Evaluation of mismatch discrimination
There is no obvious ‘best way’ to measure and report mismatch discrimination. Every duplex with a single base mismatch can be compared to two different perfectly matched duplexes. For example, a probe-target duplex having an A•C mismatch could be compared with perfectly matched duplexes containing either an A•T or a G•C base pair at that position. Within this study, we used the average Tm
of both of these perfectly matched duplexes as the reference Tm
and defined the mismatch discriminating temperature difference, ΔmdTm
, as the difference between average melting temperature of the two possible ‘parent’ duplexes and the melting temperature of the mismatch duplex. Perfectly matched ‘parent’ duplexes are obtained when one of two bases at the mismatch site is replaced to create Watson–Crick base pair. For example, mismatch discriminating temperature difference for A•C mismatch is,
(A•C) is the melting temperature of the duplex with A•C mismatch. Tm
(A•T) and Tm
(G•C) are the melting temperatures of two perfectly matched duplexes when cytosine base of the mismatch base pair is replaced with thymine, and adenine base of the mismatch base pair is replaced with guanine, respectively. Mismatch discriminating free energy, enthalpy and entropy changes accompanying melting transitions can be similarly defined.
Mismatch discrimination of LNA probes
Sets of sequences are presented in . ‘Probe’ sequences contained LNA bases at the positions indicated and were hybridized to unmodified DNA oligomers. Sequence Sets 1, 2 and 3 contained centrally placed mismatches in the context of the trinucleotides -tx
a- and -ax
a-, respectively. For designs in Sets 1 and 2, each ‘Set’ includes 16 distinct duplexes comprising all possible match (4
) and mismatch (12
) base pairs at the ‘x
’ position. Data for sequence Sets 3 were obtained from Koshkin et al
). Because the probes have different numbers of base pairs and different sequence context within their mismatch positions, it was possible to discern ‘general effects’ that are not specific to a given sequence context.
Figure 1 Sets of DNA and LNA sequences. Mismatch sites ‘x’ are highlighted in red, DNA bases are lowercase, LNA monomers are uppercase and are underscored with a gray box for easy identification. Uppercase letter C denotes 5-methylcytosine LNA (more ...)
Examples of experimental melting profiles are plotted in for DNA–DNA and LNA–DNA duplexes. Melting profiles of all duplexes displayed single, cooperative transitions. Heating and cooling melting curves were reversible confirming that transitions were at equilibrium. illustrates that introduction of LNA modifications stabilizes the perfectly matched duplex more than the mismatched duplex (compare differences between 1p, 1m and 2p, 2m curves), which results in improved mismatch discrimination potential. Melting temperatures of more than 300 duplexes were determined and are reported in and and in the Supplementary Data. Mismatch discrimination varies with position of LNA residues and base identity. Significant differences were observed. For example, comparison of Sets 1DNA and 1T in shows that LNA modifications increase ΔmdTm value of an A•A mismatch from 8.4 to 12.3°C, while discrimination of a G•T mismatch decreases from 6.3 to 5.5°C.
Figure 2 Experimental melting profiles of perfectly matched C•G (1p) and mismatched A•G (1m) duplexes for native 1DNA Set are plotted with melting profiles of isosequential C•G (2p) and A•G (2m) duplexes of the 1T Set, which contains (more ...)
Melting temperatures, Tm, (°C) and mismatch discriminating temperature differences, ΔmdTm, of native and LNA-modified duplexes based on the sequence Set 1 in 1 M Na+ buffera
Melting temperatures, Tm, (°C) and mismatch discriminating temperature differences, ΔmdTm, of native and LNA-modified duplexes based on the sequence Set 2 in 1 M Na+ or 2 mM Mg2+ buffersa
Comprehensive analysis of discrimination for 12 kinds of mismatched base pairs is presented in where differences of mismatch discrimination between LNA-modified [ΔmdTm(LNA)] and unmodified DNA [ΔmdTm(DNA)] probes are graphically displayed. Neighboring bases can influence discrimination. The majority of mismatches exhibit positive or insignificant differences of ΔmdTm(LNA) − ΔmdTm(DNA) values, indicating that the addition of LNA bases either improves mismatch discrimination or is neutral. An important exception is the G•T mismatch where LNA modifications of guanine (at the mismatch) or even at neighboring residues are detrimental and decrease mismatch discrimination. This problem can be minimized if G•T mismatch assays are designed to detect the complementary strand, where the LNA probe contains the thymine residue of the mismatch. Alternatively, LNA nucleotides can be introduced two or more bases away from the mismatch guanine residue where they would at least not worsen mismatch discrimination.
