Atomically precise binding conformations of adenine and its variants on gold using single molecule conductance signatures

Detecting the binding conformation of biologically functional molecules, such as nucleobases, on noble metal surfaces is important for biosensor development, biotic–abiotic interface engineering, and electronic measurements of biomolecular conductivity.¹ Of the nucleobases, adenine, in particular, has demonstrated the strongest binding affinity for gold.^2,3 Developing an understanding of gold–adenine chemistry and interactions has the potential to advance the development of biosensors that effectively detect and distinguish biologically relevant adenine derivatives. Numerous studies on the adsorption orientation of adenine on noble metal surfaces have been performed in the last several decades. These include surface^4–7 and tip-enhanced^8,9 Raman scattering methods (SERS and TERS, respectively) and other surface-enhanced vibrational spectroscopic techniques.^5,10,11 In addition, density functional theory (DFT)^4,6,12,13 has been widely used to simulate the surface interaction and determine the most probable absorption orientations of adenine on gold or silver surfaces. Collectively, these studies have established that the interaction of adenine with the gold surface occurs primarily through the lone pairs on the nitrogen atoms within the molecule. However, the presence of numerous nucleophilic nitrogens within adenine can potentially compete for the binding to gold and make the determination of the most likely configuration of adenine on gold challenging.¹⁴ 

Here, we tackle the long-standing question on the most likely binding orientation of adenine to gold using a combination of single adenine–gold junction conductance measurements and DFT calculations of the junction structures and transport. We also demonstrate how single molecule conductance measurements can be used to distinguish the binding of adenine from its biologically relevant variants, 6-methyladenine and 2′-deoxyadenosine. In the context of single molecule conductance measurements, donor–acceptor binding of the nitrogen lone pairs to the gold electrodes results in metal–molecule–metal junctions whose conductance can serve as a signature of molecule–metal binding geometry. Lone pairs on amine,¹⁵ pyridine,¹⁶ and imidazole¹⁷ molecular moieties have been shown to form stable and selective donor–acceptor bonds to undercoordinated gold atoms on metal electrodes. Single molecule conductance measurements of molecules containing two such linkers result in molecular junctions whose conductance signatures are exquisitely sensitive to the binding chemistry and orientation, allowing for a conductance-based mapping of binding geometry with atomic sensitively to interface structure.¹⁸ 

Adenine (Ade) is one of the purine nucleobases, containing two fused nitrogen-containing heterocycles. The five-membered imidazole ring consists of a pyridine and a pyrrole nitrogen N7 and N9, which can tautomerize into both the N7H and N9H forms shown in Fig. 1(a).¹⁹ The six-membered pyrimidine ring contains two pyridine-like nitrogen atoms, N1 and N3, and an exocyclic amino functional group with the nitrogen atom N10. Within the DNA double helix, the pyrrole-like N9 is bound to the deoxyribose moiety, which then links to the phosphodiester backbone. The Watson–Crick pairing between the adenine and thymine within the double helix occurs through the hydrogen bond between N1 and N10 of the pyrimidine ring. The molecule is entirely planar, which makes it compatible with the rung-on-a-ladder structure in the DNA.²⁰ The isosurface plots of the frontier orbital electron density in Fig. S1 show that the electron-rich areas on the molecule correspond to the π system as well as the nitrogen lone pairs. These features enable the base-pairing and the π–π stacking within the DNA structures.

Here, we use the Scanning Tunneling Microscope-based Break Junction (STMBJ) technique to measure the conductance of single Ade and Ade-derivative molecules on gold. By recording and analyzing statistically significant datasets of single molecule conductance measurements, we identify reproducible conductance signatures corresponding to distinct binding configurations of this molecule on metal electrodes.^21,22 We compare conductance features of Ade and other structurally similar molecules that lack one or more of the candidate binding sites of Ade. Through this systematic comparison, we assign individual molecular features to specific metal–molecule binding configurations. Our results indicate that Ade binds with gold electrodes through N7, N3, and N9 but not through N1 or N10. We identify N9, for the first time, as a strong binding site in slightly basic conditions when the molecule becomes deprotonated. DFT calculations of the junction structures and charge transport support these conclusions and identify the importance of charge transfer in the binding strength of different conformations. More broadly, our combined experiments and theoretical approach provides a detailed picture of the most prevalent Ade–gold binding conformations and how they depend on the solvent environment and electrode geometry. We also demonstrate that conductance signatures unambiguously distinguish Ade from its biologically relevant derivatives, 6-methyladenine and 2′-deoxyadenosine, in single molecule junctions. These results establish single molecule conductance measurements as a powerful new tool for characterizing the atomic structure of metal–organic interfaces and for biosensing applications.

