Multiplex Amplification Coupled with COLD-PCR and High Resolution Melting Enables Identification of Low-Abundance Mutations in Cancer Samples with Low DNA Content

The Journal of Molecular Diagnostics, Vol. 13, No. 2, March 2011 Copyright © 2011 American Society for Investigative Pathology and the Association for Molecular Pathology. Published by Elsevier Inc. All rights reserved. DOI: 10.1016/j.jmoldx.2010.10.008 Multiplex Amplification Coupled with COLD-PCR and High Resolution Melting Enables Identification of Low-Abundance Mutations in Cancer Samples with Low DNA Content

Coren A. Milbury,* Clark C. Chen,*† prove mutation detection sensitivity via COLD-PCR Harvey Mamon,* Pingfang Liu,* amplification. (J Mol Diagn 2011, 13:220–232; DOI: Sandro Santagata,‡ and G. Mike Makrigiorgos* 10.1016/j.jmoldx.2010.10.008) From the Department of Radiation Oncology,* Dana-Farber/Brigham and Women’s Cancer Center, Harvard A vast collection of markers have been identified as im- Medical School, Boston; the Division of Neurosurgery,† portant determinants of cancer type, treatment course, Beth–Israel Deaconess Medical Center, Boston; and the acquired resistance, and survival. However, the ability to Department of Pathology,‡ Brigham and Women’s Hospital, screen for an extensive panel of markers can be difficult Harvard Medical School Boston, Massachusetts when the specimen sample is low in DNA concentration or tumor abundance. For example, biopsies, fine needle aspirations, circulating tumor cells, plasma-circulating Thorough screening of cancer-specific biomarkers, DNA, buccal swab specimens, filter paper-collected such as DNA mutations, can require large amounts blood, and histological slide sections may contain impor- of genomic material; however, the amount of tant genetic markers; however, the DNA that can be genomic material obtained from some specimens retrieved from such samples may be low in concentra- (such as biopsies, fine-needle aspirations, circulat- tion. As a result, the number of biomarkers that can be ing-DNA or tumor cells, and histological slides) may analyzed is limited and the amount of information that can limit the analyses that can be performed. Further- be obtained is restricted. Whole genome amplification more, mutant alleles may be at low-abundance rel- (WGA) is often used to overcome sample limitations.1–6 ative to wild-type DNA, reducing detection ability. Although it is broadly useful, WGA often requires a min- We present a multiplex-PCR approach tailored to imum of 10 ng of DNA per reaction if one is to capture amplify targets of interest from small amounts of intratumoral heterogeneity and low-abundance events.7 precious specimens, for extensive downstream de- For target-specific applications, multiplex-PCR pre-am- tection of low-abundance alleles. Using 3 ng of DNA plification of the target of interest is an alternative that has (1000 genome-equivalents), we amplified the 1 cod- certain advantages over WGA, depending on subse- ing exons (2-11) of TP53 via multiplex-PCR. Follow- quent sample analysis (discussed herein). ing multiplex-PCR, we performed COLD-PCR (co- For downstream mutation screening, high resolution amplification of major and minor alleles at lower melting (HRM) curve analysis is simple, rapid, and inex- denaturation temperature) to enrich low-abun- pensive to perform,8–22 and exhibits high sensitivity for dance variants and high resolution melting (HRM) to screen for aberrant melting profiles. Mutation- scanning for unknown low-abundance mutations and positive samples were sequenced. Evaluation of mu- variants. The reported sensitivity of HRM is largely deter- tation-containing dilutions revealed improved sen- sitivities after COLD-PCR over conventional-PCR. Supported by grants T32-CA009078 from the NCI (C.A.M.), and NIH COLD-PCR improved HRM sensitivity by approxi- grants CA-138280 and CA-111994. mately threefold to sixfold. Similarly, COLD-PCR The contents of this manuscript are the responsibility of the authors and improved mutation identification in sequence-chro- do not necessarily represent the official views of the NIH. matograms over conventional PCR. In clinical spec- Accepted for publication October 18, 2010. imens, eight mutations were detected via conven- Supplemental material for this article can be found at http://jmd. tional-PCR-HRM, whereas 12 were detected by amjpathol.org or at doi:10.1016/j.jmoldx.2010.10.008. COLD-PCR-HRM, yielding a 33% improvement in mu- Address reprint requests to G. Mike Makrigiorgos, Dana-Farber/ tation detection. In summary, we demonstrate an ef- Brigham and Women’s Cancer Center, Brigham and Women’s Hospital, ficient approach to increase screening capabilities Level L2, Radiation Therapy, 75 Francis Street, Boston, MA 02115. E-mail: from limited DNA material via multiplex-PCR and im- [email protected]. 