Structural diversity of target-specific homopyrimidine peptide nucleic acid–dsDNA complexes

Abstract

Sequence-selective recognition of double-stranded (ds) DNA by homopyrimidine peptide nucleic acid (PNA) oligomers can occur by major groove triplex binding or by helix invasion via triplex P-loop formation. We have compared the binding of a decamer, a dodecamer and a pentadecamer thymine–cytosine homopyrimidine PNA oligomer to a sequence complementary homopurine target in duplex DNA using gel-shift and chemical probing analyses. We find that all three PNAs form stable triplex invasion complexes, and also conventional triplexes with the dsDNA target. Triplexes form with much faster kinetics than invasion complexes and prevail at lower PNA concentrations and at shorter incubation times. Furthermore, increasing the ionic strength strongly favour triplex formation over invasion as the latter is severely inhibited by cations. Whereas a single triplex invasion complex is formed with the decameric PNA, two structurally different target-specific invasion complexes were characterized for the dodecameric PNA and more than five for the pentadecameric PNA. Finally, it is shown that isolated triplex complexes can be converted to specific invasion complexes without dissociation of the Hoogsteen base-paired triplex PNA. These result demonstrate a clear example of a ‘triplex first’ mechanism for PNA helix invasion.

INTRODUCTION

Efficient and specific targeting of predefined chromosomal DNA sequences by synthetic ligands is a major goal in chemical biology. A variety of agents are available for this purpose including engineered zinc finger proteins (1), synthetic polyamides (2) triplex forming oligonucleotides (TFOs) (3,4) and peptide nucleic acids (PNAs) (5,6). PNA oligomers are synthetic DNA mimics containing a pseudopeptide backbone consisting of an N-aminoethyl glycine polymer to which the nucleobases are connected via methylene carbonyl linkers (7). PNA binds sequence complementary single-stranded nucleic acid targets with high affinity and selectivity (8). Particularly stable ‘P-loop’ triplex invasion structures can form upon binding of homopyrimidine PNAs to double-stranded (ds) DNA containing a sequence complementary purine target via formation of an internal PNA–DNA–PNA triplex involving combined Watson–Crick and Hoogsteen base pairing (9). In this four-stranded structure, the unbound DNA strand is displaced in a single-stranded conformation. A number of variant P-loop complexes have been described, the formation of which depends on the PNA oligomer and DNA sequence in question (5,6). P-loop complexes have been used in a number of applications for instance for interference with (10–15), or activation of (15–17) transcription.

Because two PNA strands are involved in the triplex invasion P-loop complex, bis-PNA constructs containing two PNA oligomers connected via a flexible linker were generated (18). Bis-PNAs generally bind complementary DNA targets with superior efficiency as compared with conventional ‘mono’-PNAs (19,20) because of the pseudo-first-order reaction of binding for bis-PNAs as compared with the pseudo-second order of binding for mono-PNAs. Furthermore, the strand polarity can be optimized for bis-PNAs to allow the preferred parallel (i.e. the PNA N-terminus facing the 5′ end of the bound DNA strand) and anti-parallel (i.e. the PNA N-terminus facing the 3′ end of the bound DNA strand) orientation of the PNA strand bound via Hoogsteen and Watson–Crick base pairs, respectively. The binding can be further enhanced at physiological pH by the substitution of pseudoisocytosine for cytosine (pseudoisocytosine does not require low pH for efficient Hoogsteen base pairing to guanine) (18). Using a mono-PNA the two strands of the P-loop are identical and consequently the polarity of one or the other must be suboptimal unless a symmetrical sequence is targeted. Bis-PNA binding to a cognate target can yield structural isomers of perfectly matched P-loops. Such isomers are the consequence of alternative trajectories of the bis-PNA linker relative to the DNA strands in kinetically trapped complexes (21).

Because the dissociation rate of triplex invasion complexes is exceedingly slow (19,22) (and often not directly measurable at physiological temperature), helix invasion using decamer homopyrimidine PNA oligomers is kinetically controlled (23). Consequently, single mismatch specificity is also kinetically controlled, i.e. depending on the on-rate only (24). Typically, the association rate for fully matched complexes is significantly higher than that for mismatched complexes. However, once formed, even mismatched complexes remain highly stable so prolonged incubation times or increased PNA concentrations will eventually result in the formation of mismatched complexes (23).

Different mechanisms have been proposed for P-loop formation (23,25,26). Current evidence favour a ‘Hoogsteen first’ mechanism in which one PNA strand initially binds as a non-invasion triplex in the major groove via Hoogsteen base pairs followed by binding of another PNA strand via helix invasion and Watson–Crick base pairing. The evidence for this mechanism is indirect, however, and the molecular details remain poorly understood.

