Toward a rational approach to design split G-quadruplex probes

Abstract

Hybridization probes have become an indispensable tool for nucleic acid analysis. Systematic efforts in probe optimization resulted in their improved binding affinity, turn-on ratios, and ability to discriminate single nucleotide substitutions (SNSs). Split (or multicomponent) probes is a promising strategy to improve probe selectivity and enable analysis of folded analytes. Here, we developed criteria for the rational design of split G-quadruplex (G4) peroxidase-like deoxyribozyme (sPDz) probe that provides visual output signal. The sPDz probe consists of two DNA strands that hybridize to the abutting positions of a DNA/RNA target and form a G4 structure catalyzing, in the presence of a hemin cofactor, H2O2-mediated oxidation of organic compounds into their colored oxidation products. We have demonstrated that probe design becomes complicated in case of target sequences containing clusters (two or more) of cytosine residues, and developed strategies to overcome the challenges to achieve high signal-to-noise and excellent SNS discrimination. Specifically, to improve selectivity, a conformational constraint that stabilizes the probe’s dissociated state is beneficial. If the signal intensity is compromised, introduction of flexible non-nucleotide linkers between the G4-forming and target-recognizing elements of the probe helps to decrease the steric hindrance for G4 PDz formation observed as a signal increase. Varying the modes of G4 core splitting is another instrument for the optimal sPDz design. The suggested algorithm was successfully utilized for the design of the sPDz probe interrogating a fragment of Influenza A virus genome (subtype H1N1), which can be of practical use for flu diagnostics and surveillance.

Hybridization probes use synthetic oligonucleotide strands that specifically bind to nucleic acids targets through Watson-Crick base pairs. They have been used in PCR, microarrays, and fluorescent in situ hybridization, to name a few, for diagnosis of infectious diseases and human genetic disorders. Along with diagnostic applications, understanding of nucleic acid hybridization is important for such technologies as PCR, antisense, siRNA, and CRISPR/Cas. Optimization of hybridization probes has been a subject of extensive research, which lead to establishing the nearest neighbor hybridization parameters for determination of binding affinity and melting temperature of the probe-analyte hybrids.¹ Optimization of the hybridization selectivity of DNA microarray probes has been extensively studied and optimized.^2,3 Soon after introduction of nanostructured molecular beacon probes, a series of studies on the probe optimization has revealed the relationship between probe performance and the length of its stem and loop parts, and, in general, the effect of the degree of the conformational constraint on the probe’s selectivity.^4–6 These studies established the guidelines for the molecular beacon probe design, which contributed to both widespread use of the probes, as well as our general understanding of hybridization thermodynamics.⁷ Recent developments of molecular-beacon-based split (or multicomponent) probes demonstrated how some of the probe limitations⁸ can be overcome.^9–14 The multicomponent probe approach promises to add an important advantage to the existing hybridization tools by enabling recognition of structured nucleic acids with high affinity and selectivity.^14,15 Systematic investigation on the design of multicomponent probes can facilitate their practical applications.

One of the advantages of split hybridization probes is their ability to reliably discriminate single nucleotide substitutions (SNS) in nucleic acid targets under ambient conditions.^{10,11,14–16} Analysis of SNS patterns in human genes is relevant to the diagnostics of genetic diseases,¹⁷ determining people’s predisposition to diseases,¹⁸ or response to therapy,¹⁹ and can assist in studying the evolution of human populations,²⁰ as well as in forensic applications.²¹ In addition, SNSs can serve as important markers for genetic studies of human disease vectors,²² and for prediction of pathogen drug-resistant phenotypes.²³ The potential of hybridization probes for SNS analysis has been widely explored in genotype-based drug susceptibility testing of pathogenic bacteria, which is considered a faster alternative to routinely used phenotypic tools.²⁴ Most molecular approaches for drug susceptibility testing rely on fluorescent labels covalently attached to the probes to reveal the presence of a genetic signature of interest. In addition to the high cost of fluorescent probes, such tests require instrumentation to read the signal, which limits their use to laboratory settings. The use of color change as a signal readout instead of fluorescence would make molecular drug susceptibility testing more affordable and suitable for point-of-care settings (e.g. clinics or doctor’s office), since in this case the signal can be read by the naked eyes without the need for instrumentation.²⁵ This would make molecular drug susceptibility testing more affordable to the patients, which would allow for more efficient and timely treatment and prevention of antibiotic misuse.

Optical sensors based on peroxidase-like deoxyribozymes (PDz) have been wildly explored for the detection of various analytes including nucleic acids.^26–42 The color is generated due to the catalytic activity of G-quadruplex (G4) DNA,^43–47 which uses hemin as a cofactor to catalyze peroxidation of colorless organic molecules to produce a colored oxidation product. Previously, several probes utilizing split PDz (sPDz) for colorimetric nucleic acid detection have been reported, including those targeting SNS sites.^34–42 However, to the best of our knowledge, nucleic acid detection tests based on PDz probes have yet to reach commercial success. High background signals were previously reported as a major deterrent,^48,49 and such issues are further investigated in this study.

In the PDz design, twelve guanine residues can be distributed between the probe’s strands either symmetrically (each strand contains two G-triads, 6:6)^34–39 (Scheme 1, middle) or asymmetrically (one strand includes a single G-triad, while another contains three triads, 3:9 or 9:3).^40–42 The two strands are equipped with DNA sequences complementary to adjacent positions on the target (“arms” in Scheme 1). In the presence of the complementary DNA or RNA target, the G4 structure is formed due to the proximity of the two PDz subunits resulted from their hybridization to the target (Scheme 1, right). In the case of mispairing between the target and strand S (e.g. in case of SNS) the catalytic G4 structure fails to assemble, and the signal remains low (Scheme 1, left). The differential binding capabilities of the probe components allow for their fine-tuning and ensure maintaining both high affinity and selectivity – the combination of properties generally unattainable by conventional hybridization probes.^15,50

Scheme 1.

