Interfacial Water Ordering Is Insufficient to Explain Ice-Nucleating Protein Activity

Ice-nucleating proteins (INPs) found in bacteria are the most effective ice nucleators known, enabling the crystallization of water at temperatures close to 0 °C. Although their function has been known for decades, the underlying mechanism is still under debate. Here, we show that INPs from Pseudomonas syringae in aqueous solution exhibit a defined solution structure and show no significant conformational changes upon cooling. In contrast, irreversible structural changes are observed upon heating to temperatures exceeding ∼55 °C, leading to a loss of the ice-nucleation activity. Sum-frequency generation (SFG) spectroscopy reveals that active and heat-inactivated INPs impose similar structural ordering of interfacial water molecules upon cooling. Our results demonstrate that increased water ordering is not sufficient to explain INPs’ high ice-nucleation activity and confirm that intact three-dimensional protein structures are critical for bacterial ice nucleation, supporting a mechanism that depends on the INPs’ supramolecular interactions. T formation of ice is thermodynamically favored in water at temperatures below 0 °C, but the crystallization is kinetically hindered owing to the energy barrier associated with creating the initial ice seed. As a result, pure water droplets can, depending on their size and cooling rate, be supercooled to temperatures as low as −38 °C. Ice crystals can be formed either by homogeneous nucleation at lower temperatures or by heterogeneous nucleation catalyzed by compounds that serve as ice nucleators (IN). The most effective biological IN known are ice-nucleating proteins from bacteria such as Pseudomonas syringae. Bacterial INPs can have different sizes but are typically large macromolecules that are anchored to the outer cell membrane of the bacterial cell wall. They are typically present as monomers but have repeatedly been shown to aggregate in the bacterial outer membranes. Large INP aggregates are thought to be responsible for freezing at temperatures between −2 and −4 °C and smaller INP aggregates at temperatures between −7 and −12 °C. Computer-based homology modeling proposed the bacterial INP structure to be β-helical (Figure 1A) with similarities to hyperactive insect antifreeze proteins (AFPs). More recent models based on molecular dynamic simulations further suggest that a highly conserved threonine-X-threonine motif is used to interact with ice, which again is similar to some AFPs. On the molecular scale, the INPs are believed to work by organizing water into preordered patterns, which increase in size as the temperature decreases until they are large enough to form a stable embryonic crystal, leading to ice growth. However, the role of the INP structure, the interaction of INPs with water, and the underlying working mechanism remain largely unknown. Here, we study the effects of temperature on the structure, hydration shell, and ice-nucleation efficiency of purified proteinaceous IN of P. syringae. We performed purification of fragmented P. syringae (Snomax) solutions using falling water ice affinity and rotary ice-shell purification (see the Supporting Information (SI) for details). Both purification methods use the unique property of the INPs to interact with ice and have previously been used to purify AFPs from natural sources. The purification process involved the incorporation of the INPs into the slowly growing ice phase and the exclusion of other biomolecules and impurities. Using this ice-affinity purification, we obtained a mixture of all the INPs present in P. syringae, including residual protein-associated lipids. The success of the purification was assessed by determining the ice-nucleation activity of the purified INPs using the high-throughput Twinplate Ice Nucleation Assay (TINA) (Figure 1 and Figure S2). In the following, we will refer to these purified samples as “purified INPs”. Figure 1B shows typical statistical freezing curves of aqueous solutions of fragmented P. syringae and purified INPs with 0.1 mg/mL, while Figure 1C compares their freezing behaviors inferred from freezing curves recorded for concentrations spanning from 0.1 mg/mL to 1 ng/mL (see also Figure S3). Received: October 19, 2020 Accepted: December 3, 2020 Published: December 16, 2020 Letter pubs.acs.org/JPCL © 2020 The Authors. Published by American Chemical Society 218 https://dx.doi.org/10.1021/acs.jpclett.0c03163 J. Phys. Chem. Lett. 2021, 12, 218−223 This is an open access article published under a Creative Commons Attribution (CC-BY) License, which permits unrestricted use, distribution and reproduction in any medium, provided the author and source are cited. D ow nl oa de d vi a U N IV A M ST E R D A M o n Fe br ua ry 1 , 2 02 1 at 0 8: 37 :2 1 (U T C ). Se e ht tp s: //p ub s. ac s. or g/ sh ar in gg ui de lin es f or o pt io ns o n ho w to le gi tim at el y sh ar e pu bl is he d ar tic le s.

