The shape of neurotransmitters in the gas phase: A theoretical study of adrenaline, pseudoadrenaline, and hydrated adrenaline

The conformational landscapes of adrenaline and pseudoadrenaline have been explored by electronic structure computation at the B3LYP/6-31+G*, MP2/6-31+G* and MP2/aug-cc-pVDZ levels of theory.

Based on the relative energies of the different conformers, we conclude that the extended AG1a and folded GG1a conformers are the most likely candidates for detection in spectroscopy experiments using a supersonic expansion.

Predictions for the infrared spectra of the AG1a and GG1a conformers and their 1∶1 hydrates are presented.

We explored the suitability of a rigid-body DMA-based model potential to locate the most stable 1∶1 and 1∶2 adrenaline-water clusters.

The model potential was able to locate all relevant 1∶1 clusters, but failed to find the most stable 1∶2 cluster.


This paper is part of a combined theoretical/experimental study of the catecholamine neurotransmitters.

Previous work in this series include the investigation of the conformational landscapes of noradrenaline (NA),1 the simplest catecholamine neurotransmitter, and of singly hydrated NA.2

The current work focuses on the adrenaline (A)/pseudoadrenaline (PA) pair.

Noradrenaline has one chiral centre (the Cβ atom, see Fig. 1 for the labelling of the atoms), and thus occurs in two spectroscopically identical chiral forms (R and S).

Adrenaline/pseudoadrenaline is formed by the replacement of one of the hydrogens of the terminal NH2 group by a methyl group.

This makes the N atom chiral as well, which leads to the existence of two diastereoisomers: adrenaline (1R2S/1S2R) and pseudoadrenaline (1S2S/1R2R).

A biomolecule's molecular shape plays a crucial role in its transport properties, its selectivity and function, and its receptor binding properties.

Its shape and conformation result from a delicate balance of intramolecular and environmental influences.

The study of gas-phase biomolecules is essential in order to differentiate the relative importance of the intrinsic and environmental influences on the biomolecule's conformation.

In this article we provide a full theoretical characterisation of the neutral conformers of adrenaline and pseudoadrenaline in the gas phase, and provide a first investigation of the 1∶1 and 1∶2 hydrates of adrenaline.

The functional groups of the catecholamine neurotransmitters (the catecholic hydroxyls, the chain OH and NH groups, and the π electron cloud) provide many possible water-binding sites.

For the most stable NA conformer (AG1a), we located as many as eleven different AG1a–H2O structures.2

To find the most stable hydrates, one should not only investigate clusters involving the most stable conformer, but one has to consider several of the low-lying conformers, as the interaction with water may change the relative stability of the conformers.

In pseudoephedrine, for example, hydration results in a substantial rearrangement of its conformational landscape, altering the conformation at the global minimum from an extended (Aga) to a folded (Gga) conformation.3

In addition, the number of local minima increases steeply with the number of constituents in the cluster.

Consequently, a full study of the adrenaline hydrates will be a formidable task.

In this paper, we present the initial investigation of the 1∶1 and 1∶2 adrenaline hydrates.

We investigate the different water-binding sites, and we comment on the most appropriate theoretical methodology to compute the relative stability of hydrates containing different adrenaline conformers.

In addition, we explore the suitability of force-field calculations using a DMA (distributed multipole analysis)-based rigid-body model potential for scanning the potential energy surface of the hydrates.


Electronic structure calculations

The catechol OH groups have two possible orientations forming an intramolecular hydrogen bond: Cδ-syn (the catechol hydrogens are syn with respect to the Cδ–H group) and Cδ-trans (the catechol hydrogens are trans with respect to the Cδ–H group).

There is, in fact, a third orientation for the catechol OH groups.

In this one the catechol hydrogens point away from each other, preventing the catechol OH groups to form an intramolecular hydrogen bond.

This orientation was also identified as a local minimum on the potential energy surface of catechol,4–6 and caffeic acid (which consists of a catechol moiety and a propenoate side chain).7

It is however much less favourable than the orientations forming an intramolecular hydrogen bond.

We have therefore not considered this particular orientation of the catechol OH groups.

We first explored the conformational landscapes of adrenaline and pseudoadrenaline holding the catechol OH groups in the Cδ-trans orientation.

To provide starting structures for geometry optimisations, the γ1 = γ(Cδ–Cγ–Cβ–Cα), γ2 = γ(Cγ–Cβ–Cα–N), γ3 = γ(Cβ–Cα–N–C) torsion angles were initially set to 0°, 60°, 120°, 180°, 240° and 300°, and γ4 = γ(Cγ–Cβ–O–H) was set to 60°, 180° and 300°.

For γ1, γ2 and γ3 we initially used torsion angle step sizes of 60°.

As the eclipsed conformations were found not to be local minima on the potential energy surface, part of the conformational search considered only staggered conformations.

The resulting conformations were subjected to HF/6-31G* geometry optimisations.

