Quinones play an important role in biological proton and electron transfer processes that occur in both respiration and photosynthesis (Trumpower, 1982). In type II photosynthetic reaction centers two quinone molecules act as terminal electron acceptors (Ke, 2001a,b). The two quinones are often termed Q_A and Q_B. In this manuscript we will refer to the quinone binding site as Q_A and Q_B, however. The quinones that occupy the Q_A and Q_B binding sites have very different functions. The Q_A quinone is an intermediary cofactor involved in transferring electrons from (bacterio) pheophytin to Q_B, while the Q_B quinone couples proton and electron transfer processes (Ke, 2001b,c,d).

In this manuscript we focus on the Q_A binding site. The quinone occupying the Q_A binding site is species dependent. In Rhodobacter (Rb.) sphaeroides purple bacterial reaction centers (PBRCs) a ubiquinone (UQ) molecule occupies the Q_A binding site. In PBRCs from Blastochloris viridis (Shopes and Wraight, 1985) and Chloroflexus aurantiacus (Hale et al., 1983) a menaquinone occupies the Q_A binding site. In photosystem II reaction centers from oxygen evolving organisms, a plastoquinone (PQ) molecule occupies the Q_A binding site. In photosystem I the secondary electron acceptor, termed A₁, is a vitamin K₁ (VK) molecule (also called phylloquinone). Menaquinone and VK are both naphthoquinone (NQ) moieties that differ only in the degree of saturation of the tail at C₆.

Figure 1 shows the structure, numbering and abbreviations we will use for the various quinones discussed in this manuscript. MQ₀ and DMNQ are UQ and VK analogues, respectively, in which the hydrocarbon tail has been replaced with a methyl group. DQ is a PQ analogue in which the hydrocarbon chain at C₆ is replaced with a methyl group.

It has been suggested, at least for the UQ that occupies the Q_A site in PBRCs, that the role of the hydrocarbon chain at C₆ is to anchor and orient the quinone head-group in a specific way (Warncke et al., 1994). Data is available that may argue against this proposal, however, (McComb et al., 1990; Srinivasan and Golbeck, 2009; Wraight and Gunner, 2009).

Comparison of the properties of PBRCs with MQ₀ and UQ_n, or VK and DMNQ, incorporated into the Q_A binding site will allow one to assess how or if the hydrocarbon chain at C₆ modifies the quinones functional properties. Similarly, comparison of the properties of PBRCs with MQ₀ and DQ incorporated into the Q_A binding site will allow one to assess how or if the methoxy groups at C₂ and C₃ modifies the quinones functional properties. Calculated spectra for Rb. sphaeroides PBRCs with VK in the Q_A binding site can be compared to experimental spectra. These calculated spectra may also serve as a useful model for B. viridis PBRCs that naturally have VK incorporated into the Q_A binding site.

Fourier transform infrared (FTIR) difference spectroscopy (DS) is a sensitive molecular-level probe of pigment-protein interactions, and it is widely used to study both the neutral and reduced states of quinones in PBRCs (Breton and Nabedryk, 1996) and in photosystem II (Noguchi and Berthomieu, 2005). In this manuscript we focus on Q⁻_A/Q_A FTIR DS. Many molecular species contribute to Q⁻_A/Q_A FTIR DS, and in the past it has been difficult to identify which bands are associated specifically with UQ in the Q_A site. However, fully functional quinones can be incorporated into Q_A depleted PBRCs, and by collecting Q⁻_A/Q_A FTIR DS using PBRCs with unlabeled and isotopically labeled quinones incorporated, so called isotope edited FTIR difference spectra can be constructed, and from these spectra it has proven possible to separate contributions of the quinones from those of the protein in Q⁻_A/Q_A FTIR DS (Breton and Nabedryk, 1996). Previously, a variety of unlabeled and ¹⁸O labeled quinones have been incorporated into the Q_A binding site in PBRC's, and ¹⁸O isotope edited FTIR DS have been obtained (Breton et al., 1994a). The goal in this manuscript is the simulation of these experimental ¹⁸O isotope edited FTIR DS associated with the neutral state of the quinone in the Q_A binding site. Calculated IR spectra associated with the quinone anion state are considerably more complicated (Lamichhane and Hastings, 2013) and are currently being undertaken.

