FOM-Institute AMOLF, Amsterdam, the Netherlands (bonn@amolf.nl) Water at biological membranes: structure, dynamics and biomolecular sensing Mischa Bonn FOM-Institute AMOLF, Amsterdam, the Netherlands (bonn@amolf.nl) Stockholm, 15/5/08

Acknowledgements Avishek Ghosh, Marc Smits, Jens Bredenbeck, Maria Sovago, Sjors Wurpel, Martin Sterrer, Michiel Muller, Huib Bakker &co

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

Biological membranes Outside cell Inside cell lipids proteins water hydrophobic alkyl chains hydrophilic “polar” head group Phospholipids: building block of cell membrane

Biological membranes Outside cell Inside cell water proteins lipids Important biomolecular interactions at cell membrane Functionality = interplay between lipids/proteins/water

Membrane-bound water Most information on membrane-bound water has been obtained through MD simulations: residence time: exchange with bulk on ps-ns timescales preferential orientation heterogeneity?? MD simulation (~100 ps) Schulten group UIUC

the challenge Membrane-bound water The one-molecule thick layer of water (~3 Å) at the membrane Many biological reactions happen at membrane surfaces: structure and dynamics of membrane-bound water important for these processes the challenge Direct probing of membrane-bound water

How to investigate structure of 1 ML water? Model system: Lipid monolayer on water Self-assembled monolayers of lipids: good membrane model Compare lipid-water interface with air-water interface: distinguish effects due to termination of bulk from lipid- specific effects water Dimyristoylphosphatidylserine (DMPS) air How is the water at the water-air and water-lipid interface different? To understand the difference we need a technique to specifically probe the interfacial water- here we´re talking about water layers as thick as ~3 A° from the surface! This is a challenge! So how do we go about this? water How to investigate structure of 1 ML water?

How to investigate water structure? Water displays strong variation in H-bond strengths, which affect O—H stretch vibration O—H stretch vibration is a marker of local water environment 3600 3500 3400 nO-H (cm-1) 3300 3200 3100 2.6 2.8 3 O—O distance (Å) Traditionally, vibrational IR spectroscopy has been used to understand the H-bond network in the system, which is a signature of the molecular structural environment. Stronger the H-bond, weaker is the O-H vibration as you can see in this cartoon. So using vibrational IR spectroscopy, this O-H vibration can be observed, and thus observing the H-bond. The plot on the right shows how the O-H frequency depends on the O—O bond distance. Make transition – move this whole slide up . Show cartoon of lipid-water, air-water and “how to detect that ONE molecular thick water?” How to detect ONE MOLECULAR layer of water?

Sum Frequency Generation (SFG) provides vibrational spectrum of surface monolayer IR vis SFG H c(2)=0 O H c(2)=0 The sum-frequency process is a second-order nonlinear process. This can only occur in media lacking inversion symmetry. In the bulk of the water, the environment is centrosymmetric, and there will be no SFG generation. At the surface, however, the symmetry is broken per definition, and we observe a sum frequency signal. This allows us to record ~1 monolayer thick water: of the order of about 3 angstroms. Second water molecule in bulk  “forbidden in centrosymmetric” pops up and cross on the energy level diagram.. v = 1 forbidden in bulk water v = 0

SFG spectrum of water/air interface vs bulk IR spectrum Hydrogen bonded strong ‘ice-like’ weak ‘water-like’ Free OH 0.6 0.5 0.4 I SFG OH-stretch vibrations of H2O 0.3 0.2 For pure D2O, the linear, bulk vibrational spectrum consists of broad overlapping bands that respresent different distributions of hydrogen-bonded water. In the SFG spectrum, the most notable feature is the sharp peak at high frequencies. The frequency is comparable to that of gas-phase water and can therefore be attributed to OD groups that are not hydrogen-bonded. In other words: it is OD bonds “sticking out of the surface”. The other peaks can be assigned to water in two different environments: either weakly or strongly hydrogen bonded. 0.1 Bulk IR 0.0 3000 3200 3400 3600 3800 -1 IR frequency (cm )

