The physical state of water in bacterial sporesErik P. Sundea, Peter Setlowb, Lars Hederstedtc, and Bertil Hallea,1

aDepartment of Biophysical Chemistry, Center for Molecular Protein Science, Lund University, SE-22100 Lund, Sweden; bDepartment of Molecular, Microbial,and Structural Biology, University of Connecticut Health Center, Farmington, CT 06030-3305; and cDepartment of Cell and Organism Biology, LundUniversity, SE-22362 Lund, Sweden

Edited by Richard M. Losick, Harvard University, Cambridge, MA, and approved September 23, 2009 (received for review July 31, 2009)

The bacterial spore, the hardiest known life form, can survive in ametabolically dormant state for many years and can withstand hightemperatures, radiation, and toxic chemicals. The molecular basis ofspore dormancy and resistance is not understood, but the physicalstate of water in the different spore compartments is thought to playa key role. To characterize this water in situ, we recorded the water2H and 17O spin relaxation rates in D2O-exchanged Bacillus subtilisspores over a wide frequency range. The data indicate high watermobility throughout the spore, comparable with binary protein–water systems at similar hydration levels. Even in the dense core, theaverage water rotational correlation time is only 50 ps. Spore dor-mancy therefore cannot be explained by glass-like quenching ofmolecular diffusion but may be linked to dehydration-induced con-formational changes in key enzymes. The data demonstrate that mostspore proteins are rotationally immobilized, which may contribute toheat resistance by preventing heat-denatured proteins from aggre-gating irreversibly. We also find that the water permeability of theinner membrane is at least 2 orders of magnitude lower than formodel membranes, consistent with the reported high degree of lipidimmobilization in this membrane and with its proposed role in sporeresistance to chemicals that damage DNA. The quantitative resultsreported here on water mobility and transport provide importantclues about the mechanism of spore dormancy and resistance, withrelevance to food preservation, disease prevention, and astrobiology.

Bacillus subtilis � hydration � magnetic relaxation dispersion �spore dormancy � spore resistance

When deprived of nutrients, Bacillus and Clostridium bac-teria form endospores that can survive in a metabolically

dormant state for years yet can transform into vegetative cellswithin minutes, triggered by environmental signals. Bacterialspores not only survive starvation, but are also highly resistantto heat, radiation, and toxic chemicals (1, 2). Because of theirexceptional ability to survive adverse conditions for long periods,bacterial spores are major causes of food spoilage and food-borne disease, and they might even serve as vehicles for transferof life between planets (3). Furthermore, Bacillus anthracisspores pose a threat as a potential agent of biological warfare andterrorism. For these reasons, it is important to elucidate themolecular mechanisms of spore dormancy and resistance. Thistask is challenging because it involves several fundamentalbiophysical phenomena which are themselves not fully under-stood, including protein stability and mobility, membrane per-meability, and control of enzymatic activity.

The thermal stability of native protein conformations is largelygoverned by solvent-mediated interactions (4), and enzymaticactivity relies on conformational changes and molecular associa-tions that are strongly coupled to solvent mobility. Both the heatresistance and the metabolic dormancy of spores may therefore belargely controlled by the physical state and chemical composition ofthe solvent. Anhydrobiotic organisms—including plant seeds, yeastcells, and certain plants and invertebrates—achieve dormancy byreplacing most of the cell water by compatible osmolytes, such astrehalose or sucrose, thereby transforming the cytoplasm into ametastable glassy state (5, 6). Although bacterial spores do notaccumulate glass-forming osmolytes, it has been proposed that thespore’s core region is also in a glassy state (7–9). If so, spore

dormancy could simply be attributed to the extreme retardation ofdiffusive molecular motions in the highly viscous glass state. Fur-thermore, heat resistance might then be explained in kinetic ratherthan thermodynamic, terms. An endothermic transition observedby differential scanning calorimetry on Bacillus subtilis spores wasinitially assigned to a glass transition (9), but this interpretation hasbeen questioned (10). On the other hand, at least some corecomponents appear to be immobilized. Dielectric permittivity data(11) and electron paramagnetic resonance spectra (12) suggest thatmost ions in the core are immobilized, 13C NMR spectroscopyindicates that the core’s large depot of pyridine-2,6-dicarboxylicacid (dipicolinic acid, DPA) is in a solid-like state (13), andfluorescence measurements show that green fluorescent protein isat least 4 orders of magnitude less mobile in the dormant core thanin the cytoplasm of the vegetative B. subtilis cell (14). Althoughthese observations are consistent with a glassy core, attempts using1H NMR linewidths to directly measure water mobility in sporeshave not been conclusive (15, 16), partly because of the confound-ing effect of paramagnetic Mn(II) ions.

Whether glass-like or not, the core, even in fully hydrated spores,has a remarkably low water content, which correlates with thespore’s heat resistance (17). The core contains the genome, abouthalf of the spore’s proteins, and a high concentration of calcium andother divalent ions chelated by DPA (1, 2). The core is surroundedby the cortex, a highly hydrated, cross-linked peptidoglycan matrix(18) that is thought to maintain the dehydrated state of the core(Fig. 1). The spore’s outermost protective barrier, known as thecoat, is composed of multiple layers of cross-linked proteins (19,20). The coat provides resistance to chemicals and protects thecortex from hydrolytic enzymes. The spore has two membranes.The outer membrane, located just inside the coat, may not retainits integrity in the mature spore and, in any case, it is not a significantpermeability barrier (2). The inner membrane (IM) that envelopsthe core may exist partly in a gel state with largely immobile lipids(21). The low permeability of the IM to ionic species and even tosmall nonionic solutes like methylamine has been linked to its rolein protecting the core DNA from chemical attack (2).

Nuclear magnetic relaxation is among the few techniques that canyield quantitative information about the physical state of water incomplex biological systems, but the interpretation of the measuredspin relaxation rates is rarely straightforward (22). To separatecontributions from water motions on different time scales, it isnecessary to measure the longitudinal spin relaxation rate R1 overa wide range of resonance frequencies �0. By using an array of NMRinstruments, including fast field-cycling magnets as well as conven-tional high-field magnets, a so-called magnetic relaxation dispersion(MRD) profile, R1(�0), can be recorded that spans 5 orders ofmagnitude in frequency (103-108 Hz for 2H). The MRD profile is

Author contributions: E.P.S., P.S., L.H., and B.H. designed research; E.P.S. performed re-search; P.S. and L.H. contributed new reagents/analytic tools; E.P.S. and B.H. analyzed data;and E.P.S. and B.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

1To whom correspondence should be addressed. E-mail: bertil.halle@bpc.lu.se.

This article contains supporting information online at www.pnas.org/cgi/content/full/0908712106/DCSupplemental.

19334–19339 � PNAS � November 17, 2009 � vol. 106 � no. 46 www.pnas.org�cgi�doi�10.1073�pnas.0908712106

essentially a frequency mapping of the distribution of rotationalcorrelation times for all water molecules in the sample. Recent 2HMRD studies of living bacteria (23) have shown that cell waterconforms to the expectations based on MRD studies of modelsystems (24). Here we use the same approach to characterize waterdynamics in fully hydrated dormant spores of B. subtilis. In addition,we have measured the high-frequency relaxation rate of water 17Ospins. In general, the 2H and 17O nuclides are preferable to 1Hbecause they relax by a single-spin mechanism (involving thenuclear electric quadrupole) and hence provide direct informationabout the molecular rotational mobility (25), which is closely relatedto the (local) viscosity. In addition, the 2H and 17O relaxation ratesare less affected than the 1H rate by paramagnetic ions. Finally, the17O relaxation rate reports exclusively on water molecules, whereasthe 1H and (to a lesser extent) 2H rates may contain a pH-dependent contribution from macromolecular OH and NH hydro-gens that exchange with water hydrogens on the NMR relaxationtime scale (22).

Because the key to dormancy and heat resistance is believed toreside in the core, the mobility of core water is of particular interest.H2O/D2O exchange studies demonstrate that external water ex-changes with at least 96% of spore water (26), but our data showthat water exchange across the IM is slow on the 2H NMRrelaxation time scale (20 ms). This allows us to separately assesswater mobility inside and outside the core. We use the Mn(II)-induced 2H spin relaxation to identify the magnetization compo-nent associated with the core, but we remove Mn(II) ions outsidethe core by EDTA to reveal the water dynamics in the cortex andcoat. Further information was obtained by examining coat-deficientspores from a B. subtilis strain with mutations in the cotE and gerEgenes (27). Unexpectedly, we found that the water permeability ofthe IM is greatly enhanced in these coat-deficient spores.