Design of split peroxidase-like deoxyribozyme (sPDz) probe. The G4 core is split into two halves, with the target recognition sequences (“arms”) added to each half to constitute strands U and S. The color change is triggered only in the presence of the fully matched target due to the formation of the catalytically active G4-hemin complex promoting peroxidation of colorless substrates (e.g. ABTS²⁻) to colored products (ABTS^−.).

Despite an abundance of data reflecting the high selectivity of the sPDz probes interrogating various targets,^{35,37,39–42} systematic investigation on the probe design is still lacking. Previously,⁴⁸ the effect of the G4-splitting mode on the ability of the probe to bind a fluorogenic hemin analogue, PPIX, has been investigated. It has been demonstrated that the background caused by symmetric splitting of the G4-forming sequence is higher than that for asymmetric splitting. This conclusion was, however, made based on results with one model target only. At the same time, primary and/or secondary structure of the target may have a significant effect on the probe performance. Unintended interactions of the G4-forming nucleotides with the target, especially if the target is C-rich, may be more pronounced for certain modes of G4 splitting, thus governing the probe design. Indeed, for the reported use of the sPDz probe,^35,39,41 thorough optimization of the probe’s design and/or addition of the competition probe was required to achieve optimal SNS discrimination, especially for the targets with high GC content. In this study, we demonstrated the effect of the target sequence (the presence of cytosine clusters and type of SNS) on the signal turn-on ratio and SNS discrimination ability of the probe. We also explored a repertoire of strategies to overcome target-associated challenges and highlighted the strategies for the design of efficient sPDz probes for SNS discrimination analysis. The findings are summarized into empirical guidelines for rational design of the sPDz probes.

RESULTS AND DISCUSSION

Target selection and probe design.

Generally, design of a split hybridization probe for discrimination of SNS is straightforward: the sequence of a signal reporter is divided into two halves, which are elongated with the target-complementary fragments to ensure formation of the signal-generating structure in the probe-target complex.¹⁵ The sequences of the signal reporter fragments are kept unchanged for all the targets to be interrogated, with only the target-interacting fragments tailored to the target’s sequence. In case of the sPDz probes, there is some degree of freedom in the sequence of signal-generating fragments: the twelve guanine residues of the G4-forming sequence can be distributed equally to both strands of the probe, or one of the probe may have more G-triplets than another. This G-rich sequence can abruptly interact with the target to disturb the probe’s performance, especially if a target is G- and/or C-rich, which is often the case for targets of bacterial origin.⁵¹ Varying the number of G-triplets in the probe’s strands can provide a simple solution for the improved performance of the sPDz probes. Here, we focused on the targets with two or more consecutive C residues in the fragments complementary to one or both strands of the sPDz probe.

Four practically significant model nucleic acid targets were chosen (Supplementary Table 1). Three target sequences (T1-T3) corresponded to the gene fragments of mycobacterial species (M. tuberculosis and M. abscessus) with point mutations conferring antibiotic-resistant bacterial phenotypes. Bacterial species belonging to M. tuberculosis (Mtb) complex are the main causative agent of tuberculosis. Species of M. abscessus (Mabs) complex can cause pulmonary disease, and skin and soft tissue infections.⁵² Targets T1 and T2 corresponded to the fragments of the katG and rpoB genes of Mtb, respectively. Point mutations in an 81-bp fragment of the rpoB gene are responsible for 95% cases of resistance to a first line anti-tuberculosis drug, rifampin.⁵³ Point-mutations in the katG gene contribute to resistance to another first line antibiotic – isoniazid.⁵⁴ T1 contained the “wild type” sequence of codon 315 of the katG gene (a genotype causing isoniazid-susceptible phenotype). T1C was a “mutant” corresponding to isoniazid-resistant phenotype with an AGC>ACC SNS.⁵⁴ T2C contained a genome fragment from rifampin-susceptible bacteria (with CAC sequence at codon 526), while T2T represented a sequence with a CAC>TAC substitution (Supplementary Table 1).⁵⁴ Targets of the T3 group contained a sequence of the M. abscessus rrl gene with four possible nucleotides at each of the positions 2058 and 2059 (E. coli numbering). Target T3 contained adenine residues at both positions and represented the drug-susceptible genotype. Any nucleotide substitution at each potion renders bacteria resistant to a macrolide antibiotic clarithromycin.⁵⁵ Six targets (T3GA, T3TA, T3CA, T3AG, T3AT, or T3AC) corresponding to all possible SNS-containing drug-resistant genotypes were used in our experiments (Supplementary Table 1). The last target corresponded to a fragment of segment 7 from the genome of the Influenza A virus (IAV). T4 mimicked H1N1 subtype of IAV, which is one of the most prevalent virus genotypes infecting humans.⁵⁶ H3N2, the second most common IAV subtype circulating in humans,⁵⁷ was represented by T4T. T4A corresponded to a fragment from H5N1 genome (Supplementary Table 1). In this case, differentiation between the SNSs in target T4 would allow IAV subtyping rather than drug susceptibility testing.

The targets were selected based on the presence of three or more consecutive cytosine residues, in or near the site complementary to the arm of strand S, as well as the type of SNS (Supplementary Table 1). (i) Target T1 had no apparent clusters of cytosine (C) residues, and there was a G>C substitution in T1C resulting in a C-C mismatch in the probe-T1C complex. We anticipated no problems with the probe’s selectivity and/or turn-on properties. (ii) Target T2 contained a C-cluster in the middle of the fragment complementary to strand S, and the SNS site was within the cluster. Taking into account the G-rich sequence of the sPDz probe, we hypothesized interfering target-probe interactions involving the C-cluster. In addition, a C>T SNS in a mismatched target T2T would cause a G-T wobble base formation between the mismatched target and strand S, thus compromising the probe’s selectivity. (iii) Target T3 had a C-cluster flanking the fragment interrogated by strand S (shown in grey italics in Supplementary Table 1). It could allow an extended hybrid between the target and a G-triplet of strand U, affecting signal intensity and/or selectivity. There were two adjacent SNS positions, with adenine changed into all three possible SNSs in each of the two positions of mismatched targets. Along with a challenge of discriminating the SNSs at both position using the same probe, two mismatched targets (T4GA and T4AG) would form G-T wobble base pairs with strand S, which would compromise the probe selectivity. (iv) Finally, target T4 also did not contain apparent C-clusters with no problems with the probe performance envisioned. It was mainly used as a “case study” to experimentally verify an algorithm for the probe design suggested based on the findings from the data analysis for T1, T2, and T3.