Clearly, kinetic control is of paramount importance in helix invasion of dsDNA by PNA. Moreover, the extent to which kinetics and thermodynamics each contribute to the control of PNA–dsDNA recognition is most likely shifted further towards kinetics for PNA oligomers of increased length. This is particularly pertinent for targeting unique sequences in the human genome, which from statistical considerations would require recognition of 15–17 bp. We therefore asked what might be the structural consequences of using such extended PNA oligomers for P-loop formation. To avoid the structural and kinetic complications associated with bis-PNAs (see above), we used conventional mono-PNAs. We analysed PNA–dsDNA complexes resulting from incubation of 10mer, 12mer and 15mer homopyrimidine PNAs with dsDNA containing a cognate sequence target. The results show that increased PNA oligomer length causes increased complexity of the generated sequence-targeted complexes. Furthermore, the data reveal the existence of distinct pathways for helix invasion by homopyrimidine PNA.

MATERIALS AND METHODS

PNA oligomers

PNA oligomers were synthesized as described (18,27). Conjugation of the nitrilotetraacetic acid (NTA) group to the N-terminus PNA was previously described (28). The following PNAs were used: PNA1707 (H-TTTTTCTCTC-Lys-NH₂) (measured mass: 2761; calculated mass: 2763), PNA1708 (H-TTTTTCTCTCTC-Lys-NH₂) (measured mass: 3276; calculated mass: 3280), PNA2046 (NTA-Lys-TTTTTCTCTCTC-Lys-NH₂) (measured mass: 3584; calculated mass: 3583), PNA1940 (NTA-Lys₂-TTTTTCTCTCTCTCT-Lys-NH₂) (measured mass: 4492; calculated mass: 4495). PNA concentrations were determined by spectrophotometry at 260 nm using the following molar extinction coefficients: ɛT (8800 M⁻¹ cm⁻¹) and ɛC (7300 M⁻¹ cm⁻¹). Low binding tubes (Sorenson Biosciences Cat. no. 11720 or Eppendorf Cat. no. 22 43 108-1) and tips (Sorenson Biosciences Cat. no.: 35010 and 35090) were used.

DNA

All DNA manipulations were done using standard methods (29). The p322 vector is a pUC19 derivative containing the PNA target sequence 5′-AAAAAGAGAGAGAGA-3′/5′-TCTCTCTCTCTTTTT-3′) in the polylinker region. The PNA target is flanked on one side by a unique engineered XbaI site and on the other side by an EcoRI restriction site from the vector (see Supplementary Material). The plasmid was propagated in Escherichia coli strain XL1-Blue (Stratagene), amplified by maxipreparation (Jet Star, Genomed) and verified by sequencing at DNA Technology, Denmark.

The relevant DNA restriction fragment of p322 was 5′-labelled using polynucleotide kinase and [γ-³²P]ATP or 3′-labelled using the Klenow fragment of DNA polymerase and [α-³²P]dATP at the desired end as indicated in the figure legends using standard methods (29). The DNA fragments were resolved using 5% polyacrylamide gels (30:1 in acrylamide to bisacrylamide), and the relevant fragment was excised and eluted overnight into 0.5 M ammonium acetate, 1 mM EDTA. Recovered ³²P-DNA was precipitated with 96% ethanol, washed with 70% ice-cold ethanol and air dried. The DNA was resuspended in H₂O and stored at −20°C.

Analyses of PNA binding to a target in dsDNA by gel-shift and chemical probing

³²P-DNA fragment (10–20 nM) was incubated with the indicated amount of PNA for 1 h in 20 μl of 10 mM sodium phosphate at pH 6 or 6.5 as indicated (slightly acidic due to the requirement of Hoogsteen base pairing on protonization of cytosine position N3) at 37°C. Resulting complexes were resolved by electrophoretic gel-shift analysis using native TAE-buffered 10% PAGE and visualized by autoradiography.

For in situ chemical probing with DMS and KMnO₄, varying PNA concentrations (0.25–10 μM) were used to establish the desired PNA–dsDNA complexes. Individual PNA–dsDNA complexes were excised from native polyacrylamide gels and processed basically as described (5,21) using optimized in situ probing reaction times of 1 min (DMS) or 2 min (KMnO₄).

Fe/NTA affinity cleavage was done in 100 μl sodium phosphate buffer. Eight μl of 10 mM (NH₄)₂[Fe(SO₄)₂] was added and incubation continued for 5 min at 37°C followed by addition of 20 μl 50 mM DTT, and incubation for 1 h at 37°C. The reactions were terminated by addition of 10 μl of 0.5 M EDTA at pH 8. Adenine/guanine-specific sequence reactions were performed as described (5).

Kinetic analyses were conducted by setting up a master mix (V = 220 μl) containing PNA1940 at a final concentration of 5 μM and ³²P-labelled target DNA fragment in 10 mM sodium phosphate (pH 6.5). At the indicated length of time, 20 μl sample was withdrawn, combined with TAE loading buffer and immediately resolved by 10% TAE-buffered PAGE. The data were quantified using a phosphorimager and Image Quant software. When appropriate, the ‘quench’ oligonucleotide 5′-dAGAGAGAGAGAAAAA-3′ was added to a 2- to 4-fold molar excess relative to PNA1940.