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

SFG spectrum of the Air-D2O Interface OH-stretch vibrations of D2O Hydrogen bonded Free OD free 0.6 0.5 0.4 I SFG 0.3 For pure D2O, the linear, bulk vibrational spectrum consists of broad overlapping bands that respresent different distributions of hydrogen-bonded water. In the SFG spectrum, the most notable feature is the sharp peak at high frequencies. The frequency is comparable to that of gas-phase water and can therefore be attributed to OD groups that are not hydrogen-bonded. In other words: it is OD bonds “sticking out of the surface”. The other peaks can be assigned to water in two different environments: either weakly or strongly hydrogen bonded. 0.2 0.1 0.0 2200 2400 2600 2800 -1 IR frequency (cm )

SFG spectrum of D2O-lipid interface terminal-CH3 CH3 as CH3 ss water/ lipid CH3 ss - + all-trans CH2: SFG invisible water/air - Negative phosphate Positive ammonium Negative carboxylate OD ss Dimyristoylphosphatidylserine Why is the signal so much larger at the water-lipid interface?

We can use the large signal from membrane-bound water…… water-lipid water-air

Large signal also observed for cationic lipid monolayer D2O + DPTAP (cationic lipid) Large signal related to charge? Lipid CH Here we see the water spectrum again. When we form a charged lipid monolayer at the air-water interface we note a dramatic spectral change. The lipid we used is DPTAP, which is a lipid with a positively charged headgroup. Apart from the signals that can be ascribed to the CH groups of the alkyl chains, we observe an order of magnitude increase in water signals. D2O/air dipalmitoyl trimethyl ammonium propane

Water Alignment + + + + + + + + + + + + + + + + + + There are two possible explanations for the dependence of the SFG water signal on the surface potential. The first is that the electric field of the positively charged interface is able to orient the water molecules. For our system, it can be calculated that the electric field might orient at most two water layers. As we remember, the SFG signal, is only generated in non-centrosymmetric medium, so by alligning the water molecules more signal can be generated. In centro-symmetric environment (bulk water)

Water Alignment + + + + + + + + + + + + + + + + + + in headgroup region due to water alignment  large SFG signal!

Adding NaCl Lowers Water Signal The water signal was found to depend on salt concentration in the subphase. As shown here, adding mM conc of salt reduces the water signal. This shows that the strength of the water signal is determined by the strength of the electric field generated by the cationic headgroups. The added salt screens the headgroup charges and thereby reduces the surface potential, which in turn has an effect on the water signal. We will discuss the mechanism behind this shortly, but first, let’s see if the water signal strength indeed is consisten with charge screening.

Water Alignment + + + + + + + + + + + + + + + + + + in headgroup region due to water orientation  large SFG signal!

Water Alignment + + + + + + + + + Cl- Cl- Cl- Cl- Cl- + + + + + + + + + Cl- Cl- Cl- Cl- Cl- There are two possible explanations for the dependence of the SFG water signal on the surface potential. The first is that the electric field of the positively charged interface is able to orient the water molecules. For our system, it can be calculated that the electric field might orient at most two water layers. As we remember, the SFG signal, is only generated in non-centrosymmetric medium, so by alligning the water molecules more signal can be generated. Lipid field is shielded by Cl- ions: disorder is restored

Water Signal Depends on Surface Potential Gouy-Chapman: Water Signal surface charge density conc. electrolyte Here, the amplitude of the OD stretch vibrations is plotted as a function of salt concentration. We also calculated for our system the surface potential, which in the simplest approximation is given by the Gouy-Chapmann theory. This relation gives the surface potential as a function of the charge density of the interface (which is fixed for our case) and the concentration electrolyte. Indeed, our data corresponds well with this charge screening model. We can use the water signal for DNA detection: DNA is a poly-electrolyte

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

Water Signal Depends on Surface Potential NaCl Here, the amplitude of the OD stretch vibrations is plotted as a function of salt concentration. We also calculated for our system the surface potential, which in the simplest approximation is given by the Gouy-Chapmann theory. This relation gives the surface potential as a function of the charge density of the interface (which is fixed for our case) and the concentration electrolyte. Indeed, our data corresponds well with this charge screening model.

