www.sciencemag.org/cgi/content/full/317/5845/1732/DC1

Supporting Online Material for

Free-Solution, Label-Free Molecular Interactions

Studied by Back-Scattering Interferometry

Darryl J. Bornhop,* J. C. Latham, Amanda Kussrow, D. A. Markov,

Richard D. Jones, Henrik S. Sørensen

*To whom correspondence should be addressed. E-mail: darryl.bornhop@vanderbilt.edu

Published 21 September 2007, Science 317, 1732 (2007)

DOI: 10.1126/science.1146559

This PDF file includes:

Materials and Methods

SOM Text

Figs. S1 to S7

References

Other Supporting Online Material for this manuscript includes the following:

(available at www.sciencemag.org/cgi/content/full/317/5845/1732/DC1)

Movie S1

1

Supporting Online Material

Materials and Methods

Chemicals and reagents

Interactions of Calmodulin (Sigma, St. Louis, MO) with a small molecule

inhibitor, a small peptide, and a binding protein were performed in a calcium containing

buffer system (i.e. 0.1 M HEPES, 0.1 M KCl, and 0.2 mM CaCl2 @ pH = 7.5) while

reactions of Calmodulin to Ca2+

were investigated using a non-calcium containing buffer

(i.e. 0.1 M HEPES, 0.1 M KCl, and 0.1 mM EGTA @ pH = 7.5). The pH of buffer

solutions was adjusted to the required pH by addition of 1M HCl or 1M KOH. Solutions

were filtered and degassed prior to binding experiments. The small molecule inhibitor,

trifluoperazine dihydrochloride (480 g/mol, Sigma, St. Louis, MO), was varied in

concentration from 5-25 μM using the calcium containing buffer. Calmodulin, held

constant at 2 M, and the inhibitor were both introduced on-chip and mixed in-line.

CaM’s interaction with each concentration of inhibitor was monitored and the real-time

association observed.

The small peptide used was a 17 residue (Arg-Arg-Lys-Trp-Gln-Lys-Thr-Gly-

His-Ala-Val-Arg-Ala-Ile-Gly-Arg-Leu) peptide from the sequence of myosin light chain

kinase (MLCK). The MLCK peptide, (2074.5 g/mol, Calbiochem, La Jolla, CA) was

aliquoted into concentrations ranging from 5-50 nM using the calcium containing buffer.

Calmodulin, held constant at 5 nM, was mixed on-chip with each concentration of the

MLCK peptide. The kinetics of Calmodulin binding MLCK peptide was monitored and

recorded for later analysis.

2

The CaM and Calcineurin interactions were performed in similar conditions to

those of the MLCK peptide. CaM was again buffered at a pH = 7.5 with 0.1M HEPES

and 0.1 M KCl. CaM concentration was held constant throughout the experiment at

10 nM. The CaM solution once more contained 0.2 mM CaCl2 to ensure CaM was in its

active conformational state. The interaction of CaM with various concentrations (10 –

100 nM) of Calcineurin was monitored by BSI. All concentrations of Calcineurin used

for the association curves were made using the same buffer and held at the same pH as

CaM. The temperature was held constant at 25ºC throughout the entire experiment. A 10

nM solution of CaM and a 100 nM solution of Calcineurin both in the absence of Ca2+

were mixed to serve as a control. The control showed < 1.5% of the signal observed at

the equivalent CaM and Calcineurin concentrations when Ca2+

was present.

Calmodulin was also reacted to a small metal ion, Ca2+

. CaM was buffered at a

pH = 7.5 with 0.1M HEPES and 0.1M KCl with its concentration held constant

throughout the experiment at 5 M. The CaM solution contained a small amount of

EGTA to chelate any free Ca2+

. BSI monitored, in real-time, the sequential reactions of

5 M CaM with concentrations of Ca2+

ranging from 12.5 – 100 M.