220 COLD-PCR and HRM Mutation Detection 221 JMD March 2011, Vol. 13, No. 2 mined by fragment length, sequence composition, muta- Pyrosequencing,30,32 real-time TaqMan analyses,33 single- tion identity, PCR quality, and equipment.8–22 Although strand conformation polymorphism,34 PCR-based mutation- recent publications report the ability to detect Ͻ1% specific restriction enzyme digestion,35 and HRM.29,36 mutant in wild-type DNA,23,24 most applications of When combined with HRM, COLD-PCR can improve muta- HRM-based assays exhibit a detection capability of tion detection and identification by approximately sixfold to approximately 5% to 10% mutant among wild-type 20-fold.29 alleles.14,17,18,24–28 In this study, we present a multiplex-PCR approach that Although HRM mutation scanning is highly sensitive can be specifically tailored to targets of interest and low- and efficient, HRM lacks the ability to identify the specific abundance alleles when DNA obtained from clinical sam- nucleotide change; this is a particularly important issue ples is limited. The multiplex-PCR amplification that we when mutations or variants are not known a priori and are present here is selected to amplify all exons and splice likely to occur at any position within the amplicon se- regions of the tumor suppressor gene TP53 in a single step. quence. Furthermore, HRM is generally more sensitive Subsequent application of COLD-PCR allows for the enrich- than conventional Sanger sequencing analysis of PCR ment of low-abundance mutations, increasing the sensitivity amplicons; as such, discrete identification of low-abun- of the assay and the accuracy of mutation detection. We dance mutations that are unknown a priori requires addi- demonstrate that subsequent screening of COLD-PCR am- tional analyses. Pyrosequencing could potentially be used plicons with HRM allows for the identification of aberrant to screen for low-abundance mutations (lower limit of 2% to melting profiles in a rapid and inexpensive manner29; down- 5%); however, it is most efficient in sequencing short reads stream sequencing of these products allows for the discrete of predefined regions or mutational hotspots. Confirmation identification of individual mutations. by single-strand conformation polymorphism, denaturing gradient gel electrophoresis, or high pressure liquid chromatography (HPLC) fractionation, followed by either band Materials and Methods selection or targeted collection of a specific peak, and final analysis via Sanger sequencing, may be performed, al- DNA and Tumor Samples though these methodologies are often time-intensive. Simi- larly, more complex approaches such as digital PCR, single Genomic DNA from cell lines with defined TP53 mutations molecule sequencing, or cloning followed by sequencing22 (T47D, SNU-182, HCC2157, MDA-MB-435, and HCC2218; could also be applied. Table 1) was purchased from American Type Culture We recently demonstrated that amplifying DNA with Collection (ATCC); the cell line SW480 was purchased COLD-PCR (co-amplification of major and minor alleles at from ATCC and genomic DNA was extracted from cul- lower denaturation temperature) before screening with tured cells. Genomic male DNA (Promega Corporation, HRM allows for the detection of low-abundance mutations Madison, WI) served as the wild-type control. with increased sensitivity and the ability to specifically iden- Clinical colorectal (N ϭ 12) and glioma tumor tissue tify the mutation via Sanger Sequencing.29 COLD-PCR specimens (N ϭ 12), snap-frozen in liquid nitrogen within is a recently-developed PCR-based approach to en- 1 to 2 hours following surgery, were obtained from the rich low-abundance mutations and minor allele vari- Massachusetts General Hospital Tumor Bank and from ants.30 COLD-PCR often reduces the limitations in low- the Division of Neurosurgery at the Beth Israel – Deacon- abundance mutation detection and identification by ess Medical Center, respectively. Paired formalin-fixed enriching unknown mutations at any position within the am- paraffin-embedded (FFPE) histology slides were also ob- plicon through the use of a critical denaturation temperature tained for the glioblastoma specimens. Patient speci- during PCR amplification. The critical denaturation temper- mens were used according to Internal Review Board ature (Tc) is lower than standard denaturation temperatures approval and standards. and preferentially denatures heteroduplexed molecules Following manual macrodissection of the tissue, genomic (those formed by hybridization of mutant and wild type DNA was isolated from the colorectal specimens using the sequences) and amplicons possessing mutations that DNeasy tissue kit (Qiagen Inc., Valencia, CA) according to lower the amplicon melting temperature (Tm), such as manufacturer’s instructions. Genomic DNA quality was as- G:CϾA:T or G:CϾT:A. Minor allele enrichment by COLD- sessed and quantified using the NanoDrop1000 spectro- PCR has been demonstrated in combination with several photometer (Thermo Scientific, Inc., Wilmington, DE); pu- downstream approaches such as Sanger sequencing,29–31 rified DNA was stored at Ϫ80°C. denaturing high-performance liquid chromatography/Sur- Glioblastoma tumors were independently analyzed veyor (Transgenomic Inc., Omaha, NE),30 MALDI-TOF,30 from both frozen and FFPE forms. For each, histologi-

