Quantifying Biomolecular Binding Constants using Video Paper Analytical Devices

Abstract A novel ultra‐low‐cost biochemical analysis platform to quantify protein dissociation binding constants and kinetics using paper microfluidics is reported. This approach marries video imaging with one of humankind's oldest materials: paper, requiring no large, expensive laboratory equipment, complex microfluidics or external power. Temporal measurements of nanoparticle–antibody conjugates binding on paper is found to follow the Langmuir Adsorption Model. This is exploited to measure a series of antibody–antigen dissociation constants on paper, showing excellent agreement with a gold‐standard benchtop interferometer. The concept is demonstrated with a camera and low‐end smartphone, 500‐fold cheaper than the reference method, and can be multiplexed to measure ten reactions in parallel. These findings will help to widen access to quantitative analytical biochemistry, for diverse applications spanning disease diagnostics, drug discovery, and environmental analysis in resource‐limited settings.

Abstract: An ovel ultra-low-cost biochemical analysisp latform to quantify protein dissociation bindingc onstants and kinetics using paper microfluidics is reported. This approach marriesv ideo imaging with one of humankind's oldest materials:paper,requiring no large, expensive laboratory equipment, complex microfluidics or external power.T emporal measurements of nanoparticle-antibody conjugates binding on paper is found to follow the Langmuir Adsorption Model.T his is exploited to measure a series of antibody-antigen dissociation constantso n paper,s howingexcellent agreement with ag old-standard benchtop interferometer. The concept is demonstrated with ac ameraa nd low-end smartphone, 500-fold cheaper than the reference method, and can be multiplexedt o measure ten reactions in parallel. These findings will help to widen access to quantitativea nalytical biochemistry, for diverse applicationsspanning diseased iagnostics, drug discovery,a nd environmental analysisi nr esource-limited settings.
To day,anew generationo fl ow cost consumer electronicbased biosensors is emerging [1,2] with the potentialt od ramatically widen access to analyticalc hemistry capabilities in resource-limited settings. [3][4][5] This emerging fields eeks to harness:m ass manufactured sensors foundw ithin smartphones, such as cameras,t oe lectronically capture test results;p hone battery to powere xternal devices;p rocessing power to analyze results; screens to display results;a nd connectivity to transmitg eo-located results to central databases. Therei si n-creasingi nteresti nt he use of smartphonest od etect results from lateral flow tests. Lateralf low tests, also known as microfluidicp aper-baseda nalyticald evices (mPADs), including 2D [6] and 3D structures, [7][8][9] and paper origami, [10][11][12][13] are opening up new capabilities for multiplexed analysis with small sample volumes and on-test sample handling. The merits of mPADs are their compatibilityw ith ab road range of chemical and biological molecules,l ow non-specific interactions, low manufacturing cost (as little as $0.001 [7] ), portability,l ow sample volumes, safe disposala nd power-free fluid pumping, exploiting the natural capillarity of paper. [14] To date, the use of smartphone cameras to interpretl ateral flow tests has focusedo ni ndividuals till image end-point readings to interpret paper-based tests for diagnostics, [15][16][17][18][19][20][21][22][23] chemical sensing, [24,25] and drug monitoring. [26,27] Cameras and smartphones have also been used with other microfluidic techniquest oq uantify biologicalr eactions, such as the detection of nucleic acid sequences [28] and E. coli detection using quantum dots. [29] Video tracking of biological interactions hasa lso been used forr eal-time recording polymerase chain reaction amplification using ad igital single lens reflex camera, [30] glucose sensing with am obile phone, [25] and the use of ac omplementarym etal-oxide semiconductor image sensor to track the motiono fs perm cells. [31] Te mporal surface plasmon resonance protein detection has also been demonstrated with as martphone [32] using ap olydimethylsiloxane microfluidic device.
Here, for the first time in the literature, we progress beyond still images of lateral flow tests to video analysis in order to investigate whether dynamic ligand-receptor binding on mPADS followst he Langmuir Adsorption IsothermM odel, [33] and whether mPADs could quantifyf undamentalb iomolecular parameters, namely,t he thermodynamic equilibrium dissociation constant, K D and kinetic k on and k off rates. We overcome potential barriers associated with quantitative analysiso nl ateral flow tests cited in previousw ork, including sample volume limitations, [34] color inhomogeneity, [2] reproducibility issues, [35] such as surfacef low and inconsistentm embranes, [2] porous 3D surface, protein dissociation over long periods, and possible reaction-limiting local sample depletion due to flow rate.
