Article

Microfluidic Cold-Finger Device for the Investigationof Ice-Binding Proteins

Lotem Haleva,1 Yeliz Celik,2,3 Maya Bar-Dolev,1 Natalya Pertaya-Braun,2 Avigail Kaner,1 Peter L. Davies,4

and Ido Braslavsky1,2,*1Institute of Biochemistry, Food Science and Nutrition, Robert H. Smith Faculty of Agriculture, Food and Environment, The Hebrew Universityof Jerusalem, Rehovot, Israel; 2Department of Physics and Astronomy, Ohio University, Athens, Ohio; 3Department of Physics and PhysicalSciences, Marshall University, Huntington, West Virginia; and 4Department of Biomedical and Molecular Sciences, Queen’s University,Kingston, Canada

ABSTRACT Ice-binding proteins (IBPs) bind to ice crystals and control their structure, enlargement, and melting, thereby help-ing their host organisms to avoid injuries associated with ice growth. IBPs are useful in applications where ice growth control isnecessary, such as cryopreservation, food storage, and anti-icing. The study of an IBP’s mechanism of action is limited by thetechnological difficulties of in situ observations of molecules at the dynamic interface between ice and water. We describe hereina new, to our knowledge, apparatus designed to generate a controlled temperature gradient in a microfluidic chip, called a micro-fluidic cold finger (MCF). This device allows growth of a stable ice crystal that can be easily manipulated with or without IBPs insolution. Using the MCF, we show that the fluorescence signal of IBPs conjugated to green fluorescent protein is reduced uponfreezing and recovers at melting. This finding strengthens the evidence for irreversible binding of IBPs to their ligand, ice. Wealso used the MCF to demonstrate the basal-plane affinity of several IBPs, including a recently described IBP from Rhagiuminquisitor. Use of the MCF device, along with a temperature-controlled setup, provides a relatively simple and robust techniquethat can be widely used for further analysis of materials at the ice/water interface.

INTRODUCTION

Ice-binding proteins (IBPs) are found inmany organisms thatresist or tolerate freezing, including fish (1), insects (2,3),plants (4), and microorganisms (5–8) inhabiting subzero en-vironments. Despite their different structures and biologicalroles, IBPs are thought to have a universal mechanism ofaction, which is to adsorb directly to ice crystals (9–11). Indoing so, IBPs depress the growth of ice crystals, leading todepression of the freezing point of the solution below themelting point, i.e., thermal hysteresis (TH). The term ‘‘anti-freeze protein’’ (AFP) has traditionally been used for IBPswith high TH activity. These AFPs act to depress the freezingpoint of an organism and prevent its body fluids from freezingat subzero temperatures. Additional functions of other IBPsinclude inhibition of ice recrystallization (12) and ice adhe-sion (13,14). The ability of IBPs to prevent freezing andfreeze injury in nature has current and potential uses. Theseinclude applications in the food industry (15); cryopreserva-

Submitted August 17, 2015, and accepted for publication August 1, 2016.

*Correspondence: ido.braslavsky@mail.huji.ac.il

Lotem Haleva, Yeliz Celik, and Maya Bar-Dolev contributed equally to this

work.

Editor: Elizabeth Rhoades.

http://dx.doi.org/10.1016/j.bpj.2016.08.003

� 2016

tion and chill-storage of cells (16–18), tissues, and organs fortransplantation (19); cryosurgery; preventing frost damage toplants (20); and anti-icing (21).

A gross classification of IBPs divides them into those thatcan adhere to the basal plane of ice and those that cannot.For example, fish AFPs typically bind prism and/or pyrami-dal planes of ice, leading to the growth of bipyramidal iceshapes (22). Insect AFPs usually bind basal planes of icein addition to other planes, leading to various bipyramidalor truncated bipyramidal shapes (23). Basal-plane affinitywas shown to be the basis of the extremely high TH levelsof insect AFPs (24), leading to their classification as hyper-active AFPs (25). Nevertheless, some IBPs with low THlevels can also bind the basal planes (26).

Investigators have used many different methods toaddress the mechanism by which IBPs depress ice growth,and have suggested several theories to explain it (reviewedin (10)). The challenges involved in handling ice, which isthe natural ligand of IBPs, in aqueous solutions, and inspect-ing ice at the molecular level have made it difficult to followthe dynamics of ice binding by the proteins. We previouslyused microfluidic technology and fluorescently tagged IBPsto directly study the dynamics of the interactions of IBPs

Biophysical Journal 111, 1143–1150, September 20, 2016 1143

FIGURE 1 MCF and an experimental cell. (A) Design of the microfluidic

chip with three inlets and one outlet. The copper tip (orange circle) acts as a

cold finger. The scale bar is 1 mm. (B) Experimental setup. TheMCF device

is placed on a copper plate, and two thermoelectric coolers shown in blue

determine its temperature (T1). The cold finger (orange) is cooled by a third

thermoelectric cooler shown in purple (T2) that is externally controlled

independently of T1. The scale bar is 1 cm. Prepared on the basis of the

ISOCOMID figure (27) with Google sketchup software.

Haleva et al.

with ice crystals (27–30). Our previous apparatuses wereisothermally cooled microfluidic devices (ISOCOMIDs),which had microfluidic channels and chambers that wereplaced on a temperature-controlled copper plate forisothermal cooling. The temperature of these chips couldbe lowered to ~�25�C for initial freezing. Once the wholesample was frozen, copper wires embedded close to theinlet and outlet were connected to a second temperaturecontroller to heat and melt peripheral ice blocking the chan-nel (27). To improve the performance of the first version ofthe ISOCOMIDs, we introduced the use of infrared lasersthat could melt ice at any chosen site for the second versionof ISOCOMIDs. The third versions of the ISOCOMIDswere even more complicated devices with valves that al-lowed the complete arrest of fluid flow (28,30). Using theseISOCOMIDs, we grew single ice crystals in protein solu-tions and followed the dynamics of ice binding by IBPs.We showed that fish IBPs adsorb on ice surfaces within sec-onds, whereas insect IBPs continue to accumulate for hours(28). In a different study (30), we washed the crystals fromthe protein-containing solution. In this way, we were able toevaluate the influence of free protein in the solution on theability of the surface-bound protein to prevent ice growth.We found that the insect IBPs could depress ice growth tothe same extent with and without protein in solution. Incontrast, fish proteins were much less able to prevent anice crystal from growing when protein was absent fromthe solution. This finding was explained by the ability ofinsect IBPs to bind the basal plane of ice.

