4290
J. Phys. Chem. B 2008, 112, 4290-4297
Molecular Mechanisms of Cryoprotection in Aqueous Proline: Light Scattering and Molecular Dynamics Simulations R. Z. Troitzsch,† H. Vass,† W. J. Hossack,† G. J. Martyna,*,‡ and J. Crain†,§ School of Physics, The UniVersity of Edinburgh, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom, IBM T.J. Watson Research Center, Yorktown Heights, New York 10598, and National Physical Laboratory, Hampton Road, Teddington TW11 0LW, United Kingdom ReceiVed: August 21, 2007; In Final Form: October 11, 2007
Free proline amino acid is a natural cryoprotectant expressed by numerous organisms under low-temperature stress. Previous reports have suggested that complex assemblies underlie its functional properties. We investigate here aqueous proline solutions as a function of temperature using combinations of Raman spectroscopy, Rayleigh-Brillouin light scattering, and molecular dynamics simulations with the view to revealing the molecular origins of the mixtures’ functionality as a cryoprotectant. The evolution of the Brillouin frequency shifts and line widths with temperature shows that, above a critical proline concentration, the water-like dynamics is suppressed and viscoelastic behavior emerges: Here, the Landau-Placzek ratio also shows a temperature-independent maximum arising from concentration fluctuations. Molecular dynamics simulations reveal that the water-water correlations in the mixtures depend much more weakly on temperature than does bulk water. By contrast, the water OH Raman bands exhibit strong red-shifts on cooling similar to those seen in ices; however, no evidence of ice lattice phonons is observed in the low-frequency spectrum. We attribute this primarily to enhanced proline-water hydrogen bonding. In general, the picture that emerges is that aqueous proline is a heterogeneous mixture on molecular length scales (characterized by significant concentration fluctuations rather than well-defined aggregates). Simulations reveal that proline also appears to suppress the normal dependence of water structure on temperature and preserves the ambient-temperature correlations even in very cold solutions. The water structure in cold proline solutions therefore appears to be similar to that at a higher effective temperature. This, coupled with the emergence of glassy dynamics offers a molecular explanation for the functional properties of proline as a cryoprotectant without the need to invoke previously proposed complex aggregates.
1. Introduction and Motivation For several decades it has been known that accumulation of free proline amino acid in plants occurs upon exposure to lowtemperature stress.1 In numerous cases, a direct relationship has been established between the accumulation of proline during cold acclimation and freezing tolerance.2,3 It is also known that certain transgenic plants having enhanced proline levels show higher freezing tolerance than do the corresponding wild types.4 In this sense, proline is considered a common natural cryoprotectant (osmolyte) and its expression under adverse conditions has also been observed in bacteria, invertebrates, protozoa, and algae.5 Considerable attention has now been focused on the biochemical and signal transduction pathways governing proline level regulation under low-temperature conditions. However, the molecular properties of simple aqueous proline amino acid solutions at low temperatures have been largely unexplored. As a result, the mechanisms underlying its protective properties are not understood from this perspective. Under ambient conditions, there are disputed claims that the mixtures have a complex mesoscale structure in which proline molecules form semiordered aggregates in aqueous solution. These associations, * Address correspondence to this author. † The University of Edinburgh. ‡ IBM T.J. Watson Research Center. § National Physical Laboratory.
it has been proposed, resemble polymer-like “hydrophilic colloids”.6 The basic structural elements of these assemblies are assumed to comprise a hydrophobic backbone of stacked pyrrolidine rings with solvent-exposed hydrophilic groups on the surface.6 Some indirect support for this idea has come also from independent measurements including calorimetric data7 which revealed evidence for low-temperature eutectic phase separation in aqueous proline at a range of concentrations. A similar inference is drawn from Fourier-transform infrared (FTIR) spectra which show a splitting of the COO- asymmetric stretch band on increasing proline concentration implying that the carboxylate group exists in distinct local environments one of which is possibly the proposed pyrrolidine stack.7 There are also reports of a significant increase in light scattering in proline solutions above 1.5 M concentration8 that may be explained by proline clustering. The same authors also report an increase in emission intensity of a hydrophobic dye as a function of proline concentration in aqueous solution. This was interpreted as additional evidence for the formation of supramolecular assemblies of proline having distinct hydrophobic regions.8 However, to our knowledge the presence of such aggregates, even if they are present, does not provide molecular-level insight into the protective properties of proline. Moreover, recent computer simulations and neutron diffraction data (over intermediate and low Q ranges) at ambient temperature do not reveal evidence for such structures.9 Numerous local motifs, however,
10.1021/jp076713m CCC: $40.75 © 2008 American Chemical Society Published on Web 03/15/2008
Cryoprotection in Aqueous Proline
J. Phys. Chem. B, Vol. 112, No. 14, 2008 4291
have been identified including dimers and short oligomeric chains formed from conventional as well as bifurcated hydrogen bonds. In this paper we explore aqueous proline solutions at low temperature using a combination of light scattering, Raman spectroscopy, and computer simulation. The objectives are to elucidate the evolution of low-temperature structure and dynamics and thereby to reveal aspects of molecular scale phenomena that may be relevant to the rudimentary function of proline as a natural cryoprotectant. The paper is organized as follows: we first outline the experimental and computational methods used. The results of Rayleigh-Brillouin light scattering (long wavelength excitations) are discussed next, which are interpreted by using generalized hydrodynamics methods to fit to the scattered spectrum. We then present Raman data at high frequency to explore the temperature evolution of the water OH stretching bands (a reporter of hydrogen bonding) and at low frequency to investigate the lattice mode region on cooling. Finally, we introduce computer simulation based on classical empirical potentials.