ResultsSpatial Distribution and Hydration of Mn(II) Ions. Fig. 2 (blacksymbols) shows a water-2H MRD profile spanning five frequencydecades, measured on a sample of packed, fully hydrated WT B.subtilis spores. The MRD profile has two prominent features: alarge-amplitude dispersion below �1 MHz and a broad maximumnear 6 MHz. As discussed below (Internal Water in Coat Proteins),the low-frequency dispersion is produced by 0.14% of the sporewater, transiently trapped in protein cavities. The MRD maximumis the hallmark of a paramagnetic relaxation enhancement (PRE)induced by the magnetic dipole–dipole coupling between unpairedelectron spins of slowly tumbling paramagnetic ions, such asprotein-bound Mn(II), and 2H (or 1H) spins in directly coordinatedwater molecules (28, 29). Manganese is present in our sporepreparation at a high level of 84 �mol (g dry mass)�1 (Table S1 ofthe SI Appendix). Other paramagnetic ions are much less abundant(Table S1 of the SI Appendix) and produce (at a given concentra-tion) a much smaller PRE than Mn(II) (29).

The observed 2H relaxation rate is the sum of a quadrupolar

contribution, which contains the desired information aboutspore-water rotational dynamics (22), and a PRE contribution.To eliminate the Mn(II)-induced PRE contribution, we treatedthe spores with the chelating agent EDTA. The MRD profile forEDTA-treated spores (Fig. 2, blue symbols) lacks the charac-teristic PRE maximum, and it can be represented with a sum ofLorentzian spectral densities, as for diamagnetic samples ofimmobilized proteins (23, 24).

Elemental analysis showed that EDTA treatment removes only32% of spore manganese (Table S1 of the SI Appendix). Because theMRD profile from EDTA-treated spores does not have a PREcontribution (Fig. 2), we conclude that 68% of the Mn(II) ions inthe spore are ‘‘invisible’’ to water 2H relaxation. Two well-established facts strongly suggest that these invisible and nonex-tractable Mn(II) ions are located in the core. First, microanalyticalstudies identify the core as the major locus of mineral ions, withlesser amounts present in the coat (1). Second, the IM is essentiallyimpermeable to ionic species (2). Neither Mn(II) ions nor EDTAare therefore expected to cross the IM on the 4-h time scale of theEDTA treatment. For the same reason, EDTA treatment removedonly 10% of the large Ca2� content of the spore (Table S1 of theSI Appendix).

The PRE contribution, obtained from the difference of the blackand blue MRD data in Fig. 2, was analyzed quantitatively with theaid of the conventional Solomon–Bloembergen–Morgan theory(29, 30). A satisfactory agreement was obtained with parametervalues in the expected ranges (see Text and Fig. S1 in the SIAppendix). In particular, the analysis indicates that the ‘‘visible’’Mn(II) ions, located outside the core, have one water ligand onaverage.

Decomposition of the 2H Relaxation Rate. The water-2H magneti-zation relaxes biexponentially in both untreated and EDTA-treated spores (Fig. 3), implying slow water exchange betweentwo compartments (31). Biexponential fits to the 2H magneti-zation recovery at 92.1 MHz yield local relaxation rates for thetwo regions and the water fraction in each region (Table 1). Thewater distribution is the same for untreated and EDTA-treatedspores, with 12% of sample water in the minor fraction and 88%

Fig. 1. Transmission electron micrograph of a wild-type (WT) B. subtilisendospore (60), with the principal regions color coded: core (red), cortex(white), and coat (green).

Fig. 2. Water-2H MRD profiles from B. subtilis spores at 27° C and pH 7.6:untreated WT (black), EDTA-treated WT (blue), and EDTA-treated mutant(red). All R1 data refer to a water content of 1.20 g D2O (g dry mass)�1. Thecurves are multi-Lorentzian numerical representations used for the model-free analysis. The blue-colored area is the contribution from internal watermolecules with residence time �160 ns. The difference between the upper(black) and lower (blue) curves represents the PRE produced by Mn(II) ionsoutside the core.

Sunde et al. PNAS � November 17, 2009 � vol. 106 � no. 46 � 19335

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

in the major fraction. Because the IM is the only strongpermeability barrier in the spore (2), we associate these waterfractions with the core (cr) and noncore (ncr) regions of thesample. The ncr region comprises the cortex (cx), the coat (ct),and the external (ex) interstitial volume in the spore pellet (Fig.1). Furthermore, we assign the minor fraction to the core regionfor the following reasons. First, 12% is in the range expected forthe core-water fraction of B. subtilis spores (17, 32). Second, theminor fraction has the largest relaxation rate (Table 1), asexpected from the lower water content of the core (17, 32).Third, because EDTA removes Mn(II) ions from the ncr regionbut not from the core (see Spatial Distribution and Hydration ofMn(II) Ions), EDTA-treatment should reduce the 2H relaxationrate for the major (ncr) fraction but should not affect the rate forthe minor (core) fraction. As seen from Table 1, this is indeedthe case.

Spatially Resolved Spore Water Dynamics. The biexponential 2Hrelaxation caused by slow water exchange across the IM makes

it possible to separately assess water mobility in the differentspore compartments. We quantify water mobility in region k bythe dynamic perturbation factor (DPF), �k � ��R

k �/�R0 , that is, the

rotational correlation time averaged over all water molecules inregion k divided by the same quantity in bulk water. To anexcellent approximation (23), the DPF can be obtained in amodel-independent way as the ratio of the high-frequency 2Hrelaxation rate in region k, R1

k(�0�), and the (frequency-

independent) relaxation rate in bulk water, R10, that is �k �

R1k(�0

�)/R10. By using the highest experimental frequency, �0

� �92.1 MHz, we include in the average ��R

k � all water molecules withcorrelation times shorter than 1/(2��0

�) � 2 ns (23). The DPFthus includes essentially all water molecules in the hydrationlayers of proteins and other solutes (33), but not internal watermolecules buried in small polar cavities inside proteins or at theinterfaces of tightly associated macromolecules.

By using the local relaxation rates R1k(�0

�) in Table 1 and R10 �

2.175 s�1 (measured for bulk D2O at 27° C), we can calculate theDPF for the core and ncr regions (Table 2). For core water, wethus obtain �cr � 30.8 � 0.5. The ncr region includes the exvolume in the spore pellet. This volume is essentially bulk water,so �ex � 1. Subtracting this trivial contribution from �ncr, we find�cx�ct � 4.8 � 0.4 for water in the cortex � coat region (see theSI Appendix). The smaller DPF for this region is expectedbecause the water content is higher than in the core. Because offast water exchange between cortex and coat, we can only obtainthe population-weighted average of the DPFs for these regions.We expect that �cx is close to 1 because of the high water contentin the cortex, whereas �ct may be similar to �cr.

The DPFs given in Table 2 should be regarded as upper boundson the relative slowing down of water in the different sporecompartments because the high-frequency 2H relaxation rate mayinclude a contribution from labile biopolymer deuterons in solvent-exposed OD and ND groups exchanging with D2O deuterons on themillisecond time scale at pH 7.6 (34, 35). To assess the importanceof labile deuterons, we measured the water-17O relaxation rate at81.3 MHz for the EDTA-treated sample. Because the relaxationtime scale is much shorter for 17O than for 2H (22), the 17Omagnetization should also exchange slowly between the core andncr regions. Nevertheless, the 17O magnetization recovery appearsto be monoexponential (Fig. S2A in the SI Appendix), presumablybecause fast transverse 17O relaxation and/or dephasing due to localfield inhomogeneities in the core causes the minor magnetizationcomponent to be lost during the 90-�s acquisition delay. Theobserved 17O R1 rate, which thus pertains to the ncr region, yieldsa DPF of �ncr � 2.69 � 0.02 and, after correction for the ex fraction(as for the 2H data), �cx�ct � 3.6 � 0.3. This value is 25% smallerthan the 2H-derived DPF (Table 2). In contrast, for Escherichia colicells at the same temperature and pH as here, the 2H and 17O DPFsdo not differ significantly (23).

For the core, only the 2H-derived DPF is available (Table 2). Thelabile-deuteron contribution in the core is probably negligibly smallbecause the core pH is about one unit lower than in the ncr region(36) and because the DPF is much larger. Nevertheless, the �cr value

Fig. 3. Water-2H magnetization recoveries at 92.1 MHz from untreated (A)and EDTA-treated (B) WT B. subtilis spores. The lower frames show experi-mental data and biexponential fits with parameter values as given in Table 1.The upper frames show the residuals for monoexponential (solid blue circles)and biexponential (open red circles) fits, with the former divided by a factor50 to fit on the same scale.

Table 1. Results of fits to water-2H magnetization recovery at92.1 MHz

Sample fcr R1cr (s�1) R1

ncr (s�1)

WT 0.121 (4) 68.6 (3) 19.40 (1)WT � EDTA 0.116 (4) 67.1 (5) 7.56 (1)Mutant � EDTA – 16.33 (1)

Errors (one standard deviation) in the last digit are given in parentheses.

Table 2. Water fraction and dynamic perturbation factor fordifferent spore regions

Strain Region fk* �k†

WT cr 0.12 (1) 30.8 (5)WT ncr 0.88 (1) 3.48 (1)WT cx � ct 0.58 (5) 4.8 (4)cotE gerE mutant cr � ncr 1 7.51 (1)

Errors (one standard deviation) in the last digit are given in parentheses. cr,core; cx, cortex; ct, coat; ex, external interstitial space; ncr, cx � ct � ex.*Fraction of sample water in region k.†Dynamic perturbation factor for region k.