RESULTS AND DISCUSSION

Gel-shift analysis of PNA–dsDNA complexes

The PNA decamer gave rise to a single gel-shift [Designated PD¹⁰—the designation ‘PD’ identifies a complex as consisting of PNA and dsDNA. To keep track of gel-shifted complexes a suffix is added that includes a roman number identifying the gel-shift complex in question and a superscript identifying the PNA oligomer length used. The nomenclature should not be confused with the ‘PD-loop’ (30).] upon binding to a dsDNA fragment harbouring one sequence complementary target (Figure 1A). In contrast, incubation of the PNA dodecamer or the PNA pentadecamer with the dsDNA fragment harbouring the sequence complementary target resulted in multiple retarded gel migration species designated PDI¹²–PDIII¹² and PDI¹⁵–PDVII¹⁵, respectively (Figures 2A and 3A).

Figure 1.

P-loop formation on binding of PNA decamer to dsDNA: (PD¹⁰). (A) P-loop formation as investigated by gel-shift analysis. P-loops were generated using the following concentrations of PNA 1707: lane 1, w/o PNA; lane 2, 0.25 μM; lane 3, 0.5 μM; lane 4, 0.75 μM; lane 5, 1 μM; lane 6, 2 μM; and lane 7, 4 μM. P-loop (PD) and free dsDNA (D) are indicated. (B) In situ DMS probing of P-loop complex identified in (A): lane 1, dsDNA + DMS; lane 2, dsDNA w/o DMS; lane 3, PD¹⁰ + DMS; lane 4, A/G sequence. (C) In situ permanganate probing of P-loop complex identified in (A): lane 1, DNA + KMnO₄; lane 2, DNA w/o KMnO₄; lane 3, PD¹⁰ + KMnO₄; and lane 4, A/G sequence. (D) DMS probing in solution at elevated high ionic strength (+90 mM KCl) and using the following PNA1707 concentrations: lane 1, w/o PNA; lane 2, 0.13 μM PNA; lane 3, 0.25 μM PNA; lane 4, 0.5 μM PNA; lane 5, 1 μM PNA; lane 6, 2 μM PNA; lane 7, 4 μM PNA; lane 8, 8 μM PNA; and lane 9, A/G sequence. (E) Structural assignment as based on the data in (A–C). (F) Structural assignment based on the data in (D and manuscript in preparation). We used the 168 bp XbaI–PvuII restriction fragment of p322 either 5′-kinase labeled (A and D) or 3′-Klenow labeled (C) with ³²P-phosphate at the XbaI site. The 223 bp PvuII–EcoRI restriction fragment of p322 was 3′-Klenow labeled and used in (B). PNA incubations were at pH 6 as described in Materials and Methods.

Figure 2.

Binding of PNA dodecamer to dsDNA (PD¹²). Autoradiographs showing binding of the indicated PNAs to dsDNA as investigated by gel-shift analysis (A and B), in situ chemical probing (C and D) and affinity cleavage (E). (A) Titration using the following PNA1708 concentrations: lane 1, w/o; lane 2, 0.25 μM; lane 3, 2 μM; and lane 4, 6 μM. Free dsDNA (D) and the formed complexes PDI¹², PDII¹² and PDIII¹² are indicated. Analogous PD¹² complexes were obtained using NTA-conjugated PNA2046 (Supplementary Figure S1). (B) PD¹² complex formation as a function of ionic strength using 6 μM PNA1708 and supplemented with the following amounts of KCl: lane 1, w/o; lane 2, 5 mM; lane 3, 10 mM; lane 4, 20 mM; lane 5, 40 mM; lane 6, 80 mM; lane 7, 160 mM; and lane 8, 320 mM. (C) DMS probing of gel-purified PD¹² complexes involving PNA2046: lane 1, dsDNA + DMS; lane 2, dsDNA, w/o DMS; lane 3, PDI¹² + DMS; lane 4, PDII¹² + DMS; lane 5, PDIII¹² + DMS; and lane 6, A/G sequence. (D) Permanganate probing of gel-purified PD¹² complexes involving PNA2046: lane 1, dsDNA + KMnO₄; lane 2, dsDNA, w/o KMnO₄; lane 3, PDI¹² + KMnO₄; lane 4, PDII¹² + KMnO₄; lane 5, PDIII¹² + KMnO₄; and lane 6, A/G sequence. (E) Fe/NTA affinity cleavage of gel-purified PD¹² complexes involving PNA2046: lane 1, dsDNA + Fe/NTA; lane 2, dsDNA, w/o Fe/NTA; lane 3, PDI¹² + Fe/NTA; lane 4, PDII¹² + Fe/NTA; lane 5, PDIII¹² + Fe/NTA; and lane 6, A/G sequence marker. (F) Structural assignment as based on the probing results [arrow (KMnO₄ hypersensitivity), perpendicular symbol (DMS protection)]. The 168 bp XbaI–PvuII restriction fragment of p322 was used either 5′-kinase labelled (A–C, E) or 3′-Klenow labelled (D) with ³²P-phosphate at the XbaI site. PNA incubations were at pH 6.5 as described in Materials and Methods.