The advantages of microfluidics technology make it anattractive strategy for the further study of IBPs. The abilityto control physical parameters such as temperature, liquidcomposition, and flow rate while following a multistepprocess in microliter volumes under the microscope hasgreat potential for facilitating basic research into IBP mecha-nisms; however, it was difficult to control the ice usingISOCOMIDs. Sophisticated measures were necessary tomelt the bulk ice and to leave a single crystal in the microflui-dic devices. Furthermore, the ice crystals obtained in these de-viceswerevery small (typically 50mm)and itwas challengingto sustain them in the channel during solution exchange.

Here, we describe the fabrication and application of a new,to our knowledge, microfluidic device to enable more robustcontrol over ice growth. This device, named the microfluidiccold finger (MCF), is based on a cell designed by Wilen andDash (31) to study the physical characteristics of ice and theflow in surface-melted ice layers. The MCF device has twoindependent temperature-controlled units: one from the stagebelow the chip as in the ISOCOMID, and one from themiddleof the chamber in which the ice is grown. By cooling the coldfinger more than the stage, we can form a temperaturegradient that determines the size of the ice that grows aroundthe cold finger. A simulation of the spatial distribution of thetemperature in the systemwas used to determine the gradientprofile. The MCF configuration allows an even flow of solu-

1144 Biophysical Journal 111, 1143–1150, September 20, 2016

tion without melting the ice in themiddle of the channel. Onecan easily grow single ice crystals from pure water before in-jecting proteins into the channel, and thus can exclude anypossible effect of protein embedded in the crystal. UsingtheMCF,we recently studied the binding of anAntarctic bac-terium to ice (14). In the study presented here, we used theMCF to follow an intriguing phenomenon of spatial meltingof the basal plane of ice when it is covered by IBPs. We showthat the fluorescence of green fluorescent protein (GFP)-labeled AFPs drops when the IBP-covered basal plane isovergrown by a new ice front. When the new ice layer ismelted back, the fluorescence signal of the original basalplane is regained from the sides to themiddle, until thewholeplane is melted. We also show that the MCF device can beused as a fast means of determining basal-plane affinity.

MATERIALS AND METHODS

Fabrication of the MCF device

MCFs were fabricated using soft lithography and replica molding with pol-

ydimethylsiloxane (PDMS) (32). Details are provided in Supporting Mate-

rials and Methods in the Supporting Material. The microfluidic channel

network contained one to three inlets and one outlet. A copper wire serving

as the cold finger was embedded in the center of an elliptical pool (Fig. 1 A).

The volume of the device is on the order of 0.2 mL and the height of the de-

vice is 20 mm, as verified by a surface profiler (Dektak 150 Stylus; Veeco,

Tucson, AZ). The low height of the device is important for fluorescence

imaging with low background signal.

Microfluidic Cold Finger for IBP Research

Experimental setup

All experiments were conducted with a custom-designed, temperature-

controlled cell as previously described (14,27) and presented in Fig. 1 B.

The cell includes a thermistor in combination with two thermoelectric cool-

ing elements connected to a copper plate. The thermoelectric coolers are

driven by a temperature controller (models 3150 and 3040; Newport, Irvine,

CA). This setupworks in the range from room temperature to�30.0�Cwith a

precision of50.01�C. TheMCF is positioned on the copper plate, which has

a 3-mm-diameter opening in the center for microscopic visualization. The

copper tip is connected to another thermoelectric cooler, which is operated

by a second temperature controller (ILX Lightwave or as above). Thus, the

ice crystal growth is controlled by forming a temperature gradient between

the copper cold finger embedded in the microfluidic device and the copper

plate the device rests upon.Typically, 10mLsyringeswere used as buffer res-

ervoirs that were connected to the device by Tygon tubing and blunt needles.

The proteins were injected via a syringe needle through one of the inlet lines,

and the flow to the cell was controlled by gravity (27,33). We used solutions

with 4 mMof protein unless mentioned otherwise. The cell was placed on an

Eclipse Ti microscope equipped with an EGFP filter (ET 49002; Chroma,

Bellows Falls, VT) connected to an Intensilight C-HGFIE fiber illuminator

(Nikon, Tokyo, Japan) and a Neo sCMOS camera (ANDOR Technology,

Belfast, UK). The analysis was performed using NIS Elements software

(Nikon). The images in Figs. 2, 4, B and E, and S3 were taken on a Nikon

TE2000-U epifluorescent microscope (Nikon, Japan) equipped with

633 nm He-Ne and 488 nm Ar laser illumination lines and a 488 TIRF filter

(C-45647; Chroma). For observation of the spectral shift of the GFP fluores-

cence under freezing conditions, we used a Leica sp8 confocal microscope

and the same cell and system used for cooling the MCF. In this experiment,

the sample was placed between two coverslips.

Growth and melting in the MCF

Initially, we set out to form a single ice crystal with a clear basal-plane

orientation. For the experiment involving ice growth and basal-plane for-

mation without protein, water supplemented with 1 mM cyanine 5 (Cy5)

fluorescent dye was injected through the flow line. All experiments with

proteins were started by flowing BSA solution for 20 min to block the

PDMS walls. Next, pure water was flown into the MCF. The sample in

the MCF was then frozen by cooling the stage to ~�20�C. Then, the cell

was warmed slowly to the ice melting point while the temperature of the

copper tip (cold finger) was kept a few degrees below melting temperature.