Figure 3 Difference of ΔmdTm values between LNA and DNA probes for various mismatches. Sequence Set 1 (panel A and B), and Sets 2 and 3 (panel C) are plotted. Positions of LNA residues and set names are shown in Figure 1. A positive difference indicates (more ...)
To find the best modification pattern, Set 1T (triplet of LNA bases), 1M (LNA only at the mismatch site) and 1NN (LNA at nearest-neighbor bases) were compared ( and ). The largest ΔmdTm
(LNA) − ΔmdTm
(DNA) differences are consistently seen in the 1T Set. When the LNA triplet at the mismatch site is reduced to doublets (Sets 1MA and 1BM), the discriminatory power is generally reduced compared to the 1T Set. When only the mismatched nucleotide is LNA-modified (Set 1M), the best discrimination is achieved when the LNA-modified bases are purines. If nearest-neighbor residues are LNAs (Set 1NN), the largest discriminatory power is observed when probes contain a pyrimidine base at the mismatch site. These results agree with the genotyping study of Chou et al
), which employed LNA probes to detect a C•A mismatch resulting from 1691G > A mutation in the Factor V Leiden gene. They reported that LNA probes exhibited the highest discriminatory power when a nucleoide flanking cytosine residue of the mismatch site was modified. Examination of Sets 2T, 2M, 2MA and 2NN gives similar results ( and ) in spite of the fact that probes for these Sets employ a different sequence. The triplet LNA design (Set 2T) again has the highest discriminatory power. Mismatch discrimination for Sets 2M and 2NN follows similar dependence on purine and pyrimidine bases, although Set 2NN also shows improved mismatch discrimination for some purine mismatches. Therefore, with the exception of a G•T mismatch, the best discrimination is achieved with probes containing a triplet of LNA-modified nucleotides where the middle base of the triplet is at the mismatch site.
The beneficial effect of LNA modification decreases with distance. Set 1NNN places LNA residues at the next-nearest-neighbor position () and differences of ΔmdTm
values between LNA and DNA probes are mostly within the limits of experimental error. Therefore, the LNA modifications of next-nearest-neighbor nucleotides have little effects on mismatch discrimination. The thermodynamic effects of isolated LNA substitutions are local and do not extend beyond bases immediately flanking the modifications. Because similar ΔmdTm
values are observed for 1M and 1ET Sets, this is also true for probes where every third base of the probe is modified with LNA (). These results also suggest that the nearest-neighbor model is adequate to predict duplex thermodynamics for LNA-containing sequences. When the fraction of nucleotides modified with LNA is higher, e.g. Set 1ES where every second nucleotide is a LNA, synergistic energetic effects are observed; values of ΔmdTm
are different in 1M and 1ES Sets. Kierzek et al
) found LNA effects to be approximately additive when every second base is LNA-modified in 2′-O
-methyl RNA/RNA duplexes. Our results for LNA–DNA duplexes suggest that this additivity may break down when the density of LNA residues is lower. NMR experiments have shown (15
) that the structure of the duplex and sugar puckering are altered surrounding a LNA base and that this can impact thermodynamic parameters for both neighboring DNA and LNA residues. However, these structural effects are local and only the nucleotides immediately flanking a LNA modification are affected (15
). The most pronounced structural changes were seen for the nucleotide that follows the LNA residue in 3′ direction (15
). This is in agreement with our observations that Set 1MA enhances mismatch discrimination more than Set 1BM (). Thermodynamic and structural changes are related.
