by the treatment of sodium arsenite for 30 min, and directly visualized with the TD labeled primary antibody for G3BP1 protein (Fig. 4d) by hybridization to the antibody’s conjugated oligonucleotide. This enabled clear visualization of stress granules in cells without a secondary antibody, which can reduce the potential cross-reactivity in multianalyte imaging.

Turnover detection of target DNA

We found that selective properties of repair enzymes can be taken advantage of in labeling strategies. For example, MPG, a bacterial enzyme that excises alkylated and deaminated purines, recognizes DNA lesions only in dsDNA, not in ssDNA32. Based on this dsDNA specificity, we designed a DNA detection system with the potential for turnover (Fig. 5a). In this detection system, a DNA probe containing dI, a substrate of MPG, in the center of the strand is designed to hybridize with a target DNA. In the absence of target DNA, the probe itself is a very poor substrate for MPG as it is single-stranded, yielding no fluorescence enhancement (Fig. 5b). On the other hand, in the presence of complementary target DNA, it forms an enzyme-competent duplex, and then is repaired and labeled by the MPG/UBER combination through BETr, affording strong fluorescence enhancement over time. Additionally, we tested varied bases opposite dI to optimize the activity of MPG (Supplementary Fig. 18), revealing the highest fluorescence enhancement when dI was paired opposite dT or dG, consistent with the reported kinetics of MPG32. Moreover, the detection system can also be applied to RNA detection (Supplementary Fig. 19). We found that the MPG enzyme also functions to excise the base from dI in DNA hybridized to RNA, enabling the detection of the target RNA sequence.

Fig. 5: Detection of a target through BETr with isothermal turnover and single nucleotide specificity.
figure 5

a Illustration describing the mechanism of turnover detection strategy. b Fluorescence responses of labeling probe oligo (11mer) in the absence and presence of target DNA (Template dT). c Fluorescence enhancement upon the detection of target DNAs containing single mismatch in varied positions. d Fluorescence response depending on the lengths of probe oligos when detecting target DNA (Template dT). e, f Fluorescence response while detecting target DNA (Template dT) with 11mer depending on the relative amount of probe oligo and target DNA. [Target DNA] = 2 µM in be, 10 nM in f. [Probe oligo] = 10 µM (bd), 200 nM–80 µM (ef), [UBER] = 10 µM in be, 100 µM in f, [MPG] = 100 U/mL. The fluorescence intensities were measured on a Fluoroskan Ascent microplate reader (485 nm/538 nm) at 37 °C.

We further tested the single-mismatch discrimination of this strategy by preparing mismatched targets with mismatch positions varied 1–4 nt from the labeling site. Single mismatches 1 nt or 2 nt from the labeling site resulted in no fluorescence enhancement, thus showing very high specificity (Fig. 5c). Thus, the turnover detection system can be used to discriminate single nucleotide polymorphisms within this range. We hypothesize that the discrimination arises from the mismatch destabilization of the helix, preventing the formation of a stable duplex substrate for MPG. Mismatch stabilities depend both on the position as well as the specific bases involved, both of which were varied here. We note that the “gamma” site, which displayed less discrimination, was originally an AT pair, mutated to a relatively stable AG mismatch, while the others were GC pairs, likely resulting in larger Tm differences.

Given that the dye-labeled position destabilizes the duplex, a DNA probe having optimized length may be long enough to form a semi-stable duplex that can act as a substrate for MPG, yet short enough to dissociate after the dye labeling destabilizes the duplex (Fig. 5a). We designed 7–11mer dI-containing probe strands and tested them for target DNA detection over time (Fig. 5d). The results revealed that at 37 °C a 7-mer strand was too short to act as a substrate, while an 11-mer exhibited robust repair activity with MPG and significantly greater fluorescence enhancement. To evaluate an application of the turnover design, fluorescence enhancement in this detection system was observed with different ratios of probe strand (11mer) to target DNA (Fig. 5e). The results showed that one copy of target DNA can label more than one copy of probe ODN. Moreover, an excess amount of probe ODN showed identical initial kinetics with an equivalent amount of probe ODN, indicating that only hybridized dsDNA was subject to BETr labeling and that the labeled single-stranded probe ODNs can turn over (Supplementary Fig. 17b). Additionally, we also investigated the detection of a sequence longer than 11 nt by introducing multiple deoxyinosines in the probe strand (Supplementary Fig. 20). The results showed that introducing multiple dI nucleotides with four base pairs between the lesions facilitated isothermal turnover detection for a longer sequence.

Testing with a lower 10 nM target DNA concentration (0.5 picomole) showed that five units of MPG facilitates ca. 2000-fold turnover of probe ODN in this condition (Fig. 5f), yielding greatly amplified signals relative to copy number. An advantageous feature of this turnover system is that the probe strand is completely non-fluorescent and bio-compatible, which enables the use of relatively high concentrations to detect a target at much lower concentration. The detection limit, in this case, was 0.5 femtomole of target (10 pM) using 10 µM probe DNA (Supplementary Fig. 21).

Leave a Comment