We probe the formation and conductance of metal–molecule–metal junctions using our homebuilt STMBJ set up under ambient conditions at room temperature as described in detail previously.^15,23,24 The conductance measurements are performed by repeatedly bringing a gold tip in and out of contact at 16 nm/s with a gold substrate to form, stretch, and break atomic junctions. We record conductance during the junction stretching process to obtain a conductance-displacement trace as shown in Figs. 1(b) and 1(c). A constant bias of 500 mV is applied during junction formation. A conductance plateau at 1G₀ indicating the formation of a single gold atom point contact can be observed in sample traces shown in Fig. 1(c).²⁵ The rupture of the gold atomic contact opens a tunnel gap where molecules can bind and persist during junction stretching, resulting in molecular conductance plateaus. As the junction is stretched, the distance between the electrodes increases and the molecule may change binding orientation to remain in the junction before detaching from the contacts as shown in Fig. 1(b). The resulting conductance plateaus below 1G₀ appear in a fraction of traces measured in the presence of molecules containing linker groups, such as amine, pyridine, imidazole, or others; only molecules with at least two attachment sites can effectively bridge the gap and persist in the junction. Sample conductance traces measured in the presence of Ade are shown in Fig. 1(c). All molecules measured here, depicted in Fig. 2(a), are purchased from Sigma-Aldrich. We deposit the molecules onto a gold substrate using the dip-coating procedure described previously¹⁷ and detailed in the supplementary material.

Thousands of traces measured in the presence of the molecule of interest and containing conductance and electrode displacement information are collected and then binned into conductance histograms using logarithmic binning without any data selection.²² Reproducible molecular conductance plateaus appear as peaks in conductance histograms shown in Figs. 2(b) and 2(c) and correspond to conductance signatures of molecular binding geometries.

DFT calculations are performed to evaluate the interactions between Ade and the gold electrode. Since Ade is deprotonated under basic conditions (Fig. S2), we use the Ade anion with a deprotonated N9 atom in the calculations to account for the pH effect observed in the experiments. An Au₂₀ pyramid with or without an adatom on the tip is used to represent the gold electrode in the sharp or blunt configuration, respectively. The electronic structures of the systems are described using the PBE0 density functional,²⁶ the D3 dispersion correction,²⁷ and the 6-311G(d,p) basis set for Ade and the LanL2DZ basis set for the Au atoms. All the calculations are performed using the conductor-like polarizable continuum model (PCM) with a dielectric constant of 78.4 to mimic the aqueous environment. Here, we do not treat the water molecules explicitly due to the high computational costs associated with accurately describing the fluctuating solvation structures around the Ade anion using quantum chemistry methods, which is beyond the scope of the current work. Despite its lack of molecular details, we expect the PCM to capture the solvation environment in experimental measurements and give qualitatively correct predictions to the binding geometries and affinities.

We first compute the binding energies between Ade and the sharp or blunt gold tip using the Gaussian 16 program.²⁸ We optimize the structures of the sharp and blunt Au₂₀ pyramids in the aqueous environment and fix the positions of the gold atoms in the following analysis of the Ade–gold complexes. To construct the initial structures of the complexes, we place the Ade anion close to the Au₂₀ pyramid in the sharp or blunt configuration, with a distance of 2.5 Å between the N1, N3, N7, or N9 atom and the tip of the pyramid. After optimizing the geometry of each Ade–gold complex, we obtain the binding energy as $ΔEINT(s)=Ecomplex−(EAde+Egold)$⁠, where E_Ade and E_gold are the energies of the isolated Ade anion and Au₂₀ pyramid, respectively. Note that E_Ade is calculated from the geometry in the optimized Ade–gold complex so that $ΔEINT(s)$ can be used in the energy decomposition analysis. To assess the validity of this approach, we repeat the calculations by optimizing the structure of the isolated Ade and find that the differences in the resulting binding energies are within 0.7 kcal/mol (Table S1). To ensure that the calculations reach the optimal complex geometries, we vary the initial structures by keeping the corresponding N–Au distance at 2.5 Å and rotating the Ade anions by 30°, 45°, and 60°. In addition, we vary the initial positions of Ade around the blunt Au₂₀ pyramid by placing its N1, N3, N7, and N9 atoms at a distance of 2.5 Å away from 1, 2, or 3 Au atoms on the top surface of the pyramid. We then repeat the optimization processes; the reported $ΔEINT(s)$ values are from the most stable Ade–gold complexes. We also carry out the same calculations in vacuum with counterpoise corrections²⁹ and find that the relevant basis set superposition errors in the binding energy calculations are 3.3–3.9 kcal/mol. We then perform energy decomposition analysis on the optimized Ade–gold complexes using the ALMO-EDA(solv) method³⁰ as implemented in the Q-Chem 5.3 software.³¹ 

