Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to...

9
P.M. Kosaka, V. Pini, J.J. Ruz, R. A. da Silva, M. Ujue-González, D. Ramos, M. Calleja and J. Tamayo This file includes: S1. Materials and Methods of the Sandwich Assay S2. Materials and Methods of the Measurements S3. Analysis of the density and nanoparticle distribution on the surfaces S4. Statistical study of the reliability of the dual sensor S5. Results with the prostate specific antigen (PSA). Supplementary Figures 1, 2, 3 and 4. Supplementary references. Detection of cancer biomarkers in serum using a hybrid mechanical and optoplasmonic nanosensor SUPPLEMENTARY INFORMATION DOI: 10.1038/NNANO.2014.250 NATURE NANOTECHNOLOGY | www.nature.com/naturenanotechnology 1 © 2014 Macmillan Publishers Limited. All rights reserved.

Transcript of Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to...

Page 1: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S1

SUPPLEMENTARY INFORMATION

Detection of cancer biomarkers in serum using a hybrid mechanical and optoplasmonic nanosensor

P.M. Kosaka, V. Pini, J.J. Ruz, R. A. da Silva, M. Ujue-González, D. Ramos, M. Calleja and J. Tamayo

This file includes:

S1. Materials and Methods of the Sandwich Assay

S2. Materials and Methods of the Measurements

S3. Analysis of the density and nanoparticle distribution on the surfaces

S4. Statistical study of the reliability of the dual sensor

S5. Results with the prostate specific antigen (PSA).

Supplementary Figures 1, 2, 3 and 4.

Supplementary references.

Detection of cancer biomarkers in serum using a hybrid mechanical and optoplasmonic nanosensor

SUPPLEMENTARY INFORMATIONDOI: 10.1038/NNANO.2014.250

NATURE NANOTECHNOLOGY | www.nature.com/naturenanotechnology 1

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 2: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S2

S1. Materials and Methods of the Sandwich Assay

S1.1. Substrates, proteins and reagents

Sulphuric acid (ACS reagent, 95-98%), hydrogen peroxide (H2O2 30%), (3-glycidyloxypropyl)trimethoxysilane (98%), Nα,Nα-Bis(carboxymethyl)-L-lysine hydrate (97%, TLC), dry toluene (99.8%), N-hydroxysulfosuccinimide sodium salt (sulfo-NHS), N-(3-dimethylamino propyl)-N’-ethylcarbodiimide hydrochloride (EDC), (Aminoethyl)polyethylene glycol (5,000 Da), 2-(N-Morpholino)ethanesulfonic acid (MES), Anti-Peroxidase antibody produced in rabbit (anti-HRP), bicarbonate buffer solution, phosphate buffered saline (PBS), sodium chloride, sodium phosphate dibasic, sodium phosphate monobasic and Tween® 20 were purchased from Sigma-Aldrich ( St. Louis, USA). Dialysis Kit with 8 kDa molecular weight cut-off was purchased from GE Healthcare (Little Chalfont, United Kingdom). Coomasie Plus (Bradford) Assay Kit was purchased from Thermo Scientific (Waltham, USA). Fetal bovine serum (FBS), Brazilian origin, sterile filtered for cell culture, was purchased from Lonza (Verviers, Belgium). Spherical 100-nm-diameter gold nanoparticles coated with 5 nm thick carboxyl polymer (C11-100-TC-50) were acquired from NanopartzTM (Loveland, USA). The gold nanoparticles were supplied with a given concentration of 6.34 x 109 nanoparticles/mL in 18 MEG DI water. The silicon cantilever arrays were purchased from Concentris GmbH (Basel, Switzerland). Monoclonal mouse antibodies against carcinoembryonic antigen (3C6 and 3C1), carcinoembryonic antigen (CEA) from single patient source colon carcinoma liver metastatic tissue, monoclonal mouse antibodies against prostate specific antigen (1H12 and 5A6) and prostate-specific antigen (PSA) from human seminal fluid were purchased from HyTest (Turku, Finland). In the sandwich immunoassay used for the detection of CEA, monoclonal antibodies 3C1 and 3C6 are the detection and capture antibodies, respectively. For PSA detection, the detection and capture antibodies are the monoclonal antibodies 5A6 and 1H12, respectively. S1.2. Antibody conjugation to carboxyl-polymer coated spherical gold nanoparticles The monoclonal mouse detection antibody, anti-carcinoembryonic antigen 3C1 (MAb3C1) or anti-prostate specific antigen 5A6 (MAb5A6), was immobilized onto the surface of the carboxyl-polymer coated gold nanoparticles following the procedure provided by NanopartzTM (http://www.nanopartz.com/). First, 100 µL of 2.5 mg/mL Au nanoparticles in PBS were mixed with 100 µL of 1 mg/mL of the detection antibody in PBS. Then, 1 mL of Milli-Q water was added and the solution was vortexed for 1 minute at 25 °C (Solution A). Immediately 100 µL of 1 mg/mL of EDC in Milli-Q water was added to the Solution A. The mixture was incubated at 25 °C for 1 hour and then span in a microcentrifuge until a nice pellet was formed. The supernatant was removed to less than 50 µL and the remaining solution was refilled to 1.5 mL with PBS buffer. This step was repeated twice. After, the nanoparticle was spinned again and the pellet formed was mixed with 1.5 mL of 1 mg/mL BSA in PBS and incubated for 1 hour at 25 °C to block the uncoated voids on the gold nanoparticle and avoid non-specific interaction. The solution was then spun in a microcentrifuge until a nice pellet was formed. The supernatant was removed to less than 50 µL and the remaining solution was refilled to 1.5 mL with PBS buffer. This step was repeated twice. The sample was stored in refrigerator at 4 ⁰C

