Proton transfer reactions and hydrogen-bond networks in protein environments

Abstract

In protein environments, proton transfer reactions occur along polar or charged residues and isolated water molecules. These species consist of H-bond networks that serve as proton transfer pathways; therefore, thorough understanding of H-bond energetics is essential when investigating proton transfer reactions in protein environments. When the pK_a values (or proton affinity) of the H-bond donor and acceptor moieties are equal, significantly short, symmetric H-bonds can be formed between the two, and proton transfer reactions can occur in an efficient manner. However, such short, symmetric H-bonds are not necessarily stable when they are situated near the protein bulk surface, because the condition of matching pK_a values is opposite to that required for the formation of strong salt bridges, which play a key role in protein–protein interactions. To satisfy the pK_a matching condition and allow for proton transfer reactions, proteins often adjust the pK_a via electron transfer reactions or H-bond pattern changes. In particular, when a symmetric H-bond is formed near the protein bulk surface as a result of one of these phenomena, its instability often results in breakage, leading to large changes in protein conformation.

1. Introduction

Proton transfer reactions play key roles in energy conversion processes in biological systems, including photosynthesis and respiration. In the O₂-evolving complex, photosystem II (PSII), removal of subproduct protons from the catalytic site within the protein (2H₂O → O₂ + 4H⁺ + 4e⁻) is essential for the water-splitting reaction to proceed (see [1] for the most recently reported crystal structure). Proton pump proteins, bacteriorhodopsin for example, are linked to ATP synthase, as a proton gradient is the driving force for ATP synthesis [2–4]. In contrast to the bulk solvent, where proton carriers, for example water molecules, are readily available, in protein environments, the presence of such molecules is strictly limited. In order to overcome this, proteins use a proton transfer pathway comprised ionizable residues, polar residues and water molecules. These polar groups form an H-bond network. Thus, it is essential to understand H-bond energetics before we focus on specific proton transfer pathways in protein environments.

2. Classification of H-bonds and their characteristics

A typical H-bond consists of donor and acceptor moieties, with their characters differentiated by their pK_a values (table 1). As widely observed in ionizable groups, a proton is more likely to populate the moiety with the higher pK_a value out of the two, with this serving as the H-bond donor, and that with the lower pK_a serving as the H-bond acceptor.

Table 1.

Comparison between symmetric H-bonds and asymmetric H-bonds.

	single-well H-bond, LBHB	standard H-bond
potential shape	symmetric	asymmetric
pKa difference	close to zero	present
donor–acceptor length	short	long
coupling	large	small
proton transfer	most efficienta	unfavourable
salt-bridge character	low	high
stability (in protein)	less stableb	stable
	in particular when ‘pKa matching’ is satisfied transiently near the protein bulk surface → often leading to conformational changesc

^aSee [5–7].

^bSee [8,9].

^cE.g. [10,11].

2.1. H-bond length and NMR chemical shift

It has been suggested that a strong H-bond results in a more downfield ¹H nuclear magnetic resonance (NMR) chemical shift. According to the classification of H-bonds by Jeffrey [12] or Frey [13], single-well H-bonds (or symmetrical H-bonds [14]) are typically very short, with O–O distances of 2.4–2.5 Å, and display ¹H NMR chemical shifts (δ_H) of 20–22 ppm [13] (figure 1). Low-barrier H-bonds (LBHBs) are longer at 2.5–2.6 Å, with δ_H values of 17–19 ppm [13], whereas weak H-bonds are even longer, with δ_H values of 10–12 ppm [13].

Figure 1.

Overview of typical potential energy profiles: (a) standard H-bonds (asymmetric double-well), typically with an O_donor–O_acceptor distance greater than 2.6 Å; (b) LBHB, typically with an O_donor–O_acceptor distance of 2.5–2.6 Å and (c) single-well (ionic) H-bonds, typically with an O_donor–O_acceptor distance of less than 2.5 Å [8]. The corresponding O–N distances are generally greater than O–O distances.

2.2. pK_a values

According to Perrin & Nielson [8] or Schutz & Warshel [15], the definition of an LBHB is vague. Schutz & Warshel [15] concluded that LBHBs cannot be defined solely by their length or strength and that only energy-based evaluations can be used to determine the type of H-bond that is present. In particular, the pK_a values of the donor and acceptor moieties are important in determining the energy barrier required for moving an H atom between them [15]. In the original reports by Frey et al. [16] and Cleland & Kreevoy [17], it was stated that an LBHB may form when the pK_a difference between donor and acceptor moieties is nearly zero (matching pK_a).

Interestingly, it is noteworthy that the matching pK_a condition is the situation where proton transfer reactions between the two moieties are most efficient [5–7] (table 1). Under this condition, coupling of the proton donor/acceptor moieties is the greatest, leading to the formation of a short H-bond and a decrease in the energy barrier for the proton transfer.

On the other hand, it is widely recognized that a salt bridge is a strong H-bond and plays a key role in electrostatic interactions at the protein surface. In many cases, salt bridges are formed between acidic and basic residues, with low and high pK_a values, respectively. Thus, the matching pK_a condition required for LBHB formation is exactly opposite to that required for salt-bridge formation (table 1). This implies that LBHBs, which have less salt-bridge-like character (i.e. large and small (nearly zero) pK_a difference for salt-bridge and LBHB, respectively), are not necessarily strong, even if the donor–acceptor distance is short, because concentrated charge, which is more pronounced in salt bridges owing to the large pK_a difference for the H-bond donor and acceptor moieties, is solvated more strongly than a distributed charge, which is more pronounced in LBHBs [9,18]. In particular, when the matching pK_a condition is only transiently satisfied in the protein environment, for example, in intermediate states that are formed by oxidation/reduction or photoisomerization, LBHBs are unstable, where protein preorganized dipoles (e.g. backbone dipoles) are present [9].

2.3. What determines pK_a values in proteins?

2.3.1. Solvation

‘pK_a’ corresponds to proton affinity of the donor/acceptor moieties in H-bonds. In some cases, one might consider that to refer to H-bonds, a proton affinity description rather than a pK_a would be more apt for H-bonds that are not exposed to the solvent. On the other hand, it should also be noted that the solvation energy term is also included in pK_a. pK_a can be defined for any titratable site in proteins, regardless of whether it is practically difficult to measure in experiments (in particular, if it is allowed to define a proton affinity of the same sites). For a single molecule in the bulk solvent, the pK_a value is predominantly determined by the molecular structure and the solvation energy. Thus, the pK_a value can be calculated quantumchemically, by using the atomic coordinates of the molecule and considering the solvation energy [19,20], as is the same for the redox potential for a redox-active site [21]. In a protein environment, although one also has to consider (i) the availability of the solvation and (ii) electrostatic interactions with other groups in addition to its own pK_a value [22], it is also possible to calculate the pK_a value at any titratable site as far as the reasonable atomic coordinates are available.

