Label-Free Method Using a Weighted-Phase ... - ACS Publications

Subscriber access provided by UNIV OF CALIFORNIA SAN DIEGO LIBRARIES

Article

A label-free method using a weighted-phase algorithm to quantitate nanoscale interactions between molecules on DNA microarrays Qi Li, Rongxin Fu, Junqi Zhang, Ruliang Wang, Jiancheng Ye, Ning Xue, Xue Lin, Ya Su, Wupeng Gan, Ying Lu, and Guoliang Huang Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.6b04596 • Publication Date (Web): 23 Feb 2017 Downloaded from http://pubs.acs.org on February 23, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

A label-free method using a weighted-phase algorithm to quantitate nanoscale interactions between molecules on DNA microarrays Qi Li,† Rongxin Fu,† Junqi Zhang,† Ruliang Wang,† Jiancheng Ye,† Ning Xue,† Xue Lin,† Ya Su,† Wupeng Gan, ‡ Ying Lu,† and Guoliang Huang*, †, ‡ †

Department of Biomedical Engineering, Tsinghua University School of Medicine, Beijing 100084, PR China

‡

National Engineering Research Center for Beijing Biochip Technology, Beijing 102206, PR China

ABSTRACT: White light interference is used as a label-free method to detect nanoscale changes on surfaces. However, the signalto-noise ratio of the white light interference method is very low, thus resulting in inaccurate results. In this paper, we report a corrected label-free method based on hyperspectral interferometry to overcome the shortcoming of the white light interference method. A platform based on hyperspectral interferometry was established, and a DNA hybridization microarray was constructed to quantitate thickness variation of molecules on a solid surface. We used fluorescence resonance energy transfer (FRET) to validate the results of our method. Compared to conventional fluorescence-labeled method like FRET, our method has advantages because it does not require a fluorescent label, has a detection limit of 1.78 nm, a high accuracy and wide detection range (5-64 bp).

With life science research increasingly trending from the macro to micro perspective, researchers are eager to understand the mechanism of biochemical processes at the molecular level and locations or interactions of different parts of cells. In this context, molecular detection technology has emerged and been rapidly developed. The measurement of interactions between molecules is an important field in the research of molecular detection technology. In recent years, these interactions have been used to study the structure and conformation of proteins and nucleic acids,1 as well as interactions between single molecules.2-4 Common methods for studying interactions between molecules include fluorescence resonance energy transfer (FRET), surface plasmon resonance (SPR), single molecule fluorescence microscopy, and atomic force microscope (AFM). The development of microarray chips over the past 20 years has enabled the high-throughput detection and analysis of many biological specimens. Microarray chips can be divided into nucleic acid, protein, and tissue chips according to the different types of biological probes used. Nucleic acid chips are the basis of microarray chips, whose principle is base pair hybridization between a capture probe and a target probe. Nucleic acid chips can be applied to genome structures in cells or at the tissue level for gene function and gene expression analysis. To facilitate these analyses, it is significant to understand the mechanism of the biochemical reaction to optimize the reaction system and improve the efficiency of hybridization to study the conformation of nucleic acids on solid surfaces and interactions between capture probes and target probes. For oligonucleotide sequences < 50 bp on microarrays, the length of the probe is within a few nanometers to tens of nanometers. However, due to the optical diffraction limit, the resolution of commonly used fluorescence imaging techniques is ~200 nm, which is far behind these requirements for detection. There-

fore, a novel nanoscale detection technique is needed to study the interactions between molecules on solid surfaces. FRET is a process that transfers energy from a donor fluorophore to an acceptor in the excited state, which can be used as an optical ruler to measure distances between 1 and 10 nm. The FRET method can also be applied to a wide range of different optical configurations and has been broadly applied for the detection of molecular interactions.5-11 Zhu et al.12 collected three-channel fluorescence signals, eliminated the crosstalk of the fluorescence signal using a correction algorithm, and measured the efficiency of DNA hybridization. Using this method, the signal is collected with a high signal-to-noise ratio, the FRET efficiency on a solid surface can be calculated accurately, and the distance between the labeled fluorescent probes can be acquired. However, this method requires fluorescent markers, which have many limitations such as biological toxicity and bleaching. Further, fluorescently labeled molecular groups may also unintentionally affect the structure of nucleic acids, and the multi-channel fluorescence signal acquisition system is complicated. Indeed, many system parameters must be separately measured in every experiment. Therefore, the method is tedious, time consuming, and impractical. To overcome the problems of fluorescent labeling methods, label-free detection methods are being developed. Label-free detection is a method in general that does not require the test objects to be marked. Common label-free detection methods include optical interference, SPR, and the photo-acoustic method13. Compared to fluorescence-labeled detection, labelfree detection is a non-invasion method, and it does not change the physical and chemical properties of the measured object. On the nanometer scale, label-free methods have many applications. Moiseev et al.14 used white light interference to measure the DNA conformation of a silicon surface, and compare their results to fluorescence spectral self-interference