19336 � www.pnas.org�cgi�doi�10.1073�pnas.0908712106 Sunde et al.

in Table 2 should still be regarded as an upper bound because it maycontain a small (outer-sphere) PRE contribution from the nonex-tractable Mn(II) ions in the core (see the SI Appendix).

Internal Water in Coat Proteins. We return now to the dominantlow-frequency dispersion, focusing on the profile from EDTA-treated spores (Fig. 2, blue symbols). Similarly large low-frequency2H dispersions are obtained from immobilized globular proteins incross-linked protein gels (24) or in bacterial cells (23) but not fromfreely tumbling proteins (22). This dispersion has been quantita-tively linked to the escape of buried water molecules from smallpolar cavities inside the protein, rate-limited by intermittent proteinconformational dynamics on the 10�8 � 10�5 s time scale (24).

As shown above, the MRD profile from WT spores reports onwater dynamics in the ncr region outside the core. Because thecortex peptidoglycan network (18) is not expected to trap watermolecules for periods �10 ns (35), the low-frequency dispersionmust be produced by internal water molecule in the coat layers,comprising �55% of total spore protein (37). The coat proteins,organized in dense, cross-linked layers (20, 27), are expected to berotationally immobilized and they should therefore contribute tothe low-frequency dispersion (23, 24).

We determine the internal water fraction, fint, by numericalintegration of the dispersion profile (Fig. 2, blue area). Because thisintegral is rigorously independent of the unknown internal-waterresidence time distribution, it yields fint in an essentially model-independent way (23, 38). With an upper integration limit of 1MHz, fint represents internal water molecules with residence timeslonger than 160 ns and shorter than �10 �s (23). In this way, weobtain fint � (1.0 � 0.1)�10�3. We can then estimate the averageprevalence of internal water in coat proteins, expressed as thenumber, �int, of internal water molecules per 100 aa residues (seethe SI Appendix). The result, �int � 2.5 � 0.6, is similar to the value3.6 � 1.0 derived in the same way for immobilized E. coli proteinsand to the value 3.4 obtained from analysis of 842 protein crystalstructures (39). The slightly lower value obtained here may reflecta lower internal-water content of coat proteins or a larger fractionof internal water molecules with residence times �10 �s (makingthem unobservable) in the extensively cross-linked (and hence lessflexible) coat proteins. In any case, our analysis shows that the largelow-frequency 2H dispersion can be accounted for by a tiny fraction(0.14%) of spore water trapped inside proteins and that it does notindicate a drastic slowing down of a significant water populationelsewhere in the spore.

Water Exchange Across the Inner Membrane. The finding that waterexchange between the core and ncr regions is slow on the 2Hrelaxation time scale provides an upper bound on the waterpermeability of the IM. Quantitatively, the observation of biexpo-nential 2H relaxation implies that the mean residence time of awater molecule in the core obeys the inequality (31) �cr � [(1 � fcr)(R1

cr � R1ncr)]�1. Inserting values from Table 1, we obtain �cr � 20

ms. This result can be converted into an upper bound on the passive(diffusive) water permeability of the IM through the relation Pw ��w

cracr/(3�cr), with the effective radius acr � 3Vcr/Acr determined bythe volume and surface area of the core. Taking the dimensions ofthe assumed prolate-ellipsoidal core from Fig. 1, we find acr � 0.20�m. By using �w

cr � 0.55 � 0.1 for the volume fraction of water inthe core (14, 17, 32), we thus obtain Pw 1.8 � 0.6 �m s�1.

We also recorded the water-2H MRD profile (Fig. 2, redsymbols) from spores of a cotE gerE B. subtilis strain lacking mostof the proteinaceous coat (27). The fraction of missing coat protein,87 � 16%, was estimated from protein analysis of the WT andmutant spore preparations (see the SI Appendix). The mutantspores were also treated with EDTA, which removed the PRE peakat 6 MHz. For WT spores, the MRD profile reports on waterdynamics outside the core. Because most of the protein in thisregion is concentrated in the coat, we expected the MRD profile for

the mutant spores to fall well below that of the WT spores. [The R1data in Fig. 2 have been normalized to the same water mass fractionto compensate for the slightly different total water content in thesamples (22).] But the 2H relaxation rate for the mutant spores ishigher at all frequencies. This paradox is resolved by examining themagnetization recovery at high frequencies.

In contrast to WT spores, mutant spores yield monoexponential2H relaxation at 92 MHz (Fig. S2B in the SI Appendix), implying fastwater exchange across the IM (core water residence time 20 ms).R1 thus reports on water dynamics in the entire spore, including thecore region. If water exchange had been fast in the WT spore, wewould have observed monoexponential 2H relaxation at 92 MHzwith R1 � fcr R1

cr � (1 � fcr) R1ncr � 14.5 �1, not far from the value

measured for the mutant spores (Table 1). (Full agreement is notexpected because of the loss of coat protein, which should decreaseR1

ncr and increase fcr.)The fast-exchange condition established at 92 MHz does not

necessarily apply at low frequencies, where the much larger relax-ation rate makes the condition more restrictive. Although thebiexponential magnetization recovery expected under slow-exchange conditions would not have been observed with thefield-cycling instrument (see the SI Appendix), the low-frequencydispersion is far too large to be produced by internal watermolecules in the proteins of the severely decimated coat of themutant spore. Indeed, a quantitative analysis shows that waterexchange across the IM must be fast compared with the 2Hrelaxation rate also at low frequencies (see the SI Appendix). We canthen infer a more stringent upper bound on the mean residencetime of a water molecule in the core: �cr 0.8 ms (see the SIAppendix). Converting the residence time into IM water perme-ability as before, we find Pw � 46 � 14 �m s� 1.

DiscussionBarriers to Water Transport Within the Spore. As compared tovegetative cells, spores are highly resistant to a wide range of toxicchemicals (2). The spore’s first line of defense is the coat, themultiple protein layers of which act as a chemical filter. In addition,the core is physically protected by the low permeability of the IM.Whereas single-component phospholipid bilayers typically havewater permeabilities in the range 50–300 �m s�1 (at 25° C) (40, 41),we find that Pw 1.8 � 0.6 �m s�1 (at 27° C) for the IM of B. subtilisspores. The low water permeability of the IM is consistent with theslow uptake of methylamine by the core of B. megaterium spores(42). From the reported time, 2.5 h, for half maximal uptake (at24° C) (42), we estimate a permeability for (deprotonated) meth-ylamine of 0.1 �m s�1. Because methylamine is less polar thanwater, this result suggests that the water permeability of the IM maybe 1–2 orders of magnitude below our upper bound and thus 3–4orders of magnitude lower than in simple bilayer membranes(40, 41).

According to the solubility-diffusion model of permeation, Pw isproportional to the product of diffusion coefficient, Dw, and par-tition coefficient of water in the membrane (43). The extremely lowPw thus indicates a drastic reduction of either or both of thesefactors in the IM. Diffusion measurements employing fluorescentlabels have shown that 70% of the IM lipids in B. subtilis spores areimmobile on a time scale of seconds, whereas the diffusion coef-ficient, Dlipid, of the remaining 30% is reduced by an order ofmagnitude compared with vegetative cells (21). The drasticallyreduced fluidity of the IM implied by these results is qualitativelyconsistent with the low water permeability inferred here becauseboth Dw and Dlipid should be inversely proportional to the effectiveviscosity of the membrane. Furthermore, the relatively high core-water mobility found here indicates that the low lipid mobility iscaused by membrane compression (and perhaps gel-state ordering)rather than by headgroup interactions with a glassy core interior.

The much larger water permeability Pw � 46 � 14 �m s�1, foundhere for the IM of the coat-deficient mutant spores, is unexpected

Sunde et al. PNAS � November 17, 2009 � vol. 106 � no. 46 � 19337

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

and can only be explained by invoking major structural changes inthe membrane. Presumably, the IM of mutant spores is also morepermeable to other small molecules. Although the sensitivity ofcotE gerE spores to hypochlorite has been linked to the loss of coatproteins (27), the enhanced permeability of the IM may alsocontribute to the impaired chemical resistance.

The spore coat, including the outer membrane, is not consideredto be a significant permeability barrier for small molecules (2).Consistent with this view, our data indicate fast water exchangeacross the coat on the 2H and 17O relaxation time scales. Scanningmicroscope observation of B. thuringiensis spores exposed to vari-able humidity identified two distinct time scales, 50 and 500 s, forspore swelling, tentatively identified with water influx into the coat� cortex and into the core, respectively (44). These time scales referto swelling of dried spores and they are much longer than the timescales found here for water exchange within fully hydrated spores.The longer time scale (500 s) corresponds to a water permeabilityof the IM of order 10�4 �m s�1, 3 orders of magnitude less than formethylamine (see above). The shorter time scale (50 s) is 4 ordersof magnitude longer than the upper bound for the water exchangetime across the coat implied by the present NMR data.