We can use the water signal for DNA detection Important for life sciences, forensic and medical diagnostics Most common approach: labeling with (fluorescent) marker  optical detection with high sensitivities, but requires extensive (bio-)chemical treatment.  much interest in label-free DNA detection schemes

Label-free DNA detection schemes publication approach sensitivity detection Fritz, 2002 PNAS 99, 14142 lithography 1 nM electronic Pouthas, 2004 APL 84, 1594 nano-lithography 10 µM electronic Hahm, 2004 Nano Lett. 4, 51 nanowires 10 fM electronic Star, 2006 PNAS 103, 1594 carbon nanotubes 1 nM

General principle of label-free schemes Field-effect transistor-type geometries that rely on changes in surface fields due to adsorption of anionic DNA Si depletion region =PO- 3 P Still requires extensive effort – no longer on DNA side, but on detection side (lithographical and biochemical) P P P P P P Fritz, PNAS 99, 14142 (2002). Copyright ©2002 by the National Academy of Sciences

Our scheme Also employ poly-anionic character of DNA Detect optically – not electrically – using water as the reporter for the presence of DNA

Adding DNA a polyanion -DNA: 48502 bp 0 pM 12 pM 47 pM 94 pM Instead of a simple 1:1 electrolyte, we were also interested in the interaction between the charged lipid monolayer and DNA, which is a polyanionic molecule. As shown here, adding DNA to the subphase also decreases the water signal, but already at pM concentrations?

Water Signal Depends on Surface Potential Gouy-Chapman: Water Signal charge density conc. electrolyte Here, the amplitude of the OD stretch vibrations is plotted as a function of salt concentration. We also calculated for our system the surface potential, which in the simplest approximation is given by the Gouy-Chapmann theory. This relation gives the surface potential as a function of the charge density of the interface (which is fixed for our case) and the concentration electrolyte. Indeed, our data corresponds well with this charge screening model.

DNA Concentration Dependence: picomolar sensitivity! NaCl Gouy-Chapman Boltzmann Langmuir If we look at the concentration dependence, we see that the the water signal is 9 orders of magnitude more sensitive to the DNA concentration than for NaCl concentration. We could model this behaviour by assuming a cooperative Langmuir adsorption of DNA bases to cationic sites in the monolayer. The surface concentration of “ligands” L is given by a Boltzman equation. From this model we derive an association constant and cooperativity factor that are in good agreement with thermodynamic measurements on bulk complexation. association constant K  represents the unbound fraction of the cationic lipid = fraction of available ‘adsorption’ sites [DNA]0=near-surface concentration of DNA [DNA]=bulk concentration of DNA K=association constant n= cooperativity constant Following Miranda, CPL 1998

DNA Concentration Dependence: picomolar sensitivity! NaCl -DNA If we look at the concentration dependence, we see that the the water signal is 9 orders of magnitude more sensitive to the DNA concentration than for NaCl concentration. We could model this behaviour by assuming a cooperative Langmuir adsorption of DNA bases to cationic sites in the monolayer. The surface concentration of “ligands” L is given by a Boltzman equation. From this model we derive an association constant and cooperativity factor that are in good agreement with thermodynamic measurements on bulk complexation. association constant K  represents the unbound fraction of the cationic lipid per base Cooperativity n=3 Highly sensitive DNA detection, readily made specific

Conclusion (intermediate…) Sensitive, labelfree DNA detection by optical detection of water vibrations (also useful for toxins). What determines the shape of the SFG spectra; what does it tell us about the membrane-bound water? water-lipid water-air

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

Comparison of air-water and lipid-water interfaces H-bonded OH H-bonded OH SFG intensity SFG intensity 3000 3200 3400 3600 3000 3200 3400 3600 IR frequency (cm-1) IR frequency (cm-1) Similar spectra: same interfacial water?

SFG spectrum of water/air interface vs bulk IR spectrum ? ? Hydrogen bonded free strong strong ‘ice-like’ weak ‘water-like’ 0.6 Free OH weak 0.5 0.4 I SFG 0.3 0.2 Interpretation in terms of quasi-static sub-structures1 at surface doubtful, given recent experimental results for bulk water2 For pure D2O, the linear, bulk vibrational spectrum consists of broad overlapping bands that respresent different distributions of hydrogen-bonded water. In the SFG spectrum, the most notable feature is the sharp peak at high frequencies. The frequency is comparable to that of gas-phase water and can therefore be attributed to OD groups that are not hydrogen-bonded. In other words: it is OD bonds “sticking out of the surface”. The other peaks can be assigned to water in two different environments: either weakly or strongly hydrogen bonded. 0.1 Bulk IR 0.0 3000 3200 3400 3600 3800 -1 IR frequency (cm ) 1 Shen et al., PRL 1993; 2Woutersen, Nature, 1999; Cowan, Nature 2005.