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

fluorescence microscopy (SSFM) results and theoretical calculations. However, the DNA molecular thickness (≤ 10 nm) is relatively smaller than the interference layer thickness (500 nm), so the signal-to-noise ratio of white light interference is very low. Indeed the results of DNA layer thickness differed greatly from both the SSFM and theoretical results. Zhang et al.15 used a dual-spectral interferometric imaging biosensor to quantitate the conformation of protein-DNA binding. However, the DNA concentration proposed was 30 µM, which resulted in low detection sensitivity of sample. In this paper, we propose a modified label-free method based on hyper-spectral interference to overcome the shortcomings of fluorescence detection and a weight-phase algorithm to increase the sensibility and accuracy of existing label-free detection methods. We validated that the method could be used as an accurate, comprehensive, and convenient quantitative measurement for nanoscale molecular interactions without fluorescent labels, complex systems, and tedious parameter measurements. Compared to existing label-free methods, our technique achieved the accuracy of FRET detection. In addition, our method overcame the 1-10 nm range limitation of the FRET method. Thus, the dynamic range of our method is large, and it detected the thickness change of molecules in the range of 1-4000 nm16. We verified that this method could detect DNA hybridization-induced changes in the thickness accurately, quantitatively, and conveniently, and it could calculate the efficiency of DNA hybridization in the constructed DNA microarray chip. The results of our method were verified by FRET measurement results and theoretical values.

MATERIALS AND METHOD Oligonucleotide Synthesis. All oligonucleotides were synthesized using basic phosphoramidite chemistry and purified by high-performance liquid chromatography (Sangon Biotech, Beijing, China). The full details are provided in Table S1 of the Supporting Information. Four sets of oligonucleotides of different length (8, 16, 32, and 64 nt) were synthesized and named target-1, target-2, target-3 and target-4, respectively. For quantitative fluorescence studies, a cyanine 3 (Cy3) donor fluorophore and a cyanine 5 (Cy5) acceptor fluorophore were used. Three sets of oligonucleotides that differed with respect to the attached fluorophore were also synthesized. These were respectively, Cy5-(target-n) with a Cy5 attached to the 5’ terminus; (target-n)-Cy3 with a Cy3 attached to the 3’ terminus, and Cy5-(target-n)-Cy3 with a Cy5 attached to the 5’ terminus and a Cy3 attached to the 3’ terminus, where n is the number of target probe. Another four sets of oligonucleotides of different length (8, 16, 32, and 64 nt) were synthesized and named capture-1, capture2, capture-3, and capture-4, respectively. These were complementary to the corresponding target probe. These four sets of oligonucleotides were modified with a 3’ amidocyanogen and named capture-n-NH2, where n is the number of the capture probe. For quantitative fluorescence studies, we designed another capture probe named Cy3-capture-n, with a Cy3 attached to the 5’ terminus and 3’ terminal amidocyanogen, where n is the number of the capture probe. DNA Microarray Construction. A silicon wafer, whose surface was coated with a 505.4-nm-thick silica film by thermal oxidation (Tebo, Harbin, China), was used to make the microarray chip. After surface hydroxylation modification by oxygen plasma, the chip was cut into a 4×4 mm square and placed

Page 2 of 9

in 6% 3-aminopropyltriethoxysilane (APTES) solution for 1 h to form an amino-modified chip. Then, the chip was treated with glutaraldehyde and NaBH4 solution to formal aldehydemodify the chip, which is mainly used to firmly fix the capture probe with its 5' amino-modified end on the surface of the chip. Finally, the capture probe was dissolved in the DNA spotting buffer (CapitalBio, Beijing, China) and diluted to five different concentrations ranging from 4-0.25 µM. The capture probe was spotted and fixed on the surface of the chip using a SmartArrayer48 microarray spotter (CapitalBio, Beijing, China). Two types of microarray chip were fabricated by spotting capture-n-NH2 and Cy3-capture-n, respectively. As shown in Figure 1, each type of microarray chip contained five subarrays of five different capture probe concentrations, and each sub-array contained eight duplicate spots.