The geometries of all conformers with energies less than 20 kJ mol−1 above the global minimum were re-optimised with density functional theory (DFT) using the B3LYP8,9 functional and the 6-31+G* basis set.

The Cδ-syn conformers of the ten most stable A and of the ten most stable PA conformers were located using B3LYP/6-31+G*.

One of the A Cδ-syn conformers (AG3a-2) appeared to be unstable.

The relative energies of the resulting nineteen A and twenty PA conformers were evaluated by single-point MP2/6-31+G* and MP2/aug-cc-pVDZ calculations.

Zero-point energies (scaled by 097610). and scaled (097610. for OH, 095611. for NH) harmonic vibrational frequencies were calculated at the B3LYP/6-31+G* level of theory.

Calculation of interaction energies

The A–H2O and PA–H2O structures considered in this work were optimised with B3LYP/6-31+G*.

The interaction energies were corrected for BSSE (basis set superposition error) by using the counterpoise procedure.12

The deformation energies were taken into account as well.

These are computed as the difference of the energy of A/PA (or H2O) at the geometry the monomer adopts in the complex, and the energy of the free molecule at its equilibrium geometry re.

As discussed in our recent paper on NA–H2O,2 there are two possible choices for the geometry of the free adrenaline.

If one is interested in how strongly water binds to a particular conformer, then the equilibrium geometry of this particular conformer should be used.

The resulting interaction energy is called the specific water interaction energy (ΔEH2O).

However, if one wants to compute relative interaction energies of A–H2O complexes consisting of different A conformers, then one should use the equilibrium energy of the most stable conformer.

Relative energies calculated like this are corrected for BSSE, in contrast to relative energies based on the A–H2O total energies.

Unless specified otherwise, we will use the second definition in the current paper.

To provide the interaction energies at 0 K, ΔE0, the (harmonic, scaled by 0.976) zero-point energies (ZPE) of AW and the separated fragments (computed in the harmonic approximation) were taken into account: ΔE0 = ΔE + EZPEAW − EZPEA(global)(re) − EZPEW(re)(For the specific water interaction energy , one should take the zero-point energy of adrenaline at the equilibrium geometry of the particular conformer one is interested in, instead of that of the global minimum).

In the current paper, the interaction energies are listed as De (=−ΔE) and D0 (=−ΔE0), so that positive values denote an attractive interaction.

The interaction energies of π-type hydrogen bonds are difficult to evaluate accurately, due to the large contribution of the dispersion energy to the interaction, which is known to converge slowly with addition of high angular momentum polarisation functions.13

Current standard density functionals cannot evaluate the dispersion energy quantitatively.14

MP2 in principle allows the calculation of dispersion-type interactions; however, MP2 calculations with large basis sets readily become intractable for the size of systems studied in this work.

An earlier study on indole-H2O and methylindole-H2O15 showed that MP2 calculations employing the compact interaction-optimised DZPi basis set16,17 do give accurate interaction energies, for conventional and π-type H-bonded complexes.

We have therefore computed the structures and geometries of the π-bonded complexes with MP2/DZPi.

The effect of BSSE on the intermolecular geometry can be quite large15,18 when using the DZPi basis set.

To correct for this in an approximate way, the hydrogen-bond distances in these complexes were corrected for BSSE by numerically locating the distance for which the counterpoise-corrected De has its maximum.1,15,17,18

The electronic-structure calculations were performed with Gaussian 98 (revisions A.9 and A.11)19 and NWChem (version 4.1)20 on clusters of 1.7 GHz Pentium 4 and 2.0 GHz dual Xeon PCs running Linux.

An Origin 2000 and a cluster of Sun 900 MHz V880 servers at the HiPerSPACE Computing Centre at University College London were used as well.

As usual, only the valence electrons were correlated.

Gaussian's “ultrafine” integration grid was used for the DFT calculations.

The optimisations were converged using Gaussian's “tight” criteria for the cutoffs on forces and step size.

Model potential calculations

The model potential used, dubbed MP2fit/DMA, was developed in a study on uracil-water,21 and used recently in an investigation of protonated nicotine interacting with water and a third body.22

It consists of an atom–atom 6–exp potential to describe the repulsion and dispersion terms, and a DMA (distributed multipole analysis) model for the electrostatic contribution.

All individual terms of this potential are pairwise additive, and thus, it can readily be extended to complexes containing more than two molecules.

The potential has different parameters for each atom type (C, N, O, H), and it distinguishes between hydrogens attached to C, polar hydrogens (attached to O or N), and water hydrogens.

The DMA model uses atomic multipoles up to hexadecapole, derived from ab initio wavefunctions (MP2/6-311++G(2d,2p) for water, MP2/6-311G** for adrenaline), and includes all terms in the atom–atom multipole expansion up to Rab−5.

The DMA was derived from the wavefunctions calculated with Gaussian 98 by using the GDMA23 program.

