Brought to you by Portland Press Ltd.
Published on behalf of the International Federation for Cell Biology
Cancer Cell death Cell cycle Cytoskeleton Exo/endocytosis Differentiation Division Organelles Signalling Stem cells Trafficking
Cell Biology International (2006) 30, 66–73 (Printed in Great Britain)
An NMR method to characterize multiple water compartments on mammalian collagen
Gary D. Fullertona*, Elena Nesa, Maxwell Amuraoa, Andres Rahala, Lada Krasnosselskaiaa and Ivan Cameronb
aRadiology Department, University of Texas HSCSA, Floyd Curl Drive, San Antonio, TX 78229-3900, United States
bCellular and Structural Biology Department, University of Texas HSCSA, Floyd Curl Drive, San Antonio, TX 78229-3900, United States


A molecular model is proposed to explain water 1H NMR spin-lattice relaxation at different levels of hydration (NMR titration method) on collagen. A fast proton exchange model is used to identify and characterize protein hydration compartments at three distinct Gibbs free energy levels. The NMR titration method reveals a spectrum of water motions with three well-separated peaks in addition to bulk water that can be uniquely characterized by sequential dehydration. Categorical changes in water motion occur at critical hydration levels h (gwater/gcollagen) defined by integral multiples N=1, 4 and 24 times the fundamental hydration value of one water bridge per every three amino acid residues as originally proposed by Ramachandran in 1968. Changes occur at (1) the Ramachandran single water bridge between a positive amide and negative carbonyl group at h1=0.0658g/g, (2) the Berendsen single water chain per cleft at h2=0.264g/g, and (3) full monolayer coverage with six water chains per cleft level at h3=1.584g/g. The NMR titration method is verified by comparison of measured NMR relaxation compartments with molecular hydration compartments predicted from models of collagen structure. NMR titration studies of globular proteins using the hydration model may provide unique insight into the critical contributions of hydration to protein folding.

Keywords: Collagen, Protein hydration, NMR relaxometry, MRI, Proton relaxation, Tendon, Hydration compartments.

*Corresponding author.

1 Introduction

The scientific community generally accepts that protein hydration and protein induced water structuring are fundamental sources of protein structural behaviors that support life functions. Collagen is frequently selected as the model protein to elucidate structural and hydration relationships (Leikin et al., 1994, 1995, 1997, 2002; Privalov, 1982). Tendon has high concentration of Type I collagen approaching 100% of the dry biomass in some instances. Molecular tropocollagen crystallizes spontaneously and as a result significant new structural information is available with ever improving accuracy from X-ray diffraction studies of both native and molecular analogues of collagen (Bella and Berman, 1996; Bella et al., 1994, 1995, 1996; Berisio et al., 2002; Fraser et al., 1979; Kramer et al., 1998, 1999, 2000, 2001; Miller and Scheraga, 1976; Okuyama et al., 1977; Ramachandran, 1967; Rich and Crick, 1961; Yonath and Traub, 1969). The more recent, high-resolution studies consistently show a water bridge network surrounding the collagen molecule.

Structural studies provide a molecular model of collagen supporting the existence of three distinct categories of water bridges. The most tightly bound consists of one highly immobilized “water bridge” per every three protein residues hb=18Da/(3×91.2Da)=0.0658gwater/gprotein as originally proposed by Ramachandran (1967), Fullerton and Amurao (2005), and Ramachandran and Ramakrishnan (1968). A second less immobilized water fraction consists of three additional cleft waters per tripeptide unit residing in the three groove-like depressions between the peptide chains of the triple helix hc=3×hb=0.197gwater/g. The water bridge and three cleft waters complete a chain of four water molecules per tripeptide. These chains form a triple helix of water in the three clefts of the α-protein chains of the collagen triple helix. Cleft water was proposed by Berendsen (1962) and discussed by others (Lim, 1981; Privalov, 1982). Recently our laboratory showed that the remaining water on native collagen (&007E;1.62gwater/gprotein) is in the first interfacial monolayer him=5×0.263=1.315gwater/gprotein at higher energy relative to the first two compartments but still lower than the free energy of bulk water (Fullerton and Amurao, 2005). As a result of these three water compartments we predict NMR detectable changes in water motion at h1=hb=0.0658, h2=hb+hc=0.264 and h3=hb+hc+him=1.58gwater/gprotein or at hi=Ni×0.0658 for Ni=1, 4 and 24.

