**2.2.3 Hydrogen bonds**

Hydrogen bonds are non-covalent, attractive interactions between a hydrogen covalently bonded to some electronegative group ("donor"), and another electronegative atom, such as oxygen or nitrogen ("acceptor"). The hydrogen bond can be described as an electrostatic dipole-dipole interaction. However, it also has some features of covalent bonding: it is specific, directional, it produces interatomic distances shorter than sum of van der Waals radii, and usually it involves a limited number of interaction partners, which can be interpreted as a type of valence.

Proteins contain ample hydrogen bond donors and acceptors both in their backbone and in the side chains. The environment (aqueous solvent, protein-protein network, lipid bilayers) in which proteins of interest are immersed also contains numerous proton donors and acceptors – be it water molecule, interacting proteins, lipid headgroups, or DNA/RNA. Hydrogen bonding, therefore, occurs not only between ligand and protein and within the protein itself, but also within the surrounding medium.

Like all non-covalent interactions, hydrogen bonds are fairly weak: in biological conditions, the strength of hydrogen bonds varies between 5-30 kJ/mol (outside of biological systems, the strength of hydrogen bonds may vary from 2 kJ/mol to even 155 kJ/mol for HF2-) (Emsley, 1980), which is weaker than ionic or covalent bonds. However, because of their relative weakness, they can be formed and broken rapidly during binding event, conformational changes, or protein folding. Thus, hydrogen bonds in biological systems may be switched on or off with energies that are within the range of thermal fluctuations. This is one of the prime factors that facilitates macromolecular association events, and biological activity. Another key factor is related to the strict geometric rules, followed by hydrogen bonds in biological systems. Namely, their orientations, lengths, and angular preferences, which make hydrogen bonding very specific. Due to these properties, the role of hydrogen bonds in governing specific interactions in biological recognition processes is absolutely crucial. Hydrogen bonds, both intra-and inter-molecular, are partly responsible for the secondary, tertiary, and quaternary structures of proteins, nucleic acids, and also some synthetic polymers. They play a pivotal role in molecular recognition events, and they tune the properties of the macromolecular system (e.g. mechanical strength, binding specificity). These geometric rules were among the first to be extracted from crystal structure databases (Bissantz *et al.*, 2010). While the preferred geometries of hydrogen bonds are easily defined, their contributions to binding free energy are system-specific (Davis and Teague, 1999, Williams and Ladbury, 2003). Hydrogen bonds always convey specificity to a recognition process but do not always add much binding free energy (Bissantz *et al.*, 2010).

Hydrogen bonds can vary quite considerably in their strength. Often, a stronger hydrogen bond implies higher penalty of desolvation, so the net free energy gain of a stronger hydrogen bond might be seriously compromised. However, such a picture is not always the case. Hydrogen bond strength, in the context of the free energy changes, should be carefully examined, as it is likely to vary considerably from one ligand-protein system to another one (Barratt *et al.*, 2005, 2006).

Regarding weak hydrogen bonds, the most prominent donor is the CH group. These interactions, despite of their weakness, play an important role in stabilising appropriate conformations of ligand-protein complexes, for instance among the complexes between protein kinases and their inhibitors (Bissantz *et al.*, 2010). Protonated histidines can also act as strong CH donors (Chakrabarti and Bhattacharyya, 2007). Weak hydrogen bonds, their nature, and their role in ligand-protein interactions have been extensively reviewed by Panigrahi and Desiraju (2007).

## **2.2.4 Halogen bonds and multipolar interactions**

6 Thermodynamics – Interaction Studies – Solids, Liquids and Gases

(Donchev, 2006) indicate that multi-body effects amount to only 5% of the total energy (Finkelstein, 2007). However, Finkelstein (2010) shows that those largely ignored multi-atom Van der Waals interactions may lead to significant changes in free energy in the presence of covalent bonds. Those changes can be comparable to those caused by the substitutions of one atom by another one in conventional pairwise Van der Waals interactions. Thus, the