Microfluidic chip fabrication

CleWin 2.7 layout editor software was used to design the fluidic network. A soda

lime/chrome mask was generated from this design by Delta Mask (The Netherlands)

where the chrome layer was ~ 100 nm in thickness. Master molds were created from the

lithographic mask by standard lithographic techniques. First, three inch silicon wafers

3

(P <100>) were cleaned by sonication in acetone followed by treatment with piranha

solution. Then, they were rinsed, dried under N2 gas and prior to deposition baked on a

hot plate at 95°C for 5 minutes. The negative photoresist SU-8 2050 from Microchem

(Newton, MA) was evenly deposited on the Si wafer using a Laurell WS-400 Bench-top

single wafer spinner with 10 s @ 500 rpm spread cycle and 40 s @ 3000 rpm spin cycle.

The wafer was then soft baked on a hot plate for 3 min @ 65°C followed by 9 min @

95°C. The wafer was then allowed to cool to room temperature. UV exposure through

the photolithographic mask was accomplished using NOVACURE 2100 curing system

coupled to a liquid wave guide and a collimating lens. Following irradiation, a post

exposure bake (PEB) was performed (1 min @ 65°C 7 min @ 95°C). The wafer was

then again cooled to room temperature. The wafer was then placed in SU-8 organic

developer from Microchem to remove unexposed SU-8 photoresist, then rinsed with an

isopropyl alcohol (IPA) and dried under N2 gas. If during IPA rinse a milky white

substance appears, the developing time should be extended. As a result a positive relief

mold is created. To further increase mechanical stability of the mold it was hard baked

for approximately 5 hours @ 200°C . An Alphastep 200 stylus surface profiler (Tencor

Instruments) was used to accurately measure the height of the standing relief structures.

All binding assays were performed in microchips created by cast molding onto the

master mold fabricated above. Cast molding was done in a silicon elastomer,

polydimethylsiloxane (PDMS), purchased as Sylgard 184 (Dow Corning, Midland, MI).

Prior to casting, the PDMS was mixed in a 10 : 1 ratio (base : curing agent) and degassed.

PDMS was cast over the master that had been placed into a 100 15 mm Falcon Petri

dish (Becton Dickinson, Franklin Lakes, NJ) such that the height of the PDMS was

4

approximately 2 mm. The Petri dish was placed into a desiccator and a vacuum was

applied for further degassing. Once no air bubbles were visibly present, the Petri dish

was removed from the desiccators and set in a large convection oven for roughly 8 hours

@ 65°C.

After the curing process was complete, the Petri dish was removed from the oven

and allowed to cool briefly. The PDMS microchip device was physically removed from

the Si master mold by fine precision scalpel and tweezers. Access ports for sample

introduction (2 ports) and applied vacuum/waste removal (1 port) were mechanically

punched out by stainless steel capillary tubing. PDMS, with the fluidic network facing

up, was then plasma oxidized for about 30 seconds along with a 3 1 1 mm glass

microscope slide (Fisher Scientific) cleaned in the same fashion as the bare Si wafer.

Following oxidation, the PDMS was irreversibly sealed to the microscope slide so that

the fluidic network was in contact with the glass. Water was kept in the channels until

experiments were run to help maintain the hydrophilic surface created by plasma

oxidation.

Optical alignment and temperature control

The laser and temperature controller were powered on and allowed to equilibrate

at least one hour prior to starting the experiments. The mirror above the microfluidic

chip was positioned so that the incident beam was directed onto the flow channel

orthogonal to fluid flow. The centroid of the backscattered interference pattern was

located just above the focusing lens on the end of a single-mode optical fiber, insuring

that the alignment of the system was along a central plane. The appropriate working

5

distance (283 mm) of the lens required to guarantee parallel rays impinge the channel was

measured physically and by adjusting the reflected laser spot size in the far field. A

Garry 3000 linear array CCD camera (Ames Photonics, Hurst, TX) was positioned near

direct backscatter in order to obtain a high-contrast fringe pattern with a dominant

Fourier frequency, generally near the fifth or sixth fringe from the centroid. The phase of

this dominant frequency is the BSI fringe shift signal, in radians.

The “fringe shift” is illustrated by the simple function sin2(x

2) shown in a

graphical movie (S.mov) accompanying this document. While we emphasize this is not a

real fringe pattern, it does show what is meant by “shift” of the pattern as the phase of the

sine argument is incremented (e.g. a binding event induced RI change).