The Core Is Not in a Glassy State. To compare the core-water DPF,�cr � 30.8 � 0.5, with DPFs for water in other biological systems,we must take into account differences in water content. For thiscomparison, we quote hydration levels in units of g H2O (g dryprotein)�1, conventionally denoted by h. For the core of the WTspores, we estimate a hydration level of 0.6 � 0.1 h (see the SIAppendix). Monolayer coverage of a 20 kDa protein corresponds to0.7 h (23). Because some of the core water hydrates nonproteincomponents, we conclude that the core proteins are surrounded byat most one water layer. With the possible exception of Ca2� ions(see the SI Appendix), nonprotein components are not expected tocontribute significantly to the core-water DPF (23).

The core-water DPF is twice as large as the DPF of 15.6 �3 for water in the macromolecular hydration layers of an E. colicell (23), and it differs even more from the DPFs of 5–10obtained for the hydration layer of proteins in dilute aqueoussolution (45). These DPF differences can be understood byrealizing that the DPF is an average over a wide distributionof hydration sites, with water correlation times ranging from afew ps to 2 ns. Hydration sites with correlation times approach-ing the upper limit of this range contribute disproportionatelyto the DPF (33). The defining feature of such stronglyperturbing hydration sites is geometrical confinement, whichprohibits collective rearrangement of the hydrogen-bond net-work in liquid water (33). On an ‘‘isolated’’ protein surface,these secluded hydration sites, usually deep pockets, are few innumber, but additional secluded hydration sites are createdwhen a (partial) hydration layer is confined between twoprotein surfaces. The increasing DPF trend in going fromprotein solution (10–100 h) to bacterial cell (3.2 h) to sporecore (0.6 h) can thus be explained by the increasing prevalenceof confined hydration layers as the water content decreases.This notion is consistent with water 2H relaxation data fromprotein crystals and rehydrated protein powders. For proteincrystals, DPF values range from 9 for tetragonal hen lysozymeat 0.52 h (46) to 40 for crambin at 0.34 h (47), whereas a DPFof 46 was reported for amorphous hen lysozyme powderhydrated to 0.25 h (48). In conclusion, the core-water DPF isconsistent with DPF values from binary protein-water systemsat comparable hydration levels. As for other systems, themajority of core water molecules are likely to be considerablymore mobile than indicated by the DPF, which is dominated bya minority of water molecules in confined hydration sites.Furthermore, a significant contribution to the core-water DPFcould come from relatively slow (but 2 ns) symmetric f lip

motions of water molecules coordinated to Ca2� ions in thecore (49).

If the core were in a glassy state, as conjectured (7–9), then corewater would be in the ‘‘rigid lattice’’ NMR regime (correlation time�� 10�6 s), producing a �200 kHz wide 2H NMR spectrum thatwould not contribute to the 2H NMR signal detected here. How-ever, we find that, on average, core water is ‘‘only’’ 30-fold lessmobile than bulk water. Because the rotational correlation time forbulk H2O is 1.6 ps at 27° C, �cr � 30.8 corresponds to a meanrotational correlation time of 30.8 � 1.6 � 50 ps for core water. Wetherefore conclude that the core is not in a glassy state. Fullyhydrated proteins undergo a broad intramolecular dynamical tran-sition to a glass-like state but only at low temperatures (180–220 K)(50). A similar transition is observed near room temperature forprotein powders (50) but only at water contents (�0.1 h) muchlower than in the core (�0.6 h). Our results on core-water mobilityare therefore consistent with the behavior of simpler systems.

Water behavior in biological systems is often described in termsof ‘‘free’’ and ‘‘bound’’ water, with the tacit assumption that boundwater is largely immobile. These terms lack precision and are oftenmisleading. If a water molecule is extensively hydrogen-bonded,whether to proteins or to adjacent water molecules, it stronglyprefers this environment to a noninteracting state (water vapor).However, hydrogen bonds per se do not immobilize a watermolecule, as seen by considering bulk water. Core water is bothbound and mobile, and there is no contradiction in this.

Origin of Dormancy and Heat Resistance. In binary protein–watersystems, protein tumbling is effectively quenched at hydration levels1.5 h (51). In the core (at 0.6 h), proteins are thus not expectedto undergo translational or rotational diffusion at measureablerates. This expectation is consistent with fluorescence studies (14)and with our analysis of the low-frequency 2H dispersion (Fig. 2) interms of internal water molecules in rotationally immobilizedproteins. However, protein immobilization does not preclude en-zyme activity as long as substrates can diffuse within the core. Asnoted above, most core water is expected to be much more mobilethan indicated by the DPF. The rate of substrate diffusion in thecore of the dormant spore may therefore be of the same order ofmagnitude as in the vegetative cell. Unlike eukaryotic organismsthat achieve dormancy by replacing water with glass-forming dis-accharides (5, 6), bacterial spores can apparently inhibit metabolismwithout immobilizing the solvent. This conclusion may seem sur-prising because purified enzymes may show significant activity athydration levels as low as 0.2 h (50). However, water is probably notevenly distributed within the core. Key enzymes may therefore besufficiently dehydrated that they cannot access conformationsrequired for activity (52).

Heat inactivation of spores typically requires temperatures 40° Cabove what is needed to kill the corresponding vegetative cells (1,2). Moreover, for Bacillus spores, heat resistance decreases expo-nentially with increasing core water content in the range 28–57%(17). Core dehydration might improve heat resistance by stabilizingproteins against thermal denaturation (50, 53), presumably bydisfavoring the unfolded state entropically (by excluding extendedpolypeptide conformations) as well as energetically (by restrictingavailability of water molecules for replacing intraprotein hydrogenbonds). However, to raise the heat denaturation temperature by40° C in binary systems requires dehydration to �0.1 h, whereaslittle or no stabilization is seen at the hydration level (0.6 h) of thespore core (50, 53).

If core dehydration does not confer heat resistance directly bystabilizing proteins against thermal denaturation, it might actindirectly by immobilizing proteins, thereby preventing irreversibleprotein aggregation. As long as heat denaturation is reversible, itshould not kill the spore. Indeed, protein lyophilization may induce(partial) unfolding (54), but this is usually a reversible process. Onlyat hydration levels high enough to permit intermolecular disulfide

19338 � www.pnas.org�cgi�doi�10.1073�pnas.0908712106 Sunde et al.

exchange or entanglement does unfolding lead to irreversibleprotein aggregation (55), which would compromise the viability ofthe spore. This view is also consistent with Raman spectra of singleB. subtilis spores, indicating that heat activation at a sublethaltemperature of 70° C is accompanied by partial but reversibleprotein denaturation (56).

Materials and MethodsB. subtilis Strains. The two B. subtilis strains used here are isogenic derivativesof strain PS832, a prototrophic laboratory derivative of strain 168. The WTstrain PS533 (57) carries a plasmid pUB110 encoding resistance to kanamycin,and the coat-deficient strain PS4150 (27) is deleted for most of the cotE andgerE coding sequences. The CotE protein is essential for assembly of many coatproteins and of the outer layer and GerE is a DNA-binding protein that, inaddition to other activities, regulates transcription of several coat proteingenes (19, 20). In the text, spores from the strains PS533 and PS4150 arereferred to as WT and mutant spores, respectively.

Spore Preparation. Spores were prepared at 37° C on 2�SG medium agar plates,and were harvested, cleaned and stored as described (58, 59). All spore prepara-

tions used were free (� 99%) from growing or sporulating cells, germinatedspores and cell debris as determined by phase contrast microscopy.

Preparation of NMR Samples. The spores were dissolved in 5 mM phosphate/D2O buffer at pH 7.6. After centrifugation at 6,800 � g, the spore mass wastransferred to an NMR tube insert. Manganese depleted samples were firstsuspended in 10 mM EDTA and washed twice. After completion of NMRmeasurements, the samples were dried to determine the water content andthey were then subjected to elemental analysis (Table S1 of the SI Appendix)and complete amino acid analysis (Table S3 of the SI Appendix). For furtherdetails, see the SI Appendix.

NMR Experiments. The longitudinal relaxation rate R1 of the water-2H mag-netization was measured from 1.5 kHz to 92.1 MHz by using six different NMRinstruments, and the water-17O R1 was measured at 81.3 MHz. All NMRexperiments were performed within 3 days at a temperature of 27 � 0.1° C.Biexponential magnetization curves were fitted with the function Mz(t) �Mz() � [Mz (0) � Mz()] [fcr exp(�R1

crt) � (1 � fcr) exp(�R1ncrt)]. For further

details, see the SI Appendix.

ACKNOWLEDGMENTS. This work was supported by grants from the SwedishResearch Council (to B.H. and L.H.), the Knut and Alice Wallenberg Foundation(to B.H.), and the US Army Research Office (to P.S.).

1. Gerhardt P, Marquis RE (1989) Spore thermoresistance mechanisms. In Regulation ofProcaryotic Development, eds Smith I, Slepecky RA, Setlow P (Am Soc Microbiol,Washington, DC), pp 43–63.

2. Setlow P (2006) Spores of Bacillus subtilis: Their resistance to and killing by radiation,heat and chemicals. J Appl Microbiol 101:514–525.