In relaxing liquids, spontaneously generated sound waves transfer energy to the nonpropagating modes thereby enhancing the central line. The simple hydrodynamic theory is inadequate for this case and the expression must be modified to include a frequency-dependent, longitudinal modulus according to RLP ) γ0(M′(ω)/M0) - 1, where M′(ω) and M0 are the moduli at the Brillouin and zero frequency regimes, respectively.11 In mixtures, concentration fluctuations arising from transient aggregates of the components are also present. Like entropy fluctuations, these relax by diffusive processes and are nonpropagating. They contribute to the central Rayleigh line and thereby increase RLP. Measurements of the LP ratio in various binary mixtures have been made and used to infer qualitative information on local aggregation in alcohols12 and the theory of the Rayleigh-Brillouin spectra and Landau-Placzek ratio for multicomponent liquids has been considered by a number of authors.13-15 The basic expression for the LP ratio of a multicomponent mixture is:16
RLP ) (γ - 1)[1 + Kx1x2] where
2. Methods 2.1. Information Content and Theoretical Background. Brillouin spectroscopy measures the inelastic scattering of light from propagating adiabatic density fluctuations and is a measure of acoustic excitations at small wave-vectors (q ≈ 10-3 Å-1) and high frequencies. Nonpropagating entropy fluctuations contribute to Rayleigh scattering. In this simplest formulation, an approximate form of the dynamic structure factor gives the triplet line-shape for Rayleigh-Brillouin scattering I(ν) as
ΓR ΓBνB2 I(ν) ) AR 2 + A B ν + ΓR2 (νB2 ( ν2)2 + (νΓB)2
(1)
where the first term is the central Rayleigh component and the second term is the Brillouin doublet. The Rayleigh line width is determined by the thermal diffusivity ΓR ) DTq2. The lifetime of the propagating acoustic waves is (ΓBq2)-1, where ΓB is the acoustic attenuation. These relationships are obtained from linearized forms of hydrodynamic constitutive equations that express conservation of mass energy and momentum. The Brillouin frequency shift frequency νB (typically in the frequency range 1-10 GHz) is given by
νB ) ((2nVs/λ0) sin(θ/2)
(2)
in which the speed of hypersonic acoustic waves (Vs) depends on the density (F) and adiabatic compressibility (χS) according to VS ) 1/xFχS. The ratio of the integrated elastic to inelastic scattered intensity defines another important quantity, the LandauPlaczek ratio,10 RLP. It is obtained experimentally from RLP ) (IR/2IB) ) (AR/AB). The factor of 2 arises because of the Brillouin doublet. The value of RLP is therefore determined by thermodynamic fluctuations including microscopic aggregation and it also represents a measure of nonideal behavior: For example, in monatomic, nonrelaxing liquids, RLP is completely prescribed by the constant pressure and constant volume specific heats (cp and cv, respectively) according to RLP ) cp/cv - 1 ) γ0 - 1. For a derivation of this relationship directly from classical scattering theory (based on entropy and pressure fluctuations) the reader is referred to the paper by Cummins and Gammon.10
(3)
K)
Cp (∂n/∂x2)2
(4)
RT2 (∂n/∂T)2
where n is the refractive index of the material and the derivatives represent variation of n with respect to solute mole fraction, x2 (concentration fluctuation), and temperature, T. In general, agreement between theory and experiment is fairly good though there remains some controversy over the intensity of the Landau-Placzek ratio in glass formers,17 which tends to be very severely underestimated by theory relative to light-scattering measurements. 2.2. Experimental Methods. Samples were obtained from commercial suppliers and used without further purification, but were passed through 0.22-Millipore filters to remove any residual dust. Brillouin spectra were collected with a Burleigh Fabry-Perot interferometer set for high contrast in five-pass mode. The interferometer was actively stabilized against drift due to thermal and mechanical disturbances, thereby allowing data collection to be performed over long periods. All spectra were recorded in right-angle scattering geometry, using the 514.5 nm excitation line of a single-mode argon ion laser. The line-spread function G(ω) (used later in the deconvolution of the scattered spectrum) in five-pass mode is given approximately by a Lorentzian line-shape raised to the fifth power. That is,
G(ω) )
(
( ))
1
1 + 4F2
5
ω ∆ωp
2
(5)
Here the free spectral range is set to ∆ωp ) 66π G Rad-1. The instrument finesse has been measured to be F ) 15.9 ( 0.1 by least-squares fitting of the above expression to the central line of a reference solution at -90 °C where there was no detectable Brillouin spectrum. This gives an effective instrument finesse of 43.1 and full-width at half-height of approximately 2.5π G Rad-1, which is 30% of the width of typical Brillouin peaks obtained from the proline samples. This experimentally measured finesse is lower than the theoretical reflection finesse of ∼95 as a result of jitter introduced by the active stabilization system. Thus, to obtain realistic results for the physical
4292 J. Phys. Chem. B, Vol. 112, No. 14, 2008
Troitzsch et al.