Figure 3.

Binding of PNA pentadecamer to dsDNA (PD¹⁵). Autoradiographs showing binding of PNA1940 to dsDNA as investigated by gel-shift analysis (A and B), in situ chemical probing (C and D) and affinity cleavage (E). (A) PNA titration using the following concentrations: lane 1, w/o; lane 2, 0.25 μM; lane 3, 0.5 μM; lane 4, 0.75 μM; lane 5, 1 μM; lane 6, 2 μM; lane 7, 4 μM; and lane 8, 8 μM. Free DNA (D) and the formed complexes PDI¹⁵–PDVII¹⁵ are indicated. (B) PD¹⁵ complex formation as a function of ionic strength. Incubation was carried out using 6 μM PNA and supplemented with KCl as follows: lane 1, w/o; lane 2, 5 mM; lane 3, 10 mM; lane 4, 20 mM; lane 5, 30 mM; lane 6, 40 mM; lane 7, 50 mM; lane 8, 60 mM; lane 9, 70 mM; lane 10, 80 mM; lane 11, 100 mM; lane 12, 120 mM; lane 13, 140 mM; lane 14, 160 mM; lane 15, 180 mM; lane 16, 300 mM; and lane 17, 320 mM. (C) DMS probing of the complexes identified in (A). Lane 1, dsDNA + DMS; lane 2, dsDNA w/o DMS; lane 3, PDI¹⁵ + DMS; lane 4, PDII¹⁵ + DMS; lane 5, PDIII¹⁵ + DMS; lane 6, PDIV¹⁵/PDV¹⁵ + DMS; lane 7, PDVI¹⁵ + DMS; lane 8, PDVII¹⁵ + DMS; and lane 9, A/G sequence marker. (D) KMnO₄ probing of the complexes identified in (A). Lane 1, dsDNA + KMnO₄; lane 2, dsDNA w/o KMnO₄; lane 3, PDI¹⁵ + KMnO₄; lane 4, PDII¹⁵ + KMnO₄; lane 5, PDIII¹⁵ + KMnO₄; lane 6, PDIV¹⁵/PDV¹⁵ + KMnO₄; lane 7, PDVI¹⁵ + KMnO₄; lane 8, PDVII¹⁵ + KMnO₄; and lane 9, A/G sequence marker. (E) Autoradiograph showing PDI¹⁵ and PDII¹⁵ subjected to Fe/NTA affinity cleavage. Lane 1, dsDNA + Fe/NTA; lane 2, PDI¹⁵ + Fe/NTA; lane 3, PDII¹⁵ + Fe/NTA; and lane 4, A/G sequence marker. (F) Structural assignment (arrow, KMnO₄ hypersensitivity; perpendicular symbol, DMS protection). The 168 bp Xba–PvuII restriction fragment of p322 was used either 5′-kinase labeled (A–C, E) or 3′-Klenow labeled (D) with ³²P-phosphate at the XbaI site. PNA incubations were at pH 6.5 as described in Materials and Methods.

Decamer PNA–dsDNA complexes

P-loop structures can be probed using potassium permanganate (KMnO₄) and dimethyl sulphate (DMS) (9). KMnO₄ preferentially oxidizes the 5–6 double bond of unstacked (single-stranded) thymine nucleobases, which poises the DNA strand for alkali-induced backbone cleavage at such reacted sites (31). Hypersensitivity towards KMnO₄ can thus be used to reveal the single-stranded DNA of a P-loop (9). DMS methylates guanine at the N7 position (and to a much lesser extent adenine at the N3 position) (32). Because the guanine N7 is engaged in Hoogsteen hydrogen bonding, PNA interference with DMS methylation can be used to reveal Hoogsteen base pairing, e.g. during P-loop formation.

Chemical probing of the gel-excised PD¹⁰ complex revealed protection from DMS methylation of residues G₆, G₈ and G₁₀ (Figure 1B) and hypersensitivity towards KMnO₄ oxidation of T₁–T₅, T₇ and T₉ (Figure 1C). These results are compatible with a triplex invasion complex containing two PNA strands bound via combined Watson–Crick and Hoogsteen base pairs. Because the DNA lacks G-residues as a reporter for DMS probing in the 5′-target region, we cannot formally rule out alternative modes of binding for the Hoogsteen base-paired PNA strand. However, we favour the parallel alignment shown in Figure 1E because this is the optimal binding orientation (18) and because incomplete base pairing would involve as few as 5 bp of PNA–DNA hybrid that would be very unstable (33).

When the PNA binding reaction was carried out at elevated ionic strength conditions no hybridization of conventional unmodified decamer homopyrimidine PNA was detected in gel-shift analyses. In contrast, probing directly in solution revealed a PNA concentration-dependent protection from DMS methylation in reactions supplemented with 90 mM KCl (Figure 1D). These results [consistent with the previous results (11,24,34–36)], show that invasion by simple PNA oligomers is severely compromised at elevated ionic strength and suggests that a conventional triplex structure involving PNA decamer bound via Hoogsteen base pairs is formed, which is not sufficiently stable for gel-shift analysis (Figure 1F).