As the MP2fit/DMA potential is a rigid-body potential, the DMA for adrenaline needs to be recomputed for each different conformer.

The monomer geometries used were obtained by geometry optimisations at the same level of theory as used to calculate the DMAs.

The model potential calculations were done with the program Orient.24

Further details on the potential model are provided in ref. 25.

Ab initio and density functional theory results

Adrenaline and pseudoadrenaline

The structures of the sixteen most stable A and PA conformers (based on the single-point MP2/aug-cc-pVDZ energies) are shown in Fig. 2.

To classify the A and PA conformers we used the same notation as employed in our previous study on NA,1 which is based on previous studies on 2-amino-1-phenylethanol10,26 and the ephedra:27–29

(i) The A/G notation denotes the arrangements (anti or gauche) of the Cγ–Cβ–Cα–N and O–Cβ–Cα–N atom chains, respectively.

(ii) In NA, which contains an NH2 group instead of the NH(CH3) group in A and PA, the AG and GG conformers have either a short OH⋯N (AG1, GG1), or a short NH⋯O (AG2/3, GG2/3) contact.

(The AG2/3 and GG2/3 families differ depending on which amino H atom is involved in the H-bonding).

In A and PA, the corresponding AG2/3 and GG2/3 families will have an NH⋯O (if the free hydrogen in the corresponding NA structure is methylated) or a CH⋯O contact (if the hydrogen-bonding H atom is methylated).

(iii) The catechol OH groups can have two different orientations (Cδ-syn and Cδ-trans) and the side chain can be above or below the ring.

This is notated by “a” (Cδ-syn, above plane), “b” (Cδ-trans, above plane), “c” (Cδ-trans, below plane) and “d” (Cδ-syn, below plane).

Cδ-syn and Cδ-trans conformations corresponding to the same side chain conformation differ slightly in energy due to interaction of the catechol hydroxyl groups with the side chain functional groups.

This was also observed for caffeic acid7 for which the Cδ-syn conformers were found to have a slightly lower energy than the Cδ-trans conformers.

In A and PA the “a” conformers, which have a Cδ-syn orientation of the catecholic hydroxyls, appear to be favoured.

The conformers with an intramolecular OH⋯N hydrogen bond (AG1, GG1) are more stable than the NH⋯O hydrogen-bonded structures (AG2, AG3, GG2).

This is in agreement with the larger polarity of OH as compared to NH, as well as with studies on the hydrogen-bonding abilities of oxygen and nitrogen in different hydrogen-bonding environments (oxygens covalently bound to two non-hydrogen atoms of which at least one is sp2 hybridised30 and hydrogen bonds to monocyclic aromatic heterocycles31), which showed that nitrogen atoms are stronger hydrogen-bond acceptors than oxygens.

Rablen et al. likewise found that nitrogens are better hydrogen-bond acceptors than sp3-hybridised oxygens.32

(However, Vargas et al. found that NH⋯O interactions are more stable than NH⋯N interactions,33 which is in contradiction with the presumed larger hydrogen-bond acceptor ability of N as compared to O).

In our previous study on noradrenaline we found that the twenty most stable NA conformers are of the AG or GG types (AG1, GG1, AG2, AG3 and GG2).1

The twenty most stable A conformers (based on the B3LYP energies) include the equivalents of the AG1, GG1 and AG3 conformers found for NA, a second AG3 family and a GA family.

For completeness, we have also optimised the structures of the AG2 and GG2 adrenaline conformers with B3LYP/6-31+G* (followed by single-point MP2/6-31+G* and MP2/aug-cc-pVDZ calculations).

The relative energies of the resulting 27 A conformers, listed in decreasing order of stability based on the MP2/aug-cc-pVDZ energies, are collected in Table S1 of the ESI.

The twenty most stable PA conformers (based on the B3LYP energies) include the equivalent conformers of the AG1, GG1 and AG2 conformers found for NA, as well as a GA and GG3 family, which were not among the most stable NA conformers.

We have also calculated the relative energies of the AG3 and GG2 families, which were not among the twenty most stable PA conformers.

The relative energies Δ(Ee) and Δ(E0) of the resulting 28 PA conformers, in decreasing order of stability based on the MP2/aug-cc-pVDZ Δ(Ee) values, are listed in Table S2 of the ESI.

At the B3LYP level of theory, PA is slightly more stable than A; this order is reversed however at the MP2 level of theory.

Even though A and PA are roughly isoenergetic, it is not likely that there will be interconversion between the two diastereoisomers in supersonic expansion experiments, as interconversion via umbrella-inversion of the chiral amino group is expected to have a high barrier.

We have calculated the HF/6-31G* relaxed potential energy profile for the umbrella inversion starting from the AG1a conformer, and optimised the geometry of the transition state for this process (see Fig. 3).

The barrier height for this process is calculated to be 26.6 kJ mol−1.

