Acceleration of Surface-Based Hybridization Reactions Using

Feb 11, 2014 - We present a theoretical model and experimental demonstration of a novel method for acceleration of surface-based reactions using isota...
0 downloads 9 Views 2MB Size
Article pubs.acs.org/ac

Acceleration of Surface-Based Hybridization Reactions Using Isotachophoretic Focusing Merav Karsenty,§ Shimon Rubin,§ and Moran Bercovici* Faculty of Mechanical Engineering, Technion − Israel Institute of Technology, Haifa 32000, Israel ABSTRACT: We present a theoretical model and experimental demonstration of a novel method for acceleration of surface-based reactions using isotachophoresis (ITP). We use ITP to focus a sample of interest and deliver a high concentration target to a prefunctionalized surface, thus enabling rapid reaction at the sensor site. The concentration of the focused analyte is bound in space by the ITP interface and, upon reaction with the surface, continues electromigrating downstream, removing any contamination or reacted sample molecules from the surface. This constitutes a one-step reactand-wash assay which can be performed in a simple channel and does not require flow control elements or moving parts. We designed a novel microfluidic chip where reaction surfaces are formed by paramagnetic beads, immobilized at desired sites by an external magnetic field. Using this chip, we compared ITP-based surface hybridization to standard continuous flow-based hybridization and experimentally demonstrated a 2 orders of magnitude improvement in limit of detection (LoD) in a 3 min nucleic acid hybridization assay. The simple analytical model we present allows prediction of the rate of surface reaction under ITP and can be used to design and optimize such assays as a function of the physical properties of the system, including buffer chemistry, applied voltage, analyte mobility, analyte concentration, probe density, and surface length. The method, model, and experimental setup can be applied to various forms or surface reactions and may serve as the basis for highly genetic analysis and immunoassays.

S

Isotachophoresis (ITP) is an electrophoretic technique in which analytes of interest can be focused at the interface of two distinct electrolyte solutions characterized by high and low electrophoretic mobilities. Using ITP, Jung et al. demonstrated a concentration factor of a million-fold, in less than 2 min.15 Bercovici et al. demonstrated the use of ITP for accelerating the hybridization reaction of ionic species cofocused at the ITP interface. Cofocusing target DNA with molecular beacons probes, the authors demonstrated 10 000 fold acceleration of hybridization kinetics.16 Persat et al. and Bercovici et al. demonstrated the use of molecular beacons and ITP for rapid detection of miRNA and 16S rRNA, respectively.17,18 While molecular beacons are a highly attractive mechanism for detection of nucleic acids in the bulk, the majority of bioassays require surfaces to separate free probes from reacted probes and wash any nonspecific molecules. Schudel et al.6 recently reported the use of surface-bound molecular beacons for multiplexed detection of viral sequences. The authors reported a limit of detection of 10 nM LoD after a 1 h hybridization time and showed that the limiting step in such assays is the hybridization rate. We here present a first proof of principle for accelerated reactions on surfaces using ITP, demonstrating 2 orders of

urface-based biosensors are some of the most common type of sensors for biological targets such as nucleic acids and proteins. In most implementations, they are based on a “capturing probe” (e.g., an antibody or synthetic DNA sequence) which is immobilized on a surface and to which targets specifically bind. Detection of the binding events can then be obtained in various ways, including, for example, fluorescence, electrochemical signals, or surface plasmon resonance (SPR).1−3 Regardless of the binding or transduction mechanism, the sensitivity of all surface biosensors is fundamentally limited by the rate at which target molecules bind to the surface. Several factors, namely diffusion, transport, and reaction rates, limit hybridization or binding at low concentrations.4−6 While diffusion and transport limitation can be effectively overcome by use of devices such as mixers and flow channels,7,8 reaction rates remain a major bottleneck toward achieving rapid binding of biomolecules at low concentrations. This is because hybridization and binding typically take the form of second-order reactions, with reaction time inversely proportional to the concentration of the reactants.4,5 A typical example where such surface reactions are employed is in microarrays, where thousands of biomarkers can be probed simultenously,3,9−11 but due to reaction kinetic limitations12,13 require long incubation times, up to several hours.14 There is thus a growing need for methods that significantly accelerate reaction rates and lower detection time. © 2014 American Chemical Society

Received: November 26, 2013 Accepted: February 11, 2014 Published: February 11, 2014 3028

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

magnitude improvement in signal compared to a standard flowthrough reaction. We use ITP to deliver a highly focused sample to a prefunctionalized surface, thus enabling rapid reaction at the sensor site. Importantly, since the sample is delivered to the surface in a finite “packet”, no additional wash steps are required, and any contamination or unhybridized molecules continue electromigrating downstream with the moving ITP. This constitutes a one-step react-and-wash assay which can be performed in a simple channel and does not require flow control elements or moving parts. To facilitate rapid turnaround time in surface hybridization experiments, we designed a novel microfluidic chip where wellordered and confined reaction surfaces are formed by paramagnetic beads, immobilized at desired sites by an external magnetic field. The beads are prelabeled with desired captureprobes, eliminating the need for surface modifications, and enables reusability of the chip. In support of our experimental results, we provide in the theory section a detailed, yet simple, analytical model which can be used to predict the rate of surface reaction under ITP as a function of the physical properties of the system, such as ITP chemistry, applied voltage, analyte mobility, probe density, and length of the surface.

Figure 1. Schematic illustration of the surface hybridization assay. An analyte is focused at the sharp ITP interface between the L-ion and Tion, which acquire constant concentration values at the far LE and TE regions, and electromigrates at a velocity VITP. At a distance x0 from the T-ion reservoir, the analyte encounters a surface containing immobilized probes having a surface density b0. As the concentrated analyte passes over the surface, its high concentration rapidly reacts with the surface probes with on- and off- rates of kon and kof f, respectively. The analyte overlaps with the surface for a duration of w/VITP, after which it continues electromigrating downstream.