This study tests the hypothesis that that the NMR spin-lattice relaxation time of water on native collagen is not characteristic of any of the three molecular compartments but a weighted average due to fast exchange between three hydration fractions or water phases on the protein surface (Zimmerman and Brittin, 1957). The relaxation characteristics of all three water phases can, however, be extracted and characterized by sequential NMR measurements using a titration of water content (Fullerton et al., 1982, 1986). Tendon is uniquely suited to identifying the molecular origin of the three hydration fractions that may be typical of proteins in general. The collagen molecule is highly immobilized and aligned in protein fiber structure that suppresses proton NMR signal from the protein due to static dipole coupling. The study confirms that proton NMR signal emanates only from the liquid water that shows the presence of three distinct hydration fractions in fast exchange equilibrium. The compartments extracted using the NMR titration method quantitatively confirm the three molecular water bridge environments identified in the preceding paragraph. The correlation of protein NMR titration measurements with a molecular model of water bridges holds promise as a general model to evaluate hydration of both fibrillar and globular proteins.

2 Materials and methods

2.1 NMR titration measurements

Samples of bovine flexor tendon from three animals (age and sex unknown) were obtained from a local slaughter house, dissected on site, diced to 3mm3 cubes and placed in separate pre-weighed 1″ diameter NMR sample tubes for each animal. Tubes were immediately sealed to prevent dehydration. The initial sample volume was approximately 6cc when wet and approximately 3cc when dry. Preliminary measurements of T1 relaxation reported here were made at 10.7MHz on a Praxis II pulsed NMR device with a primary magnetic field strength of 0.25T. The initial research plan was to use these less accurate but rapid measurements at 10.7MHz to provide pilot data for designing of more accurate and complete experiments as a function of temperature at 500, 700 and 900MHz. Results at 0.25T are, however, sufficient to substantiate the hypothesis and are therefore presented as stand alone experiments. Praxis measurements of T1 use the saturation recovery (90-τ-90) pulse sequence with a repetition time TR exceeding five times the longest T1 relaxation time for the sample. The 90° pulse is approximately 13μs (bandwidth in excess of 100kHz). The sample chamber was equilibrated in the instrument at 22±1°C for at least 1h prior to each measurement. Samples were out of the vacuum for approximately 1–4h. Typical measurement times in the NMR unit were approximately 5–10min. Random orientation of the tendon with respect to the magnetic field was used to eliminate possibility of residual orientation dependence. The signal after a recovery delay τ was measured as an amplitude of the free induction decay, A(τ), with the first time point at 26μs. Each data point samples signal for 1μs and 10 points following the peak signal were averaged to reduce the measurement noise.

Each relaxation measurement consisted of 31 amplitude measurements with a total of 30 sequential delay times with even increments of &007E;20ms each for a maximum delay time of 30×Δτ as shown in Fig. 1. Although delay times were adjusted to evaluate signal amplitude variation over as much as two orders of magnitude only single exponent T1 decay was observed in all instances. Measurements were repeated three times to improve the signal-to-noise ratio to 300:1 (0.3% noise) at the initial hydration level. A non-linear regression fit to the data was calculated as shown with a single exponent decay using two fitting constants, A(∞) and T1, using the equation A(∞)A(τ)=A(∞)(eτ/T1) and non-linear regression from the Graph Pad program (Motulsky, 2003). The T1 was measured for the tendon samples over the entire range of hydration from native hydration to completely dry protein using the procedure described below. The tissue mass Mt was measured at each hydration level to allow calculation of h=Mw/Mp following measurement of the dry protein mass Mp (see next paragraph). The mass of salt and other cosolutes was neglected for these calculations.

Fig. 1

T1 analysis of a saturation recovery curve for bovine tendon at the native hydration level is shown here. Solid circles stand for data points. Solid line is a fit to the function ln(A()A(τ))=aτ, where τ is a delay time, A(τ)i is an FID amplitude after the delay τ and A()=A(τT1) and a is the regression fit constant , , R2=0.9992, signal-to-noise ratio was never lower than 20:1.

The original sample weights were approximately 5.4g of tendon which when fully dehydrated decreased to approximately 2.3g of dry mass. Initial dehydrations were achieved by placing the open NMR tubes in a vacuum chamber at approximately 0.25atmospheric pressure. Samples were removed at intervals ranging initially from approximately 1h to several days to achieve significant reduction in the sample water content. After achieving equilibrium mass at room temperature the temperature of the vacuum oven was increased to 90°C and the dehydration extended for another 30 days with intermittent measurements of T1 taken until a new equilibrium and final dry mass for the protein were achieved. The dehydration and T1 measurement protocol for the tendon extended over a total period of approximately 3 months. An extended drying period was undertaken to assure hydration equilibrium over the entire volume of the tendon. Due to the slow dehydration rate no measurable dependence on the time interval between removal from the vacuum oven and NMR measurement was detectable between 1 and 24h (data not shown). The measurement process was repeated for all three samples. Results for all three samples were similar but only two are reported as one sample tube broke during the final heating phase of the experiment.