As studies on conformational relaxation in supersonic jets indicate that barriers larger than 6 kJ mol−1 are sufficient to prevent conformational relaxation,34,35 it is very unlikely that A ↔ PA interconversion will take place at the low temperatures of supersonic jet experiments.

There are significant differences between the B3LYP and MP2 relative energies of A and PA.

This is demonstrated more effectively in Figs. 4 and 5, which show the relative energies Δ(E0) of the A and PA conformers grouped together in their conformational families.

B3LYP underestimates the relative stability of the GG1 conformers, which contain a short N–H⋯π (A) or CH⋯π (PA) contact, of the GG2 conformers, which also contain a short N–H⋯π (PA) or CH⋯π (A) contact, and of the GA conformers, which contain a short NH⋯π interaction.

Thus, for conformers that contain π-type interactions, B3LYP is less reliable for calculating relative energies.

Although B3LYP predicts the adrenaline GG1a conformer to be less stable than AG1a by 2.4 kJ mol−1 (3.0 kJ mol−1 after zero-point vibrational corrections), at the single-point MP2/aug- cc-pVDZ level of theory this energy gap has decreased to a mere 0.49 kJ mol−1 (1.06 kJ mol−1 after vibrational corrections).

The narrow energy gap between the adrenaline AG1 and GG1 conformers was also observed for NA1 and APE (2-amino-1-phenylethanol),10 the benzene analogue of NA.

However, whereas both AG1 and GG1 APE conformers were observed experimentally, the NA experiments, surprisingly showed only one NA conformer, which was assigned as AG1a.1

(In the NA experiments, gas-phase NA was generated via laser ablation rather than evaporation using a heated nozzle, which may influence the observed conformer distribution).

Similarly, the most intense band in the UV spectrum of APE was assigned to the extended AG conformer, even though MP2 calculations predicted the GG structure to be slightly more stable.10

As different levels of theory yield different results on the relative stability of AG1a and GG1a, assignment of AG1a or GG1a as the global minimum cannot be based on the calculated energy differences alone.

It is likely that also for adrenaline the AG1a conformer will be the preferred structure.

However, as for NA, this issue cannot be resolved without spectroscopic investigation.

To aid future spectroscopic assignment, we have listed the harmonic frequencies of the hydride stretches of the AG1a and GG1a conformers, calculated at the B3LYP/6-31+G* level of theory, in Table 1.

Frequencies of the other conformers are available upon request.

The NH stretch modes have near-zero intensity, and cannot be expected to be observed experimentally.

One of the catechol OH groups is hydrogen bonded to its neighbour, resulting in a red-shifted stretch mode.

The “free” catechol OH stretch band, (OH)free, cat, and the stretch band of the hydrogen-bonded catechol OH, (OH–O)cat, occur at nearly identical frequencies/intensities in AG1a and GG1a, and therefore cannot be used to distinguish between the two conformers.

Unfortunately, also the frequency of the hydrogen-bonded side-chain OH stretch mode, OH–N, is very similar in AG1a and GG1a, though the AG1a mode has a larger intensity.

The PA OH–N mode, on the other hand, is sufficiently different in AG1a and GG1a to aid assignment of future observed conformers.

Fig. 6 compares the relative energies Δ(E0) of the corresponding NA, A and PA conformers obtained from single-point MP2 calculations.

The adrenaline AG2 conformers are much less stable than the corresponding NA and PA conformers.

This is because the A conformer has a CH⋯O instead of an NH⋯O interaction.

Similarly, the pseudoadrenaline AG3 conformers are destabilised by the replacement of an intramolecular NH⋯O by a less favourable CH⋯O interaction.

For the GG2 family both A and PA are destabilised with respect to the corresponding NA conformers, as the NH⋯π interaction is replaced by a CH⋯π contact in A, while in PA the NH⋯O interaction is replaced by a CH⋯O contact.


We have computed ten different AG1c–H2O and ten different AG3b–H2O structures.

AG1c belongs to the most stable conformer family, AG1, whereas AG3b belongs to the first family of conformers with an intramolecular NH⋯O hydrogen bond.

Table 2 shows the computed interaction energies of the AG1c–H2O hydrates.

Their structures, located with B3LYP/6-31+G*, are shown in Fig. 7.

Previous studies on the hydration of ethanolamine neurotransmitters and their analogues found two different categories of 1∶1 hydrates: insertion structures, in which the water inserts between two functional groups of the host molecule, and addition structures, formed by the addition of a water molecule to one of the functional groups.

The global minimum is an insertion structure for 1∶1 clusters of 2-aminoethanol with phenol,36 and for the 1∶1 hydrates of 2-phenoxy-ethanol,37 APE,30 and the ephedrine/pseudoephedrine pair.3

However, the most stable 1∶1 hydrate of NA (which has a catechol ring instead of the benzene ring in 2-phenoxy- ethanol, APE and the ephedra), was found to be an addition structure; the water molecule binds to one of the catechol OH groups,2 indicating that the hydroxyl groups of the catechol ring provide a more attractive water-binding site than the ethanolamine side chain.