The ability to measures uch fundamental chemical binding constantsa nd kinetic reaction rates lies at the heart of chemistry and traditionally relies on access to sophisticated laboratory instrumentation, such as surface plasmon resonance [36] and interferometry,u sed here as ag old-standard referencem ethod, typicallyu sing instruments costing in excesso f£100 000. Other label-free methods that also require specific instrumentation are dynamic light scattering [37] and isothermal titration calorimetry. [38] There are av ariety of fluorescence-labeling techniques such as fluorescencep olarization, [39] fluorescencec orrelation spectroscopy, [40] total internal reflection fluorescencem icroscopy, [41] and Fçrster resonance energy transfer. [42] These methods all requiref luorescencer eadout, such as af luorescence microscope or spectrometer.I nc ontrast, the method presented here is ultra-low cost, simply requiring ad igital camera/smartphone, giving equipment costs of just £500/£214 respectively,a nd per assay paper microfluidic strip and consumables costs of approximately £1.20 (Supporting Information (SI), Ta ble S1).
Our low-cost techniqueu ses as imple set-up consisting of a consumer camera or smartphone, as eries of direct-detection mPADs, and a9 6-well plate. Lines of antigen are immobilized on nitrocellulose paper strips. When the antibody-functionalized gold nanoparticles (Ab-AuNP) flow along the membrane, they bind to the test line (Scheme 1a), generating ar ed-color, the intensity of which is proportional to the number of gold nanoparticles (SI, Figure S1), and therefore the number of bound antibody-antigens. The nitrocellulose strips are mounted together,w ith al arge absorbent pad to prevent saturation, and dipped into a9 6-well plate, where each wellc ontains a differentc oncentration of Ab-AuNP solution (Scheme 1b). The camera videos the mPADe xperiment running, as shown in Scheme1c. An excesss olution volume is used in order to mimic an infinite solution (see SI, Figures S2-S5 for flow rate analysis). Scheme1ds hows as eries of video frames to show the temporal development of a mPADt est line. Video analysis (Wolfram Mathematica) is used to extract changes in colorimetric intensity, I. The pixel values are extracted and averaged acrosst he width of the mPADt or educe noise, creating al ine profilea long the strip. The peak heighti st hen outputted as a functiono ft ime for each mPAD. Here, the green channel of the RGB color-space is used to match the absorptionp eak of 20 nm AuNPs, but this can be tailored to the type of nanoparticles employed. Due to the timescales considered here, as ampling rate of 1Hzi su sed;h owever,t his could be increased up to 30 Hz to measuref aster biochemical reactions. Figure 1a shows an example set of video-mPADt ime-intensity plots to track the binding of am onoclonala ntibody to the influenzah emagglutinin H5 antigen test line. As eries of eight different Ab-AuNP concentrations are measured (0.9pMt o Scheme1.Te mporal consumer-electronic camera video analysisofmPADs. where I is intensity (test line peak height), I 1 is the equilibrium intensity value, k obs is the fitted observed binding rate, and t is time. As t !1, I ! I 1 .T he I 1 values are determined for each concentration of analyte, and fitted to aL angmuir model shown below as Equation (2) [33] (Figure 1b): where a is the saturation intensity when all availableb inding sites are occupied, and C is Ab-AuNP concentration.
To determine the k on and k off rates, the values of k obs are extracted from the fits of Equation (1) and plotteda gainst C. They are then fitted to followingt he relationship in order to extract k on and k off ,s hown below as Equation (3): An example of this is shown in Figure 1b (inset).
Our results show stronga greement between the binding kinetics measured by video analysis and the Langmuir model.