2.2 NMR titration analysis of spin-lattice relaxation data

The water bridge hydration hypothesis identifies three hydration water fractions in addition to a bulk water phase as shown in Fig. 2. For the protein in dilute solution the hydration fractions are assumed to be in fast exchange equilibrium on T1 relaxation scale such that the spin-lattice relaxation rate iswhere fi is the fraction of water and 1/T1i is the spin-lattice relaxation rate of the ith water fraction. There are three water fast exchange zones (Zones 1, 2 and 3) and one protein fast exchange zone (Zone 4) to express relaxation rate as a function of inverse hydration 1/h over the entire range h=0 (dry protein) to infinity (infinitely dilute protein solution). The definition of the hydration zones is summarized as follows:

Zone 1 (linear): all hydration compartments full and in fast exchange with a variable bulk water compartment;

Fig. 2

(a) Conceptual cartoon for Zone 1 includes bulk water and shows the Gibbs free energy relationships of water molecules in different water bridge bonding arrangements on the collagen molecule. At low hydration levels water molecules accumulate in the “Water Bridges” at the lowest free energy level. Higher energy compartments are filled at higher hydrations by sequential filling of the “Cleft Water” and “Interfacial Monolayer Water until water spills over into the “Bulk” water compartment. (b) There is fast exchange of protons between these compartments when the tissue contains sufficient water to fill all the hydration compartments and maintains a bulk water fraction as well. As protons sample all compartments during decay the NMR spin-lattice decay is monoexponential.

Zone 2 (linear): bulk water depleted leaving cleft water and water bridge compartments full and in fast exchange with a variable interface monolayer compartment;

Zone 3 (linear): bulk water and interface monolayer compartments depleted, water bridge compartment full and in fast exchange with a variable cleft water compartment; and

Zone 4 (non-linear): bulk, interface monolayer and cleft water compartments depleted leaving tripeptide segments on the protein either occupied or not occupied (with or without water bridges). Each tropocollagen has many water bridge sites (&007E;1000) to yield mean relaxation rate by spin-diffusion during transition from full bridges to completely dry protein relaxation rates.

The dilute protein case is shown in Figs. 2 and 3 as Zone 1 where 1/h1/h3. The water consists of two fractions, hydration water h3=hb+hc+him=1.58gwater/gprotein and bulk water of mass Mo. The total hydration water mass M3=h3×Mp allows calculation of the fraction of water f3=h3×Mp/Mw that interacts directly with the protein surface while the bulk water fraction fo=Mo/Mw has the motional properties of pure bulk water. The sum of all fractions f3+fo=1. The relaxation rate (R=1/T1) of water is the weighted sum of the relaxation rates for bulk water, Ro, and the mean relaxation rate for hydration water R3 where h=Mw/Mp=the measured hydration of the sample.

Fig. 3

“Zonal Analysis” of relaxation rate versus the reciprocal of hydration 1/h. A multi-segment non-linear least squares regression fit (R2=0.9916) for n=74 measurements over the entire range of hydration from native to completely dry protein residue demonstrates the ability of the Molecular Model to explain relaxation behavior. (A) Shows entire range of experiments. (B) Focuses on data from Zone 3. (C) Focuses on data from Zone 2. Each zone is identified as a range of hydrations by the molecular model as follows: Zone 1: h>1.584gwater/h dry mass, Zone 2: 1.584g/g>h>0.264g/g, Zone 3: 0.264>h>0.0658g/g and Zone 4: h<0.0658g/g. Extrapolation of Zone 2 data to Mp/Mw=0 shows that the relaxation rate for bulk water Ro=0.37s−1 is significantly different p<0.05 than Rim=1.354 for surface water (see Table 2). We note that although Zone 2 is well described by the linear relationship used in the present analysis. There are possible 2nd order regularities near Mp/Mw=1.3 and 3.0gsolids/gwater. This could reflect small changes in water organization in the interfacial monolayer fraction as neighboring water molecules are removed. Note the lack of data in Zone 1 for native tendon which indicates absence of bulk-like water on this tissue (see text for details).