Remember the H2 molecule? asym O-D O-D symm (same water molecule)

Hypotheses Which hypothesis is correct? ‘Ice/Liquid-like’ hypothesis O-H coupling hypothesis Strong Weak SS AS Which hypothesis is correct?

‘Ice/Liquid-like’ hypothesis Testing the hypothesis using isotopic dilution experiments ‘Ice/Liquid-like’ hypothesis Strong Weak D2O HDO the amplitude of the two-peaks should change over the whole spectrum

Testing the hypothesis using isotopic dilution experiments ‘Ice/Liquid-like’ hypothesis O-H coupling hypothesis Strong Weak SS AS D2O D2O HDO HDO SS AS SFG spectrum should change from double-peak to a single-peak structure

H-bonded doublet is due to substructures?  Isotopic dilution pure D2O H2O:D2O 2:1 H2O:HDO:D2O 4:4:1 ? ice- like water- like H I SFG 2800 2600 2400 2200 IR frequency (cm-1) 2800 2600 2400 2200 IR frequency (cm-1) inconsistent with substructures I SFG D I SFG consistent with substructures Splitting of peak by distinct substructures or intramolecular coupling (symm. & asym. stretches)??

H-bonded doublet is NOT due to substructures O-D 2 water/ lipid ÷10 C-H pure D2O SFG intensity (arb. units) 1 H2O:D2O 2:1 water/air pure D2O H2O:D2O 2:1 2000 2200 2400 2600 2800 3000 3200 -1 IR frequency (cm )

D2O  HOD switches off coupling H-bonded doublet is due to Fermi resonance bend overtone (dOH=2) * * water/air 2000 2200 2400 2600 2800 3000 -1 IR frequency (cm ) asym H O-D O-D * symm dOH=2 D * D2O  HOD switches off coupling

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction Structure Sensing Less structure Dynamics Structure from dynamics Conclusions

Double-peaked structure is due to Fermi resonance… lipid air water water H-bonded OH H-bonded OH SFG intensity SFG intensity 3000 3200 3400 3600 3000 3200 3400 3600 IR frequency (cm-1) IR frequency (cm-1) …and contains little info on structure. Broad and featureless spectra hide water dynamics.

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction A bit of structure Sensing Structure?? Dynamics Structure from dynamics Conclusions

Fs interfacial vibrational dynamics: From c(2) to c(4) Surface-specific femtosecond time-resolved spectroscopy* IR pump (100 fs) IR (100 fs) vis (ps) SFG The SFG signal decreases due to depletion of ground state. Recovery reflects vibrational relaxation. wait SFG pump v = 1 twait v = 0 *Backus… Bonn, Science 2005 ; McGuire, Science 2006; Smits… Bonn PRL 2007

Comparison air-water and lipid-water interfaces SFG intensiteit IR frequency (cm-1) 3400 3200 3000 3600 H-bonded OH air water H-bonded OH SFG intensiteit 3000 3200 3400 3600 IR frequency (cm-1) Record changes in SFG signal at specific frequencies as a function of time after vibrational excitation

Dynamics at the water-air interface 1.3 1.2 1.1 1.0 0.9 DSFG signal 1500 1000 500 pump-probe delay (fs) 3500 cm-1 3400 cm-1 3300 cm-1 3200 cm-1 Dynamics at water/air interface: Distinct spectral response Two time constants No pump-polarization dependence just like bulk!

