s 70 (2007) 773–777www.elsevier.com/locate/jbbm

J. Biochem. Biophys. Method

Novel fluorescence detection technique for non-contact temperaturesensing in microchip PCR

Sudip Mondal ⁎, V. Venkataraman

Department of Physics, Indian Institute of Science, Bangalore-560012, India

Received 5 January 2007; received in revised form 30 April 2007; accepted 1 May 2007

Abstract

DNA amplification using Polymerase Chain Reaction (PCR) in a small volume is used in Lab-on-a-chip systems involving DNA manipulation.For few microliters of volume of liquid, it becomes difficult to measure and monitor the thermal profile accurately and reproducibly, which is anessential requirement for successful amplification. Conventional temperature sensors are either not biocompatible or too large and hencepositioned away from the liquid leading to calibration errors. In this work we present a fluorescence based detection technique that is completelybiocompatible and measures directly the liquid temperature. PCR is demonstrated in a 3 μL silicon-glass microfabricated device using non-contactinduction heating whose temperature is controlled using fluorescence feedback from SYBR green I dye molecules intercalated within sensorDNA. The performance is compared with temperature feedback using a thermocouple sensor. Melting curve followed by gel electrophoresis isused to confirm product specificity after the PCR cycles.© 2007 Elsevier B.V. All rights reserved.

Keywords: Fluorescence detection; Microchip-PCR; SYBR Green I; Non-contact temperature sensor

1. Introduction

Polymerase chain reaction (PCR) is one of the unavoidablecomponents in Lab-on-a-chip devices that require amplificationof a selective part of a DNA. PCR involves cycling a smallvolume of liquid through three temperatures: (i) denaturationtemperature (∼95 °C) for melting the double-stranded DNA(ds-DNA) into two single stranded DNA (ss-DNA), (ii)annealing temperature (∼50 °C) for annealing primer (shortss-DNA) to the ss-DNA template and (iii) extension tempera-ture (72 °C) for extension of the primer annealed ss-DNAtemplate to double the amount of DNA after the end of thecycle. In conventional PCR protocols a large volume of reactionmixture (∼25 μL) is cycled through the three temperaturesusing a commercial thermocycler. With the development of bio-micro-electromechanical systems (bio-MEMS), the amount ofreaction volume required has reduced (b3 μL), leading to areduction of total analysis time. On the other hand, miniaturiza-

⁎ Corresponding author. Tel.: +91 80 2293 2316; fax: +91 80 2360 2602.E-mail address: sudipm@physics.iisc.ernet.in (S. Mondal).

0165-022X/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.jbbm.2007.05.001

tion has also increased the surface to volume ratio causing non-uniformity in temperature distribution within the small volumecompared to large volume conventional reactions. It is achallenging task to measure and calibrate the thermal profileinside a small volume of liquid in the microchip. In previousreports, conventional temperature sensors such as platinumresistor [1–3] and thermocouple [4] were integrated andpackaged on the chip using standard CMOS technologies.Integrated sensors require complex microfabrication steps andhave to be calibrated individually for every chip design [5]. Insome cases, such as platinum resistors, it is too expensive tointegrate the sensor for low cost disposable chip applications[5]. Conventional sensors are either too bulky or non-biocompatible to be placed inside the reaction mix formeasuring in-situ temperature. Attempts have been madeduring the last decade to measure in-situ temperature of asmall volume of liquid in microchannels using non-contactsensors such as Raman [6], NMR [7] and fluorescence dyes[8,9]. Though micro-Raman has been used to demonstrate non-contact mode of temperature measurement for PCR in amicrochip [10], it requires a bulky and complicated opticalarrangement and therefore not suitable for miniaturization.

774 S. Mondal, V. Venkataraman / J. Biochem. Biophys. Methods 70 (2007) 773–777

Fluorescence detection in microchip PCR has been used inmany earlier reports to monitor DNA amplification andcalibrate the chip using melting curve analysis [2,11]. Themost common dye used for PCR is ds-DNA binding SYBR®green I (SG) [12,13]. It is well known that fluorescence of mostdye molecules is temperature dependent [14]. Therefore onemay expect that the SG fluorescence signal can directly be usedto monitor and control the PCR mix temperature. There are,however, three problems in implementing this idea in conven-tional real time microchip PCR systems. First, fluorescence istypically recorded only once per cycle and therefore notsufficient for continuous temperature monitoring. Second, thefluorescence is proportional to ds-DNA amount and changesfrom cycle to cycle depending on the amplification. Lastly, thechange in fluorescence with temperature is very small at thebeginning of the PCR cycles due to the low concentration ofinitial template DNA.