In general, in the inner hydrophobic core of a protein, the absence of the availability of solvation (rather than repulsive interactions) is the major contributor towards destabilizing the charged groups [23]. In a hydrophobic environment with less chance of solvation for a charged species, the less charged state is energetically more favourable, i.e. protonated state for acidic residue and deprotonated state for basic residues. Thus, the absence of the solvation energy leads to an increase and a decrease in pK_a for acidic and basic groups, respectively (e.g. [22–27]). For example, a catalytic lysine (Lys115) in the active site of the acetoacetate decarboxylase (AADase) has long been known to be deprotonated. Using a reporter group that can approach the catalytic site in the protein, the pK_a value of Lys115 (pK_a(Lys115)) was formerly measured to be 5.96 [28]. More recently, using atomic coordinates of the X-ray crystal structures [26] with consideration of protonation states of all of the titratable sites in the protein, pK_a(Lys115) values were calculated to be 5.73 for the Clostridium acetobutylicum AADase (CaAAD) and 5.96 for the Chromobacterium violaceum AADase [27], significantly low with respect to pK_a(Lys) = 10.4 in water [29]. In general, protein atomic charges can shift the pK_a in the inner core of the protein more significantly than in the bulk water. In CaAAD, the protein atomic charges (including H-bond interaction and long-distance electrostatic interaction) contributed to the increase in pK_a(Lys115) of 3.6 [27]. However, the loss of solvation energy more significantly contributed to the decrease in pK_a(Lys115) of 8.3 [27]. As a consequence, pK_a(Lys115) is considerably low, approximately 6 in AADase [26–28]. The same reason also holds true for the decrease in pK_a(NH₃/NH₄⁺) (i.e. leading to formation of NH₃) at the binding sites along the inner pore of the ammonia transport protein AmtB [24,25].

2.3.2. Protein charges (electrostatic interactions including H-bonds)

pK_a is also affected by electrostatic interaction including H-bonds that originate from the atomic charges of a protein. As donation of a single H-bond typically shifts the redox potential by approximately 60–120 mV, so does the pK_a by approximately 1–2 unit in a protein environment. Functionally important redox/titratable active sites are often H-bonded in the protein environment. Donation of an H-bond to a titratable site leads to stabilization of the deprotonated (reduced) state, resulting in the decrease in the pK_a (increase in the redox potential). Flavodoxin from Clostridium beijerinckii has a flavin mononucleotide (FMN) as a redox-active group. FMN has also the N5 atom as a protonatable site. pK_a(N5) has been measured to be greater than 13.9 [30] or greater than 13 [31] in the native flavodoxin. On the other hand, the experimentally measured value of pK_a(N5) is significantly lowered to 11.3 in the G57T mutant flavodoxin [30]. To understand the difference in pK_a(N5), by solving the linear Poisson–Boltzmann equation with consideration of the protonation states of all titratable sites in the entire flavodoxin, pK_a(N5) has been calculated to be 13.9 for the native flavodoxin and 11.7 for the G57T mutant flavodoxin [32]. The difference in pK_a(N5) has been identified to be the difference in the backbone conformation near the flavin-binding site for the native and G57T mutant proteins [32]. The N5 atom has the backbone carbonyl O atom as an H-bond acceptor in the native crystal structure, whereas the backbone carbonyl O atom is flipped away from the N5 atom in the G57 mutant crystal structure [30]. The incapability of forming an H-bond between N5 and the backbone O atom leads to destabilization of the protonated FMN form, which lowers pK_a(N5) in the G57 mutant by approximately 2 pK_a units [32]. Thus, H-bond patterns determine the pK_a value of the active site in proteins.

pK_a or redox potential is also affected by electrostatic interactions other than H-bonds. In mammals, xanthine oxidoreductase can exist as xanthine dehydrogenase (XDH) and xanthine oxidase (XO). The two enzymes possess common redox-active cofactors, which form an electron transfer pathway terminated by a flavin cofactor. In spite of identical protein primary structures, the redox potential difference between XDH and XO for the flavin is approximately 170 mV, a striking difference [33,34]. The new crystal structures for XDH and XO confirmed the side chain orientation in detail [35]. In particular, the two crystal structures revealed the difference in the protein environment of the FAD-binding site and the H-bond network for XDH and XO. The majority of the redox potential difference between XDH and XO originates from a conformational change in the highly charged loop at positions 423–433 near the flavin-binding site, causing the differences in stability of the semiquinone state. The difference in the redox between XDH and XO is 150 mV when calculated over the region of residues 422–433. The influence of the protein volume (that prevents access to the flavin from water) on the redox potential is equal in XDH and XO [35].

In summary, pK_a values (and redox potential values) in a protein environment can be sufficiently described by the protein geometry as long as the reliable atomic coordinates are available. The calculated value may tend to deviate from the experimentally measured value, in case the geometry of the crystal structure does not represent the functionally relevant conformation. When a number of different conformations are energetically possible, as typically indicated by large disorders of the atomic coordinates (i.e. large B-factor), each conformation has each pK_a value, resulting in large deviation of the pK_a values. Of course, this is often not the case for enzymatic active sites, where the function is realized by specific H-bond patterns and a specific pK_a (or redox potential) value has been experimentally identified (e.g. proteins discussed above).

2.4. Proposed roles of low-barrier H-bonds

The catalytic power of enzymes is owing to the stabilization of the transition state relative to bulk water [18]. Thus, H-bonds in the catalytic site play an important role in the stabilization of the transition state. The LBHB was originally proposed to possess covalent bond-like character, significantly stabilizing the transition state, and facilitating enzymatic reactions [16,17]. In such a bond, the atomic charges of the H-bond donor and acceptor moieties will be more delocalized than those in a conventional H-bond. An advantage that the catalytic site of the protein possesses in contrast to the bulk water is the availability of preorganized dipoles, for example polar or charged side chains, in addition to the protein backbone itself, which can electrostatically stabilize the transition state. If the H-bond is an LBHB (i.e. has low polarity), it will lose the electrostatic advantage owing to its more delocalized atomic charge [15,18].

2.5. δ_H for compounds

δ_H is large for O_donor–H···O_acceptor H-bonds, where the H atom migrates more towards the acceptor moiety. In general, migration of the H atom is more pronounced as the O_donor–O_acceptor distance is reduced [36]. Thus, δ_H is a useful parameter for evaluating the strength of H-bonds.