In summary, a triplet of LNA modifications positioned centered at the mismatch site gives the highest difference of melting temperatures between matched and mismatched LNA–DNA duplexes. As an alternative strategy, a LNA modification can be selectively placed at a purine nucleotide in a mismatch site and will also improve mismatch discrimination. If a pyrimidine base is present at the mismatch site, nearest-neighbor nucleotides flanking the pyrimidine base should be modified. An important exception to these rules is the G•T mismatch where any LNA modifications of the guanine nucleotide or either of its nearest-neighbor bases should be avoided.
Effects of Na+ and Mg2+ ions on duplex stability and mismatch discrimination
Because probe hybridizations are done in buffers of various composition, we studied whether different ionic environments could influence these results. Microarray experiments use buffers containing relatively low concentrations of sodium salts in the final wash steps. PCR and other enzymatic experiments commonly employ buffers where magnesium concentrations range from 1 to 5 mM. Therefore, several sequence sets were melted in a 69 mM Na+
) and in a 2 mM Mg2+
buffer and mismatch discrimination was compared with the original melting data collected in 1M Na+
buffer. plots the differences observed in mismatch discrimination between sodium and magnesium buffers. In most cases, the ΔmdTm
) − ΔmdTm
) values are negligible, i.e. the mismatch discrimination in Na+
buffers are similar for the same probes (see also and the Supplementary Data). Probes for the most destabilizing mismatches, A•C and C•C, exhibit slightly smaller discrimination power in the Mg2+
buffer than in the 1 M Na+
buffer. However, similar decreases of mismatch discrimination are observed for both unmodified and LNA-modified probes. Mismatch discrimination in 69 mM Na+
buffer is similar to mismatch discrimination in 1 M Na+
(data not shown). Therefore, the same guidelines for probe design can be applied in magnesium buffers as well as sodium buffers of various concentrations. Interactions of Na+
ions with probe-target duplexes are not significantly mismatch specific. The mode of binding to nucleic acids is comparable for both ions and depends essentially on backbone negative charge density, i.e. distribution of phosphate groups (25
). Since LNA residues do not introduce any significant charges, this result is not unexpected.
Figure 4 Differences of ΔmdTm values between 2 mM Mg2+ and 1 M Na+ buffers for unmodified and LNA-modified probes. Dashed lines denote the range (±0.8°C), which is within the experimental error of the measurements. See for sequences (more ...)
To further investigate effects of LNA modifications on counterion binding, perfectly matched duplexes from Sets 1DNA, 1T, 2T, 1ET, 1ES were melted in NaCl/phosphate buffers (19
) at sodium ion concentrations of 69, 119, 220 and 621 mM Na+
. plots dependence of melting temperatures on sodium concentrations. When the percentage of LNA-modified bases was increased from 0 to 47% in these sets, the duplex stability increased. The solid lines in were not obtained from fits to experimental data, rather they are calculated from published Tm
salt correction (19
) using melting temperatures in 1 M Na+
as a reference,
salt correction function was derived for native DNA–DNA duplexes and is observed to also apply to LNA–DNA duplexes, an important result that simplifies Tm
predictions for LNA duplexes in various ionic environments.
Figure 5 Effects of sodium ions on melting temperatures of DNA–DNA and LNA–DNA duplexes. Duplexes were investigated that had a T•A base pair in the X•Y mismatch site. All duplexes had the same base sequence, 5′-ggtcctttcttggtg-3′/3′-ccaggaaagaaccac-5′, (more ...)
Assuming a two-state melting transition, the number of sodium counterions released from the duplex, Δn
, can be estimated from the experimental Tm
is the ideal gas constant, and α is the correction term for the sodium ion activity coefficient and is about 0.92 at 100 mM Na+
. Transition enthalpies, ΔH
°, were estimated from fits of melting profiles to two-state model (26
). The values of Δn
calculated at 100 mM Na+
for T•A duplexes are 3.5, 3.5, 3.6, 3.6 for Sets 1DNA, 1T, 1ET, 1ES, respectively. The Δn
can be predicted from the published empirical relationship (19
), which was obtained from a large set of 92 duplex DNAs,
The predicted value of 4.1 is in reasonable agreement with experimentally measured Δn
values. The results show that LNA modifications do not change amount of counterions that are released when a duplex is melted.
