Peptide Nucleic Acid: Synthesis and Hybridisation Behaviour

The original PNA design detached the deoxyribose phosphate backbone of DNA in a computer model and replaced it with an achiral polyamide; the first oligomers were made from thymine-linked aminoethylglycyl units ^[1]. The finished structure is repeating N-(2-aminoethyl)glycine units linked by amide bonds, with the purine and pyrimidine bases attached to that backbone through methylene carbonyl linkages, and no pentose sugar or phosphate anywhere in the molecule ^[3]. Two consequences dominate everything downstream: the backbone carries no charge, and the oligomer is assembled by peptide chemistry, not phosphoramidite chemistry.

What neutrality buys, and where the salt goes

Those first homothymine oligomers recognised their complementary target inside double-stranded DNA by strand displacement — made possible, the authors argued, by the extraordinarily high stability of the PNA-DNA hybrids ^[1]. With all four nucleobases available, PNA was shown to hybridise to complementary oligonucleotides obeying the Watson-Crick base-pairing rules, making it a true DNA mimic in base-pair recognition ^[2].

Differential scanning calorimetry puts numbers on it. For 10 bp PNA/DNA duplexes in 10 mM sodium phosphate at pH 7.0, transition temperatures ran 329-343 K with calorimetric transition enthalpies of 209 ± 6 to 283 ± 37 kJ mol^-1; the corresponding DNA/DNA transition temperatures were 7-20 K lower, enthalpies 72 ± 29 to 236 ± 24 kJ mol^{-1 [12]}. Isothermal titration calorimetry on thirteen hybridisation reactions of 10-base sequences of design 5'-ATGCXYATGC-3' found that most PNA sequences bound more tightly than their corresponding DNA sequences, and attributed that to smaller entropy changes in the PNA/DNA reactions ^[13]. The advantage is not primarily a bigger binding enthalpy — the PNA/DNA binding enthalpies in that set spanned -194 to -77 kJ mol^-1, against -223 to -124 kJ mol^-1 for the DNA/DNA reactions.

"Salt-independent" is the shorthand you will hear, close enough to be useful and wrong enough to trip you. The absence of electrostatic repulsion between the strands leaves the T_m of a PNA/nucleic acid duplex largely independent of salt concentration, but the operational statement concerns the gap between two curves: the PNA/DNA advantage over DNA/DNA is significant at physiological salt, and the melting temperatures are essentially the same above 500 mM sodium ^[3]. Single-molecule association and dissociation measurements across a range of ionic strengths and temperatures locate the effect. Unlike DNA:DNA duplexes, PNA:DNA duplexes are more stable at lower ionic strength, and that is due to a higher association rate; the dissociation rate of PNA:DNA duplexes is largely insensitive to ionic strength, though significantly lower than that of DNA:DNA duplexes having the same number and sequence of base pairing interactions ^[15]. Salt acts on how fast the PNA closes on its target, not on how long the closed duplex survives. The payoff is that a PNA probe can be hybridised in the absence of salt, despite competing secondary structure in the target ^[3].

Orientation, and RNA versus DNA

PNA can bind complementary nucleic acids in both antiparallel and parallel orientation, but the antiparallel orientation — the 3' end of the DNA facing the amino terminus of the PNA — is strongly preferred, and the parallel duplex has been shown to have a different structure ^[3]. Unless a paper says otherwise, "PNA/DNA duplex" means the antiparallel complex; a probe laid out by writing the PNA left-to-right alongside DNA written 5' to 3' is the parallel one.

RNA targets bind slightly harder. For a mixed-sequence pentadecamer PNA containing all four nucleobases, T_m values for PNA·RNA duplexes were on average 4 °C higher than for PNA·DNA duplexes, with a faster association and a slower dissociation rate constant ^[10]. That study is worth reading for its own limitation: the authors reported that determination of k_a seemed prone to artefacts of the method and was not considered capable of providing absolute values representing the association rate constant in bulk solution ^[10]. Surface kinetics on PNA tethered through a biotin linker to streptavidin at a dextran/gold surface are not solution kinetics.