Calculations of metal–molecule–metal junction geometries are performed in the AITRANSS package of the Fritz Haber Institute ab initio molecular simulation (FHI-aims)^33–34 with the Perdew–Burke–Ernzerhof (PBE) density functional for the exchange-correlation.³⁵ The Kohn–Sham states are computed with an optimized all-electron numeric atom-centered tight basis set (similar to a double zeta plus polarization basis set) for the Ade molecule.³⁶ In this set of calculations, we use the Au₁₈ pyramid, which is a standard geometry within the FHI package, and relax its structure prior to forming the junction. The gold electrode atoms are calculated using the loose basis set (similar to a double zeta basis set).^33,37 The final structures are achieved by freezing all atoms except the Ade molecule and the top two layers of gold atoms, which are allowed to relax until the atomic forces reach 0.01 $eV×Å−1$ or lower. The nonequilibrium Green’s function method implemented in the AITRANSS package^32,34,38 is used to calculate the energy-dependent electron transmissions across the relaxed junction geometries at zero bias with the same level of theory.

A conductance histogram made without data selection from at least 6000 consecutively measured traces in the presence of Ade is shown in Fig. 2(b) (green). We observe several clear peaks in this unfiltered histogram data that arise from several bridging configurations of Ade on gold. We hypothesize that the multiple amine and imine moieties on the molecule, specifically N1, N3, N7, N9, and the exocyclic N10 amine, allow for multiple distinct bridging configurations in the junction. The binding of several molecules in the junction in parallel can also lead to additional conductance peaks in the histogram.¹⁷ 

To disentangle the adsorption configurations corresponding to the distinct conductance signatures, we select molecules that are structurally similar to Ade but lack one or more of its candidate binding sites, and we probe their conductance. As shown in Fig. 2(a), all the molecules are derivatives of benzimidazole (BIm, pink), which contains a fused system of imidazole and a six-membered benzene ring with no nitrogen lone pairs for binding to gold. Similar to Ade, in purine (Pu, black), the six-membered ring is a pyrimidine that contains two nitrogens: N1 and N3. To distinguish between these two binding sites, we consider two other control molecules, CN1 and CN3, shown in Fig. 2(a), which contain one pyridine nitrogen in positions analogous to the N1 or N3 positions in Ade, respectively. The full compound names and numbering of atoms within molecules in Fig. 2(a) are shown in Fig. S3. As we vary the presence of the nitrogen binding sites by tuning the structure of the molecules, we compare the reproducible conductance features appearing in corresponding histograms.

We aim to distinguish the role of the imidazole, the pyrimidine ring, and the exocyclic amine on the binding of Ade to gold. First, we compare the conductance histograms of BIm, Pu, and Ade in Fig. 2(b). We observe that all three molecules display a conductance peak near 1 × 10⁻²G₀. We labeled this peak as HG1 and determined the most likely HG1 conductance value by fitting a Gaussian to the linear histograms for each molecule as shown in Fig. S4. The resulting HG1 values are listed in Table I. We compare the HG1 peak to the most likely conductance of imidazole, which is a common component of all the molecules in the sample and has a value of 1.9 × 10⁻²G₀.¹⁷ Imidazole bridges the tip–sample junction only in basic conditions when both nitrogens can be deprotonated and bind to gold.^17,39