parameters of interest, this instrument line-shape must be deconvolved during the fitting process. In the relaxing regime, we can attempt to fit to the RayleighBrillouin spectra using a Debeye-type relaxation model. In this simplest case, we assume a single relaxation time, τ, and the memory function takes the form M(ω) ) M′(ω) + iM′′(ω) ) iτ/(1 - 1ωτ). Then the isotropic scattered intensity (proportional to the dynamic structure factor) takes the form
γ0 + M′′(ω) I(ω) ) I0 2 [ω - ω02 + ωM′(ω)]2 + ω2[γ0 + M′′(ω)]2
(6)
where
M′ )
∆2τ -∆2ωτ2 and M′′ ) 1 + (ωτ)2 1 + (ωτ)2
Figure 1. Raman spectrum of aqueous proline at two concentrations showing the OH water band region of the spectrum.
where ∆ is the relaxation strengthsa measure of the coupling between the structural relaxation and the longitudinal acoustic wave. The expression has the form of the power spectrum of a harmonic oscillator with frequency-dependent damping. The γ0 parameter is obtained from the low-temperature limiting line width where structural relaxation effects are negligible. In this limit, we assume that all residual processes are fast relative to the acoustic wave period. The value is set to γ0 ) 0.6π G Rad-1. The measured line-shape is also subject to arbitrary brightness and variable constant offset, so the full function characterizing the Brillouin line-shape is given by
F(ω) ) aI(ω) + b
(7)
thereby requiring a fit over the five parameters ω0, τ, ∆, a, and b. If we also include the instrument line-shape, then the total function to be fitted is H(ω) ) F(ω) X G(ω). The convolution does not have an analytical solution, so must be calculated numerically by using a least-squares fitting process. Least-squares fits of the above function showed very strong correlation between the required τ relaxation time and ∆, which resulted in instability. For the temperature range of interest, we therefore assume that ∆ is approximately constant and, as previously discussed, the maximum ΓB occurs at T ) 0 °C where ωτ ≈ 1. This gives a value of ∆ ) 42.3 Rad-1, which was then fixed for the other numerical fits. In the Raman experiments described here we use a Coderg T800 triple-grating spectrometer. The 514.5 nm line of an Ar + laser was used as the excitation source and scattered light was collected at right angles to the excitation. In both Raman and Brillouin measurements temperature was controlled to better than 0.1 °C. 2.3. Computational Methods. Classical molecular dynamics (MD) simulations were performed on the 2.75 M aqueous proline solution at T ) -45, -15, and +15 °C, using the Charmm2218,19 force field and the TIP4P20 empirical potentials for water. TIP4P is a well-known four-site model for water that is rigid and nonpolarizable. The additional site associated with the oxygen charge is displaced along the bisector of the HOH angle. A primitive cell containing 6860 and 343 water and proline molecules, respectively, was initially placed on a cubic lattice with randomized molecular orientations in a simulation box of edge length 62.8 Å. The system was then equilibrated to anneal out any unphysical high-energy contacts, using the canonical ensemble, for 300 ps at a temperature of T ) 27 °C. To improve equilibration and minimize configurational sampling errors at
low temperature we cooled the system by gradual quenching (at a rate of -1 K/ps) to arrive at the three required temperatures. Each system was then relaxed for a further 60 ps in the isothermal-isobaric ensemble at P ) 1 bar, before production runs of 8 ns were performed in the microcanonical ensemble. Relaxation of the low-temperature systems was tested by computation of the time correlation functions for S(Q), with 10 different values of Q chosen to probe relevant length scales. The time scale for decorrelation was found to be much shorter than the production run time of 8 ns. Therefore we assume that the local hydration structure we observe at low temperature is properly equilibrated. Nose´-Hoover chain thermostats21,22 and barostats23,22 were used to create the canonical and isothermal-isobaric ensembles from the MD simulations. Multiple time step integrators,21-23 incorporating constraints on the high-frequency OH and NH bond vibrations (SHAKE24 and RATTLE25), together with a 15.999 au hydrogen mass, allowed for a 6.0 fs “outer” timestep21-23 to be employed for the canonical and isothermalisobaric ensembles. For the microcanonical ensemble, a hydrogen mass of 1.0 au together with a time-step of 0.25 fs was employed to capture correctly the dynamics of bond vibration. Periodic boundary conditions were assumed and Ewald summation was employed to compute long-range interactions, enabled