See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264796478

Microfluidic Genotyping by Rapid Serial PCR and High-Speed Melting

Analysis

Article  in  Clinical Chemistry · August 2014

DOI: 10.1373/clinchem.2014.223768 · Source: PubMed

CITATIONS READS

21 1,063

9 authors, including:

Scott O Sundberg Carl T Wittwer

Canon Virginia University of Utah
11 PUBLICATIONS   268 CITATIONS    305 PUBLICATIONS   36,837 CITATIONS   

SEE PROFILE SEE PROFILE

Renee Howell Robert Pryor

Siemens University of Utah
16 PUBLICATIONS   550 CITATIONS    24 PUBLICATIONS   3,093 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Unlabeled probe assays View project

All content following this page was uploaded by Renee Howell on 03 January 2017.

The user has requested enhancement of the downloaded file.

Clinical Chemistry 60:10 Molecular Diagnostics and Genetics
1306–1313 (2014)

Microfluidic Genotyping by Rapid Serial PCR and

High-Speed Melting Analysis
Scott O. Sundberg,1,2 Carl T. Wittwer,1* Renée M. Howell,3 Jarkko Huuskonen,3 Robert J. Pryor,1
Jared S. Farrar,1 Heather M. Stiles,3 Robert A. Palais,4 and Ivor T. Knight3

BACKGROUND: Clinical molecular testing typically batches Most genetic diseases are rare. Laboratories typically lack
samples to minimize costs or uses multiplex lab-on-a- technology to offer rapid testing for rare diseases or geno-
chip disposables to analyze a few targets. In genetics, mul- types. Individual samples are seldom processed. Instead,
tiple variants need to be analyzed, and different work multiple samples are collected for batch analysis, often
flows that rapidly analyze multiple loci in a few targets are resulting in slow turnaround times varying from a few
attractive. days to several weeks. Despite the extraordinary amount
of information collected by massively parallel sequencing,
METHODS: We used a microfluidic platform tailored to the focus is not on turnaround time, and costs remain
rapid serial PCR and high-speed melting (HSM) to geno- high for single-gene analysis compared with scanning by
type 4 single nucleotide variants. A contiguous stream of high-resolution melting (1 ).
master mix with sample DNA was pulsed with each Turnaround times are often slow in PCR-based
primer pair for serial PCR and melting. Two study sites genetic tests because multiple reactions are performed
each analyzed 100 samples for F2 (c.*97G⬎A), F5 in parallel by use of programmable thermal cyclers.
(c.1601G⬎A), and MTHFR (c.665C⬎T and c.1286A⬎C) These methods are typically restricted to a single pro-
after blinding for genotype and genotype proportions. tocol, placing constraints on assay design. A DNA sam-
Internal temperature controls improved melting curve ple may need to be processed multiple times to accom-
precision. The platform’s liquid-handling system auto- modate varying protocols for a panel of tests. As a
mated PCR and HSM. result, consumables, turnaround times, and laboratory
errors all increase, thus decreasing throughput. In ad-
RESULTS: PCR and HSM were completed in a total of
dition, some protocols require additional reflex testing
12.5 min. Melting was performed at 0.5 °C/s. As ex- based on the initial result outcome. This typically re-
pected, homozygous variants were separated by melt- quires more processing that may require additional
ing temperature, and heterozygotes were identified by batching and delay. Some molecular tests with a lim-
curve shape. All samples were correctly genotyped by ited number of targets use single-use individual car-
the instrument. Follow-up testing was required on tridges. A few can be rapidly performed, but at a cost
1.38% of the assays for a definitive genotype. that may not be acceptable to the laboratory.
Fast molecular testing usually requires specialized
sample containers. For example, rapid-cycle PCR and
CONCLUSIONS: We demonstrate genotyping accuracy on
high-resolution melting can be performed quickly in
a novel microfluidic platform with rapid serial PCR
capillaries (2 ), but individual capillaries are difficult to
and HSM. The platform targets short turnaround handle in high numbers.
times for multiple genetic variants in up to 8 samples. It We have developed an integrated microfluidic
is also designed to allow automatic and immediate reflex- platform with serial, rapid-cycle PCR and high-speed
ive or repeat testing depending on results from the melting (HSM)5 analysis. Sequential discrete reactions
streaming DNA. Rapid serial PCR provides a flexible ge- within microfluidic channels allow variable thermal
netic work flow and is nicely matched to HSM analysis. cycling and melting parameters for each reaction. The
© 2014 American Association for Clinical Chemistry
volume scale-down of microfluidics enables rapid heat
transfer for faster thermal cycling (3 ) as well as rapid

1
Department of Pathology, University of Utah Health Sciences Center, Salt Lake City, UT; Received February 22, 2014; accepted July 17, 2014.
2
current address: Canon U.S. Life Sciences, Newport News, VA; 3 Canon U.S. Life Previously published online at DOI: 10.1373/clinchem.2014.223768
Sciences, Rockville, MD; 4 Department of Mathematics, Utah Valley University, Orem, UT. © 2014 American Association for Clinical Chemistry
5
* Address correspondence to this author at: Department of Pathology, University Nonstandard abbreviations: HSM, high-speed melting; ARUP, Associated Re-
of Utah Medical School, 50 N Medical Drive, Salt Lake City, UT 84132. Fax gional and University Pathologists; Tm, melting temperature; LED, light-emitting
801-581-6001; e-mail [email protected]. diode; HRM, high-resolution melting.