Dodecamer PNA–dsDNA complexes

Three complexes designated PDI¹², PDII¹² and PDIII¹² were identified by gel-shift analysis upon incubation of the PNA dodecamer with the dsDNA fragment containing the cognate sequence target (Figure 2A). The relative ratio of these complexes was PNA concentration dependent. PDI¹² was the predominant species at low PNA concentrations and the amount of this complex decreased with increasing PNA concentration. In contrast, the amount of PDII¹² and PDIII¹² increased with increasing PNA concentration and these complexes were the sole species observed at elevated PNA concentrations.

We anticipated that ionic strength would affect triplex formation much less than helix invasion. Thus to distinguish possible triplex and invasion type complexes, a salt titration was conducted. At the employed PNA concentration, low ionic strength conditions produced only PDII¹² and PDIII¹², but as the salt concentration was increased, PDI¹² emerged at the expense of the former (Figure 2B) suggesting that PDII¹² and PDIII¹² are invasion complexes whereas PDI¹² is a conventional triplex.

To enable an investigation of the PNA strand polarity of gel-shifted complexes, we chose to employ NTA-conjugated PNA oligomers (see below). Using such conjugated dodecamer PNA (PNA2046), gel-shifts PDI¹², PDII¹² and PDIII¹² were obtained analogously to those obtained with unmodified PNA dodecamer (Supplementary Figure S1).

Chemical probing of gel-excised PDI¹² involving NTA-conjugated PNA dodecamer, showed that guanine residues G₆ through G₁₂ are protected from DMS methylation (Figure 2C, lane 3), and all thymine residues T₁ through T₁₁ are resistant to KMnO₄ oxidation (Figure 2D, lane 3). These results show PNA binding to residues G₆–G₁₂ compatible with a structure containing a single PNA oligomer bound via Hoogsteen base pairs to target residues A₁–G₁₂ (since there are no G-residues in the 5′-proximal target, DMS probing cannot be used to reveal PNA binding to this region) (Figure 2F, PDI¹²).

To investigate the orientation of PNA strands in the generated PD complexes, we employed Fe/NTA affinity cleavage. Ferrous ions are well-known DNA cleaving reagents that function via activation by molecular oxygen. Such reactions are particularly useful in conjunction with a bifunctional molecule containing a DNA-binder linked to a metal chelator because this yields DNA cleavage proximal to the DNA binder (37,38). Lohse et al. (28) previously reported the synthesis and properties of such a bifunctional molecule based on a NTA–PNA conjugate in which ferrous iron was chelated to the NTA group.

PDI¹² subjected to Fe/NTA affinity cleavage showed strand scission 5′-distal to the PNA target (Figure 2E, lane 3). Because the NTA moiety is conjugated via the N-terminus of the PNA (see Materials and Methods), these data are compatible with a parallel orientation of the PNA oligomer and formally establish the binding of the PNA strand to the entire target (residues A₁–G₁₂) (Figure 2F, PDI¹²).

DMS probing of gel-excised PDII¹² and PDIII¹² revealed protection from methylation of guanine residues G₆ through G₁₂ (Figure 2C, lanes 4 and 5) similar to that of PDI¹². Thus both PDII¹² and PDIII¹² contain a PNA strand bound via Hoogsteen base pairs to the target (see below).

Permanganate probing of gel-excised PDIII¹² revealed hypersensitive thymine residues along the entire target region including T₁–T₁₁ (Figure 2D, lane 5). Consequently, the PDIII¹² complex must constitute a helix invasion type structure in which a Watson–Crick base-paired PNA strand is hybridized to all target residues A₁–G₁₂ in a parallel orientation (Figure 2F, PDIII¹²). Permanganate probing of gel-excised PDII¹² revealed an unequal hypersensitivity of the thymines within the target: T₁ through T₃ were almost entirely resistant to KMnO₄ oxidation, T₄ and T₅ showed intermediate KMnO₄ hypersensitivity, and finally T₇, T₉ and T₁₁ were most hypersensitive (Figure 2D, lane 4). Thus PDII¹² represent an invasion type structure, with only partial target occupation, in which the helix invasion PNA is bound to DNA target residues G₆–G₁₂. Such partial Watson–Crick base pairing can occur with the optimal anti-parallel strand orientation (Figure 2F, PDII¹²).

Induction of Fe/NTA affinity cleavage in PDII¹² and PDIII¹² revealed DNA strand scission 5′ to the PNA target in both cases (Figure 2E, lanes 4 and 5). In PDII¹², the 5′-cleavage is accounted for by a parallel PNA bound via Hoogsteen base pairs. The PNA strand engaged in Watson–Crick base pairing, is not expected to yield efficient 3′-dsDNA strand scission if the Fe/NTA group is displaced from the target via a segment of unbound PNA (Figure 2F). Finally, in the case of PDIII¹², selective Fe/NTA affinity cleavage at the 5′-dsDNA target is accounted for by a parallel orientation of both Hoogsteen and Watson–Crick base-paired PNA strands, which co-locates the reactive Fe/NTA groups to this position.