In order to evaluate the accuracy of the quantumchemically calculated δ_H, we calculated δ_H for maleate and compounds 1–4 (figure 2), which are all thought to contain an LBHB [37]. The calculated δ_H values were found to be close to the experimentally measured values, with discrepancies of only approximately 1 ppm or less (table 2) [38]. Hence, the calculated δ_H values should be considered at this accuracy level. There is a tendency that the discrepancy between the measured values and the calculated values is small for short H-bonds [38]. This predominantly originates from the relevance of the quantumchemically optimized geometries of the molecules used for the δ_H calculations. In H-bonds with long donor–acceptor distances, the two moieties are less strongly coupled, allowing more deviations in the H-bond geometries from those that were quantumchemically optimized.

Figure 2.

Compounds listed in table 2.

Table 2.

Experimental [37] and calculated [38] values of δ_H for compounds in ppm.

entry	exp. δH	calc. δH
maleate	21.5	21.8
1	11.0	12.1
2	12.2	13.1
3	14.9	15.7
4	15.6	15.9

The calculated OHO-bond geometries and the NMR chemical shifts were evaluated using the correlation proposed by Limbach et al. [36]. The geometric correlation of the O_acceptor···H–O_donor bond between the acceptor···hydrogen (O_acceptor···H) distance r₁ and the donor–hydrogen (O_donor–H) distance r₂ was obtained by

2.1

where q_2min represents a minimum value corresponding to the minimum O_acceptor···O_donor distance in the case of a linear H-bond, and r₀ is the equilibrium distance in the fictive free diatomic unit OH [36].

The correlation between the OHO-bond geometry and the ¹H NMR chemical shift was obtained from

2.2

where δ_ΟH⁰ and Δ_H represent the limiting chemical shifts of the separate fictive OH groups and the excess chemical shift of the quasi-symmetric complex, respectively, and m is an empirical parameter. q₂ is taken from equation (2.1).

Interestingly, the obtained H-bond geometries of all compounds fitted exactly with the proposed correlation curve (equation (2.1)) [36] (figure 3), demonstrating that the methodology from [36] is able to reasonably reproduce the quantumchemically optimized O_acceptor···H and O_donor–H distances once the O_acceptor–O_donor distance is specified.

Figure 3.

Calculated δ_H (vertical axis) in ppm and the (O_acceptor···H)/(O_donor–H) difference (horizontal axis) in ångström. r₁ = (O_acceptor···H); r₂ = (O_donor–H); correlation curve (solid curve, equations (2.1) and (2.2) [36]); calculated geometry (open squares) [38]. Vertical dotted lines indicate the (r₁ – r₁)/2 values of O_Glu46–O_pCA and O_Tyr42–O_pCA in the neutron diffraction geometry [39]. Note that only equations (2.1) and (2.2) were used to depict the correlation curve (independent of the calculated geometries).

3. Is a low-barrier H-bond stable or unstable? Short H-bonds in photoactive yellow protein

Photoactive yellow protein (PYP) serves as a bacterial photoreceptor, in particular, as a sensor for negative phototaxis to blue light [40]. The photoactive chromophore of PYP is p-coumaric acid (pCA), which is covalently attached to Cys69 [41]. In the PYP ground state, the pCA chromophore exists as a phenolate anion [42–44], and analysis of its crystal structure has revealed that it is H-bonded with protonated Tyr42 and protonated Glu46 (figure 4). Tyr42 is further H-bonded with Thr50. Structural analysis has also suggested that Glu46 is protonated and pCA is ionized in the PYP ground state, pG [46,47].

Figure 4.

Geometry of the photoactive site in PYP. Only the H atom position of the H-bonds between Tyr42 and pCA, and between Glu46 and pCA are shown (green or cyan spheres). O and N atoms of the side chains are depicted as red and blue balls, respectively. (a) Neutron diffraction analysis [39] (PDB; 2ZOI). (b) QM/MM optimized structure [45] based on the X-ray crystal structure (PDB; 2ZOH). (Online version in colour.)

Recently, neutron diffraction analysis has been used to assign the H (or deuterium) atom positions of PYP [39]. It was found that in the case of the Glu46–pCA pair, an H atom was situated 1.21 Å from Glu46 and 1.37 Å from pCA, almost at the midpoint of the O_Glu46–O_pCA bond (2.57 Å) (figure 4). From this unusual H atom position, the H-bond between Glu46 and pCA was interpreted as an LBHB [17] in [39].

As mentioned earlier, an LBHB forms when the pK_a difference between donor and acceptor moieties is nearly zero. If this is the case, the identification of an LBHB with a single minimum potential can be valid only if the minimum is at the centre of the O_Glu46–O_pCA bond (i.e. the pK_a values of the two moieties are nearly equal) as suggested by Schutz & Warshel [15]. The H atom position in the O_Glu46–O_pCA bond in the neutron diffraction study appears to satisfy the criterion for an LBHB, which should correspond to similar pK_a values for Glu46 and pCA. However, if this was the case, then it would contradict the Glu46 being protonated and the pCA being deprotonated in the PYP ground state, as suggested in a number of previous experimental studies [42–44,48]. In NMR studies, a δ_H of 15.2 ppm was assigned to protonated Glu46 [49]. This is smaller than that for single-well H-bonds (20–22 ppm [13]) or even for LBHB (17–19 ppm [13]). The actual H atom position that corresponds to a δ_H of 15.2 ppm in O_Glu46–O_pCA was not clarified.

3.1. δ_H for photoactive yellow protein

Using the QM/MM optimized geometry, we calculated the δ_H value for the O_Glu46–O_pCA bond and found it to be 14.5 ppm (PDB, 2ZOH [39]) or 14.6 ppm (PDB, 1OTB [50]) [38]; these values differ by 0.6–0.7 ppm from the experimental values of 15.2 ppm [49]. This discrepancy may also reflect the distribution of H-bond lengths, even in these high-resolution (approx. 1 Å) crystal structures of PYP (reviewed in [50]).

We analysed the dependence of δ_H on the H atom position. The origin of the downfield character for the chemical shift is considered to be owing to attenuation of the electronic shielding around the proton owing to the two electronegative donor and acceptor atoms [51]. The maximum δ_H value of approximately 20 ppm was observed near the centre of O_Glu46–O_pCA [38]. The δ_H of approximately 15 ppm that was obtained from solution ¹H NMR studies of PYP [49] could not be obtained near to the centre of O_Glu46–O_pCA, but only at the Glu46 or pCA moieties [38].