⎛ μTE ⎞ dNa = (μaTE − μTTE )ETEcaTEA = ⎜⎜ aTE − 1⎟⎟caTEVITPA dt ⎝ μT ⎠

THEORY AND PRINCIPLE OF THE ASSAY In ITP sample ions are focused under a spatially nonuniform electric field between a high mobility leading ion (L) and a low mobility trailing ion (T). In contrast to plateau mode ITP, where species achieve sufficiently high concentrations to form distinct and separate zones,19 peak mode ITP describes cases in which one or more species focus at the sharp electric field gradient between the leading electrolyte (LE) and trailing electrolyte (TE) zones and have sufficiently low concentrations such that their effect on the electric field is negligible.20 Peak mode ITP is the typical focusing mode for biological macromolecules such as proteins and nucleic acids. To achieve focusing, sample analytes can be mixed with the L-ion or T-ion buffers (typically referred to as “infinite sample injection”) or injected in an independent zone between these electrolytes (referred to as “finite sample”). For simplicity, we will here address the case of infinite sample injection in which sample ions are mixed with the T-ion and continuously accumulate at the ITP interface. However, the method and analysis can be directly applied to other injection modalities. As illustrated in Figure 1, we consider a simple, straight microfluidic channel having a cross section area A, height h and containing a finite segment of immobilized probes centered at a distance x0 from the West (TE) reservoir. We denote the surface density of the probes as b0, and the length of the reactive surface as S . The initial conditions for ITP are formed simply by filling the channel and East reservoir with LE solution and filling the West reservoir with a mixture of TE and target molecules. Under an applied electric field, the ITP front, containing focused analyte molecules, electromigrates from West to East. At steady state, both the L- and the T-ions electromigrate at LE TE LE TE = μTE the same velocity VITP = μLE L E T E , where μL and μT are, respectively, the effective mobilities of the L- and T-ions in the LE and adjusted TE zones; ELE and ETE are the electric fields in the two zones. Consequently, the rate of analyte ions entering the ITP interface is governed by the difference in velocity between the analyte (in the adjusted TE) and the interface.

(1)

which, by relating location and time through x = VITPt, leads to ⎛ μTE ⎞ Na(x) = ⎜⎜ aTE − 1⎟⎟caTEAx ⎝ μT ⎠

(2)

Here cTE a is the total analyte concentration in the adjusted TE zones, which differs from its concentration in the well. Utilizing mass conservation at steady state, the concentrations can be related by equating the fluxes in the two zones, TE TE w w w LE LE μTE a E ca = μa E ca . Further utilizing the ITP condition, μL E LE LE w w TE TE = μT E , and current conservation σ E = σ E , we may write the concentration of the analyte in the adjusted TE zone as caTE =

μaw Ew

μaw Ew μTTE w μaw σ LE μTTE w w c = c = ca a a μaTE ETE μaTE E LE μLLE μaTE σ w μLLE

(3)

where σLE and σw are, respectively, the conductivies in the LE and in the well. Substituting into eq 2, we finally obtain

Here ηT is a dimensionless buffer dependent gain factor, which determines the gain in the number of target molecules accumulated at the ITP interface in the infinite injection case, as compared to a finite injection volume Ax. The gain factor may be obtained from electrophoresis simulations such as Spresso21 or Simul22 or from buffer calculators such as Peakmaster.23 Further simplification of eq 4 is however possible when T, L, and C are monovalent ions and when the same C-ion is used in both the LE and the TE. The LE conductivity can then be (fi) (fi) (fi) (f i) expressed as σLE = F(μ(fi) denote fully L cL + μC cC ), where μi (f i) ionized mobilities and ci denote ionic concentrations. Applying net neutrality, we may then write 3029

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry σ LE = F(μL(fi) + μC(fi))cL(fi)

Article

have sufficient opportunity to “sample” the surface. Clearly, for higher ITP velocities or deeper channels, this assumption no longer holds. Under this assumption, we may write the area averaged Nernst−Planck equation for the target molecule as

(5)

The total concentrations, ci, can be related to the fully ionized concentrations, ci(f i), via ci(f i) = γ ici, where γ i are the corresponding degrees of dissociation expressed as a function of the dissociations constants and the local pH, explicitly given by 1 1 γT = pKL − pH 1 + 10 1 + 10 pKT − pH 1 γC = 1 + 10 pH − pKC

∂ca ∂ca ⎤ ∂ ⎡ + ⎢⎣(zμE − VITP)ca − Deff ⎥⎦ ∂x ∂t ∂x 1 = [−koncab + koff b] h

γL =

where, consistently with the discussion above, all ca molecules in a given cross section are available for reaction with the surface, and the 1/h factor on the right-hand side transfers the surface density b to a volume concentration b/h. Def f is the effective diffusivity of the analyte, which, in the general case, may deviate from the molecular diffusivity, D, due to dispersion effects, and in particular Taylor−Aris type dispersion.25 In our experiments the relatively low height-based Peclet number, Pe = hU/D ≅ 0.25, indicates that the Taylor−Aris effective diffusivity coincides with the molecular diffusivity, Def f ∼D. Notably, the continuum description of our eqs 2 and 3 is expected to break down below critical concentration, corresponding to one target molecule, or less, per sensor volume. For example, for a typical sensor volume of 20 μm × 50 μm × 100 μm, equivalent to 100 pL, the critical concentration is of the order ccrit ∼10 fM. (ii) The relative change in the concentration of the target molecule as it passes over the surface is of the order Δca/ca ∼ konb/hΔt ∼ konbmS /hVITP. Taking kon ∼ 103−104 [M−1 s−1], bm ∼ 1012[copy/cm2], yields Δca/ca ∼ 10−3−10−2; i.e., during the relevant time scale, the amount of target bound to the surface is 0.1−1%. Furthermore, the typical length of the reactive surface (order ∼10−100 μm) is significantly shorter than the typical distance from the TE reservoir to the reaction site (order ∼1−10 cm), S ≪ x0. Hence, as the ITP interface passes over reactive surface, the relative change in analyte concentration, due to ITP accumulation, is of the order Δca/ca ∼ (S + w)/x0 ∼10−2; i.e., during τITP, the concentration of target ions increased by just 1%. Thus, the target concentration profile can be assumed to have a constant number of molecules and take the form of a translating wave of the type26

