Supplementary Materials for

Transcription

1 advances.sciencemag.org/cgi/content/full/2/4/e /dc1 Supplementary Materials for Ice-nucleating bacteria control the order and dynamics of interfacial water Ravindra Pandey, Kota Usui, Ruth A. Livingstone, Sean A. Fischer, Jim Pfaendtner, Ellen H. G. Backus, Yuki Nagata, Janine Fröhlich-Nowoisky, Lars Schmüser, Sergio Mauri, Jan F. Scheel, Daniel A. Knopf, Ulrich Pöschl, Mischa Bonn, Tobias Weidner Published 22 April 2016, Sci. Adv. 2, e (2016) DOI: /sciadv The PDF file includes: Bicinonic Acid Assay (BCA) Surface tension experimental details Surface tensions for control samples at RT XPS experimental details fig. S1. Fitting results for SFG spectra. fig. S2. SFG spectra recorded in the amide I region for P. syringae at the air-water interface at RT and 5 C. fig. S3. Time traces of surface tension changes recorded while the temperature was changed in the trough for D2O and Snomax at the air-water interface. fig. S4. Surface tension changes relative to the tension at 22 C for pure D2O and the Snomax sample for sample temperatures studied with SFG. fig. S5. XPS spectra of P. syringae films adsorbed at the water-air interface at 22 and 5 C and transferred to gold-coated silicon chips following the Langmuir- Schaefer (54) method. fig. S6. XPS survey spectrum of the Snomax sample. fig. S7. Droplet freeze assay for the P. syringae sample used in this study. fig. S8. Illustration of the simulation box. (2) fig. S9. Calculated SFG intensity χ 2 xxz for the O D stretch chromophores within 15 Å of the center of mass of the IN dimer in the z direction. fig. S10. Snapshot of the MD simulation near the Thr-rich region (Fig. 3, region 2, main text) of an IN site. fig. S11. Time-resolved difference sum frequency spectra for control substances.

2 fig. S12. Energy transfer processes at the water-afp interface. table S1. Elemental compositions (atom %) determined by XPS of protein monolayers deposited onto gold-coated silicon chips with the Langmuir-Schaefer technique. table S2. Time scales (in femtoseconds) extracted from the coupled differential equation fits (only down was allowed to float; the others were fixed). References (55 68)

3 Supplementary Materials: Bicinonic Acid Assay (BCA) Total amount of protein in samples were estimated using the BCA TM Protein Assay Kit from Thermo Scientific (Rockford, IL USA). All procedures were performed according to the manual instructions. Briefly, a series of standards were generated using bovine serum albumin. Snomax, lyophilized P. syringae (DSM10604) and the standards were treated with bicinonic acid reagent and measured in a Multiskan Go spectrophotometer (Thermo Scientific). The estimated protein concentrations per 0.01 g were very similar for Snomax and P. syringae cells, 5.3 mg/ml and 5.5 mg/ml, respectively. Surface Tension Experimental Details Surface tension has been measured using a Langmuir tensiometer (Kibron, Finland). P. syringae solutions at 0.1mg/ml have been prepared in D2O. The temperature controlled trough was thoroughly cleaned with acetone, ethanol and milli-q water, and dried under nitrogen stream prior to measurements. The temperature dependent surface tension measurement of D2O was repeated five times, and the surface tension measurement of Snomax solution was repeated 4 times. The tensiometer was calibrated using pure D2O at room temperature (22 C). The temperature dependence was recorded after the surface tension has reached equilibrium within the accuracy of the measurement. Surface tensions for control samples at RT Sample Surface tension Lysozyme 7 mn/m Protein extract 14 mn/m DPPG 14 mn/m AFPIII 13 mn/m INP fragment 14 mn/m XPS Experimental Details Gold-coated silicon chips used as substrates were produced in-house. Prior to the 40 nm gold layer a 1 nm chromium adhesion layer was evaporated onto the chips to increase the stability of the gold film. Surface elemental compositions of the samples were analyzed with X-ray photoelectron spectroscopy (XPS) (Axis Ultra spectrometer, Kratos, Manchester, England) using a monochromatic Al X-ray source. All spectra were acquired in the hybrid mode with a 0 take-off angle, which is defined as the angle between the surface normal and the axis of the analyzer lens. The compositional survey scans and carbon, nitrogen and oxygen narrow region scans were acquired using pass energies of 80 ev. C 1s, N 1s and O 1s high-resolution spectra were collected using pass energies of 20 ev. Each scan was performed on three different spots per sample. Surface elemental compositions were determined using the CasaXPS software (Casa Software Ltd.). High-resolution spectra were fitted using the XPSpeak package.