Fluorescence of 2-aminopurine reveals changes in stacking interactions introduced by LNA modifications
Stacking interactions within the mismatch site were investigated using a 2-aminopurine base (p). Fluorescence of this base is sensitive to local environment and is quenched when the base is stacked to neighboring bases (27
). Magnitude of the fluorescence signal correlates with the relative extent of stacking interactions in duplex oligomers. The 2-aminopurine base is complementary to thymine (29
) and forms a hydrogen bonded base pair that has similar dimensions to Watson–Crick T•A base pairs.
DNA oligomer target 5′-gcgaggpggctt-3′ with a central 2-aminopurine base was hybridized to four different strands (). Fluorescence of the target oligomer and the four duplexes was determined. shows the results of UV melting and fluorescence measurements. Standard UV melting experiments demonstrate that a triplet of LNAs at the mismatch site stabilizes the duplex and increases mismatch discrimination in comparison with unmodified DNAs. Specifically, Tm difference between matched (t•p) and mismatched (g•p) duplexes is 9.5°C for native DNA oligomers and 15.9°C for the LNA probes. Fluorescence of the single stranded target (black line) was much larger than fluorescence of any of the duplexes, indicating that the 2-aminopurine base remains substantially stacked in all of the mismatched base pair combinations. The fluorescence spectra showed greater quenching for the DNA t•p base pair and the greatest quenching (and therefore base stacking) was observed for the LNA T•p base pair. Introduction of a mismatch increases fluorescence and decreases stacking interactions. The LNA-modified duplex with G•p mismatch resulted in more fluorescence than the native g•p DNA duplex, which suggests that the improved mismatch discrimination seen with LNA bases originates from subtle changes of base stacking interactions. We propose that LNA modifications may enhance stacking interactions of matched base pairs and suppress stacking interactions of mismatch base pairs relative to base stacking in native DNA duplexes.
Figure 6 Fluorescence emission spectra of target oligomer, 5′-gcgaggpggctt-3′, with single 2-aminopurine (p) reveal the magnitude of stacking interactions. Duplexes containing a H-bonded t•p base pair are compared with mismatched g•p (more ...)
LNA modification introduces an O
-methylene group, which ‘points’ into the minor groove (15
). A stabilizing water bridge between 2′ and 3′ oxygen atoms of the ribose moiety has been proposed for various 2′-O
). It is likely that the LNA modification alters the hydration pattern in the DNA minor groove. Stable water structure in the vicinity of the O
-methylene bridge might contribute to increased duplex stability. Perturbations of hydration patterns and a direct role for hydration energetics in discriminatory effects of LNAs have yet to be proven.
Mismatch discrimination is dependent on oligomer length and mismatch position
and show the effects that various mismatches have on Tm
values for 15 and 23 bp oligomers. Values vary with oligomer length and the impact of a single mismatch on overall duplex stability decreases with length. illustrates the dependence of mismatch discrimination on duplex length. Duplexes with a single g•t or a•c mismatch were examined, because these mismatches represent extremes, being the least and the most destabilizing mismatches, respectively. We assumed that average transition enthalpy and entropy of perfectly matched duplexes containing 50% g•c base pairs follow these relationships (22
which correspond to average values derived from the nearest-neighbor parameters (32
). shows that a single a•c mismatch always destabilizes the duplex more than a single g•t mismatch. Discrimination is strongly dependent on length for short oligomers and decreases as oligomer length is increased. Decrease of discriminatory power is very significant in the 10–25 bp length range. For example, the ΔmdTm
values for an a•c mismatch decrease from 16 to 4.5°C as duplex length is increased from 12 to 25 bp. Accurate predictions of mismatch discrimination can be calculated from the nearest-neighbor model for the particular DNA sequence and mismatch studied (32
). Since LNA substitutions increase duplex stability, LNA probes can be designed to be shorter than unmodified DNA probes at the same site, allowing use of shorter probes and further enhancing mismatch discrimination.