(6)

Similarly, the effective mobilities are related to the fully ionized mobilities via μi = γiμ(fi i), and eq 4 can then be recast as

where the fully ionized mobilities and the dissociation can, in general, depend also on ionic strength.24 In a particular case of a strong acid, characteristic to the case of nucleic acid relevant to this work, γwa = 1, and the expression can be further simplified to the form

Here all mobilities are fully ionized, all concentrations are total concentration and pH dependence enters only through the dissociation constant γTE T . This high concentration “packet” of target molecules steadily electromigrates down the channel and reaches the sensor surface which contains immobilized capture probes. Assuming Langmuir kinetics at the surface, and following the notation of Squires et al.,5 the concentration of surface probes bound by target molecules follows (in dimensional form) ∂b(x , t ) = kon[bm − b(x , t )]ca(x , t ) − koff b(x , t ) ∂t

(10)

(9)

ca(x , t ) = Na

where kon M−1 s−1 and koff s−1 are, respectively, the on- and offrates of the reaction, [bm − b(x,t)] is the surface density of available (unbound) probes, and ca(x,t) is the target molecule concentration at the surface. We now turn to a scaling analysis which shows (i) the problem may be assumed as one-dimensional, i.e., that the analyte concentration at the surface can be effectively represented by its cross-section average, and (ii) that the analyte concentration, i.e., its spatial distribution, remains practically constant during its reaction with the surface. (i) Assuming an ITP velocity of 50 μm/s, and a characteristic analyte zone width of w = 100 μm (typical in our experiments), we find that the interface transverses any point on the reacting surface at a characteristic time of

sin(πL/La) exp[(x − VITPt )/La] ALπ 1 + exp[(x − VITPt )/L]

(11)

where the interface and analyte length scales are defined, −1 respectively, as L−1 = FVITP/RT(μ−1 − μ−1 = T L ) and La −1 FVITP/RT(|za|μ−1 − μ ), F is Faraday’s constant, R is a gas T a constant, T is absolute temperature, and VITP is the velocity of the ITP interface. In our experiments, the analyte concentration profile is highly symmetrical, showing excellent fit (R2 value of 0.9928) to a Gaussian. The case of a symmetric analyte is characterized by μa ∼ μ(sym) = 2μLμT/(μL − μT), and expansion a of eq 11 to second order around its maximum reveals it can be well approximated as a Gaussian profile

τITP = w/VITP ∼ 2 s At the same time, assuming a channel height of h = 20 μm, and a diffusion coefficient of D = 5 × 10−9 m2/s, the characteristic diffusion time is of order τd = h2/D ∼ 0.1 s. Hence, while passing over the surface, target molecules at the ITP interface

where cmax is the peak concentration, expressed in terms of the initial concentration in the well, c0, and the preconcentration factor α. The width of the analyte distribution, w (which we 3030

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

the TE well to the reaction site, τtot = x0/VITP. The exact governing equation for the case of reaction under standard flow conditions (no ITP) is identical to eq 9, except with the preconcentration factor set to unity, fa = 1, indicating that the concentration of the analyte in contact with the surface remains c0. To obtain a fair comparison between the two reaction modalities, we compare the final normalized number of hybridized probes, b/bm, after an identical total time τtot. In the standard, pressure-driven flow case, the arrival time of the target molecules to the reacting surface is negligible, and τtot plays the role of the total reaction time. The ratio of these signals, indicating the enhancement due to ITP, is given by

define as the width enclosing 95% of the Gaussian area), is related to the buffer properties via FVITP μL − μT 1 = w 8RT μL μT (13) We note that while eq 13 provides the correct scaling for the analyte width, in practice, the width can be very sensitive to dispersion effects.25 Thus, throughout the work, we use the width measured in experiments for quantitative analysis. For each point on the surface (i.e., constant x), eq 9 takes the following form: db(t ) = konca(t )[bm − b(t )] − koff b(t ) dt

(14)

R(c0) = fα

where the concentration that the point experiences over a duration 0 ≤ t ≤ τITP is given by 2 ⎡ ⎛ ⎤ ⎛τ ⎞⎞ ca(t ) = cmax exp⎢− 8⎜VITP ⎜ ITP − t ⎟⎟ /w 2 ⎥ ; ⎝ 2 ⎠⎠ ⎢⎣ ⎝ ⎥⎦

(21)

(15)

R → fα

corresponding to transition of 95% of the sample between t = 0 and t = τITP. Replacing the concentration ca(t), in eq 7, by its time average ⟨ca⟩ = 1/τITP∫ τ0ITP ca(t)dt, explicitly given by

introduces negligible, relative error of an order kof fτITP to b(τITP) (as may be seen by integrating directly both sides of eq 7) yet leads toward the following simpler reaction equation:

where the corresponding time scales are given by τoff−1 = koff (18)