Thus, PDII¹² and PDIII¹² are structurally different. The PDII¹² complex exhibit optimal strand polarity at the expense of a reduced number of base pairs (only 7 out of 12 possible invasion base pairs), whereas PDIII¹² complex has compromised strand polarity (parallel Watson–Crick base-paired PNA strand) but an optimal number of base pairs (12 invasion base pairs).

Pentadecamer PNA–dsDNA complexes

As suggested by the increase in product complexity on going from PNA decamer to PNA dodecamer, the number of PNA–dsDNA complexes increased even further when using the PNA pentadecamer now counting at least 7 complexes: PDI¹⁵–PDVII¹⁵ (Figure 3). In analogy to the results for the PNA dodecamer, the distribution among the pentadecamer PNA–dsDNA complexes is highly PNA concentration dependent (Figure 3A) and sensitive to ionic strength (Figure 3B): PDI¹⁵–PDII¹⁵ and PDIII¹⁵–PDVII¹⁵ are prominent at low and high PNA concentrations, respectively. On increasing the ionic strength, PDI¹⁵ and PDII¹⁵ appear at the expense of PDIII¹⁵–PDVII¹⁵. This suggests that PDI¹⁵ and PDII¹⁵ represent triplex type structures whereas PDIII¹⁵ through PDVII¹⁵ most likely correspond to various distinct invasion complexes.

DMS probing of gel-excised complexes PDI¹⁵ and PDII¹⁵ showed protection of all target guanines (G₆ through G₁₄; Figure 3C, lanes 3 and 4) and KMnO₄ probing revealed resistance to oxidation of all target thymines (T₁–T₁₅; Figure 3D, lanes 3 and 4) consistent with these complexes being triplex type structures containing a single PNA strand bound via Hoogsteen base pairs to the DNA target. These data could be compatible with the existence of two triplexes, one with the PNA bound parallel to the entire target (A₁–A₁₅), and another with the PNA bound anti-parallel to only part of the target (A₅–A₁₅). However, Fe/NTA affinity cleavage unambiguously establishes that the PNA strand bound via Hoogsteen base pairs is hybridized to the DNA target in a parallel orientation for both complexes (Figure 3C, lanes 2 and 3).

Enigmatically, complexes PDI¹⁵ and PDII¹⁵ are clearly different structures as based on their electrophoretic mobility differences in native gels (Figure 3A and B)—yet they are indistinguishable via chemical probing, at least using the present methods. This result could, in principle, be caused by an inhomogeneous PNA oligomer preparation containing truncated or modified PNA molecules. However, although small amounts of impurities were seen by HPLC, mass spectrometry revealed only a single peak of the correct mass arguing against this (data not shown). Furthermore, since PDI¹⁵ and PDII¹⁵ show similar affinity cleavage intensities (Figure 3E), both complexes must contain full-lenth PNA oligomers because the NTA group is conjugated to the PNA oligomer in the last coupling step.

Chemical probing of gel-excised PDVI¹⁵ (i.e. in fact two poorly resolved complexes) and PDVII¹⁵ revealed that both complexes show protection towards DMS methylation of all DNA target guanines (G₆ through G₁₄; Figure 3C, lanes 7 and 8). Thus both complexes contain a PNA strand bound via Hoogsteen base pairs to at least part of the target, but most likely to the entire target (A₁ through A₁₅). Although the Fe/NTA affinity cleavage yielded inconclusive data for invasion complexes involving PNA pentadecamer due to interference with gel-mobility by the exceedingly stable PNA₂–DNA triplex (data not shown), we favour a structure in which the Hoogsteen base-paired PNA strand binds parallel to the target DNA strand because this is observed in the case of all PD¹² complexes as well as for PDI¹⁵ and PDII¹⁵, and is the preferred orientation for PNA Hoogsteen hybridization (18).

As in the case of PDII¹² and PDIII¹², probing of gel-excised PDVI¹⁵ and PDVII¹⁵ revealed notable differences in hypersensitivity to permanganate oxidation of target thymines (Figure 3D). In the case of PDVI¹⁵, target residues T₁ through T₄ showed minimal reactivity, T₅ exhibit intermediate hypersensitivity, and T₇–T₁₅ were all strongly reactive towards permanganate. These results are consistent with partial invasion by a PNA strand bound via Watson–Crick base pairs to target residues A₇ through A₁₅, most likely in an anti-parallel orientation. Permanganate probing of the PDVII¹⁵ complex revealed strong and almost equal hypersensitivity along the entire DNA target compatible with invasion of the target by a PNA strand bound to the entire target (A₁ through A₁₅) in a parallel orientation.