4 cm-1 OD 2880 cm-1 CH Lipid 2880 cm-1 CH3 Normalized SFG Intensity (a.u.) Normalized SFG Intensity (a.u.) Temperature / C Temperature / C fig. S1. Left: Changes of the 2880 cm -1 CH3 symmetric stretching mode (red) and the 2390 cm -1 OD stretching mode (black) with decreasing temperature. The CH mode is significantly less affected by the temperature changes compared with the water mode. Right: Analysis of the lipid resonance near 2880 cm -1 related to the CH3 symmetric stretching mode.

5 SFG Intensity / a.u IR Frequency / cm -1 fig. S2. SFG spectra recorded in the amide I region for P. syringae at the air-water interface at RT and 5 C. The amide region is representative of protein backbone modes ( cm -1 ) and lipid carbonyl groups ( cm -1 ). The protein signal visible in the spectra remains unchanged when the samples is cooled to 5 C, which indicates the protein structure and composition is not affected by the lower temperature. The absence of lipid signals indicates that any lipids present at the air-water interface are in a disordered state (20) o C to 5 o C Surface tension / mn/m o C to 10 o C 22 o C to 15 o C Time / s fig. S3. Time traces of surface tension changes recorded while the temperature was changed in the trough for D2O and the Snomax at the air water interface. The data were normalized to unity for better comparison. The D2O sample shows the tension response expected for temperature changes (67). D 2 O Snomax

6 Temperature-induced changes or film reassembly takes place on a timescale of hours and would be expected to significantly change the rate observed in the surface tension curve for Snomax. Since the Snomax time trace is almost identical to the D2O traces, we can conclude that additional adsorption or exchange of molecules with the subphase does not occur within the error margin of the experiment. This is in good agreement with the XPS results shown below. Normalized surface tension / m/nm D 2 O Snomax Temperature / o C fig. S4. Surface tension changes relative to the tension at 22 C for pure D2O and the Snomax sample for sample temperatures studied with SFG. Given the differences in ph and ionic strength between P. syringae and the control, which affect the temperature dependence of the surface tension, the changes of the surface tension are remarkably similar. This supports the view that there is no detectable additional adsorption of material to the air water interface due to changes in temperature (68).

7 Normalized XPS intensity O 1s 22 C 5 C N 1s 22 C 5 C C 1s XPS 22 C 5 C Binding Energy (ev) fig. S5. XPS spectra of P. syringae films adsorbed at the water air interface at 22 C and 5 C and transferred to gold coated silicon chips following the Langmuir-Schaefer (54) method. The data was acquired to test whether the surface coverage or composition changes with temperature. The highresolution spectra, together with the composition tables (table S1) show that within the accuracy of the XPS measurements, there are no differences between the films at different temperatures. The data show that temperature-related changes within the SFG spectra are not caused by changes in the amount or type of molecules present at the air-water interface. Au 4f Intensity (a.u.) O KLL Au 4s Au 4p Au 4p O 1s N 1s Au 4d C 1s Au 5s Au 5p Binding Energy (ev) fig. S6. XPS survey spectrum of the Snomax sample.