Although experimental Q⁻_A/Q_A FTIR DS have been obtained using PBRCs with various unlabeled and isotope labeled quinones incorporated, virtually no computational work aimed at modeling the experimental FTIR DS have been undertaken. Calculations aimed at modeling the vibrational properties of quinones in the Q_A binding site must account for the protein environment. Quantum mechanical (QM) calculations of such a large molecular system (pigment plus protein environment) are unfeasible. To circumnavigate this problem methods have been developed that allow one to separate the molecular system into distinct layers that can be treated at different levels of theory. In one layer the chemical properties of the principle species of interest (the pigment) can be calculated using “high-level” QM methods. The surrounding protein environment is included in the calculation but it is treated using computationally less expensive molecular mechanics (MM) methods.

Recently, we have undertaken QM:MM calculations for UQ in the Q_A binding site using the ONIOM method (Vreven et al., 2006). ONIOM is an acronym for: our Own N-layered Integrated molecular Orbital + Molecular mechanics package. In these calculations we showed that we could simulate experimental isotope edited FTIR difference spectra obtained using PBRCs with neutral UQ in the Q_A binding site. Here we extend upon these previous studies by attempting to simulate experimental isotope edited FTIR difference spectra obtained using PBRCs with symmetric tail-less and corresponding tail containing quinones incorporated into the Q_A binding site. We show that the calculated spectra agree well with the experimental spectra, further supporting the notion that the ONIOM method is a useful approach for understanding complex FTIR difference spectra associated with pigments in protein binding sites. We are also able to assess to what extent the quinone hydrocarbon chain may influence the calculated spectra.

Previously we have shown that ONIOM methods can be used to calculate isotope edited difference spectra for UQ₁ in the Q_A binding site, and that these calculated spectra model very well the corresponding experimental spectra (Lamichhane and Hastings, 2011). Here we considerably extend these studies, and show that ONIOM calculated isotope edited spectra for different quinones in the Q_A binding site model very well the corresponding experimental spectra. We further show that the calculated spectra for quinones in the gas phase are totally inappropriate for modeling the vibrational properties of quinones in the Q_A binding site. This is also very likely to be the case for modeling the properties of any protein bound pigment.

Without normal mode vibrational frequency calculations interpretation of experimental spectra is limited, and here we clearly show that our computational methods lead to a greatly increased understanding of the normal modes that contribute to the bands in the experimental spectra.

In our ONIOM calculations for neutral UQ₁ in the Q_A binding site we considered all amino acids included in the model only at the molecular mechanics level of computation. Previously, ONIOM calculations have been undertaken in order to model EPR data associated with the UQ anion in the Q_A binding site (Lin and O'Malley, 2008). In these calculations key amino acids, such as HisM219 and AlaM260, were considered at the higher quantum mechanical level of calculation. Taken together these results may indicate differences in the nature of H-bonding for the neutral and anion UQ species. Changes in H-bonding upon radical formation may be a mechanism for (de)stabilizing cofactors to fine-tune electron transfer processes in biological systems. Clearly it will be useful to undertake calculations similar to that presented here (modeling isotope edited FTIR difference spectra) for the UQ anion in the Q_A binding site, and such calculations are underway in our lab. Based on the above one could argue that the methods used here, with only the quinone treated at the QM level, will be inadequate to simulate the experimental spectra associated with quinone anions in the Q_A binding site. Or, for simulating the vibrational properties of quinone anions in the Q_A binding site it will be necessary to treat key amino acids at a quantum mechanical level. Again, calculations are underway in our lab to test this proposal.