The temperature of the stage was then reduced to allow growth of the crys-

tal. Partial warming and recooling of the ice was conducted to reduce the

number of ice grains with different orientations. During growth, ice crystals

formed flat faces typical of the basal planes that were lost at melting. The

warming/cooling procedure was repeated until only one or two parallel flat

planes were observed, indicating the presence of a single crystal. At this

point, protein was injected. We controlled the growth and melting of the

crystal by independently adjusting the temperatures of the cold finger and

the metal stage beneath the MCF, which formed a temperature gradient.

Typically, the stage temperature varied between 0�C and þ1.5�C while

the cold finger was kept at �2�C, as measured by a thermistor located on

the top of the microfluidic device, 5 mm above the microfluidic channel.

AFPs

All proteins used in this study were recombinant fusions containing GFP

and a 6-His tag. Proteins were produced in Escherichia coli and purified

by affinity chromatography on Ni-NTA Sepharose followed by ice-affinity

purification (34). Spruce budworm AFP (sbwAFP) isoform 501 and fish

ocean pout type III QAE isoform conjugated to GFP were prepared as pre-

viously described (24,35,36). Marinomonas primoryensis bacterial AFP

(MpAFP) was prepared as described in (37). Plasmid bearing Rhagium

inquisitor AFP (RiAFP; a gift from A. Hakim (Yale University)) was trans-

formed into E. coli and expressed as described previously (38). GFP-linked

TmAFP was prepared as described previously (39,40). It was previously

shown that the activity of AFPs is not negatively affected by conjugation

to GFP (24,35). GFP-sbwAFP and GFP-AFPIII were kept in 10 mM

Tris-HCl, 1 mM EDTA, 10 mM ammonium bicarbonate (pH 8) buffer

solution, and GFP-TmAFP was kept in 20 mM ammonium bicarbonate

(pH 8). GFP-MpAFP was kept in a buffer solution containing 20 mM

CaCl2 and 25 mM Tris-HCl (pH 8). RiAFP was used in 150 mM NaCl

and 50 mM NaPO4 (pH 7.5).

Simulation of the temperature profile in MCFs

Temperature profiles in the microfluidic device were simulated using Com-

sol Multiphysics software. The computational setup includes domains with

the properties and dimensions of the PDMS chip, cover glass, oil, ice,

copper plate, and copper tip. The temperature field was simulated in three

dimensions. We analyzed the two-dimensional temperature profile of the

main chamber with and without an ice crystal right underneath the cold

finger. The temperature of the copper plate was set to 273.2 K and the

top of the cold finger was set to 267.5 K or 258.2 K. The thickness of the

wire was 0.8 mm or 0.4 mm. The wire was narrow at its end and the tip

touched the top of the microfluidic channel.

RESULTS AND DISCUSSION

Formation of a single ice crystal in themicrofluidic device

Fig. 2 presents a single ice crystal in the MCF during acourse of melting (Fig. 2 A) and growth (Fig. 2 B). Ice crys-tals formed a circular shape when melting, but two flat basalplanes parallel to each other developed during growth anddisappeared upon remelting. The twofold symmetry andthe flatness of these surfaces are characteristic of basal

FIGURE 2 Ice crystal grown in the MCF device.

(A and B) Images showing a single ice crystal

grown from Cy5 solution in the MCF. (A) The

ice crystal is above the melting point. During

melting, the crystal is round. (B) The temperature

is lowered below the melting point, and flat basal

planes develop during crystal growth. The basal

planes are observed when the ice is oriented with

the c-axis normal to the objective. (C) Representa-

tion of an ice crystal with hexagonal symmetry.

The c-axis direction is noted. To see this figure in

color, go online.

Biophysical Journal 111, 1143–1150, September 20, 2016 1145

Haleva et al.

planes of ice grown from pure water by a layer-by-layergrowth mechanism (31). Since there were no AFPs or anyother ice-affecting substance in the solution, the crystalgrowth was completely determined by the temperaturegradient in the cell and the properties of free growth ofthe ice crystal. Under these conditions, the basal planesgrow flat, whereas the prism surfaces tend to grow with cur-vature and their shape is determined by the temperaturefield. In this experiment, Cy5 was used for better visualiza-tion of the ice crystal in the solution since it is excluded dur-ing ice growth. In a different study (14), we used the MCF toobserve how the Antarctic bacteriumM. primoryensis bindsice. After growing an ice crystal in the microfluidic channel,we injected fresh bacteria and observed their interactionswith ice. M. primoryensis have low viability after freezing,and the MCF allowed us to introduce the bacteria into pre-formed ice without subjecting them to a freeze-thaw cycle.For further details, see (14).

Simulating the temperature gradient in the MCF

A simulation of the temperature gradient in the cell is pre-sented in Fig. 3. The simulation shows that when the stageis set to þ0.2�C and the top part of the cold finger is keptat 6�C below the melting point, the temperature of themelting point is obtained at a distance of 0.25 mm fromthe center, representing an ice crystal of 0.5 mm diameter(Fig. 3 B). These parameters are in accordance with ourexperimental setup. A general increase in the temperatureof the sample is due to the observation hole (2.5 mm

1146 Biophysical Journal 111, 1143–1150, September 20, 2016

diameter) in the underlying cold copper plate that servesas the first cooling element. The hole creates a gap inthe cooling plate that leads to a temperature rise. We furthershow that if the cold finger had been 0.4 mm thick and not0.8 mm as in our design, ice would not have existed with acold finger set 6�C below the melting point (Fig. 3 D). In asimulation using a 0.4-mm-thick wire, a temperature differ-ence of 15�C would have been needed to obtain an icecrystal of 0.5 mm in diameter. The simulation takes intoconsideration the differences in heat conductivity betweenice and water, although this factor results in only a negli-gible difference on the temperature gradient (Fig. 4 C).Overall, the temperature gradient allows for easy replace-ment of the liquid around the ice, as the inner ice is keptat a low temperature and thus is quite stable.