Four sets of sPDz probes targeting T1, T2, T3, and T4 were designed. The probes interrogated 26–34 nt target fragments and contained the arms of strands S and U fully complementary to their correspondent targets. To ensure SNS discrimination, the target recognition fragment of the SNS-discriminating strand S (Scheme 1) was designed to be 8–9 nt long (T_m=19–36°C). This is long enough to enable the formation of a stable complex between the sPDz probe and the fully complementary target at room temperature. At the same time, if a single nucleotide in the target was mis-paired with the probe, strand S would fail to bind to the target, thus preventing the G4 structure from assembling (Scheme 1, left).

The arm of strand U was chosen to be longer (17–24 nt, T_m=56–66°C) to tightly bind to the target and assist in unwinding of its secondary structure (Supplementary Figure 1). For successful design of strand U, it should be complementary to a single-stranded target fragment, which serves as a “toe-hold” for strand U binding to the target. In all sPDz designs that recognize the same target, the arms were the same (Supplementary Tables 2–5).

(1). Mode of the G4 core splitting.

The sequence of the G4 signal transducer was distributed between strands U and S in three different ways (Table 1): (i) each strand of the probe contained two G-triplets (6:6 split, probe designation P^6:6); (ii) one G-triplet was placed in strand U, while strand S had three triplets (3:9 split, probe designation P^3:9); or (iii) there were three and one G-triplets in strand U and strand S, respectively (9:3 split, probe designation P^9:3). Correspondent strands U and S are designated according to the number of the G-residues they contain indicated in superscript (Table 1). For most of the probes, we explored all three G4-splitting patterns. Depending on the number of consecutive C residues in the target, the probes with the same arms but different splitting of the G4 core differed not only in the background and signal intensity but also in selectivity. Therefore, on a quest for the most optimum sPDz probe, it is worth considering different ways of G4 core splitting, and either symmetric or asymmetric splitting mode can be advantageous depending on the target sequence.

Table 1.

Splitting of the G4 core in the sPDz probes^a

Splitting mode	Number of G-triplets in the signal transducing elements
	Strand U	Strand S
asymmetric, 3:9 (P3:9)	One G-triplet (3U)	Three G-triplets (9S)
symmetric, 6:6 (P6:6)	Two G-triplets (6U)	Two G-triplets (6S)
asymmetric, 9:3 (P9:3)	Three G-triplets (9U)	One G-triplet (3S)

^aThe notations used for the probes and probe strands are given in parenthesis. Full sequences of the probe strands are listed in Table S2.

(2). Linkers between the G4-forming and target-recognizing fragments.

The most cost-efficient option is to use unmodified oligonucleotides as probe components. In this case, the two elements of each probe strands would be connected via a phosphodiester bond, with an option to include a bridging nucleotide as a spacer. At the same time, a flexible non-nucleotide linker between these elements can aid to the formation of a stable G4 structure, while having the probe tightly bound to the target.³⁵ This can result in higher signal intensity, but compromise SNS discrimination ability. Here, we experimentally verified this hypothesis.

(3). Conformational constraint in strand S.

One strategy to increase the selectivity of a hybridization probe is to introduce a conformational constraint in the form of a stem-loop on one or both strands of the probe.^4,7 For “signal-on” probes, such a constraint may improve selectivity and reduce the background signal, but may also reduce the signal in the presence of the desired target. For the sPDz probes, the conformational constraint, if not formed intrinsically, can be introduced by adding a C-rich “tail” to the arm(s) of the strand(s). It would help sequester one or more of the G-triads into a stem-loop structure in the absence of the target. In this work, the advantageous effect of the “tail” in strand S was experimentally tested.

Performance of the sPDz probes

Performance of the sPDz probes was evaluated based on the color intensity for the samples containing the probe components in the absence of a target (blank samples, background intensity), or in the presence of either a fully matched (specific signal intensity) or SNS-containing (non-specific signal intensity) target. The color intensity was visually monitored and quantified by measuring the absorbance of the samples at 420 nm (one of the maxima for ABTS^−. absorbance).⁵⁸ The following characteristics of the probes were determined. Signal-to-background ratios (S/B) were calculated by dividing the absorbance for the target-containing sample by the blank absorbance (sPDz in the absence of the target). Higher S/B values predispose the probe to have lower detection limits. In addition, selectivity of the probes was tested by comparing the signal triggered by the fully complementary target with that by an SNS-containing target. A selectivity factor (SF) was calculated using the following equation:

$$SF=\left[1-\frac{A_{SNS}-A_0}{A_{wt}-A_0}\right]\text{*}100\text{\%},$$

where A_wt, A_SNS, and A₀ are values for absorbance at 420 nm for the samples containing the probes in the presence of the fully complementary target, SNS-containing target, and in the absence of the target, respectively. Based on the equation, the selectivity factor can be between 0%, when a probe fails to discriminate SNSs, and 100%, when the SNS-containing target does not trigger the signal above the background. By comparing the absorbance values for the signal of the sPDz probes with visual observation of the color intensity, we empirically derived an S/B≥3 and SF≥90 as criteria which would indicate an excellent probe performance. For clear visual observation of the color change, absolute absorbance of the target-containing and blank samples should be higher than 0.8 and less than 0.3 o.u., respectively. We further referred to the probes with an S/B≥3 and SF≥90 as “well-performing” ones (Tables 2 and 3, values in bold).

Table 2.