On the other hand, the H atom positions calculated from the neutron diffraction study [39] yielded a δ_H value of 19.7 ppm for O_Glu46–O_pCA (figure 3) [38]. This value satisfies the criterion of LBHBs proposed by Frey (δ_H of 17–19 ppm [13]). The fact that an H atom position near the midpoint of O_Glu46–O_pCA resulted in a δ_H of 19.7 ppm for a typical LBHB [13] is also a clear validation of the criterion proposed by Schutz & Warshel [15], i.e. the minimum of the potential energy curve for an LBHB is at the centre of the O_Glu46–O_pCA bond. However, the δ_H of 19.7 ppm is clearly larger than the value of 15.2 ppm obtained from the NMR studies [49]. Hence, the H atom positions obtained in the neutron diffraction study [39] resulted in an overestimation of the chemical shift for O_Glu46–O_pCA, and an underestimation of that for O_Tyr42–O_pCA.

Notably, Steiner [52] and Limbach et al. [36] also proposed the correlation between δ_H and the H-bond geometry; δ_H could be reproduced from (r₁ – r₂)/2 (or alternatively r₁ + r₂), using equations (2.1) and (2.2). The calculated δ_H values were in agreement with those estimated from the proposed correlation curve [36], demonstrating that δ_H can be reproduced if a reasonable H-bond geometry is provided (figure 3). From the δ_H correlation curve, the r₁ of 1.37 Å, and r₂ of 1.21 Å reported for O_Glu46–O_pCA in the neutron diffraction study [39] yielded a δ_H of 20.2 ppm (figure 3) [38]. It should be noted that none of the H-bond geometries investigated (even maleate) possessed an H atom at the centre of the O–O bond (figure 3), as previously reported by Perrin et al. [53,54].

4. Formation of a short H-bond photoinduced by trans–cis isomerization, leading to proton transfer: the photoactive chromophore in photoactive yellow protein

Upon exposure to blue light, PYP undergoes the following photocycle: pG (ground state) → P*–(trans–cis isomerization) → I₀ → I₀^‡ → pR–(proton transfer and large conformational change) → pB → pG [55–57]. The pR to pB transition has been suggested to involve protonation of pCA (i.e. proton transfer) and a large structural change in the protein [55,56]. Although time-resolved Laue diffraction studies proposed structural models of the intermediates [58], the relevance of the suggested pB structure (PDB: 1TS0) as an intermediate of the photocycle is a matter for debate. From the Laue diffraction studies, the pB intermediate has an H-bond between Arg52 and pCA [58]; whereas, solution structures of the pB state indicate a high degree of disorder in residues 42–56 [59] (discussed in [60]).

On the other hand, the time-resolved Laue diffraction studies identified the pR_CW intermediate [58], with a structure proposed to correspond to the pR species [55–57] observed in the spectroscopic studies. The pR state decays to the pB state as a result of proton transfer from Glu46 to pCA, with a rate coefficient of 250 μs [55,56], which is consistent with that of the 333 μs found for the pR_CW decay [58]. To the best of our knowledge, the solution structure of the pR state has not been reported to date.

Interestingly, the O_Glu46–O_pCA bond is unusually short at 2.47 Å in the pR_CW structure (1.60 Å resolution) [58], which may argue against the presence of an LBHB in the ground state that is proposed in [39]. In general, an H-bond donor–acceptor distance can be at its shortest when the pK_a difference between donor and acceptor moieties is nearly zero, owing to maximized coupling between them. This is why LBHB and single-well H-bonds are shorter than standard (asymmetric double-well) H-bonds (figure 1) [8,9,15,54]. If the presence of the shorter O_Glu46–O_pCA bond in the pR_CW state relative to the ground state is plausible, this would suggest that the matching pK_a condition between the H-bond donor and acceptor moieties is not satisfied in the ground state, or is at least less likely than in the pR_CW state.

The QM/MM calculations reproduced the unusually short H-bond distance (2.49 Å) on the basis of the pR_CW structure, only with the assumption that Tyr42 does not donate an H-bond to pCA (figure 5) [10]. The standard O_Glu46–O_pCA H-bond geometry (i.e. Tyr42 donates an H-bond to pCA) yielded a bond length of 2.60 Å, even in QM/MM calculations of the pR_CW structure. These results confirm that in the pR_CW crystal structure, Tyr42 is flipped away from pCA, rather than donating an H-bond to it.

Figure 5.

Changes in the H-bond pattern photoinduced by trans–cis isomerization of pCA, and associated changes in the potential energy profile of the Glu46–pCA H-bond in PYP. (Online version in colour.)

The existence of the unusually short H-bond appears to be plausible not only in the pR_CW structure [58], but also in the pR species [55–57] observed in the spectroscopic studies. FTIR analysis has suggested that the H-bond between Glu46 and pCA becomes stronger in the pR relative to the pG, as suggested by the downshift in the C=O stretching frequency of protonated Glu46 [43]. As the shortening of an H-bond donor–acceptor distance leads to migration of the H atom towards the acceptor moiety (e.g. [38,61]), the observed downshift in the C=O stretching frequency of Glu46 is consistent with the presence of the unusually short O_Glu46–O_pCA bond in the pR_CW structure. The H-bond pattern of Tyr42 and pCA in the O_Glu46–O_pCA length is also evident in studies of the Y42F mutant. The Y42F crystal structure was found to have a shorter O_Glu46–O_pCA bond (2.51 Å [62]) compared with that of the native PYP (2.59 Å [50]), and the C=O stretching frequency of protonated Glu46 in the Y42F mutant was downshifted relative to the wild-type PYP, as seen in the FTIR studies [63].

4.1. How does the short H-bond appear in FTIR studies?

Interestingly, the potential energy curve of the O_Glu46–O_pCA bond (2.49 Å) in the pR_CW crystal structure resembles that of a typical single-well H-bond; the barrierless potential for the proton transfer is an indication of the pR_CW intermediate being ready for the proton transfer (figure 5). As observed in FTIR studies, the C=O stretching frequency for protonated Glu46 is downshifted to 1732 cm⁻¹ in pR relative to 1740 cm⁻¹ in pG, suggesting that the H atom in the O_Glu46–O_pCA bond remains in the Glu46 moiety (i.e. can interact with Glu46), but simultaneously migrated towards the pCA moiety [43], as in the case for a single-well H-bond. Indeed, in FTIR studies, the existence of a single-well H-bond has already been proposed [43]; a stronger H-bond in pR relative to pG lowers the energy barrier for proton transfer from Glu46 to pCA (see fig. 4 in [43]). Note that the O_Glu46–O_pCA H-bond is absent in the pB state [43], and in the solution structures of the pB state [59]. This indicates that the unusually short H-bond is not stable in the protein environment, in particular if the matching pK_a condition for the H-bond donor and acceptor moieties is satisfied only transiently near the protein surface. (For comparison, see also the case of an unusually short H-bond between D1-Tyr161 and D1-His190 in PSII [1,64]; here the matching pK_a condition is satisfied easily in the protein inner core, without tuning the original pK_a values of the donor and acceptor moieties. Discussed later, see ‘Presence of an unusually short, but stable H-bond in redox-active D1-Tyr161 (TyrZ) in photosystem II’.)