Hydrogen bonds are non-covalent, attractive interactions between a hydrogen covalently bonded to some electronegative group ("donor"), and another electronegative atom, such as oxygen or nitrogen ("acceptor"). The hydrogen bond can be described as an electrostatic dipole-dipole interaction. However, it also has some features of covalent bonding: it is specific, directional, it produces interatomic distances shorter than sum of van der Waals radii, and usually it involves a limited number of interaction partners, which can be

Proteins contain ample hydrogen bond donors and acceptors both in their backbone and in the side chains. The environment (aqueous solvent, protein-protein network, lipid bilayers) in which proteins of interest are immersed also contains numerous proton donors and acceptors – be it water molecule, interacting proteins, lipid headgroups, or DNA/RNA. Hydrogen bonding, therefore, occurs not only between ligand and protein and within the

Like all non-covalent interactions, hydrogen bonds are fairly weak: in biological conditions, the strength of hydrogen bonds varies between 5-30 kJ/mol (outside of biological systems, the strength of hydrogen bonds may vary from 2 kJ/mol to even 155 kJ/mol for HF2-) (Emsley, 1980), which is weaker than ionic or covalent bonds. However, because of their relative weakness, they can be formed and broken rapidly during binding event, conformational changes, or protein folding. Thus, hydrogen bonds in biological systems may be switched on or off with energies that are within the range of thermal fluctuations. This is one of the prime factors that facilitates macromolecular association events, and biological activity. Another key factor is related to the strict geometric rules, followed by hydrogen bonds in biological systems. Namely, their orientations, lengths, and angular preferences, which make hydrogen bonding very specific. Due to these properties, the role of hydrogen bonds in governing specific interactions in biological recognition processes is absolutely crucial. Hydrogen bonds, both intra-and inter-molecular, are partly responsible for the secondary, tertiary, and quaternary structures of proteins, nucleic acids, and also some synthetic polymers. They play a pivotal role in molecular recognition events, and they tune the properties of the macromolecular system (e.g. mechanical strength, binding specificity). These geometric rules were among the first to be extracted from crystal structure databases (Bissantz *et al.*, 2010). While the preferred geometries of hydrogen bonds are easily defined, their contributions to binding free energy are system-specific (Davis and Teague, 1999, Williams and Ladbury, 2003). Hydrogen bonds always convey specificity to a recognition process but do not always add much binding free energy

Hydrogen bonds can vary quite considerably in their strength. Often, a stronger hydrogen bond implies higher penalty of desolvation, so the net free energy gain of a stronger hydrogen bond might be seriously compromised. However, such a picture is not always the case. Hydrogen bond strength, in the context of the free energy changes, should be carefully

currently used force fields (applied in MD simulations) need to be revised.

**2.2.3 Hydrogen bonds** 

(Bissantz *et al.*, 2010).

interpreted as a type of valence.

protein itself, but also within the surrounding medium.

The concept of halogen bonds is similar to hydrogen bonds: both types of interactions involve relationships between an electron donor and electron acceptor. In hydrogen bonding, a hydrogen atom acts as the electron acceptor and forms a non-covalent bond by accepting electron density from an electronegative atom ("donor"). In halogen bonding, a halogen atom is the donor.

Despite of their prevalence in complexes between proteins and small organic inhibitors (many of them contain halogen atoms due to solubility and bioavailability) and their importance for medicinal chemistry, the significance of halogen bonds in biological context has been overlooked for a long time (Zhou *et al.*, 2010). For a number of years, halogen atoms were regarded as hydrophobic appendages, convenient – from the molecular design point of view - to fill apolar protein cavities. The nature of halogen interactions (such as directionality, sigma-holes) was not studied in detail and not regarded as very important. Indeed, halogen bonds are, in general, fairly weak interactions. On the other hand, in some cases they can compete with hydrogen bonds, thus should be considered in more details, given the importance of hydrogen bonds in ligand-protein interactions and given that many of synthesised small organic compounds contain halogen bonds in their structure (Bissantz *et al.*, 2010, Zhou *et al.*, 2010).