Accumulation of hyperactive AFPs on the basalplane of ice

Fig. 4 presents the interactions of a set of GFP-labeled AFPsfrom various organisms with the basal plane of ice that wasgrown from pure water. In Fig. 4 A, a single crystal isobserved in the presence of GFP alone. The two parallelbasal planes of the crystal are clearly noted without anyfluorescent signal on them, showing that the basal planesfail to accumulate the protein during growth. In the experi-ment shown in Fig. 4 B, a similar crystal was grown, butwith GFP-TmAFP solution introduced into the channel.The high fluorescence signal emanating from the flat basalplanes clearly shows the affinity of the protein for this

FIGURE 3 Simulation of ice grown in the MCF

device. (A) Illustration of the design used in the

simulation, including the two cooling units (cold

stage and cold finger). The channel is shown in

light brown. Below the main chamber is a hole in

the cold stage for observation by the inverted

microscope. The simulation box dimensions are

20 mm � 20 mm � 10 mm. (B) Two-dimensional

temperature gradient of the microfluidic chamber,

where the temperature of the stage is þ0.2�C and

the cold finger is set to�6.2�C. At this temperature

gradient, an ice crystal 0.5 mm in diameter is

formed, marked as a black circle. (C) A one-

dimensional temperature distribution along the

middle of the chamber and cold finger (marked

by a red line in B), comparing the simulation con-

ducted using water parameters (as in B and D) with

ice parameters. (D) One-dimensional temperature

distribution along a line through the middle of

the chamber, comparing the designed cold finger

(0.8 mm in diameter) with a theoretical cold finger

(0.4 mm in diameter). The former yields 0.5 mm

ice. In the latter case, the temperature is also above

zero in the location under the cold finger, so there

is no ice in the chamber.

FIGURE 4 Basal-plane affinity of hyperactive AFPs compared with type

III AFP. The direction of the c-axis is noted by a white arrow on the basal

plane. (A) eGFP alone. (B) TmAFP. (C) RiAFP. (D) sbwAFP. (E) MpAFP.

(F) Type III AFP. The small black spot in the center is a part of the

PDMS channel (a 50 mm pillar supporting the microfluidic well). (G)

The same crystal in (F) after reduction of the temperature, showing the

basal plane of the crystal after growth. (H) Enlargement of the rectangular

section in (G), showing the spicular growth pattern. Scale bars indicate

100 mm. To see this figure in color, go online.

FIGURE 5 Accumulation of GFP-sbwAFP on the basal plane of ice. (A)

Ice crystal grown with a flat basal plane after injection of GFP-sbwAFP

(4 mM) at the beginning of the accumulation period (time ¼ 0 s). The

basal plane is indicated by an arrow. The scale bar indicates 100 mm.

(B) The same crystal in (A) after 30 min of incubation. (C) Fluorescence

intensity profile on the ice rim. The graph represents the difference be-

tween the average fluorescence signal of the box on the rim shown in

(A) and (B) and the fluorescence of the solution, which is taken as the

box at the top-right corner. The solution fluorescence was taken far

from the ice rim to avoid bias from depletion of protein in the area next

to the rim. A white arrow points to the basal plane. To see this figure in

color, go online.

Microfluidic Cold Finger for IBP Research

specific surface. Fig. 4, C–E, show similar experiments inwhich hyperactive AFPs from other insects and an Antarcticbacterium,MpAFP, also demonstrate basal-plane affinity. InFig. 4 C, five layers of basal planes were grown one abovethe other in the presence of Rhaguim inquisitor AFP(RiAFP). We were able to obtain such a crystal by loweringthe temperature in steps after protein accumulated on a basalplane. Five steps of accumulation and overgrowth wereconducted to form the five layers of fluorescent basalplanes. A similar experiment with sbwAFP is presented inMovie S1. These findings are in accordance with previousstudies using fluorescence microscopy (10,24,28,29) andice hemisphere (41) techniques, which showed the directbasal-plane affinity of hyperactive AFPs. When GFP-labeled type III AFP, a moderately active AFP, was injectedinto a channel containing basal ice planes grown from purewater, there was no accumulation of fluorescent signal,as shown in Fig. 4 F. When we allowed this crystal togrow further, many spikes grew on top of the basal plane

(Fig. 4, G and H), resembling the spicular growth patternpreviously observed with AFP type III (23). This growthsession is also presented in Movie S2. In the movie, onecan note that during crystal growth, protein from the solu-tion is pushed by the growing ice front, and for a few sec-onds it appears like a layer of protein on top of the icerim. When the crystal grows a little further, the spikes startto grow and this layer disappears, indicating that this proteinis not bound to the basal plane of the ice. These findingsemphasize the usefulness of the MCF for determining thebasal-plane affinity of AFPs.

In the presence of AFPs that accumulate on the basalplanes, these planes are stable and do not grow even ifthe temperature is slightly reduced. Fig. 5 presents an icecrystal in GFP-sbwAFP solution over time. This crystalwas observed from the instant of formation of the basalplane (Fig. 5 A) for 30 min (Fig. 5 B). The intensity profileof the GFP-sbwAFP on the basal plane (ice rim indicated bythe white arrow) is shown in Fig. 5 C. In the presence of4 mM protein, accumulation started at a fast rate and thenslowed down, as indicated by the incline of the fluorescence

Biophysical Journal 111, 1143–1150, September 20, 2016 1147

Haleva et al.

signal profile (Fig. 5 C), but continued for several hours. Athigh protein concentrations (20 mM) the accumulation ofAFPs reached a saturation level after which no further accu-mulation was observed. This accumulation is in accordancewith our previous results showing both accumulation of pro-tein on ice surfaces using ISOCOMIDs (28) and an increasein TH activity upon longer time exposure of ice to a proteinsolution (27,28). Prolonged time periods of incubation ofthe ice in protein solution leads to more coverage of theice surface by protein and better protection of the ice fromsecondary nucleation events (29). This results in depressionof the freezing point to lower temperatures and subsequentlyhigher TH values. The basal plane in the presence of type IIIAFP was not stable due to the protein’s lack of basal-planeaffinity (Movie S2) (30), so prolonged accumulation exper-iments could not be conducted. The prolonged accumulationtimes of the hyperactive AFPs on the basal planes are inaccordance with our own previous studies on the differentbinding dynamics of insect and fish AFPs (28), as well asrecent work on TH measurements by sonocrystallization.In the latter type of measurement, a sonication pulse inducesinstant nucleation in a supercooled solution. It was previ-ously shown that the TH level of fish type III AFP underthese conditions was similar to values obtained using ananoliter osmometer (42). In contrast, hyperactive AFPsdid not hold high TH by sonocrystallization relative to theiractivities measured by a nanoliter osmometer. This wasattributed to the fast binding of type III AFP to ice relativeto the slow binding kinetics of the hyperactive AFPs. Here,we used the MCF device to exchange the solution around anice crystal (Fig. S1), and found that crystals covered withfish or insect AFPs were stable over time even when thecrystals were in protein-free solutions. Clearly, the MCF is

A

B

1148 Biophysical Journal 111, 1143–1150, September 20, 2016

a more robust device than the ISOCOMID for proteinexchange experiments (28).