by the Smooth Particle Mesh26 approximation to the reciprocal space part. 3. Results and Discussion We first show the Raman spectra of the OH stretching region for dilute and concentrated proline mixtures in Figure 1. The Raman spectra of liquid water in this spectral region, which includes two primary but very broad and partially overlapping bands, has been studied very extensively. There remains some debate as to the physical origin of these signature bands of water,27 however, it is generally understood that they arise from distinct hydrogen-bonded configurations.27 In particular, Hbonding tends to displace electron density from the OH-covalent bond leading to lower Raman frequencies. Therefore, the higher frequency OH bands are assigned to the least associated molecules and the lowest frequency sub-bands to those most strongly hydrogen bonded. Assuming these general assignments, it is evident that under ambient conditions very little perturbation to these bands occurs upon introduction of proline for concentrations up to the solubility limit. We therefore assume that either the local H-bond environment of the water is not significantly
Cryoprotection in Aqueous Proline
Figure 2. Landau-Placzek ratio of aqueous proline solutions at several concentrations.
disrupted by proline under ambient conditions or that proline substitutes for water-water interactions with little discernible effect on the spectral signature. We show the behavior of the Landau-Placzek (LP) ratio as a function of composition and temperature in Figure 2. It is evident here that the LP ratio shows a complex behavior. Near ambient conditions, there appears to be a single maximum at 0.055 mole fraction solution where RLP ≈ 5.0. The concentration corresponding to maximum RLP ratio does not depend sensitively on temperature. In addition to this temperature-independent peak, which appears at low concentrations, we also note that the LP ratio increases very dramatically in the more concentrated solutions on cooling. This increase eventually leads to very large RLP ratios at low temperature that exceed the value of the temperature-independent maximum. As stated previously, RLP for a binary mixture arise from two sources-concentration fluctuations and structural relaxation. We will return to the behavior of RLP(T) and the separation of these two contributions, but will first examine the Brillouin frequencies and line widths. As the samples are cooled, a number of features of the Rayleigh-Brillouin spectra are notable: First, in the more dilute proline regime, we note that decreasing temperature leads to a red-shift of the mode frequency. This decrease in frequency (sound speed) is a unique property of pure liquid water connected to the high values of compressibility and density at low temperatures.28 In this dilute concentration regime, it appears that proline does not disrupt the network sufficiently to perturb the structural origin of this peculiar dynamical property (see Figure 4). Moreover, in this dilute regime we note that the lowest temperature reached is approximately -20 °C before crystallization occurs. Near concentrations of 0.07 mole fraction of proline we find that the Brillouin mode frequency (≈6.50 GHz) is temperature independent down to the lowest temperatures reached in these measurements. Here we can estimate the sound speed (Vs) and hypersonic adiabatic compressibility (χS) from eq 2. For our right-angle scattering geometry and excitation wavelength, and assuming a refractive index of 1.5 and a density of 1.1 g/cm 3 (based on computer simulations), we obtain a hypersonic sound speed of Vs ≈ 1.58 × 10 5 cm/s. This is only slightly higher than that reported for pure water by Cummins and Gammon.10 From this the computed hypersonic adiabatic compressibility at this concentration is found to be χS ≈ 36.5
J. Phys. Chem. B, Vol. 112, No. 14, 2008 4293
Figure 3. Rayleigh-Brillouin spectra for a concentrated proline solution at several temperatures. At this concentration the frequencies increase monotonically with decreasing temperature whereas the widths go through a maximumsbehavior characteristic of viscoelastic liquids. Three diffraction orders are shown within the free spectral range. At the lower temperatures the Brillouin peaks from other orders overlap.
Figure 4. Brillouin frequencies as a function of temperature at several proline concentrations in water. In all cases the lowest temperatures shown correspond to the point at which crystallization occurs with the exception of the most concentrated sample. The concentration near 0.0741 mole fraction appears to separate water-like behavior from normal liquid-like behavior.