until its use.

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 3: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S3

S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution (3 H2SO4 : 1 H2O2) to remove all the organic residues on the surface (caution: piranha solution is

extremely corrosive, reactive and potentially explosive) for 15 minutes at room temperature (RT). The cantilevers were rinsed three times with Milli-Q water and dried under a stream of nitrogen. The cantilevers were dipped into a 0.2% solution of (3-glycidyloxypropyl)trimethoxysilane in dry toluene overnight at room temperature. The samples were then washed with toluene, Milli-Q water and dried under N2. A 100 mM NTA solution in 50 mM carbonate buffer (pH 9.5) was prepared and the cantilevers were incubated overnight at 25 °C under gentle agitation. The microcantilever functionalization with NTA is a two-step method1 that comprises (i) grafting the epoxysilane 3-(2,3-epoxypropoxy) propyltrimehoxysilane to the clean microcantilever surface and (ii) linkage of the NTA through the epoxide groups (the epoxide groups on the surface reacts with the amine groups of the NTA at basic pH). The cantilevers were then rinsed with 50 mM carbonate buffer pH 9.5, Milli-Q water and dried under N2. The carboxyl groups at the cantilever surface were activated by immersion in a mixed solution of 100 mM EDC and 150 mM sulfo-NHS both dissolved in 10 mM MES pH 5.5. The cantilevers were incubated for 45 minutes at 37 °C under gentle agitation. Immediately, the samples were extensively rinsed with 10 mM MES. S1.4. Immobilization of the capture and control antibodies on the cantilever Prior to the immobilization of antibodies onto cantilevers, 1 mL of a 4 mg/mL anti-HRP solution in Milli-Q water was dialyzed overnight at 4°C. The concentration of the anti-HRP solution after the dialysis was determined using Bradford assay2. A calibration curve was made using serum bovine albumin (BSA) as protein standard. The range of linearity of the assay was from 5 μg/mL to 2500 μg/mL. The anti-HRP was immobilized onto the surfaces of the cantilevers used as control samples following the same protocol applied for the capture antibodies and described as follows. Right after the surface activation step, the antibodies were immobilized only on the top side of the cantilevers using the drop method described elsewhere3. The functionalization procedure involves covalent and oriented immobilization of the antibody with a recognition capability higher than 90%3. A solution of 50 μg/mL of the control or capture antibodies (monoclonal mouse anti-carcinoembryonic antigen 3C6 (MAb3C6) or monoclonal mouse anti-prostate specific antigen 1H12 (MAb1H12)), was prepared in 10 mM MES (pH 5.5). The cantilevers were incubated for 2 hours at 37 °C. After that, the samples were washed with 10 mM MES (pH 5.5) and incubated for 45 minutes at 37 °C with 10 mM sodium phosphate buffer (pH 8.0) with 0.3 M NaCl to desorb antibodies that are not covalently attached to the surface. The cantilever surface was subsequently blocked to prevent nonspecific adsorptions; the cantilevers were dipped into a 1 mg/mL (aminoethyl)polyethylene glycol (PEG) solution in 10 mM MES with 0.05% Tween® 20 (pH5.5) overnight at 4°C. Thereafter, the samples were washed with MES pH 5.5 with 0.05% Tween® 20 (pH5.5).