Reduced GFP fluorescence intensity due tofreezing

Fig. 6 and Movie S1 show a sequence of growth and meltingof ice in a solution of GFP-sbwAFP. The fusion protein wasflown into the channel after an initial ice crystal was grownas described above, and then allowed to grow further. Oncethe temperature was stabilized within the TH gap, crystalgrowth stopped and protein accumulated on the ice surface(Fig. 6 A, i and ii). After a short period of accumulation, thetemperature was slightly lowered and the previous surfacewas overgrown by another layer of ice (Fig. 6 A, iii). Atthe moment of overgrowth of the surface-bound AFP, wenoted a clear reduction in the fluorescence signal, which isillustrated in Fig. 6 A, iii, and B. The signal was recoveredwhen the new ice layer was melted. At the point of melting,we kept the temperature only slightly above the meltingpoint to ensure a slow melting process that would allow usto follow the fluorescence signal. Still, the recovery was afast process of a few seconds, after which the protein imme-diately diffused away to the bulk solution. Our camera wasfast enough to catch the peak fluorescence upon melting.Fig. 6 B presents the fluorescence profile at four positionson the ice rim, where position 1 is close to the edge ofthe basal plane and position 4 is close to the center of theplane. The observed delay in the melting peak betweenthe different positions indicates that the basal plane meltedfrom the edges to the center. The height of the melting peakslightly increases as melting advances from the edge to themiddle of the basal plane due to fluorescence of protein that

FIGURE 6 Ice growth and melting with en-

gulfed GFP-labeled AFPs in the MCF device. (A)

Sequence of images taken during the course of an

experiment, showing (i) a stabilized crystal from

the melt, (ii) accumulation of protein on a stable

ice rim, and (iii) overgrowth by a second ice layer

(marked by a white arrow) upon temperature

reduction. The white numbers in the top-right

corner indicate the time at which the image was

captured. The signal observed on the ice rim in ii

is reduced. (iv–vi) The new ice layer was allowed

to melt and the fluorescence signal was restored

sequentially after the basal plane melted from the

corners to the center. The time of each frame is

shown in the top-right corner. (B) Fluorescence in-

tensity profile of the areas boxed in (A) over time.

The position of the boxed areas was constant along

the experiment, with the blue box (position 1)

being closest to the edge of the basal plane, fol-

lowed by red, purple, and orange boxes (positions

2, 3 and 4, respectively). Their colors match

the four traces. The accumulation peak, the signal

reduction upon overgrowth, and the recovery

peak are denoted as a, b, and c, respectively. See

Movie S1.

Microfluidic Cold Finger for IBP Research

diffuses from the melting ice front. To verify this point, wereduced the fluorescence of the solution near the basal planefrom the fluorescence of the ice surface (Fig. S2). The fluo-rescence profile in this case was indeed even at all four po-sitions. We note that the recovery peaks are not as high asthe signal before growth, most probably because some ofthe protein diffused away faster than our imaging could cap-ture the fluorescence. These findings indicate that the fusionprotein was strongly attached to the ice surface, such that itwas incorporated into the ice upon growth of the new layer.The phenomenon was highly reproducible in all of thehyperactive AFPs tested, including TmAFP, MpAFP, andRiAFP. One might consider that the observed reduction insignal during growth of the new ice surface could be dueto desorption of protein molecules from the surface. How-ever, the recovery of the signal during melting occurredmuch faster than the accumulation rate during growth. Wecalculated the slope of the intensity profile to estimate therates of fluorescence regain. The intensity increase uponaccumulation was 6 units/s, whereas during melting itreached 19 units/s. This >3-fold increase in the rate ofsignal gain emphasizes that the protein was already on theice surface when melting occurred. The overall reductionof the fluorescent signal upon growth of a new ice layer pre-sumably stems from a slight distortional stress of the GFPchromophore during freezing. This distortion is removedwhen the GFP molecules are in solution again. To determinewhether the reduction of fluorescence was accompanied bya spectral shift, we conducted a spectral analysis of GFPbefore and after freezing in a confocal microscope. Wefound that the fluorescence peak had a reduction in intensity,but no shift (Fig. S3). In accordance with our findings, a pre-vious study of mechanical perturbation of GFP moleculesusing atomic force microscopy showed a reduction of thefluorescence signal upon compression of the GFP barrelby the atomic force microscope tip, without a shift of thespectral peak (43). After the reduction of stress, the fluores-cence was recovered, even though the recovery time was notimmediate. The recovery of the GFP signal shows that thestructure of the molecule was not irreversibly disrupted,and it is reasonable that the minor distortion due to thefreezing quickly rebounded in our experimental system.We note that the recovery peak increased from position 1to position 4 due to signal from protein that was releasedinto the medium. Overall, our results support the modelof irreversible binding of AFPs to ice (27,35). Indeed,adsorption and overgrowth of AFPs on ice is the basis ofice-affinity purification (34,44), which has been used withgreat effectiveness to isolate numerous different AFP typesfrom crude protein solutions (45,46).