Figure 7 Dependence of melting temperature and mismatch discrimination on oligomer length. Average melting temperatures were calculated for 50% g•c DNA duplex oligomers that did not contain any mismatched base pairs as well as oligomers with single g•t (more ...)
Location of a mismatch within the probe sequence is also important. UV melting experiments have demonstrated (32
) that terminal and penultimate mismatches have much smaller discriminating power than internal mismatches. Similar results were obtained using DNA microarrays (33
). Duplexes with mismatches in the third to fifth position from the 5′ end were reported to have significantly lower Td
s than those with mismatches in the first or second position. If the mismatch site is located 3 or more base pairs from the ends of oligomer, nearest-neighbor model assumes that energetic effects of the mismatch are position independent. To check these assumptions, we prepared a series of duplexes that contained a G•A mismatch at different positions and compared these to a perfectly matched T•A pair. Mismatch sites occurred either 3 bp from the 5′ end, 3 bp from the 3′ end or in the center of the duplex. In all cases, irrespective of location, the mismatches were embedded within the same sequence context, -TAXGA-, so any sequence-dependent effects on duplex stability were minimized. UV melting experiments were conducted and Tm
values were determined for each duplex; results are presented in . As expected, LNA triplets improved mismatch discrimination by ~3.0°C. The position of the mismatched base pair had a small influence on mismatch discrimination. The mismatch located in the center of the sequence showed slightly improved discrimination compared to mismatches located closer to the ends of the duplex. Interestingly, this position preference was more pronounced for DNA probes than for LNA probes. Therefore, whenever possible positioning the mismatch site close to the center of the probe-target duplex is preferred to achieve maximum discriminatory power. Letowski et al
) observed similar positional effects using microarrays. They reported that 50 bp long probes showed better specificity when five distributed mismatches were located in the middle of the probe. Probes with mismatches located at either end of the molecule were less discriminating.
Comparison of melting temperatures and mismatch discrimination for duplexes containing the same G•A mismatch either in the middle versus near the ends of the sequence
Stability predictions for LNA–DNA duplexes
Stability of perfectly matched
LNA–DNA duplexes can be estimated from published nearest-neighbor parameters (32
) for mixed sequences containing LNA bases interspersed among DNA bases. LNA parameters (35
) predict energetic differences between LNA-modified and unmodified DNA duplexes. We therefore tested if the parameters were accurate in predicting changes of melting temperatures due to LNA substitutions. Experimental Tm
(LNA) − Tm
(DNA) values of perfectly matched duplexes are presented in and are compared with values calculated using the ‘unified’ (32
) and LNA (35
) parameter sets. The Tm
increases caused by LNA modifications were predicted for a combined set of all LNA probes with an average error of ±0.7°C, which is within the experimental error. However, predictions of Tm
changes were less accurate for the 1ET Set (±1.6°C) and even less accurate for the 1ES Set (±2.3°C), where half of the nucleotides were LNA-modified. The published LNA thermodynamic parameters (35
) give a more accurate prediction of duplex stabilization when LNA residues are present in low densities.
Comparison of predicted and experimentally measured melting temperatures (°C) for perfectly matched DNA and LNA probe-target duplexes in 1 M Na+ buffer
Since nearest-neighbor parameters for LNA mismatches are unknown, thermodynamic stability of mismatched LNA–DNA duplexes cannot be accurately predicted. Melting data presented in and are not sufficient to derive these parameters, but will help in future investigations. Here, two base sequences were employed and the relative positions of LNA modifications were varied. To obtain nearest-neighbor parameters, it will be necessary to study a much larger set of sequences with the same kind of modification pattern. This study suggests that it would be most useful to determine parameters that can predict thermodynamics for triplets of LNA residues. Accurate stability predictions would make the process of designing LNA probes more reliable.