S1.5. Biomarker recognition and Sandwich assay The cantilevers were incubated for 1 hour at 37°C in 1 mL solutions of CEA or PSA with concentrations ranging from 1 pg/mL to 10 ag/mL in PBS solution with 0.05% Tween® 20 pH 7.4 (PBST) and FBS to simulate a real sample. Stringent control experiments were performed with the control cantilevers functionalized with the anti-peroxidase antibody and very high

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 4: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S4

CEA or PSA concentrations of 1 μg/mL. Right after, the cantilevers were washed twice with PBST and once with PBS pH 7.4. After that, the samples were rinsed with Milli-Q water and dried under a stream of N2. For the sandwich assay, the cantilevers were dipped into 1 mL of a 1 μg/mL solution of spherical gold nanoparticles functionalized with the detection antibody

(MAb3C1 or MAb5A6) prepared in MES pH 5.5 with 0.05% Tween® 20 (MEST). The samples were incubated at 37°C for 1 hour under gentle agitation, washed three times with MEST, twice with MES, extensively rinsed with Milli-Q water and dried under a stream of N2.

S2. Materials and Methods of the Measurements

S2.1. Microcantilever specifications Monocrystalline silicon microcantilever arrays were used in this work (Concentris GmbH). The nominal length, width and thickness of the cantilever are 500, 100 and 1 µm, respectively. The edges are oriented along the <110> direction, and that the Young’s modulus relevant for the

calculation of the flexural eigenfrequencies is E=169 GPa. The density of silicon is 2330 Kg/m3. The fundamental resonance frequency and quality factor are 5.080.18 kHz and 17.01.8, respectively (statistics from 400 cantilevers). S2.2. Resonance frequency measurements The cantilever resonance frequency was measured after the cancer biomarker recognition on the cantilever surface and after the sandwich assay with the gold nanoparticles. The measurements were performed in air using the optical beam deflection technique described elsewhere4. The readout technique combines the optical beam deflection method and the automated 2D scanning of a single laser beam by voice-coil actuators. A 3 mW red laser diode (Schafter-Kirchhoff, 635 nm) is mounted on two perpendicular linear voice coil actuators (Physik Instrumente GmbH& Co.). The laser spot size is 1 µm. The measurements were carried out in less than 2 minutes, in order to keep the biological activity of the proteins on the cantilever surface. S2.3. Bright/Dark field microscopy The bright field images of the cantilever at different wavelengths were performed with a Nikon microscope (Nikon, ECLIPSE LV150A) equipped with an ultrahigh-definition cooled color camera (Nikon, DS-RI1). Several light filters placed along the illumination arm of the microscope were used to tune the illumination wavelength. Dark field imaging was carried out with a Zeiss microscope (Axioskop 2 MAT) equipped with a high-definition color camera (Zeiss, AxioCam MRc 5). Measurements were performed in dark-field reflection configuration with an Epiplan Neofluar® 50X objective (Zeiss, N.A. 0.8). The size of the images were 220x297 µm2. The red channel intensity of the darkfield images was used to quantify the scattering intensity in a typical 8-bit scale (from 0 to 255). The raw data were processed with a custom routine developed with Matlab® that calculates the mean red component of each dark-field image normalized to the mean red component of a clean silicon surface. The plotted intensity values in Fig. 5a are the mean value of the intensity in whole surface of the cantilever (500x100µm2), and the mean value in a region of the