The corresponding solution of eq 9, which applies to times 0 ≤ t ≤ τITP and practically coincides with the exact numeric solution at t = τITP, is given by ⎛ t ⎞⎞ b(t ) τR ⎛ ⎜⎜1 − exp⎜ − ⎟⎟⎟ bm τon ⎝ ⎝ τR ⎠⎠

(19)

where the time scale which governs the reaction rate may be rewritten as

τR =

1 kon(fαc0 + Kd)

1 − exp( −koff τtot )

∼ fα

τITP τtot

(22)

(23)

representing the fact that at sufficiently high concentrations (with respect to Kd), which saturate the sensor, no significant gain is obtained from ITP. Clearly, eq 16 shows that the typical time needed for the standard flow method to match the ITP method. Specifically, for times greater than, τtot > fατITP, and low concentrations, the standard flow method becomes superior to the ITP method. Figure 2 presents this ratio of signals, as a function of the initial concentration at the well, for two typical values of Kd. Here the total reaction time for the standard flow-through conditions is taken as τtot = 180 s, whereas the ITP reaction time is taken as τtot = 2 s. To simulate conditions similar to those used in our experiments (see Experimental Section), we performed a simulation using Spresso21,27 from which we extracted the L-ion effective mobility as 69 × 10−9 m2/(V s) (corresponding to Cl−), the T-ion effective mobility as 18 × 10−9 m2/(V s) (corresponding to Tricine), and the counterion effective mobility as 8 × 10−9 m2/(V s) . We set the analyte mobility as 38 × 10−9 m2/(V s) (corresponding to DNA in free solution)28 and the ITP velocity and interface width as 50 μm/s and 100 μm, respectively. We position the reactive surface at a distance of 2 cm from the TE reservoir, yielding (using eqs 12, 13, and 16) α = 13 500, and fα ≃ 8000. Hence, while ITP increases the effective concentration over the surface by a factor 8000, it spends only 1/90 of the time over the surface, and thus the total enhancement, at low concentrations, is of the order R ∼ 90 Equation 21 and Figure 1 reveal that ITP-based reactions are particularly advantageous at low concentration, where c0 ≪ Kd. In fact, eq 14 implies that in the regime Kd/fα ≪ c0 ≪ Kd, sample focusing shifts the characteristic time scale from off-rate dominated to on-rate dominated, resulting in both faster reactions and a higher fraction of reacted sites. At lower concentrations, c0 ≪ Kd/fα, both the standard and the ITPbased reactions are governed by the off rate. Nevertheless, the

(17)

τon−1 = fαc0kon ;

1 − exp( −koff τITP)

where we assumed τITP ≪ τof f, τtot≪ τoff and expanded the exponents up to the first order in their arguments. At high concentrations, set by the corresponding equilibrium constant, the ratio decays to a lower bound of R→1

τR −1 = τon−1 + τoff−1;

koff + fαc0kon 1 − exp( −(koff + c0kon)τtot )

Clearly, at low concentrations, we obtain an upper bound on the signal ratio given by

0 ≤ t ≤ τITP

b db(t ) b(t ) = m − dt τon τR

koff + c0kon 1 − exp( −(koff + fαc0kon)τITP)

(20)

with the reaction constant Kd = kof f/kon setting a typical concentration scale in our problem. The latter introduces a natural scale which separates the low concentration and high concentration regimes, governed by the off-rate and by the onrate, respectively. It is worth noting that owing to the fact that in our setup the diffusion time, τd, is at least 4 orders of magnitude smaller than the reaction time, τR, the corresponding surface reactions are not diffusion limited. We further note that τITP corresponds to the duration in which the focused sample overlaps with the reactive surface. The total assay time is however typically much longer and is dominated by the electromigration time of the interface from 3031

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry



Article

EXPERIMENTAL SECTION

Magnetic Bead-Based Reactive Surfaces. The common methods for patterning of capture probes on a surface require multiple well-controlled chemical steps and strongly depend on the substrate used.29 The majority of surface chemistries are performed by silanol chemistries. Prerequisite for grafting of silane is the availability of surface hydroxyls groups, which can readily create covalent bonds with such silanes. Silicon substrates, whose widespread use originates from the semiconductor industry, has thus gained popularity as a substrate for immobilization of biomoleculs, owing to its high density of surface hydroxyls. Unfortuantely, silicon surfaces are inherently conducting, and thus are incompatible with electrokinetic assays in which high electric fields are used. Nonconductive substrates, such as glass, PDMS, and plastics are suitable for electrokinetic assays but have comparatively lower density of free surface hydroxyls. Obtaining high quality and repeatable patterning on such a surface is thus far more challenging.29 Furthermore, enclosing finite length, welldefined, functionalized surfaces within microfluidic channels is a challenging task, as the majority of microfabrication bonding techniques (e.g., plasma treatment, thermal bonding) are destructive to organics.30 Lastly, microfluidic channels containing a functionalized surface can be used only once and must be replaced between experiments. While this is acceptable and even required for diagnostic products, in a laboratory setting it results in long turnaround time per experiment and decreased reproducibility. To overcome these diffulties, we designed a microfluidic chip in which reaction surfaces are formed by streptavidin-coated paramagnetic beads, immobilized at desired sites by an external magnetic field. The beads are prelabeled with capture-probes (e.g., biotinylated oligonucleotides or biotinylated antibodies), eliminating the need for surface modifications and enabling reusability of the chip. As illustrated schematically in Figure 4 and experimentally in Figure 5, our microfluidic chip includes a 5 μm deep “trench” of dimensions 100 μm × 30 μm, in which magnetic beads are concentrated and trapped, creating a uniform and well-defined surface. Figure 3a and 3b presents experimental images of captured magnetic beads in a standard flat channel and in the trench design, respectively. Without the trench, while beads can be held in place, they are randomly and uncontrollably scattered within a long region of the channel. Furthermore, we found that beads held in this configuration are often displaced or removed upon applying an electric field. In contrast, the trench design confines the beads to a well-defined region, creating