In the current study we find that this addition structure is also the most stable 1∶1 hydrate of the AG1c adrenaline conformer (see Table 2 and Fig. 7): in this structure (labelled catOH–Ow), the water binds to one of the catechol OH groups to form an OH⋯OH⋯OwHw daisy chain of hydrogen bonds.

A related structure (catO–HwOw) with a similar hydrogen-bond chain but with water acting as the hydrogen bond donor is distinctly less stable.

There are two different insertion-type structures, with water being located either between the chain OH and NH2 groups (OH–OwHw–N) or between the two catechol OH groups (catOH–OwHw–catO).

The two insertion structures have large monomer deformation energies, reflecting the energy penalty associated with opening up the dihedral angle between the two functional groups, to make space for the water molecule.

Despite this energy penalty, the OH–OwHw–N insertion structure is only slightly less stable than the global minimum (catOH–Ow).

The water molecule can further bind to the side-chain NH group (NH–Ow), to either of the lone pairs of the side-chain oxygen (the two O–HwOw structures), and to the π-electron cloud of the catechol ring.

In the first π-bonded structure, OH–OwHw–π, the water forms an additional hydrogen bond with this OH group.

In the second π-bonded structure, OwHw–π, the water binds to the other face to the catechol ring, where there is no possibility to form an additional strong hydrogen-bonded contact (although there is some additional stabilisation due to the two CH⋯Ow interactions), and this structure is therefore less stable than the OH–OwHw–π minimum.

Table 2 shows that the O–HwOw–π structure is not stable on the B3LYP/6-31+G* potential energy surface.

As the effect of BSSE on the intermolecular geometry can be quite large in π-bonded complexes,15,18 we re-optimised the hydrogen-bond distances of the two π-bonded structures with counterpoise corrections.

The resulting interaction energies are given in parentheses in Table 2.

This re-optimisation has negligible effect on the B3LYP results, but it significantly increases the MP2/DZPi interaction energy, thereby further increasing the discrepancy between the MP2 and DFT results.

Fig. 7 shows that the re-optimisation significantly increases the MP2 hydrogen-bond distances.

Fig. 7 shows two nearly iso-energetic catO–HwOw structures, just differing in the position of the free (not hydrogen-bonded) water hydrogen, which can be on either side of the aromatic ring.

To investigate the barrier for interconversion between the (a) and (b) structure, we have calculated the relaxed potential energy profile for rotation of the free water hydrogen around the Ocat–Ow axis.

As can be seen in Fig. 8, the catO–HwOw (a) minimum is very shallow, with an almost non-existent barrier for conversion to the (b) structure, and it is therefore surprising that the (a) structure was found at all!

Also shown in Fig. 8 is the counterpoise-corrected potential energy profile.

The counterpoise-corrected barrier, though larger than the uncorrected barrier, is still only 0.8 kJ mol−1.

It can therefore be expected that the two minima will be indistinguishable even at the very low vibrational temperatures of supersonic jet experiments.

Indeed, plots of the H-density derived from diffusion Monte Carlo studies on hydrogen-bonded complexes like uracil–(H2O)n (n = 1–3)21 and cytosine–(H2O)238 show a large-amplitude, banana-shaped, zero-point vibrational motion of the free water hydrogen atoms, covering the corresponding minima differing only in the position of the free water hydrogen (above or below the plane of the biomolecule).

This delocalised zero-point motion indicates that even at 0 K these minima cannot be differentiated experimentally.

Table 3 shows the computed interaction energies of the 1∶1 AG3b–H2O hydrates.

Their structures, located with B3LYP/6-31+G*, are displayed in Fig. 9.

The AG3b hydrates in which water binds to the catechol OH groups are very similar in structure and interaction energy to their corresponding AG1c hydrates.

However, because AG3b has an NH⋯O hydrogen bond instead of the OH⋯N bond in AG1c, hydrates in which water binds to the side chain can be radically different.

Thus, the insertion structure in which the water molecule is located between the side-chain OH and NH2 functional groups contains an OH⋯OwHw⋯N hydrogen-bond chain in AG1c–H2O, but an O⋯HwOw⋯NH hydrogen-bond chain in AG3b–H2O resulting in a 5 kJ mol−1 weaker interaction.

The N⋯HOw interaction in the AG1c NH–Ow hydrate, on the other hand, is replaced by a stronger N⋯HwOw hydrogen bond in AG3b–H2O.

The specific water interaction energy (32 kJ mol−1) shows that this is the most favourable water-binding site.

The N–HwOw AG3b–H2O structure is of comparable stability to the most stable AG1c–H2O structure (catOH–Ow), despite the 2.5 kJ mol−1 smaller stability of the AG3b conformer.