4.2. Is the short H-bond in the pR_CW state unusually strong?

If the short O_Glu46–O_pCA H-bond was extremely strong, the pR_CW intermediate would be highly stable and the subsequent pB state would not form on such a timescale. It should also be noted that the lifetime of hundreds of microseconds for the pR_CW state is owing to the large structural change rather than proton transfer from Glu46 to pCA. In addition, in a single-well H-bond, movement of a proton between the donor and acceptor moieties is not directly associated with breakage of the H-bond. Breakage of the short O_Glu46–O_pCA H-bond can occur as a result of the large structural change, which is driven by the photon energy stored in the system [65]. Hence, the pR intermediate can lower the energy required to proceed to the pB state by eliminating the unusually short O_Glu46–O_pCA H-bond of less than 2.5 Å. In terms of the local H-bond network of pCA, formation of the unusually short H-bond is energetically allowed (or favoured) at the stage of the pR intermediate; however, this is not energetically the lowest state of the entire protein, which explains why the pB state is formed.

5. Formation of a short H-bond induced by electron transfer, leading to proton transfer: electron acceptor quinone in photosystem II

The core of the PSII reaction centre is composed of D1/D2, a heterodimer of protein subunits that contains the cofactors that are involved in photochemical charge separation, quinone reduction and water oxidation (figure 6). These reactions are driven by the absorption of light with a wavelength of approximately 680 nm by the P680 pigment. P680 is composed of four chlorophyll a (Chla) molecules, P_D1/P_D2, Chl_D1/Chl_D2, and two pheophytin a molecules (Pheo_D1/Pheo_D2). Excitation of P680 initially leads to the formation of a range of charge-separated states, with the Chl_D1^·+ Pheo_D1^·− state dominating. After a short time, the secondary radical pair, [P_D1/P_D2]^·+ Pheo_D1^·−, is formed in nearly all of the centres. This state is stabilized by electron transfer to the first quinone, Q_A, and by electron donation from a Tyr residue, D1-Tyr161 (TyrZ), to P_D1^·+. TyrZ^·, then oxidizes the Mn₄CaO₅ cluster, which catalyses the subsequent water-splitting reaction. Q_A/Q_A^·− acts as a 1-electron redox couple, accepting electrons from Pheo_D1^·−, and donating to the second quinone, Q_B, without undergoing protonation itself. By contrast, Q_B reduction involves two consecutive 1-electron reduction reactions, with a series of associated proton uptake reactions (reviewed in [66–71]).

Figure 6.

Photoinduced electron transfer (red arrows) and associated proton transfer (blue arrows) reactions in the D1 (red) and D2 (blue) heterodimer protein subunits of PSII. (Online version in colour.)

Q_B is located near the non-haem Fe²⁺ and the ligand to the Fe²⁺, D1-His215, donates an H-bond to the Q_B carbonyl O atom that is nearer to the Fe complex (O_QB(proximal)) (figure 7). The Q_B carbonyl O atom distal to the Fe complex (O_QB(distal)) accepts an H-bond from D1-Ser264, which itself accepts an H-bond from D1-His252 (figure 7), which is located on the protein surface in contact with the aqueous medium [1,70–73]. It is known that Q_B^·− formation is linked to proton uptake [74,75], and comparisons with the structure of the bacterial reaction centre led to the first suggestion that the D1-His252 was the residue undergoing protonation in response to Q_B^·− formation [76]. In theoretical studies, it has been proposed that proton uptake by D1-His252 causes reorientation of the hydroxyl group of D1-Ser264 towards the distal Q_B carbonyl group, and stabilizes Q_B^·−, facilitating the initial electron transfer from Q_A to Q_B [73].

Figure 7.

H-bond partner side chains of Q_A and Q_B in PSII. The backbone amide group of D2-Phe261, which forms an H-bond with Q_A, is omitted for clarity. (Online version in colour.)

5.1. First protonation step: conversion of Q_B⁻ to Q_BH via D1-His215 and D1-Ser264

In order to elucidate how the conversion of Q_B^·− to Q_BH^· occurs, we analysed the potential energy profiles of the two H-bonds of Q_B^·−, O_QB(distal)···H–O_D1-Ser264 and O_QB(proximal)···H–Nδ_D1-His215. In general, Ser is unlikely to deprotonate; however, in the Q_B^·− state, the potential energy profile indicates that transfer of a proton from D1-Ser264 to O_QB(distal) occurs very easily in an energetically downhill process (figure 8) [11]. This reaction is accompanied by a concerted proton transfer from protonated D1-His252 to D1-Ser264, resulting in the formation of Q_B(distal)H^·, deprotonated (neutral) D1-His252, and reoriented D1-Ser264 (figure 8). The QM/MM-optimized geometry indicates that the two H-bonds of D1-Ser264, O_QB(distal)···H–O_D1-Ser264 (2.48 Å) and O_D1-Ser264–H···Nδ_D1-His252 (2.51 Å), are unusually short, especially in the Q_B^·− state [11]. These two short H-bonds were only present before the initial proton transfer occurred but they lengthened (to 2.73 and 2.67 Å, respectively [11]) immediately after proton transfer had occurred. Therefore, the presence of an unusually short H-bond indicates that proton transfer between the donor and acceptor moieties is about to occur.

Figure 8.

Potential energy profiles of the H-bond donor–acceptor pairs in PSII: (right) H-bond between D1-Ser264 and the distal Q_B carbonyl (purple (in the neutral Q_A state) and blue (in the reduced Q_A state) curves); (left) H-bond between D1-His215 and the proximal Q_B carbonyl (red curves). Arrows indicate the directions of proton transfer. (Online version in colour.)

By contrast, the potential energy profile of the O_QB(proximal)···H–Nδ_D1-His215 resembles that of a standard asymmetric double-well H-bond [8], suggesting that the proton transfer from D1-His215 to O_QB(proximal) is an energetically uphill process (figure 8). This is primarily because proton release from the singly protonated (neutral) His (pK_a approx. 14 for imidazole [77]) is unfavourable, unlike the doubly protonated (positively charged) His, for which the pK_a is approximately 7. While the pK_a for neutral His is expected to be lowered to some extent by the positive charge and environment around Fe (see below), it is still likely to be relatively high, and thus unfavourable on this step.