The monolayer hydration h3=1.58gwater/gdry protein is known from the molecular model while mass of the tissue Mt, dry mass of the protein Mp and relaxation rate of the tissue R1(h) are measured directly. The mass of water Mw=MtMp is calculated. Linear regression fits allow extraction of the fitting constants Ro, the relaxation rate of the outer water fraction (bulk water for Zone 1) and the mean relaxation rate R3 of the hydrated protein. The same method applied to Zone 2 gives the relaxation rate of the interfacial monolayer water Rim and Zone 3 analysis gives the relaxation rate for cleft water Rc.

Zone 4 differs from the other zones because it describes the transition from collagen with water bridges to collagen without bridges. There is a broad spectrum of NMR signal from protons (hydrogen nuclei) covalently bound in the protein. A fraction of this protein signal appears in the water resonance range. In Zone 4 the proton signal comes from solid-like protein in two possible states either with or without water bridges. The observed change in relaxation rate indicates that protein with water relaxes slightly more rapidly than collagen without water bridges. The water bridges accelerate relaxation for protein without structural water. This is likely through the well known mechanism of spin-diffusion in solids (Edzes and Samulski, 1977). Thus there are two protein categories: category A with water bridges and category B without water. As the signal from “solid like” water is much smaller than from protein we can neglect the signal amplitude due to the water and calculate the relaxation rate as follows:

RA=the relaxation rate of protein hydrogen nuclei with water bridges

RB=the relaxation rate of the protein hydrogen nuclei without water bridges

fA=the fraction of the protein having one bridge per three residues

fB=the fraction of protein without a water bridge per three residues

h=the hydration=Mw/Mp

hb=the single bridge hydration fraction


This last equation R1 is a linear function of h while the other three equations are linear functions of 1/h. We plot all zones as a function of 1/h as this gives three linear sections and one curved line for Zone 4. The four equations must agree at the intersections of adjacent sections. This allows simplification to reduce the number of fitting constants to five to yield the relationships summarized in Table 1.

Table 1.

Multi-segment or zonal fast exchange equations for spin-lattice relaxation for water on collagen

ZoneHydration constraintRelaxation rate (s−1)
1h > 24hb1/T1 = Ro + (24hb/h) (5/6Rim + 1/8Rc + 1/24Rbp − Ro)
224hb ≥ h ≥ 4hb1/T1 = Rim + (4hb/h) (3/4Rc + 1/4Rbp − Rim)
34hb ≥ h ≥ hb1/T1 = Rc + hb/h (Rbp − Rc)
4hb ≥ h1/T1 = Rpr + h/hb (Rbp − Rpr)

2.3 Statistical analysis

The 1/T1 relaxation rate as a function of 1/h was evaluated using non-linear least squares regression fit to the four zone multi-segment equations defined in Table 3 to evaluate the ability of the proposed molecular model to describe relaxation results over the entire hydration range (Motulsky, 2003).

3 Results

The results from samples S1 are shown in Fig. 3. Results on S2 were similar but not shown for brevity.

The details of the least squares fits are summarized in Table 2 for both the tendon S1 and S2 measurement series. The goodness of fit for tendon S1 was R2=0.9916 and R2=0.9910 for tendon S2. The best fit value, standard error and 95% confidence interval for Rim, Rc, Rbp and Rpr are presented as well as the direct measure of the spin-lattice relaxation rate for water which was measured independently. The correlation times calculated from the mean relaxation rates are summarized in Table 3.

Table 2.

Fitting constants for statistical fits to NMR titration studies

Table 3.

Correlation times for water calculated from data in Table 2 using BPP theory (Bloembergen et al., 1948)

Bulk water1/T1 = 0.347 s−1τ = 6.7 × 10−12 (Reference)
Orientational water1/T1 = 1.352 s−1τ = 2.6 × 10−11 (Bulk × 4)
Cleft water1/T1 = 33.1 s−1τ = 6.55 × 10−10 (Bulk × 100)

Fitting constantBest fitStandard error95% Confidence interval
Rb (s−1)0.347Measured separatelyNot visible on collagen
Rim (s−1) – summer 11.3540.19790.9588–1.749
Rim (s−1) – summer 21.3490.21020.9292–1.770
Rc (s−1) – summer 132.840.314232.21–33.47
Rc (s−1) – summer 233.350.386932.58–34.12
Rbp (s−1) – summer 116.060.309915.44–16.67
Rbp (s−1) – summer 214.010.411913.18–14.83
Rpr (s−1) – summer 14.8420.40944.025–5.660
Rpr (s−1) – summer 27.0520.41276.227–7.877