FIGURE 1. Spotting pattern of the DNA microarrays. C1, 4 µM; C2, 2 µM; C3, 1 µM; C4, 0.5 µM; and C5, 0.25 µM.

All of target probes were diluted in hybridization buffer (5× Denhardt’s solution, 0.2% SDS, and 3×SSC in deionized water) to 1 µM. As shown in Figure 2, for label-free measurement, 5 µL of target-n probe was hybridized to the capture-nNH2 chip to fabricate the label-free chip. For FRET measurement, 5 µL of target-n probe was hybridized to the capture-n chip and capture-n-NH2 chip to fabricate four types of DNA microarray chips: Cy5-DNA chip (n), Cy3-DNA chip (n), Cy5-DNA-Cy3 chip (n), and Cy5-DNA-excessCy3 chip (n), respectively.

FIGURE 2. Schematic of the different DNA microarrays. (A) Figure key. (1) Chip structure; (2) capture-n-NH2 chip; (3) Cy3capture-n chip; (4) Cy5-(target-n) probe; (5) (target-n)-Cy3 probe; and (6) Cy5-(target-n)-Cy3 probe. (B) Chip (n) for label-free measurement after hybridization. (C) Chip (n) for FRET measurement after hybridization.

After hybridization at 42°C for 2 h, these microarrays were washed for 4 min, first in washing buffer I (0.2% SDS and 2× SSC) and then with washing buffer II (0.2× SSC) at 42°C.


Page 3 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Finally, they were dried by centrifugation at 1600 rpm for 1 min. Data Acquisition and Processing. FRET chips were scanned using a LuxScan HT24 (CapitalBio, Beijing, China) laser confocal microarray scanner that was equipped with three laser channels (a new channel compared to conventional dualchannel scanner, which combined the laser in the Cy3 channel and the emission filter of the Cy5 channel) to detect the FRET emission signal. The microarray images were generated and analyzed by LuxScan v3.0 imaging and analysis software (CapitalBio, Beijing, China). Label-free chips were scanned using a hyper-spectral interference platform. The incident light that is reflected by the top surface of the specimen layer or the top surface of the silica and the silica-silicon interface create interference.16, 17As shown in Figure 3, a hyper-spectral interference platform was built to acquire the reflected spectrum and process the spectrum data. The incident light is provided by a halogen lamp (Saifan, Beijing, China) and passes through a condenser, beam splitter, and objective lens. Then, the light that is reflected by the top surface of the specimen layer or the top surface of the silica interferes with the light that is reflected by the silicasilicon interface. Subsequently, the reflected light passes through the objective lens, beam splitter, and the tube lens, and the reflection spectrum is acquired by a spectrometer (Ocean Optics, U.S.A.) through an optical fiber. Finally, the images are generated and analyzed using custom software that we developed with C sharp language and MATLAB.

Weighted-phase Algorithm for the DNA Microarrays. According to the conventional theory of light interference, the intensity of reflected light from a specimen can be resolved by the following equation:

i(λ ) = is (λ ) + ir (λ ) +  4π n(d + x)  2 ⋅ ( ρ e) ⋅ s(λ ) ⋅ σ k ⋅ ∆t ⋅ is (λ ) ⋅ ir (λ ) ⋅ cos   (1) λ   where λ is the wavelength of light, i ( λ ) is the intensity of the reflected light, is (λ ) is the intensity of the light reflected from the silica-silicon interface, and ir (λ ) is the intensity of the light reflected from the top surface of the specimen layer. ρ is the detector response factor, e is the electronic charge, s (λ ) is the density function of the light source, σ k is the spectral resolution of the spectrometer, and ∆t is the integral time of the spectrometer. n is the index of refraction, d is the thickness of the silica layer, and x is the thickness of the specimen. Considering that is (λ ) and ir (λ ) are constant parameters, we focused

on

the

variable

term.