Thus PDVI¹⁵ and PDVII¹⁵ are distinct structures. Most likely, both complexes have a fully hybridized parallel Hoogsteen base-paired PNA strand (i.e. bound to residues A₁–A₁₅). PDVI¹⁵ has a reduced number of Watson–Crick base pairs (9 out of 15 bp), presumably reflecting an optimal binding orientation of the invasion PNA strand. In contrast, PDVII¹⁵ has an optimal number of Watson–Crick base-pairs, which can only take place with suboptimal PNA strand polarity. In other words, PDVII¹⁵ appears to be a conventional P-loop structure with a 15 bp internal PNA–DNA–PNA triple helix and a corresponding stretch of displaced DNA, whereas PDVI¹⁵, due to an internal duplex–triplex hybrid shows only ∼9–10 bp of looped out DNA (Figure 3F).

The least prominent complexes PDIII¹⁵ through PDV¹⁵ remain incompletely understood in part because PDIV¹⁵ and PDV¹⁵ could not be separated. Consequently, specific structural assignments have been omitted for these complexes and the probing data are merely summarized (Figure 3F, PDIII¹⁵ or PDIV¹⁵/PDV¹⁵). Probing of gel-excised PDIII¹⁵ or the PDIV¹⁵/PDV¹⁵ mixture showed that these different complexes have in common that all target guanine residues except G₁₄ (the 3′-ultimate guanine) show protection towards DMS methylation indicating that a PNA strand binds incompletely to the target sequence via Hoogsteen base pairs (Figure 3C, lanes 5 and 6). Further chemical probing of PDIII¹⁵ and PDIV¹⁵/PDV¹⁵ revealed maximal permanganate hypersensitivity in the middle of the target (T₇ through T₁₂). Towards the 3′ end of the target of both PDIII¹⁵ and PDIV¹⁵/PDV¹⁵ showed reduced hypersensitivity towards KMnO₄ at position T₁₅. Towards the 5′ end of the target, PDIII¹⁵ and PDIV¹⁵/PDV¹⁵ differed with respect to permanganate reactivity. PDIII¹⁵ showed only marginal reactivity at residues T₁ through T₄ and intermediate hypersensitivity at position T₅. The PDIV¹⁵/PDV¹⁵, however, showed minimal permanganate reactivity at position T₁ and T₂ and intermediate hypersensitivity at positions T₃–T₅. Thus PDIII¹⁵ and PDIV¹⁵/PDV¹⁵ are clearly distinct structures with respect to the helix invasion PNA strand. Together these results show that stable invasion complexes involving structures with incompletely bound Hoogsteen and/or Watson–Crick base-paired PNA strands can result upon binding of a PNA pentadecamer to a cognate dsDNA target.

Conversion of triplex to invasion complex

Previous studies have addressed the mechanism of PNA helix invasion of a dsDNA target suggesting the triplex as an intermediate (23,25,26). The present results reveal distinct and stable triplex structures (Figures 2 and 3), and therefore it should be possible to study their possible interconversion and conversion into invasion complexes by incubation of isolated triplex with additional PNA oligomer. Hence gel-purified PDI¹⁵ and PDII¹⁵ triplexes were challenged with additional PNA pentadecamer (Figure 4). Indeed, PDI¹⁵ and PDII¹⁵ were transformed into various new species. Surprisingly, each specific triplex yielded unique invasion complexes: PDI¹⁵ produced PDIII¹⁵ and PDV¹⁵ while PDII¹⁵ gave rise to PDIV¹⁵, PDVI¹⁵ and PDVII¹⁵ (Figure 4). Furthermore, the two triplexes PDI¹⁵ and PDII¹⁵ were not interconvertible. These results provide compelling evidence that a triplex is indeed a substrate for invasion and also supports that the triplexes PDI¹⁵ and PDII¹⁵ are structurally different. Irrespective of the nature of this difference and of its cause, each triplex produces a distinct family of invasion complexes indicating that binding of the ‘Watson–Crick’ PNA strand occurs via the pre-existing ‘Hoogsteen type’ triplex—and does not require or promote its dissociation. We also note that base pair alterations apparently take place in the Hoogsteen base-paired-PNA strand upon transformation from PDI¹⁵ to PDIII¹⁵ (Figure 3C, cf. lanes 3 and 5) (see below).

Figure 4.

Chasing isolated PD¹⁵ triplexes into structurally distinct invasion complexes. Upper panel: Autoradiograph showing conversion of triplex to invasion complex. The triplexes PDI¹⁵ and PDII¹⁵ were generated using 6 μM PNA1940 and isolated by excision and crushing of the gel-slices. PDI¹⁵ (lanes 2–5) and PDII¹⁵ (lanes 6–9) were incubated with fresh PNA1940 in the gel-slurry and resolved by gel-shift analysis. The following PNA concentrations were used: w/o (lanes 1, 2 and 6), 1 μM (lanes 3 and 7), 3 μM (lanes 4 and 8) and 6 μM (lanes 5 and 9). The PDI¹⁵ triplex is transformed into the invasion species PDIII¹⁵ and PDV¹⁵ whereas PDII¹⁵ is chased into PDIV¹⁵, PDVI¹⁵ and PDVII¹⁵. The 168 bp XbaI–PvuII restriction fragment of p322 5′-kinase labeled with ³²P-phosphate at the XbaI site was used. Lower panel: Diagram showing the relationship between complexes. PNA incubations were carried out at pH 6.5.