We hypothesize that the observed conductance features of BIm are due to binding through imidazole nitrogens and will occur at pH conditions above the pKa of BIm. We perform single molecule conductance measurements on BIm in a range of pH conditions and plot the resulting conductance histograms in Fig. S4. We observe that the molecular signatures of BIm are abrogated in acidic environments, consistent with binding through imidazole. These experimental results allow us to assign the HG conductance features of BIm, and by analogy of the other molecules in our sample including Ade, to a configuration where the molecule bridges the tip–sample gap through the imidazole moiety nitrogen atoms N7 and N9. Slight shifts in the HG1 conductance values listed in Table I are consistent with the effect of substituents and chemical structure variations on orbital energies as previously documented in the literature.^41–42 

All three molecules plotted in Fig. 2(b) also show peaks at values of conductance higher than HG1, which we term HG2. The results of Gaussian fits to HG2 are listed in Table I and occur near an integer multiple of the main HG1 feature. As with imidazole, multiple molecules can bridge the junction in parallel to result in these secondary conductance features.¹⁷ We also note that both BIm and Pu display a higher junction yield as indicated by the higher amplitudes of HG signal than Ade. This result indicates that Ade bridges a smaller fraction of the junctions in the HG configuration than the BIm and Pu variants possibly because of the steric interference of the amine or the reduced charged density at the binding sites due to the electron withdrawing effects of the other pyridine-like nitrogens in Ade.^15,40,43

In addition to the HG peaks discussed above, Pu and Ade display a lower conductance signature, at 3.6 × 10⁻³G₀, which we term the LG peak. We investigate the role of the exocyclic amine N10 in Ade binding in the LG configuration. We note that Ade and Pu differ only in the presence or absence, respectively, of N10. However, the conductance histograms of the two molecules share the same conductance features. These results suggest that neither the HG nor the LG features observed in Ade are attributable to N10 binding. To verify this result, we measure the conductance of 6-methylpurine (6MePu) where the exocyclic amine group is replaced by a methyl group. The resulting conductance histogram in Fig. S6 shows the same number and position of conductance features, confirming that N10 does not serve as an anchor point for Ade on gold electrodes. We conclude that the pyrimidine nitrogens N1 and N3, instead of the N10 amine, must participate in binding to result in the LG configuration.

The lack of amine binding in the junction is surprising in light of the established view in the literature that amine-linked molecules can effectively bridge metal–molecule junctions through a donor–acceptor bond between the electron lone pair on the amine and an undercoordinated gold atom on the electrode.¹⁸ We probe the difference in the electronic structure between the N10 atom in Ade and the amine linkers used typically in break junction single molecule conductance measurements, such as in 1,4 benzenediamine (BDA).^45–46 We compare the DFT-calculated gas phase structure of Ade and BDA in Fig. S7. The N–C bond length of Ade and BDA is 1.35 and 1.40 Å, respectively. Importantly, the dihedral angle between the two H-atoms on the amine, which is defined as the angle between the two C–N–H half planes, is 180° in Ade and 129° in BDA. These differences suggest that the N10 atom in Ade has sp² character, whereas it is predominantly sp³-hybridized in BDA. The planar configuration of N10 in Ade results in a delocalized nitrogen lone pair, which is unavailable for binding to gold.

To distinguish which of the two sites, N1 or N3, on the pyrimidine results in the LG conductance configuration, we collect thousands of conductance traces in the presence of two Pu derivatives: CN3 and CN1. The corresponding conductance histograms are shown in Fig. 2(c). Both molecules display HG conductance features, consistent with the shared imidazole moiety between all the molecules in the sample. However, CN3, but not CN1, binds in the LG conductance configuration. In addition, CN3 has a higher HG1 and HG2 conductance signal than CN1, suggesting that the N3 nitrogen participates in anchoring the molecule in the HG configuration as well. We can conclude that the N3 site, rather than the N1 site, in Ade is necessary for the LG conductance signature and is directly involved in the molecule–gold binding.