Table 1. Cell-Line DNA Used to Evaluate the Efficacy of COLD-PCR Enrichment

TP53 Amplicon Cell-line Mutation (nt) Mutation (aa) Exon 6 6A T47D c.580CϾT p.L194F Exon 7 7 HCC2157 c.742CϾT p.R248W Exon 8 8A MDA-MB-435 c.797GϾA p.G266E Exon 8 8B HCC2218 c.847CϾT p.R283C Exon 9 9 SW480 c.925CϾT p.P309S 222 Milbury et al JMD March 2011, Vol. 13, No. 2

cal analysis of slides was completed before tissue se- To determine the sensitivity of our approach, 3 ng lection and genomic extraction. Frozen glioblastoma tu- genomic DNA from cell lines (T47D, SNU-182, mor specimens were embedded in Tissue-Tek Optimal HCC2157, MD-MBA-435, HCC2218, and SW480) was Cutting Temperature (OCT) Compound (Sakura Finetek serially diluted into wild-type DNA and amplified by the Inc., Torrance, CA). The OCT compound is composed of TP53 multiplex-PCR approach. These sets of serial water soluble glycol and resins, and is an appropriate dilutions served for estimating method sensitivity in specimen matrix for cryostat sectioning at temperatures exons 6, 7, 8, and 9. The specific mutant fractions of Ϫ10°C and lower. At the Brigham and Women’s Hos- examined were: 0.1%, 0.25%, 0.5%, 1.0%, 2.0%, 3.0%, pital, Specialized Histology Longwood Core facility, sev- 4.0%, 5.0%, 6.0%, 8.0%, and 10% mutant-to-wild-type eral 4 ␮mol/L and 10 ␮mol/L sections were cut from the ratios. In addition, several replicates of wild-type hu- OCT-embedded tissue for (1) staining with hematoxylin man genomic DNA (0% mutant; male G1471, Promega and eosin, and (2) genomic DNA extractions, respec- Corp) were amplified and evaluated in parallel. Serial tively. Stained sections were examined and the percent dilutions of the mutant cell lines were used in down- tumor content was determined. Samples containing stream assays to validate method sensitivity and en- Ն70% to 80% tumor content were selected for analysis. richment selectivity. Additionally, 3 ng of colon tumor OCT was removed via three washes with 5 ml of phosphate (N ϭ 12), as well as paired frozen and FFPE glioblas- buffered saline (PBS) before isolating genomic DNA with toma tumor (N ϭ 12 for each respectively) genomic the DNeasy tissue kit (Qiagen Inc.). DNA were amplified by the TP53 multiplex-PCR For the FFPE glioblastoma tumor specimens, approx- method, along with additional genomic male reference imately five 4 ␮mol/L sections were selected from un- DNA (n ϭ 6). Subsequently, conventional PCR ampli- stained histological slides. Genomic DNA was isolated fication was applied for each TP53 exon. from the FFPE sections using a QIAamp DNA FFPE tissue kit (Qiagen Inc.) following manufacturer’s protocols. For both sets of genomic DNA (from frozen and FFPE spec- Conventional PCR-HRM and Sequence Analysis imens), quality was assessed and quantified using the NanoDrop1000 spectrophotometer (ThermoFisher Scien- The 10 coding exons (exons 2 through 11) of TP53 tific, Inc.). Purified DNA was stored at Ϫ80°C. were amplified in 14 individual reactions from the mul- The protocol presented below was first validated for tiplex-PCR products (Table 2). The primers selected varying amounts of snap-frozen tumor DNA as starting amplified each exon of TP53 including the intron-exon material. To determine and assure