Kinetics

Further investigation of the conversion of triplex to invasion complexes was done by monitoring the time course of the reactions in a low ionic strength buffer (Figure 5). The triplex is formed very rapidly and the amount of this complex reaches its maximal value already at the first time point (30 s), and then diminishes as invasion complexes emerge (Figure 5). A control experiment in which surplus of an oligonucleotide complementary to the PNA was added at 30 s of incubation completely inhibited the formation of invasion complexes presumably by trapping free PNA. In contrast, the amount of triplex produced at this time point remained constant over the entire experiment indicating no dissociation of this complex (data not shown). Therefore, these results confirm that a direct transformation of a PNA–dsDNA ‘Hoogsteen’ type triplex into a PNA–DNA–PNA, DNA invasion complex does indeed take place.

Figure 5.

Kinetic analysis of triplex and invasion complex formation. Amount of triplex or invasion complex as a function of time. PNA–dsDNA complex formation was conducted using PNA1940 followed by gel-shift and phosphorimaging analysis. The 168 bp XbaI–PvuII restriction fragment of p322 was used 3′-Klenow labeled with ³²P-phosphate as described in Materials and Methods.

Although the investigated triplexes are stable in the absence of excess PNA oligomer, the base-pair alterations of the Hoogsteen PNA strand observed upon transformation of PDI¹⁵ to PDIII¹⁵ suggest that remodelling of the Hoogsteen PNA strand may take place upon helix invasion by a Watson–Crick binding PNA strand.

The experiments in this study were performed at slightly acidic pH to facilitate protonation of cytosine N3. To demonstrate that the results extend to a physiologically relevant pH, an analogous experiment was carried out at pH 7.3 using a pentadecamer homopyrimidine PNA oligomer containing pseudoisocytosine (18) instead of cytosine. Binding of this PNA to the target gave rise to a multitude of complexes analogously to what was observed for the corresponding cytosine PNA (Supplementary Figure S2). Moreover, the results underline that the specific complexes formed as well as their migration in the gel depend on the specific PNA oligomer used. In particular, the sequence will of course dictate which complexes are stable, but other chemical modification (such as charge) may also be of importance.

CONCLUSION

The first conclusion drawn from the present work is that extending the strand length of homopyrimidine PNA oligomers increases the structural heterogeneity of resulting sequence targeted PNA–dsDNA complexes and this increased product complexity is due to the formation of kinetically trapped (but thermodynamically suboptimal) complexes involving compromised PNA strand orientation and incomplete base pairing. Thus, target selectivity of helix invasion is expected to be reduced when addressing longer than decamer sequences because kinetically stable complexes that do not involve recognition of the entire target can be formed. Indeed, we observe only limited differences of binding efficiency when targeting decamer, dodecamer and pentadecamer dsDNA targets by a pentadecamer homopyrimidine PNA via helix invasion (unpublished data). Such extended dsDNA sequences may, however, be targeted using modular constructs containing PNA oligomer conjugated to small molecule DNA binding agents such as the minor groove binder Hoechst (39). In this approach, the sequence preference of the minor groove binder selects a subset of targets via equilibrium interactions with its own target, which is followed by PNA helix invasion at sites that contain an adjacent PNA target.

It is important to emphasize that the binding diversity and target discrimination described for the PNA oligomers is a consequence of kinetic control and not of thermodynamic equilibrium because of the very high stability and thus long life times of the complexes (40). This is analogous to the trapping of meta-stable intermediates in RNA folding (41) as well as protein folding (42), which can result in incorrectly folded and dysfunctional structures. In most protein–DNA interactions, the affinity has been tuned by evolution (perhaps to avoid this type of kinetic trapping) such that sequence specificity is controlled by thermodynamics.

The second conclusion is that it is possible to stably target single sites in dsDNA via a triplex strategy when using homopyrimidine PNA oligomers longer than 12mer. Previous investigations have reported the formation of triplexes between PNA oligomers and complementary dsDNA targets (26,43) but so far there has been no structural characterization of such complexes. Detailed studies addressing the specificity and affinity of such PNA–dsDNA triplex interactions are under way in order to explore whether this approach has advantages over other triplex targeting strategies (3,4).

The third conclusion from the present work is that a stable non-invasion ‘Hoogsteen type’ triplex homopyrimidine PNA–dsDNA intermediate can be directly converted into a helix invasion P-loop complex involving combined Hoogsteen and Watson–Crick base pairs. Thus the ‘Hoogsteen first’ mechanism is valid at least for the employed pentadecamer under the present conditions. However, we are not able to conclude that this is the general mechanism for triplex invasion, since PNA sequence (especially C-content) and strand length as well as template topology and dynamics (e.g. resulting from the activity of DNA processing enzymes) may alter the mechanism of P-loop formation.