1306
Microfluidic Genotyping by High-Speed Melting Analysis

melting rates for HSM. HSM, a faster variant of high-

resolution melting, provides a simple method to ana- Table 1. DNA duplex length, guanine-cytosine (GC)
lyze PCR products by use of a saturating fluorescent content, and measured Tm values.
dye to detect sequence variation. HSM analysis mini-
mizes time and cost after PCR (1 ), allowing genotyping GC
Length, content, Tm, ⌬Tm,
(4 ), variant scanning (5 ), and other applications DNA duplex bp % °Ca °Cb
(6, 7 ). In the present study, we demonstrate the geno-
Ultraconserved sequence 81 44 80.5
typing accuracy of a custom prototype instrument that
analyzes up to 8 samples in parallel by small amplicon F2 c.*97G⬎A 48 50 78.8 0.8
genotyping (8 ) and HSM. F5 c.1601G⬎A 43 49 76.6 0.8
MTHFR c.665C⬎T 48 52 79.6 0.9
Materials and Methods MTHFR c.1286A⬎C 46 46 76.6 ⫺1.1
Internal temperature control 45 71 85.0
CLINICAL SAMPLES Low Tm calibrator 100 15 69.9
We obtained previously genotyped and deidentified High Tm calibrator 200 61 90.0
(University of Utah Institutional Review Board #7275)
blood samples from Associated Regional and Univer- a
Measured Tm of the wild-type sequence.
b
sity Pathologists (ARUP) Laboratories. After enrich- Tm of the wild-type minus the Tm of the homozygous variant.
ment for rare F5 [coagulation factor V (proaccelerin,
labile factor)] homozygotes,6 F2 [coagulation factor II
(thrombin)] heterozygotes, and F2 homozygotes, quired for future system use. Control DNA, wild-type
DNA was extracted (Gentra Puregene Blood Kit, Qia- at F2, F5, and the MTHFR loci, was extracted from the
gen) and diluted to 200 ng/␮L with 10 mmol/L Tris, pH GM11254 cell line (Coriell) by column absorption
8.0, 0.1 mmol/L EDTA with A260 (Nanodrop 1000, (Blood and Cell Culture DNA Kit, Qiagen). We de-
Thermo Scientific). We selected 105 samples with an signed primers for high-temperature (90.0 °C) and
A260/A280 ratio between 1.7 and 2.0, genotyped them by low-temperature (69.9 °C) calibration of the instru-
high-resolution melting assays on commercial instru- ment to amplify artificial plasmid sequences not anal-
ments, and used ⱖ1 of the following: small amplicon ogous to the human genome. Primers, controls, and
genotyping (8 ), unlabeled probe genotyping (9 ), or calibrator sequences were synthesized, HPLC purified,
snapback primer genotyping (10 ). Samples were then and resuspended in 10 mmol/L Tris, 0.1 mmol/L EDTA
blinded into two 100-sample lots, with 1 lot tested at at pH 8.0 (Integrated DNA Technologies). DNA du-
Canon US Life Sciences (Rockville, MD) and the other plex length, guanine-cytosine content, and measured
at the University of Utah (Salt Lake City, UT). melting temperature (Tm) values are shown in Table 1,
and primer and control sequences are given in Supple-
DNA CONTROLS AND OLIGONUCLEOTIDES mental Table 1, which accompanies the online version
We designed primers for 4 small amplicon genotyping of this article at http://www.clinchem.org/content/
assays to detect common single nucleotide variants as- vol60/issue10.
sociated with thrombophilia: F2 c.*97G⬎A (11 ), F5
c.1601G⬎A (12 ), and MTHFR [methylenetetrahydro- RAPID-CYCLE PCR AND HSM ANALYSIS
folate reductase (NAD(P)H)] c.665C⬎T (13 ) and PCR was performed within microchannels that con-
c.1286A⬎C (14 ). An internal temperature control of tained approximately 50-nL reaction volumes. The re-
complementary oligonucleotides was included in each actions included 1⫻ Titanium威 Taq DNA Polymerase
assay for better temperature precision (15 ). An ultra- (Clontech), 3.0 mmol/L MgCl2, 50 mmol/L Tris-HCl
conserved element (16 ) on chromosome 17 was also (pH 8.0), 0.04% Tween威 20 (Thermo Fisher Scien-
amplified and melted in all channels 1 time before the tific), 1.0 mol/L betaine monohydrate (Sigma-
panel of 4 loci in this study. It was used as an additional Aldrich), 2% DMSO, 50 mmol/L KCl, 0.2% BSA, 375
control during instrument development but is not re- ␮mol/L of each dNTP, 1⫻ LCGreen威 Plus (BioFire
Diagnostics), 1.0 ␮mol/L of each test primer, 0.5
␮mol/L internal temperature duplex, and 18 ng/␮L
human genomic DNA. The ultraconserved element
6
Human genes: F5, coagulation factor V (proaccelerin, labile factor); F2, coag-
PCR did not include the internal temperature control
ulation factor II (thrombin); MTHFR, methylenetetrahydrofolate reductase duplex. The PCR protocol included an initial denatur-
(NAD(P)H); PAH, phenylalanine hydroxylase; CFTR, cystic fibrosis transmem- ation step of 95 °C for 30 s followed by 40 cycles of
brane conductance regulator (ATP-binding cassette sub-family C, member 7);
BRCA1, breast cancer 1, early onset; KRAS, Kirsten rat sarcoma viral oncogene 95 °C for 1.5 s (50 °C/s ramp rate), 60 °C for 2.5 s
homolog; BRAF, v-raf murine sarcoma viral oncogene homolog B. (20 °C/s ramp rate), and 72 °C for 2.5 s (2 °C/s ramp