Binding experiments were monitored in real-time at camera frame rates of 50-100

Hz recorded by the three thousand (7 m wide by 200 m tall) pixels. All experiments

were conducted at 25.00 ± 0.01°C maintained by a thermoelectric cooler (Peltier device)

connected to a Melcor MTCA 6040 temperature controller (MELCOR, Trenton, NJ).

Supporting Text

The source of BSI signal is not T

It is common practice to employ the second and third laws of thermodynamics to

describe changes in the system and its surroundings when a chemical reaction takes place

between two or more compounds. The function that describes the system itself as well as

the spontaneity of the reaction is that of Gibbs energy: G = H –TS, where H is enthalpy,

T is temperature, and S is entropy. Gibbs free energy can be described as the maximum

obtainable non-PV work, at constant temperature and pressure. For a system at constant

6

pressure and temperature this function has a minimum value when that system is at

equilibrium, leading to the fact that there must be a decrease in G for a reaction to occur

spontaneously (S1). The change in Gibbs energy, G, when two different species react

with each other (e.g. Protein A and IgG Fc ) can be calculated as:

aKRTG ln= , (1)

where R is the gas constant: 8.314 J/(mol•K); T is temperature in kelvins; and Ka is the

association constant, which is on the order of 108 M

-1 for the PA – IgG pair. By

substituting these values into Eq. 1 and using T = 298.15 K, we obtain G = - 45661

J/mol. This number is comparable to the G and H values published in the literature for

IgG and other compounds with similar Ka value (S2-S6). In the 350 pL probe volume

containing [Fc] = 40 nM (the highest concentration used in these experiments), there

would only be 14 10-18

mol of IgG Fc. The total change in energy due to the binding

event is E = - 45661 J/mol • 14 10-18

mol = -0.64 10-12

J. If we assume the worst case

scenario, when all of this energy is used to change the temperature of the solution (E=Q),

which can be calculated as

TmcQ = , (2)

where Q is heat energy in joules; m is mass of water in the probe volume (m = 0.35 10-6

g); c is the specific heat capacity of water (c = 4.186 J/(g•K)), then the corresponding

temperature can be calculated as follows

T =Q

mc= 0.44 μK . (3)

7

Most of the fluids exhibit large temperature dependence of the refractive index,

measured in RIU. For water dn dT = 1 104 RIU/K. Thus the change in the refractive

index induced by this T is found as n = 0.44 10-6

K • 1 10-4

RIU/K = 44 10-12

RIU.

This refractive index change is four orders of magnitude smaller than the best detection

limit obtainable with BSI in PDMS of 5 10-7

RIU. Note, we typically operate at RI =

10-6

. It is important to keep in mind that since we used G in these calculations and not

H, the observable in calorimetry, the estimated temperature change due to the binding

event is significantly overestimated (approximately one order of magnitude) and does

represent the worst-case scenario in our experiments. These calculations illustrate that

the signal that was detected for the free solution binding event was generated and

detected as a permanent refractive index change due to the reaction itself and not as a

heat of reaction. BSI measurements are made under isothermal conditions maintained

by a temperature-controlled 300 g aluminum heat sink beneath the chip.

Kinetic determination of binding affinity

Kinetic data recorded during BSI experiments were analyzed according to the

following derivation. Since two reactants are mixed on-chip in solution phase BSI

molecular interactions, a quantitative solution to a generic, bimolecular reversible

reaction was modeled by the analytical solution of a homogeneous linear first order

ordinary differential equation (ODE). The generic reaction is presented below:

[R] + [L] [R•L]k1

k2

[R] + [L] [R•L]k1

k2

(4)

8

where 1k and 2k are the forward and reverse rate constants, respectively, [ ]R and [ ]L are

the concentrations of the two reactant species, and [ ]LR • is the concentration of the

product formed. The forward reaction rate at any time, t , proceeds as k1

R[ ]t

L[ ]t

while the reverse reaction rate of the product progresses by k2

R • L[ ]t. The net rate of

this reaction is defined by:

d R • L[ ]

dt= k

2R • L[ ]

t+ k

1R[ ]

tL[ ]

t.