K = ( ρ e ) ⋅ s ( λ ) ⋅ σ k ⋅ ∆ t and k =

2π

λ

Moreover,

we

let

because all of the pa-

rameters above are constant. Therefore, the third term of equation 1 can be simplified as the following equation:

i ' ( λ ) = is' ( λ ) + ir' ( λ ) +2 ⋅ K ⋅

is' ( k ) ⋅ ir' ( k ) ⋅ cos [ 2 nk ( d + x ) ]

(2)

where k is the wave number of light, is' (k ) and ir' (k ) are deformed from is (λ ) and ir (λ ) due to variable substitution. After Fourier transform, we obtained the following equation:

I ' (k ) = I s' (k ) + I r' (k ) K⋅

FIGURE 3. Schematic graph of the hyper-spectral interference platform.

Three-channel FRET Method for DNA Microarray. The detection principle of the three-channel FRET method is explained elsewhere.5, 18, 19 In short, three different filter sets were used to generate fluorescent signals and calibrate the sensitized-emission FRET. Each filter set consists of an excitation bandwidth filter that selectively excites the donor or acceptor and an emission bandwidth filter that selectively collects the donor or acceptor emission. However, the sensitizedemission consists of acceptor emission and FRET from the excited donor to the acceptor. To obtain the accurate FRET emission that can be used to calculate the FRET efficiency, one must eliminate the crosstalk between the donor and acceptor emission. According to the three-channel FRET method, three kinds of reference specimens were used to calibrate the system parameters and calculate the crosstalk factors. The detailed calculation process is presented in the Supporting Information.

I s' ( k ) ⋅ I r' (k ) ⋅ E (2nk0 d n ) ⋅ exp(± j ⋅ 2k0 n ⋅ xn ) (3)

Equation 3 is the expression of equation 2 in the frequency domain. From equation 3, we know that different thicknesses of the silica layer will represent different frequency peaks in the frequency domain ( E (2 nk 0 d n ) ). However, the limitation of the bandwidth of the light source and sampling window decrease the resolution of the frequency domain. Traditionally, when the width of white light is 400-1000 nm and the center wavelength is 532 nm, the thickness resolution of the frequency domain is ~50 nm. Moreover, the thickness of d and x in time domain will change to d n and xn in the frequency domain, and these parameters follow the function d + x = d n + xn . At this point, we can obtain the expression of reflected light in the frequency domain, which contains the thickness signal of the specimen. However, in the actual experiment, the light spot is not infinitesimal, which means that the reflected light may contain different thickness signals. Therefore, we suppose the reflected light spectrum is mixed with both specimen spectrum and pure silica-silicon spectrum and define the following function: Y ( k ) = (1 − m ) P exp( ± jQxn ) + mP ⋅ exp [ ± jQ ( xn + δ x ) ] (4)



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

where

P = I s' ( k ) + I r' ( k ) + K ⋅

I s' ( k ) ⋅ I r' ( k ) ⋅ E (2nk0 d n )

and Q = 2 k 0 n . m represents the linear-weighed variable for specimen spectrum, while (1 − m ) represents the linearweighed variable for pure silica-silicon spectrum. δ x is the thickness of the specimen in the frequency domain, x n is the thickness of a pure silica-silicon layer in the frequency domain, and x n is a constant for the same silica-silicon layer. As for a given value of k , we can easily determine that the function Y ( k ) is a binary monotonic function of variable m and δ x by derivation, which means we can obtain the only numerical solution under the provided condition. In addition, the actual reflected light spectrum acquired for one pixel by the spectrometer is composed by series of discrete Y ( k ) . In order to quantitate the difference between actual spectrum and theoretical spectrum, we construct the following function D : D = ∑ A (k ) − Y (k )

(5)

k

where A ( k ) is the actual value of light intensity acquired by the spectrometer for a given value of k . In the Equation 5, we subtract A ( k ) to the theoretical Y ( k ) and take the absolute value, then summarize all values for all discrete k of the detection range and get D , which is also a binary monotonic function of variable m and δ x . We can calculate the only solution of m and δ x by solving the minimum valve of function D . In this way, we can get the value of m and δ x for one pixel. Then, we can get a group of value of m and δ x by scanning the specimen pixel by pixel, and get the mean value and standard deviation (S.D.) of δ x .