Because PNA/DNA melting does not follow the same rules as DNA/DNA melting ^[3], prediction is done by empirical correction rather than by a nearest-neighbour model for PNA itself. The published formula takes the T_m calculated for the corresponding DNA/DNA duplex using a nearest-neighbour approach and adds terms for the pyrimidine content and length of the PNA, accounting for the increased thermostability of PNA/DNA hybrids and the asymmetry of the heteroduplex. Challenged with an independent data set not used for model building, it predicted the T_m of more than 90% of sequences within 5 K and 98% within 10 K ^[9]. That is a design filter, not a measurement.

Boc, Fmoc, and why couplings stall

PNA is prepared by sequential coupling of protected monomers on a solid support using an approach similar to solid-phase peptide synthesis, but it is more challenging: the monomers are difficult to prepare and poorly soluble, and elongation is jeopardised by intra- and inter-chain aggregation and side reactions. The protecting-group scheme chosen for the monomer then dictates the approach used to build the oligomer ^[6].

The Boc route was the first made routine. Activation by HBTU in combination with in situ neutralisation of the resin allowed efficient coupling of all four Boc-protected PNA monomers within 30 minutes, and HPLC of the crude product from a fully automated synthesis of the model 17-mer H-CGGACTAAGTCCATTGC-Gly-NH₂ indicated an average yield per synthetic cycle of 97.1%. N1-benzyloxycarbonyl-N6(3)-methylimidazole triflate substantially outperformed acetic anhydride as a capping reagent, and the resin-bound PNA was cleaved by the "low-high" trifluoromethanesulphonic acid procedure ^[4]. That 97.1% is an average across one model sequence terminating in glycinamide — not a per-monomer coupling efficiency, and not a forecast for a purine-rich target.

The strong acid at the end is why Fmoc chemistry is attractive: TFMSA cleavage is not available in every laboratory and is unkind to acid-sensitive reporter groups. The Fmoc-protected backbone is not trivially accessible either. A scalable route to Fmoc-AEG-OBn mono-Boc protects ethylenediamine, alkylates it with benzyl bromoacetate, then removes the Boc group and replaces it with Fmoc; from 50 g of Boc anhydride it gave 31 g of product in 32% overall yield ^[5]. That yield belongs to a backbone building block — a key intermediate for nucleobase-modified monomers, not an assembled oligomer.

Purine-rich sequences, aggregation, and the length ceiling

The ceiling on PNA length is set by aggregation of the growing chain on the solid support, particularly for purine-rich sequences, and most reported solid-phase methodology suffers from poor coupling efficiency that limits production to sequences of less than 15 residues ^[7]. In a study of backbone modification with Hmb (2-hydroxy-4-methoxybenzyl) and Dmb (2,4-dimethoxybenzyl) to ameliorate difficult couplings and reduce on-resin aggregation, a dimer library identified Hmb as superior in ease of removal, and selective Hmb incorporation then allowed synthesis of a 22-mer purine-rich PNA that could not be made by standard coupling methodology ^[7]. Backbone protection interrupts the aggregate keeping the next monomer from the free amine.

Solubility fails on the same sequences. PNA oligomers generally show good aqueous solubility, but some PNAs in certain buffers — phosphate especially — pose solubility problems, especially PNAs longer than 12 units with purine content higher than 60%, G in particular. Labels may aggravate the problem (fluorescein, biotin) or alleviate it (rhodamine), extra ethylene glycol linkers most often remedy it, and introducing a few lysine-derived backbone monomers has been demonstrated to greatly increase aqueous solubility ^[3]. A single lysine is routinely used to improve solubility and prevent aggregation of the neutral, hydrophobic amide backbone ^[8].

Where that lysine sits matters less than intuition suggests. A systematic study of lysine location (N- versus C-terminus) and chirality (D- versus L-) on triple-helical binding of PNA to double-stranded RNA and DNA confirmed that lysine conjugation significantly increased triplex stability while conjugation with noncharged amino acids decreased affinity, and found neither the location nor the chirality of lysine had significant impact on affinity for either target — the positive charge appears to be the main driving force ^[8]. The same work found PNA affinity for dsRNA about an order of magnitude higher than for the same dsDNA sequence, a triplex result against double-stranded targets that does not transfer to duplex hybridisation.