5.2. Second proton transfer and an unusually short H-bond distance between Q_BH⁻ and D1-His215

As O_QB(distal) is protonated upon Q_BH^· formation, the second protonation, i.e. the conversion of Q_BH^·− to Q_BH₂, must occur at O_QB(proximal), which is H-bonded by Nδ of D1-His215 (figure 7). The QM/MM-optimized H-bond distance between O_QB(proximal) of Q_BH⁻ and Nδ of D1-His215 was found to be unusually short (2.47 Å) in the Q_BH⁻ state [11]. Intriguingly, this distance is identical to that in the 1.9 Å structural model of the PSII monomer unit A of the PSII complexes [1]. The corresponding O_QB(proximal)–Nδ_D1-His215 distances were found to be 2.77 Å in the Q_B state and 2.68 Å in the Q_B^·− state [11], both being significantly longer than that in the Q_BH⁻ state. In addition, the corresponding Q_A-side distance (O_QA(proximal)–Nδ_D2-His214) was 2.78 Å in the 1.9 Å structure [1] and approximately 2.8–2.9 Å in the purple bacterial reaction centre [78]. Given that the Q_B geometry was less well defined than Q_A geometry in the 1.9 Å structure [1], the significance of the short O_QB(proximal)–Nδ_D1-His215 distance (2.47 Å) should be treated with caution. Furthermore, the Q_BH⁻ state is expected to be a short-lived intermediate and not a state that would be present in PSII under normal circumstances.

In a typical H-bond with an O–O distance longer than approximately 2.6 Å, the H atom is located near to the donor moiety owing to the larger pK_a value of the donor relative to the acceptor (with an asymmetric double-well potential H-bond [8], figure 1). On the other hand, according to the classification of H-bonds by Jeffrey [12] or Frey [13], short H-bonds with O–O distances of 2.4–2.5 Å can be classified as single-well (ionic) H-bonds [8] (figure 1). Because O–N distances are generally greater than O–O distances, the O_QB(proximal)–Nδ_D1-His215 of 2.47 Å is an unusually short H-bond and may possess a single-well potential. Remarkably, the calculated potential energy profile for the O_QB(proximal)–Nδ_D1-His215 in the Q_BH⁻ state was found to resemble that of a barrierless single-well (ionic) H-bond, suggesting that the second proton transfer could occur isoenergetically at O_QB(proximal) (figure 8). The significantly elongated H–N bond of D1-His215 (1.15 Å) in the Q_BH⁻ state implies that further migration of an H atom towards the acceptor O_QB(proximal) moiety (i.e. proton transfer) can easily occur. Indeed, the single-well potential obtained for O_QB(proximal)–Nδ_D1-His215 is symmetric (figure 8), implying that the pK_a difference [15] between D1-His215 deprotonation and Q_BH⁻ protonation is close to zero.

The pK_a for the QH⁻ to QH₂ protonation for plastoquinone is expected to be similar to that measured for ubiquinone in aqueous solution, i.e. 10.7, which is significantly higher than the pK_a of 4.9 found for the protonation of the semiquinone, Q^·− to QH^· [79,80]. The pK_a for deprotonation of a neutral His is expected to be similar to that for imidazole, i.e. approximately 14 [77]. In PSII, however, the ligation of D1-His215 to the positively charged Fe²⁺ should lower the pK_a of neutral D1-His215. The pK_a of the neutral His ligand to Fe²⁺ in the Rieske [2Fe–2S] cluster has been measured to be approximately 12.5 rather than approximately 14 [81,82]. The ligand environment of Fe²⁺ in PSII is more positively charged than that in the Rieske cluster; thus, the pK_a of neutral D1-His215 deprotonation is expected to be lower than approximately 12.5. In agreement with this, FTIR studies have indicated that deprotonation of D1-His215 occurs in response to pH changes [83]. Overall, the literature suggests that the pK_a value of D1-His215 is likely to be close to that for the Q_BH⁻ to Q_BH₂ protonation, in accordance with the single-well potential obtained here (figure 8).

6. Presence of an unusually short but stable H-bond in redox-active D1-Tyr161 (TyrZ) in photosystem II

The 1.9 Å structure confirmed that the OH group of TyrZ is situated an H-bond distance from the Nε of D1-His190. Most strikingly, the donor–acceptor distance of this H-bond (O_TyrZ–N_ɛ,His distance) is very short at 2.46 Å [1] (figures 6 and 9). The assignment of the atoms of TyrZ and D1-His190 appears to be quite reliable because the B-factors of TyrZ (22.4 for hydroxyl O), D1-His190 (22.4 for Nɛ and 23.3 for Nδ) and water molecules in the surrounding area (25.5 for W7 [72]) were relatively small for membrane proteins. In comparison with the corresponding distances of 2.78 Å in the previous 3.0 Å structure [84]; 2.67 Å in the previous 2.9 Å structure [85]; and 2.82 Å [86], 2.83 Å [87] and 2.88 Å [88] in the geometry-optimized Tyr-His model complexes, the O_TyrZ–N_ɛ,His distance of 2.46 Å in the 1.9 Å structure [1] is unusually short. It would therefore be interesting to elucidate how this short H-bond distance is energetically possible and what makes the distance so small.

Figure 9.

H-bond network around (a) TyrZ and (b) TyrD in PSII. Red and green spheres indicate water O and Cl atoms, respectively. (Online version in colour.)

Intriguingly, QM/MM geometry optimization of the neutral [TyrZ-O···H···N_ε-His-N_δH···O=Asn] state resulted in a O_TyrZ–N_ε,His length of 2.47 Å [64], which was in agreement with the corresponding distance of 2.46 Å found in the 1.9 Å structure [1]. By altering the O_TyrZ–N_ɛ,His distance and analysing the potential energy profile, it was confirmed that the O_TyrZ–N_ɛ,His distance of 2.47 Å exactly corresponded to the energy minimum of the H-bond, with this increasing both above and below the O_TyrZ–N_ɛ,His of 2.47 Å (figure 10).

Figure 10.

Energy profiles along proton transfer coordinate the H-bond between D1-Tyr161 (TyrZ) and D1-His190 in PSII. ΔE describes the difference in energy relative to the energy minimum. The red arrow indicates the energy difference from the energy minimum, although not always shown. (Online version in colour.)