New design guidelines and biological applications
Although LNA synthesis methods have improved in recent years, LNA phosphoramidite monomers still exhibit lower coupling efficiency than DNA monomers. In particular, coupling efficiency of guanine LNA monomer is lower than the other LNA residues. It is better to avoid synthesis of long stretches of consecutive LNA bases because such sequences give lower synthesis yields, lower purity, and may therefore require PAGE or HPLC purifications.
Any probe candidates, including LNA probes, should be examined for their tendency to fold into hairpins, homodimers and other undesirable secondary structures. Stability of such secondary structures need to be minimized to prevent competition with desired probe-target duplex formation. Thermodynamic experiments have demonstrated (36
) that LNA–LNA base pairs are more stable than LNA–DNA base pairs, which are more stable than DNA–DNA base pairs. It is therefore likely that problems arising from secondary structures will be more significant when LNA residues are present.
Combining our experimental observations with previous suggestions (1
), we propose the following rules to maximize single mismatch discrimination using LNA probes:
- Place a triplet of LNA modifications with the central base of the triplet at the mismatch site, unless the probe contains the guanine base of a G•T mismatch.
- LNA modification of the guanine nucleotide or either of its nearest-neighbor bases should be avoided in a G•T mismatch site.
- Shorter probes improve mismatch discrimination.
- For best synthesis quality, it is preferable to avoid consecutive sequence of more than four LNA residues.
- Better discrimination seems to be achieved when the position of the mismatch site is close to the center of the probe. Discrimination is significantly decreased if mismatches are located at the first or second base from either end of the duplex.
- If a mismatch type and location are unknown, the best approach is to modify all adenine nucleotides until the optimal Tm is achieved. The rest of the rules should be taken into account.
- Probes should not fold into stable, undesired self-complementary secondary structures or form self-dimers, especially when these structures contain LNA–LNA base pairs.
Rules (iii) and (iv) should be balanced depending on the precise needs of specific applications. Fully modified short LNA probes (6–8 bp) have been used successfully in SNP detection (13
) experiments. They are less suitable for some high-throughput screenings where each experiment cannot be thoroughly optimized, and varying synthesis yields as well as extensive purifications are incompatible. Longer LNAs (>14 bp) may possess too high affinity for complementary strands when they are fully modified. Additional experimentation is needed to obtain more insight about advantages and disadvantages of short fully LNA-modified probes versus chimeric LNA/DNA probes. The guidelines presented herein are available in the IDT SciTools OligoAnalyzer calculator (freely accessible at www.idtdna.com
), which can predict stability of LNA–DNA duplexes and assist with probe design.
DNA microarrays are a powerful technology for parallel high-throughput analysis of gene expression and genotyping. Since a single set of hybridization and wash conditions (temperature, ionic strength) is typically employed, design of microarrays that can reliably distinguish single base variations on a genomic scale represents a real challenge. LNAs can be used to improve microarray probes because they allow use of shorter and therefore better discriminating sequences. Ideally, at the hybridization temperature employed, perfectly matched probe-target duplexes should be stable and mismatched duplexes should be unstable. In this setting, the increased discriminatory power and specificity of LNA probes may enable more definitive results. The stability of duplexes increases with increasing G•C content. LNA bases can be selectively incorporated to fine-tune and control thermal stability, serving to normalize the Tm of large numbers of heterogeneous probe sequences. Probes with low f(G•C) could contain more LNA residues than probes with high f(G•C). The number of LNA residues and lengths of probes could be adjusted, so that various probe-target duplexes achieve similar stability.
LNA probes have been successfully used in quantitative real-time PCR applications (1
). Improved sensitivity and specificity have been observed in genotyping experiments and SNP polymorphism analysis. The number of LNA modifications can be adjusted to optimize melting temperatures of probe-target duplexes. Primers are usually designed to have very similar melting temperatures in the range of 55–60°C and T
of the probe is recommended to be about 7–10°C higher (3
). LNA modifications allow use of shorter fluorescent probes, improve mismatch discrimination and increase the window of annealing temperatures in which accurate genotyping results can be achieved (1
). Various formats of detection are possible. Besides dual-labeled probes, unlabeled LNA probes with LCGreen or SYBR Green dyes can be used to improve sensitivity of melting curve analysis (4