× 106 bar-1. This is somewhat lower than the isothermal or adiabatic value for pure water, which is reported as 45.5 × 106 bar-1.10 More concentrated solutions no longer show water-like behavior and crystallization is inhibited. Rather, the mode frequencies increase on cooling. In the more concentrated mixtures this increase is very large at the lowest temperatures. The mode line widths are shown as a function of temperature in Figure 5 for several concentrations. It is clear that the Brillouin line width ΓB increases on cooling for all solutions including pure water. In the most concentrated solution, we note that the line width reaches a maximum value at T ) 0 °C and then narrows on further cooling. This non-monotonic temperature evolution of the Brillouin line width is consistent with the high-frequency dynamics expected for viscoelastic liquids: It occurs because the characteristic time for structural relaxation increases on cooling. Specifically, at ambient temperatures, the structural relaxation time is short relative to the inverse Brillouin frequency, ωBτ , 1. The relaxation does not couple effectively to the sound wave and the Brillouin line width is therefore
4294 J. Phys. Chem. B, Vol. 112, No. 14, 2008
Troitzsch et al.
Figure 6. Plot of relaxation time τ for temperatures in the range -50 to 40 °C at a concentration of 0.1818 mole fraction. Figure 5. Brillouin mode line width (Γ) as a function of temperature at several concentrations of proline in water.
relatively sharp and its frequency relatively low. As the sample is cooled, the structural relaxation slows down and its characteristic time scale begins to overlap with the acoustic wave period. The lifetime of the acoustic wave therefore decreases and the related line width broadens to a maximum width reached when the condition ωBτ ≈ 1 is reached. This relation allows a simple estimate of the relaxation time to be made. For the most concentrated solution, the maximum in ΓB occurs almost exactly at the freezing temperature of pure water T ) 0 °C at which point we obtain a value of τ ≈ 17 ( 1 ps based on this simple estimate. In a slightly less concentrated solution (1:5 mole ratio or 0.17 mole fraction), the maximum line width is not reached until the lower temperature of -15 °C at which point the relaxation time is estimated to be similar at τ ≈ 15.8 ( 1 ps. Compared to other systems, this value is very close to that estimated in the same way for glassy ethanol-water mixtures at T ) 150 K (-123 °C) by Ko and Kojima.29 In the case of glassy ethanol, these authors also point out that a relaxation time in this range is faster than that expected for dielectric R or β processes and speculate that the origin of the fast process in the proline glass may be the coupling of intramolecular degrees of freedom (including low-amplitude libration) with the acoustic waves. However, the apparent relaxation time for proline solutions is somewhat faster than that reported in aqueous LiCl, which has a line width maximum also near T ) 0 °C that is reported as 8.2 ps. On continued cooling, the relaxation time slows further, exceeding that of the acoustic-wave period (ωBτ . 1). In this regime the liquid supports long-lived acoustic excitations accounting for the observed recovery of narrow line widths. These features of the Brillouin light-scattering spectra observed in aqueous proline are therefore classic signatures of viscoelastic behavior and have been observed in numerous glass formers.30,31 Figure 8 shows the spectra fit obtained via the generalized hydrodynamics model according to eq 6. From the fits we obtain estimates of the relaxation time τ as a function of temperature that is shown in Figure 6 shows the relaxation time τ against the same temperature range giving a τ ≈ 19 ps at 0 °C. This is in good agreement to the value of τ ≈ 17 ps estimated from the condition ωτ ≈ 0. The results show an approximately linear relation with decreasing temperature until approximately -30 °C, after which there is an apparent rapid increase. At temperatures lower than -50 °C the Brillouin signal was too small and dominated by noise to obtain reliable data fits.
Figure 7. Raman frequencies in the OH stretching region as a function of temperature showing a pronounced red-shift on cooling. The inset shows the low-frequency lattice mode region of the spectrum (0-200 cm-1) in which no ice-like phonons are observed.
Returning now to the complex behavior of RLP, we note that the temperature-independent peak occurs in a concentration range where the Brillouin line widths are not significantly broadened at ambient temperature; neither are the frequencies much increased. In this regime, structural relaxation does not appear to make a significant contribution to the LP ratio. We therefore assume the temperature-independent LP ratio maximum arises primarily from concentration fluctuations. These results extend significantly the room temperature light scattering results of Samuel8 which showed increased scattering intensity above 1.5 M. These saturated above 2.5 M but the conclusion is consistent with the early notion that loose aggregates due to concentration fluctuations are present in the solution. The behavior is also similar to the data on aqueous propanol.12 In that system, a peak in RLP ≈ 5 at concentrations of about 12 mol % alcohol is seen,12 which is consistent with the results of small-angle X-ray diffraction.12 By contrast, no peak in RLP is observed for methanol at any concentration. Only a weak monotonic increase with concentration is found12 even though molecular scale segregation has been suggested by both neutron diffraction and computer simulation.32,9 Interestingly, we note that the concentration range at which the water-like behavior is suppressed (sound speed ceases to decrease on cooling) in the proline mixtures is very near that at which the temperature-independent peak in RLP is found. That
Cryoprotection in Aqueous Proline
J. Phys. Chem. B, Vol. 112, No. 14, 2008 4295
Figure 10. Association of water and proline at low temperature. Notably the polar groups in proline are used as pseudo-water sites in the tetrahedral structure of water. We suggest this to be the origin of the ice-like red-shifts in the Raman data, without the concurrent occurrence of ice-like phonons. Figure 8. Spectra and fits according to the generalized hydrodynamics model described by eq 6 at select temperatures and a concentration of 0.1818 mole fraction.