Clinical Chemistry 60:10 (2014) 1307

 Fig. 1. (A), Photograph of the fully assembled 8-channel microfluidic cartridge containing the core glass chip; (B),
Schematic of the glass chip labeled with primary components.

rate). After PCR, a denature/renature step was imple- station that was used to house the pipette tips and foam
mented at 95 °C for 1.5 s (50 °C/s ramp rate) and 50 °C rollers that removed excess fluid from the pipette tips
for 2 s (20 °C/s ramp rate) before melting the product. in between fluid deliveries (Fig. 2).
The product was then melted from 65 °C to 95 °C with Two fluid types, either the PCR reagents or the
a 0.5 °C/s ramping rate to produce HSM curves, thus blanking solution, are delivered to 8 fused silica capil-
providing a temperature resolution of approximately laries (Polymicro Technologies) assembled within the
0.02 °C. Melting curves were interpreted with custom cartridge by use of 2 different liquid-handling robots.
high-resolution melting software (17 ). Briefly, the flu- One robot and set of pipette tips is dedicated for PCR
orescence melting curve data was first extracted by re- reagents, whereas the other delivers blanking solution.
moving the exponential background. Then, the nega- First, the blanking solution containing 35 ␮mol/L Al-
tive derivative of the melting curve was obtained by exa 647 fluorescent dye is delivered to the cartridge (see
differentiating the second-degree polynomial best fit online Supplemental Fig. 2A). The solution moves into
with Savitzky–Golay moving windows (18 ). The tem- the cartridge and through the microchannels in the
perature control peak maximum (Tm) was used to glass chip, where it is optically monitored via illumina-
overlay each curve so that the control Tm was translated tion with a red light-emitting diode (LED). Further-
to the mean control Tm of all curves. Homozygous more, the fluid motion is controlled by detecting the
samples were differentiated by Tm, by use of knowledge fluorescent edge of the blanking solution and giving
of the measured Tm difference outlined in Table 1 and feedback to the peristaltic pumps (Watson Marlowe) to
the wild-type control as a reference point along with control and confirm the solution position. Every other
the aid of clustering software (19 ) to help define wild- fluid delivery to the cartridge contains this blanking
type and homozygous variant groupings. Heterozy- solution, whereas alternating deliveries contain the re-
gous samples were distinguished by shape differences, agents for PCR/HSM (see online Supplemental Fig. 2).
caused by the presence of heteroduplexes. This delivery pattern forms the basis of a serial PCR
stream where PCR reagents with specific primers are
MICROFLUIDIC PLATFORM AND WORK FLOW delivered to the cartridge in a sequential fashion. These
Before starting the instrument run, DNA samples were specific reactions on the same DNA sample are sepa-
mixed with the PCR reagents and placed on the 384- rated by blanking solution.
well plate, as indicated in online Supplemental Fig. 1. A maximum of 8 different samples are used in a
The calibrator was also placed on the plate as indicated, run within the same cartridge (1 DNA sample/chan-
and the blanking solution was pipetted onto the car- nel). Cartridge heater calibration is performed at the
tridge. The cartridge has 2 components, a “world-to- beginning of the run by melting a mixture of high- and
chip” poly(methylmethacrylate) plastic interface that low-Tm preamplified products as calibrators (Table 1).
surrounds a glass chip component where PCR and Then, the ultraconserved element PCR is (optionally)
HSM are performed (Fig. 1). These consumables were performed as a control. This is followed by multiple
placed in the instrument drawer together with a tip serial tests, where each locus to be genotyped is sequen-

1308 Clinical Chemistry 60:10 (2014)

Microfluidic Genotyping by High-Speed Melting Analysis

Fig. 2. Prototype instrument.