The time-dependent concentrations of R and L make this a coupled differential equation

to be simplified. Mass balance requires the amount of free R at any time, R[ ]t, be equal

to the difference in its initial concentration, R[ ]0

, and the amount consumed in the

formation of the product, expressed by Equation 6:

R[ ]t= R[ ]

0R • L[ ]

t.

Likewise, the mass balance equation for L is:

L[ ]t= L[ ]

0R • L[ ]

t.

Equations 6 and 7 are substituted into Eq. 5 to yield the following nonlinear ODE:

d R • L[ ]dt

= k2

R • L[ ]t+ k

1R[ ]0

R • L[ ]t{ } L[ ]

0R • L[ ]

t{ }.

The product of time-dependent terms in curly brackets makes Eq. 8 nonlinear. While

solutions exist for Eq. 8, they are not in a form amenable for fitting to data to obtain

values for the rate constants k1 and k2 , and ultimately the equilibrium dissociation

constant,

Kd=k2

k1

.

(5)

(6)

(7)

(8)

(9)

9

A suitable fitting equation can be obtained by choice of experimental conditions that

permit the approximation, to be justified at the end of the analysis, that L remains in

excess of product R•L to the extent that [L] is essentially constant at all times during the

reaction. That is, L[ ]t L[ ]0

R • L[ ]max R • L[ ]t L[ ]0

R • L[ ]t{ } L[ ]0

,

allowing the rightmost curly bracketed term of Eq. 8 to be set to L[ ]0

. (Quantitative

justification of approximation appears below.) Algebraic manipulation generates:

d R • L[ ]dt

+ (k1

L[ ]0 + k2 ) R • L[ ]t= k

1R[ ]0 L[ ]0 .

Eqn. 10 is a non-homogeneous linear first order ODE of the form

f (t) + A f (t) = B,

where A and B are constants, A = k1

L[ ]0+ k

2 and B = k

1R[ ]0

L[ ]0

. The associated

homogeneous (right hand side equals zero) ODE is f (t) + A f (t) = 0 , which can be re-

arranged to df f = d ln( f ) = A dt and integrated to give

ln( f ) = A t + C`

f (t) = C eA t,

where C is the single integration constant of the first order ODE. Knowing the solution

to the homogeneous ODE, and that linear combinations of ODE solutions are themselves

solutions (S7, S8), any particular solution of the non-homogeneous ODE can be added to

the general solution of the homogeneous ODE to form a general solution to the non-

homogeneous ODE. A particular solution of the non-homogeneous ODE for this system

can be found at the equilibrium state, where the net reaction rate is zero, f ( ) = 0 ,

yielding f ( ) + A f ( ) = B f ( ) = B A , which is then added to Eq. 12,

resulting in

(10)

(11)

(12)

10

f (t) = C eA t

+B

A.

Finally, the initial boundary condition, f (0) = 0 , can be imposed on Eq. 13 such that

f (0) = 0 = C e0+ B A C = B A , giving the desired general solution in Eq. 14,

f (t) =B

A1 e

A t( ).

This is recognized as an exponential rise to maximum of B/A with characteristic rise

constant of A. The two parameters A and B will be used to fit Eq. 14 to BSI kinetic data.

By comparing Eq. 14 to Eq. 10, the two parameters can be equated to chemical quantities:

A = (k1

L[ ]0 + k2 ) kobs

is called the “observed rate constant”, and B = k1

R[ ]0

L[ ]0

is

the forward reaction rate identified by Eq. 5. By substituting values of A and B into Eq

14 we obtain

[R • L]t=k1[R]

0[L]

0

k1[L]

0+ k

2

1 ek1 [L]0 +k2( ) t{ },

or, in briefer form:

R • L[ ]t = R • L[ ]max

1 ekobs

t{ }.

This result is similar to one derived by Atkins and de Paula (S9) for what is sometimes

referred to as pseudo-first-order kinetics.

Nonlinear regression is used to fit Eq. 16 to time-dependent BSI signals from

molecular interactions, providing quantitative kinetics parameters. B/A = [R•L]max is the

maximum product formed, and kobs is the exponential rise time constant related to the

binding affinity for the molecular interaction. Plotting the observed rate (kobs) determined

from multiple experiments with varied initial concentrations of L should exhibit a linear

(14)

(15)

(16)

(13)

11

relationship over the concentration range of L, as depicted in Fig. S1, where each point

represented by a square results from a fitting of parameters in Eq. 16 to time-dependent

BSI data. The linear equation kobs = k1 L[ ]0+ k2 has slope of 1k and y-intercept of k2.