TABLE 1: Ratio of 4-Fold Coordinated Water Molecules to Total Number of Water Molecules in Solution and Pure Liquid As Obtained from Simulationsa pure water 2.75 M proline 2.75 M proline, inc. polar groups
-45 °C
-15 °C
+15 °C
+27 °C
0.92 0.68 0.86
0.89 0.68 0.84
0.85 0.68 0.83
0.84 0.68 0.83
a For the purpose of this ratio, a water molecule was deemed to be 4-fold coordinated when it had at least four bonding partners within a cutoff of 3.3 Å.
Figure 9. Radial distribution functions for water oxygen contacts in solution (main panel) and pure water (inset) at -45, -15, and 15 °C. In the pure liquid a distinct sharpening of the peaks is observed with decreasing temperature. This behavior is absent in the solution where we find very modest structural changes for the lowest temperature, and perhaps more surprisingly, a decrease in structure for the intermediate temperature. The results suggest that proline suppresses the normal lowtemperature evolution of water structure. The low-temperature structure therefore corresponds to a higher effective temperature.
these two concentrations should be similar is not immediately obvious. One might, however, argue that long-range correlations in water are necessary for the unique decrease in sound speed on cooling. The point at which the concentration fluctuations are maximal may correspond to the point at which long-range correlations begin to be disrupted by proline. The room temperature Raman data, however, are sensitive only to shortrange correlations and indicate that the presence of the proline either maintains the local OH bonding connectivity in water or that proline-water hydrogen-bonding substitutes for waterwater hydrogen bonding without any obvious spectral change. By contrast, the range of concentration over which RLP depends strongly on temperature corresponds to the range where the Brillouin line width increases on addition of proline and
further increases on cooling. We assume therefore that the increase in LP ratio in this concentration regime is primarily due to structural relaxation contributions to the central line. In binary mixtures these slowing processes are likely to lead to very slowly relaxing concentration fluctuations at low temperature until they effectively appear as static inhomogeneities leading to very large values of RLP at low temperatures in the concentrated regime. Finally, in Figure 5 we show the variation in the OH Raman bands on cooling a concentrated proline mixture to temperatures well below the Brillouin line width maximum. Whereas we previously noted no discernible perturbation to these bands, here we find a continuous and large red-shift on cooling of approximately 140 cm-1. These shifts are consistent with the behavior of water/ices at low temperature in which the strength and population of tetrahedrally coordinated H-bonded molecules increases as the number of under coordinated molecules falls. However, there is no spectroscopic evidence for the formation of crystalline ices in the mixtures. The inset of Figure 5 shows the low-frequency Raman spectrum over the range 0 to 200 cm-1 showing a characteristic glassy Boson peak. There is no discernible signal from lattice phonons in this range. The only clear peaks are assignable to low-frequency intramolecular proline vibrations. These results again point to one of two possible structural origins: (1) that the water-water correlations are maintained and only the relative populations of associated to nonassociated molecules change or (2) that water-proline hydrogen bonding leads to spectroscopic signatures in this spectral region that are very similar to that observed in pure ices.
4296 J. Phys. Chem. B, Vol. 112, No. 14, 2008
Troitzsch et al.
Figure 11. Trajectory snapshot from molecular dynamics simulations of aqueous proline (water not shown) at a concentration of 1:20 molar ratio. The slices are made up of the full x-y plane and a fixed 15 Å clipping in the z-direction, corresponding to a quarter of the box edge. The two panels for each temperature are taken at times separated by at least 15 fs and no more than 25 fs. Evident on this time scale are strong concentration fluctuations which are assumed to be the molecular origin of the behavior of the Landau-Placzek ratio at this concentration.
Figure 12. Close-up of the MD trajectory showing a transient hydrogen-bonded hexamer in the solution. While these structures are observed, they are not long-lived enough, even on the scale of our simulations, to be deemed a deciding factor in the water structure.