(A), Schematic shows the liquid handling robots, a 384-reagent plate, a microfluidic cartridge, and an optical imaging system. The
components are controlled and coordinated by custom LabVIEW software (National Instruments). The automated pipette system has
2 independent pipette heads that use linear actuators (Oriental Motor) to move custom pipette tips. One pipette head handles PCR
reagents and moves between a 384-well microtiter plate (Greiner Bio-One), a tip station, and the microfluidic cartridge. The tip station
contains custom pipette tips and a disposal area. The other pipette head handles a blanking solution, moving between different
locations on the microfluidic cartridge. Syringe pumps with 50-␮L capacity (TriContinent) are used for each pipette tip to aspirate and
dispense fluids. Three microliters are aspirated from the 384-well microtiter plate, with 2 ␮L dispensed on the end of the tip forming
a hemispherical droplet of reagent. This droplet is brought in contact with the capillary tubes on the cartridge to introduce the reagent
to the microfluidic cartridge. (B), Photograph of the assembled prototype. (C), Photograph of the inside of the prototype. Inside the
drawer from left to right are the microtiter plate, the tip station, and the microfluidic cartridge.

tially PCR amplified and melted. In our experimental Optics), positioned 45° incident to the glass chip on the
setup, the 2 outer channels in each run were negative fluidic outlet side, with a 632DF22 excitation filter
controls without genomic DNA, and 1 of the remain- (Semrock) to illuminate the fluorescent dyes within the
ing 6 channels contained wild-type control DNA. Each glass chip. Emitted fluorescent light was collected in an
of the microfluidic channels has individual heat con- EF 50-mm F/1.4 USM lens (Canon U.S.A.), passed
trol through thin-film platinum heaters embedded un- through a 510DF80/685DF50 dual bandpass filter
der each sample channel in the glass chip. The heaters (Semrock), and imaged, with an acquisition rate of 30
are isolated from the fluid within the microchannel by Hz for HSM and 1 Hz for other system operations, by
a passivation layer of SiO2. The glass chip design in- use of a complementary metal– oxide semiconductor
cludes gold metal traces that connect to flex circuits on sensor (EOS 5D Mark II, Canon U.S.A.). Heaters used
the cartridge and interface with electrical components proportional/integral/derivative loop control with
on the instrument to enable temperature control. Fans pulse width modulation. The thin-film platinum heat-
in the instrument move air across a heat sink attached ers in the microfluidic chip served as both heaters and
to the glass chip to aid in the passive cooling of the temperature sensors.
platinum heaters during PCR. Additional technical The microfluidic design combined with individual
specifications have been described (20 ). heating allows for rapid-cycle PCR (see online Supple-
The optics to monitor fluid control and perform mental Fig. 3). We used a PCR protocol with 40 cycles
HSM were matched to LCGreen Plus and Alexa 647 of 15 s each. This protocol, coupled with an HSM pro-
fluorescent dyes. We used a 445-nm blue LED (Inno- tocol with a 0.5 °C/s ramp rate from 65 °C to 95 °C,
vation In Optics), positioned 45° incident to the glass allowed a single PCR/HSM test to be completed in
chip on the fluidic inlet side, with a 438DF24 excitation about 12.5 min, not including time for fluid delivery to
filter (Semrock) and a 629-nm red LED (Innovation In the chip. Performing the panel of 5 PCRs used here (1

Clinical Chemistry 60:10 (2014) 1309

 Fig. 3. Example melting curves for 6 samples at each locus clustered by genotype.
The Human Genome Variation Society nomenclature for the variants are (A), F2 (c.*97G⬎A); (B), F5 (c.1601G⬎A); (C), MTHFR
(c.665C⬎T); and (D), MTHFR c.1286A⬎C. The legacy names for these variants are F2 20210G⬎A, F5 1691G⬎A, MTHFR
677C⬎T, and MTHFR 1298A⬎C. The negative derivative melting plots are displayed after exponential background removal,
normalization, and temperature overlay with the internal temperature control (ITC). The black curves are wild-type, the blue
curves are homozygous variant, and the red curves are heterozygous variant. The y axis is the negative derivative of fluorescence
with respect to temperature (– dF/dT ).

control and 4 genotyping loci) currently requires 2.5 h, channels decreased from 0.18 °C before temperature
including fluid delivery, system initialization, and cal- correction to 0.07 °C after temperature correction, al-
ibration, steps that have not yet been optimized for lowing good separation of both homozygotes in all as-
speed. At the end of the run, the instrument generates a says. The entire process (PCR, HSM, and melting curve
data output file that contains fluorescence vs tempera- display) required 12.5 min per assay. Representative
ture information. This file is then analyzed for visual melting curves, shown as their negative derivatives, are
genotyping (Fig. 3). displayed in Fig. 3.
As expected, different homozygotes were sepa-
Results rated by Tm, and heterozygotes had a second low tem-
perature peak consisting of heteroduplexes (8 ). An addi-
The small amplicon genotyping products were 43– 48 tional melting transition from the internal temperature
bp, with experimental Tm values of 76.6 – 80.5 °C (Ta- control at 85.0 °C was used to overlay control Tm values
ble 1). The temperature difference between the 2 ho- across samples to increase temperature precision. The
mozygous genotypes (⌬Tm) at each locus varied from minimum difference in temperature between homozy-
0.8 to 1.1 °C as expected for class 1 and class 2 single gous genotypes at any locus was 0.8 °C compared with
base variants (8 ). The synthetic internal temperature a peak SD of 0.07 °C, enough separation that genotyp-
control used to increase temperature precision melted ing errors between homozygotes should be very rare.
5.4 °C higher than any genotype at the 4 loci and did Genotyping accuracy was evaluated in 2 blinded
not interfere with genotyping. The mean Tm SD across studies. In each study, 100 samples were selected from