Performance of sPDz probes of the P1 and P2 sets^a

Probe	P1 set
	P16:6	P13:9	P19:3	P16:6-teg	P13:9-teg	P19:3-teg
S/B	4.26±0.05	4.5±0.1	5.8±0.1	5.42±0.04	4.0±0.1	10.47±0.06
SFb	98	100	99	90	51	98
Probe	P1 set with a stem-loop
	sl-P16:6	sl-P13:9	sl-P19:3	sl-P16:6-teg	sl-P13:9-teg	sl-P19:3-teg
S/B	1.82±0.03	4.63±0.05	1.31±0.03	3.84±0.06	6.78±0.08	3.32±0.04
SFb	100	99	69	97	91	95
Probe	P2 set
	P26:6	P23:9	P29:3	P26:6-teg	P23:9-teg	P29:3-teg
S/B	1.22±0.03	0.9±0.1	1.21±0.02	1.31±0.08	ND	1.92±0.05
SFb	65	66	100	30	ND	66
Probe	P2 set with a stem-loop
	sl-P26:6	sl-P23:9	sl-P29:3	sl-P26:6-teg	sl-P23:9-teg	sl-P29:3-teg
S/B	1.49±0.04	0.8±0.3	1.28±0.07	6.95±0.08	1.1±0.4	3.22±0.08
SFb	24	0	16	73	0	70

^aThe values for S/B and/or SF indicating poor performance of the probe are in grey. Acceptable probes are bold. “ND” stands for “not determined”;

^bIn case of a mathematical value for SF>100%, the SF was made equal to 100%. When the signal triggered by the mismatched target was higher when the signal in the presence of the fully complementary target, the SF was 0.

Table 3.

Performance of sPDz probes of the P3 set^a

Probe	S/B	SFb
		SNS at nt2058			SNS at nt2059
		G	T	C	G	T	C
P36:6	5.8±0.3	26	39	87	50	24	86
P33:9	5.2±0.2	0	0	11	2	3	10
P39:3	7.5±0.2	72	80	100	87	71	100
P36:6-teg	5.3±0.1	0	ND	ND	0	ND	ND
P33:9-teg	11.81±0.05	3	ND	ND	0	ND	ND
P39:3-teg	10.3±0.2	22	ND	ND	24	ND	ND
sl-P36:6	12.1±0.2	75	95	100	90	96	100
sl-P33:9	11.48±0.08	10	ND	ND	20	ND	ND
sl-P39:3	9.5±0.3	100	98	99	100	97	100
sl-P36:6-teg	8.4±0.2	89	ND	ND	95	ND	ND
sl-P33:9-teg	16.94±0.04	43	ND	ND	65	ND	ND
sl-P39:3-teg	11.6±0.2	97	ND	ND	100	ND	ND

^aThe values for S/B and/or SF indicating poor performance of the probe are in grey. Acceptable probes are bold. “ND” stands for “not determined”;

^bIn case of a mathematical value for SF>100%, the SF was made equal to 100%. When the signal triggered by the mismatched target was higher when the signal in the presence of the fully complementary target, the SF was 0.

Target without apparent cytosine clusters.

The sequence of the target T1 representing a fragment of the Mtb katG gene does not have more than two consecutive C residues in the probe-interrogated region (Supplementary Table 1). Therefore, we expected the simplest design for the PDz probes (without non-nucleotide linkers and conformational constraints in the probe strands) being efficient in interrogating the target and discriminating the G>C substitution. As expected, the P1 probes exhibited excellent selectivity disregarding the G4 core-splitting mode (Figure 1, panel a, Table 2). For the three probes, the signal triggered by SNS-containing T1C was at the background level. The excellent SNS discrimination ability of the P1 probes (with SF of 98–100%) is contributed to by the changes in the secondary structure in the fragment complementary to strand S caused by the G>C substitution (Supplementary Figure 1). In T1C, the SNS site is in a stem, which further stabilizes the dissociated state of the target in comparison with the associated state when the target is in complex with strands U and S.

Figure 1.

Performance and selectivity of the sPDz probes interrogating a fragment of Mtb katG gene. a) Probes P1^6:6, P1^9:3 and P1^3:9 in the absence of the targets, or in the presence of the specific target T1 or SNS-containing T1C. b) Proposed interactions between T1 and either ⁶U1, ⁹U1, or ³U1. The structures are drawn as predicted by NUPACK (https://www.nupack.org). The G4-forming nucleotides are shown in green, and the arm of strand U is in blue; the SNS position is highlighted in orange. c) Probes P1^6:6, P1^9:3 and P1^3:9 pre-annealed in the absence or presence of T1 or T1C before the signal generation step. d) Complex of T1 with ⁹U1^-teg (top), and performance of P1^6:6-teg, P1^9:3-teg and P1^3:9-teg in the absence of presence of T1 or T1C (bottom).

Even though P1^6:6 produced the lowest signal, all three probes enabled reliable visual detection of T1 (Figure 1, panel a), with S/B in the range of 4–6 (Table 2). Somewhat lower intensity of the T1-triggered signal for P1^6:6 and P1^9:3 relative to P1^3:9 can be attributed to an extended hybrid between the target and strand ⁶U1 or ⁹U1 beyond their arms, which is absent in case of ³U1 (Figure 1, panel b). This partially sequesters the G4-forming nucleotides and likely interferes with T1-strand S interaction, thereby decreasing the target-induced signal. Indeed, annealing of strands U and S with T1 prior to the peroxidation reaction resulted in the same signal intensity for all three probes regardless of the G4 splitting mode (Figure 1, panel c). The same effect was observed in the case of prolonged incubation of strands U and S with T1 prior to the peroxidation reaction (data not shown), as opposed to immediate signal generation (Figure 1, panel a). Our data agrees with similar signals observed for symmetrical and asymmetrical designs of the split G4-based probes reported previously.⁴⁸ The extended hybrid formation can be impeded if a triethylene glycol (teg) linker is placed between the G4-forming sequences and “arms” of the probes (Figure 1, panel d, top). We hypothesized that the linker would prevent the formation of a T-G wobble base-pair, which otherwise contributes to the extended hybrid stabilization. Thus, it would be easier for strand S to interact with the target complexed with teg-containing rather than teg-free strand U. Indeed, for all teg-containing probes, the T1-induced signal was more intense than for their linker-less counterparts (Figure 1, panels a and d). This agrees with the previously reported improvement of the target-specific signal for the 6:6 sPDz probe upon teg-linker incorporation.³⁵ For the P1^3:9-teg probe, however, improvement of the T1-induced signal was not significant, while the non-specific signal in the presence of T1C increased to the point that the color was clearly observed (Figure 1, panel d), which resulted in about 2-fold drop in the selectivity factor (Table 3).