method sensitivity, the splicing regions. Two amplicons were required to span assays were evaluated using three different starting the entire length of the exon and splice region of four amounts of DNA, 30 ng, 10 ng, and 3 ng. For each exons (exons 4, 5, 6, and 8). Primers were designed or starting amount, serial dilutions were created; multiplex- selected23 such that the PCR amplicons would pos- PCR was performed, followed by conventional PCR and sess a single melting domain, amplify efficiently, and COLD-PCR in separate analyses, and finally screened generate amplicons of less than 200 bp in length. via HRM. The amplification was robust in all cases (data Conventional PCR amplicons were amplified on a Mas- not shown). As clinical patient samples are precious and terCycler using thermocycling conditions as defined in often in limited amounts, we proceeded to develop the Table 3. Amplifications were completed in triplicate to approach using as little DNA as necessary. For all re- ensure reproducibility. PCR reactions were performed maining reactions, 3 ng of DNA (ϳ1000 genome equiv- using 1X manufacturer-supplied high-fidelity buffer, 1.5 ␮ alents) was used. mmol/L MgCl2, 0.2 mmol/L dNTPs, 0.2 mol/L primers, 1.0X LCGreenϩ dye (Idaho Technologies Inc., Salt Lake City, UT), 5 U/␮L Phusion high fidelity polymerase, and Multiplex-PCR for TP53 1 ␮L of a 200-fold diluted multiplex-PCR product. Ampli- fied products were evaluated by electrophoresis on an Multiplex-PCR for TP53 was designed using guidelines agarose gel with ethidium-bromide staining to confirm presented by Fredriksson et al.37 Seven primer sets were robust amplification of the correct length. used to amplify all exons and splice regions of TP53 To evaluate the sensitivity of our approach, and to (Table 2). Amplifications (reaction volume 15 ␮L) were screen for the presence of mutations, approximately 10 performed on a MasterCycler EP (Eppendorf Inc., Haup- ␮L of each PCR product, with a 20 ␮L mineral oil pauge, NY) with thermocycling conditions performed as overlay, was subjected to HRM analysis on the Light- described in Table 3. Multiplex-PCR was performed us- Scanner HR96 system (Idaho Technologies Inc.). Melt- ing a high-fidelity polymerase (Phusion DNA Polymerase; ing curves were analyzed via the LightScanner soft- Thermo Scientific Inc., Lafayette, CO) to prevent PCR ware with Call-IT 2.0 (Idaho Technologies Inc.) to error as follows: 1X manufacturer-supplied high-fidelity discern the presence of a mutation. Products amplified ␮ buffer, 2.0 mmol/L MgCl2, 0.2 mmol/L dNTPs, 0.1 mol/L from the colorectal and glioblastoma tumor specimens each primer (14 in total), 5 U/␮L Phusion high fidelity were directly compared with each other and with sev- polymerase, and 3 ng of DNA. Exonuclease I (1 ␮L) was eral wild-type, genomic male DNA reference samples added to each multiplex-PCR product and incubated at in parallel. Samples exhibiting aberrant melting profiles 37°C for 30 minutes, followed by 95°C for 5 minutes, to were subsequently subjected to Exonuclease I/shrimp degrade any remaining unincorporated primers. alkaline phosphatase digestion and sequenced at COLD-PCR and HRM Mutation Detection 223 JMD March 2011, Vol. 13, No. 2

