Combined microfluidic-optical DNA analysis with single-base-pair sizing capability

DNA sequencing by microchip capillary electrophoresis (CE) enables cheap, highspeed analysis of low reagent volumes. One of its potential applications is the identification of genomic deletions or insertions associated with genetic illnesses. Detecting single base-pair insertions or deletions from DNA fragments in the diagnostically relevant size range of 1501000 base-pairs requires a variance of  < 10. In a microfluidic chip post-processed by femtosecond-laser writing of an optical waveguide we CE-separated 12 blue-labeled and 23 red-labeled DNA fragments in size. Each set was excited by either of two lasers powermodulated at different frequencies, their fluorescence detected by a photomultiplier, and blue and red signals distinguished by Fourier analysis. We tested different calibration strategies. Choice of the fluorescent label as well as the applied fit function strongly influence the obtained variance, whereas fluctuations between two consecutive experiments are less detrimental in a laboratory environment. We demonstrate a variance of   410, lower than required for the detection of single base-pair insertion or deletion in an optofluidic chip. © 2016 Optical Society of America OCIS codes: (130.3120) Integrated optics devices; (170.3890) Medical optics instrumentation. References and links 1. A. Manz, N. Graber, and H. M. Widmer, “Miniaturized total chemical analysis systems: a novel concept for chemical sensing,” Sens. Actuator BChem. 1(1-6), 244248 (1990). 2. R. D. Reyes, D. Iossifidis, P. A. Auroux, and A. Manz, “Micrototal analysis systems


Introduction
A lab on a chip [13] squeezes the functionalities of a biological or chemical laboratory onto a single substrate.Its advantages are speed of analysis, low reagent volumes, and measurement automation and standardization.Capillary electrophoresis (CE) is a powerful method for biomolecule separation and analysis [46].The sorting and sizing of DNA molecules within the human genome project [7] has been enabled largely by CE separation and analysis [8] and has led to the genetic mapping of various human illnesses [9].On-chip integration of DNA sequencing [1013] and genetic diagnostics [14,15] have become feasible.Laser-induced fluorescence exploiting fluorescent dye labels is the most popular microchip CE monitoring technique, allowing a low limit of detection of 210 fM [16].
Micro-optical components have been integrated in microfluidic labs on a chip to perform on-chip optical detection [1727].Integrated optical waveguides allow one to confine and transport light in the chip, directing it to a small volume of the microfluidic channel and collecting the emitted/transmitted light, as has recently been applied to monitor on-chip DNA sequencing [11,28] in a now commercialized DNA sequencer [13].
Sizing accuracy poses an inherent challenge to CE and particularly microchip CE [2931].Application of CE-based DNA sequencing in a lab-on-a-chip to identify genomic deletions or insertions associated with genetic illnesses critically depends on the detection of single basepair insertions or deletions from DNA fragments in the diagnostically relevant range of 1501000 base-pairs, i.e., it requires a variance of  2 < 10 -3 , while only  2 < 10 -2 were reported [32,6,33].Here we test different calibration strategies and demonstrate CE-based DNA analysis in an optofluidic chip with sub-base-pair resolution ( 2  410 -4 ) of low concentrations of DNA molecules, thereby paving the way for the envisaged application.

Experimental
A schematic of the commercial microfluidic chip [34] is displayed in Fig. 1.The microfluidic channel network and reservoirs were patterned photolithographically and wet-etched in fused silica glass and then sealed off by bonding another piece of fused silica glass on top.The chip has dimensions of 55 mm  5.5 mm  1 mm and the microfluidic channels have a crosssection of ~110 µm width and ~50 µm depth.The optical waveguide was inscribed into the bulk of the fused silica chip by fs-laser writing [35] at a translation speed of 20 µm/s, using a Ti:Sapphire laser operating at 800 nm wavelength with 150 fs, 4 µJ pulses at a repetition rate of 1 kHz.Employing tunable, astigmatic beam shaping [36], an elliptical cross section of the written waveguide was obtained, with a major diameter of ~50 µm in the vertical direction, while the minor diameter in the horizontal direction is ~12 µm in order to retain a high spatial resolution along the direction of DNA migration and separation.Prior to the experimental runs the inner walls of the microfluidic channel network were coated with an epoxy-poly-(dimethylacrylamide) (EPDMA)-based polymer [37].Subsequently, the channels were filled with a sieving gel matrix consisting of hydroxyethylcellulose (HEC) (2% wt./vol.)[38], dissolved in 20 mM MES / 20 mM His buffer (pH 6.2).The reagents were sterilized, filtered (0.22 µm pore size), and stored at 269 K.
The CE sample loading and separation protocol was based on optimized actuation voltages of up to 1.5 kV, delivered by Pt electrodes integrated into the microfluidic reservoirs.The entire on-chip migration was controlled with a LabVIEW (National Instruments, Inc.) script steering a microfluidic control system [39].Application of a high voltage forced the analyte molecules to migrate into the CE injection channel from sample reservoir 1 to waste reservoir 2 (see Fig. 1).By switching the voltages at all four reservoirs simultaneously a wellconfined plug containing the mixture of analyte molecules -with a volume of ~605 picoliters at the crossing junction of the two microfluidic channels -was injected into the CE separation channel, from the microfluidic crossing junction toward waste reservoir 4, and the analyte molecules contained in the plug volume were separated according to their size.The excitation waveguide intersects the CE separation channel orthogonally at a distance of ~3.6 cm from the microfluidic crossing junction at which the separation commences, toward the end of the CE separation channel, close to reservoir 4, where the plug separation is highest.
Chip surfaces not required to transmit light were blackened to suppress reflections of parasitic light.A photon-counting photomultiplier tube (H7421-40, Hamamatsu) cooled to 203 K was built onto the output port of an inverted microscope (DM5000, Leica Microsystems) and aligned to collect the fluorescence signal through the detection window.Combination of excitation/emission band filters (K3, Leica Microsystems, and XF57, Omega Optical) in the fluorescence collection path ensured that only fluorescence signals emitted by labeled DNA molecules reached the photomultiplier tube, while other sources of background signal, including scattered laser light, were rejected.The fluorescence intensity, recorded as a function of time, was further numerically post-processed to obtain a high signal-to-noise ratio.
Two sets of DNA molecules with known base-pair sizes were covalently and permanently end-labeled with different dyes during a polymerase chain reaction.Previous attempts with intercalating dyes resulted in less good sizing accuracy than reported here for covalent dyes.