Probe P1^3:9 had higher background than the other two probes. This most likely reflects the folding of strand ⁹S into a G4-structure with at least two planes of G-tetrads (Supplementary Figure 2, panel a), which is supported by the circular dichroism (CD) spectroscopy analysis (Supplementary Figure 2, panel b). The background can be decreased by introduction of a conformational constraint in the form of a stem-loop (sl) into one or both strands of the probe.^16,59,60 Therefore, 4–5-nt C-rich “tails” complementary to a portion of the G4-forming sequences were introduced next to the arm of strands U and S to make probes sl-P1^6:6, sl-P1^3:9, and sl-P1^9:3 (Supplementary Table 2). As expected, the background for sl-P1^3:9 decreased to that same level as was observed for the probes P1^6:6 and P1^9:3 (Supplementary Figure 3, compare with the data in Figure 1, panel a). At the same time, the stabilization of the dissociated state of the probe compromises the interaction between the probe and desired target. Indeed, decreased signal was observed for all constrained probes (sl-P1^6:6, sl-P1^3:9, sl-P1^9:3), with sl-P1^9:3 showing no signal higher than the background (Supplementary Figure 3). The signal could be partially restored if the conformational constraint introduced together with a teg-linker between the G4-forming fragments and arms (Supplementary Figure 3, sl-P1^u:s-teg probes).

In several previously reported sPDz probes,^37,38,40,41 the target-recognizing elements were designed to leave a one-nucleotide gap separating them. Therefore, we tested if this gap affects the performance of P1^9:3 probe. For this purpose, strand U was “moved” one nucleotide away from strand S, so that the probe-target complex had a C residue in the target free from interaction to either of the strands (Supplementary Figure 4). The gap resulted in a slight drop in the specific target-induced signal, while the high selectivity of SNS differentiation was retained (Supplementary Figure 4, panel c).

In summary, for a targeted nucleic sequence without three or more subsequent C-residues and with an SNS other than C>T or A>G, any mode of the G4 splitting would produce a well-performing teg-free sPDz probe. The presence of a cluster of two C residues in the target’s fragments complementary to strand S may increase the signal intensity for P^3:9 both in the absence and presence of the target due to folding of strand ⁹S into a catalytically active two-tiered G4 stricture. The teg-linker improves the S/B through background reduction but increases the commercial cost for the probe strands. In the case of T1 analysis, probe P^9:3 seems to provide the optimal cost-efficiency. Similar designs have been used previously for visual detection and/or discrimination of point-mutations in nucleic acid targets. In the reported cases, however, probe selectivity down to single mismatch was achieved only due to the addition of competition sequences.^40,41

SNS site within a cluster of cytosine residues.

In some cases, the presence of several consecutive C residues in the probe-interrogated fragment of the targeted nucleic acid is unavoidable. For example, when an SNS site is within or next to a C-cluster. To demonstrate approaches for the sPDz design in this case, we used a model target T2 with a CCC-sequence and C>T substitution of the last nucleotide of the C-triplet in its SNS-containing target T2T (Supplementary Table 1). The SNS site was located next to a stable stem (Supplementary Figure 1), which is generally considered a challenging target for interrogation by hybridization probes.^9,11 To be able to unwind the target, strand U was designed to have a fragment complementary to the stem-loop forming nucleotides of T2, as well as beyond the stem toward the 5’-end of the target (Supplementary Figure 1). In this case, the intramolecular interactions in the arms of strand U (Supplementary Figure 5, panel a) may hinder its intermolecular interactions with the target.¹¹ At the same time, minimum energy secondary structure of strand U reveals an intrinsic conformational constraint that can help in reducing the background. To preserve the conformational constraint while mitigating undesirable intramolecular interactions, we introduced a G>A substitution in strand U (Supplementary Table 3 and Supplementary Figure 5, panel b), which would not significantly compromise the strand-target hybridization due to the length of the hybrid. The same substitution was preserved strands U of all the probes in the P2 set. Because of the intrinsic complementarity between the arms and G4-forming fragments, even the “open” sPDz probes (designated as P2^6:6, P2^3:9, P2^9:3, P2^6:6-teg, P2^3:9-teg, or P2^9:3-teg in Supplementary Table 3) utilized a conformationally constrained strand U.