The potential energy profile of the unusually short H-bond between TyrZ and D1-His190 resembles that of a single-well H-bond (figure 10). The presence of a symmetric H-bond between TyrZ and D1-His190 indicates that the pK_a values for the two moieties are similar [64]. These features are in contrast to those observed for the short O_Glu46–O_pCA bond in PYP (figure 11). As suggested by Schutz & Warshel [15], the identification of an LBHB with a single minimum potential can be valid only if the minimum is at the centre of the O_Glu46–O_pCA bond (i.e. the pK_a values of the two moieties are nearly equal). In PYP, even if the donor–acceptor distance is forced to shorten, the shape of the potential curve does not become symmetric and the energy minimum becomes even higher (figure 11) [45]. These comparisons indicate that the short H-bond in PYP is less likely to be an LBHB with respect to the short H-bond in PSII.

Figure 11.

Dependence of the potential energy profiles at O_Glu46–O_pCA = 2.32, 2.57 and 2.77 Å in PYP. The red arrow indicates the energy difference from the energy minimum obtained at O_Glu46–O_pCA of 2.57 Å. Compare the significantly different energetic properties of the O_Glu46–O_pCA bond from those of the O_TyrZ–N_D1-His190 bond in PSII, i.e. figure 10. In contrast to the symmetric shape of the potential energy curves irrespective of the donor–acceptor distances in PSII (figure 10), the shape of the potential energy curve never becomes symmetric in PYP, even if the distance is forced to be short. (Online version in colour.)

These observations raise the question of how the pK_a values of TyrZ and D1-His190 are equalized, making the formation of the unusually short H-bond energetically possible. In the PSII protein environment, two water molecules, W4 (O_TyrZ–O_W4 = 2.87 Å) and W7 (O_TyrZ–O_W7 = 2.62 Å), are at H-bond distances from the hydroxyl O atom of TyrZ [72]. W4 is one of the two Ca-ligated water molecules of the O₂-evolving complex, and W7 is H-bonded to W6 and another Ca-ligated water W3. Both W3 and W6 are connected with W5. Thus, the four water molecules W3, W5, W6 and W7 form a diamond-shaped water cluster, where W7 is situated at the interface between the water cluster and TyrZ (figure 9). Upon removal of all water molecules from PSII, the O_TyrZ–N_ɛ,His distance significantly increases to 2.71 Å [64], and remarkably, the highly anharmonic single-well potential curve (figure 12) obtained for the neutral [TyrZ-O···H···N_ɛ-His-N_δH···O=Asn] state (O_TyrZ–N_ɛ,His = 2.47 Å) is dramatically altered to an asymmetric double-well potential curve (figure 12). These results indicate that the presence of the water cluster contributes to matching of the two pK_a values in the PSII protein environment, thus facilitating the formation of the unusually short, symmetric H-bond.

Figure 12.

Alteration of a significantly short, single-well H-bond (O_TyrZ–N_D1-His190 = 2.47 Å) to a standard, asymmetric double-well H-bond (2.71 Å) upon removal of water molecules near the water-splitting catalytic site, Mn₄CaO₅ in PSII. (Online version in colour.)

7. Proton transfer via an H-bond network: mechanism of TyrD (D2-Tyr160) oxidation in photosystem II

In PSII, there are two redox-active Tyr residues [89–91], D1-Tyr161 (TyrZ) and D2-Tyr160 (TyrD), that can provide electrons to [P_D1/P_D2]^·+. TyrZ, which has D1-His190 as an H-bond partner, is the kinetically competent Tyr that mediates proton-coupled electron transfer from Mn₄CaO₅ to [P_D1/P_D2]^·+ (P680^·+). TyrD is not kinetically competent and plays no obligatory role in enzyme function; indeed, when TyrD is replaced by Phe, enzyme function appeared to be unaffected [90,91]. Nevertheless, TyrD is likely to play specific roles that are beneficial for PSII function (reviewed in [92]).

The neutral radical, TyrD-O^·, is formed upon oxidation of TyrD-OH by [P_D1/P_D2]^·+ [92–95], with this occurring in the tens of millisecond timescale [96] (see also [97]), which is slow enough that it does not compete with the rapid electron donation from TyrZ (which occurs on the tens to hundreds of nanosecond timescale). In the functional enzyme, TyrD oxidation occurs when the reversible intermediates of the water-splitting cycle, the so-called S2 or S3 states, equilibrate with TyrZ-O^·, and thence [P_D1/P_D2]^·+ allowing the slow electron donation from the TyrD, forming the stable radical, TyrD-O^·.

Once formed, TyrD-O^· is highly stable for many hours under physiological conditions, giving rise to the term ‘Signal II slow/dark’ describing the EPR signal from TyrD-O^· [92–94]. By contrast, TyrZ-O^· is reduced by Mn₄CaO₅ on the tens of microsecond to millisecond timescale [93,94]. It has been suggested that the proton released from TyrD-OH upon oxidation remains near to TyrD-O^· (e.g. [92]), with the crystallographic models showing a hydrophobic environment that appears to be consistent with this suggestion [1,84,98]. Until fairly recently, it was generally assumed that both TyrZ and TyrD underwent oxidation with the simultaneous transfer of the phenolic proton to the bases, D1-His190 and D2-His189, respectively, as originally suggested by Debus et al. [90]. However, subsequent FTIR studies have suggested that a proton carrier other than D2-His189 [99] could play a role in the redox properties of TyrD and that water could accept the proton from TyrD [100]. Exchangeable protons near TyrD-O^· have also been detected by ENDOR and ESEEM studies (e.g. [101,102]). The TyrD-O^· EPR signal was lost and/or significantly modified, and PSII photochemistry perturbed, when D2-Arg180 was mutated and this led Manna et al. [103] to propose that D2-Arg180 could accept or stabilize a proton from TyrD.

A recently resolved crystal structure has demonstrated the presence of a cluster of water molecules near TyrZ, but no corresponding cluster near TyrD [1], just a single water molecule. Curiously, this molecule is seen to occupy two different positions, proximal (H₂O_prox) and distal to the TyrD (H₂O_dist), separated by 1.8 Å, with B-factors of 20.1 and 19.3, respectively (figure 9). The proximal position is at H-bonding distance with the phenolic O atom of TyrD (O_TyrD–O_H2Oprox = 2.73 Å), whereas the distal position is beyond it (O_TyrD–O_H2Odist = 4.30 Å) and is instead at H-bonding distance with the guanidinium N atom of D2-Arg180 (O_H2Odist–N_D2-Arg180 = 3.01 Å).

The H-bond geometry of the water in the distal position (H₂O_dist) was obtained from QM/MM calculations, with TyrD taken as a deprotonated, neutral radical (TyrD-O^·) [104]. In the presence of TyrD-O^·, the water molecule would not remain in the H-bonding, proximal position (H₂O_prox), instead it would move to the distal position, even if it initially donated an H-bond to the phenolic O atom of TyrD-O^·. Thus, H₂O_dist and H₂O_prox correspond to deprotonated TyrD-O^· radical and protonated TyrD-OH states, respectively [104] (figure 13).