Halogens involved in halogen bonds are chlorine, bromine, iodine, and fluorine (not very often). All four halogens are capable of acting as donors (as proven by computational and experimental data) and follow the general trend: F < Cl < Br < I, with iodine normally forming the strongest bonds, as the strength increases with the size of the halogen atom. From the chemical point of view, the halogens, with the exception of fluorine, have unique electronic properties when bound to aryl or electron withdrawing alkyl groups. They show an anisotropy of electron density distribution with a positive area (so-called -hole) of electrostatic potential opposite the carbon-halogen bond (Clark *et al.*, 2007). The molecular origin of the -hole can be explained quantum chemically and the detailed description is provided in the work by Clark and coworkers (2007). Briefly, a patch of negative charge is formed around the central region of the bond between carbon and halogen atom, leaving the outermost region positive (hence the "hole").

Available experimental data show the strong influence of halogen bonds on binding affinity. Replacement of hydrogen by halogen atom is often used by medicinal chemists in order to increase the affinity. Indeed, in a series of adenosine kinase inhibitors, a 200-fold affinity gain from hydrogen to iodine has been observed (Iltzsch *et al.*, 1995). Another spectacular, 300-fold affinity difference upon iodine substitution was observed in a series of HIV reversetranscriptase inhibitors (Benjahad *et al.*, 2003). Unsurprisingly, substitution of hydrogen by iodine typically leads to the largest affinity gain, since the strength of the halogen bond increases with the size of halogen atom.

An introduction of heteroatoms into aromatic ring affects the ratio of both geometries. The preference to perpendicular interactions increases when the acidity of the interacting "side" atoms increases; this happens upon introduction of a strongly electron-withdrawing substituent in either ortho- or para-position. This was demonstrated by high-accuracy quantum chemical calculations by Sinnokrot and Sherrill (2004): The interaction between benzene as a donor and fluorobenzene as the acceptor, while both compounds were perpendicular to each other, was ~0.3 kcal/mol weaker than that of the benzene dimer. With reverse of roles (fluorobenzene as the donor), the interaction became ~0.6 kcal/mol

For perpendicularly-oriented aromatic-aromatic interactions, studies on several model systems showed that aliphatic-aromatic interactions in the same orientation provide a favourable contribution to the free energy of the same magnitude as aromatic-aromatic interactions (Turk and Smithrud, 2001). For aliphatic-aromatic interactions, interactions energy becomes more favourable when acidity of the interacting CH unit of aliphatic counterpart increases. Study conducted by Tsuzuki *et al*. (2000) showed that ethane (sp3 hybridisation of carbon atom, less acidic) is a worse binder of benzene than acetylene (sp hybridisation of carbon atom, more acidic), and the difference in dissociation energies between acetylene-benzene and ethane-benzene complexes is around 1 kcal/mol. In ligandprotein complexes, this type of interaction can be found in interactions between aromatic side chains and methyl groups. The strength of such interactions depends on the group to which the interacting methyl group is bound: the more electronegative the group, the more the preference towards perpendicular geometry of interacting methyl-aromatic side chain is

protein backbone (namely, their pi faces) and ion pairs - interactions between acidic (Asp,

 

favourable interactions between heavier halogens and aromatic rings has been studied, in particular in the context of halogen bonds. C-H - halogen interactions can be regarded as

Any binding event displaces water molecules from the interaction interface or from the binding pocket, while simultaneously desolvating the ligand (or a part of it). Although most of those waters are disordered and loosely associated with protein structure, such displacement affects the whole solvation shell around the ligand-protein complex

While the vast majority of those water molecules are mobile and easily displaceable, some are tightly bound to the protein structure. Tightly bound water molecules are often conserved across multiple crystal structures of ligand-protein complexes (Poornima and Dean, 1995c). Often, those water molecules play an important role in tuning the biological activity of the protein, as in the case of many enzymes (Langhorst *et al.,* 1999, Nagendra e*t al.*, 1998, Poornima and Dean, 1995a). Those water molecules may be regarded as part of the protein structure. Ligand-protein interactions are often mediated by water molecules buried in the binding site and forming multiple hydrogen bonds with both binding partners (Poornima and Dean, 1995a-c). In other cases, those bound water molecules are released to

interactions are also displayed by amide bonds of

interactions. Recently, the nature of

 

stronger as compared to the benzene dimer.

pronounced (Bissantz *et al.*, 2010).

(Poornima and Dean, 1995b).