We re-optimised the hydrogen-bond lengths with counterpoise corrections for a selection of the AG3b–H2O minima (see Table 3 for the resulting interaction energies and Fig. 9 for the hydrogen-bond distances).

The effect is very small for the B3LYP calculations, but the MP2/DZPi interaction energies are increased by 2.2–2.4 kJ mol−1.

Unexpectedly, the effect is of similar magnitude for the π-bonded and conventional hydrogen-bonded structures.

The catOH–Ow hydrates of AG1a and GG1a, the two most stable adrenaline conformers, are the most likely candidates for experimental observation, since the catOH–Ow hydrate was found to be the observed NA–H2O structure.2

Table 4 lists the calculated infrared frequencies of these two structures.

(Infrared frequencies of the other hydrates are available upon request).

Comparison of the frequencies of the bare molecule (Table 1) with those of the hydrates (Table 4) reveals a strong red shift of the (OH)cat frequency, accompanied by a large increase in intensity.

This is due to hydrogen bonding with the water molecule, as was also observed for NA–H2O.2

Finally, we would like to comment on the calculation of the interaction energies.

As explained in the Methodology section, the interaction energies should be computed relative to the most stable conformer, more precisely, the most stable conformer for the method and basis set used.

For B3LYP/6-31+G* and single-point MP2 calculations with the 6-31+G* or aug-cc-pVDZ basis sets this is the AG1a conformer, but for MP2/DZPi this is the GG1a conformer.

Thus, the B3LYP/6-31+G* interaction energy of a particular AG1a–H2O hydrate will be the same as its specific water interaction energy, but the corresponding MP2/DZPi interaction energy will contain an energy penalty term reflecting the difference in stability of the AG1a and GG1a conformers.

In Table 5 we have listed the specific interaction energies, and De, for the catOH–Ow hydrates of AG1a and GG1a.

As can be seen, B3LYP/6-31+G* favours the AG1a hydrate, whereas MP2/DZPi favours the GG1a hydrate.

Results from model potential calculations


To understand the effect of water on the geometry of the biomolecule, one has to study clusters containing numerous water molecules.

However, there are several major problems with geometry optimisations of molecular clusters consisting of more than two constituents.

Firstly, the number of local minima increases steeply with the number of constituents in the cluster, and secondly, the convergence of the calculations slows down due to the coupling of inter- and intramolecular degrees of freedom.39

In addition, the individual energy calculations become much more computationally demanding.

Thus, to find all possible minima for the A–(H2O)n clusters of several low-lying adrenaline conformers using methods like DFT or MP2 becomes a near-impossible task.

We therefore need to find a computationally cheap, but reliable, method for exploring the potential energy surface of the A–(H2O)n (n > 1) clusters.

The semi-empirical AM1 and PM3 methods are not suitable for this, as these methods failed to accurately predict the structure and relative stabilities of the different serotonin conformers.40

In this section, we explore the suitability of force-field calculations using the MP2fit/DMA21 rigid-body model potential for scanning the potential energy surface of the hydrates.

The DMA electrostatic model of this potential accurately describes non-spherical features of the charge distribution, such as π-electron clouds and lone pairs, which play an important role in establishing the water-binding sites in the biomolecule.

The model potential yields the interaction between two rigid monomer fragments, and thus, the MP2fit/DMA results should be compared to the B3LYP and MP2 specific water interaction energies.

We will first investigate the performance of the model potential for locating the 1∶1 hydrates of the AG1c and AG3b conformers.

The resulting interaction energies are included in Tables 2 and 3.

The results are very encouraging.

All minima are found (we failed to locate one of the two O–HwOw AG1c hydrates, but as mentioned above, structures that only differ in the free Hw orientation are basically the same structure).

The model potential generally yields interaction energies that are 2–4 kJ mol−1 larger than the corresponding B3LYP values, except for the π-bonded structures and the AG3c N–HwOw minimum, for which the model potential yields much larger interaction energies.

The discrepancy between the DFT and MP2fit/DMA results for the π-bonded structures is due to the inability of DFT to properly describe these structures, rather than deficiencies in the model potential.

This is confirmed by the good agreement between the MP2fit/DMA and MP2/DZPi results for the π-bonded structures.

The discrepancy for the N-HwOw structure indicates a possible deficiency in the potential's parameters for N. The repulsion and dispersion parameters of O and H were adjusted to yield good agreement with MP2 data on the minima and transition states of uracil-water,25 but the N parameters were not attuned to ab initio data (they were derived from empirical fits to organic crystal structures and heats of sublimation41–43), and thus, the N parameters may not be optimal for reproducing the gas-phase structures.

Of particular interest are the insertion-type structures.

Because the MP2fit/DMA is a rigid-body potential, the adrenaline molecule cannot adjust its geometry to facilitate the water molecule to migrate between two of its functional groups.