Since Kd= k

2k1

, the y-intercept divided by the slope of the best fit line for the graph

of obsk versus L[ ]0

determines the equilibrium dissociation constant.

Interactions of CaM with all ligands were analyzed in this manner. Observed

rates determined from exponential fits of kinetic traces (Fig. 3 in the main article) were

plotted versus the concentrations of the various ligands and displayed in Fig. S2. All

plots exhibit excellent linear response to concentration with coefficients of correlation

ranging from 0.989 (Fig. S2C) to 0.997 (Fig. S2D). Equilibrium dissociation constants

(Kd) can be determined from each graph to yield the binding affinity of each interaction.

Division of the y-intercept by the slope of each best fit line yields the Kd values listed in

Table 2 of the main article.

Validation of approximation

The approximation of constant ligand concentration during a reaction is affirmed

from the linearity of all four data sets shown in Fig. S2. If the model equation were not

valid, significant departures from linearity at low ligand concentrations would betray the

approximation. A quantitative test of the approximation is performed by using the rate

constants obtained from the linear least-squares results, along with the initial

concentrations in the experiments, to compare [R•L]t and [L]0. In each of the four binding

assays there is a constant initial receptor concentration [R]0, and a minimum initial ligand

concentration, denoted here as [L]0,min on the left end of each line. The most testing

12

(‘worst’) case in each of the four can be compared with the ratio [L]0,min/ [R]0, which are

12/5, 5/2, 10/10, and 5/5 for Fig. S2A, S2B, S2C, and S2D, respectively. Of these, the

last two are most susceptible to ligand depletion (violation of the approximation) during

their reactions. Fig. S2C, CaM + Calcineurin, is chosen for this test since it has the

poorest R2 coefficient of linear correlation and therefore represents the worst case. In the

next paragraph, it is only this reaction to which reference is made.

The initial CaM concentrations are all [R]0 = 10 nM, [L]0,min = 10 nM, the linear

fit equation can be written as kobs

= 0.006 L[ ]0,min + 0.100 kobs

= 0.160 at [L]0,min ,

the slope, k1 = 0.006, and from Eqs. 12 and 13, R • L[ ]max

= k1

R[ ]0

L[ ]0( ) k

obs.

Finally, the numerical value of R • L[ ]max

= 3.74 nM, which is less than [L]0,min = 10 nM,

supporting the approximation under test. Moreover, during the binding reaction,

0 R • L[ ]t 3.74 nM , further validating the approximation.

Thus, since the approximation is valid for the single most testing initial conditions

of the 24 time-dependent BSI determinations of CaM binding, it is justified for all these

experimental conditions.

End-point analysis

Saturation binding is a second method for the determination of the equilibrium

dissociation constant Kd, useful when the reaction kinetics are at extrema in the time

domain (i.e. too fast for accurate kinetic measurements or too slow for expeditious

results). Based on the law of mass action ([L] = [L]0 – [R•L] and [R] = [R]0 – [R•L] )

substituted into the equilibrium expression for KD, and assuming the reaction reached

13

equilibrium, the maximum binding signal, [R•L] , (calculated as the relative change in

phase) can be expressed in terms of the concentration of the binding partner, [L]0 (ligand,

substrate, etc.) resulting in Eqn. 17, sometimes called ‘hyperbola’.

R • L[ ] =R[ ]

0L[ ]

0

L[ ]0+ K

d

.