4 Discussion

4.1 NMR titration method

Comparison of the NMR titration measurements shown in Fig. 3 (data points) with predictions of the four zone multi-component molecular model of collagen hydration (solid line) calculated with parameters from Table 1 demonstrates the capacity of fast exchange relationships to accurately describe the relaxation rate of water on collagen. The model predictions of sharp demarcations at h1=hb=0.0658, h2=hb+hc=0.264 and h3=hb+hc+him=1.58gwater/gprotein or at hi=Ni×0.0658 for Ni=1, 4 and 24 is confirmed. Thus all of the hydration fractions for collagen can be described as integer multiples of h1=0.0658 as proposed by Ramachandran and shown in Fig. 4. We refer to this fundamental quantity as the Ramachandran hydration value. It can be calculated for any protein for which molecular weight A and number of residues n are known using the equation as demonstrated for lysozyme below.

Fig. 4

The Ramachandran proposal of one direct hydrogen bond per tripeptide, one water bridge and a second non-charge bridge water molecule was necessary to complete the two-bond model proposed by his group. (a) Shows configuration when peptide three is other than hydroxyproline. (b) Shows the alternative tripeptide hydrogen bonding configuration when hydroxyproline occupies position three in Chain A. In both instances hRa=18Da/(3×mean peptide molecular weight)=0.0658g/g for collagen. Reproduced with permission (Ramachandran and Chandrasekharan, 1968).

4.2 Water bridges

As shown in Fig. 5 the function of the water bridge and direct hydrogen bond is to serve as mechanisms to reduce the electrostatic energy of the protein by transferring positive charge from amide sites to compensate negative charge from carbonyl groups. Direct hydrogen bonds occur when the separation is less than 3Å while water bridges are necessary when separations approach 5Å sufficient to allow insertion of a water molecule. Separations are controlled by stereotactic restriction when the protein is folded. Hydrogen exchange experiments (Yee et al., 1974) showed extremely long exchange times (hours to days) for amide hydrogen atoms participating in either direct or water bridge bonds to confirm the two-bond model proposed by Ramachandran. The NMR titration measurements extend this conclusion by confirming as proposed that a water bridge provides the second bond per tripeptide.

Fig. 5

Molecular water serves as a dielectric and thereby reduces the in vacuo electrostatic energy of the collagen molecule by aligning with the electric field generated between positive amide and negative carbonyl charge sites on the neighboring α-protein chains. Water bridges are highly immobilized such that the relaxation rates of protein with water bridge Rbp approach that of the solid protein Rpr without water as shown in Table 2.

4.3 Cleft water

We define cleft water as the additional three water molecules per tripeptide that bind in a hydrogen bonded network to the protein as well as to the structural water bridge (also located in the cleft) to complete the four-water molecule chain per every three peptide residues in the molecular grooves of the collagen triple helix proposed as by Berendsen (1962) and Fullerton and Amurao (2005). We reason on the basis of geometry that cleft waters participate in double water bridges between the remaining main chain amide group and neighboring carbonyl groups; recall one of the three amides participates in a direct bond (Rich and Crick, 1955), a second in a single water bridge (Ramachandran and Chandrasekharan, 1968) and only one remains. As discussed by Westerhof (1993) a single water binds to the positive amide but branches into double water bridges to bind with three water molecules to two negative carbonyl groups when separations are too great for single water bridges. Thus direct bonds (shortest separation), structural waters (intermediate separation) and cleft water (greatest separation) use four waters to fill all the available hydrogen bonding sites on the collagen protein main chain. All remaining water bridges originate from less confined side chains such as the positive hydroxyl group on hydroxyproline.

As shown in Table 2 the Rc relaxation rates for cleft water on samples 1 and 2 are not significantly different. The correlation time calculated for the mean relaxation rate given in Table 3 shows that cleft water tumbles at a rate 100 times slower than bulk water. Slowed tumbling is attributed to two factors: (1) the water molecule are participating in double water bridges between sites separated by distances &007E;6.5Å and (2) the waters are mechanically compressed between protein surfaces (Jeffrey and Saenger, 1991; Westerhof, 1993). The intramolecular double water bridges form cleft water. Cleft water is the most rapidly relaxing water fraction and thereby serves as the relaxation sink for the entire water/protein system through the mechanism of fast proton exchange. The motion of water in cleft water is sufficiently slowed that water tumbling resonates with the proton resonant frequency. As protons exchange positions rapidly and relaxation in the cleft fraction is so rapid, most protons undergo spin-lattice relaxation while in the cleft compartment. Thus the double water bridge cleft water is the relaxation sink for the entire water/protein proton ensemble. Cleft water hydration hc is three times the Ramachandran hydration but the peak in the observed spin-lattice relaxation occurs at h2=hb+hc=0.264gwater/gprotein. This for the first time identifies the source of short T1 relaxation in tissue.