8 table S1. Elemental compositions (atom % a ) determined by XPS of protein monolayers deposited onto gold coated silicon chips with the Langmuir-Schaefer technique. atom % Sample Au C N O 22 C 29.5(0.6) 49(0.4) 9.1(0.1) 12.4(0.2) 5 C 29.7(0.6) 48.5(0.4) 9.4(0.2) 12.4(0.4) a Average elemental compositions (three spots per sample) with standard deviations in parentheses. Droplet Freezing Assay of P. syringae g of Snomax was resuspended in 1 ml of ultrapure water (18.2 MΩ cm) from a water purification system (Thermoscientific Barnstead GenPure xcad plus). Additionally the water was autoclaved and filtered to a 0.1 µm PES syringe filter (Corning ). The resuspended Snomax was then filtrated through a 0.1 µm PES syringe filter (Acrodisc ). A dilution series was made from 10-6 to and every dilution was spotted in 50 µl droplets on a 96 well PCR tray (Axon ). A customized PCR-plate thermal block (VWR ) was tempered by a cooling bath (Julabo Presto A30) and the frozen droplets were counted. For every dilution 32 droplets were counted resulted in a total amount of 184 single droplets. The concentration of IN in the suspension was examined after the modified Vali formula (see Eq. S1, (64)) (1n m [g 1 ] = ln (1 f ice ) Vwash V drop F dil m Snomax (S1) f ice, fraction of frozen droplets V wash, volume of water V drop, droplet volume F dil, dilution m myc, weight (g) Characterization of the Snomax IN To characterize the IN of the Snomax we applied a customized droplet freezing assay and measured the frozen droplets dependent on the temperature in water. The frozen droplets were plotted by the number of IN to the temperature (fig. S7). The resulting graph showed a typical freezing curve for Snomax with two freezing optima corresponding to the class A IN of Snomax at K and class C IN at K respectively. The resulting curve and average values are comparable to frequently published data (24). The total amount of measured droplets was 184 in serial dilutions from 10-6 to Thus possible contents of salt or other low molecular contaminants resulting from generation of Snomax are insignificant in terms of freezing point depression.

9 1E17 1E15 IN/g [Snomax] 1E13 1E11 1E9 1E T [ K] fig. S7. Droplet freeze assay for the P. syingae sample used in this study. Details of the MD Simulation The molecular dynamics simulations were performed with Gromacs version (59). The inaz model was simulated with the AMBER99SB-ILDN force-field, 4894 TIP4P-Ew water molecules, and 16 Na atoms for charge neutralization. TIP4P-Ew was chosen since it is one of the few water models that correctly predicts ice Ih to be the thermodynamically stable phase (60, 61). Particle-mesh Ewald was used for the Coulomb interactions with a cut-off of 1 nm, while a 0.9 nm cut-off was used for the van der Waals interactions. The size of the simulation cell is set to Å Å Å with vacuum space included at the top and bottom region of the simulation cell to avoid spurious periodicity effects. The inaz model was completely hydrated in the box (see fig. S8). The geometry of the system was first relaxed with 5000 steps of a steepest descent algorithm. From this point all our simulations were run in triplicate, each starting from the same energy-minimized geometry but independent thereafter. Equilibration was achieved in two steps. First an NVT ensemble was simulated for 100 ps with a global stochastic thermostat (62) and a relaxation time of 0.1 ps. This was followed by 100 ps of simulation in the NPT ensemble with the global stochastic thermostat and a Berendsen barostat at 1 bar with a relaxation time of 1 ps. The initial box dimensions were 4.44 nm by 5.72 nm by 7.25 nm, and position restraints were used on the inaz atoms. Following equilibration, 50 ns of NVT simulation were performed with the backbone atoms of the inaz model frozen. A 2 fs time step and a leapfrog integrator were utilized for all simulations. Snapshots from these trajectories were then used to calculate the SFG response.

10 fig. S8. Illustration of the simulation box. The inaz protein is completely hydrated. The size of the simulation cell was set to Å Å Å with vacuum space included. The vacuum space is not shown here for clarity of presentation. Calculation of SFG Spectra The SFG spectra for D2O were calculated using the trajectories obtained from the MD simulations for the dimer of the IN active site of inaz and H2O. The resonant part of the SFG intensity with x, x, and z polarization directions for the SFG response signal, visible pulse, and IR pulse, respectively, can be calculated by (2) (ω) = a n,xx (0)m n,z (0) n (S2) χ xxz ω n (0) ω i 2T 1 where a n,xx (0), m n,z (0), and ω n (0) are the xx component of the transition polarizability, z component of the transition dipole moment and transition frequency of chromophore n respectively. The excited vibrational lifetime T1, which causes the broadening of signals, was assumed to be 1.3 ps. The z component of the transition dipole moment is represented as m z = μ x 10 u bond u z (S3) where μ, x 10, u bond, and u z are the magnitude of the dipole derivative, the 0-1 matrix element of the O- D stretch coordinate, the normalized O-D vector, and the unit vector in the z-direction, which is defined as the surface normal. Meanwhile the xx component of the transition polarizability is calculated by