We note that although N1 does not participate in binding to gold, it decreases the amplitude of conductance signatures in Pu, CN1, Ade, and 6MePu compared to BIm and CN3, as seen in Figs. 2(b) and 2(c) and S6, reflecting lower formation and persistence probability of these molecular junctions.^22,47 This trend is consistent with the electron withdrawing effect of a pyridine nitrogen in the N1 position, which can weaken the donor–acceptor bond of the molecules to gold. With Ade, junction formation is further impeded by the steric repulsion between the N10 amine group and the gold electrode. A similar steric effect is observed with 6MePu in Fig. S6, where the amine is replaced by a methyl group. However, the LG peak of Ade is broader than that of 6MePu or of Pu. We point out that two tautomeric forms of Ade molecule exist involving the exchange of hydrogen between the N1 and N10 sites. This is shown in Fig. S8. These distinct forms would have slightly different electronic structures and may result in broadened conductance peaks observed here for Ade. Further analysis of LG conductance plateaus and broadening is described in the supplementary material and shown in Fig. S8.

To summarize our results so far, we list all the molecules in our sample and the most likely conductance values obtained from Gaussian fits to the linear histograms in Table I. Our results indicate that Ade has well-defined bridging geometries in the metal–molecule–metal junction and it binds on gold through the imidazole moieties, N7 and N9, and through pyrimidine N3. We cannot discount the possibility of simultaneous binding to all three sites.

To probe these junction geometries in more detail, we track how molecular conductance signatures evolve during junction stretching. Two-dimensional (2D) conductance histograms allow us to retain both the conductance and displacement information and correlate conductance to molecular binding configurations in the junction.²² A 2D histogram for Ade, Fig. 2(d), indicates that the LG plateaus occur at longer elongation compared to the HG features. The HG features, on average, begin immediately following G₀ rupture and last for ∼1.5 Å. The LG conductance plateaus appear after an additional ∼0.5 Å of stretching following gold point contact rupture and persist for ∼2 Å on average. We note that the N7–N9 and N7–N3 distances in Ade are 2.2 and 3.6 Å, respectively, as calculated by DFT methods. These results indicate that at shorter tip–sample distances, the conductance path across the molecular circuit occurs through the more proximal N7–N9 sites. The binding through the imidazole N7 and N9 atoms is stronger than a typical donor–acceptor bond due to the deprotonation of the imidazole moiety and added electrostatic interactions.^17,40 As the junction is stretched and the tip–sample distance widens, a reorganization of the molecule in the junction occurs to yield a binding configuration with a conducting pathway that includes N3, as illustrated in Fig. 1(b).

Complementary to the single molecule conductance measurements, we carry out DFT calculations to investigate the electronic structures of the molecular junctions and the charge transfer process that affects the junction formation and conductance characteristics. Here, we use the Au₂₀ pyramid structures both with and without an adatom on the tip to approximate the sharp and blunt electrodes, respectively. We find it essential to perform the calculations in an aqueous environment, which we represent using the PCM, as it stabilizes the charged species in the molecular junctions and reduces the binding energies between the Ade anion and the gold electrode by over 24 kcal/mol as compared to those in vacuum (Table S1). Since it is generally accepted in the field that the blunt tips containing two or more apex atoms predominate in break junction single molecule measurements, especially in room temperature conditions when reorganization and flattening of the tips typically occurs immediately following rupture of the 1G₀ contact,^25,47–51 we will mainly discuss the calculation results using the blunt gold electrode in the following. The optimal binding geometries of deprotonated Ade on the electrode are shown in Figs. 3(a)–3(d), and the binding affinity of the different nitrogen sites follows the trend of N3 $≈$ N9 > N7 > N1. In all cases, one or more nitrogen atoms in Ade adsorb on the electrode with a distance of 2.2 Å from the closest Au atoms. Interestingly, the blunt tip configuration provides a large contact surface for the binding and the constrained purine ring structure allows the N3 and N9 atoms to interact simultaneously with the gold electrode, leading to nearly identical optimal structures for the binding of Ade through the N3 and N9 sites and the strongest binding energies of −40.7 kcal/mol [Figs. 3(b) and 3(d)]. This is followed by the N7 site with an interaction energy of −30.6 kcal/mol between Ade and the Au₂₀ pyramid. Consistent with the experimental observations, binding through the N1 site is energetically least favorable, possibly because the exocyclic amino group in Ade hinders its contact with the gold electrode. We observe a similar trend in the binding of Ade on the sharp Au₂₀ pyramid, as shown in Fig. S9 and Table S2. However, the differences between the binding energies are much smaller than those in the blunt cases because only one of the N1, N3, N7, and N9 atoms of Ade can bind to the tip of the gold electrode in each configuration.