For the neutral state of quinones occupying the Q_A binding site, calculations with the H-bonding amino acids treated at the MM level lead to calculated spectra that are in excellent agreement with experimental spectra (Figures 3–5). Clearly, treating H-bonding amino acids quantum mechanically will not lead to improved modeling of the experimental spectra. We have in fact undertaken calculations in which neutral UQ and the H-bonding amino acids are treated using QM, and we have found that the calculated spectra are very similar to that obtained when only neutral UQ is treated using QM (and the surrounding amino acids are treated using MM) (Zhao et al., J. Phys. Chem. B in press). Thus, it is very clear that, at least for the case of neutral quinones occupying the Q_A binding site, and as far as modeling isotope edited FTIR spectra is concerned, QM/MM calculations with only the quinone treated at the QM level need to be considered.

An experimental ¹⁸O isotope edited FTIR difference spectrum for UQ in the Q_A binding site is shown in Figure 5B. Three positive bands at 1660, 1629, and 1601 cm⁻¹ are observed. By considering FTIR difference spectra obtained using PBRCs with unlabeled and specifically ¹³C₁ and ¹³C₄ labeled UQ occupying the Q_A binding site, it was concluded that the 1660 and 1601 cm⁻¹ bands are due to the C₁=O and C₄=O vibrations of unlabeled neutral UQ, respectively. It was also concluded that the 1628 cm⁻¹ band is due to a UQ C=C vibration (Breton et al., 1994c). Since the C₄=O mode was so massively downshifted (from ~1660 cm⁻¹ for UQ in solvent to 1601 cm⁻¹ for UQ in the Q_A binding site) it was suggested that this group must be engaged in very strong hydrogen bonding, presumably with HisM219 (Figure 2) (Breton et al., 1994c). This conclusion is difficult to rationalize based on the crystal structural data and other experimental data [see Wraight and Gunner (2009) for a review]. Such a conclusion is also not supported by the data presented here. Specifically, the C₁=O and C₄=O modes of DMNQ and VK are found at 1653 and 1643 cm⁻¹ (Table 2), respectively, compared to ~1662 cm⁻¹ for the coupled C=O vibration in solution (Bandaranayake et al., 2006). Thus, the C₁=O/C₄= O mode of DMNQ or VK in the Q_A site is downshifted 9/19 cm⁻¹, respectively, compared to that found in solution. Such shifts suggest that both C=O modes of DMNQ or VK are H-bonded in the Q_A site, albeit quite weakly.

From the experimental spectrum of VK (or DMNQ) in Figure 3B two positive bands are observed at 1651 and 1640 cm⁻¹ and one negative band at 1620 cm⁻¹. The two C=O modes of unlabeled VK give rise to the positive bands at 1651 and 1640 cm⁻¹, but only a single band is observed at 1620 cm⁻¹ upon ¹⁸O labeling. Two interpretations for these observations have been proposed (Breton et al., 1994b). One suggestion is that upon ¹⁸O labeling the 1640 cm⁻¹ band downshifts to ~1620 cm⁻¹, while the 1651 cm⁻¹ band downshifts to near 1640 cm⁻¹ and decreases in intensity. The negative band near 1640 cm⁻¹ due to a C = ¹⁸O group of VK is then masked by the positive band (also at 1640 cm⁻¹) due to the unlabeled C=O group. A second hypothesis is that the two C=O modes of unlabeled VK (at 1651 and 1640 cm⁻¹) both downshift to ~1620 cm⁻¹ upon ¹⁸O labeling. The different ¹⁸O induced shifts of the two C=O modes results from differential coupling to C=C modes.