3. Nicholson WL, Munakata N, Horneck G, Melosh HJ, Setlow P (2000) Resistance ofBacillus endopsores to extreme terrestrial and extraterrestrial environments. MicrobiolMol Biol Rev 64:548–572.

4. Dill KA (1990) Dominant forces in protein folding. Biochemistry 29:7133–7155.5. Crowe JH, Carpenter JF, Crowe LM (1998) The role of vitrification in anhydrobiosis.

Annu Rev Physiol 60:73–103.6. Buitink J, Leprince O (2008) Intracellular glasses and seed survival in the dry state.

Comptes Rendus Biologies 331:788–795.7. Gould GW (1986) Water and the survival of bacterial spores. In Membranes, Metabo-

lism, and Dry Organisms, ed Leopold AC (Cornell Univ Press, Ithaca), pp 143–156.8. Sapru V, Labuza TP (1993) Glassy state in bacterial spores predicted by polymer

glass-transition theory. J Food Sci 58:445–448.9. Ablett S, Darke AH, Lillford PJ, Martin DR (1999) Glass formation and dormancy in

bacterial spores. Int J Food Sci Technol 34:59–69.10. Leuschner RGK, Lillford PJ (2003) Thermal properties of bacterial spores and biopoly-

mers. Int J Food Microbiol 80:131–143.11. Carstensen EL, Marquis RE, Child SZ, Bender GR (1979) Dielectric properties of native

and decoated spores of Bacillus megaterium. J Bacteriol 140:917–928.12. Johnstone K, Stewart GSAB, Barratt MD, Ellar DJ (1982) An electron paramagnetic

resonance study of the manganese environment within dormant spores of Bacillusmegaterium KM. Biochim Biophys Acta 714:379–381.

13. Leuschner RGK, Lillford PJ (2000) Effects of hydration on molecular mobility in phase-bright Bacillus subtilis spores. Microbiology 146:49–55.

14. Cowan AE, Koppel DE, Setlow B, Setlow P (2003) A soluble protein is immobile indormant spores of Bacillus subtilis but is mobile in germinated spores: Implications forspore dormancy. Proc Natl Acad Sci USA 100:4209–4214.

15. Maeda Y, Fujita T, Sugiura Y, Koga S (1968) Physical properties of water in spores ofBacillus megaterium. J Gen Appl Microbiol 14:217–226.

16. Bradbury JH, Foster JR, Hammer B, Lindsay J, Murrell WG (1981) The source of the heatresistance of bacterial spores. Study of water in spores by NMR. Biochim Biophys Acta678:157–164.

17. Beaman TC, Gerhardt P (1986) Heat resistance of bacterial spores correlated withprotoplast dehydration, mineralization, and thermal adaption. Appl Environ Micro-biol 52:1242–1246.

18. Popham DL (2002) Specialized peptidoglycan of the bacterial endospore: The innerwall of the lockbox. Cell Mol Life Sci 59:426–433.

19. Driks A (1999) Bacillus subtilis spore coat. Microbiol Mol Biol Rev 63:1–20.20. Henriques AO, Moran CP (2000) Structure and assembly of the bacterial endospore

coat. Methods 20:95–100.21. Cowan AE, et al. (2004) Lipids in the inner membrane of dormant spores of Bacillus

species are largely immobile. Proc Natl Acad Sci USA 101:7733–7738.22. Halle B, Denisov VP, Venu K (1999) Multinuclear relaxation dispersion studies of

protein hydration. In Biological Magnetic Resonance, eds Krishna NR, Berliner LJ(Kluwer Academic, New York), Vol 17, pp 419–484.

23. Persson E, Halle B (2008) Cell water dynamics on multiple time scales. Proc Natl AcadSci USA 105:6266–6271.

24. Persson E, Halle B (2008) Nanosecond to microsecond protein dynamics probed by mag-netic relaxation dispersion of buried water molecules. J Am Chem Soc 130:1774–1787.

25. Abragam A (1961) The Principles of Nuclear Magnetism (Clarendon Press, Oxford).26. Marshall BJ, Murrell WG (1970) Biophysical analysis of the spore. J Appl Bact 33:103–

129.27. Ghosh S, et al. (2008) Characterization of spores of Bacillus subtilis that lack most coat

layers. J Bacteriol 190:6741–6748.28. Koenig SH, Brown RD (1985) Relaxation of solvent protons and deuterons by protein-

bound Mn2� ions. Theory and experiment for Mn2�-concanavalin A. J Magn Reson61:426–439.

29. Bertini I, Luchinat I, Parigi G (2005) 1H NMRD profiles of paramagnetic complexes andmetalloproteins. Adv Inorg Chem 57:105–172.

30. Kowalewski J, Kruk D, Parigi G (2005) NMR relaxation in solution of paramagneticcomplexes: Recent theoretical progress for S � 1. Adv Inorg Chem 57:41–104.

31. Zimmerman JR, Brittin WE (1957) Nuclear magnetic resonance studies in multiplephase systems: Lifetime of a water molecule in an adsorbing phase on silica gel. J PhysChem 61:1328–1333.

32. Nakashio S, Gerhardt P (1985) Protoplast dehydration correlated with heat resistanceof bacterial spores. J Bacteriol 162:571–578.

33. Mattea C, Qvist J, Halle B (2008) Dynamics at the protein-water interface from 17O spinrelaxation in deeply supercooled solutions. Biophys J 95:2951–2963.

34. Denisov VP, Halle B (1995) Hydrogen exchange and protein hydration: The deuteronspin relaxation dispersions of bovine pancreatic trypsin inhibitor and ubiquitin. J MolBiol 245:698–709.

35. Vaca Chavez F, Hellstrand E, Halle B (2006) Hydrogen exchange and hydration dynam-ics in gelatin gels. J Phys Chem B 110:21551–21559.

36. Setlow B, Setlow P (1980) Measurements of the pH within dormant and germinatedbacterial spores. Proc Natl Acad Sci USA 77:2474–2476.

37. Goldman RC, Tipper DJ (1978) Bacillus subtilis spore coats: Complexity and purificationof a unique polypeptide component. J Bacteriol 135:1091–1106.

38. Halle B, Johannesson H, Venu K (1998) Model-free analysis of stretched relaxationdispersions. J Magn Reson 135:1–13.

39. Park S, Saven JG (2005) Statistical and molecular dynamics studies of buried waters inglobular proteins. Proteins 60:450–463.

40. Huster D, Jin AJ, Arnold K, Gawrisch K (1997) Water permeability of polyunsaturatedlipid membranes measured by 17O NMR., Biophys J 73:855–864.

41. Mathai JC, Tristram–Nagle S, Nagle JF, Zeidel ML (2008) Structural determinants ofwater permeability through the lipid membrane. J Gen Physiol 131:69–76.

42. Swerdlow BM, Setlow B, Setlow P (1981) Levels of H� and other monovalent cations indormant and germinating spores of Bacillus megaterium. J Bacteriol 148:20–29.

43. Finkelstein A (1987) Water Movement Through Lipid Bilayers, Pores, and PlasmaMembranes. (Wiley, New York).

44. Westphal AJ, Price PB, Leighton TJ, Wheeler KE (2003) Kinetics of size changes ofindividual Bacillus thuringiensis spores in response to changes in relative humidity.Proc Natl Acad Sci USA 100:3461–3466.

45. Halle B (2004) Protein hydration dynamics in solution: a critical survey. Philos Trans RSoc London Ser B 359:1207–1224.

46. Usha MG, Speyer J, Wittebort RJ (1991) Dynamics of the hydrate and amide groups ofcrystalline ribonuclease and lysozyme. Chem Phys 158:487–500.

47. Usha MG, Wittebort RJ (1989) Orientational ordering and dynamics of the hydrate andexchangeable hydrogen atoms in crystalline crambin. J Mol Biol 208:669–678.

48. Peemoeller H, Yeomans FG, Kydon DW, Sharp AR (1986) Water molecule dynamics inhydrated lysozyme. A deuteron magnetic resonance study. Biophys J 49:943–948.

49. Denisov VP, Halle B (1995) Direct observation of calcium-coordinated water in calbi-ndin D9k by nuclear magnetic relaxation dispersion. Biochemistry 117:8456–8465.

50. Gregory RB (1995) Protein hydration and glass transition behavior. In Protein-SolventInteractions, ed Gregory RB (M Dekker, New York), pp 191–264.

51. Kimmich R, Gneiting T, Kotitschke K, Schnur G (1990) Fluctuations, exchange processes,and water diffusion in aqueous protein systems. A study of bovine serum albumin bydiverse NMR techniques. Biophys J 58:1183–1197.

52. Setlow P (1994) Mechanisms which contribute to the long-term survival of spores ofBacillus species. J Appl Bact 76:49S–60S.

53. Rupley JA, Careri G (1991) Protein hydration and function. Adv Prot Chem 41:37–172.54. Desai UR, Osterhout JJ, Klibanov AM (1994) Protein structure in the lyophilized state:

A hydrogen isotope exchange / NMR study with bovine pancreatic trypsin inhibitor.J Am Chem Soc 116:9420–9422.