Data analysis software can be used to determine the fitting parameters Kd and [R]0

with their uncertainties at any confidence interval. Saturation binding experiments, also

called end point analysis or steady-state binding, are routinely used for the determination

of Kd. Kinetic traces for molecular interactions detected during BSI experiments are fit

to a single exponential of the form: y = a(1 ex) . A nonlinear least squares regression

algorithm is used to analyze kinetic data and is easily accessible by multiple programs

including: SigmaPlot®, Origin

®, Prism

®, etc. Nonlinear regression uses iterative

optimization procedures to compute the parameters a and for each molecular

interaction curve. The parameter a describes the asymptotic region of the binding curve

where steady-state conditions have been reached. The end point values, a , determined

by nonlinear regression exhibit a hyperbolic relationship as a function of the

concentration of ligand. Fitting Eq. 17 to these data allows a second determination of the

equilibrium dissociation constant. The biostatistical software package Prism® (Graphpad,

San Diego, CA) was employed to fit the end point data and compute Kd (Fig. S3). The

equilibrium dissociation constants determined by steady-state analysis were found to be

in excellent agreement with our kinetically obtained values as well as previously

published results. Analysis of these plots determined Kd values of 17.77 M for Ca2+

,

7.64 M for TFP, 11.57 nM for CaN, and 11.13 nM for the M13 peptide.

(17)

14

IL-2 calibration curve

The Interleukin-2 (IL-2, R&D Systems, Minneapolis, MN) calibrations was

performed in solution (4 mM HCL with 0.1% FBS) using concentrations varying from

10-100 pM . Each concentration of IL-2 was introduced into the channel, allowed to

pressure and temperature equilibrate and resulting signal was recorded. The average of

three runs for each concentration is plotted in Fig. S4. The linear least-squares fit yields a

3 detection limit of 40 pM with R2 = 0.985. At this detection limit, there are about

10,800 molecules or 270 attograms of protein in the probe volume (490 pL).

The interaction between IL-2 and its antibody (IL-Ab, Abcam, Cambridge, MA)

was studied in solution. The IL-Ab concentration was held constant at 2 nM (in PBS, pH

7.4) and the IL-2 concentration was varied from 10-100 pM. The IL-2 and IL-Ab were

introduced and mixed on-chip and the interaction was monitored in real-time (Fig. S5).

The IL-2 – IL-Ab interaction assay was then repeated in cell media (RPMI 1640

with 1% FBS and 10 g/mL Cipro). The IL-Ab concentration was again held constant at

2 nM and the IL-2 concentration was varied from 10-100 pM. A kinetic analysis of this

assay was performed in the same fashion as described above and yielded a linear plot

(Fig. S6) from which the Kd was determined.

On-chip mixing and BSI signal

Prior to performing the sequential binding assays the appropriate range finding

experiments are done to optimize the flow rate to produce a homogeneous mixture of

reactants, without allowing appreciable progression of the reaction. Next, the lowest

15

concentration of the binding pair is introduced into the system, the flow is stopped and

the reaction is observed. Fig. S7 is an example of a real-time trace of the signal before

and after stopping the vacuum-induced flow. The time scale is arbitrarily set to zero

when flow is stopped. At negative times, freshly mixed reactants are passing through the

laser-illuminated interaction region.

16

0 1 2 3 4 5 60

2

4

6 k2 = y-intercept

k1 = slope

Concentration of L

ko

bs (

)

Figure S1. Expected linear relationship of the observed rate versus the initial

concentration of L.

17

0 10 20 30 40 50 600.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

R2 = 0.998

KD = 2.89 ± 0.737 nM

Concentration of M13 (nM)

ko

bs

0 25 50 75 100 1250.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

R2 = 0.990

KD = 3.40 ± 2.86 μM

Concentration of Ca2+

(μM)

ko

bs

C. Calcineurin D. M13

A. Calcium B. Trifluoperazine Dihydrochloride

0 25 50 75 100 1250.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

R2 = 0.989

KD = 15.64 ± 3.25 nM

Concentration of CaN (nM)

ko

bs

0 5 10 15 20 25 300.0

0.1

0.2

0.3

0.4

0.5

0.6

R2 = 0.996

KD = 4.73 ± 0.483 μM

Concentration of TFP (μM)

ko

bs

Figure S2: Observed rates determined by exponential fits to kinetic traces were plotted

versus various concentrations of Ca2+

(A), Trifluoperazine dihydrochloride (B),

Calcineurin (C), and the peptide sequence from myosin light chain kinase (D).