Figure 2. Analytical model results showing the ratio of surface hybridization fractions between ITP-based and standard flow hybridization, as a function of the initial target concentration. For illustration of the physical behavior, we present the results for two different kon values, 102 and 106 M−1 s−1 (resulting in different Kd values). ITPbased reactions provide significant signal enhancement in the regime c0 ≪ Kd, where, in addition to a higher total number of target molecules delivered, sample focusing accelerates the characteristic reaction rate from τof f = 1/koff to τon = 1/( fαc0kon). At higher concentrations the initial concentration in the well (i.e., before focusing) is sufficient to saturate the sensor, and no significant gain is obtained from ITP. Here koff = 10−6 s−1, α = 13500, the total reaction time for the standard flow-through conditions is τtot = 180 s, and the ITP reaction time is τtot = w/VITP = 2 s (S = 10−4 m; w = 10−4 m; VITP = 0.5 × 10−4 m/s).

gain from ITP is maintained as the number of target molecules delivered to the surface is higher by a factor of fα. In contrast, at higher concentrations, c0 ≫ Kd, the gain from ITP quickly diminishes. This is because, even without focusing, and given the allocated reaction time, the initial concentration is sufficient to saturate the sensor. Thus, while at higher concentrations ITP still enables faster reaction rates, they are no longer important when the allowed assay time is much longer than both reaction times.

Figure 3. Raw fluorescence images of 2.8 μm magnetic beads immobilized in a microchannel by an external cylindrical NdFeB magnet (1/8“X3/8” grade N52): (a) beads trapped in a standard flat bottom microchannel. The beads are successfully immobilized but create a nonuniform and nonrepeatable distribution. (b) The same magnet is used to capture the same beads within a microfluidic trench, creating a uniform and well-defined surface. 3032

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

Figure 4. Schematic illustration of the assay. (a) A microfluidic channel connecting two reservoirs is initially filled with LE and prelabeled paramagnetic beads. The solution is flowed through the channel by applying pressure, and beads are trapped in the designated trench under the magnetic force of a permanent magnet placed on top of the chip. (b) A mixture of TE and sample is injected in the West reservoir, and an electric field is applied between the two reservoirs to initiate ITP. (c) Highly focused target molecules are delivered to the reactive surface by ITP and rapidly react with the probes on the surface. (d) Unbound target molecules leave the reaction site and continue electromigrating with the ITP interface, leaving the surface embedded in TE buffer.

reaction spot which is convenient to analyze. Using this configuration, we were able to apply electric fieds as high as 300 V/cm, without any interruption to the beads’ location. Experimental Setup. An SU8 mold was fabricated at Stanford Microfluidic Foundry (Stanford University, Stanford, CA, http://www.stanford.edu/group/foundry/) using standard lithography. A two-layer design was used: the first layer was spun to 30 μm in thickness, and a mylar mask was used to define a 50 μm wide, 40 mm long straight channel. The second layer was spun to 5 μm and was used to define the 30 μm wide × 100 μm long bead capturing trench, located at the center of the channel. Using the mold, microfluidic chips were fabricated in-house from PDMS, using a cross-linker to monomer ratio of 1:10 and using standard protocols.31 Importantly, before baking, the PDMS was spun to a thickness of approximately 300 μm using a commercial spincoater (Larurell WS-650Mz23NPP, North Wales, PA). The PDMS thickness dictates the distance between the external magnet and the channel and thus directly affects the magnetic force. We obtained images using an inverted epifluorescent microscope (Ti−U, Nikon, Tokyo, Japan) equipped with a metal halide light source (Intensilight, Nikon Japan), a 10× (NA = 0.45, WD = 4 mm) Nikon PlanApo objective and Chroma 49006 filter-cube (620/60 nm excitation, 700/75 nm emission, and 660 nm dichroic mirror). Images were captured using a 16 bit, 2560 × 2160 pixel array CMOS camera (Neo, Andor, Belfast Ireland) cooled to −40 °C. Images were taken before and after the hybridization process, using an exposure time of 100 ms. During the hybridization process itself, the light source was shuttered to prevent photobleaching of the

immobilized probes. We controlled the camera using NIS Elements software (v.4.13, Nikon, Japan) and processed the images with MATLAB (R2011b, Mathworks, Natick, MA). All ITP experiments were performed using constant voltage (defined in each of the figures) using a high voltage sequencer (HVS3000D, Labsmith, Livermore, CA). Standard flowthrough hybridization experiments were performed using a water column connected to one of the channel’s reservoirs. Magnetic Beads and Labeling Protocol. We used 2.8 μm paramagnetic beads (Dynal M280, Life Technologies, NY) and applied a standard protocol, based on the manufacturer’s instructions, to wash and label the beads. Briefly, 10 μL of the beads solution was mixed with a 90 μL of a 2X wash buffer containing 10 mM Tris, 20 mM HCl, 1 mM EDTA, and 2 M NaCl. A magnet was used to hold the beads in place, the fluid was discarded, and we repeated this process four times. We then mixed 10 μL of the washed beads with 10 μL of DI for a final salt concentration of 1 M and added 1 μL of 20 μM probes. We incubated the beads for 10 min, washed them with a 1X wash buffer, and finally suspended them in 100 μL of DI. We estimate the final concentration of the labeled beads as 1 mg/mL, or 6 × 107 beads/mL. Biotinylated Molecular Beacons and Target Sequences. The biotinylated molecular beacon probes were purchased from IDT (Coralville, Iowa). We designed a 22-mer probe sequence which is complementary to a preserved section of bacterial 16S rRNA.32 To the main probe sequence, we added six base-pairs on either side to form the molecular beacon stem. The 5′ terminus was labeled with Cy5, and the 3′ terminus was labeled with Black Hole Quencher 2 (BHQ2). On the basis of 3033