55. Klibanov AM, Schefiliti JA (2004) On the relationship between conformation andstability in solid pharmaceutical protein formulations. Biotechnol Lett 26:1103–1106.

56. Zhang P, Setlow P, Li Y (2009) Characterization of single heat-activated Bacillus sporesusing laser tweezers Raman spectroscopy. Optics Express 17:16480–16491.

57. Setlow B, Setlow P (1996) Role of DNA repair in Bacillus subtilis spore resistance. JBacteriol 178:3486–3495.

58. Nicholson WL, Setlow P (1990) Sporulation, germination and outgrowth. In BiologicalMethods for Bacillus, eds Harwood CR, Cutting SM (Wiley, Chichester, UK), pp 391–450.

59. Paidhungat M, Setlow B, Driks A, Setlow P (2000) Characterization of spores of Bacillussubtilis which lack dipicolinic acid. J Bacteriol 182:5505–5512.

60. Erlendsson LS, Moller M, Hederstedt L (2004) Bacillus subtilis StoA is a thiol-disulfideoxidoreductase important for spore cortex synthesis. J Bacteriol 186:6230–6238.

Sunde et al. PNAS � November 17, 2009 � vol. 106 � no. 46 � 19339

BIO

PHYS

ICS

AN

DCO

MPU

TATI

ON

AL

BIO

LOG

Y

1

Supporting Information

The physical state of water in bacterial spores Erik P. Sunde, Peter Setlow, Lars Hederstedt, and Bertil Halle

SI Text

Preparation of NMR Samples. Freeze-dried wild-type, WT (240 mg) or mutant (220 mg)

spores were suspended in 180 ml 50 mM Tris/HCl buffer containing 10 mM EDTA in

distilled water at pH 7.2. After 4 h incubation on a rocking table at 4 °C, the sample was

centrifuged for 15 min at 3500 × g and then washed twice with 180 ml distilled water. The

EDTA-treated WT and mutant spore pellets and 201 mg untreated WT spores were each

dissolved in 10 ml 5 mM phosphate buffer in 99.9 % D2O (Spectra Stable Isotopes) at pH

7.6 and incubated overnight at 4 °C. (The quoted pH is the uncorrected value measured

with a glass electrode calibrated in H2O buffers.) The samples were finally centrifuged for

20 min at 6800 × g. The supernatant was discarded and the spore mass transferred to an 8 ×

20 mm NMR tube insert and sealed with parafilm. The water content of the NMR samples,

determined gravimetrically after drying for 16 h at 105 °C, was 1.20 g D2O (g dry mass)–1

for the untreated and EDTA-treated WT samples and 0.863 g D2O (g dry mass)–1 for the

mutant sample. The dried spore mass was used for elemental analysis by inductively

coupled plasma sector field mass spectrometry (performed at ALS Analytica, Luleå,

Sweden) (Table S1) and for complete amino acid analysis (performed at the Amino Acid

Analysis Center, Dept. of Biochemistry and Organic Chemistry, Uppsala University,

Sweden) (Table S3).

2

NMR Experiments. The longitudinal relaxation rate R1 of the water-2H magnetization was

measured from 1.5 kHz to 92.1 MHz by using six different NMR instruments: a Stelar

Spinmaster 1 Tesla fast field-cycling (FC) spectrometer (1.5 kHz – 5.8 MHz); a Tecmag

Discovery spectrometer equipped with an iron-core magnet (Drusch EAR-35N), variable-

field lock and flux stabilizer (11.9 MHz); and four spectrometers equipped with

conventional cryomagnets: Bruker Avance DMX 200 (30.7 MHz), Varian Unity Plus 400

operated at 55.6 MHz, Varian Unity Inova 500 (76.7 MHz) and 600 (92.1 MHz). The

relaxation rate R1 of the water-17O magnetization (at natural abundance, ~ 0.04 %) was

measured at 81.3 MHz. For the FC measurements, the pre-polarized and non-prepolarized

sequences were used with pre-polarization and detection at 6.14 and 4.80 MHz,

respectively. The recovery and polarization times were set to 4 T1. In the non-FC

experiments, standard phase-cycled inversion recovery pulse sequences were used. All the

NMR experiments were performed within 3 days at a temperature of 27 0.1 °C,

maintained by a thermostated air flow. The temperature was checked before and after each

relaxation experiment with a thermocouple referenced to an ice-water bath.

Whereas 2H relaxation is clearly bi-exponential at frequencies above 10 MHz (Fig.

3), it appears mono-exponential below 6 MHz (Fig. S2), because the rapidly relaxing

magnetization component escapes detection with the field-cycling instrument used at low

frequencies. The R1 values reported in Fig. 2 thus pertain to the slowly relaxing water

fraction. This is approximately true also at high frequencies, because the effective R1

obtained from a mono-exponential fit differs by < 10 % from the relaxation rate of the

dominant slowly relaxing component.

3

Analysis of Mn(II)-Induced Paramagnetic Relaxation Enhancement. The longitudinal

relaxation rate of the water 2H magnetization in a sample containing Mn(II) ions is a sum of

two contributions:

�

R1 = R1Q + R1

PRE [1]

�

R1Q is the purely quadrupolar relaxation rate (1) that would be measured in the absence of

Mn(II) ions. Under the conditions of the present study, the paramagnetic relaxation

enhancement (PRE) may be expressed as

�

R1PRE =

qMnNMn

R1, MnDD [2]

The first factor represents the fraction water directly coordinated to Mn(II) ions, expressed

in terms of the number, qMn, of water ligands per Mn(II) ion and the water:Mn mole ratio,

NMn. The intrinsic relaxation rate of Mn(II)-coordinated water deuterons,

�

R1, MnDD , is induced

by the magnetic dipole-dipole coupling between the 5 unpaired electron spins of Mn(II) and

the nuclear spin of 2H. The Fermi contact and Curie spin relaxation contributions are

negligibly small for Mn(II) at the magnetic fields of interest here (2-4), and the so-called

outer-sphere contribution (the PRE for water molecules not directly coordinated to the

Mn(II) ion) is also negligible (2). Equation [2] is valid in the fast-exchange limit, where the

mean residence time, τMn, of the Mn(II)-coordinated water deuteron is much shorter than its

intrinsic relaxation time,

�

1/R1, MnDD (here, of order 10 µs or longer).

The intrinsic dipolar relaxation rate takes the form (3, 4)

�

R1, MnDD = ωDD

2 0.3τD11 + (ωI τD1)

2 +0.7 τD2

1 + (ωS τD2)2

⎡

⎣ ⎢

⎤

⎦ ⎥ [3]

with the dipolar coupling frequency given by

4

�

ωDD = 2 µ04π γ I γ S

S(S +1)3

⎡ ⎣ ⎢

⎤ ⎦ ⎥ 1/2rIS−3 [4]

Here,

�

γ I = 4.1065×107 rad s–1 T–1 and

�

γ S = –1.7609×1011 rad s–1 T–1 are the 2H and

electron magnetogyric ratios, respectively, S = 5/2 is the electron spin quantum number of

Mn(II), and rIS is the Mn–2H separation. The Lorentzian spectral density functions in Eq.

[3] are evaluated at the 2H resonance frequency

�

ωI (denoted by ω0 in the main text) and at

the electron Larmor frequency

�

ωS = − 4288ωI . The I–S dipole coupling is modulated by

three independent processes: rotational diffusion of the Mn–2H vector with correlation time

τR, exchange of the Mn(II)-bound water molecule with residence time τMn, and electron

spin relaxation, with longitudinal and transverse relaxation times

�

T1S and

�

T2S . The effective

correlation times appearing in Eq. [3] are given by (3, 4)

�

1τDk

=1τR

+1

τMn+

1TkS , k = 1, 2 [5]

The electron spin relaxation of Mn(II) is induced by the transient zero-field splitting (ZFS)

Δ, modulated by vibrational motions in the coordination shell with correlation time τV.

Although the electron spin relaxation of Mn(II) is tri-exponential when

�

ωS τV > 1, it can be

approximately characterized by single longitudinal and transverse relaxation times given by

(3, 4)

�

1T1S =

Δ2

54 S(S +1) − 3[ ] 0.2 τV

1 + (ωS τV)2 +

0.8 τV1 + (2ωS τV)

2⎡

⎣ ⎢

⎤

⎦ ⎥ [6a]

�

1T2S =

Δ2

54 S(S +1) − 3[ ] 0.3τV +

0.5 τV1 + (ωS τV)

2 +0.2 τV

1 + (2ωS τV)2

⎡

⎣ ⎢

⎤

⎦ ⎥ [6b]

The model defined by Eqs. [2] – [6] predicts that the PRE exhibits two dispersion

steps and that

�

R1PRE(ωI ) increases with the 2H frequency

�

ωI between these steps. For

Mn(II) at room temperature,

�

T2S is of order 10–9 s, while τMn is of order 10–9 – 10–7 s (2, 5).

5

If the Mn(II) ion tumbles slowly, e.g. because it is protein-bound, so that τR >

�

T2S , then Eq.