T he formation of ice is thermodynamically favored in water at temperatures below 0°C, but the crystallization is kinetically hindered owing to the energy barrier associated with creating the initial ice seed. 1 As a result, pure water droplets can, depending on their size and cooling rate, be supercooled to temperatures as low as −38°C. 2 Ice crystals can be formed either by homogeneous nucleation at lower temperatures or by heterogeneous nucleation catalyzed by compounds that serve as ice nucleators (IN). The most effective biological IN known are ice-nucleating proteins from bacteria such as Pseudomonas syringae. 3,4 Bacterial INPs can have different sizes but are typically large macromolecules that are anchored to the outer cell membrane of the bacterial cell wall. They are typically present as monomers but have repeatedly been shown to aggregate in the bacterial outer membranes. 5−8 Large INP aggregates are thought to be responsible for freezing at temperatures between −2 and −4°C and smaller INP aggregates at temperatures between −7 and −12°C. 9 Computer-based homology modeling proposed the bacterial INP structure to be β-helical ( Figure 1A) with similarities to hyperactive insect antifreeze proteins (AFPs). 10 More recent models based on molecular dynamic simulations further suggest that a highly conserved threonine-X-threonine motif is used to interact with ice, which again is similar to some AFPs. 11 On the molecular scale, the INPs are believed to work by organizing water into preordered patterns, which increase in size as the temperature decreases until they are large enough to form a stable embryonic crystal, leading to ice growth. 12 However, the role of the INP structure, the interaction of INPs with water, and the underlying working mechanism remain largely unknown. Here, we study the effects of temperature on the structure, hydration shell, and ice-nucleation efficiency of purified proteinaceous IN of P. syringae.
We performed purification of fragmented P. syringae (Snomax) solutions using falling water ice affinity and rotary ice-shell purification (see the Supporting Information (SI) for details). 13,14 Both purification methods use the unique property of the INPs to interact with ice and have previously been used to purify AFPs from natural sources. 15 The purification process involved the incorporation of the INPs into the slowly growing ice phase and the exclusion of other biomolecules and impurities. Using this ice-affinity purification, we obtained a mixture of all the INPs present in P. syringae, including residual protein-associated lipids. The success of the purification was assessed by determining the ice-nucleation activity of the purified INPs using the high-throughput Twinplate Ice Nucleation Assay (TINA) (Figure 1 and Figure  S2). 16 In the following, we will refer to these purified samples as "purified INPs". Figure 1B shows typical statistical freezing curves of aqueous solutions of fragmented P. syringae and purified INPs with 0.1 mg/mL, while Figure 1C compares their freezing behaviors inferred from freezing curves recorded for concentrations spanning from 0.1 mg/mL to 1 ng/mL (see also Figure S3).
The curve of P. syringae shows two substantial increases in the cumulative number of IN per unit mass, N m (T) ( Figure 1C) around −3.0 and −7.5°C with plateaus between −4.5 and −7.0°C and below −9.5°C. At the plateaus, at temperatures below each increase of N m (T), fewer IN are active. 17 The two rises in the curve reveal that the ice-nucleation activity stems from two distinct classes of IN with different activation temperatures. We attribute the observed rise at −3.0°C to large assemblies of INPs (class A IN) and the rise at −7.5°C to smaller assemblies of INPs (class C IN) in accordance with previous studies. 18 −23 The freezing curve of the purified INPs looks similar to the nonpurified INP solution, with a change in the ratio of the INP number in the two classes at −3.0 and −7.5°C (see also Figure S2). Clearly, the purification process was successful and yielded active INPs. The reduction of class A IN activity for the purified sample indicates that the purification reduced the number of the larger INP aggregates compared to the nonpurified solution. This observation is in line with the hypothesis that the bacterial membranes are involved in the formation of larger functional INP aggregates, 5,10,24−28 and we expect bacterial membrane fragments to have no ice affinity.
Heat-treated INP solutions (see the SI for details on heat treatment) behave fundamentally differently. As apparent from comparing the droplet freezing statistics of the highest dilution concentrations shown in Figure 1B, the rises at −3.0 and −7.5°C are completely absent. Instead, we observe activity only around −25°C, which corresponds to background freezing of pure water in our system. Evidently, the heat treatment of the purified INPs completely inactivates their ice-nucleation abilities.
Using SFG spectroscopy, Pandey et al. reported that fragmented P. syringae bacteria (Snomax) show an increased capability to order water in their vicinity when cooled to temperatures close to the melting point of deuterated water . 29 Control experiments using misfolded and denatured INP fragments, lipids, and the protein lysozyme did not show this effect. The alignment of water into an ordered structure was concluded to be a condition that will promote interfacial ice nucleation.
Here, we conducted further SFG experiments with active and heat-inactivated INPs to determine whether there is a direct causal correlation between enhanced SFG water signals at low temperatures and bacterial ice-nucleation activity. In SFG, a broadband infrared pulse resonant with the probed molecular vibrations and a visible pulse are combined at a surface to generate light at the sum frequency of the two incident fields. The SFG process is bulk-forbidden in isotropic media, and only ensembles of molecules with a net orientation, e.g., at an interface, can generate a detectable signal. Figure 2A shows the temperature-dependent SFG spectra of aqueous solutions of purified INPs. The broad response from the O−D stretching bands of interfacial water molecules appears at frequencies below 2700 cm −1 and is affected by their interactions with the INPs adsorbed to the air−liquid interface. In the frequency region of 2800−3000 cm −1 , the SFG spectra show strong signals that we attribute to C−H stretching vibrations.