In this work we show how these problems can be overcomein a 3 μL silicon-glass microchip that is heated by non-contactinduction heating [15]. The key idea is to add non-specific“sensor” DNA to the PCR mix whose melting point andconcentration are carefully controlled. The fluorescence, whichis continuously measured using fast data acquisition with alock-in amplifier, is shown to be reproducibly temperaturedependent for all PCR cycles. The calibrated intensity is used tomeasure and control the in-situ temperature using computer-controlled set points. The amplification of Lambda DNA iscompared with standard temperature feedback using meltingcurve analysis and gel electrophoresis.

2. Materials and methods

2.1. PCR samples

The PCR mix was prepared using 2× DyNAmo™ SYBR®green qPCR kit (Finzymes, Espoo, Finland) containing SG andThermus brockianus DNA polymerase. A 50 μL mix wasprepared with 100 ng of Lambda DNA and 0.5 μM of forward(5′-AGT GTC GAATTC TGATCG TGG TGATAT CCG -3′)and reverse primer (5′-AGT GTC AAG CTT CAG CTT CAGTTC TCT -3′) to produce 311 base pair (bp) of product. 6 U ofThermus aquaticus polymerase was added in excess to over-come adverse surface adsorption effects in the microchip. Forfluorescence feedback experiments with single sensor, we addedseparately 920 ng of a 1024 bp DNA (called “sensor 1”) in 50 μLreaction volume apart from the template DNA and primers. Fordual sensor experiments, 230 ng of sensor 1 and 800 ng of a100 bp DNA (called “sensor 2”) were added in 50 μL reactionmix. Both sensor 1 and sensor 2 are non-specific to the aboveprimers and show no amplification. Their melting points andconcentrations are chosen suitably as explained below.

2.2. Microchip and thermocycler

PCR was carried out in a silicon-glass microchip fabricatedusing MEMS techniques [15]. One mm thick Pyrex #7740 glasswas bonded to a 500 μm oxidized silicon wafer by anodic

bonding and sawed to 3.5 mm×3.5 mm size. A 2 mm diameterhole in the glass formed the reservoir to hold the mix. For PCRreactions, 3 μL volume of the mix was covered with 4 μL ofmineral oil to prevent evaporation at high temperature and allowoptical access for fluorescence measurements. Use of mineraloil instead of permanent sealing with bonded glass allowed easyrecovery of the mix after the reaction. The microchip was placedon the secondary of the induction heater using a thin layer ofheat sink compound for better thermal contact. Thermal cyclingwas performed using induction heater with a K-type thermo-couple attached at the back of the secondary for temperaturefeedback as described earlier [15]. PCR using temperaturefeedback was carried out starting with an initial denaturation at95 °C for 1 min followed by 30 cycles of 95 °C for 4 s, 50 °C for8 s and 72 °C for 30 s. The cycles were completed with a 5 minof final extension at 72 °C.

2.3. Fluorescence detection

SG fluorescence detection system consisted of a 470 nm blueLED source (Roithner Lasertechnik, Vienna, Austria) projectedusing a dichroic mirror and focused on the microchip with a10x objective. The emission at 520 nmwas collected, filtered andfocused on a silicon photodiode. The LED intensity wasmodulated at 190 Hz and the photocurrent detected synchro-nously using a Lock-in amplifier SRS-830 (Stanford ResearchSystems, Sunnyvale, CA, USA) [16]. The signal, acquired every0.7 s, when plotted against time yields the continuousfluorescence emission during PCR. For fluorescence feedback,voltage equivalent of the photocurrent from the Lock-in was usedto compare with the voltage set point specified by a computer.

2.4. Melting curve

For melting curve the heater power was ramped slowly usingan uncalibrated thermocouple attached close to the microchipand the fluorescence values were recorded as a function oftemperature. The peak in the derivative of the melting curve wasdefined as the melting point TM. In case of PCR usingfluorescence feedback the mix prepared with the well-characterized sensor DNA was melted in the beginning byramping up the heater power slowly. Ramping was carried outuntil the sensor DNA melted completely and fluorescenceshowed no change. Once the fluorescence at room temperatureand at the sensor melting temperature was measured, thefluorescence at the required annealing, extension, and denatura-tion temperatures of the template could be determined by linearinterpolation. Thus the fluorescence intensity was determinedfor all three PCR setpoints without depending on an accuratelycalibrated temperature sensor.