Simulation and Validation of the Weighted-phase Algorithm. To test the feasibility of our method, we simulated the spectrum of different sample thicknesses and calculated results

Page 4 of 9

using different algorithms, such as Fast Fourier Transformation14, 17 (FFT), phase algorithm18, and differential algorithm16 in Figure 4. In Figure 4A, we simulated reflected spectrums of samples with 510, 515, and 520 nm thickness (where m is 0.5), and as the sample thickness increased, the reflected spectrum experienced a red shift. As shown in Figure 4B, we compared the calculated results with the actual sample thickness using different algorithms (where m is 0.5). We found that the results of the weighted-phase algorithm developed in this paper fit well with the actual sample thickness, but the differential algorithm, the phase algorithm, and FFT displayed significant errors. As shown in Figure 4C, we compared m with the calculated sample thickness using different algorithms (where the actual sample thickness is 520 nm). The results of the weighted-phase algorithm developed in this paper fit well with the actual sample thickness when m changed, but the results of the differential algorithm and the phase algorithm only fit well when m = 1. Moreover, FFT showed significant errors. The reason for this phenomenon is that first, the differential algorithm simplifies the Gaussian term in the formula. Thus, when the sample thickness increased, the error also increased. Second, the phase algorithm assumed that m is one, so it caused error when m was < 1. Finally, the FFT algorithm was limited by the frequency-sampling window, so the results displayed a step-like distribution. Moreover, our method also takes the size of light spot into consideration. In the actual experiment, the size of the light spot is not infinitely small, so the signal may contain more than one thickness value per sample. When the light spot reaches the edge of the sample, the reflected spectrum is mixed with sample spectrum and silica layer spectrum. The conventional white light interference method uses the mixed spectrum to compute the thickness, which is actually the average thickness of the sample and silica layer14. In our method, we can acquire more accurate thickness results by discriminating the mixed spectrum using equation (4).

FIGURE 4. Theoretical simulations using different algorithms. (A) Simulation of the reflected spectrum of samples with different thicknesses (where m is 0.5). (B) Simulation of calculated results of different algorithms for sample thickness (where m is 0.5). (C) Simulation of calculated results of different algorithms when m ranges from 0-1 (where sample thickness is 520 nm).


Page 5 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


RESULTS Measurement results of the three-channel FRET method. FRET is a classical measurement approach to detect the interaction distance between two fluorophores. As mentioned in Figure 2, we used four types of microarray chips to measure the interaction distance of DNA hybridization for each basepair. We present the fluorescent images of 16 bp in Figure 5. Moreover, we used the three-channel FRET method to calibrate the system parameters and calculate the FRET efficiency of each Cy3-Cy5 probe set. Last, we calculated the interaction distance of Cy3 and Cy5 by using the following equation:

E=

1 R 1 + ( )6 R0

FIGURE 6. Results of the label-free method for DNAs of different length (from 8 bp to 64 bp, probe concentration for hybridization is 0.5 µM, and minimum spotting concentration is 0.25 µM at the bottom row).

(6)

where E is the FRET efficiency of the probe set, R0 is the distance at which energy transfer is 50% efficient (which is 6.01 nm for Cy3-Cy5), and R is the actual distance of the probe set.

Comparison of the two methods on different DNA concentrations. The results of different spotting probe concentrations are shown in Figure 7, and both FRET and label-free experiment on DNA microarray was repeated six times. The details of data statistical analysis are provided in Table S2 and S3 of the Supporting Information.

FIGURE 5. Scanned microarray images. The respective images were obtained by scanning the Cy5-DNA chip, Cy3-DNA chip, and Cy5-DNAexcessCy3 chip through the Cy5 channel, the Cy3 channel, and the FRET channel. The excitation power of the green and red lasers was set to 50 and 30, respectively, and the photomultiplier tube (PMT) gain was set to 530.

Measurement images from the label-free method. The images from the label-free method were generated and analyzed via software that we developed. Figure 6 shows the processed images of 0.5 µM probe concentration for hybridization, which indicate that our method had good repeatability. In the processed image, we have visually displayed the thickness variation of different DNA lengths. Compared to the conventional fluorescent method, our method is simpler and more intuitive.