The presence of a C-triplet in the fragment of T2 interacting to strand S (Supplementary Table 1) can cause several issues in the design of the sPDz probe. First, for the P2^3:9 probe, due to an additional G-triad in the arm (complementary to the CCC-stretch of T2), folding of strand ⁹S2 into a unimolecular G4 structure is expected, which, in case of parallel G4 topology (Supplementary Figure 6, panel a), would cause high signal of the probe disregarding the target presence. This hypothesis was experimentally confirmed by CD analysis (Supplementary Figure 6, panel b) and in the colorimetric assay for G4 peroxidase-like activity (Supplementary Figure 6, panel c). If “closing” of ⁹S2 is implemented to over compete the G4 formation, a very stable stem-loop structure would be required, which may compromise probe interaction with the target. Previously,⁶¹ a stem-loop structure with ΔG ~ −12 kcal/mol was needed to inhibit a thermodynamically favored unimolecular G4 structure. Second, the presence of the C-triplet in T2 makes the formation of a target-strand U extended hybrid highly probable for both P2^6:6 and P2^9:3 probes (Figure 2, panel a). This hybrid would mitigate the target-specific signal for the indicated probes, which was experimentally confirmed (Figure 2, panel b). Indeed, no visible difference between the color intensities of the samples containing the probe alone or in the presence of either T2 or T2T target was observed for either probe. As expected, introduction of a teg-linker resulted in an increase of the signal intensity for all the samples (blank, specific and non-specific signal) with P2^6:6 and P2^9:3 probes (Figure 2, panel c), resulting in the SF of 30 and 66, respectively. For both high intensity of the target-specific signal and acceptable selectivity, simultaneous presence of the non-nucleotide linker and the “closing tail” was required (Figure 2, panel d). In this case, only the P2^6:6 probe exhibited an appreciable color change in the presence of the specific target (Figure 2, panel d). Due to the presence of a wobble base-pair in the T2T-probe complex, a slight increase in absorbance triggered by T2T was observed, resulting in a SF of 75 (Table 2). Nevertheless, visual differentiation between T2 and T2T was achieved (Figure 2, panel d). In the case of the sl-P2^6:6-T probe specific to T2T, the specific target triggered the signal of S/B=7.6 ±0.1, while the non-specific signal was at the background level, resulting in a SF of 97 (Figure 2, panel e). Target T2 corresponding to a fragment of the rpoB gene from Mtb is among the most challenging targets to interrogate with the sPDz probe. Previously,⁴¹ discrimination between SNSs in rpoB targets has been achieved only in the presence of the competition probe that would bind to the mismatched target, thus mitigating the non-specific signal.

Figure 2.

Performance and selectivity of the sPDz probes interrogating a G/C-rich fragment of Mtb rpoB gene. a) Proposed interactions between T2 and either ⁶U2 or ⁹U2 strands. The structures are drawn as predicted by NUPACK (https://www.nupack.org). The G4-forming nucleotides are shown in green, and the arm nucleotides of strand U are in blue; the SNS position is highlighted in orange. b) Probes P2^6:6 and P2^9:3 in the absence of the targets (“Blank”), or in the presence of the specific target T2 or SNS-containing T2T. c) Probes P2^6:6-teg and P2^9:3-teg in the absence of the targets, or in the presence of the specific target T2 or SNS-containing T2T.D. Probes sl-P2^6:6-teg and sl-P2^9:3-teg in the absence of the targets, or in the presence of the specific target T2 or SNS-containing T2T. e) Probe sl-P2^6:6-teg-T designed to specifically recognize T2T in the absence of the targets, or in the presence of either T2 or T2T.

To conclude, for a target with an SNS within or immediately next to a C-cluster, the sPDz probe with strand S containing three G-triplets will likely to fold into a G4 structure, thus generating high background that could not be mitigated by introduction of a stem-loop constraint into the strand. Such catalytically active monomolecular G4 structure can be also formed if G-triplets are brought in proximity as a result of a stem-loop formation in a strand, which has been previously demonstrated.^49,62,63 A non-nucleotide linker will be beneficial to prevent formation of a too stable extended hybrid between strand U and the target. For the probe with symmetric G-core splitting, it is less probable that the extended hybrid formation completely prevents the target to bind to both probe strands instead of just strand U, but this would depend on other nucleotides in the sequence of the fragment interrogated by strand S.

Clusters of cytosine residues near the SNS site.

In the structure of T3 (Supplementary Table 1), there is a long (4 nt) stretch of C residues downstream the SNS sites. It is expected that the stretch is too far from the junction between the strands of the sPDz probe to cause the formation of a monomolecular G4 structure by ⁹S3 strand. At the same time, the presence of the stretch can affect the performance of the probes containing strands ⁶U3 and ⁹U3 with the G4-forming fragments long enough to reach to the C4 stretch on either the fully complementary or SNS-containing target (Supplementary Figure 7, panel a). This extended hybrid is less likely in case of strand ³U3, so the P3^3:9 probe is expected to have poor selectivity, which was experimentally verified (Supplementary Figure 7, panel b). Between the probes P3^6:6 and P3^9:3, the extended hybrid of T3 with ⁹U3 is expected to be slightly more stable than that with ⁶U3 due to the presence of additional T and G residues of strand U that can base-pair with hanging C and G residues of the target (though not predicted by NUPACK⁶⁴ algorithm). Therefore, the signal (both specific and non-specific) for P3^6:6 should be higher than that for P3^9:3, experimentally confirmed (Supplementary Figure 7, panel b). This significantly compromised selectivity of both P3^6:6 and P3^9:3. Out of six SNS-containing targets used (three for each of the two SNS positions – nt 2058 and nt 2059), only T3CA and T3AC did not stabilize the catalytic G4-core well enough for visually observed color even for more selective P3^9:3 (Supplementary Figure 7, panel b). This can be attributed to even more stable hybrid between strand U and T3CA or T3AC than between strand U and the fully complementary T3 (Supplementary Figure 7, panel a). By introducing a 5-nt stem in strand ¹S3, it was possible to achieve excellent selectivity and high S/B value simultaneously (Figure 3, panel a, Table 3). For the conformationally constrained probe sl-P3^6:6, though, even the presence of the “tail” to stabilize the probe’s dissociated state did not help discriminate an A>G mismatch (Figure 3, panel a), due to the wobble-base in the S-target complex.

Figure 3.

Performance and selectivity of the conformationally constrained sPDz probes interrogating targets T3 and T4. a) Probes sl-P3^6:6 and sl-P3^9:3 in the absence of the targets (“Blank”), or in the presence of the specific target T3 or SNS-containing targets indicated. b) Probes sl-P4^3:9, sl-P4^3:9-T, and sl-P4^3:9-A are designed to be complementary to the targets T4, T4T, and T4A, respectively. The targets represent genomes of IAV subtypes H1N1, H3N2, and H5N1, respectively. Absorbance at 420 nm and tube images are shown for the samples containing the probes in the absence of the targets (“Blank”), or in the presence of the correspondent specific target or one of the SNS-containing targets indicated. Images of the sample tubes are taken immediately before absorbance measurement.