1310 Clinical Chemistry 60:10 (2014)

Microfluidic Genotyping by High-Speed Melting Analysis

ysis of the stock DNA. Finally, 1 sample had a melting

Table 2. Number of clinical samples by locus and curve between wild-type and heterozygous curves and
genotype that were correctly genotyped without could not be genotyped. Subsequent targeted sequencing
repeats or additional analysis at Canon US Life of its stock DNA revealed 80%–90% C allele and 10%–
Sciences. 20% T allele (data not shown), suggesting a mixed DNA
sample or product contamination. Overall, correct geno-
Heterozygote Homozygote
Target Wild-type mutant mutant
types were obtained on the first pass 98.63% of the time,
with 0.75% requiring 1 repeat and 0.63% of samples com-
F2 c.*97G⬎A 79/79 20/20 1/1 promised by exchange or mixing.
F5 c.1601G⬎A 75/75 22/22 3/3
MTHFR c.665C⬎T 54/56a,b 39/40b 4/4 Discussion
MTHFR c.1286A⬎C 23/23 53/53 24/24
a
We used a prototype microfluidic instrument for
One sample had a melting curve between wild-type and heterozygous
genotypes that precluded genotyping. Sequencing of the purified DNA
rapid-cycle, serial PCR and HSM to genotype 100 sam-
stock of this sample showed about 90% C allele and 10% T allele ples at 2 laboratories (800 analyzed loci) with 100%
(presumably resulting from DNA or amplicon contamination). The melting instrument accuracy. The platform was designed to
curves of the other 3 loci were successfully genotyped on the first run. perform multiple sequential assays on 8 different sam-
b
Two samples of MTHFR c.665C⬎T showed nonspecific amplification in
their melting curves that precluded genotyping on the initial run. These
ples (1 sample/channel). Within the 8-sample limit, the
samples were repeated once and correctly genotyped. number of samples and positive and negative controls
is flexible, although running ⬎8 samples requires ad-
ditional microfluidic cartridges and runs. For each se-
the 105 available DNA samples and analysis was per- rial assay, different PCR protocols can be run and the
formed at either Canon US Life Sciences (Table 2) or number of genotyping assays can be increased to at
the University of Utah (Table 3) for a total of 800 ge- least 25 on existing instrumentation.
notype analyses. In both blinded studies, no genotyp- Rapid-cycle PCR was introduced in 1990 (21–23 )
ing errors could be attributed to the instrument or and defined as 30 cycles of PCR in 10 –30 min. The
HSM (100% accuracy). However, the melting curves of microfluidic instrument presented here performs near
6 assays (0.75%) were not definitive on the initial run the fast end of this range. Furthermore, with the micro-
and needed to be repeated once to verify genotype. In fluidics on this instrument, we have initial evidence
addition, 2 samples next to each other in the sample that the time for PCR can be reduced even further.
tray were manually exchanged by mistake, resulting in High-resolution melting (sometimes abbreviated
4 discrepant genotypes. This was verified by repeat anal- HRM) was introduced in 2003 (4, 24 ). Accurate geno-
typing of the thrombophilia targets presented here has
been reported on more conventional instruments
(8, 25 ). The maximum speeds reported to date for
Table 3. Number of clinical samples by locus and
high-resolution melting are 0.3 °C/s (26, 27 ). The pro-
genotype that were correctly genotyped without
totype presented here routinely melts at 0.5 °C/s, 67%
repeats or additional analysis at the University of
faster than previously reported rates. Because we have
Utah.
observed improvements in heterozygote detection at
Heterozygote Homozygote
faster rates, we are introducing the new term, high-
Wild-type mutant mutant speed melting, and its abbreviation, HSM, to differen-
F2 c.*97G⬎A 78/78 20/20 2/2
tiate faster methods from slower instruments that
a,b a,c
typically perform high-resolution melting at around
F5 c.1601G⬎A 72/75 20/22 3/3
0.01– 0.04 °C/s (27 ). The prototype presented here com-
MTHFR c.665C⬎T 55/56a 39/40a 4/4 bines and extends the speed advantages of both rapid-
MTHFR c.1286A⬎C 24/24 51/52c 24/24 cycle PCR and high-resolution melting. The system is tar-
a
Genotypes at F5 c.1601G⬎A and MTHFR c.665C⬎T of 2 neighboring DNA
geted toward analyzing a few DNA samples at multiple
samples appeared to be exchanged on analysis of the first run. After repeat variant loci, not parallel analysis on many DNA samples.
analysis using the stock DNA samples, the initial exchange (which most likely Use of internal temperature controls decreased the
occurred during manual loading of the 384-well plate) was confirmed.
b
Tm SD from 0.18 to 0.07 °C, allowing wild-type and
Two samples did not amplify well at F5 c.1601G⬎A (low signal strength),
precluding genotyping on the initial run. These samples were repeated homozygous genotypes to be easily distinguished, sim-
once and correctly genotyped. ilar to other studies (15 ). The Tm range of the 4 single
c
One F5 c.1601G⬎A and 1 MTHFR c.1286A⬎C could not be evaluated on nucleotide polymorphisms studied here was 76.6 –
the initial run because of irregular fluid flow in one cartridge channel.
These samples were repeated once and correctly genotyped.
79.6 °C, with the internal temperature control well out-
side this range at 85.0 °C to prevent interference. For