[5] shows that

�

τD2 ≈ T2S . The second term of Eq. [3] then gives rise to a dispersion step

near 10 kHz 2H frequency, where

�

ωS τD2 ≈

�

ωS T2S ≈ 1. Above 1 MHz 2H frequency, where

�

ωS τV > 1,

�

T1S and

�

τD1 increase with frequency. As a result, the first term of Eq. [3]

increases with frequency, passes through a maximum when

�

ωI τD1 ≈ 1, and then falls to

zero.

As seen from Fig. S1, Eqs. [2] – [6] account rather well for the PRE contribution

from B. subtilis spores, extracted as the difference in R1 between the untreated and EDTA-

treated WT samples (black and blue symbols in Fig. 2). A quantitative agreement is not

expected since the model is approximate. In particular, it ignores heterogeneity in Mn(II)

coordination (qMn) and dynamics (τRM) and it does not include the effect of the static ZFS

expected for a Mn(II) ion with asymmetrical coordination (3, 4). The effect of a static ZFS

is to attenuate and upshift the low-frequency dispersion step (6), thereby bringing the model

in closer agreement with the data below ~ 1 MHz (Fig. S1). Although we have not analyzed

the data with the extended model, which involves 3 additional parameters and has little

effect on the PRE maximum (6), the fit in Fig. S1 yields parameter values within the

expected ranges. The ZFS parameters Δ = (1.20 ± 0.07)×109 s–1 (or 0.040 ± 0.002 cm–1) and

τV = 30 ± 3 ps are similar to those reported for Mn(II) chelates and protein-bound Mn(II)

ions (2, 7). With the Mn – 2H separation rIS = 2.90 Å (4) and the water:Mn mole ratio NMn =

1.20×0.88 / [20.02× (84.3–57.0) ×10–6] = 1932 in the non-core region of the spore sample

(Tables 2 and S1), the composite parameter

�

(qMn /NMn) rIS−6 = (1.0 ± 0.2)×10–6 Å–6 yields a

water coordination number qMn = 1.1 ± 0.2. This result indicates that the Mn(II) ions have

5 – 6 non-water ligands. Finally, the parameter τRM ≡ (1/τR + 1/τM)–1 = 4.9 ± 0.1 ns is in the

range 1 – 10 ns of τM values reported for Mn(II) complexes with a single water ligand (5).

6

This agreement indicates that

�

τR >> 5 ns, as expected for the largely immobilized proteins

in the coat layers (see main text).

Estimate of PRE Contribution to 2H Relaxation Rate of Core Water. If the core Mn(II)

ions coordinate one water molecule on average, as for the Mn(II) ions outside the core (see

above), then, because of the 15-fold lower water:Mn mole ratio in the core, we would

expect a PRE contribution to

�

R1cr of ~ 200 s–1, exceeding the observed value of 68.6 s–1

(Table 1). However, bi-exponential magnetization recovery fits in the frequency range 11.9

– 92.1 MHz show that

�

R1cr varies by a factor 2.9 only, much less than expected if

�

R1cr

contained a dominant PRE contribution (with a maximum near 6 MHz). All these fits yield

the same core-water fraction, fcr = 0.12. These observations argue against a significant

inner-sphere PRE contribution to

�

R1cr . This conclusion is consistent with the observation

that the manganese EPR spectrum from B. megaterium spores closely resembles that from

an anhydrous 10:1 Ca:Mn DPA chelate model system (8) and with 13C NMR spectra

indicating that the DPA in the core is in a solid-like state (9). Even if the core Mn(II) ions

have one or more water ligands, they would not affect the observed water-2H relaxation if

these water ligands have residence times

�

>> 10 µs, as might be the case if the core Mn(II)

ions are confined to a solid-like subdomain in the core. An outer-sphere PRE contribution

cannot be ruled out, but it is likely to be small in comparison with the large observed

�

R1cr

(2).

Since we find fast water exchange across the IM in coat-deficient spores, we should

have observed a PRE maximum near 6 MHz from this sample if the core Mn(II) ions gave

rise to a PRE for the core water. The absence of a PRE maximum for this sample thus

supports the conclusion (see above) that the core Mn(II) ions are 'invisible' to the water 2H

relaxation.

7

DPF Contribution from Interstitial Water. To subtract the contribution to ξncr from

water in the external interstitial region (ex), we need to know the fraction fex of sample

water in this region. From equilibrium permeation measurements with tritiated water and 14C-labeled glucose on fully hydrated B. subtilis spores, the fraction of spore water present

in the core, that is fcr /(1–fex), was determined to be 0.17 (10). We have determined the

fraction of sample water present in the core, fcr = 0.12 (Table 2). Assuming that the former

quantity (which does not depend on spore packing) is the same for our spores, we deduce

fex = 0.30 ± 0.05 with an uncertainty that allows for some variation in water distribution

between the spore preparations. The same authors also determined the volume fraction

water within the spore,

�

φwspore = 0.78 (10). We can then calculate, for our sample, the

volume fraction external interstitial space as

�

φex = fex / [ fex + (1− fex ) /φwspore] = 0.25. This

is a reasonable result, close to the value (0.26) for close-packed spheres. Since water

exchange is fast within the ncr region, we have

�

fncr ξncr = fcx+ct ξcx+ct + fex ξex . Inserting

values from Table 2, we thus obtain ξcx+ct = 4.8 ± 0.4.

Internal Water Content of Coat Proteins. The number of internal water molecules per

100 amino acid residues in proteins outside the core in WT spores may be written as

�

νint =mintncr

mpncr ×

100 MaaMw

[7]

where Mw = 20.02 g mol–1 is the molar mass of D2O and Maa = 120.3 g mol–1 is the mass-

weighted molar mass of an amino acid residue in the spore (Table S2). Furthermore, the

mass of internal water outside the core is given by

�

mintncr = f int mw

ncr , where

�

mwncr is 88 %

(Table 1) of the total sample water content, 1.20 g D2O (g dry mass)–1. Finally, the mass of

immobilized protein outside the core is given by

�

mpncr = ximmob fp

ncr mp , where mp = 0.5046

g protein (g dry mass)–1 is the total protein content of the dried sample (Table S2),

�

fpncr =

0.55 ± 0.1 is the fraction of spore protein located outside the core (11), and ximmob, taken to

8

be 0.9 ± 0.1, is the fraction immobilized protein within the ncr region. Combining all this,

we obtain νint = 2.5 ± 0.6 for the internal water content of the coat proteins.

Protein Deficiency in Coat of Mutant Spores. The fraction missing coat protein in the

mutant, ζ, can be obtained from the relation

�

ζ fpncr = 1 −

1/ ′ x p −11/ xp −1

[8]

where xp and x'p are the protein mass fractions in dried WT and mutant spores, respectively,

and

�

fpncr is the fraction of WT spore protein located outside the core. With the measured

protein mass fractions (Table S2) inserted into Eq. [8], we obtain ζ = 0.48/

�

fpncr . This

relation implies that ζ > 0.48 and that

�

fpncr > 0.48. With

�

fpncr = 0.55 ± 0.1 (11), we find ζ =

0.87 ± 0.16, consistent with other estimates (12). From a comparison of the amino acid

compositions of WT and mutant spores (Table S3), we can also conclude that the core

proteins have substantially more Ala and (Glu+Gln) and substantially less Gly, Phe and Tyr

than the coat proteins.

Demonstration of Fast Water Exchange in Mutant Spores. The fast-exchange condition

established at 92 MHz does not necessarily apply at low frequencies, where the much larger

relaxation rate makes the condition more restrictive. Although the bi-exponential

magnetization recovery expected under slow-exchange conditions would not have been

observed with the field-cycling instrument (see above), the low-frequency dispersion is far

too large to be produced by internal water molecules in the proteins of the severely

decimated coat of the mutant spore.

To demonstrate that water exchange across the IM in mutant spores is fast also on

the shorter time scale of 2H relaxation at the kHz frequencies, we show that the low-

frequency dispersion from the mutant is incompatible with slow-exchange conditions. In

9

the case of slow water exchange across the IM, the low-frequency dispersion is produced

by internal water molecules in (coat) proteins outside the core, just as for WT spores.

Equation [7] then applies, except that

�

′ m pncr = ximmob ′ f p

ncr ′ m p (where primes refer to the

mutant spore). The fraction of mutant spore protein located outside the core,

�

′ f pncr , can be

obtained from the corresponding quantity for WT spores by noting that

�

′ f pncr =

(1−ζ ) fpncr

1−ζ fpncr [9]

Numerical integration of the mutant MRD profile up to 1 MHz yields the internal-water

fraction fint = (1.2 ± 0.1)×10–3. With ζ = 0.87 ± 0.16 (see above) and

�

′ m p = 0.3461 g protein

(g dry mass)–1 (Table S2), we then obtain νint = 18 ± 4. This value is unreasonably high

compared to the values (2.5 – 3.6) obtained from WT spores (see above), from bacterial

cells (13) and from protein crystal structures (14).

In the case of fast water exchange across the IM, the low-frequency dispersion is

produced by internal water molecules all spore proteins. To calculate νint for this case, we

replace

�

mintncr in Eq. [7] by

�

f int mw and

�

mpncr by

�

ximmobmp . We thus obtain a result in the

expected range: νint = 2.7 ± 0.4. We therefore conclude that, in the coat-deficient spores,

water exchange across the IM is fast even at low frequencies.