4.4 Interfacial monolayer water

As seen in Fig. 3 and Tables 2 and 3 the relaxation times of the outer most water compartment on the two native tendon samples measured in Zone 2 do not differ significantly. However, the mean Rim=1.352s−1 differs significantly from Ro=0.347s−1 measured separately for bulk water at the same temperature. Hydration force measurements (Leikin et al., 1994) show that large mechanical or osmotic pressures are necessary to provide the work to remove interfacial water in the hydration range h=0.26–1.1gwater/gcollagen. In a related paper in this issue we show that water in this hydration range is contained in the first water monolayer on collagen (Fullerton and Amurao, 2005). Monolayer coverage on neighboring molecule provides double layer separation of approximately 6.3Å between neighboring collagen surfaces. The separation accommodates a rapidly shifting hydrogen bond network (Bella et al., 1994, 1995) consisting of intermolecular double and triple water bridges from positive charge sites on the peptide side chains to nearby negative charge sites. We refer this water as interfacial monolayer water and exclude from it the more confined cleft and structural water fractions that are strictly intramolecular and associated with the protein main chain. Thus hydration force, NMR titration and molecular structure calculations all agree with the assertion that interfacial monolayer water is in a lower free energy state than bulk water and there is little if any bulk water present on native mammalian tendon.

The maximum extent of interfacial monolayer water is difficult to determine accurately from the measurements in Fig. 3. Prior experiments on other protein solutions (Hallenga and Koenig, 1976) show that under dilute conditions of Zone 1 the straight line extrapolates to the relaxation rate measured for bulk water or Ro=0.347s−1 on our instrument. The fact that the data in Zone 2 extrapolate to a number significantly different therefore implies that there is a break in the line somewhere between the lowest value of Mp/Mw and zero which is consistent with the dotted line at h=1.58 predicted by the model. Experiments using tendon expansion measurements and relation to molecular collagen separation models also reported in this volume indicate that monolayer hydration h=1.62±0.16 SDg/g (Fullerton and Amurao, 2005).

4.5 Protein relaxation

The protein relaxation rates for S1 and S2 are significantly different even though the relaxations of the water fractions are to all purposes identical. This difference will require further investigation but is likely due to difference in collagen crosslinking due to life-time stress and other animal specific biomolecular factors.

4.6 Application of NMR titration to lysozyme

The NMR titration method was developed originally as a phenomenological method without theoretical basis in an attempt to understand and predict MR image contrast. Lysozyme was studied as model globular protein system (Fullerton et al., 1986). The re-evaluation of the published results allows testing of the molecular hydration model. The molecular weight of lysozyme is A=13 270 and there are n=129 peptide residues. Thus the mean residue molecular weight is 102.9Da/residue. This is slightly higher than collagen due primarily to the high concentration of glycine in collagen. The Ramachandran hydration constant is calculated for lysozyme hRa=18/(3×102.9)=0.0583g/g. This model prediction compares well with the NMR titration measurement of 0.055g/g and agrees well with more than 10 other measurements of structural water by isopiestic, infrared, heat capacity and a variety of calculation methods quoted in the paper (see Table IV in Fullerton et al., 1986). The molecular model predicts that the peak in the lysozyme titration study should occur at h2=4×hRa=0.233gwater/gprotein. The model value compares well with the peak values measured by titration which ranged from 0.22 to 0.27 with mean 0.253gwater/gprotein. This amount of bridge and cleft water agrees with more than 15 papers giving measures of what was previously called “bound” water using (1) nonfreezing fraction, (2) dielectric measures, (3) sedimentation, (4) compositional calculation, (5) isotherm measurements, (6) isopiestic measurements, (7) BET isotherm, (8) X-ray diffraction as well as (9) NMR titration (see Table V in Fullerton et al., 1986). This implies that cleft hydration hi=0.175g/g is predicted by the collagen based set of protein molecular hydration rules. Finally the extent of interfacial monolayer water on lysozyme extends to at least him=1.4gwater/gprotein and possibly more. All imply large amount of water on internal surfaces of the protein rather than multilayer of structured water originally hypothesized to explain results.