FIGURE 7. Results of different spotting probe concentrations. (A) FRET efficiency of the three-channel FRET method. (B) Thickness results of the label-free method. (From 8 bp to 64 bp, probe concentration for hybridization is 0.5 µM, and minimum spotting concentration is 0.25 µM).

In Figure 7A, FRET efficiency increases as the spotting probe concentration increases. This phenomenon is caused by intermolecular FRET on the solid surface. Intermolecular FRET is unique on solid surfaces. Due to hybridization, cleaning, and



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

drying, the molecules on the solid phase cannot freely diffuse and tend to gather, resulting in close proximity of the doublestranded DNA (dsDNA) molecules to one another. When the distance between the different fluorophores of the dsDNA is < 10 nm, intermolecular FRET can occur. Therefore, to accurately measure the FRET efficiency of the fluorophores on the same piece of dsDNA, we must eliminate intermolecular FRET. One possible solution is decreasing the density of the fluorophores by decreasing the probe concentration. From Figure 7A, one can see that when the probe concentration is > 0.5 µM, the FRET efficiency of the different samples tended to increase. This is because the increase of concentration decreases the distance between every DNA probe and leads to intermolecular FRET, which is considered noise when attempting to measure accurate intramolecular FRET. Therefore, we used the FRET efficiency of 0.5 µM to calculate the interaction distance of the fluorophores. Moreover, the FRET efficiency was near zero for 64 bp because the interaction distance was > 10 nm, a distance at which the FRET effect is extremely weak. In Figure 7B, the results of thickness show no remarkable change with changes in probe concentration. This is because our method considers the mixed reflected spectrum theoretically. In equation 4, two variables, m and δ x , are included. When the probe concentration decreases, the value of m decreases, and the value of δ x remains constant. Thus, the experimental results matched the theoretical evaluation. Compared to the fluorescent method, our label-free method can eliminate the effect of probe concentration and had a wider detection range. Comparison of the two methods on different DNA lengths. In Table 1, we present the thickness measurement results of the two tested methods and the theoretical values. For the theoretical value of the DNA length, we used 0.34 nm for each bp20 and considered the tilt angle of dsDNA on a solid surface21 (48°). We found that our method matches well with the fluorescent method for DNA lengths of 16 and 32 bp. However, the FRET method cannot detect distances > 10 nm, and the sensitivity of the FRET method is limited when the interaction distance of the fluorophores is very low (< 3.5 nm), where the FRET efficiency is > 96%. Therefore, the FRET efficiency is near zero for 64 bp and displayed a marked difference between our label-free method for 8 bp. Table 1 shows that the results of our label-free method correlate with the theoretical value and FRET results. We can see from the contrast results at Table 1 that FRET thickness error to the theoretical values for 8 bp is about 59%, 16 bp about 8%, 32 bp about 15%, and 64 bp about 39%. And the thickness error to the theoretical values of our label-free method at Table 1 for 8 bp is less than 11%, 16 bp less than 7%, 32 bp less than 11%, and 64 bp less than 13%. And we can conclude that the results of the label-free method are more close to the theoretical values. Table 1. Results of two measurement methods on different DNA lengths. DNA lengths 8 bp Methods

16 bp

32 bp

64 bp

Theoretical value

2.02 nm

4.04 nm

8.08 nm

16.16 nm

Label-free method

2.24 nm

4.31 nm

7.23 nm

14.17 nm

Relative error (%)

10.9%

6.68%

10.5%

12.3%

Page 6 of 9

FRET

3.21 nm

4.35 nm

6.83 nm

9.85 nm

Relative error (%)

58.9%

7.67%

15.5%

39.0%

Above all, it was obvious that our method was capable of measuring the thickness increases exactly, as opposed to the FRET method. Compared with FRET method, our method had higher accuracy and broader dynamic range of detection, as the FRET method cannot detect distances > 10 nm. Detection sensitivity and detection limit of the label-free method. To measure the detection sensitivity and detection limit of our method, we conducted further experiments with probe lengths of 5, 6, and 7 bp and plotted a graph of theoretical values and calculated results of the label-free method (probe concentration for hybridization is 0.5 µM). The details of data statistical analysis are provided in Table S4 of the Supporting Information. Figure 8 shows a linear plot, and the slope of the linear fit is 0.837. Thus, the detection sensitivity was 0.837. Moreover, we used the following equation to calculate the detection limit of our method:

L=

3σ k

where L is the detection limit,

(7)

σ

is the standard deviation of

the calculated result, and k is the slope of the linear curve. We used the standard deviation of 8 bp, 0.5 µM for σ , which is 0.498. Ultimately, we calculated the detection limit using equation 7, which was 1.78 nm.