Table 2. Multiplex TP53 Primer Sequences from Fredriksson et al.37

Exon Sequence Size (bp) Tm°C Tc°C Ta°C Multiplex primers 2-3 Forward-5=-ATGCTGGATCCCCACTTTTC-3=* 350 Reverse-5=-GACCAGGTCCTCAGCCC-3=* 4 Forward-5=-GACAAGGGTTGGGCTGG-3=* 486 Reverse-5=-CCAAAGGGTGAAGAGGAATC-3=* 5-6 Forward-5=-TCTTTGCTGCCGTCTTCC-3=* 517 Reverse-5=-AGGGCCACTGACAACCAC-3=* 7 Forward-5=-TGCTTGCCACAGGTCTCC-3=* 235 Reverse-5=-GTCAGAGGCAAGCAGAGGC-3=* 8-9 Forward-5=-GGACAGGTAGGACCTGATTTCC-3=* 441 Reverse-5=-AAACAGTCAAGAAGAAAACGGC-3=* 10 Forward-5=-AACTTGAACCATCTTTTAACTCAGG-3=* 243 Reverse-5=-GGAATCCTATGGCTTTCCAAC-3=* 11 Forward-5=-AGGGGCACAGACCCTCTC-3=* 222 Reverse-5=-AGACCCAAAACCCAAAATGG-3=* Nested primers 2 Forward-5=-GCAGCCAGACTGCCTTCCG-3= 134 90.6 89.6 57 Reverse-5=-GTGGGCCTGCCCTTCCAAT-3= 3 Forward-5=-gtaaaacgacggccagtTGGGACTGACTTTCTGCT-3=† 108 86.8 85.9 58 Reverse-5=-tcccgcgaaattaatacgacGCCCAACCCTTGTCCTTA-3=† 4a Forward-5=-GTCCTCTGACTGCTCTTTTCACCC-3= 181 90.8 89.8 60 Reverse-5=-GGTGTAGGAGCTGCTGGTGC-3= 4b Forward-5=-CCCGTGGCCCCTGCACC-3= 186 88.8‡ 87.8 70 Reverse-5=-AGCCAGCCCCTCAGGGCAA-3= 5a Forward-5=-TGTGCCCTGACTTTCAACTCTGTCTC-3= 120 88.8 87.7 68 Reverse-5=-GGGTGTGGAATCAACCCACAGC-3= 5b Forward-5=-TTCCACACCCCCGCCCG-3= 148 94.5 93.5 63 Reverse-5=-GCCCTGTCGTCTCTCCAGCC-3= 6a Forward-5=-GCCTCTGATTCCTCACTGATTG-3= 129 86.9 86.0 57 Reverse-5=-TAGGGCACCACCACACTATG-3= 6b Forward-5=-TGCGTGTGGAGTATTTGGATGAC-3= 105 88.5 87.5 57 Reverse-5=-CCCTCCTCCCAGAGACCC-3= 7 Forward-5=-CCAAGGCGCACTGGCCTCA-3= 185 91.0 90.0 57 Reverse-5=-GCCAGTGTGCAGGGTGGCAA-3= 8a Forward-5=-TGCCTCTTGCTTCTCTTTTC-3=† 128 88.8 88.0 57 Reverse-5=-CTTTCTTGCGGAGATTCTCTTC-3=† 8b Forward-5=-GAACAGCTTTGAGGTGCGTGTTT-3= 155 91.4 90.5 57 Reverse-5=-TGGTCTCCTCCACCGCTTC-3= 9 Forward-5=-GGTGCAGTTATGCCTCAGAT-3= 176 87.2 86.1 57 Reverse-5=-GTTAGACTGGAAACTTTCCACTTGATA-3= 10 Forward-5=-ATATACTTACTTCTCCCCCTCCTCTGTTGC-3= 172 92.8 91.8 70 Reverse-5=-TAGGGCCAGGAAGGGGCTGA-3= 11 Forward-5=-CTCACTCATGTGATGTCATCTCT-3= 165 88.1 87.2 57 Reverse-5=-GGGAGGCTGTCAGTGGG-3=