3. Results

3.1. Conventional PCR

For conventional PCR, 3 μL of mix (without the sensorDNA) was filled in the silicon-glass microchip and overlaid

775S. Mondal, V. Venkataraman / J. Biochem. Biophys. Methods 70 (2007) 773–777

with mineral oil. The chip was cycled in the induction heaterwith temperature feedback using the protocol mentionedpreviously. The fluorescence signal and the controlled tem-perature were recorded every 0.7 s. Fluorescence profile of 30PCR cycles amplifying Lambda DNA is shown in Fig. 1A usingthe thermal profile shown in the inset. The increase influorescence envelope with cycle number indicates successfulamplification of the template. During the initial cycles,denaturation and extension fluorescence are very close to eachother but become well separated at the later cycles when theamount of DNA increases. The product specificity is confirmedusing melting curve and agarose gel electrophoresis.

For a more quantitative analysis of the fluorescenceenvelope, we assume that the molecular weight and concentra-tion of the stock solution of SG to be 509.27 g/mol and 10 mg/ml, respectively [17]. For our calculation we assume that SG isused in 1:10,000 dilution in the PCR mix i.e. ∼6×1013

molecules of SG. All numbers henceforth will be presentedreferring to 50 μL volume of the mix. The initial number ofDNA molecules is ∼2×109. During PCR, the fluorescenceincreases with the accumulation of amplified product. Thefluorescence envelope saturates when the copy number reachesa value such that all the SG molecules can bind to available sitesand no SG molecules are left free in the solution. Beyond thispoint the envelope does not change even if more product DNAmolecules are formed. In our case, the fluorescence is observedto saturate after ∼20 cycles (Fig. 1A). Since there is one SGbinding site for every 10 bp assuming minor groove binding[17], the number of amplified base pairs is expected to be∼6×1013 ×10 which is 6×1014 bp or 2×1012 productmolecules.

3.2. PCR with single sensor DNA

Fig. 1A shows an inverse relation between the fluorescenceand temperature, which can be used to track the in situtemperature of the mix. However, as emphasized in theintroduction, the fluorescence changes with amplification inthe initial cycles and therefore cannot be used for reproducibletemperature sensing. During the later cycles when the amount ofDNA increases and the envelope saturates, the fluorescencevalues are well separated at denaturation, extension andannealing temperatures and remain stable with cycling. This

Fig. 1. Fluorescence data during PCR of lambda DNA. Amplification carried out witcycles using temperature feedback. In (B) 920 ng of 1024 bp sensor DNA was adde

kind of behavior in fluorescence is suitable to be used as asensor response. In order to extend this response to the initialcycles, we add an excess amount of sensor DNA in the mix. Thenumber of sensor DNA molecules remains unchanged as theyare chosen to be nonspecific to the primers and hence do notamplify.

Experimentally we find that 920 ng of a 1024 bp DNA(sensor 1) when used as a sensor in the mix saturates theenvelope from the first cycle onwards without affecting thePCR. The envelope remains constant (as shown in Fig. 1B) eventhough the primer amplifies 311 bp fragments from lambdaDNA, as verified by agarose gel electrophoresis. The aboveamount of sensor corresponds to ∼8×1014 number of bp thatare available for the 6×1013 SG molecules to bind. Since we getnearly saturated fluorescence at the ratio of SG to bp of 1:10, wethink the binding is preferentially by minor grove mechanismand not intercalation [17]. The melting curve of the solutionboth at the beginning and end of the PCR cycling shows a singleTM value of 90 °C, corresponding to the sensor 1. Singlemelting point for the mix before PCR is well understood, whichis due to large number of sensor DNA compared to the template.After a successful amplification producing ∼2×1012 copies ofthe product the melting is overwhelmed by the 920 ng of thesensor DNA. At the product TM of 84 °C, the fraction of SGmolecules bound to the product are released into the solutionthat should have caused the fluorescence to drop. But the hugeamount of sensor 1 is capable of providing enough sites for thefree SG molecules to bind. The slow temperature ramp ratefacilitates this process and is responsible for maintaining thefluorescence until the point where the sensor starts melting.Thereafter there are no more binding sites and the fluorescencereduces abruptly. This kind of melting behavior has beenreported for multiplex PCR, due to redistribution of dyemolecules depending on the relative concentration of SG andthe different DNA molecules present in the mix [18].