If water exchange had been fast in the WT spore, we would have observed mono-

exponential 2H relaxation at 92 MHz with

�

R1 = fcr R1cr + (1− fcr )R1

ncr = 14.5 s–1, not far

from the value measured for the mutant spores (Table 1). Full agreement is not expected,

because of the loss of coat protein, which should decrease

�

R1ncr and increase fcr.

Upper Bound on Core Water Residence Time in Mutant Spores. Under fast-exchange

conditions, we can infer an upper bound on the mean residence time of a water molecule in

the core:

10

�

τcr <1

(1− ′ f cr )(R1cr − R1

ncr )=

′ f cr + 1/(α −1)(1− ′ f cr ) R1

[10]

where the last form follows from the fast-exchange relation

�

R1 = ′ f cr R1cr + (1− ′ f cr )R1

ncr

and we have defined

�

α ≡ R1cr /R1

ncr . At low frequencies, 2H relaxation is produced by

internal water molecules so the local relaxation rates are proportional to the protein

concentration (protein:water mass ratio) in core and ncr regions. The fraction spore protein

located in the core,

�

′ f pcr = 1− ′ f p

ncr , is obtained from Eq. [9] as 0.86. The fraction sample

water located in the core,

�

′ f cr , should be higher than for WT spores (with fcr = 0.12); we set

�

′ f cr = 0.20. (This precise value of

�

′ f cr is unimportant as long as it is not close to 1.) The

protein:water ratio, and the local 2H relaxation rate, is thus higher in the core than in the ncr

region by a factor α = (0.86/0.20) / (0.14/0.80) = 25. Inserting α and

�

′ f cr as well as R1 =

380 s–1 (at low frequencies, Fig. 2) into Eq. [10], we obtain the desired upper bound: τcr <

0.8 ms.

Chemical Composition of the Core. The EDTA-treated WT spore sample contains 1.20 g

D2O (g dry mass)–1 or 59.9 mmol water (g dry mass)–1, 12 % of which resides in the core

(Table 1), and 0.689 mmol Ca (g dry mass)–1 (Table S1), virtually all of which is confined

to the core. The water:Ca mole ratio in the core is thus 0.12 × 59.9 / 0.689 = 10.4.

Assuming that the core contains 45 ± 10 % (11) of the total protein mass in the sample

(Table S2), we obtain a protein hydration level of 0.12 × 1.20 × (18.02 / 20.02) / (0.45 ×

0.5046) = 0.57 ± 0.13 g H2O (g dry protein)–1. If 10 % of the sample dry mass is DPA (15)

our Ca analysis (Table S1) implies a Ca:DPA mole ratio of 0.689 / (100 / 167.1) = 1.15.

Since a DPA molecule only occupies 3 of the 6 – 8 coordination sites of a Ca2+ ion (16), as

much as 100 x (7 – 3 / 1.15) / 10.4 = 42 % of the core water might be Ca2+ coordinated. If

Ca2+ ions also coordinate proteins, nucleic acids and/or phospholipids, this fraction would

be smaller.

11

References

1. Halle B, Denisov VP, & Venu K (1999) Multinuclear relaxation dispersion studies

of protein hydration. Biological Magnetic Resonance., eds Krishna NR & Berliner

LJ (Kluwer Academic / Plenum, New York), Vol 17, pp 419–484.

2. Koenig SH & Brown RD (1985) Relaxation of solvent protons and deuterons by

protein-bound Mn2+ ions. Theory and experiment for Mn2+-concanavalin A. J.

Magn. Reson. 61:426–439.

3. Bertini I, Luchinat I, & Parigi G (2005) 1H NMRD profiles of paramagnetic

complexes and metalloproteins. Adv. Inorg. Chem. 57:105–172.

4. Kowalewski J, Kruk D, & Parigi G (2005) NMR relaxation in solution of

paramagnetic complexes: Recent theoretical progress for S ≥ 1. Adv. Inorg. Chem.

57:41–104.

5. Helm L, Nicolle GM, & Merbach AE (2005) Water and proton exchange processes

on metal ions. Adv. Inorg. Chem. 57:327–379.

6. Strandberg E & Westlund P-O (1999) Paramagnetic proton nuclear spin relaxation

theory of low-symmetry complexes for electron spin quantum number S = 5/2. J.

Magn. Reson. 137:333–344.

7. Aime S, et al. (2002) Relaxometric evaluation of novel manganese(II) complexes

for application as contrast agents in magnetic resonance imaging. J. Biol. Inorg.

Chem. 7:58–67.

12

8. Johnstone K, Stewart GSAB, Barratt MD, & Ellar DJ (1982) An electron

paramagnetic resonance study of the manganese environment within dormant spores

of Bacillus megaterium KM. Biochim. Biophys. Acta 714:379–381.

9. Leuschner RGK & Lillford PJ (2000) Effects of hydration on molecular mobility in

phase-bright Bacillus subtilis spores. Microbiol. 146:49–55.

10. Nakashio S & Gerhardt P (1985) Protoplast dehydration correlated with heat

resistance of bacterial spores. J. Bacteriol. 162:571–578.

11. Goldman RC & Tipper DJ (1978) Bacillus subtilis spore coats: Complexity and

purification of a unique polypeptide component. J. Bacteriol. 135:1091–1106.

12. Ghosh S, et al. (2008) Characterization of spores of Bacillus subtilis that lack most

coat layers. J. Bacteriol. 190:6741–6748.

13. Persson E & Halle B (2008) Cell water dynamics on multiple time scales. Proc.

Natl. Acad. Sci. USA 105:6266–6271.

14. Park S & Saven JG (2005) Statistical and molecular dynamics studies of buried

waters in globular proteins. Proteins 60:450–463.

15. Murrell WG (1967) The biochemistry of the bacterial endospore. Adv. Microbial

Physiol. 1:133–251.

16. Ikeda T, Boero M, & Terakura K (2007) Hydration properties of magnesium and

calcium ions from constrained first principles molecular dynamics. J. Chem. Phys.

127:074503.

13

Table S1. Elemental abundance [µmol (g dry mass)–1] in spore samples

Element WT WT + EDTA Mutant

Ca 768 689 1003

Fe 1.2 1.1 < 1

K 50.4 52.4 56.5

Mg 119 75 33

Mn 84.3 57.0 87.9

Si 10.4 9.5 6.0

Zn 0.67 0.32 0.34

Table S2. Results of protein analysis on spore samples

WT WT + EDTA Mutant

Protein content a / g (g dry mass)–1 0.502 0.505 0.346

Mass-weighted residue molar mass b / g mol–1 120.2 120.3 117.5

a Including Cys and Trp, with abundancies from Table S3.

b Computed from the data in Table S3.

14

Table S3. Amino acid composition (mole %) of spore samples

Amino acid WT WT + EDTA Mutant

Ala 11.6 11.5 18.9

Arg 5.2 5.2 6.5

Asx a 8.9 8.8 8.8

Cys b (0.8) (0.8) (0.8)

Glx a 13.0 12.8 20.1

Gly 12.3 12.2 7.4

His 2.9 2.9 1.3

Ile 3.6 3.5 3.6

Leu 5.0 5.0 5.3

Lys 6.3 6.1 5.3

Met 1.0 1.5 1.9

Phe 4.5 4.5 2.8

Pro 3.2 3.7 2.4

Ser 5.2 5.1 3.7

Thr 4.4 4.3 3.8

Trp b (1.0) (1.0) (1.0)

Tyr 6.0 6.0 1.6

Val 5.1 5.1 4.8

a Asx = Asn + Asp; Glx = Gln + Glu

b Cys and Trp were not analyzed and were assigned the same mole fractions as for the

average of all proteins predicted to be encoded by the genome of B. subtilis strain 168

(http://www.pasteur.fr/~tekaia/aafreq). Because these are the least abundant residues, this

approximation hardly affects the remaining mole fractions.

15

Fig. S1. Water 2H paramagnetic relaxation enhancement from WT B. subtilis spores,

obtained as the difference in R1 between untreated (black symbols in Fig. 2) and EDTA-

treated (blue symbols in Fig. 2) spores. The curve resulted from a fit of the 4 parameters in

the model defined by Eqs. [2] – [6].

16

Fig. S2. (A) Water-17O magnetization recovery at 81.3 MHz from EDTA-treated WT

spores. (B) Water-2H magnetization recovery at 92.1 MHz from EDTA-treated mutant

spores. The curves resulted from mono-exponential fits with residuals as shown in the

upper frames.

17

18

Fig. S3. Water-2H magnetization curves at 1.5 kHz (A), 20 kHz (B), 2.5 MHz (C) and 5.8

MHz (D) from EDTA-treated WT B. subtilis spores. The data were acquired on the fast

field-cycling instrument using the pre-polarized (A–C) or non-prepolarized (D) sequences.

The curves resulted from mono-exponential fits with residuals as shown in the upper

frames.