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 5: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S5

preclamping immediate to the cantilever with the same area than that of the cantilever. None data was discarded. S2.4. Light scattering spectra Darkfield spectra were acquired by means of an optical spectrometer (Andor, Shamrock SR-303i) coupled by an optical fiber (Ocean Optics) to an optical microscope (Olympus, BX51) in dark-field reflection configuration. The probe light emitted by the internal microscope light bulb, is focused onto the sample and the scattered light is collected by a dark-field 20X objective (Olympus MPlan N, N.A. 0.4). The optical signal analyzed by the optical spectrometer comes from a sample region of 40 μm in diameter. At least three different regions were measured in both the preclamping and cantilever for each analyzed cantilever. Each dark-field spectrum was acquired in the spectral interval from 450 nm to 900 nm with an acquisition time of 5 s. CCD dark current was removed from all raw darkfield sample spectra by subtracting the spectrum obtained when the microscope illumination is blocked by an obturator (acquisition time 5 s). Each dark-current corrected spectra were finally normalized to the dark-field spectrum of an unfunctionalized silicon surface. S2.5. Scanning electron microscopy (SEM) The surface density of gold nanoparticles on the cantilever and preclamping surfaces were accurately measured by SEM (Hitachi S-800 FESEM). About 100 images, as shown in Suppl. Fig. 1, were acquired at different regions of each cantilever and the preclamping for each concentration. A signal contrast-based algorithm implemented in Matlab software was used to evaluate the gold nanoparticle surface density and analyze the distribution of the nanoparticles into clusters.

Supplementary figure 1. SEM images of a region of the cantilever and preclamping after the sandwich

assay in a control experiment and in a detection assay of 1 fg/mL of CEA in serum.

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 6: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S6

S3. Analysis of the density and nanoparticle distribution on the surfaces Supplementary figure 2(a) shows that the surface density of nanoparticles on the supporting chip (that includes the preclamping region) and the cantilever obtained from the analysis of the SEM images (Section S2.5). We notice that no differences in the particle density were found between the preclamping and chip regions. The data shows no differences in the nanoparticle distribution between the cantilever and the chip, which indicates that the higher scattering signal on the cantilever arises from the fact that the cantilever acts as an optical cavity that dramatically increases the plasmonic signal of the gold nanoparticles. Another feature obtained from the data shown in Suppl. Fig. 2(a) is that the number of nanoparticles quickly saturates for concentrations higher than 1 fg/mL. Supplementary figure 2(b) shows the nanoparticle relative amount of nanoparticles distributed as monomers and clusters (dimers, trimers etc). For concentrations lower than 1 fg/mL, most of the nanoparticles are bound to the surface as monomers (80%). However for higher concentrations, an abrupt transition in the nanoparticle distribution emerges. The nanoparticles forming clusters are about 60% of the total nanoparticles on the surface. The nanoparticle distribution follows a similar pattern in the cantilever, preclamping and chip. No differences are observed between the nanoparticle distributions obtained in PBS and serum.

Supplementary figure 2. (a) Nanoparticle density on cantilevers and chip in serum. (b) Relative

distribution of nanoparticles as monomers and clusters.

S4. Statistical study of the reliability of the dual sensor

The sensitivity and specificity of a diagnostic test are function of a chosen threshold value used to discriminate positive from false detections. Changing the threshold value, so as to increase the sensitivity, will decrease the specificity, and vice versa. The receiver operating

characteristic (ROC) curve is a plot of the true positive rate (or sensitivity) on the y-axis and the false negative rate (1-specificity) on the x-axis resulting from continuously varying the decision threshold over the entire range of results observed. The true positive rate (TPR) is the probability that a disease case will be correctly classified and true negative rate (FPR) is the probability that a healthy case will be incorrectly classified. The ROC curve can also be used to compare the performance of two or more diagnostic tests7,8. An alternative to the ROC curve is the detection error tradeoff (DET) graph, which plots the false negative rate (missed detections) vs. the false positive rate (false alarms) on log x- and y-axes. This alternative spends more graph area on the region of interest, i.e., the region with minimal false rate.

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 7: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S7

We calculate a DET curve for each CEA concentration for both nanomechanical and optoplasmonic transduction. In this calculation, we assume a normal distribution of the data for each concentration determined by the mean value and standard deviation. Supplementary figure 3 shows the DET curves for a concentration of 10 fg/mL for the plasmonic and nanomechanical transduction methods. The dashed-dotted grey line corresponds to a random

guess. Both transduction methods provide DET curves well below this non-discrimination curve. The optimal threshold signal value is given by the maximum length of the interception of the DET curve with the line at 90 deg to the non-discrimination curve. We refer this length to as interception length. Now consider the case in which our signal is a linear combination of the scattering intensity and the resonance frequency shift5. The linear combination is optimized by maximizing the interception length. In this way, the false negative and false positive rate of our detection is always minimized as shown in Suppl. Fig. 3 (black dashed line). We emphasize that for determining the optimal linear combination of the resonance frequency shift and light scattering signals, it is required a previous calibration of the sensor response to different biomarker concentrations with a statistically significant sample size as performed here. For the sake of simplicity, we plot in Fig. 5b of the main text the mean value of the false positive and negative rates at the optimal signal threshold, referred to as error rate.