Clinical Chemistry 60:10 (2014) 1311

amplicons with a Tm near to this internal control, a nal transduction gene pathways associated with preva-
second internal control with lower Tm could be used. lent mutations in colorectal cancers to determine eligi-
Many patients could benefit from more rapid genetic bility for anti– epidermal growth factor receptor
test results. One of the more compelling reasons for easy therapy. Tumor mutation analysis can first look for the
access and reduced turnaround time for genetic testing is most prevalent mutations in KRAS (Kirsten rat sar-
posed by variant-specific drug treatments such as iva- coma viral oncogene homolog) codons 12/13 and then
caftor for some cystic fibrosis variants, including reflex to the next likely BRAF (v-raf murine sarcoma
Gly551Asp (28 ). Early trials have shown improved lung viral oncogene homolog B) V600E (37 ).
function, suggesting that lung damage may be minimized Several future enhancements can be envisioned to
if patients who bear the targeted variant can be identified advance system function and performance. Genotyp-
at birth and treatment initiated. ing can be automated in software rather than the visual
Another example is illustrated by the current prac- calls used here. The Tm differences between wild-type
tice of diagnosing inborn errors of metabolism such as and homozygous variant products are large enough
medium-chain acyl coenzyme A dehydrogenase defi- (0.8 –1.1 °C) and the precision good enough (SD
ciency. This genetic disorder occurs in approximately 1 0.07 °C), that Tm values can be used for definitive geno-
in 15 000 births in the US population (29 ) and is a typing. The pipette system can be augmented to make
low-volume test for laboratories equipped to perform reagent setup completely automatic, reducing time and
it. Initial screening by mass spectrometry is confirmed operator error. Reduced time for reagent mixing
with genotyping. Because treatment to avoid severe sei- should also limit nonspecific amplification, and the
zures, coma, or death consists of simply changing to a amount of DNA can be decreased. Improving thermal
low-fat diet and avoiding periods of fasting, confirm- insulation around the microfluidic chip will improve
ing newborns with this disease before hospital dis- performance by limiting temperature gradients across
charge can be lifesaving (30 ). channels, especially the outer channels. Newly de-
Finally, rapid identification of variants that affect signed chips with additional heaters on the outside of
drug metabolism or confer drug resistance allows for channels 1 and 8 can reduce the thermal gradient ob-
better patient treatment. Identification of slow vs rapid served, and fewer no-template controls can be used.
metabolizers for cytochrome P-450 can be critical to Better temperature precision will improve the resolu-
achieve the correct dose of clopidogrel for treatment of tion between wild-type and homozygous samples,
percutaneous coronary syndromes (31 ) and may im- hopefully allowing the discrimination of class 3 and 4
prove dosing of warfarin treatment (32 ). variants (8 ) and difficult small insertions/deletions. Al-
The lack of genetic testing at local laboratories also ternatively, unlabeled probes (9 ) or snapback primers
prevents identification of individuals affected with rare (10 ) can be used to genotype difficult variants and can
diseases during an opportune moment for counseling. be performed on the same platform.
Recently Evans et al. (33 ) argued that identification of In conclusion, serial PCR and HSM improve work
individuals carrying rare mutations usually leads to fa- flow for low-volume genetic tests. When combined with
milial evaluation, identifying carriers and additional rapid PCR and HSM, turnaround time is reduced and
affected individuals, thereby increasing the overall ben- multiple tests can be performed on the same samples. The
efit. As new genetic information becomes available, the platform also has the potential for customized repeat and
ability to rapidly develop and validate additional ge- reflexive testing to minimize sample reprocessing and
netic tests on a fast serial platform is attractive. generation of extraneous data.
An additional benefit of serial PCR is automatic
reflexing to additional assays depending on the results
of a preceding test. This can provide “smart” through- Author Contributions: All authors confirmed they have contributed to
put while eliminating generation of extraneous infor- the intellectual content of this paper and have met the following 3 re-
mation. Such a simplified work flow does not require quirements: (a) significant contributions to the conception and design,
additional setup. For example, exon scanning with acquisition of data, or analysis and interpretation of data; (b) drafting
or revising the article for intellectual content; and (c) final approval of
HSM can identify the presence of variants, followed by the published article.
reflex genotyping for the most common variants in the
region of the positive scan. Such an approach has been Authors’ Disclosures or Potential Conflicts of Interest: Upon man-
uscript submission, all authors completed the author disclosure form.
successfully demonstrated for PAH (phenylalanine hy- Disclosures and/or potential conflicts of interest:
droxylase) (34 ), CFTR [cystic fibrosis transmembrane
conductance regulator (ATP-binding cassette sub- Employment or Leadership: S. Sundberg, Canon US Life Sciences; C.T.
Wittwer, BioFire Diagnostics and Clinical Chemistry, AACC; R. Howell,
family C, member 7)] (1 ), and BRCA1 (breast cancer 1, Canon US Life Sciences; J. Huuskonen, Canon US Life Sciences; H.
early onset) (35 ). Also, reflex testing was used by Stiles, Canon US Life Sciences; I.T. Knight, Canon US Life Sciences.
Guedes et al. (36 ) by scanning multiple regions in sig- Consultant or Advisory Role: None declared.