Though the product specificity cannot be determined fromthe above melting curve experiment, advantage lies in the factthat the fluorescence remains unchanged until the end of theprocess, independent of the amplification. In other wordsfluorescence set points corresponding to temperature will be aconstant value. This is a simple way to incorporate fluorescencefeedback in a small volume to measure and monitor thermalresponse. In case of microchip PCR with integrated capillary

hout addition of any sensor DNA (A), inset showing thermal profile of six initiald to the mix.

Fig. 2. Comparison between temperature and fluorescence feedback control. Fluorescence (right) and temperature (left) profiles are compared for first five cycles ofPCR with temperature feedback control (A and B) with fluorescence feedback (C and D). 800 ng of 100 bp (sensor 2) and 230 ng of 1024 bp (sensor 1) DNAwereadded in the mix for amplification of 311 bp of Lambda DNA.

776 S. Mondal, V. Venkataraman / J. Biochem. Biophys. Methods 70 (2007) 773–777

electrophoresis (CE), the above idea can be used for carryingout PCR using in situ fluorescence feedback control while theproduct can be analyzed using CE.

3.3. PCR with dual sensor DNA

The use of 920 ng of sensor 1 masks the product meltingsignature and needs post-PCR electrophoresis to check productspecificity. To avoid this additional step and confirm amplifica-tion using only product melting we reduce the effective numberof binding sites from sensor 1 (by lowering the amount of sensor1 DNA in the mix) below 10 times the total number of SGmolecules. This will release considerable amount of free SGmolecules into the solution at the product TM, leading to a dropin the fluorescence signal. At this condition we expect to get aproper melting even if the product TM is lower than the sensor,but, on the other hand, it will cause an increase in fluorescenceenvelope with amplification that is undesirable for temperaturesensing. To overcome this problem, an additional sensor DNAis used. We add 800 ng of 100 bp sensor DNA (sensor 2) in the

Fig. 3. Confirmation of PCR with dual sensor DNA. Melting curve is comparedfor the 100 bp and 1024 bp sensor DNA before amplification with 311 bp PCRproduct after 30 cycles. The melting point of 100 bp sensor, 311 bp product and1024 bp sensor is denoted by TM2, TMP and TM1, respectively.

mix along with 230 ng of sensor 1 to restore back thefluorescence envelope to a constant level. The large number of7×1014 bp due to sensor 2 is preferentially responsible forsaturation of the fluorescence and produces a constant envelopecompared to the lower number of 2×1014 bp due to sensor 1.The mix, when melted before PCR, shows a strong melting at80 °C for sensor 2 and a weak melting at 90 °C for sensor 1.

The lambda DNA was amplified in presence of the twosensors for 30 cycles. The melting curve shows a gradualdecrease in fluorescence until 84 °C, where the product melts.Though sensor 2 has a lower TM value (80 °C), fluorescencedoes not drop because of the presence of large number of sitesprovided by the amplified product. But as the temperaturereaches 84 °C, the fluorescence drops due to denaturation of3×1012 copies of product DNA. The lower amount of availablesites from sensor 1 is unable to maintain the fluorescence.Finally the fluorescence reaches its background value at 90 °Cwhen sensor 1 also melts.

The background fluorescence when calibrated against actualtemperature from well-characterized TM value of the sensorDNA was used to control the cycling parameters during PCR.Fluorescence values are used as set points in the feedbackcontroller. Sensor 2 maintains the saturated fluorescenceenvelope during PCR while sensor 1 allows control oftemperature up to ∼90 °C. In Fig. 2, we compare thetemperature and fluorescence data from the chip usingtemperature control from a thermocouple (Fig. 2A and B)with our novel fluorescence feedback control (Fig. 2C and D).The products are found to be identical for both temperature andfluorescence feedback in both melting curve analysis as shownin Fig. 3 and gel electrophoresis (data not shown).

4. Discussion

In summary we have demonstrated a novel non-contactmethod of measuring and controlling temperature usingcontinuous in-situ fluorescence detection. This method usesthe same SG dye for monitoring temperature and checking

777S. Mondal, V. Venkataraman / J. Biochem. Biophys. Methods 70 (2007) 773–777

product specificity from melting, thus simplifying the opticaldetection unit. We have demonstrated both single sensor anddual sensor protocols that can be used to measure and controlthe in-situ temperature. For single sensor protocol, the TM of thesensor should be higher than that of the product and the totalnumber of bp from the sensor DNA must be higher than 10×SGmolecules used in the mix. In the case of dual sensor protocol,the product should melt between the two sensors TM. Theconcentration of the sensor with lower TM should be more than10×SG molecules while the sensor with higher TM should beless than 10×SG molecules. Both hardware and protocols caneasily be integrated into microchip PCR systems.