We now turn to the results of the molecular dynamics simulations focusing on structure: Shown in Figure 9 is the radial distribution function between water oxygen atoms in the proline solution and, inset, pure water at three temperatures. A qualitatively unusual feature is evident: It is that the temperature evolution of the water structure in proline solution at this concentration is not as expected for pure water. In the pure water case, one expects the O-O correlations to become sharper as the local structure evolves toward more well-developed tetrahedral coordination. The peaks in the distribution should therefore narrow and the minima should become deeper as hydration shells become better defined. This is illustrated for simulated systems of pure water in Figure 9. This behavior is
usually attributed to the increase in tetrahedral alignment and 4-fold coordination of water molecules upon cooling, with the eventual formation of ice as a result. For comparison, we therefore quote the ratio of at least 4-fold coordinated water molecules to the total number of water molecules as found in the simulations in Table 1. What appears to occur in the simulated proline solutions is that the OW-OW correlations show exceptionally weak dependencies on temperature. In fact, the evolution appears to be weakly non-monotonic in the sense that on quenching from 15 to -15 °C, the OWOW correlations appear slightly understructured relative to the higher temperature system, thus displaying an inverse behavior to the pure water. Upon further quenching to -45 °C, the water correlations appear to reacquire structure to the extent that the corresponding radial distribution function is almost indistinguishable from those in the solution at ambient conditions. It is, however, very different from that of pure water under the same conditions. To support this, we present statistics for the ratio of 4-fold coordinated molecules to total number of water molecules in the solution in Table 1 as well. On the basis of this evidence, it is tempting to speculate that water structure in this proline solution appears to correspond to an effective temperature that is considerably higher than the physical temperature and that its normal temperature dependence is strongly suppressed by the presence of proline. Clearly, such a property does provide a molecular basis for the apparent biological function of proline as a cryoprotectant. However, care must be taken to reconcile this suggestion to the experimental data: The Rayleigh-Brillouin spectra are completely (but only indirectly) consistent with such a scenarios indicating only that proline induces glassy dynamics at the point where concentration fluctuations appear to be highest. Beyond
Cryoprotection in Aqueous Proline that, no specific structural conclusions can be drawn. To link the Raman evidence with the computer simulations is more problematic. On one hand, the temperature evolution of the Raman data may be taken as evidence that the water is strongly segregated from proline and therefore the Raman spectra are consistent with the transition to ice-like structures (albeit without sufficient long-range order to give rise to discernible lattice phonons). According to this interpretation of the experimental data, one would expect the water structure to behave very similarly to pure water and to show typical structural trends leading to enhanced correlations at low temperature. This is clearly not the scenario implied in the simulated system. However, the Raman results are open to a somewhat more subtle interpretation in which the water-proline hydrogen bonding is significant and upon cooling becomes stronger leading to the observed red-shifts in the OH spectrum. The water that remains unbonded to proline in the real system would then contribute only the spectral signature characteristic of normal water and the observed spectral changes would be dominated by waterproline hydrogen bonding. We offer a snapshot in Figure 10 from the molecular dynamics trajectories to illustrate an example of the hydrogen-bonding motifs present in the low-temperature solution. Here, the four potential water H-bonds are satisfied by a combination of water-water and water-proline contacts. To support this, we show the fraction of at least 4-fold coordinated water molecules in Table 1, this time including the polar groups of the proline molecules as potential H-bonding acceptors. It is apparent that under ambient conditions the coordination of water molecules is the same under this definition as in the case of pure water. The trend upon cooling is therefore particularly interesting, as a change can indeed be observed. Based on these ratios, we find the suggestion of a