11 a xx = x 10 ((α α )(u bond u x ) 2 + α ) (S4) where α and α are the magnitudes of the polarizability derivatives parallel and perpendicular to the O- D bond, respectively. μ, x 10, α, α, and ω n are calculated with the method of map developed by Skinner and co-workers (63, 64), where these values are represented as a function of the electric field, E, at the D atom in the direction of the O-D bond. α /α was set to 5.6. E was calculated by the O-D stretch chromophores within 15 Å of the center of mass of the IN dimer in the z-direction. The parameters in the maps for the TIP4P water model were adopted from Ref. (65). The intramolecular coupling ω 12 was also taken into consideration. The frequency shift caused by ω 12 is obtained by Δ = (ω 1 ω 2 ) 2 4 ω 2 12 ω 1 ω 2 2 (S5) where the coupled frequencies of ω 1 + Δ and ω 2 Δ when ω 1 > ω 2. The simulated SFG spectra are displayed in fig. S9. To exclude effects of D2O distant from the water/protein interface, only O D chromophores within 15 Å of the center of mass of the IN dimer in the z-direction were taken into consideration. fig. S9. Calculated SFG intensity χ xxz (2) (ω) 2 for the O D stretch chromophores within 15 Å of the center of mass of the IN dimer in the z direction. The condition is taken to exclude water molecules distant from the water/protein surface.

12 fig. S10. Snapshot of the MD simulation near the Thr-rich region (Fig. 3, region 2, main text) of an IN site. In agreement with previous reports (14, 18), a number of water molecules (magenta) are captured by the threonine side chains and highly ordered. At the same time, the threonine interaction isolates these clathrate water molecules from the adjacent water layer. Thus, it is highly unlikely that clathrate water molecules will be good templates for ice nucleation. This view is also supported by the comparatively low intensity of the calculated SFG spectra for this site. Model for the time-resolved data analysis Insights into the energy transfer dynamics can be obtained from the time resolved data by modeling the population of the more strongly and more weakly hydrogen bonded water molecules in the spectrum as a function of pump-probe time delay. Although, it is known that there is a continuum of vibrational states in the water band going from weakly to strongly hydrogen bonded water molecules, we split the ensemble here into two parts representing the more weakly and strongly hydrogen bonded water molecules around 2370 and 2530 cm -1. Coupled differential equations are used, assuming that each peak corresponds to a vibrational state (1 and 1 corresponding to the excited states with their peak positions of 2530 cm -1 and 2370 cm -1, respectively), and these states can couple to each other, and also decay to the hot ground states (0** and 0** ) via an intermediate state (0* and 0*, generally thought to be the OH bending mode). The lines in Fig. 4b and 4c in the main text show the result of this model. The population of the pumped, 2530 cm -1 state is largest close to zero pump-probe time delay. Equation (S6) shows the set of equations used when the molecules are excited to the 2530 cm -1 excited state 1 (55, 56). dn 0 dt dn 1 dt dn 1 dt dn 0 dt dn 0 dt dn 0 dt = σg(t)(n 0 N 1 ). = σg(t)(n 0 N 1 ) + N 1 τ down N 1 τ up N 1 τ 1. = N 1 τ up N 1 τ down N 1 τ 1 = N 1 τ 1 N 0 τ eq. = N 1 N 0. τ 1 τ eq = N 0 τ eq. (S6)