CONCLUSIONS

This study describes a microfluidic device embedded ina temperature-controlled cell that was used to grow single

ice crystals with a clear basal-plane orientation. We usedthis MCF device to follow two interesting phenomena: 1)the apparent reduction of GFP signal upon freezing of amolecule, probably as a result of temporal perturbation ofthe GFP structure that is removed upon melting, and 2)the basal-plane affinity of a variety of proteins. We showedthat the MCF device can be used as a fast means of deter-mining basal-plane affinity, in a process that takes minutescompared with more than a day using the fluorescence-based ice plane affinity technique (41). We have also usedthis device to investigate the adhesion of Antarctic bacteriato ice (14). The MCF offers a robust method for solution ex-change around ice crystals, and may provide new insightsinto IBPs as well as other ice-active substances.

SUPPORTING MATERIAL

Supporting Materials and Methods, three figures, and two movies are avail-

able at http://www.biophysj.org/biophysj/supplemental/S0006-3495(16)

30660-9.

AUTHOR CONTRIBUTIONS

L.H. performed research and analyzed data. M.B.-D. designed research,

analyzed data, and wrote the manuscript. Y.C. designed and performed

research, analyzed data, and wrote the manuscript. N.P.-B. designed

research. A.K. performed and analyzed all of the presented simulations.

P.L.D. contributed to the design and analysis of the data, and edited the

manuscript. I.B. designed and constructed the MCF device, designed

research, analyzed data, and wrote the manuscript.

ACKNOWLEDGMENTS

The authors are grateful to Drs. Liwei Chen, Savas Kaya, Yunxiang Gao,

Larry Wilen, J.S. Wettlaufer, and Alex Groisman for useful discussions

and help during fabrication of the microfluidic device. We are also indebted

to Sherry Gauthier, Vera Sirotinskaya, and Svetlana Pen for assistance in

preparing the AFPs, and to Einat Zelinger for assisting with confocal spec-

tral analysis.

This work was supported by the National Science Foundation, the European

Research Council, the Canadian Institutes for Health Research, the

Condensed Matter and Surface Science Program at Ohio University, and

the Biomimetic Nanoscience and Nanoscale Technology Initiative at

Ohio University.

REFERENCES

1. DeVries, A. L., S. K. Komatsu, and R. E. Feeney. 1970. Chemical andphysical properties of freezing point-depressing glycoproteins fromAntarctic fish. J. Biol. Chem. 245:2901–2908.

2. Duman, J. G., and J. L. Patterson. 1978. Role of thermal-hysteresis pro-teins in low-temperature tolerance of insects and spiders. Cryobiology.15:683–684.

3. Duman, J. G. 2015. Animal ice-binding (antifreeze) proteins andglycolipids: an overview with emphasis on physiological function.J. Exp. Biol. 218:1846–1855.

4. Worrall, D., L. Elias,., D. Bowles. 1998. A carrot leucine-rich-repeatprotein that inhibits ice recrystallization. Science. 282:115–117.

Biophysical Journal 111, 1143–1150, September 20, 2016 1149

Haleva et al.

5. Singh, P., Y. Hanada,., S. Tsuda. 2014. Antifreeze protein activity inArctic cryoconite bacteria. FEMS Microbiol. Lett. 351:14–22.

6. Raymond, J. A., C. Fritsen, and K. Shen. 2007. An ice-binding pro-tein from an Antarctic sea ice bacterium. FEMS Microbiol. Ecol.61:214–221.

7. Raymond, J. A., M. G. Janech, and C. H. Fritsen. 2009. Novel ice-bind-ing proteins from a psychrophilic Antarctic alga (Chlamydomonada-ceae, Chlorophyceae). J. Phycol. 45:130–136.

8. Hoshino, T., M. Kiriaki, and T. Nakajima. 2003. Novel thermal hyster-esis proteins from low temperature basidiomycete, Coprinus psychro-morbidus. Cryo Lett. 24:135–142.

9. Davies, P. L. 2014. Ice-binding proteins: a remarkable diversity ofstructures for stopping and starting ice growth. Trends Biochem. Sci.39:548–555.

10. Bar Dolev, M., I. Braslavsky, and P. L. Davies. 2016. Ice-binding pro-teins and their function. Annu. Rev. Biochem. 85:515–542.

11. Raymond, J. A., and A. L. DeVries. 1977. Adsorption inhibition as amechanism of freezing resistance in polar fishes. Proc. Natl. Acad.Sci. USA. 74:2589–2593.

12. Knight, C. A., A. L. DeVries, and L. D. Oolman. 1984. Fish anti-freeze protein and the freezing and recrystallization of ice. Nature.308:295–296.

13. Guo, S., C. P. Garnham,., P. L. Davies. 2012. Re-evaluation of a bac-terial antifreeze protein as an adhesin with ice-binding activity. PLoSOne. 7:e48805.

14. Bar-Dolev, M., R. Bernheim, ., I. Braslavsky. 2016. Putting lifeon ice: bacteria that bind to frozen water. J. R. Soc. Interface.13:20160210.

15. Payne, S. R., D. Sandford, ., O. A. Young. 1994. The effects of anti-freeze proteins on chilled and frozen meat. Meat Sci. 37:429–438.

16. Qadeer, S., M. A. Khan, ., S. Akhter. 2015. Efficiency of antifreezeglycoproteins for cryopreservation of Nili-Ravi (Bubalus bubalis) buf-falo bull sperm. Anim. Reprod. Sci. 157:56–62.

17. Ideta, A., Y. Aoyagi,., S. Tsuda. 2015. Prolonging hypothermic stor-age (4 C) of bovine embryos with fish antifreeze protein. J. Reprod.Dev. 61:1–6.

18. Kamijima, T., M. Sakashita, ., S. Tsuda. 2013. Antifreeze proteinprolongs the life-time of insulinoma cells during hypothermic preserva-tion. PLoS One. 8:e73643.

19. Rubinsky, B., A. Arav, ., C. Y. Lee. 1994. Freezing of mammalianlivers with glycerol and antifreeze proteins. Biochem. Biophys. Res.Commun. 200:732–741.