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

Figure 5. Raw fluorescence images showing an experimental demonstration of ITP-based surface hybridization. (a) Fluorescent DNA oligonucleotides are focused at the ITP interface and electromigrate toward the reaction zone composed of immobilized magnetic beads functionalized with complementary DNA probes (here still invisible). (b) The ITP interface passes over the reaction zone, allowing rapid hybridization of complementary strands. (c) The ITP interface leaves the reaction zone, carrying unhybridized free targets and leaving the surface in a clean buffer environment.

suggestions in Tan et al.33 and Schudel et al.,6 we placed the biotin label on the quencher side of the stem but as far as possible from its 3′ terminus. The final sequence is given by 5′-/Cy5/CCGGAC{TCGTTTACRGCGTGGACTACCA} GT*CCGG/BHQ2, where T* denotes the biotin modification. Target nucleic acid, complementary to the probe sequence, TGGTAGTCCACGCYGTAAACGA, was purchased from Sigma-Aldrich (St. Louis, MO). Bead Loading and Washing Process. For bead loading, we pipetted 2 μL of the prelabeled beads (at a concentration of 6 × 107 beads/mL) into the West reservoir filled with 18 μL of LE solution. We placed a cylindrical NdFeB magnet (1/8“X3/ 8” grade N52) on top of the channel, in direct contact with the PDMS, and created negative pressure of 60 cm H2O in the LE vial using a water column. Beads flow from the TE to LE reservoirs and are captured in the trench under the magnetic field. Beads that are not captured in the trench are washed by the flow. We allowed approximately 10 min for the beads to fill the trench and subsequently flushed the channel for 3 min with pure LE solution. To release the beads, we simply lifted the magnet and flushed the channel with DI until all beads were removed. Isotachophoresis Assay and Choice of Experiment Conditions. We chose the ITP chemistry such that sample focusing into the interface is maximized, while maintaining buffer capacity, and a pH close to neutral which is important for most hybridization and binding assays. As detailed in eq 1, a maximal focusing rate will be achieved by choosing a high ratio of LE to TE reservoir concentrations and a low electrophoretic mobility TE. In all experiments, we used 100 mM HCl, 200 mM Bistris, 5 mM MgCl2, and 1% 1.3 MDa polyvinylpyrrolidone (PVP). Chlorine ions are a natural selection as leading ions, due to their high pH-independent, electrophoretic mobility. We used PVP in the LE for suppression of electroosmotic flow (EOF),34 and MgCl2 for stabilizing DNA hybridization. As shown by Khurana et al.,20 a further increase in the LE concentration does not contribute significantly to the focusing rate due to ionic strength effects. TE was composed of a 1:2 ratio of tricine to Bistris. This combination results in a near neutral pH, which is one unit below the pKa of tricine (8.1). As a result, tricine ions are weakly ionized, resulting in low electrophoretic mobility (18 × 10−9 m2/(V s)),

which benefits focusing rate. Further reduction in the T-ion mobility is in principle possible by choosing a lower pKa counterion (e.g., Pyridine, pKa ∼ 5). However, this results in significant deviation from neutral pH, which is undesirable for bioassays. As shown by eq 1, lower T-ion concentrations contribute favorably to focusing rate. However, concentrations which are too low compromise buffering capacity and cannot be used robustly. Therefore, we chose the T-ion concentration to be 10 mM, which is a balance between the two. Tricine, Bistris, and PVP were obtained from Sigma-Aldrich (St. Louis, MO). HCl was obtained from Merck (Darmstadt, Germany). All buffer stock solutions were prepared in 20 mL glass bottles and kept at room temperature. The size of the bead-capturing trench (100 × 30 μm) was determined empirically, by testing several other designs. Smaller trenches resulted in nonuniform and inconsistent distribution of beads. Larger area trenches provided uniform distributions but required long filling time. The 100 × 30 μm trench was chosen as a balance which provided a uniform, repeatable surface in acceptable filling times (several minutes). Once the beads-based surface was established (see Bead Loading and Washing Process), we cleaned the TE reservoir with DI and filled it with 18 μL of TE solution and 2 μL of the target sequences (various concentrations). For ITP experiments, we applied a high voltage of 1000 V across the channel for 1 min to focus and bring the ITP interface to the reaction site area then switched to 200 V, allowing 2 min for the ITP interface to pass over the reaction site. At this voltage, the beads remained perfectly stationary as the ITP interface transverses the surface, which allowed simple image subtraction and analysis. Performing the assay at higher voltages often results in dislocation of the beads from their original position. For standard hybridization experiments (non-ITP), we applied a negative pressure of 60 cm H2O to the LE reservoir and allowed 3 min of hybridization.