5 Conclusions

The NMR titration method provides measurements to directly relate water NMR relaxation to an underlying molecular model of collagen hydration. Although the description of hydration fractions as integer multiples of the Ramachandran hydration fraction may be merely fortuitous, the regularity of protein's primary and secondary structure provides a geometric foundation consistent with such a stoichiometric hydration relationship. In addition the hydration model encompasses a wide range of difficult to explain empirical observations of protein hydration effects that have eluded coherent molecular explanation for many years.

The NMR spin-lattice relaxation “sink” for collagen is identified as three water molecules per every three peptide forming double water bridges between a single positive amide group and two carbonyl groups on the protein main chain. Comparison with results on lysozyme implies similar intimate association of three water molecules with the main chain of globular proteins in both α-helix and β-pleat configurations. Such a regular stoichiometric association of internal water with native, folded proteins is, however, contrary to the majority view of the scientific community and should be viewed with caution.

The NMR titration method thus allows experimental confirmation of hydration model calculations and predictions of hydration properties for lysozyme and possibly other proteins directly from molecular composition. Collagen hydration rules appear more universal than originally anticipated. The agreement of predicted hydration fractions for collagen and other proteins suggests water chains associated with proteins chains in the α-helix and β-pleat configurations. This was not previously known. Preliminary NMR titration studies of multiple proteins rich in both α-helix and β-pleat in our laboratories universally follow the “Ramachandran” hydration rules developed for collagen. We tentatively propose that globular proteins obey the collagen stoichiometric hydration rules of the collagen triple helix. The topic needs further investigation.


Funding for much of this work came from discretional funds supplied by Malcolm Jones Distinguished Professorship for which I am grateful to Drs. Reuter, Dodd and my other colleagues from the Radiology Department in San Antonio.


Bella J, Berman, HM. Crystallographic evidence for C alpha-H⋯OC hydrogen bonds in a collagen triple helix. J Mol Biol 1996:264:4:734-42
Crossref   Medline   1st Citation  

Bella J, Brodsky, B, Berman, HM. Hydration structure of a collagen peptide. Structure 1995:3:9:893-906
Crossref   Medline   1st Citation   2nd  

Bella J, Brodsky, B, Berman, HM. Disrupted collagen architecture in the crystal structure of a triple-helical peptide with a GlyAla substitution. Connect Tissue Res 1996:35:1–4:401-6
Crossref   Medline   1st Citation  

Bella J, Eaton, M, Brodsky, B, Berman, HM. Crystal and molecular structure of a collagen-like peptide at 1.9Å resolution. Science 1994:266:5182:75-81
Crossref   Medline   1st Citation   2nd  

Berendsen HJ. Nuclear magnetic resonance study of collagen hydration. J Chem Phys 1962:16:12:3297-305
1st Citation   2nd  

Berisio R, Vitagliano, L, Mazzarella, L, Zagari, A. Crystal structure of the collagen triple helix model [(Pro-Pro-Gly)(10)](3). Protein Sci 2002:11:2:262-70
Crossref   Medline   1st Citation  

Bloembergen N, Purcell, EM, Pound, RV. . Phys Rev 1948:73:679
Crossref   1st Citation  

Edzes HT, Samulski, ET. Cross relaxation and spin diffusion in proton NMR of hydrated collagen. Nature 1977:265:521-3
Crossref   Medline   1st Citation  

Fraser RD, MacRae, TP, Suzuki, E. Chain conformation in the collagen molecule. J Mol Biol 1979:129:3:463-81
Crossref   Medline   1st Citation  

Fullerton GD, Amurao, MR. Evidence that collagen and tendon have monolayer water coverage in the native state. Int J Cell Biol 2005:
1st Citation   2nd   3rd   4th   5th  

Fullerton GD, Ord, VA, Cameron, IL. An evaluation of the hydration of lysozyme by an NMR titration method. Biochim Biophys Acta 1986:869:230-46
Medline   1st Citation   2nd   3rd   4th  

Fullerton GD, Potter, JL, Dornbluth, NC. NMR relaxation of protons in tissues and other macromolecular water solutions. Magn Reson Imaging 1982:1:4:209-26
Crossref   Medline   1st Citation  

Hallenga K, Koenig, SH. Protein rotational relaxation as studied by solvent 1H and 2H magnetic relaxation. Biochemistry 1976:15:19:4255-64
Crossref   Medline   1st Citation  