20. Lin, X., M. Wisniewski, and J. G. Duman. 2011. Expression of twoself-enhancing antifreeze proteins from the beetle Dendroides canaden-sis in Arabidopsis thaliana. Plant Mol. Biol. Report. 29:802–813.

21. Gwak, Y., J. I. Park, ., E. Jin. 2015. Creating anti-icing surfaces viathe direct immobilization of antifreeze proteins on aluminum. Sci.Rep. 5:12019.

22. Fairley, K., B. J. Westman, ., J. P. Mackay. 2002. Type I shorthornsculpin antifreeze protein: recombinant synthesis, solution confor-mation, and ice growth inhibition studies. J. Biol. Chem. 277:24073–24080.

23. Bar-Dolev, M., Y. Celik, ., I. Braslavsky. 2012. New insights into icegrowth and melting modifications by antifreeze proteins. J. R. Soc.Interface. 9:3249–3259.

24. Pertaya, N., C. B. Marshall, ., I. Braslavsky. 2008. Direct visualiza-tion of spruce budworm antifreeze protein interacting with ice crystals:basal plane affinity confers hyperactivity. Biophys. J. 95:333–341.

25. Graham, L. A., Y. C. Liou, ., P. L. Davies. 1997. Hyperactive anti-freeze protein from beetles. Nature. 388:727–728.

1150 Biophysical Journal 111, 1143–1150, September 20, 2016

26. Middleton, A. J., C. B. Marshall,., P. L. Davies. 2012. Antifreeze pro-tein from freeze-tolerant grass has a beta-roll fold with an irregularlystructured ice-binding site. J. Mol. Biol. 416:713–724.

27. Celik, Y., R. Drori, ., I. Braslavsky. 2013. Microfluidic experimentsreveal that antifreeze proteins bound to ice crystals suffice to preventtheir growth. Proc. Natl. Acad. Sci. USA. 110:1309–1314.

28. Drori, R., Y. Celik, ., I. Braslavsky. 2014. Ice-binding proteins thataccumulate on different ice crystal planes produce distinct thermal hys-teresis dynamics. J. R. Soc. Interface. 11:20140526.

29. Drori, R., P. L. Davies, and I. Braslavsky. 2015. Experimental correla-tion between thermal hysteresis activity and the distance between anti-freeze proteins on an ice surface. RSC Advances. 5:7848–7853.

30. Drori, R., P. L. Davies, and I. Braslavsky. 2015. When are antifreezeproteins in solution essential for ice growth inhibition? Langmuir.31:5805–5811.

31. Wilen, L. A., and J. G. Dash. 1995. Giant facets at ice grain-boundarygrooves. Science. 270:1184–1186.

32. Zaouk, R., B. Y. Park, and M. J. Madou. 2005. Fabrication of polydi-methylsiloxane microfluidics using SU-8 molds. In Microfluidic Tech-niques. S. D. Minteer, editor. Humana Press, Totowa, NJ.

33. Groisman, A., C. Lobo, ., A. Levchenko. 2005. A microfluidic che-mostat for experiments with bacterial and yeast cells. Nat. Methods.2:685–689.

34. Kuiper, M. J., C. Lankin, ., P. L. Davies. 2003. Purification of anti-freeze proteins by adsorption to ice. Biochem. Biophys. Res. Commun.300:645–648.

35. Pertaya, N., C. B. Marshall, ., I. Braslavsky. 2007. Fluorescence mi-croscopy evidence for quasi-permanent attachment of antifreeze pro-teins to ice surfaces. Biophys. J. 92:3663–3673.

36. Pertaya, N., Y. Celik,., I. Braslavsky. 2007. Growth-melt asymmetryin ice crystals under the influence of spruce budworm antifreeze pro-tein. J. Phys. Condens. Matter. 19:412101.

37. Garnham, C. P., J. A. Gilbert, ., P. L. Davies. 2008. A Ca2þ-depen-dent bacterial antifreeze protein domain has a novel beta-helical ice-binding fold. Biochem. J. 411:171–180.

38. Hakim, A., J. B. Nguyen, ., W. Meng. 2013. Crystal structure of aninsect antifreeze protein and its implications for ice binding. J. Biol.Chem. 288:12295–12304.

39. Bar, M., Y. Celik, ., I. Braslavsky. 2008. Interactions of beta-helicalantifreeze protein mutants with ice. Cryst. Growth Des. 8:2954–2963.

40. Bar, M., R. Bar-Ziv,., D. Fass. 2006. Efficient production of a foldedand functional, highly disulfide-bonded beta-helix antifreeze protein inbacteria. Protein Expr. Purif. 48:243–252.

41. Basu, K., C. P. Garnham, ., P. Davies. 2014. Determining the ice-binding planes of antifreeze proteins by fluorescence-based ice planeaffinity. J. Vis. Exp. 83:e51185.

42. Olijve, L. L. C., K. Meister, ., I. K. Voets. 2016. Blocking rapid icecrystal growth through nonbasal plane adsorption of antifreeze pro-teins. Proc. Natl. Acad. Sci. USA. 113:3740–3745.

43. Kodama, T., H. Ohtani, ., A. Ikai. 2005. Mechanical perturbation-induced fluorescence change of green fluorescent protein. Appl. Phys.Lett. 86:043901.

44. Marshall, C. J., K. Basu, and P. L. Davies. 2016. Ice-shell purificationof ice-binding proteins. Cryobiology. 72:258–263.

45. Basu, K., L. A. Graham,., P. L. Davies. 2015. Flies expand the reper-toire of protein structures that bind ice. Proc. Natl. Acad. Sci. USA.112:737–742.