high effective temperature is plausible. We also show trajectory snapshots at larger length scales illustrating concentration fluctuations in the simulated system (Figure 11) and the hydrogen-bonding patterns in a transient but long oligomer (Figure 12) in solution. These figures illustrate the molecular-scale heterogeneity; in Figure 11 the fluctuations in density on a time-scale of the order of 20 fs are evident, which are consistent with the increase in the Landau-Placzek ratio. While MD probes different length scales than the ones relevant for light scattering, it is clear that the solutions are not uniformly mixed. One motif of association is displayed in Figure 12, indicating a short-lived chain-like association of proline molecules. 4. Conclusion The molecular mechanism by which proline acts as a cryoprotectant appears to be due to the fact that proline inhibits the normal structural evolution of water upon cooling and preserves the ambient structure even at very low temperatures. Clear spectroscopic signatures of viscoelastic behavior are observed and evidence for a fast relaxation mode (τ ≈ 15 ps)
J. Phys. Chem. B, Vol. 112, No. 14, 2008 4297 in the glass is found from Rayleigh-Brillouin scattering data. Below 0.07 mole fraction bulk water behavior is observed. A complex dependence of the Landau-Placzek ratio on composition and temperature is attributed to separable contributions from concentration fluctuations in the more dilute solutions and to structural relaxation in more concentrated solutions. The contribution to the Landau-Placzek ratio from concentration fluctuations appears to be largest at the concentration where the temperature dependence of the Brillouin frequencies ceases to be dominated by water and the viscoelastic character emerges. References and Notes (1) Hare, P.; Cress, W.; van Staden, J. J. Exp. Bot. 1999, 50 (333), 413-434. (2) Paquin, R.; Pelletier, G. Physiol. Veg. 1981, 19, 103. (3) Koster, K.; Lynch, D. Plant Physiol. 1992, 98, 108. (4) Nanjo, T. M.; Kobayashi, Y.; Yoshiba, Y.; Kakubarri, K.; Yamaguchi-Shinozak, K.; Shinozaki, K. FEBS Lett. 1999, 461, 205. (5) Yoshiba, Y.; Kiyosue, T.; Kakashima, K.; Shinozake, K.; Shinozake, K. Plant Cell Physiol. 1997, 38, 1095. (6) Schobert, B.; Tschesche, H. Biochim. Biophys. Acta 1978, 541, 270. (7) Rudolph, A.; Crowe, J. Biophys. J. 1986, 50, 423. (8) Samuel, D.; Krishnaswamy, T.; Kumar, S.; Ganesh, G.; Jayaraman, G.; Yang, P.; Chang, M.; Trivedi, V.; Wang, S.; Hwang, K.; Chang, D.; Yu, C. Protein Sci. 2000, 9, 344. (9) Troitzsch, R.; Martyna, G.; McLain, S.; Soper, A.; Crain, J. J. Phys. Chem. B 2007, 111 (28), 8210-8222. (10) Cummins, H.; Gammon, R. J. Chem. Phys. 1966, 44 (7), 27852796. (11) Ma, R.; He, T.; Wang, C. J. Chem. Phys. 2003, 88, 1497. (12) Amo, Y.; Tominaga, Y. Chem. Phys. Lett. 2000, 332, 521. (13) van der Elsken, J.; Bot, A. J. Appl. Phys. 1989, 66, 2118. (14) Bohidar, H. J. Appl. Phys. 1988, 64, 3766. (15) Bot, R.; Wedgham, G. Fluid Phase Equilib. 1992, 77, 285. (16) Miller, G.; Lee, C. J. Phys. Chem. 1968, 72 (13), 4644-4650. (17) Patkowski, A.; Fischer, E. W.; Steffen, W.; Glaser, H.; Baumann, M.; Ruths, T.; Meier, G. Phys. ReV. E 2001, 63, 061503. (18) Feig, M.; Mackerell, A., Jr.; Brooks, C. L., III J. Phys. Chem. B 2003, 117 (12), 2831-2836. (19) MacKerell, A.; Bashford, D.; Bellott, M.; Dunbrack, R.; Evanseck, J.; Field, M.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.; Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F.; Mattos, C.; Michnick, S.; Ngo, T.; Nguyen, D.; Prodhom, B.; Reiher, W.; Roux, B.; Schlenkrich, M.; Smith, J.; Stote, R.; Straub, J.; Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. J. Phys. Chem. B 1998, 102, 3586. (20) Hernandes, M.; da Silva, J.; Longo, R. J. Comput. Chem. 2003, 24, 973-981. (21) Martyna, G.; Klein, M.; Tuckerman, M. J. Chem. Phys. 1992, 97, 2635. (22) Martyna, G. J.; Tuckermann, M. E.; Tobias, D. J.; Klein, M. L. Mol. Phys. 1996, 87, 1117. (23) Martyna, G.; Tobias, D.; Klein, M. J. Chem. Phys. 1994, 101, 4177. (24) Ryckaert, J.; Ciccotti, G.; Berendsen, H. J. Comput. Phys. 1977, 23, 327-341. (25) Andersen, H. J. Comput. Phys. 1983, 52, 24. (26) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.; Pedersen, L. G. J. Chem. Phys. 1995, 103, 8577. (27) Kawamoto, T.; Ochiai, S.; Kagi, H. J. Chem. Phys. 2003, 120, 5867. (28) Magazu, S.; Maisano, G.; Majolino, D.; Mallamace, F.; Migliardo, P.; Alliota, F.; Vasi, C. J. Phys. Chem. 1989, 93, 942. (29) Ko, J.; Kojima, S. J. Non Cryst. Solids 2002, 307, 154. (30) Vass, H.; Edington, D.; Crain, J. J. Chem. Phys. 2003, 118, 11066. (31) Ike, Y.; Kojima, S. J. Mol. Struct. 2005, 744, 521. (32) Dixit, S.; Crain, J.; Poon, W.; Finney, J.; Soper, A. Nature 2002, 416, 829.