13 dn 0 dt = N 0 τ eq. The population in state (x) is depicted by N x. The excitation pulse was modelled by a Gaussian (G(t)) (with a full width at half maximum of 300 fs) multiplied by the percentage bleach (σ). For the final fit the population of the different states are used as shown in equations S7 and S8 (55). The long time offset is due to the heating of the molecules. This leads to a blue-shift of the spectrum, and can be included in the fit by multiplying an offset scaling factor (O hot ) with the population of hot ground states (N 0 (t) + N 0 (t)). As we are interested in obtaining information about the coupling between weakly and strongly hydrogen bonded water molecules for the three different samples, the only floating parameter in the fit is the coupling time (τ down ) besides a simple scaling factor (S) to match the fit intensity to the data. All other parameters were fixed from experimental measurements or values known from literature (56-58). Of course the thermalization time could also be different for the three different samples. However, whether enhanced coupling is also present for relaxation from the bending mode, i.e. for the thermalization time, is more difficult to quantify, because thermal effects affect the signals at longer times where the signals in general are smaller and thus noisier. As the differences in the short time dynamics are inherently more reliable, and less dependent on the details of the modeling, than those at longer times, we decided to float only the coupling time giving information about the coupling efficiency. 2 SFG(t) = S ((N 0 (t) + N 0 (t) N 1 (t) + O hot (N 0 (t) + N 0 (t))) 1 2 ) SFG 2 (t) = S ((N 0 (t) + N 0 (t) N 1 (t) + O hot (N 0 (t) + N 0 (t))) 1 2 ) (S7) (S8) The timescales extracted from the model can be found in table S2. For τ up we calculate from τ up Boltzmann s Law that = exp ( ΔH ) = 2.3, where ΔH is the change in energy (160 τ down k B T cm-1 ), k B is Boltzmann s constant, and T is the temperature (278 K). The very fast τ down timescale for INP suggests that this coupling between the more strongly and more weakly H- bonded water molecules comes from inter- or intramolecular energy transfer via dipole-dipole couplings (Forster resonance energy transfer)(48, 56). The coupling time, τ down, for the pure water-air interface is an order of magnitude longer than that of the water-inp interface, and the water-lysozyme interface coupling time is more than twice as long. The time constants are summarized in table S2 and highlighted in figure S11g. This analysis highlights the strong differences in coupling between these different interfaces, and displays how efficiently the energy transfer process occurs in the water molecules at the water INP interface.

14 fig. S11. Time resolved difference sum frequency spectrum for (a) the water air and (b) the water lysozyme interfaces after excitation with an IR pump pulse near the weakly hydrogen-bonded water

15 resonances (dashed line). (c) Time dependent bleach integrated over two spectral regions, cm -1 (strongly H-bonded) and cm -1 (weakly H-bonded) for the air water interface and (d) the water lysozyme interface. Results of the fit are summarized in table S2. (e) Inferred time-dependent populations of the two vibrational states for the water air and (f) the water lysozyme interface. (g) Coupling times τ down extracted from the time resolved data. fig. S12. Energy transfer processes at the water-afp interface. (a) Time resolved difference sum frequency spectrum for the water-antifreeze protein interface after excitation with a 2530 cm -1 pump pulse (dashed line) near the weakly H-bonded water resonances. Just as is seen for the INP interface, the signal bleach is very intense in the low frequency water peak related to strongly H-bonded water. This shows that energy transfer is also very rapid and efficient. (b) Time dependent bleach integrated over two spectral regions, cm -1 (strongly H-bonded) and cm -1 (weakly H-bonded). Fits of the data using a coupled differential equation model reveal extremely fast (<50 fs) energy transfer between weakly and strongly H-bonded water species. (c) Time resolved populations of the two vibrational states extracted from the coupled differential equations. For the water-antifreeze protein interface the strongly H-bonded state's population is higher than that of the initially excited peak, which proves extremely efficient energy transfer, similar to the INP interface. table S2. Timescales (in femtoseconds) extracted from the coupled differential equation fits (only τ down was allowed to float; the others were fixed). Sample τ down τ up τ 1 τ 1 τ eq Water - INP x τ down Water - Lysozyme x τ down Water - Air x τ down