To uncover the origin of the observed trend that N3 $≈$ N9 > N7 > N1 in the binding affinities, we use the blunt Au₂₀ pyramid as an example and perform energy decomposition analysis for its interactions with Ade using the ALMO-EDA(solv) method.³⁰ The total interaction energy, $ΔEINT(s)$⁠, is partitioned into the frozen interaction, polarization, and charge transfer energies as follows:

The superscript (s) is included since the calculations are performed in the solvation environment.

The frozen interaction energy, $ΔEFRZ(s)$⁠, describes the energy difference between the Ade–gold complex and the isolated Ade and Au₂₀ pyramid without relaxing their orbitals. It can be further partitioned into the Pauli repulsion, permanent electrostatics and dispersion contributions in vacuum and a solvation term.³⁰ The effect of orbital relaxation is incorporated in the polarization and charge transfer contributions, $ΔEPOL(s)$ and $ΔECT(s)$⁠, respectively. As shown in Fig. 3(e), the frozen interaction weakens the binding of Ade on the gold electrode, while polarization and charge transfer act to stabilize the bound complex. From Table S2, the positive $ΔEFRZ(s)$ terms mainly come from the large Pauli repulsion (>119 kcal/mol) between Ade and its nearby Au atoms in the pyramid, with a small contribution from the solute–solvent interactions that smear out the net charge on the Ade anion. The attractive electrostatic and dispersion interactions cancel about 90% of these unfavorable interactions, leading to a $ΔEFRZ(s)$ of ∼14 kcal/mol for the N1 and N7 sites and 36 kcal/mol for the dual binding through the N3 and N9 sites. As the gold electrode approaches the Ade anion, their molecular orbitals overlap and charge reorganization occurs. Accordingly, the $ΔEPOL(s)$ and $ΔECT(s)$ terms are negative and are the main sources of stabilization in the binding process [Fig. 3(e)]. Among all the nitrogen sites, the N3 and N9 atoms of Ade can interact simultaneously with the Au₂₀ pyramid, providing the most stable Ade–gold complex structures. Compared to N7, the N1 site gives the $ΔEPOL(s)$ and $ΔECT(s)$ components that are 2.6 and 3.0 kcal/mol higher in energy, respectively, making the binding though the N1 atom least favorable among all the nitrogen sites. Similar properties are observed for the sharp Au₂₀ pyramid (Fig. S5 and Table S2). Therefore, the binding of Ade on the gold electrode results from a series of competing interactions, and electrostatics, polarization, and the purely quantum mechanical charge transfer and dispersion interactions play a major role in determining the observed preference for the binding of different N atoms.

We compare the conductance expected for Ade junctions arranged in one of the lowest energy binding configurations discussed above. Figure 4 (top inset) shows converged junction geometries where Ade bridges the gold pyramids through N7 on one side and N9 (N7/N9, red), N3 (N3/N7, green), or both (N3/N7/N9, black) on the other. The calculated transmission curves plotted in Fig. 4 for these junction configurations represent energy-dependent electron transport across the junction.^32,38 We compare the calculated transmission value at the Fermi energy (0 on the energy x-axis) to our low-bias conductance measurements reported above as is standard practice in the literature.^54–56 We observe that the binding through N7/N9 and N3/N7/N9 results in very similar predicted conductance, which we cannot distinguish in our experiment. Since the N7/N9 geometry is bound less strongly that of N3/N7/N9, we can assume the latter is more likely to occur in Ade during the HG configuration. The more stable anchoring through the three nitrogen atoms can also account for the higher probability of forming HG1 and HG2 junctions with CN3 than with CN1. However, the N3/N7 junction has a significantly lower conductance since the imidazole bridge through N9 is ruptured. The DFT-predicted conductance values for these two configurations are 2.8 × 10⁻²G₀ and 1.2 × 10⁻²G₀, which are several times higher than the measured results for the HG1 and LG peaks, respectively. The degree of DFT overestimation of conductance we observe here is consistent with established trends in the literature. The Au–Au distance in the imidazole-bridging junctions N7/N9 and N3/N7/N9 is ∼6.2 Å, which is 1 Å shorter than in the N7/N3 junction. We conclude that the identified favorable Ade binding configurations are consistent with the conductance and elongation experimental measurements presented here.