Results and Discussion
The results of migration time of the 12 blue-labeled and 23 red-labeled DNA fragments versus DNA base-pair size for both experiments are displayed in Fig. 2(a) and 2(b).Rearrangement of these experimental data to compare separation of the same set of DNA molecules in consecutive experiments, as displayed in Fig. 2(c) and 2(d), shows that the 12 blue-labeled DNA molecules in Fig. 2(c) exhibit less deviation of individual DNA molecules from the fitted temporal behavior than the 23 red-labeled DNA molecules in Fig. 2(d).It is also seen from Fig. 2(c) and 2(d) that the first experiment is better reproduced by the second one for the 12 blue-labeled molecules.We fitted the data of the logarithm of the base-pair size, S bp , as a function of migration time, t, with a linear and a quadratic dependence, respectively.Table 1 provides the least-square fits of the experimental data of Fig. 2(c) and 2(d) according to Eqs. ( 1) and ( 2).Both figures show that a quadratic fit reproduces the measured data much better than a linear fit, indicating that the assumption of a simple logarithmic dependence of migration time on base-pair size underlying the production of DNA ladders with a logarithmic increase in base-pair size is not justified by our experiment.In order to reach the goal of our investigation, the demonstration of sub-base-pair resolution during DNA separation in a simple optofluidic chip with high speed and low volume consumption, we tested different calibration strategies for the dependence of   The resulting variance  2 , which quantifies the deviation of the sample data from the fit function obtained from the reference data, is given in Table 2 and displayed in Fig. 3 for both, the linear and quadratic fits from Fig. 2(c) and 2(d) for the four different approaches a)d) proposed above.These results are obtained under the optimized experimental conditions described above.Variation of these conditions typically leads to less good sizing accuracy.From the experimental results (Fig. 2) and the analysis (Fig. 3) we conclude the following: 1) Applying a quadratic [Fig.3(b)] instead of the usual linear [Fig.3(a)] fit considerably improves the accuracy of calibration.2) Blue-labeled molecules [Fig.2(c)] are separated with higher accuracy than red-labeled molecules [Fig.2(d  opposite of two possible combinations of reference and sample, thereby demonstrating the feasibility of detecting single base-pair insertion/deletion in a lab-on-a-chip.

Summary
Choice of a suitable dye label, combined with reference calibration and sample investigation by fluorescent detection in consecutive experiments, results in capillary electrophoretic separation of fluorescent-labeled DNA molecules in the 1501000 base-pair range with subbase-pair resolution, thereby paving the way towards the detection of single base-pair insertion or deletion in a lab-on-a-chip with low reagent volumes in a few-minute experiment.

Fig. 1 .
Fig. 1.Schematic of the optofluidic chip showing reservoirs 14, sample injection channel (reservoir 1  reservoir 2) and CE separation channel (reservoir 3  reservoir 4), as well as the integrated optical waveguide and detection window.
In a single flow experiment, the 12 blue-labeled (Alexa fluor 488) and 23 red-labeled (Alexa fluor 647) DNA fragments were separated in size by microchip CE.Each set was excited exclusively by either of two lasers power-modulated at different frequencies of 17 Hz and 31 Hz and launched through the optical waveguide.The emitted and detected blue and red signals were distinguished by Fourier analysis.In this manner one can separate differently encoded fluorescence signals from each other [4042].The experiment was repeated with the same optofluidic chip without renewing the inner wall coating or sieving gel matrix.
base-pair size in a given experimental situation: a) use either set of DNA molecules as a reference to calibrate the set-up and identify the base-pair sizes of the other set in the same flow experiment, thereby eliminating variations in temperature, wall-coating and sieving-gel conditions, and actuation voltages; b) use the same molecular set as reference and sample (in a real-life experiment the reference set would be the healthy counter-part of an unknown, potentially malign sample set) with the same fluorescence label, flown in consecutive experiments; c) perform cross-experiments based on different molecular sets with different labels, flown in consecutive experiments; d) also self-calibration in the same experiment was analyzed.In the calibration procedure all base-pair sizes of the calibrating set contribute to the sizing accuracy at all base-pair sizes of the sample set, hence the sizing accuracy does not change significantly if the two sets have slightly different base-pair sizes.

Fig. 3 .
Fig. 3. Variance  2 of sample data measured in experiment ("exp.") 1 or 2 from (a) the linear fit function of Eq. (1) and (b) the quadratic fit function of Eq. (2), which were obtained from the measured reference ("ref.")data from the same or the different ("diff.")experiment.
)]; apparently, different dye labels influence the DNA flow differently.3) In a laboratory environment, different dye labels affect the microfluidic flow of DNA plugs more severely than variations in temperature, wallcoating and sieving-gel conditions, and actuation voltages between consecutive experiments.4) Choosing a single, suitable dye label (here: Alexa fluor 488), combined with reference calibration and sample investigation in consecutive experiments [left-hand side of dark-blue curve in Fig.3(b)], results in a variance of  2 = 4.2  10 -4 in one and  2 = 2.4  10 -4 in the