Nested primers were used to amplify the length of each TP53 exon via both conventional and COLD-PCR. Nested Exon 3 primers were modified to include M13 and Tag1 linker sequences (lower case font) in order to increase the product length and allow sequence analysis. The annealing (Ta), amplicon melting temperature (Tm), and critical denaturation temperature (Tc) of COLD-PCR are presented for PCR reactions performed with the Phusion high-fidelity polymerase system and a 1X concentration of LCGreenϩ dye (Idaho Technologies Inc.). When necessary (Exons 4, 5, 6, and 8), two amplicons were analyzed in order to screen the entire length of the exon; primer sets are differentiated by “a” and “b” labels. *Fredriksson et al.37 †Bastien et al.23 ‡DMSO was added at 5% to lower the amplicon melting temperature. the Dana-Farber Cancer Institute, Molecular Biology tify the experimentally-derived amplicon melting tem-

Core Facility. Sequence chromatograms were evalu- perature (Tm); the melting curve analysis was ated using BioEdit biological sequence alignment editor performed post-PCR on the SmartCycler (Cepheid (http://www.mbio.ncsu.edu/BioEdit/BioEdit.html, May 2005). Inc., Sunnyvale, CA) using LCGreenϩ as a real-time intercalating dye. The critical denaturation temperatures (T ) were defined following the general rule: T ϭ COLD-PCR-HRM and Sequence Analysis c c Ϫ Tm 1°C. Defining the Tc in this manner resulted in The same 14 regions were amplified in the serial dilu- robust PCR amplification and also demonstrated ex- tion and tumor sample sets using COLD-PCR. To de- cellent mutation enrichment. fine the critical denaturation temperatures for COLD- COLD-PCR was performed on a SmartCycler II (Ceph- PCR of a given amplicon, a melting curve was first eid Inc.). Because the critical denaturation temperature evaluated for a wild-type sample amplified via conven- during COLD-PCR has to be controlled precisely (eg, to tional PCR, followed by melting curve analysis to iden- within Ϯ0.2°C), it is important to use a thermocycler with 224 Milbury et al JMD March 2011, Vol. 13, No. 2

Table 3. PCR Thermocycling Conditions

PCR type Step Conditions Multiplex-PCR Initial denaturation 98°C for 30 s Thermocycling: 35 cycles 98°C for 10 s 55°C for 20 s 72°C for 20 s Extension 72°C for 15 s Nested conventional PCR Initial denaturation 98°C for 30 s Thermocycling: 35 cycles 98°C for 10 s

Ta* for 30 s 72°C for 10 s Nested fast-COLD-PCR Initial denaturation 98°C for 30 s Stage 1 cycling: 5 cycles 98°C for 10 s

Ta* for 20 s, fluorescent reading 72°C for 10 s

Stage 2 cycling: 20 cycles Tc* for 10 s Ta* for 20 s, fluorescent reading 72°C for 10 s Stage 3 cycling: 10 cycles 98°C for 10 s

Ta* for 20 s, fluorescent reading 72°C for 10 s Melting curve Ramping 0.2°C/s, 60°C to 90°C