Glu) and basic (Lys, Arg) side chains.

Aromatic interactions are not limited to

"very weak hydrogen bonds" (Desiraju, 2002).

**2.3 Solvent effects, structural waters, and the bulk water** 

Halogen atoms can interact with the oxygen and with the carbon atoms of C=O groups, as well. The former attributes to the halogen bond formation, the latter is a hallmark of socalled orthogonal multipolar interactions. These interactions are formed by two dipolar functional groups, which are in a close distance from each other. Only recently it received attention in the field of medicinal chemistry and ligand-protein interactions (Paulini *et al.*, 2005), even though it has been described for a long time. This interaction is known to contribute to ligand-protein stabilisation (Fischer *et al.*, 2008), and it is particularly important in the context of halogen bonds (Bissantz *et al.*, 2010 and references therein). It is worth bearing in mind that in an orthogonal (perpendicular) orientation of two dipoles, the actual dipole contribution to interaction energy is zero. Thus, higher order electrostatic and dispersion terms must be responsible for this type of interaction. The disappearance of the dipole term may turn a repulsive electrostatic interaction into an attractive one. Because of its high electron density and low polarisability, fluorine's preference for dipolar interactions is more pronounced than for the other halogens (Bissantz *et al.*, 2010). Chlorine and other heavy halogens also form multipolar interactions with carbonyl groups, but they show a tendency for the C-X bond to be parallel rather than orthogonal to the amide plane, a consequence of the -hole (Bissantz *et al.*, 2010).

### **2.2.5 Hydrophobic interactions**

The interactions between ligands and the hydrophobic side chains of proteins contribute significantly to the binding free energy. The hydrophobic residues mutually repel water and other polar groups and results in a net attraction of the non-polar groups of ligand. In addition, apolar and aromatic rings of tryptophan, phenylalanine, and tyrosine participate in "stacking" interactions with aromatic moieties of ligand. Many studies have demonstrated that the hydrophobic interactions, quantified by the amount of hydrophobic surface buried upon ligand binding, is the structural parameter correlating best with binding free energy (Bissantz *et al.*, 2010, Perozzo *et al.*, 2004). It holds well for very diverse sets of ligands (Boehm and Klebe, 1996) as well as for protein-protein interactions (Vallone *et al.*, 1998). It should be emphasised, though, that a considerable part of the affinity gain caused by hydrophobic interactions in hydrophobic binding pockets comes from sub-optimal solvation of the pocket in the unbound (apo) state.

Aromatic interactions, hydrophobic effect, and other solvent effects will be discussed further in the following parts of this chapter.

### **2.2.6 Interactions mediated by aromatic rings**

Aromatic rings deserve special attention in the context of ligand-protein interactions. Interactions between ligands and protein aromatic side chains ( Phe, Trp,and Tyr) are widespread in ligand-protein complexes (Bissantz *et al.*, 2010). The unique steric and electronic properties of these side chains, which give rise to large polarizabilities and quadrupole moments, result in preferred geometries upon interactions.

For interactions between two aromatic systems, two geometries are predominant: one, where two rings are parallel to each other, and the perpendicular, edge-to-face arrangement. High-accuracy *ab initio* CCSD(T) quantum chemical calculations of the dimerisation energy of benzene predict these two geometries to be isoenergetic (Hobza *et al.*, 1996), which agrees with experimental results qualitatively and quantitatively (Grover *et al.,* 1987, Krause *et al.*, 1991).

Halogen atoms can interact with the oxygen and with the carbon atoms of C=O groups, as well. The former attributes to the halogen bond formation, the latter is a hallmark of socalled orthogonal multipolar interactions. These interactions are formed by two dipolar functional groups, which are in a close distance from each other. Only recently it received attention in the field of medicinal chemistry and ligand-protein interactions (Paulini *et al.*, 2005), even though it has been described for a long time. This interaction is known to contribute to ligand-protein stabilisation (Fischer *et al.*, 2008), and it is particularly important in the context of halogen bonds (Bissantz *et al.*, 2010 and references therein). It is worth bearing in mind that in an orthogonal (perpendicular) orientation of two dipoles, the actual dipole contribution to interaction energy is zero. Thus, higher order electrostatic and dispersion terms must be responsible for this type of interaction. The disappearance of the dipole term may turn a repulsive electrostatic interaction into an attractive one. Because of its high electron density and low polarisability, fluorine's preference for dipolar interactions is more pronounced than for the other halogens (Bissantz *et al.*, 2010). Chlorine and other heavy halogens also form multipolar interactions with carbonyl groups, but they show a tendency for the C-X bond to be parallel rather than orthogonal to the amide plane, a