Figure 13.

H-bond patterns and concomitant water molecule positions (blue) for TyrD-OH (left) and TyrD-O^· (right) species in PSII. (Online version in colour.)

Another point of interest is where the proton from TyrD-OH goes when it is deprotonated to TyrD-O^·. Examination of the crystal structure [1] shows that an H-bond network is present beyond D2-Arg180, extending out to D2-His61 near the lumenal bulk surface via a series of water molecules [104] (figure 14). Indeed, previous electrostatic calculations have indicated that the protonated states of D2-Arg180 and D2-His61 were likely to be linked [105]. Here, QM/MM calculations based on a more recent crystal structure [1] show that the proton released from TyrD is transferred via the mobile water and D2-Arg180, all the way to D2-His61, through a concerted single-step proton transfer process [104] (figure 14). The proton does not return to the TyrD/D2-His189 moiety but goes through D2-Arg180 in an energetically favourable process irrespective of the presence of positively charged D2-Arg180. This demonstrates that there is no energy barrier for the proton transfer at D2-Arg180. As soon as the proton approaches the –NH₂ group of D2-Arg180, the NH-bond stretches towards the next water molecule, W480. Synchronizing the bond stretch, an OH-bond of W480 stretches towards W373. Similar bond stretching occurs at W373 and W783, and the proton is finally stabilized at D2-His61. Although the H-bond network terminates at W354 in the 1.9 Å crystal structure [1], the proton relay may continue further, releasing the proton into bulk water via W354. Overall, the calculations show that the proton is transferred to D2-His61 along a proton transfer pathway that involves several OH and NH-bond stretches towards acceptor moieties, without explicitly forming H₃O⁺ [104].

Figure 14.

Single-step proton transfer from TyrD to the bulk solvent and the associated H-bond patterns of the proton transfer pathway in PSII: (a) before and (b) after proton transfer. (Online version in colour.)

7.1. Energetics of the proton transfer pathway proceeding from TyrD

A concerted single-step proton transfer occurring over approximately 13 Å between TyrD and D2-His61 (figure 14) requires both a well-arranged H-bond network with appropriate distances between all donor–acceptor pairs (figures 5 and 6), and a sufficient driving force. For a proton moving along the pathway, the energy profile indicates that the reaction is sufficiently downhill from the TyrD-associated water to D2-His61, even when the D2-Arg180 moiety is included [104]. The driving force for the proton transfer towards the bulk surface appears to disfavour protons returning to TyrD-O^·, thus stabilizing the radical.

As far as we are aware, involvement of D2-His61 in the proton transfer pathway from TyrD has not been reported. The electrostatic link between D2-Arg180 and D2-His61, suggested from electrostatic calculations, is likely to be functionally relevant [105]. Remarkably, the H-bond donor–acceptor distances for W783 near D1-His61 are very short in the 1.9 Å crystal structure [1], namely O_W783–O_W354 = 2.51 Å and O_W783–N_D2-His61 = 2.58 Å (figure 14). In the QM/MM calculations, these distances were found to be 2.55 and 2.53 Å in the presence of protonated D2-His61 (i.e. after proton transfer), distances that are close to the original geometry of the crystal structure, whereas longer values were obtained (2.69 and 2.71 Å) in the presence of deprotonated D2-His61 (i.e. before proton transfer) [104]. Because a standard (asymmetric double-well) H-bond possesses O–O distances of approximately 2.8 Å [36,106], the very short H-bond distances between W783 and His61 may be an indication of a proton being present on D2-His61, or possibly shared with W783, which would thus be functioning as part of the proton transfer pathway from TyrD.

8. Concluding remarks

For a proton to be transferred, alternation of the pK_a values of the two moieties is expected to occur, otherwise, it would remain localized on the donor moiety. In order for this condition to be achieved, proteins either change the H-bond pattern, or change (equalize) the pK_a values of the proton donor and acceptor moieties (table 3). Changes in the H-bond pattern can be induced by photons, as in the example of the trans–cis photoisomerization of the double bond moiety in PYP [10], ultimately leading to equalized pK_a values of the donor and acceptor moieties and efficient proton transfer. Equalizing the pK_a values can also be achieved directly by oxidizing or reducing one of the H-bond moieties, as a result of an electron transfer.

Table 3.

Summary of the properties of short H-bonds in the crystal structures.

H-bond	Tyr-OH···N-His	His-NH···−O-QB-OH	Glu-COOH···−O-pCAa
geometry
protein	PSII	PSII	PYP
state	n.d.	QBH−/QBH2	pRCW
donor	D1-Tyr161-OH	D2-His214-NH	Glu46-OH
acceptor	D1-His190-N	−O-QB-OH	−O-pCA
distance (X-ray) (Å)	2.46b	2.47b	2.47c
distance (QM/MM) (Å)	2.47	2.47	2.49
short H-bond formation
trigger	n.d.	electron	photon
mechanism	n.d.	QBH− formation → increase in pKa(Oprox)	photoinduced trans–cis isomerization of pCA → removal of an H-bond from Tyr42
pKa (bulk solvent)	approximately 9···approximately 7	approximately 14···approximately 5d	approximately 4···approximately 9
stability (protein)	stable	transient	transient
associated conformational change	(no significant conformational change)	proton transfer → QB release from the PSII protein	proton transfer → pB state formation
location	inner core	near the bulk surface	near the bulk surface

^apK_a(Glu46) > pK_a(pCA) in the PYP protein environment (in the initial pG state).

^bSee [1].

^cSee [58].

^dAs pK_a(Q_BH^·/Q_B^·−) [79,80].

By equalizing the pK_a values, the H-bond donor and acceptor moieties can maximize proton transfer coupling, thus shortening the H-bond donor–acceptor distance. However, if the pK_a matching condition is satisfied only transiently (e.g. the pR_CW state in PYP or the Q_BH⁻ state in PSII), in particular near the protein bulk surface, the H-bond is unstable, leading to breakage and concomitant protein conformational changes (table 3).

It is also notable that the matching pK_a condition required for single-well H-bond formation is opposite to that required for salt-bridge formation, which plays a key role in protein–protein interactions. From these points of view, it is clear that symmetric, short H-bonds are not necessarily strong, in particular, when the matching pK_a condition is satisfied by strongly shifting the pK_a values of the donor and acceptor moieties in protein environments, owing to its anti-salt-bridge character.

Funding statement

This research was supported by the JST PRESTO programme (K.S.), Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan (22740276 to K.S. and 25107517 to H.I.) and Takeda Science Foundation (H.I.).