Mismatch discrimination

A mismatch in a PNA/DNA duplex is generally more destabilising than the same mismatch in DNA/DNA. Testing all single-mismatch combinations in a stretch of four bases in the middle of a 15-mer, the average ΔT_m was 15 °C in the PNA/DNA duplexes against 11 °C in the DNA/DNA duplexes; more useful than the average, the lowest measured ΔT_m was 8 °C for PNA/DNA against 4 °C for DNA/DNA — PNA is roughly twice as discriminating in the least discriminating base pair ^[3].

Position and identity both matter, not in the direction most probe designs assume. Quantifying all single-base mismatched pairs between an 11-mer PNA and its complementary DNA by real-time affinity electrophoresis in PNA-containing gels gave dissociation constants from 10^-7 M for mismatches to 10^-10 M for fully complementary DNA. Stability was at a minimum when the mutation sat four bases from either terminus — a loss of 20 °C or more in melting temperature — but regained substantial stability when the mismatch was at the centre of the duplex. The most stable mismatched pairs were G:T and T:T, whereas destabilisation was maximal for A:A and G:G ^[11]. Both the position rule and the identity ranking come from that one 11-mer against DNA; take them as the shape of the effect, not as coefficients.

The penalty also shows up kinetically: DNA and RNA sequences containing single mismatches at various positions in the centre gave increased dissociation and decreased association rate constants, and the PNA·RNA duplexes followed quantitatively the same variation with mismatches as the PNA·DNA duplexes did ^[10].

Triplex, bis-PNA, and invading duplex DNA

Homopyrimidine PNA forms a (PNA)₂/DNA triplex — an antiparallel Watson-Crick strand plus a parallel-bound Hoogsteen strand — stable enough to invade intact double-stranded DNA ^[3]. Linking the two strands into a bis-PNA improves this. Connecting two seven-nucleobase PNA segments in a continuous synthesis via a flexible linker of three 8-amino-3,6-dioxaoctanoic acid units gave triple-stranded complexes of somewhat higher thermal stability than monomeric PNA with complementary oligonucleotides, with very little hysteresis in the melting transition, and these bis-PNAs were superior to monomeric PNAs for targeting double-stranded DNA by strand invasion. Placing pseudoisocytosine (J), which mimics N3-protonated cytosine ^[3], in the strand parallel to the DNA complement — the Hoogsteen strand — made the DNA binding pH independent ^[16]. Hysteresis greater than 20 °C in a 2PNA/DNA triplex melting curve falls to a few degrees on linking ^[3], which is what makes the transition interpretable.

Invasion is slow, cation-sensitive and kinetically controlled. Strand-invasion complexes require opening the DNA double helix, so their formation is slow and very sensitive to duplex-stabilising cations such as Mg²⁺ and spermine or spermidine, as well as to medium concentrations of simple monocations; it is usually not possible to detect strand invasion at physiological salt concentrations using simple monomeric PNAs. Equipping the PNA with three or four positive charges, for example terminal lysines, increases the binding rate by three orders of magnitude, and such PNAs — especially as bis-PNAs — do invade duplex targets at physiological ionic strength. Once formed, triplex invasion complexes with 10-mer or longer PNAs are exceedingly stable, with half-lives counted in days, and resistant to more than 500 mM sodium chloride — so complexes may be pre-formed at low ionic strength, typically 10 mM Tris-HCl with 1 mM EDTA at pH 7.4, then transferred to the buffer the downstream enzymology requires. Because no significant dissociation occurs over a 1-16 hour experiment, a binding constant K = k_a/k_d is not operationally meaningful; binding efficiency, and sequence discrimination with it, directly reflects the association rate ^[3].

The competing product is a plain triplex, and it usually arrives first. Comparing a decamer, a dodecamer and a pentadecamer thymine-cytosine homopyrimidine PNA against a sequence-complementary homopurine target in duplex DNA, all three formed stable triplex invasion complexes and also conventional triplexes; the triplexes formed with much faster kinetics and prevailed at lower PNA concentrations and shorter incubation times, and increasing ionic strength strongly favoured triplex over invasion, the latter being severely inhibited by cations. Length bought heterogeneity rather than cleanliness — one triplex invasion complex for the decamer, two structurally different target-specific invasion complexes for the dodecamer, more than five for the pentadecamer. Isolated triplexes could be converted to specific invasion complexes without dissociation of the Hoogsteen base-paired triplex PNA, which the authors read as a clear example of a triplex-first mechanism for helix invasion ^[17].