The SFG intensity of the O−D bands strongly increases upon lowering the temperature close to the melting temperature (3.82°C for D 2 O), indicating an increase in the structural order of the interfacial water molecules. This effect is completely reversible, as evident from the integrals of the water (O−D) bands for two cycles shown in the insets. The observed effect is also significantly larger than the effect observed for pure water (insets in Figure 2 and Figure S4). 30 In contrast, the signal intensity of the C−H stretching vibrations remains constant upon lowering the temperature.  Figure 2B shows temperature-dependent SFG spectra of aqueous solutions of heat-inactivated INPs. Interestingly, we find that the completely inactive INPs adsorbed to the air− liquid interface cause a comparably strong increase in the SFG intensity of the O−D signals upon lowering the temperature. Thus, we conclude that the enhanced interfacial water ordering at low temperatures cannot be directly associated with the presence of ice-nucleation active sites.
Interestingly, while the water response is indistinguishable between the active and inactivated INPs, marked changes occur in the C−H stretching region. Although we cannot precisely assign the manifold C−H stretching contributions in the SFG spectra, these changes indicate that there is a substantial change in the protein structure after inactivation. Figure 3A shows SFG spectra in the amide I region, which is sensitive to the secondary structure and orientation of proteins. 31,32 The amide I SFG spectra at room temperature and close to the melting temperature look very similar and show a strong signal at ∼1645 cm −1 and a weak signal at ∼1710 cm −1 . We assign the signal at 1645 cm −1 to the protein backbone of the INPs 33−35 and the weak signal at 1710 cm −1 to carbonyl groups in lipid molecules (see also Figure S5). 36−38 The lipid signal presumably originates from membrane lipids that remain protein-associated during the purification, which is in line with the presence of class A aggregates in our freezing experiments ( Figure 1B). The observation that the amide I SFG spectra do not change upon cooling suggests no structural or conformational changes of the INPs upon approaching biologically relevant working temperatures. These conclusions are supported by temperature-dependent CD spectra that also show very little change upon lowering the temperature ( Figure  3B).
Upon heating of the purified INPs, the amide I SFG response and the CD spectrum undergo marked changes, as evident from Figures 3A and 3B. Figure 3A shows that the interfacial protein backbone SFG signal at ∼1645 cm −1 of heatinactivated INPs is increased, while the lipid signal at ∼1710 cm −1 appears unaffected. The CD spectrum of the active INPs in Figure 3B shows a maximum molar ellipticity at 195 nm and a minimum at 228 nm, after which there is a gradual return to zero from 230 to 260 nm. Increasing the temperature reduces the molar ellipticity at 195 nm, and the minimum at 228 nm is reduced (see also Figure S7). These spectral changes following heating above ∼55°C suggest significant, irreversible alterations in the secondary structure contents of the INPs. We propose that the observed irreversible conformational changes cause a loss of the proteins' native functional structure and are the origin of the complete elimination of the INP's ice-nucleation activity after heat treatment.
The CD spectrum of the purified INPs ( Figure 4A) looks unusual, and its deconvolution using the structural database does not allow a clear distinction into the common secondary structures of α-helix, β-turn, β-strand, or random coil. 41 The spectral shape, however, shows similarities with those of AFPs derived from Marinomonas primoryensis (MpAFP) and Rhagium inquisitor (RiAFP) (Figure 4B, 4C) but with slightly shifted peak positions. Both AFPs have β-helical folds, 41 which is in agreement with the current theoretical model of the INP as shown in Figure 1A and the inset of Figure 4A. 42 The negligible changes in CD spectra at low temperatures are further consistent with temperature-dependent measurements of structurally similar β-helical AFPs. 43 In summary, we purified INPs from P. syringae using iceaffinity methods and report experimental evidence that the purified INPs are ice-nucleation active and that they adopt defined solution structures, which show resemblance with βhelical AFP spectra. 41 We further show that enhanced interfacial water ordering at temperatures close to the melting point of ice is found not only for active but also for completely inactivated INPs. While protein-induced enhanced interfacial water ordering likely constitutes an essential part of INPs'  The Journal of Physical Chemistry Letters pubs.acs.org/JPCL Letter working mechanism, our results reveal that increased water ordering observed with SFG spectroscopy is, by itself, not a sufficient condition for INP activity. Instead, our results highlight that the intact three-dimensional fold is essential for the ice-nucleation activity of INPs. This observation, combined with the similarity of the protein structure of the INPs from P. syringae and other ice-binding proteins, suggests that supramolecular interactions and ordering are key to the exceptional ice-nucleation activity of bacterial INPs. 42 We hypothesize that the completely intact native structure of the INP is required for the formation of the functional aggregates that allow the formation of ice nuclei or embryos large enough to enable freezing at −2°C (∼10 4 kDa). 23