Jeffrey GA, Saenger, W. Hydrogen bonding in biological structures. 1991:
1st Citation  

Kramer RZ, Bella, J, Brodsky, B, Berman, HM. The crystal and molecular structure of a collagen-like peptide with a biologically relevant sequence. J Mol Biol 2001:311:1:131-47
Crossref   Medline   1st Citation  

Kramer RZ, Bella, J, Mayville, P, Brodsky, B, Berman, HM. Sequence dependent conformational variations of collagen triple-helical structure. Nat Struct Biol 1999:6:5:454-7
Crossref   Medline   1st Citation  

Kramer RZ, Venugopal, MG, Bella, J, Mayville, P, Brodsky, B, Berman, HM. Staggered molecular packing in crystals of a collagen-like peptide with a single charged pair. J Mol Biol 2000:301:5:1191-205
Crossref   Medline   1st Citation  

Kramer RZ, Vitagliano, L, Bella, J, Berisio, R, Mazzarella, L, Brodsky, B. X-ray crystallographic determination of a collagen-like peptide with the repeating sequence (Pro-Pro-Gly). J Mol Biol 1998:280:4:623-38
Crossref   Medline   1st Citation  

Leikin S, Parsegian, VA, Yang, W, Walrafen, GE. Raman spectral evidence for hydration forces between collagen triple helices. Proc Natl Acad Sci U S A 1997:94:21:11312-7
Crossref   Medline   1st Citation  

Leikin S, Rau, DC, Parsegian, VA. Direct measurement of forces between self-assembled proteins: temperature-dependent exponential forces between collagen triple helices. Proc Natl Acad Sci U S A 1994:91:1:276-80
Crossref   Medline   1st Citation   2nd  

Leikin S, Rau, DC, Parsegian, VA. Temperature-favoured assembly of collagen is driven by hydrophilic not hydrophobic interactions. Nat Struct Biol 1995:2:3:205-10
Crossref   Medline   1st Citation  

Leikin E, Mertts, MV, Kuznetsova, N, Leikin, S. Type I collagen is thermally unstable at body temperature. Proc Natl Acad Sci U S A 2002:99:3:1314-8
Crossref   Medline   1st Citation  

Lim VI. A novel structural model for collagen: Water–carbonyl helix. FEBS Lett 1981:132:1-5
Crossref   1st Citation  

Miller MH, Scheraga, HA. . J Polymer Sci Symp 1976:54:171-200
1st Citation  

Motulsky H. Prism GraphPad. 2003:
1st Citation   2nd  

Okuyama K, Takayanagi, M, Ashida, T, Kakudo, M. . Polym J 1977:9:341-3
Crossref   1st Citation  

Privalov PL. Stability of proteins. Proteins which do not present a single cooperative system. Adv Protein Chem 1982:35:1-104
Medline   1st Citation   2nd  

Ramachandran GN. . Treatise on collagen 1967:103-83
1st Citation   2nd  

Ramachandran GN, Chandrasekharan, R. Interchain hydrogen bonds via bound water molecules in the collagen triple helix. Biopolymers 1968:6:11:1649-58
Crossref   Medline   1st Citation   2nd  

Ramachandran GN, Ramakrishnan, C. Molecular structure. Biochemistry of collagen 1968:45-84
1st Citation  

Rich A, Crick, F. The structure of collagen. Nature (London) 1955:176:176
1st Citation  

Rich A, Crick, F. The molecular structure of collagen. J Mol Biol 1961:3:483
Medline   1st Citation  

Westerhof E. . Water and biological macromolecules 1993:1-466
1st Citation   2nd  

Yee RY, Englander, SW, Von Hippel, PH. Native collagen has a two-bonded structure. J Mol Biol 1974:83:1:1-16
Crossref   Medline   1st Citation  

Yonath A, Traub, W. Polymers of tripeptides as collagen models. IV. Structure analysis of poly(l-proly-glycyl-l-proline). J Mol Biol 1969:43:3:461-77
Crossref   Medline   1st Citation  

Zimmerman JR, Brittin, WE. Nuclear magnetic resonance in multiple phase systems: life time of a water molecule in an absorbing phase on silica gel. J Phys Chem 1957:6:1328-33
1st Citation  

Received 30 May 2005; accepted 30 September 2005


ISSN Print: 1065-6995
ISSN Electronic: 1095-8355
Published by Portland Press Limited on behalf of the International Federation for Cell Biology (IFCB)