1312 Clinical Chemistry 60:10 (2014)

 Microfluidic Genotyping by High-Speed Melting Analysis

Stock Ownership: None declared. Other Remuneration: S. Sundberg, travel between University of
Honoraria: None declared. Utah and Rockville, MD; H. Stiles, Canon US Life Sciences.
Research Funding: Grant from Canon US Life Sciences to the Uni-
Role of Sponsor: The funding organizations played a direct role in
versity of Utah, C. Wittwer, PI.
the review and interpretation of data.
Expert Testimony: None declared.
Patents: S. Sundberg, application no. PCT/US2011/050104; C.T. Acknowledgments: The authors thank ARUP laboratories for pro-
Wittwer, patent no. 7803551; R. Palais, patent nos. 8068992 B2 and viding the deidentified samples and Ling Xu, Rob Troyan, and Sand-
8296074 B2; I.T. Knight, patent nos. US7915030B2, US20110262316A1, hya Patel from Canon US Life Sciences for providing technical
and US20120288865A1. assistance.

References
1. Zhou L, Palais RA, Ye F, Chen J, Montgomery JL, artery disease. Circulation 1997;96:2573–7. scanning and genotyping. Clin Chem 2007;53:
Wittwer CT. Symmetric snapback primers for 14. Weisberg I, Tran P, Christensen B, Sibani S, Rozen 1544 – 8.
scanning and genotyping of the cystic fibrosis R. A second genetic polymorphism in methyl- 28. O’Reilly R, Elphick HE. Development, clinical util-
transmembrane conductance regulator gene. Clin enetetrahydrofolate reductase (MTHFR) associ- ity, and place of ivacaftor in the treatment of
Chem 2013;59:1052– 61. ated with decreased enzyme activity. Mol Genet cystic fibrosis. Drug Des Devel Ther 2013;7:929 –
2. Wittwer CT, Rasmussen RP, Ririe KM. Rapid PCR Metab 1998;64:169 –72. 37.
and melting analysis. In: Bustin SA, ed. The PCR 15. Seipp MT, Durtschi JD, Liew MA, Williams J, 29. Andresen BS, Dobrowolski SF, O’Reilly L, Muen-
revolution: basic technologies and applications. Damjanovich K, Pont-Kingdon G, et al. Unlabeled zer J, McCandless SE, Frazier DM, et al. Medium-
New York: Cambridge University Press 2010;48 – oligonucleotides as internal temperature controls chain acyl-CoA dehydrogenase (MCAD) muta-
69. for genotyping by amplicon melting. J Mol Diagn tions identified by MS/MS-based prospective
3. Zhang C, Xing D. Miniaturized PCR chips for 2007;9:284 –9. screening of newborns differ from those observed
nucleic acid amplification and analysis: latest ad- 16. Bejerano G, Pheasant M, Makunin I, Stephen S,
in patients with clinical symptoms: identification
vances and future trends. Nucleic Acids Res 2007; Kent WJ, Mattick JS, Haussler D. Ultraconserved
and characterization of a new, prevalent muta-
35:4223–37. elements in the human genome. Science 2004;
tion that results in mild MCAD deficiency. Am J
4. Wittwer CT, Reed GH, Gundry CN, Vandersteen 304:1321–5.
Hum Genet 2001;68:1408 –18.
JG, Pryor RJ. High-resolution genotyping by am- 17. Palais R, Wittwer CT. Mathematical algorithms
30. Roe CR DJ. Mitochondrial fatty acid oxidation
plicon melting analysis using LCGreen. Clin Chem for high-resolution DNA melting analysis. Meth-
disorders. In: Scriver CR SW, Childs B, Beaudet
2003;49:853– 60. ods Enzymol 2009;454:323– 43.
5. Reed GH, Wittwer CT. Sensitivity and specificity 18. Press W, Teukolsky S, Vetterling W, Flannery B. AL, Valle D, Kinzler KW, Vogelstein B, eds. The
of single-nucleotide polymorphism scanning by Salvitzky-Golay Smoothing filters numerical metabolic bases of inherited disease. New York:
high-resolution melting analysis. Clin Chem 2004; recipes: the art of scientific computing. New York: McGraw-Hill 2001;2297–326.
50:1748 –54. Cambridge University Press 2007;766 –72. 31. Zhang L, Cui G, Li Z, Wang H, Ding H, Wang DW.
6. Vossen RH, Aten E, Roos A, den Dunnen JT. 19. Zhou L, Wang L, Palais R, Pryor R, Wittwer CT. Comparison of high-resolution melting analysis,
High-resolution melting analysis (HRMA): more High-resolution DNA melting analysis for simul- TaqMan allelic discrimination assay, and Sanger
than just sequence variant screening. Hum Mutat taneous mutation scanning and genotyping in sequencing for clopidogrel efficacy genotyping in