The homopurine target requirement is the real restriction, and pseudocomplementary PNA is the route around it. PNAs containing diaminopurine·thiouracil base pairs bind with high specificity and efficiency to complementary targets in double-stranded DNA by double duplex invasion, in which the duplex is unwound and both DNA strands are targeted simultaneously, each by one of the two pseudocomplementary PNAs. On that basis the authors predicted that, for decameric targets, more than 80% of all sequences can be targeted by straightforward Watson-Crick base pairing in the approach's then-present form; targeting the T7 phage RNA polymerase promoter effectively inhibited transcription initiation in vitro ^[18].

Cell permeability and conjugates

Unconjugated PNA does not cross membranes usefully. In a comparison across five cell types, a 14-mer PNA attached by disulfide linkage or maleimide coupling to pAnt, a 17-residue fragment of the Drosophila protein Antennapedia, or to pTat, a 14-amino-acid fragment of the Tat protein, was readily taken up at 2 µM by SK-BR-3, HeLa and IMR-90 cells; in U937 cells the conjugates and a PNA carrying lysine backbone units gave only very weak and exclusively vesicular staining, and no evident uptake of the unmodified PNA was seen ^[19]. Two findings from that work govern how any PNA uptake image should be read. The PNA was almost exclusively confined to vesicular compartments in the cytosol — the IMR-90 cells also showed a weak diffuse cytoplasmic staining — so entering the cell is largely not reaching the cytosol or nucleus; and a cytotoxic effect resulting in artificial staining appeared above 5-10 µM in the medium, depending on cell type and incubation time, which is why the authors concluded that proper controls for cytotoxicity should always be carried out to avoid misinterpretation of visual data. Uptake was time-, temperature- and concentration-dependent and mainly endocytotic ^[19].

The conjugate can be built without leaving the resin: Hmb backbone modification allowed a difficult purine-rich PNA to be synthesised, and the chain then continued to include a cell-penetrating peptide on the same support ^[7].

Analysis: mass first, then melting

PNA behaves well in MALDI-TOF. The molecules survive the MALDI process intact and are easily ionised with almost no multiply-charged species, which lets mixtures — deletion products from a difficult purine-rich synthesis, for instance — be easily characterised. Traditional protein matrices (sinapinic acid, 2,5-dihydroxybenzoic acid, α-cyano-4-hydroxycinnamic acid) were superior to DNA matrices (trihydroxyacetophenone, 3-hydroxypicolinic acid, picolinic acid), while the DNA matrix 6-aza-2-thiothymine also worked well, and PNA purity and sequence information were obtainable at low picomole levels ^[20].

Melting curves are the standard affinity readout and the standard trap. Thermodynamic parameters obtained by absorption hypochromicity agree well with parameters from calorimetric measurements, but fluorescence-energy-transfer detection of duplex melting gives in most cases more favourable free energies of hybridisation, a discrepancy ascribed to the conjugated dyes substantially affecting duplex stability. In decameric mixed purine-pyrimidine duplexes, dianionic fluorescein on a flexible linker raised the melting temperature by 3-4 °C where it could interact with an opposing dicationic lysine on a hybridised PNA strand, whereas for the PNA/DNA duplex with fluorescein attached to the PNA strand there was a large destabilisation of ΔT_m = -12 °C relative to the unlabelled duplex, probably from electrostatic repulsion against the negatively charged DNA backbone ^[14]. A labelled probe's T_m is that labelled probe's T_m.

The van't Hoff step is the other soft joint. Comparing UV melting against isothermal titration calorimetry on the same 10-base duplexes, van't Hoff enthalpies and extrapolated ΔG values agreed more closely with the ITC values for the DNA/DNA duplexes than for the PNA/DNA duplexes ^[13]. If a real enthalpy for a PNA/DNA duplex is needed, measure it calorimetrically; if a van't Hoff enthalpy is reported, report it as one.