The calculated data presented here allow us to address which of these interpretations could be appropriate, or if either is appropriate. The ONIOM calculations show that the C₁=O and C₄=O modes of unlabeled VK occur at 1652 and 1642 cm⁻¹, and that neither of these modes are coupled to C=C modes (Table 2). Upon ¹⁸O labeling the C₁=O/C₄=O mode downshifts 24/18 cm⁻¹, respectively. The modes of ¹⁸O labeled VK also display considerable coupling with C=C modes. Upon ¹⁸O labeling the C₁=O/C₄=O group couples with C=C_ring in-phase/out-of-phase vibrations, respectively. Coupling of the C₄=O group to the out of phase C=C vibration leads to a large intensity enhancement, while coupling of the C₁=O group to the in-phase C=C vibration leads to a large intensity decrease (Table 2). So, the calculations indicate that two separate uncoupled C=O modes in unlabeled VK give rise to predominantly a single mixed mode (that carries most of the intensity) in ¹⁸O labeled VK. These calculated results indicate that neither of the two previously proposed interpretations of the experimental spectra is correct. Clearly, the calculations presented here allow a more detailed insight into the nature of the bands in the experimental isotope edited FTIR difference spectra.

Experimentally, the vibrational modes of DMNQ are at a slightly higher frequency (~3 cm⁻¹) than corresponding modes of VK. Presumably replacing the C₆ methyl group with a phytyl unit causes this difference. Interestingly, this small frequency difference in the modes of DMNQ and VK is also found in our ONIOM calculated spectra (compare spectra in Figures 5C,B). This result is not entirely specific to the ONIOM method, however, as a small shift is also found in the gas phase calculations (Figure 3D).

The C₁=O and C₄=O modes of DMNQ and VK are calculated to be separated by 10 cm⁻¹. This separation cannot be due to differences in the molecular group attached at C₆. It must be due to differences in how the two C=O groups interact with the protein. Similarly, the two C=O modes of DQ are calculated to be separated by 13 cm⁻¹, and this separation is also likely due to differences in how the two C=O groups of DQ interact with the protein.

For UQ₁ (and MQ₀) the separation of the C=O modes is ~32 cm⁻¹ (1660–1628). Some of the differences in frequency of the C=O modes of UQ are due to the different orientation of the methoxy groups. If we assume that protein interactions with the C=O groups gives rise to a 13 cm⁻¹ separation in the frequencies of the two C=O modes then this would indicate that the difference in the orientation of the methoxy groups of UQ (or MQ₀) gives rise to a frequency difference of 19 cm⁻¹ for the two C=O groups. This result is in quite good agreement with results from experimental spectra of UQ in solution, which show that the two C=O modes are separated by ~16 cm⁻¹ (Breton et al., 1994c; Brudler et al., 1994).

Normal mode vibrational frequencies are governed by molecular bonding force constants. These force constants relate to the electronic structure of the molecule. Since the calculated and experimental spectra for DMNQ and VK are virtually the same it is concluded that the replacing the phytyl unit at C₆ of VK with a methyl group does not appreciably perturb the electronic structure of the NQ ring. In addition, the tail does not perturb the protein in a way that significantly modifies any pigment protein interactions.

In VK and UQ₁, the “kink” in the hydrocarbon chain after the first carbon atom is 3.1–3.2° higher in ONIOM calculations compared to gas phase calculations. It seems therefore, that the hydrocarbon chain is somewhat constrained relative to the quinone ring when incorporated into the Q_A binding site. It is not clear if this is a significant constraint, however. The C₁=O and C₄=O bonds of MQ₀ and UQ₁ (and of DMNQ and VK) are virtually unaltered in ONIOM calculations (Table 1). The hydrogen bond lengths for the C=O groups are also little altered (Table 1). The distance of the C₄=O oxygen to the non heme iron atom is 0.029/0.055 Å longer for UQ₁/VK compared to MQ₀/DMNQ, respectively, suggesting a very small change in orientation of the UQ₁/VK head-group (since the iron atom is fixed) compared to MQ₀/DMNQ in the Q_A binding site. The hydrocarbon chain may therefore, lead to a very slight change in the orientation of the quinone ring in the Q_A binding site. Figure 2F shows that there is only a very small difference in the orientation of DMNQ relative to UQ₁.