The interactions between ligands and the hydrophobic side chains of proteins contribute significantly to the binding free energy. The hydrophobic residues mutually repel water and other polar groups and results in a net attraction of the non-polar groups of ligand. In addition, apolar and aromatic rings of tryptophan, phenylalanine, and tyrosine participate in "stacking" interactions with aromatic moieties of ligand. Many studies have demonstrated that the hydrophobic interactions, quantified by the amount of hydrophobic surface buried upon ligand binding, is the structural parameter correlating best with binding free energy (Bissantz *et al.*, 2010, Perozzo *et al.*, 2004). It holds well for very diverse sets of ligands (Boehm and Klebe, 1996) as well as for protein-protein interactions (Vallone *et al.*, 1998). It should be emphasised, though, that a considerable part of the affinity gain caused by hydrophobic interactions in hydrophobic binding pockets comes from sub-optimal solvation

Aromatic interactions, hydrophobic effect, and other solvent effects will be discussed further

Aromatic rings deserve special attention in the context of ligand-protein interactions. Interactions between ligands and protein aromatic side chains ( Phe, Trp,and Tyr) are widespread in ligand-protein complexes (Bissantz *et al.*, 2010). The unique steric and electronic properties of these side chains, which give rise to large polarizabilities and

For interactions between two aromatic systems, two geometries are predominant: one, where two rings are parallel to each other, and the perpendicular, edge-to-face arrangement. High-accuracy *ab initio* CCSD(T) quantum chemical calculations of the dimerisation energy of benzene predict these two geometries to be isoenergetic (Hobza *et al.*, 1996), which agrees with experimental results qualitatively and quantitatively (Grover *et al.,* 1987, Krause *et al.*,

quadrupole moments, result in preferred geometries upon interactions.

consequence of the

of the pocket in the unbound (apo) state.

**2.2.6 Interactions mediated by aromatic rings** 

in the following parts of this chapter.

1991).

**2.2.5 Hydrophobic interactions** 


An introduction of heteroatoms into aromatic ring affects the ratio of both geometries. The preference to perpendicular interactions increases when the acidity of the interacting "side" atoms increases; this happens upon introduction of a strongly electron-withdrawing substituent in either ortho- or para-position. This was demonstrated by high-accuracy quantum chemical calculations by Sinnokrot and Sherrill (2004): The interaction between benzene as a donor and fluorobenzene as the acceptor, while both compounds were perpendicular to each other, was ~0.3 kcal/mol weaker than that of the benzene dimer. With reverse of roles (fluorobenzene as the donor), the interaction became ~0.6 kcal/mol stronger as compared to the benzene dimer.

For perpendicularly-oriented aromatic-aromatic interactions, studies on several model systems showed that aliphatic-aromatic interactions in the same orientation provide a favourable contribution to the free energy of the same magnitude as aromatic-aromatic interactions (Turk and Smithrud, 2001). For aliphatic-aromatic interactions, interactions energy becomes more favourable when acidity of the interacting CH unit of aliphatic counterpart increases. Study conducted by Tsuzuki *et al*. (2000) showed that ethane (sp3 hybridisation of carbon atom, less acidic) is a worse binder of benzene than acetylene (sp hybridisation of carbon atom, more acidic), and the difference in dissociation energies between acetylene-benzene and ethane-benzene complexes is around 1 kcal/mol. In ligandprotein complexes, this type of interaction can be found in interactions between aromatic side chains and methyl groups. The strength of such interactions depends on the group to which the interacting methyl group is bound: the more electronegative the group, the more the preference towards perpendicular geometry of interacting methyl-aromatic side chain is pronounced (Bissantz *et al.*, 2010). interactions are also displayed by amide bonds of protein backbone (namely, their pi faces) and ion pairs - interactions between acidic (Asp, Glu) and basic (Lys, Arg) side chains.