46. Graham, L. A., and P. L. Davies. 2005. Glycine-rich antifreeze proteinsfrom snow fleas. Science. 310:461–461.

Biophysical Journal, Volume 111

Supplemental Information

Microfluidic Cold-Finger Device for the Investigation of Ice-Binding

Proteins

LotemHaleva, Yeliz Celik,Maya Bar-Dolev, Natalya Pertaya-Braun, Avigail Kaner, Peter L.Davies, and Ido Braslavsky

Microfluidic cold finger device for the investigation of ice-binding proteins

Supporting Information

Lotem Haleva*, Yeliz Celik*, Maya Bar-Dolev*, Natalya Pertaya-Braun, Avigail Kaner, Peter L. Devies, and Ido Braslavsky

1Institute of Biochemistry, Food Science and Nutrition, Robert H. Smith Faculty of Agriculture, Food and Environment, The Hebrew University of Jerusalem, Rehovot, Israel

2Department of Physics and Astronomy, Ohio University, Athens, OH, USA

3Department of Physics and Physical Sciences, Marshall University, Huntington, WV, USA

4Department of Biomedical and Molecular Sciences, Queen’s University, Kingston, ON, Canada

*Equal contribution first co-authors

MCF fabrication

The mask with the pattern of the microfluidic chip was designed using DesignCAD software (Figure 1A). For the master template, SU-8 photoresist (MicroChem Corporation, Newton, MA) was used. The fabrication started with spin coating of SU-8 2010 on a 3-inch silicon wafer (Virginia Semiconductor, Virginia). The photoresist was soft-baked at 65 oC for 1 min and hard-baked at 95 oC for 2 min before being exposed to UV light (365 nm) for 35 sec with a UV mask aligner (Karl Suss, Model MJB3) where the patterns were transferred on the silicon wafer. Following UV exposure, the unexposed parts were removed by SU-8 developer (MicroChem Corporation, Newton, MA). A soft polymer, Polydimethylsiloxane (PDMS) (Slygard 184, Dow Corning Corp., Midland, MI) was prepared with a ratio of 1:10 between the curing agent and the base. The PDMS and SU-8 mold were placed in a vacuum chamber to be degassed for 1 h with fluorinated trichlorosilane (33). Then, a copper wire (0.8 mm diameter) with a sharpened tip was placed on the mold. The tip of the wire was positioned on top of the middle of the chamber. This sharpened tip was used as the cold finger, with local temperature control in the center of the chamber. After placing the copper wire, PDMS was poured on the master mold and cured over-night at room temperature. The patterned PDMS was then irreversibly bonded to a glass slide after treating both with oxygen plasma cleaner (Harrick Plasma Cleaner, Model: PDC-32G, NY). The channel holes were punched with a 20-gauge flat-head needle. The device includes two or three inlet chambers and one outlet chamber (Figure 1A). Each chamber is approximately 600 µm in radius and they are connected to each other via an elliptical chamber. The axes of the ellipse are 1400 µm and 1700 µm in length. In order to improve the stability of the main chamber in terms of keeping its initial shape, we introduced an array of small cylindrical columns of size 40-50 µm in diameter into the main chamber. These small columns of PDMS in the main chamber were designed to increase the stability of the chamber in cases where the copper tip does not touch the bottom of the channel and puts extra stress on it. However, in most of our devices the copper tip touches the bottom of the channel. The volume of the device is on the order of 0.2 µl and the height of the device is 20 µm, verified by a surface profiler (Dektak 150 Stylus, Veeco, AZ, USA). The low height of the device is important for fluorescence imaging with low background signal.

Figure S1

Figure S1. Ice crystals before and after exchange with buffer solution. (A) An ice crystal was grown in GFP-sbwAFP and allowed to reach equilibrium. (B) The solution was exchanged with AFP-free buffer solution. Over the time period of >1 h no growth was observed. (C) Ice crystal formed in GFP-AFP-III solution. (D) After exchanging the solution with buffer the crystal in (C) was still protected and no growth was observed.

Figure S2

Figure S2. Ice growth and melting with engulfed GFP-labeled AFPs in the MCF device: modified background to figure 6. (A) The intensity profile of the same experimental data used in figure 6 with local background reduction. The fluorescence recovery is lower than in figure 6 but the peak height is similar in all positions, indicating that the differences noted in figure 6 are due to protein that diffused into the solution. The accumulation peak, the signal reduction upon overgrowth and the recovery peak are noted as 1, 2, and 3, respectively. See accompanied movie S1. (B) An image at time = 110 s from the experimental data set used for figure 6, showing the boxes used for signal and for local background extraction in A.

Figure S3

Figure S3. GFP fluorescence spectrum at different temperatures. A solution containing 40 µM eGFP was placed between two cover slips, which were placed on a cold stage and observed using a confocal microscope equipped with a spectral scan illumination and detection system (see methods). The temperature was lowered from room temperature to -24 °C and then warmed back. The signal of the complete image obtained was averaged. At 0 °C (before freezing and after freezing and thawing) the sample was liquid while at -5 °C and below the sample was frozen, as noted by white light. There is a clear reduction of the fluorescence signal upon freezing. After melting the signal is restored but not completely due to bleaching by the illuminating laser.

-10 ̊ C (frozen) -5 ˚C (frozen) 0 ˚C initial (liquid)

-24 ˚C (frozen)

0 ˚C final (liquid)

movie captions

Movie S1: Growth and melting of ice in solution of GFP-sbwAFP. The experiment started after growing an ice crystal with a clear basal plane (on the right side of the crystal) from pure water in the MCF device followed by injection of 4 µM GFP-sbwAFP. Then the crystal (in the middle of the frame) is growing and the fluorescent protein is excluded, so the new ice layer looks completely black. When growth stops, the fluorescent protein accumulates on the ice surface. The fluorescence on the rim is reduced when covered by an additional layer of ice growth. When this bound surface is exposed again during melting the signal is regained for an instant before the protein diffuses from the ice surface to the bulk solution. The movie is 8 times faster than real time (1 frame every 2 sec).

Movie S2: Growth of ice in solution of GFP-type III AFP. An ice crystal (large black on the right) grown from pure water in the MCF device after injection of 4 µl of 20 µM GFP-type III AFP is observed. The flat basal plane is in the middle of the frame. At the beginning of the movie the crystal is supercooled and stable. It is clear that protein is not accumulated on the ice surface. Then the crystal is allowed to grow and protein in the solution close to the ice surface is pushed away by the growing ice front without accululating on it. As the growth continues, spicules emerge from the basal plane, growing normal to the basal plane. The movie is in real time.