18

0 10 20 30 40 50 600.0

0.1

0.2

0.3

R2 = 0.995

KD = 9.87 ± 1.12 nM

Concentration of M13 (nM)

Ym

ax

0 5 10 15 20 25 300.0

0.1

0.2

0.3

0.4

0.5

R2 = 0.9949

KD = 7.82 ± 0.911 μM

Concentration of TFP (μM)

Ym

ax

C. Calcineurin D. M13

A. Calcium B. Trifluoperazine Dihydrochloride

0 25 50 75 100 1250.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

R2 = 0.998

KD = 11.39 ± 0.819 nM

Concentration of CaN (nM)

Ym

ax

0 25 50 75 100 1250.00

0.25

0.50

0.75

1.00

R2 = 0.9975

KD = 18.23 ± 1.43 μM

Concentration of Ca2+

(μM)

Ym

ax

Figure S3: Steady-state analysis of the CaM binding to A) Ca2+

, B) TFP,

C) Calcineurin, and D) M13.

19

0 25 50 75 100 1255.75

5.80

5.85

5.90

5.95

6.00

Concentration of IL-2 (pM)

Sig

na

l (r

ad

)

Figure S4: Calibration curve of IL-2. Each point is an average of three runs

with error bars representing standard deviation.

20

0 30 60 90 120 150 1800.00

0.02

0.04

0.06

0.08

0.10

100pM

0pM

0pM

100pM

20pM

40pM

60pM

80pM

[IL-2][IL-Ab] = 2 nM

Time (sec)

Sig

na

l (r

ad

)

Figure S5: Real-time IL-2 – Ab binding curves with interaction assay performed in free

solution. Note: no signal change observed in 3 control experiments (Black: IL-2 buffer

mixed with Ab buffer ([IL-Ab] = 0 nM), Red: max IL-2 concentration mixed with Ab

buffer ([IL-Ab] = 0 nM), and Blue: IL-2 buffer mixed with 2 nM Ab) .

21

0 10 20 30 40 50 600.000

0.025

0.050

0.075

0.100

KD = 25.91 ± 5.24 pM

R2 = 0.914

Concentration of IL-2 (pM)

ko

bs

Figure S6: Observed rates determined by exponential fits of kinetic traces for the

IL-2 – IL-Ab binding assay in cell media.

22

Vacuum Stopped

Binding Starts

-0.1 0.0 0.1 0.2 0.3 0.42.27

2.28

2.29

2.30

Time (sec)

Ph

ase (

rad

)

Figure S7: Time-dependent BSI signal before (t < 0) and after (t > 0) flow of reactants

is stopped by removing vacuum nozzle from the chip outlet port. Total time shown is

second from a reaction lasting from several seconds to 3 minutes in these experiments.

23

References:

S1. K. J. Laidler, J. H. Meiser, Physical Chemistry (Houghton Miflin Company,

Boston, 1999).

S2. M. Mutz, T. Hawthorne, S. Ferrone, G. Pluschke, Molecular Immunology 34,

695-707 (JUL, 1997).

S3. I. Jelesarov, H. R. Bosshard, Journal of Molecular Recognition 12, 3-18 (JAN-

FEB, 1999).

S4. I. Jelisarov, L. Leder, H. R. Bosshard, METHODS: A companion to Methods in

Enzymology 9, 533-541 (1996).

S5. G. Pluschke, M. Mutz, Journal of Thermal Analysis and Calorimetry 57, 377-388

(1999).

S6. R. Ghirlando et al., Biochemistry 34, 13320-13327 (OCT 17, 1995).

S7. E. W. Weisstein. (MathWorld - A Wolfram Web Resource, 2007), vol. 2007.

S8. W. E. Boyce, R. C. DiPrima, Elementary Differential Equations and Boundary

Value Problems (John Wiley & Sons, New York, ed. 4th, 1986).

S9. P. Atkins, J. de Paula, Physical Chemistry (W.H. Freeman, New York, 2001).

S.mov: This is a good description of how the fringes shift position in the x-direction,

away from the directly backscattered light situated at x=0 with increasing phase. This

simple function is in excellent agreement with the proposed model in reference 26 and

32. The phase is changed to illustrate the signal change do to changes in refractive index,

see reference 33.