Aromatic interactions are not limited to interactions. Recently, the nature of favourable interactions between heavier halogens and aromatic rings has been studied, in particular in the context of halogen bonds. C-H - halogen interactions can be regarded as "very weak hydrogen bonds" (Desiraju, 2002).

### **2.3 Solvent effects, structural waters, and the bulk water**

Any binding event displaces water molecules from the interaction interface or from the binding pocket, while simultaneously desolvating the ligand (or a part of it). Although most of those waters are disordered and loosely associated with protein structure, such displacement affects the whole solvation shell around the ligand-protein complex (Poornima and Dean, 1995b).

While the vast majority of those water molecules are mobile and easily displaceable, some are tightly bound to the protein structure. Tightly bound water molecules are often conserved across multiple crystal structures of ligand-protein complexes (Poornima and Dean, 1995c). Often, those water molecules play an important role in tuning the biological activity of the protein, as in the case of many enzymes (Langhorst *et al.,* 1999, Nagendra e*t al.*, 1998, Poornima and Dean, 1995a). Those water molecules may be regarded as part of the protein structure. Ligand-protein interactions are often mediated by water molecules buried in the binding site and forming multiple hydrogen bonds with both binding partners (Poornima and Dean, 1995a-c). In other cases, those bound water molecules are released to

The concept of the classical hydrophobic effect relies on a hydrophobic solute disrupting the structure of bulk water. This decreases entropy due to ordering of water molecules around the hydrophobic entity. Such unfavourable effects can be minimised if solute molecules aggregate. Upon aggregation, water molecules form one larger "cage" surrounding the hydrophobic aggregate, and the surface area of such aggregate is smaller than the sum of surface areas of individual (non-aggregated) solutes. This makes the entropic contribution less unfavourable and hence makes the free energy more favourable (Homans, 2007).

If this mechanism was the sole driving force for a protein-ligand interaction, all binding events involving hydrophobic ligands would be entropy-driven. This is not the case. Several years ago, in the group of Steve Homans (University of Leeds), we studied the thermodynamics signature of ligand binding by the mouse major urinary protein (MUP). This protein is characterised by a strongly hydrophobic binding pocket and it binds a handful of very different hydrophobic ligands – long-chain alcohols and pyrazine derivatives, among others. Surprisingly, the ITC data showed that the binding was enthalpy-driven (Barratt *et al.*, 2005). This was combined with a negative change in heat

In order to elucidate the molecular origin of this unusual binding signature, we employed computational methods, such as molecular dynamics (MD) simulations. I will discuss the results in more details later in this chapter. The data showed that the key to this favorable enthalpy of binding of ligands to MUP seems to be the sub-optimal solvation of the binding pocket in apo (unbound) state: only a few water molecules remained there prior to ligand binding. The favourable enthalpic component was, thus, largely determined by ligand desolvation, with only a minor contribution from desolvation of the protein. Such complexation thermodynamics driven by enthalpic components have been referred to as the

**2.5 Enthalpy-entropy compensation, binding cooperativity, and protein flexibility**  The enthalpic and entropic contributions are related. An increase in enthalpy by tighter binding may directly affect the entropy by the restriction of mobility of the interacting molecules (Dunitz, 1995). This phenomenon, referred to as enthalpy-entropy compensation, is widely observed, although its relevance is disputed (Ford, 2005). Such compensation, although frequently observed, is not a requirement: if it was, meaning that changes in

In connection to the enthalpy-entropy compensation, ligand-protein interactions can be cooperative, which means the binding energy associated with them is different than the sum of the individual contributions to the binding free energies. Cooperativity provides a medium to transfer information, enhance or attenuate a response to changes in local concentration and regulate the overall signalling/reaction pathway. Its effects are either positive (synergistic) or negative (interfering), depending on whether the binding of the first ligand increases or decreases the affinity for subsequent ligands. Noncooperative (additive) binding does not affect the affinity for remaining ligands and the subsequent binding sites

Cooperativity is often linked to pronounced conformational changes in the structure of the protein. It can be, in some cases, caused by structural tightening through the presence of additional interactions; inter-atomic distances become shorter and interaction becomes

*H*

, optimisation of binding affinities

**2.4 Classical and non-classical hydrophobic effect** 

capacity upon binding - a hallmark of the hydrophobic effect.

were always compensated by opposing changes in *T S*

would not be possible, which is clearly not the case.