FIGURE 8. Linear fitting and sensitivity measurement of theoretical valves and the calculated value of the label-free method. The probe concentration for hybridization was 0.5 µM and minimum spotting concentration is 0.25 µM.

DISCUSSION Measurement of nanoscale interactions, such as DNA hybridization on solid surfaces, is a concern in molecular biology. One of the main problems with such measurements is that the conformation of dsDNA is more complicated than in a liquid phase. To estimate the theoretical value of dsDNA on a solid surface, we used a tilt angle model and assumed that the tilt angle was 48°20. Then, we used three-channel FRET measurements, which is a commonly used method for interaction distance measurements on solid surfaces, to measure the nanoscale interaction distance. The FRET measurement results


Page 7 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


correlated with the theoretical value, which demonstrated the feasibility of the model. For label-free detection, it is important to know that the thickness variation that we measured is caused by the dsDNA molecules. We measured the thickness of a microarray chip before and after probe spotting, and the results are shown in the Supporting Information. We found that after spotting the probes, the thickness increased by ~1 nm because of the addition of the single-stranded DNA (ssDNA). However, ssDNA cannot form a tilt angle and lays flat on the surface14, 20. Therefore, the thickness variation caused by ssDNA of different lengths remains constant, and the single-strand layer was subtracted from the results of dsDNA in Table 1. Compared to traditional calculation methods of white light interference, our method has two advantages. First, traditional methods assume that the surface of each exposure is very flat, and the light spot can cover the area of the sample. However, due to the limitations of light diffraction, the light spot can be > 5 µm (as the spectrum ranges from 400-1000 nm), and the diameter of the spotted probe is ~50 µm. When the light spot scans to the edge of the spotted probe, the reflected spectrum is mixed with both sample thickness and silica layer thickness. Second, as mentioned in Figure 4, traditional methods like FFT only concern the frequency value after Fourier transformation, and the thickness variation caused by the dsDNA molecules is hidden behind the noise and cannot be clearly obtained. However, our modified method overcame the two weaknesses above to collect results that were more accurate.

CONCLUSION In this paper, we propose a modified label-free method based on hyper-spectral interference to overcome the shortcomings of fluorescence detection and existing label-free detection methods. We validated that the method can be used as an accurate, comprehensive, and convenient quantitative measurement for nanoscale molecular interactions without the need for fluorescent labels, complex systems, and tedious parameter measurements. Compared to existing label-free methods, our method achieved the accuracy of FRET detection. In addition, our method overcame the distance limitation (range of 1-10nm) of the FRET. Thus, the dynamic range of our technique was large, and it could detect the thickness changes of dsDNA in the range of 5-64 bp. We verified that this method could accurately detect DNA hybridization-induced changes in the DNA thickness quantitatively and conveniently and could calculate the efficiency of DNA hybridization in the constructed DNA microarray chip. The results of our method were verified by FRET measurement results and theoretical values. The detection limit of our method was 1.78 nm.

ASSOCIATED CONTENT Supporting Information Supporting Information Available: [Details of the oligonucleotides used and data statistical analysis.] This material is available free of charge via the Internet at http://pubs.acs.org.

Corresponding Author * E-mail: [email protected]

Author Contributions Q.L. and R.F. contributed equally.

Notes The authors declare no competing financial interest.

ACKNOWLEDGMENT This work was supported by the National Natural Science Foundation of China (81327005, 61361160418, 61575100), the National Foundation of High Technology of China (2012AA020102, 2013AA041201), the National Key Foundation for Exploring Scientific Instruments (2013YQ190467), the Beijing Municipal Natural Science Foundation (4142025), the Beijing Lab Foundation, and the Tsinghua Autonomous Research Foundation (2014Z01001).