RESULTS Figure 5 presents a sequence of raw fluorescence images of ITP-based surface hybridization using 100 nM of target. Here, for illustrative purposes only, we have also labeled the target sequence such that the ITP inteface is visible during its electormigraiton. As the ITP interface passes over the beads, 3034

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

rapid hybridization takes place causing enhanced fluorescence of the surface. At the applied voltage, the ITP interface transverses the surface in 4 s. However, as described in Theory and Principle of the Assay, despite the short time, the high local concentration results in significant hybridzation of the targets to the probes. Importantly, after the short hybridization time, the ITP interface continues to electromigrate downstream in a well-defined zone, carrying along remaining unhybridized targets or other contaminents. This serves an inherent “wash” step, which leaves the sensor area submerged in a clean buffer environment with excellent optical access and low background noise. This constitutes a one-step react-and-wash assay which can be performed in a simple channel and does not require flow control elements or moving parts. Figure 6 presents a quantitative comparison of ITP-based surface hybridization with standard flow-through hybridization,

For each experiment, regardless of the hybridization scheme, we subtract a background image (taken in LE solution, before hybridization begins) from the final data image. The range bars on the signal for all concentrations represent 95% confidence on the mean (with four realizations for each standard flow case, and three repeats for each ITP case). We note that the experiments were conducted across different microfluidic chips, multiple days, and multiple labeling batches of the beads. These, together with possible variations in the number of beads captured in each experiment, and potential residual dispersion due to EOF, are potential causes for the presented variations in signal. Across all concentrations, we observed significant improvement in signal when using ITP-based hybridization. At 10 nM, we observe a 107-fold improvement in signal due to ITP, consistent with model predictions. At the highest concentration, 100 nM, hybridization under ITP approaches saturation of the surface and thus, as predicted by the model, the gain in signal reduces to 12-fold. At a lower concentration of 1 nM, the signal from flow-based hybridization is very low, within the range bars of the control and thus does not allow for quantitative comparison against the ITP-based hybridization values. In Figure 6, we also present a comparison of the experimental data to our model, based on eq 19. A key parameter in the model, which sets the time scale τR, is the onrate (kon) of the reaction. To obtain this value, we performed a set of independent kinetic measurements where we flowed a fixed target concentration using pressure-driven flow and monitored the surface signal in time (see Figure 6b). We then used eq 19 to obtain the signal for both the standard flow (using τtot = 180 s) and ITP (using τtot = 2 s) cases. Results show good agreement with experimental values, though slightly overestimate the signal gains due in ITP (90 in theory, vs 107 in experiments). We evaluated the LoD of each of the methods by measuring the standard deviation of the signal in the control case (representing the noise) and seeking the lowest concentration for which the signal-to-noise ratio (SNR) is at least 3 standard deviations. The standard deviation of the control is 120 au, and thus the LoD was set at a signal value of 360 au. Using the model curves for interpolation, we determined the LoD of the standard flow and ITP-based techniques to be 15 nM and 150 pM, respectively. Consistent with the obtained gain in signal, this results indicates a 2 orders of magnitude improvement in SNR. We wish to highlight the importance of incorporating magnesium in such hybridization assays. Interestingly, in early experiments we performed without MgCl2, we obtained significant signals at target concentrations as low as 1 pM. Although this presumably constitutes a significant improvement in limit of detection, we believe it is nonspecific and the result of instability of the target-beacon bond. Optimization of magnesium concentration would be required for such assays, regardless of whether ITP enhancement is employed.

Figure 6. (a) Analytical and experimental results comparing ITP-based hybridization with standard flow through hybridization for initial concentrations between 1 and 100 nM. In both hybridization techniques the total assay time was 3 min. Each measurement corresponds to the difference in the average fluorescence intensity between a post- and prehybridization image, across a 80 × 20 pixels square at the center or the reactive surface shown in Figure 5. The height of each bar represents the average of at least three realizations, with range bars representing 95% confidence on the mean. Results show a 107 fold and 12-fold increase in signal at concentrations of 10 and 100 nM, respectively. At 1 nM, the improved signal in ITP is clearly evident, though direct estimation of the improvement in signal is difficult because the standard hybridization case did not result in significant signal above the baseline. Consistent with our theoretical predictions, using ITP hybridization, we obtain a 2 orders of magnitude improvement in LoD. (b) Kinetic experiment, measuring the signal as a function of time for the case of a standard pressuredriven flow reaction, using 100 nM target concentration. On the basis of this measurement, we estimate the on-rate of the reaction to be 6 ×103 M−1 s−1.



as well as validation of the analytical model. We performed the ITP assay by applying 1000 V across the channel for 1 min and then 200 V for 2 min, which results in the ITP interface arriving at the reactive surface after 3 min. For consistency in comparison of the results, we thus allowed exactly 3 min of flow in the standard hybridization case before imaging the surface. In the control cases, we followed the exact hybridization protocol, except with no sample in the reservoir.

CONCLUSIONS We presented a new method for improving the sensitivity of the surface sensors. We use ITP to focus target molecules and deliver a highly concentrated sample to the reaction site and thus accelerate reaction kinetics. Throughout the process, the sample is confined to the L-, and the T-ion interface and thus transverses the reacting surface as a finite “packet”. We have 3035