Characterization of the most optimal designs

Characterization of the most optimal designs of probe sets P1 and P2 was performed in terms of the analysis of the signal intensity dependence on the target concentration, calculation of the limit of detection (LOD) and limit of quantification (LOQ), as well as linear dynamic range (Supplementary Figure 8, panels a–d). The signal intensity increases with the concentration of the probe-specific target and then slightly declines after it exceeds the concentration of the probe. This phenomenon can be explain by the increase in abundance of the catalytically inactive complexes of the target with only one strand of the probe. The linear dynamic range was determined to be 0–500 nM, with LOD and LOQ in the range of 65–75 nM and 195–206 nM, respectively (Supplementary Figure 8, panels b and d). This concentration range is not low enough for the SPDz probe to be able to detect targets in biological fluids without target amplification. Successful use of the sPDz probes in combination with PCR^37,41 or isothermal amplification⁴² has been previously reported.

Empirical Guidelines for the Design of the sPDz probe.

Based on our results and literature data, we summarize possible complications in the probe performance depending on the presence of clusters of G- or C-residues in the interrogated target fragment (Figure 4, panels a–c). Previously, DNA duplex-assisted formation of a G4 structure,^62,63 as well as G4 tolerance to the presence of additional nucleotides and even duplexes in the G4 loop portions,⁴⁹ have been demonstrated. In case of split sPDz probes, if a target has a cluster of three or more G-residues in the fragment complementary to strand S, binding of the target to strand U can form a duplex with overhang G4 structure (Figure 4, panel a). As a result, even an SNS-containing target would support G4 formation resulting in high signal and lack of SNS discrimination. Targets with clusters of three or more C-residues can also trigger complications in the probe design. For example, if a probe’s strand recognizing the C-rich fragment of the target is designed with three G-triplets, it would fold into a unimolecular G4 structure (Figure 4, panel b), regardless the target’s presence. Extremely high background for such probes was observed in this work (Supplementary Figure 6). We also showed that even in case of two consecutive C-residues in the target, the complementary strand would fold into either a two-tiered or incomplete three-tiered G4 (Supplementary Figure 2) with catalytic activity compromising the background (Figure 1, panel a). Finally, formation of an extended hybrid between a C-rich target and strand U beyond its arm is very likely, especially if strand U has two or three G-triplets (strands ⁶U or ⁹U, respectively) (Figure 4, panel d). This may decrease the signal triggered by both matched and mismatched target.

Figure 4.

Possible caveats and recommended strategy for the sPDz probe design. a) Formation of the catalytically active G4 structure by target-strand U interactions; b) Formation of the catalytically active G4 structure by one of the probe’s strands alone; c) Formation of the extended hybrid between the target and strand U resulting; d) Proposed algorithm for target-dependent selection of the G4-splitting mode for the probe design. In some cases (indicated by the asterisk,*), sub-optimal P^6:6 design could be also acceptable. For targets with no complications (indicated by **), all G4 splitting modes can give acceptable probe performance.

Taking into account the above-mentioned complications, we suggested an algorithm for the selection of the G4-splitting mode while designing the sPDz probe (Figure 4, panel d). The algorithm takes into account the presence of two or more consecutive C or G residues (C- or G-clusters, respectively) in the probe-interrogated target fragment. Even though in this work we studied the sequence effect for the targets with at least three C–residues, we included the cases of two consecutive C-residues in the algorithms, since folding into a parallel G4 structure with an incomplete third G-tetrad has been demonstrated. For the probe recommendation indicated with an asterisk (*), the algorithm suggests the most optimal G4 split, even though alternative splitting modes can give acceptable performance. For targets with no indicated sequence features, all modes of G4 splitting are acceptable (indicated with double asterisk, ** in Figure 4, panel d).

This algorithm was run through the sequences interrogated by the sPDz probes reported in the literature (Supplementary Table 6). For most of the sequences, the G4-splitting mode the authors selected is in agreement with one recommended by the algorithm. In one study,⁴⁰ a target with C-clusters in target fragments recognized by both strands U and S was interrogated with the P^9:3 probe instead of recommended P^6:6 to prevent folding of the strands into full or partial unimolecular G4 structures. Since in the reported case, stand U contained two G-doublets in addition to its G4-forming sequence, only an incomplete G4 structure was possible, which would not significantly compromise the probe’s performance. In another study,⁴¹ the P^9:3 instead of recommended P^3:9 was used to interrogate a target with a 4-nt G-cluster complementary to strand S. Therefore, DNA duplex-assisted G4 formation in target-strand U complex is expected, leading to low selectivity. To alleviate this problem, the authors used a competition probe complementary to the SNS-containing target and, at the same time, to a portion of the G4-forming sequence of strand U.⁴¹

Further suggestions and troubleshooting.

Through careful selection of the G4 splitting mode, it is possible to design a well-performing sPDz probe with no additional modifications. However, in some cases, the performance of this probe is still far from ideal. To optimize the probe, additional tools exist that can assist in improving the ability of the probe to discriminate SNS and/or increasing the intensity of the target-specific signal. A list of suggested troubleshooting options is as follows.

1. Low signal. Nucleotide or non-nucleotide (e.g. teg) linkers between the arms and the G4-forming fragments of the probe may decrease the steric hindrance inherent to multi-stranded complexes, such as the probe-target complex of the sPDz probe. This would facilitate G4 formation, thereby increasing overall signal. However, the loss of such hindrance may reduce selectivity.

2. High background may be reduced by extending probes with C-rich sequences that are not complementary to the analyzed target. These sequences will form stem-loop structures with the G4-forming regions, thus serving as conformational constraints that stabilize the dissociated state. This would prevent association of G4-forming elements of strands U and S in the absence of the target.

3. Low selectivity. Constraining the dissociated conformation of the strands due to a stem-loop formation described for the previous problem of high background can be helpful in mitigating low selectivity. The length of the stem in the conformational constraint can act as a fine-tuning nob to maximize signal while maintaining selectivity. If the target features a G-rich region, such conformational constraint may be intrinsic to the target-binding region, thereby compromising signal intensity. Another tool is to take advantage of an intrinsic extended hybrid between the target and strand U to optimize the ability of the probe to discriminate SNSs.