5. Simplified description of the methods and its applications

Monitoring the SYBR green I fluorescence during PCR withadditional DNA fragments in the solution that is non-specific toprimers can be used for temperature sensing. This protocol canbe used to measure the local temperature of a small volume ofliquid in non-contact mode unlike conventional thermal sensors.The method when used in proper composition retains theadvantage of using intercalated dye, namely, product confirma-tion using melting curve analysis after the end of amplificationcycles.

Acknowledgements

We thank Professor V. Natarajan for his help in fluorescencesetup. This work was supported by BRNS, Department ofAtomic Energy, and the Life Science Research Board (LSRB),Government of India.

References

[1] Daniel JH, Iqbal S, Millington RB, Moore DF, Lowe CR, Leslie DL,et al. Silicon microchambers for DNA amplification. Sens Actuators, A1998;71:81–8.

[2] Ali JE, Neilsen IRP, Poulsen CR, Bang DD, Telleman P, Wolff A.Simulation and experimental validation of a SU-8 based PCR thermocyclerchip with integrated heaters and temperature sensor. Sens Actuators, A2004;110:3–10.

[3] Ke C, Kelleher AM, Berney H, Sheehan M, Mathewson A. Single stepcell lysis/PCR detection of Escherichia coli in an independentlycontrollable silicon microreactor. Sens Actuators, B 2006;120:538–44.

[4] Huhmer AFR, Landers JP. Noncontact infrared-mediated thermocyclingfor effective polymerase chain reaction amplification of DNA in nanolitervolumes. Anal Chem 2000;72:5507–12.

[5] Cho YK, Kim J, Lee Y, Kim YA, Namkoong K, Lim H, et al. Clinicalevaluation of micro-scale chip-based PCR system for rapid detection ofhepatitis B virus. Biosens Bioelectron 2006;21:2161–9.

[6] Davis KL, Liu KLK, Lanan M, Morris MD. Spatially resolved temperaturemeasurements in electrophoresis capillaries by Raman thermometry. AnalChem 1993;65:293–8.

[7] Lacey ME, Webb AG, Sweedler JV. Monitoring temperature changes incapillary electrophoresis with nanoliter-volume NMR thermometry. AnalChem 2000;72:4991–8.

[8] Ross D, Gaitan M, Locascio LE. Temperature measurement in microfluidicsystems using a temperature-dependent fluorescent dye. Anal Chem2001;73:4117–23.

[9] Jeon S, Turner J, Granick S. Noncontact temperature measurement inmicroliter-sized volumes using fluorescent-labeled DNA oligomers. J AmChem Soc 2003;125:9908–9.

[10] Kim SH, Noh J, Jeon MK, Kim KW, Lee LP, Woo SI. Micro-Ramanthermometry for measuring the temperature distribution inside themicrochannel of a polymerase chain reaction chip. J Micromech Microeng2006;16:526–30.

[11] Neuzil P, Pipper J, Hsieh TM. Disposable real-time microPCR device: lab-on-a-chip at a low cost. Mol BioSyst 2006;2:292–8.

[12] Higuchi R, Dollinger G,Walsh PS, Griffith R. Simultaneous amplificationand detection of specific DNA sequences. BioTechniques 1992;10:413–7.

[13] Wittwer CT, Herrmann MG, Moss AA, Rasmussen RP. Continuousfluorescence monitoring of rapid cycle DNA amplification. BioTechniques1997;22:130–8.

[14] Ririe KM, Rasmussen RP, Wittwer CT. Product differentiation by analysisof DNA melting curves during the polymerase chain reaction. AnalBiochem 1997;245:154–60.

[15] Pal D, Venkataraman V. A portable battery-operated chip thermocyclerbased on induction heating. Sens Actuators, A 2002;102:151–6.

[16] Mondal S, Venkataraman V. In situ monitoring of polymerase extensionrate and adaptive feedback control of PCR by using fluorescencemeasurements. J Biochem Biophys Methods 2005;65:97–105.

[17] Zipper H, Brunner H, Bernhagen J, Vitzthum F. Investigations on DNAintercalation and surface binding by SYBR Green I, its structuredetermination and methodological implications. Nucleic Acids Res2004;32:e103.

[18] Giglio S, Monis PT, Saint CP. Demonstration of preferential binding ofSYBR Green I to specific DNA fragments in real-time multiplex PCR.Nucleic Acids Res 2003;31:e136.