Ultrafast energy transfer between surface and bulk water IR pump Δt = T1 Δt = teq Förster# #Woutersen, Nature 1999;*Cowan, Nature 2005; @Lock, J. Chem. Phys 2002

Ultrafast energy transfer between surface and bulk water  (cm-1) T1 (fs) eq (fs) 3500 3400 3300 3200 190* 400@ 1.3 1.2 1.1 1.0 0.9 DSFG signal 1500 1000 500 pump-probe delay (fs) 3500 cm-1 3400 cm-1 3300 cm-1 3200 cm-1 air Due to ultrafast energy transfer, excitation samples many differently H-bonded molecules: 1 relaxation time! water IR pump Förster# Δt = 0 Δt = T1 Δt = teq #Woutersen, Nature 1999;*Cowan, Nature 2005; @Lock, J. Chem. Phys 2002

Comparison air-water and lipid-water interfaces H-bonded OH H-bonded OH SFG intensity SFG intensity 3000 3200 3400 3600 3000 3200 3400 3600 IR frequency (cm-1) IR frequency (cm-1) Similar spectra: same interfacial water?

Different dynamics for the two interfaces 1.3 3500 cm-1 1.4 3500 cm-1 1.2 1.3 3400 cm-1 3400 cm-1 1.2 1.1 DSFG signal 3300 cm-1 1.1 1.0 DSFG signal 3300 cm-1 1.0 3200 cm-1 0.9 0.9 3200 cm-1 500 1000 1500 500 1000 1500 pump-probe delay (fs) pump-probe delay (fs) Dynamics at water/air interface reflect bulk behaviour; Not so for water/lipid interface

Membrane-bound water is energetically isolated from the bulk  (cm-1) T1 (fs) 3500 3400 3300 3200 570 430 130 <100 No evidence for spectral diffusion single exponential recovery large polarization dependence 1.4 3500 cm-1 1.3 3400 cm-1 T1 (fs) 800 600 400 200 3500 3300 3100 data 1.2 Consistent with Righini results model* *Staib, A.; Hynes, J. T. Chem.Phys.Lett. 1997 1.1 3300 cm-1 DSFG signal 1.0 0.9 3200 cm-1 500 1000 1500 Vibrational frequency (cm-1) pump-probe delay (fs)

Membrane-bound water is energetically isolated from the bulk tFörster< 50 fs tFörster> 1ps How? Why? Answer lies in steady-state SFG spectra of membrane-bound water

Water bound to an anionic lipid monolayer Sign of c(2)  molecular orientation OD ss(+) CH3 ss(+) CH3 as(-) ODss (-) - CH3 ss(+) - - + ODss Ionterfentie en MEM – NIET: take my word… Water oriented with oxygen towards negative lipid??

Water bound to differently charged lipid monolayers - I SFG - - ??? - I SFG + + -1 IR frequency (cm )

Maximum Entropy Method - illustration Fitting the data is inherently ambiguous. MEM is unambiguous. For CARS: E. Vartiainen, H.A. Rinia, M. Müller and M. Bonn Optics Express 14, 3622-3630 (2006).

Orientation of water at negative glass interface Maximum Entropy Method - check Orientation of water at negative glass interface =orientation at positive lipid interface ! 20 water-glass interface (pH=10) negatively charged interface positive lipid SiO2 - - - - - - - 15 negative lipid I SFG + + 10 + + - - 5 2000 2200 2400 2600 2800 3000 3200 -1 IR frequency (cm )

Water orientation at a neutral lipid interface..? + I SFG + + O O O O H N O C H 3 P Water in contact with DPPC (neutral lipid) behaves like it’s in contact with an anionic lipid! DPPC

We are looking at headgroup water oriented the right way, but at other side of charge This is the water we are observing with SFG. Spatially isolated from bulk  slow Forster energy transfer

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction A bit of structure Sensing Less structure Dynamics Structure from dynamics

Structure from dynamics Water at biological membranes: structure, dynamics and biomolecular sensing Outline Introduction A bit of structure Sensing Less structure Dynamics Structure from dynamics Conclusions

Conclusions Membrane = lipids Membrane = lipids + water Water surface structure is simpler than has been thought (PRL 100, 173901 (2008)) At many interfaces, surface water exchanges vibrational energy rapidly with the underlying bulk (PRL 98, 098302 (2007)) In contrast, membrane-bound water does not show fast energy exchange – it does not just terminate the bulk and constitutes an intrinsic part of membrane (JACS 129, 9608 (2007)) Sensitive, labelfree DNA detection by optical detection of water vibrations (JACS 129, 8420 (2007)) Membrane = lipids Membrane = lipids + water