REFERENCES (1) Parkhurst, K. M.; Parkhurst, L. J.Biochem. 1995, 34, 285-292. (2) Best, R. B.; Merchant, K. A.; Gopich, I. V.; Schuler, B.; Bax, A.; Eaton, W. A.Proc. Natl. Acad.Sci.U.S.A. 2007, 104, 18964-18969. (3) Sun, Y.; Shopova, S. I.; Wu, C. S.; Arnold, S.; Fan, X. Proc. Natl. Acad.Sci.U.S.A. 2010, 107, 16039-16042. (4) Mou, X. B.; Li, T. T.; Wang, J. H.; Ali, Z. S.; Zhang, Y. Y.; Chen, Z.; Deng, Y.; Li, S.; Su, E. B.; Jia, Q. Y.; He, N. Y.; Ni, J.; Cui, D. X. J Biomed Nanotechnol. 2015, 11, 2057-2066. (5) Gordon, G. W.; Berry, G.; Liang, X. H.; Levine, B.; Herman, B. Biophys.J. 1998, 74, 2702-2713. (6) Ha, T.; Enderle, T.; Ogletree, D. F.; Chemla, D. S.; Selvin, P. R.; Weiss, S. Proc. Natl. Acad.Sci.U.S.A. 1996, 93, 6264-6268. (7) Koushik, S. V.; Chen, H.; Thaler, C.; Puhl, H. L.; Vogel, S. S. Biophys.J. 2006, 91, L99-L101. (8) Pelet, S.; Previte, M. J. R.; So, P. T. C. J.Biomed. Opt. 2006, 11: 034017. (9) Dong, Z.; Huang, G.; Xu, S.; Deng, C.; Zhu, J.; Chen, S.; Yang, X.; Zhao, S. J Microsc-Oxford. 2009, 234, 255-261. (10) Rizzo, M. A.; Piston, D. W. Biophys.J. 2005, 88, L14-L16. (11) Mattheyses, A. L.; Hoppe, A. D.; Axelrod, D. Biophys.J. 2004, 87, 2787-2797. (12) Zhu, J.; Deng, C.; Huang, G. L.; Xu, S. K.; Mitchelson, K.; Cheng, J. Anal. Chem. 2009, 81, 1426-1432. (13) González-Guerrero, A. B.; Maldonado, J.; Herranz, S.; Lechuga, L. M. Anal. Methods. 2016, 8, 8380-8394. (14) Moiseev, L.; Unlu, M. S.; Swan, A. K.; Goldberg, B. B.; Cantor, C. R. Proc. Natl. Acad.Sci.U.S.A.2006, 103, 2623-2628. (15) Zhang, X.; Daaboul, G. G.; Spuhler, P. S.; Droge, P.; Unlu, M. S. Nanoscale. 2016, 8, 5587-5598. (16) Wang, T. Z.; Li, Q.; Li, X.; Zhao, S. M.; Lu, Y.; Huang, G. L. Opt.Lett. 2013,38, 1524-1526. (17) Ozkumur, E.; Needham, J. W.; Bergstein, D. A.; Gonzalez, R.; Cabodi, M.; Gershoni, J. M.; Goldberg, B. B.; Unlu, M. S. Proc. Natl. Acad.Sci.U.S.A. 2008, 105, 7988-7992. (18) Zhu, J.; Lu, Y.; Deng, C.; Huang, G. L.; Chen, S. Y.; Xu, S. K.; Lv, Y.; Mitchelson, K.; Cheng, J. Anal. Chem. 2010, 82, 5304-5312. (19) Xie, F. B.; Zhu, J.; Deng, C.; Huang, G. L.; Mitchelson, K.; Cheng, J. Analyst. 2012, 137, 1013-1019. (20) Lee, N. K.; Kapanidis, A. N.; Wang, Y.; Michalet, X.; Mukhopadhyay, J.; Ebright, R. H.; Weiss, S. Biophys.J. 2005, 88, 2939-2953. (21) Ambia-Garrido, J.; Vainrub, A.; Pettitt, B. M. Comput Phys Commun. 2010, 181, 2001-2007.

AUTHOR INFORMATION



Page 8 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment

8

Page 9 of 9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


150x58mm (220 x 220 DPI)


Label-Free Method Using a Weighted-Phase ... - ACS Publications

Recommend Documents