Both insertion structures (OH–OwHw–N/O–HwOw–HN and catOH–OwHw–catO) were found with the model potential calculations, demonstrating that the rigidity of the model potential does not prevent these structures to be found.


Earlier work on uracil-(H2O)2 and uracil-(H2O)3 showed that the MP2fit/DMA potential yields good agreement with ab initio (MP2/DZPi) interaction energies and hydrogen-bond distances, despite the lack of many-body terms in the potential model.21

As the three-body contribution is about 25% for uracil-(H2O)2, this indicates that the model potential's repulsion and dispersion parameters, which have been determined by empirical parameterisation, must have absorbed some of the water-water non-additivity.44

In this section, we investigate the model potential suitability for A–(H2O)2 clusters.

We created AG3b–(H2O)2 starting structures by systematically varying the position of the two water molecules in steps of 2 Å, in three orthogonal directions.

Two different rotational orientations of the water molecules were considered.

Structures in which one of the moieties overlapped with another one (as indicated by an intermolecular distance smaller than 1.2 Å) were discarded.

Structures in which a water molecule was too far from the adrenaline molecule (closest intermolecular contact larger than 4 Å) were discarded as well.

The geometries of the resulting 33 540 geometries were optimised with the MP2fit/DMA model potential.

One geometry optimisation took approximately 1 minute of CPU time on a 1.7 GHz Pentium 4 PC.

We found as many as 170 different minima; however, some of these are very similar in structure.

Fig. 10 lists the twelve most stable structures found with MP2fit/DMA, in decreasing order of stability.

Structures 1, 4 and 11 are bridge structures: the two water molecules bridge the adrenaline chain oxygen and nitrogen functional groups.

In all these, the adrenaline chain oxygen acts as a hydrogen-bond acceptor.

Zwier suggested that such water bridges may be a preferred structural motif in cases where it can be formed.45

Structures 1 and 4 are very similar, with just small differences in the position of the free water hydrogens.

Structure 11 has a different hydrogen-bonding pattern than 1 and 4 (O⋯HO⋯HOH⋯N instead of O⋯HOH⋯OH⋯N).

In structures 2, 3, 5 and 6 the water molecules form a dimer structure, which is bonded to the adrenaline nitrogen.

We call these dimer addition structures.

The different dimer addition structures just differ in the position of the free water hydrogens, or in the direction of the hydrogen-bond chain itself.

The remaining five structures are double addition structures, formed by the addition of the two water molecules to two different functional groups of adrenaline: In structures 7 and 10 one water molecule binds to the nitrogen, while the other binds to the chain oxygen, whereas in structures 8 and 9 the second water binds to one of the catechol hydroxyl groups.

Structure 12 is a mixed insertion/addition structure: the first water molecule is inserted between the chain O and the NH group, whereas the second water molecule is added to the nitrogen.

We have reoptimised the structures of these twelve hydrates, and of two lower-lying hydrates (23 and 28), with B3LYP/6-31+G*.

In structure 23 the two waters form a ring structure involving one of the catechol hydroxyls.

Structure 28 contains a water dimer bridging the chain OH and π-electron binding sites.

Structure 23 obeys the basic structural principle of XW2 clusters in which X corresponds to an aromatic alcohol (catechol, in this case), which are equivalent to a water trimer in which one of the free hydrogens is replaced by the aromatic ring.39

Table 6 compares the MP2fit/DMA and B3LYP specific water interaction energies.

Most B3LYP-optimised structures are very similar to their corresponding MP2fit/DMA structures, but there are some important differences.

AW2-8 and AW2-9 converge to the same minimum when reoptimised with B3LYP.

Two other structures (5 and 7) are not stable on the B3LYP potential energy surface.

Most worryingly though, we have found an additional structure with B3LYP (labelled AW2-global, see Fig. 10).

In this structure, the two waters bridge the chain OH and NH functional groups.

The adrenaline chain oxygen acts as a proton donor (in contrast to the other bridge structures, 1, 4 and 11, in which the oxygen acts as an acceptor).

To allow this particular water-binding arrangement, the OH group has to rotate by as much as 60° around the Cβ–O bond, changing the hydroxyl hydrogen from a staggered to an eclipsed position with respect to the Cβ hydrogen.

This internal rotation is not possible with the rigid-body MP2fit/DMA potential.

Despite this large conformational distortion of adrenaline, costing 10 kJ mol−1, the AW2-global structure is the most stable AG3b–(H2O)2 hydrate located in this study.

In general, B3LYP gives smaller AG3b–(H2O)2 interaction energies than MP2fit/DMA.

Though it is possible that DFT underestimates the AG3b–(H2O)2 interaction energies, it is unlikely that they are underestimated by as much as 20 kJ mol−1.

Thus, it is more likely that MP2fit/DMA overestimates the interaction energies, even though the potential did not overestimate uracil-(H2O)2 interaction energies.46

As the model potential proved unsuitable to locate the A–(H2O)2 structures, further investigation of this problem has not been pursued.