The experimental ¹⁸O isotope edited spectra for MQ₀ and UQ₁ are quite different (compare Figures 5B,C). This difference cannot be modeled computationally (Figure 5A). The experimental isotope edited spectrum for MQ₀ is considerably noisier than the spectrum for UQ (Breton et al., 1994a). As far as we are aware the Q⁻_A/Q_A FTIR difference spectrum for RCs with MQ₀ in the binding site have never been reproduced, so the accuracy of the spectrum may be somewhat questionable. On the other hand MQ₀ may be able to incorporate into the Q_A binding site with the methoxy groups oriented in several different ways. As described previously, each of these methoxy group conformers will have slightly different spectra (Lamichhane et al., 2010). This heterogeneity in orientation of MQ₀ in the Q_A binding site may be a factor that contributes to differences in the experimental spectra for MQ₀ and UQ, as demonstrated in Figures 5B,C. In spite of this, there are some overall similarities in the MQ₀ and UQ₁ experimental spectra in Figures 5B,C. For the MQ₀ spectrum positive bands are found at 1665, 1631, and 1608 cm⁻¹. For UQ₁ the bands are at 1660, 1628, and 1601 cm⁻¹. For the MQ₀ spectrum negative bands are found at 1616 and 1602 cm⁻¹. For UQ the bands are at 1613 and 1586 cm⁻¹.

The ONIOM calculated and experimental isotope edited spectra for VK are very similar, although the intensity ratios of the different bands do not appear to match well. For example, the intensity ratio of 1652 and 1644 cm⁻¹ bands in the calculated spectrum (Figure 3A, dotted) is ~2.1, compared to a ratio of ~0.95 in the experimental spectrum. To investigate this further we have also calculated the ¹³C isotope edited spectrum for VK in the Q_A binding site (Figure 6A) (global ¹³C labeling of VK), and compared it to the corresponding experimental spectrum (Figure 6B). For completeness the calculated ¹³C isotope edited spectrum for VK in the gas phase is shown in Figure 6C. Again, it is evident that the calculated gas phase spectrum in no way resembles the experimental spectrum (compare Figures 6B,C). The normal modes (frequencies, intensities, and PEDs) that give rise to the various bands in the ONIOM calculated ¹³C isotope edited spectra of VK in Figure 6A are listed in Table 3.

Notice that the ONIOM calculation predicts that the two C=O modes remain completely separate (do not mix) upon ¹³C labeling (Table 3), which is unlike the behavior observed for the C=O modes upon ¹⁸O labeling (see above). Also notice that in the experimental isotope edited spectrum (Figure 6B) the 1640 cm⁻¹ band is now considerably more intense than the 1651 cm⁻¹ band, whereas in the calculated spectrum the higher frequency band is considerably more intense. The origin of these mode intensity differences are not entirely clear and are currently being investigated by considering calculations associated with VK in the presence of various types of H-bonding molecules.

We have shown that ONIOM type QM/MM calculations can be used to model experimental isotope edited FTIR difference spectra obtained using PBRCs that have had several different quinones incorporated into the Q_A binding site. The fact that the many different spectra can all be modeled is a clear indicator of the appropriateness of the approach.

The calculated spectra appear not to depend on whether the quinone incorporated has a prenyl/phytyl unit or a methyl group attached at C₆. The electronic structure of the quinone ring is therefore, not sensitive to the presence or absence of a hydrocarbon side chain at C₆. This suggests that the hydrocarbon side chain does not significantly constrain the quinone ring in the Q_A binding site.

Comparison of the calculated and experimental spectra, in combination with a consideration of the calculated PEDs of the normal modes, allows a direct assessment of the appropriateness of previous suggestions as to the origin of the bands in the experimental spectra.