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036

Analytical Chemistry

Article

(5) Squires, T. M.; Messinger, R. J.; Manalis, S. R. Nat. Biotechnol. 2008, 26, 417−426. (6) Schudel, B. R.; Tanyeri, M.; Mukherjee, A.; Schroeder, C. M.; Kenis, P. J. A. Lab Chip 2011, 11, 1916. (7) Yoon, S. K.; Fichtl, G. W.; Kenis, P. J. A. Lab Chip 2006, 6, 1516. (8) Vijayendran, R. A.; Motsegood, K. M.; Beebe, D. J.; Leckband, D. E. Langmuir 2002, 19, 1824−1828. (9) Heller, M. J. Annu. Rev. Biomed. Eng. 2002, 4, 129−153. (10) Templin, M. F.; Stoll, D.; Schrenk, M.; Traub, P. C.; Vöhringer, C. F.; Joos, T. O. Drug Discovery Today 2002, 7, 815−822. (11) Sassolas, A.; Leca-Bouvier, B. D.; Blum, L. J. Chem. Rev. 2008, 108, 109−139. (12) Gao, Y.; Wolf, L. K.; Georgiadis, R. M. Nucleic Acids Res. 2006, 34, 3370 −3377. (13) Peterson, A. W.; Wolf, L. K.; Georgiadis, R. M. J. Am. Chem. Soc. 2002, 124, 14601−14607. (14) Miles, R. FY07 Engineering Research and Technology Report; Lawrence Livermore National Laboratory, 2008. (15) Jung, B.; Bharadwaj, R.; Santiago, J. G. Anal. Chem. 2006, 78, 2319−2327. (16) Bercovici, M.; Han, C. M.; Liao, J. C.; Santiago, J. G. Proc. Natl. Acad. Sci. U.S.A. 2012, 109, 11127−11132. (17) Persat, A.; Santiago, J. G. Anal. Chem. 2011, 83, 2310−2316. (18) Bercovici, M.; Kaigala, G. V.; Mach, K. E.; Han, C. M.; Liao, J. C.; Santiago, J. G. Anal. Chem. 2011, 83, 4110−4117. (19) Everaerts, F. M.; Beckers, J. L.; Verheggen, T. P. Isotachophoresis: Theory, instrumentation, and applications, 3rd ed.; Elsevier Scientific Publishing Company: Amsterdam, The Netherlands, 1976. (20) Khurana, T. K. Anal. Chem. 2008, 80, 6300−6307. (21) Bercovici, M. J. Chromatogr., A 2009, 1216, 1008−1018. (22) Hruscaronka, V.; Jaroscaron, M.; Gascaron, B. Electrophoresis 2006, 27, 984−991. (23) Hruška, V.; Riesová, M.; Gaš, B. Electrophoresis 2012, 33, 923− 930. (24) Bahga, S. S.; Bercovici, M.; Santiago, J. G. Electrophoresis 2010, 31, 910−919. (25) Garcia-Schwarz, G.; Bercovici, M.; Marshall, L. A.; Santiago, J. G. J. Fluid Mech. 2011, 1, 1−21. (26) Rubin, S.; Schwartz, O.; Bercovici, M. Phys. Fluids 2014, 26, 012001. (27) Bercovici, M.; Lele, S. K.; Santiago, J. G. J. Chromatogr., A 2010, 1217, 588−599. (28) Stellwagen, E.; Stellwagen, N. C. Electrophoresis 2002, 23, 1935− 1941. (29) Kim, D.; Herr, A. E. Biomicrofluidics 2013, 7, 041501−041501− 47. (30) Erickson, D.; Li, D. Anal. Chim. Acta 2004, 507, 11−26. (31) Duffy, D. C.; McDonald, J. C.; Schueller, O. J. A.; Whitesides, G. M. Anal. Chem. 1998, 70, 4974−4984. (32) Mohan, R.; Mach, K. E.; Bercovici, M.; Pan, Y.; Dhulipala, L.; Wong, P. K.; Liao, J. C. PLoS One 2011, 6, e26846. (33) Tan, W.; Fang, X.; Li, J.; Liu, X. Chem.Eur. J. 2000, 6, 1107− 1111. (34) Milanova, D.; Chambers, R. D.; Bahga, S. S.; Santiago, J. G. Electrophoresis 2012, 33, 3259−3262.

demonstrated that this results in an inherent wash step, in which any unreacted species or other contaminants continue electromigrating downstream. The analytical model we developed enables direct prediction of the gains obtained by ITP-based reactions, as a function of the basic physical parameters in the system, and provides insight on the design and optimization of such assays. Importantly, the model shows that the typical gain in surfacesignal is far lower than the fold increase in sample concentration. This can be attributed to two main reasons: (i) the surface senses the average analyte concentration and not the peak concentration; (ii) the total reaction time in the ITPbased assay is far shorter than in the standard flow-through experiment. Nevertheless, very significant gains of more than 100-fold in signal and LoD can be obtained, as we have demonstrated. We demonstrated magnetic bead-based surfaces to be a highly convenient method for developing surface-based reaction assays in the lab settings. The ability to reuse each chip, while avoiding surface chemistry, was highly valuable in obtaining high turnaround time in experiments. The design of a trench in the channel for capturing the beads is crucial if welldefined and confined surfaces are desired. The main limitation of the current implementation is that the total reaction time is limited to the time it takes the ITP interface to transverse the reactive surface. Our model shows that further improvement of order 100-fold in signal may be possible if the high concentration plug could be positioned over the surface for a longer time, for example by applying counterflow and holding the ITP interface stationary. While we are currently exploring this direction, we note it requires additional flow and control mechanisms which add to the complexity of the otherwise simple setup. We have focused our proof of principle experiments on DNA hybridization with molecular beacons, as this system is relatively easy to control and monitor. However, we believe that the principles we presented may be directly applicable to other ionic targets and in particular for detection of proteins.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Author Contributions §

These authors contributed equally.

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS We gratefully acknowledge funding from the Israel Science Foundation grant no. 512/12 and 1698/12. This research was also supported by the Henri Gutwirth Promotion of Research Fund. We thank Dr. Khaled Gommed for designing the photolithography masks, and Dr. Molly Mulligan for her helpful guidance in the fabrication process.



REFERENCES

(1) Wang, J. Anal. Chim. Acta 2002, 469, 63−71. (2) Nelson, B. P.; Grimsrud, T. E.; Liles, M. R.; Goodman, R. M.; Corn, R. M. Anal. Chem. 2001, 73, 1−7. (3) Stears, R. L.; Martinsky, T.; Schena, M. Nat. Med. 2003, 9, 140− 145. (4) Gadgil, C.; Yeckel, A.; Derby, J. J.; Hu, W.-S. J. Biotechnol. 2004, 114, 31−45. 3036

dx.doi.org/10.1021/ac403838j | Anal. Chem. 2014, 86, 3028−3036