Even though environmental effects play an important role in modelling biological processes, an understanding of the intrinsic energetics of flexible biomolecules is essential for disentangling the relative importance of the intrinsic and environmental effects.

The most stable adrenaline/pseudoadrenaline conformers, in the gas phase, are those forming intramolecular hydrogen bonds: both the extended AG1 and the folded GG1 conformers contain an intramolecular OH⋯N hydrogen bond.

Naturally, under physiological conditions, adrenaline will not adopt its gas-phase geometry.

The structure of adrenaline in blood will be affected by environmental factors such as protonation, solvation and interaction with other molecules present, whereas the structure of the neurotransmitter in its receptor binding site will be influenced by competing hydrogen-bonding interactions with binding-site residues.

However, by studying the shape of biomolecules in the gas phase, the environmental effects can be eliminated, allowing the molecule-solvent interaction to be studied in a controlled environment.

A recent computational study provided a prediction for the structure and function of G protein-coupled receptors, including the β1-adrenergic receptor.47

The study shows the two catechol OH groups forming hydrogen bonds with serine residues, whereas the chain hydroxyl and the amino group form hydrogen bonds with an aspartic acid, in good agreement with earlier mutation experiments.48–51

The study predicts that the catechol hydrogens point away from each other, an unfavourable orientation not considered in the current work.

Calculations show that this catechol group orientation is about 19 kJ mol−1 higher in energy than corresponding conformers in which the catechol OH groups form an intramolecular hydrogen bond.4,52

Thus, environmental factors appear to have a profound effect on the structure of the neurotransmitter.

Our initial investigation of the adrenaline hydrates in this paper provides a first step in studying the environmental effects on the structure of adrenaline.

Hydration is believed to play an important role in the drug-receptor recognition process.

It has been proposed that the water surrounding the neurotransmitter and its receptor invokes the initial recognition process, possibly by a change in the receptor geometry induced by the structure-breaking effects of the catechol ring.53

In this study we have studied the 1∶1 hydrates of the AG1c and the 1∶1 and 1∶2 hydrates of the AG3b conformer.

For a complete study of adrenaline hydration, one would have to include several of the low-lying conformers, as the interaction with water may change their relative stability, and one would have to look at larger clusters (containing more water molecules) as well.

This is a formidable task, and would best be done in combination with an experimental study, which can provide guidance to the type of hydrates observed experimentally.

The current study shows the limitations of using a rigid-body potential for calculating the adrenaline hydrates.

While the model potential did find all relevant 1∶1 hydrates, including the insertion-type structures, it failed to locate the most stable 1∶2 hydrate, characterised by a large internal rotation around the Cβ–O bond.

A similar effect was observed for protonated nicotine-(H2O)2,22 Also for this system, the interaction with water induces a large conformational distortion, costing 12 kJ mol−1, which allows the two water molecules to bridge the pyridine and pyrrolidinium rings.

Good agreement with the DFT results was obtained only when the distorted nicotine geometry (and its corresponding DMA) was used in the MP2fit/DMA model potential calculations.22

Thus, the model potential calculations can only be used as a guiding tool to find the most stable hydrates, as its rigid-body character prevents any water-induced conformational changes.

The study of the hydrates of flexible biomolecules remains a difficult and computationally expensive task.


In this study, we have investigated the conformational structures of adrenaline (A) and pseudoadrenaline (PA) at the MP2 and DFT levels of theory.

The calculations predict the extended AG1a and folded GG1a adrenaline conformers to be nearly isoenergetic.

The pseudoadrenaline AG1a conformer on the other hand is more stable than GG1a by 3 kJ mol−1.

It would be very interesting to see which A and PA conformers are formed in supersonic jet experiments.

We have explored the molecular structures of the 1∶1 hydrates of the AG1c and the 1∶1 and 1∶2 hydrates of the AG3b adrenaline conformer.

In the most stable AG1c–H2O structure, the water molecule is bound as a proton acceptor to the p-OH catechol group, in a similar way as the most stable NA–H2O structure observed and predicted by theory.2

In the most stable AG3b–H2O structure the water molecule is bound as a proton donor to the chain nitrogen.

The MP2fit/DMA rigid-body model potential, which contains an accurate DMA model to describe the electrostatic energy contribution to the potential energy, correctly predicts all AG1c–H2O and AG3b–H2O structures.

However, the model potential fails to locate the most stable AG3b–(H2O)2 structure located with B3LYP.

This structure is characterised by a large conformational distortion of the adrenaline moiety, which cannot be described by the rigid-body model potential.

This shows the limitation of rigid-body calculations in studies of biomolecular clusters.

To aid future spectroscopic assignment, we have provided infrared frequencies of the OH and NH stretch modes of the two most stable adrenaline conformers, AG1a and GG1a, and their most plausible singly hydrated clusters.