2009;30:860 – 6. solution. Clin Chem 2005;51:1770 –7. routine molecular diagnostics. J Mol Diagn 2013;
7. Montgomery JL, Sanford LN, Wittwer CT. High- 20. Coursey JS, Hasson KC. Compound calibrator for 15:600 – 6.
resolution DNA melting analysis in clinical re- thermal sensors. US Patent Application 2012; 32. Pirmohamed M, Burnside G, Eriksson N, Jor-
search and diagnostics. Expert Rev Mol Diagn 20120051390 A1. gensen AL, Toh CH, Nicholson T, et al. A random-
2010;10:219 – 40. 21. Wittwer CT, Fillmore GC, Garling DJ. Minimizing ized trial of genotype-guided dosing of warfarin.
8. Liew M, Pryor R, Palais R, Meadows C, Erali M, the time required for DNA amplification by effi- N Engl J Med 2013;369:2294 –303.
Lyon E, Wittwer C. Genotyping of single- cient heat transfer to small samples. Anal 33. Evans JP, Berg JS, Olshan AF, Magnuson T, Rimer
nucleotide polymorphisms by high-resolution Biochem 1990;186:328 –31. BK. We screen newborns, don’t we? Realizing the
melting of small amplicons. Clin Chem 2004;50: 22. Wittwer CT, Garling DJ. Rapid cycle DNA promise of public health genomics. Genet Med
1156 – 64. amplification: time and temperature optimiza- 2013;15:332– 4.
9. Zhou L, Myers AN, Vandersteen JG, Wang L, tion. Biotechniques 1991;10:76 – 83. 34. Dobrowolski SF, Ellingson C, Coyne T, Grey J,
Wittwer CT. Closed-tube genotyping with unla- 23. Wittwer CT, Hahn M, Kaul K. Rapid cycle real- Martin R, Naylor EW, et al. Mutations in the
beled oligonucleotide probes and a saturating time PCR—methods and applications: quantifica- phenylalanine hydroxylase gene identified in 95
DNA dye. Clin Chem 2004;50:1328 –35. tion. Berlin: Springer 2004. patients with phenylketonuria using novel sys-
10. Zhou L, Errigo RJ, Lu H, Poritz MA, Seipp MT, 24. Gundry CN, Vandersteen JG, Reed GH, Pryor RJ,
tems of mutation scanning and specific genotyp-
Wittwer CT. Snapback primer genotyping with Chen J, Wittwer CT. Amplicon melting analysis
ing based upon thermal melt profiles. Mol Genet
saturating DNA dye and melting analysis. Clin with labeled primers: a closed-tube method for
Metab 2007;91:218 –27.
Chem 2008;54:1648 –56. differentiating homozygotes and heterozygotes.
35. van der Stoep N, van Paridon CD, Janssens T,
11. Poort SR, Rosendaal FR, Reitsma PH, Bertina RM. A Clin Chem 2003;49:396 – 406.
Krenkova P, Stambergova A, Macek M, et al.
common genetic variation in the 3⬘-untranslated 25. Seipp MT, Pattison D, Durtschi JD, Jama M,
Diagnostic guidelines for high-resolution melting
region of the prothrombin gene is associated with Voelkerding KV, Wittwer CT. Quadruplex geno-
elevated plasma prothrombin levels and an in- typing of F5, F2, and MTHFR variants in a single curve (HRM) analysis: an interlaboratory valida-
crease in venous thrombosis. Blood 1996;88: closed tube by high-resolution amplicon melting. tion of BRCA1 mutation scanning using the 96-
3698 –703. Clin Chem 2008;54:108 –15. well LightScanner. Hum Mutat 2009;30:899 –909.
12. Bertina RM, Koeleman BP, Koster T, Rosendaal 26. Herrmann MG, Durtschi JD, Bromley LK, Wittwer 36. Guedes JG, Veiga I, Rocha P, Pinto P, Pinto C,
FR, Dirven RJ, de Ronde H, van der Velden PA, CT, Voelkerding KV. Amplicon DNA melting anal- Pinheiro M, et al. High resolution melting analysis
Reitsma PH. Mutation in blood coagulation factor ysis for mutation scanning and genotyping: cross- of KRAS, BRAF and PIK3CA in KRAS exon 2
V associated with resistance to activated protein platform comparison of instruments and dyes. wild-type metastatic colorectal cancer. BMC Can-
C. Nature 1994;369:64 –7. Clin Chem 2006;52:494 –503. cer 2013;13:169.
13. Kluijtmans LA, Kastelein JJ, Lindemans J, Boers 27. Herrmann MG, Durtschi JD, Wittwer CT, Voelk- 37. Pritchard CC, Grady WM. Colorectal cancer mo-
GH, Heil SG, Bruschke AV, et al. Thermolabile erding KV. Expanded instrument comparison of lecular biology moves into clinical practice. Gut
methylenetetrahydrofolate reductase in coronary amplicon DNA melting analysis for mutation 2011;60:116 –29.

Clinical Chemistry 60:10 (2014) 1313

View publication stats