“Case study” to design an efficient sPDz probe for an arbitrary sequence

“Case study” to design an efficient sPDz probe for an arbitrary sequence was performed to check if the above guidelines are sufficient to design an optimal probe for SNS discrimination with minimal optimization of the probe structure. We used an arbitrary target T4, which corresponds to a fragment of IAV genome (subtype H1N1). The 5’-terminal fragment of the target contains several single nucleotide positions that are substituted in the genomes of subtypes H3N2 or H5N1 (Supplementary Table 1, targets T4T and T4A, respectively). These SNSs can be interrogated for IAV genotyping and subtype identification, which is important for epidemiological purposes.⁵⁸ For the probe design, we selected the SNS site closest to a fragment constant between the IAV subtypes (Supplementary Table 1, the targeted SNS position is in red). In this case, it is possible to use the same strand U (the strand that is not designed to interact with the SNS) for all virus subtypes, and subtype identification would rely on the target interaction with a set of subtype-specific strands S.

The selected target T4 folds into a secondary structure with two hairpins and the probed SNS site in the middle of a single-stranded region (Supplementary Figure 1). To ensure unwinding of the secondary structure, we designed strand U4 to bind to a loop region and one of the stem strands. The 5’-terminus of the arm of strand U4 is complementary to a portions of a complementary stem strand to increase the probe-target affinity. The arm of strand S was designed complementary to the single-stranded fragment of T4 containing the SNS site.

There are no C- and/or G-clusters in the strand S-complementary fragment of the target. However, there is a cluster of three cytosine residues in a T4 fragment interacting with the 5’-terminal portion of strand U4. The algorithm suggests the use of the P^3:9 probe as a starting point. We also tested P^9:3 and P^6:6 designs to confirm that the algorithm selection was indeed the most optimal design. By analyzing the folding of just the arm of strand U4, it can be predicted that strand ⁹U4 would likely form a monomolecular G4 structure due to the presence of a stem-loop element (Supplementary Figure 9, panel a). This structure itself could trigger the color change, thus causing too high background signal (Supplementary Figure 9, panel b). Both strands ⁶S4 and ⁹S4 would have an intrinsic stem-loop formed between the arms and G4-forming nucleotides of the strands, but the conformational constraint on ⁹S4 is more stable (Supplementary Figure 9, panel c), which would result in a less intensive signal but lower background and higher selectivity for P4^3:9 relative to P4^6:6, which was proven experimentally (Supplementary Figure 9, panel d). Therefore, based on the rational considerations according to our summarized guidelines, probe P4^3:9 was proven to have the most optimal design for a T4-specific sPDz probe. Similar 3:9 G4 splitting-based design was proven efficient for the T4T- specific sPDz probe (Supplementary Table 7). To discriminate A>G SNS in case of T4A-specific sPDz probe (Fig. 4B), “closing” of strand S into a stable (ΔG= −3.8 kcal/mol) stem-loop structure (Supplementary Table 7) was required.

CONCLUSIONS

Our findings indicate that there is no particular design of split G4 peroxidase-like deoxyribozyme probes that ensures optimal probe performance disregarding the target sequence. The most challenging targets to be interrogated are those with clusters of cytosine residues. Varying the distribution of G-triads may prevent excessive binding of the G4-forming fragments to the analyzed target, thus increasing the signal. Introducing non-nucleotide linkers for better flexibility may allow for an increase in signal in a similar fashion. Closing the G4-forming fragments into latent stem-loop structures reduces background and improves selectivity. Modification of the interior-loop sequences of the G4-forming regions may help reduce unwanted binding of one of the probe strands to the target, and therefore acts as a workaround for otherwise challenging targets. An algorithm for the selection of the G4-splitting depending on the target sequence features for optimal signal and SNS discrimination is suggested and supplemented with tips on how to achieve a well-performing probe using additional design elements. This algorithm can be used, in least at the initial step of the probe design, to eliminate the probe variants that would a priori cause either high signal or low selectivity. With more data on interrogation of specific targets with the sPDz probes, the algorithm should be further adjusted to strengthen the rational element in the probe design.

METHODS

Materials.

DNA oligonucleotides were purchased from IDT, Inc. and used without purification. Oligonucleotides were dissolved in DNase-free water to prepare ~100 μM stock solutions. The concentration of the stock solutions was corrected based on the absorbance of the solutions at 260 nm and the extinction coefficients provided by the vendor. ABTS and H₂O₂ were purchased from Sigma-Aldrich.

Testing and characterization of the sPDz probes.

Samples containing two probe strands (1 μM) in the absence or presence of correspondent DNA targets (1 μM) were prepared in a colorimetric buffer (50 mM HEPES-NaOH, pH 7.4, 50 mM MgCl₂, 20 mM KCl, 120 mM NaCl, 1% DMSO, and 0.03% Triton X-100) at 22 °C. Alternatively, probe strands were annealed to the target by heating at 95 °C for three minutes and slowly cooling to room temperature overnight. Hemin (375 nM), ABTS (1 mM), and H₂O₂ (1 mM) were added either immediately or following an overnight incubation at 22 °C to observe the development of a blue-green colour. Color intensity was quantified by measuring absorbance at 420 nm using a Thermo Scientific NanoDrop OneC UV-Vis Spectrophotometer (Waltham, MA, USA). Tube images were captured using a cell phone camera. For the experiments to determine limit of detection (LOD), limit of quantification (LOQ) and linear dynamic range for the probes, the concentration of DNA targets was varied (0–2 μM).

Circular dichroism spectra

Circular dichroism spectra were measured with a Jasco J-810 Spectropolarimeter. Oligonucleotides (2 μM) were prepared at room temperature in the colorimetric buffer. Measurements were conducted in a 4 mm-path cell at 22 °C, and data was obtained with a 2 nm slit width from 350 to 200 nm at 1nm intervals. CD spectra were averaged over five scans.