"non-classical hydrophobic effect".

can be regarded as independent.

the bulk upon ligand binding. Such displacement may affect the thermodynamic signature of the binding event in a dramatic way. It is generally assumed that the release of a water molecule from a rigid environment should be entropically favorable. The upper limit of the entropy gained for transferring a water molecule from a protein to bulk solvent was estimated to be 2 kcal/mol at room temperature (Dunitz, 1994). This gain would be compensated by loss of enthalpy, so the total contribution to the free energy (as a sum of its enthalpic and entropic terms) of a single water molecule released from the protein to the bulk is difficult to guess. Moreover, in order to reach this 2 kcal/mol limit the water molecule would have to be fixed very rigidly while bound. This is often not the case, and it has been observed in numerous occasions that even very tightly bound, "structural" waters may retain a significant amount of residual mobility (Denisov *et al.*, 1997, Fischer and Verma, 1999, Matthews and Liu, 2009, Smith *et al.*, 2004).

"Structural" water molecules affect their surrounding not only via direct interactions (such as hydrogen-bonding network), but also by influencing the dynamical behaviour of their environment. Numerous cases have been reported when binding of the structural water affected protein flexibility (Fischer and Verma, 1999, Smith *et al.*, 2004). The direction of such influence cannot be predicted by simple rules, as it is heavily dependent on the details of the binding site – some protein become more dynamic upon water binding (Fischer and Verma, 1999), while other ones become more rigid (Mao *et al.*, 2000). Yet ignoring those water effects is likely to lead to substantial errors in the free energy predictions. The importance of the contributions of "structural" water molecules to binding events and its implications for drug design have been emphasised in a study by Michel et al (Michel *et al.*, 2009).

The traditional, enthalpy-dominated view of ligand-protein association largely neglects solvation effects, which strongly affect the thermodynamic profile of a binding event. Recently it became clear that studying the hydration state of a protein binding pocket in the apo (unbound) state should be a routine procedure in rational drug design, as the role of solvation in tuning binding affinity is critical. Solvation costs are a plausible reason why some ligands, despite fitting into a binding site, fail during experimental tests as inhibitors. Young and coworkers showed that an optimised inhibitor of factor Xa turns virtually inactive when the isopropyl group interacting in the S4 pocket of factor Xa is substituted by hydrogen: The compound (PDB code 2J4I) is characterised by Ki of 1 nM. Replacing the isopropyl group by hydrogen reduces its affinity to 39 M. Substitution of this group by hydrogen, apart from reducing the number of favourable hydrophobic interactions, leads to unfavourable solvation of the binding pocket (Young *et al*., 2007, and references therein).

Desolvation of the ligand itself may sometimes control the binding free energy. For highly hydrophilic ligands, the desolvation costs may be very high and make unfavourable contributions to the binding (Daranas *et al.* 2004, MacRaild *et al.*, 2007, Syme *et al.*, 2010). The calorimetric study of -galactose derivatives binding to arabinose binding protein (ABP) showed dramatic differences in binding free energy between several deoxy derivatives (Daranas *et al.*, 2004). The most likely reason of 4-deoxygalactose failing to bind to ABP is the unfavourable desolvation cost (Bronowska and Homans, unpublished data).

Spectroscopic evidence shows that (1) water molecules in the first solvation shell (surrounding the hydrophobic solute) are more flexible that it was originally thought (Finney and Soper, 1994) and (2) hydrogen bonds at hydrophobic surfaces are weaker than it was assumed (Scatena *et al.*, 2001). In addition, the properties of the water molecules from first two solvation shells are very different from these of bulk water, as emerged from terahertz spectroscopy results (Ebbinghaus *et al.*, 2007, Heugen *et al.*, 2006).