Our results indicate that introducing changes to the structure of Ade through substituents that occlude sites N3, N7, or N9 will produce distinct changes in the conductance signatures. We propose that this sensitivity to molecular structure can be leveraged for biosensing applications to distinguish and detect the presence of Ade variants. To test this, we measure the conductance of 6-methyladenine (M6A) and 2′-deoxyadenosine (2′-dAdo), which are two biologically relevant Ade variants. M6A is a ubiquitous epigenetic modification and 2′-dAdo is a component of DNA and ATP with medicinal uses.⁵⁷ The structures of both molecules are shown in Fig. 5(a). Based on how Ade binds to Au, we hypothesize that in M6A, the steric hindrance from the methyl group at position 11 may block the junction formation from the N7 site. As N7 is key to anchoring Ade to one of the electrodes in the junction, we expect no conductance signatures in the presence of M6A. In 2′-dAdo, the N9 site is substituted so that imidazolate formation and binding are abrogated, leaving only the N7/N3 binding pathway characterized by the LG conductance signature.

The 1D conductance histograms constructed from at least 6000 traces measured in the presence of the variant molecules are plotted in Fig. 5(b). Consistent with our expectations, the M6A does not show any conductance signatures, whereas the 2′-dAdo displays a low amplitude peak at the LG position, suggesting that it binds in a small fraction of junctions through N3/N7. The conductance spectra of the three molecules in Fig. 5(b) are clearly distinct. While more work is needed to fully characterize the binding of these and other adenine derivatives, our results serve as a proof of principle for using conductance signatures for sensing binding orientations and distinguishing biomolecules.

In conclusion, our combined single molecule conductance and DFT computation study offers new evidence for the binding configurations of Ade on gold and establishes single molecule conductance measurements as a useful tool for biomolecule detection. We find that Ade–Au binding is robust at pH above the pKa of N7 when the imidazole moiety on Ade is deprotonated. In these conditions, our experimental measurements reveal two distinct single molecule conductance regimes: high (HG) and low (LG). Through systematic comparison of conductance signatures of Ade and its derivatives, we can unambiguously assign these conductance signals to specific binding conformations. We determine that deprotonated Ade reproducibly coordinates with gold through N3, N7, and N9 but not N10 or N1. The exocyclic amine in Ade does not bind reproducibly to gold due to the sp² character of the N10 atom and the delocalized nature of the N10 lone pair, which discourages donor–acceptor bonding to gold. DFT calculations show that binding through N7, N9, and N3 is more favorable than binding through N1 and that it is possible to form multiple contacts when Ade interacts with the blunt tip configuration, which further stabilizes the molecular junction. The stabilizing contributions to the interaction energy from the polarization and charge transfer reorganization, which occur upon binding of the deprotonated Ade to a gold electrode, are the smallest for the N1 pyridine site than for N3, N9, and N7. Calculated conductance through junctions bound through the most stable N3/N7/N9 and N3/N7 geometries is consistent with experimental measurements of the two conductance regimes. Our results provide new insights into the interaction of adenine with gold, which has been a subject of intense study and controversy. Additionally, we demonstrate how STMBJ can be used to study the interaction of biological molecules with noble metals and to detect variations in biomolecular structure by analyzing the conductance signals of single molecule–metal junctions. This work provides a blueprint for conductance-based sensing of biologically relevant molecules.

The supplementary material includes sample preparation procedures, calculated frontier molecular orbitals and structures of gas phased molecules, additional conductance measurements of control molecules, linear histograms and example Gaussian fitting, energy decomposition analysis of Ade bound on gold in aqueous and vacuum environment, and transmission spectra of Ade in different configurations.

This work was supported by the Air Force Office of Scientific Research under Award No. FA9550-19-1-0224. A.C. was supported by the Greater Boston Research Opportunities for Young Women (GROW) and Boston University Learning Resource Network (LERNet) at BU. L.W. acknowledges the Office of Advance Research Computing at Rutgers University for providing access to the Amarel server for the calculations.

The authors have no conflicts to disclose.

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

Xiaoyun Pan: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Writing – original draft (equal). Cheng Qian: Investigation (equal); Visualization (equal). Amber Chow: Data curation (equal). Lu Wang: Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Investigation (equal); Methodology (equal); Supervision (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Maria Kamenetska: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Funding acquisition (equal); Methodology (equal); Resources (equal); Supervision (equal); Validation (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.