Supplementary figure 3. DET curves for a concentration of 10 fg/mL using the nanomechanical and

plasmonic signals and an optimal linear combination of them.

S5. Results with the prostate specific antigen (PSA) Following the same procedure for detecting CEA in serum, we carried out sandwiches assays on silicon microcantilevers for detecting PSA concentrations from 10 ag/mL to 1 pg/mL. For the control assay, the capture antibody against PSA was replaced by an antibody (anti-peroxidase antibody) non specific to PSA, and the used sample was a highly concentrated PSA solution (1µg/mL). The PSA detection antibody tethered to the gold nanoparticles was kept in the control assays. Supplementary figure 4 shows the nanomechanical and optoplasmonics signals obtained for the PSA as a function of the concentration. The noise floor level due to non specific interactions obtained from the control assays is also plotted. The detection limit in the calibration curves for the nanomechanical and optoplasmonic transductions approximately are 1 fg/mL and 100 ag/mL, respectively. The detection limits are similar to those

1E-7 1E-6 1E-5 1E-4 1E-3 0.01 0.1 11E-7

1E-6

1E-5

1E-4

1E-3

0.01

0.1

1

Fa

lse

Ne

ga

tive

Ra

te

False Positive Rate

Nanomechanical

Plasmonics

Combined

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 8: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S8

obtained for the CEA. Again, the hybrid plasmonic cavity plays a determining role for detecting ultralow concentrations of this cancer biomarker6. Whereas the scattering signal in the chip lies in the region obtained in the control experiments for PSA concentrations below 1 pg/mL, the scattering enhancement due the optical cantilever cavity enables the detection of concentrations on the verge of 10 ag/mL.

Supplementary figure 4. a. Relative resonance frequency shift of the fundamental vibration mode of the

cantilever as function of the PSA concentration in serum (red symbols). The frequency shift for the control

experiments is plotted as a shaded region that represents the mean plus the standard deviation of the

data. b. Mean scattering signal in the cantilever and the preclamping rgion versus the PSA concentration

in serum. The signal is obtained from quick inspection of the cantilevers with a simple optical microscope

and darkfield objective with low magnification. The data from the cantilever are compared with the data

from the preclamping to assess the effect of the optical cantilever cavity. The scattering for the control

experiments in the cantilever and preclamping regions are plotted as a dashed region that represents the

mean value plus the standard deviation of the data.

© 2014 Macmillan Publishers Limited. All rights reserved.

Page 9: Detection of cancer biomarkers in serum using a …S3 S1.3. Cantilever functionalization Prior to their functionalization, the cantilever arrays were cleaned with piranha solution

S9

Supplementary references

1 Chevalier, S. et al. Creating biomimetic surfaces through covalent and oriented binding of proteins. Langmuir 26, 14707-14715 (2010).

2 Bradford, M. M. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Analytical biochemistry 72, 248-254 (1976).

3 Kosaka, P. M. et al. Tackling reproducibility in microcantilever biosensors: a statistical approach for sensitive and specific end-point detection of immunoreactions. Analyst 138, 863-872 (2013).

4 Martínez, N. et al. High throughput optical readout of dense arrays of nanomechanical systems for sensing applications. Review of Scientific Instruments 81, 125109-125109-125109 (2010).

5 Perkins, N. J., Schisterman, E. F. & Vexler, A. ROC curve inference for best linear combination of two biomarkers subject to limits of detection. Biometrical Journal 53, 464-476 (2011).

6 Schmidt, M. A., Lei, D. Y., Wondraczek, L., Nazabal, V. & Maier, S. A. Hybrid nanoparticle–microcavity-based plasmonic nanosensors with improved detection resolution and extended remote-sensing ability. Nature communications 3, 1108 (2012).

© 2014 Macmillan Publishers Limited. All rights reserved.