We then apply video analysist of ive different antibody-antigen pairs (SI , Table S2). In parallel, benchmarking studies are performed with the same proteins using aF ortØBio Octet RED96 benchtop interferometer.T he K D fits shown in Figure 2a demonstrate strong agreement between video-mPADs (solid lines) and interferometry (dotted lines). This is further demonstrated in Figure 2b,w here the K D values measured by video-mPADs and interferometry show al inear relationship with a gradiento f1 .1 (standard error 0.032), and an adjusted R 2 value of 0.996. For all raw data and fits, see SI, Figures S6, S7,a nd Ta ble S3.T he estimated concentration of antibodies in video-mPADa ssays assumes that all antibodies bind to AuNPs in an active, availablec onformation,e quating to % 15 antibodies per AuNP.W en ote that the measured k on and k off values differ from those measured by interferometry, although the relationship K D = k off /k on still holds (SI, Figures S8 and S9). This interesting result highlights the value of quantifying antibody-antigen reactionk inetics on mPADs in order to optimize diagnostic test performance to achieve fast kinetics and as trong K D ,g iving a sensitive, rapid test with low-sample volume.
Ap roof-of-principleo fv ideo-mPADs using as martphone is shown with an LG Nexus 5 ( Figure 3a). The performance is comparedt ot he Canon PowershotG 15 camera for am odel anti-human IgGF c-human IgG interaction (Figure 3b and SI, Figure S10). No significant difference is found between the resulting K D values (two-tailed t-test gives p-value of 0.22, t-value of 1.2, degrees of freedom = 52), confirming the LG Nexus 5 smartphone can be used for K D measurements.
Buildingo no ur work with single antibody-antigen pairs, we then sought to investigate whether video-mPADs could mea-sure multiple antibody-antigen interactions simultaneously. This could be useful for antibody and drug screening to quantify multiple antibodies' binding affinitiest oasinglet arget. Therefore, in contrast to the above,t he AuNPs are functionalized with the antigen,a nd the antibody is spotted on the mPADs. The proof-of-concept is shown in Figure 4. An array of ten antibody spots is deposited on each mPAD. Figure 4a shows af ilmstripo fam ultiplex video-mPADd eveloping over time, with ah eatm ap of pixel values shown below.T his is translated into the time-intensity graph in Figure 4b showing nine multiplex video-mPADs-eightd ifferenta ntigen-AuNP concentrations and ab uffer control. Each mPAD'st en spots are plottedo verlaid for each concentration. The low variances between spots illustrates that the kinetics are independento f spot position. Here, identicala ntibody-antigen combinations are used as ap roof-of-concept. In future, each spot could be a differentc apture ligand.W es how that the K D measured from singleplex and multiplex video-mPADs are not significantly different( two-tailed t-testg ives p-value = 0.25, t-value = 1.1, degrees of freedom = 96), validating this reversed orientation (see SI, Figure S11). Herein we harnessc onsumer electronicv ideo imaging for low-cost mPADs, creating an accurate platform for measurement of antibody-antigen dissociation constants. Our approach does not requiree xpensive, complexl aboratory-based equipment, simply using ac amera or smartphone. The measured video-mPAD K D values for five antibody-antigen pairs show excellenta greement with ar eference benchtop interferometer (Figure 2b,a djusted R 2 = 0.996). The assumptions and justificationsf or using the Langmuir model are discussed in SI Discussion and Ta ble S4. We demonstrate that as martphone can measure K D values, and create am ultiplex platform to detect multiple ligand-receptor interactions in parallel. Al-thoughA uNP labelingi su sed here, future assays could employ al abel-free competitive inhibition format.
The low-cost of video-mPADs is am ajor advantage over expensive benchtop instrumentation. SI Ta ble S1 lists the cost, size, and weight of video-mPADi nstrumentation compared to the FortØBio Octet RED96, showingt hat video-mPADe quipment costs are around2 50-580times cheaper,m aking it much more accessible to academic laboratories and resource-limited settings.L ow-ends martphones are amenable to this application, since neither high camera resolutions nor large processing power are needed. Moreover,v ideo-mPADs require6 4-fold lower amountso fc apture reagents (SI Table S1), advantageous for early stage discovery projects requiring biophysical charac-terization of reagents, where the quantities of material may be limited.
In our study,video analysis is performed off-device;however, this could be performed in real-time, even on alow-cost smartphone,w ith ac apture rate of 1Hz. In future, automatics trip detection by traditional image processing or machine learning would make the methodm ore user-friendly.
To close, this technique allows low-cost, quantitative, multiplexed analysis and is generalizable to aw ide range of biological and chemical ligand-receptor interactions,w ith many potential applicationsi na nalytical chemistry,b iomedicine,f orensics, and environmental analysis.