*Annealing temperatures and critical denaturation temperatures are defined in Table 2. high temperature precision. The Cepheid SmartCycler II can be customized and performed. In this investigation, enables temperature calibration of each individual well, we independently confirmed a GϾT mutation (c.782 ϩ thereby ensuring minimal well-to-well temperature varia- 1GϾT) in exon 7 of specimen CT11. A PspGI recognition tion and satisfactory reproducibility in the mutation en- site (CCWGG) is present in the wild-type sequence; how- richment obtained from replicate samples and across all ever, the presence of a mutation in that recognition se- wells. COLD-PCR thermocycling was performed as de- quence prevents digestion of the mutant strands (eg, a scribed in Table 3. A melting analysis was performed as “CCATG” sequence string characterizes the same region in the last step of the thermocycling on the Cepheid Smart- CT11). As such, on digestion by the restriction enzyme, the Cycler to ensure strong amplification before further HRM wild-type sequences are digested and the mutant se- evaluation. quences remain intact. We thus used this PCR – restriction COLD-PCR reactions were performed using 1X manu- fragment length polymorphisms (PCR-RFLP) approach to facturer-supplied high-fidelity buffer, 1.5 mmol/L MgCl2, 0.2 independently verify the mutation in sample CT11. mmol/L dNTPs, 0.2 ␮mol/L primers, 1.0X LCGreenϩdye In this evaluation, both genomic DNA and multiplex-am- (Idaho Technologies Inc.) 5 U/␮L Phusion (Thermo Scien- plified product were evaluated for the colon tumor speci- tific Inc.) high-fidelity polymerase, and 1 ␮L of a 200-fold men CT11 and wild-type DNA. The aforementioned 185-bp diluted multiplex-PCR product. As with conventional PCR, target region was PCR amplified using the primers defined 10 ␮L of each COLD-PCR product, with a 20 ␮L mineral oil in Table 2. The 185 bp amplicon (4 ␮L) was incubated with overlay, was subjected to high-resolution melting on the five units of the restriction enzyme PspGI (New England LightScanner HR96 system (Idaho Technologies Inc.). Melt- Biolabs, Ipswich, MA) and 1X restriction buffer #2 (New ing curves were analyzed via the LightScanner software England Biolabs), at 75°C for 2 hours. Resulting digested with Call-IT 2.0 to discern the presence of a mutation. HRM product was used directly in a subsequent nested conven- melt profiles were directly compared with the wild-type ref- tional PCR, as described above (forward primer 5=-CCT- erence genomic DNA. HRM mutation detection sensitivity CATCTTGGGCCTGTGTTAT-3=; reverse primer 5=-TGTG- was compared between conventional and COLD-PCR. Ab- CAGGGTGGCAAGT-3=; annealing temperature 57°C), to errant melting profiles were subsequently analyzed via yield a 166-bp product. The 166 bp amplicon was subse- Sanger sequencing to identify variants. quently Sanger sequenced, and the mutation was verified by visual analysis of the chromatogram. Independent Verification by Restriction Enzyme Analysis Results In scenarios when both the HRM results and the se- Multiplex-PCR and Conventional PCR Validation quence analysis reveal very low-abundance mutations, representing the borderline limit of the application, inde- Following multiplex-PCR, each of the 14 regions of interest, pendent confirmation by an alternative approach should spanning the coding and splice regions of exons 2 through be performed via restriction enzyme digestion. After a 11, were amplified via conventional PCR for validation. Am- specific nucleotide change is putatively identified via plicons were visualized via agarose gels with ethidium-bro- COLD-PCR-sequencing, a sequence-specific reaction mide staining and robust amplification of the correct length COLD-PCR and HRM Mutation Detection 225 JMD March 2011, Vol. 13, No. 2 was confirmed for each target region (see Supplemental sequencing, the 3% HCC2157 mutation abundance can Figure S1 at http://jmd.amjpathol.org). The amplification of be clearly identified (Figure 1), and results in an approx- the 14 TP53 regions of interest was robust in all cases. imately 50% mutation abundance after COLD-PCR enrichment. Analysis of the exon 8Ϫ128 bp amplicon demonstrates Mutation Detection Sensitivity via Serial-Dilution that HRM screening of the cell line MD-MBA-435 (mutation Analysis c.797GϾA, p.G266E) after conventional PCR can reliably detect a mutation above 5%. In comparison, after COLD- PCR amplicons (conventional PCR and COLD-PCR) of mu- PCR, a 2% mutation abundance of MD-MBA-435 diluted in tant serial dilutions for varying amounts of template DNA wild-type DNA can be differentiated from wild-type DNA were analyzed by HRM to determine mutation detection melting curves. Thus, for the exon 8Ϫ128 bp amplicon, sensitivity. The sensitivity was comparable for each of three COLD-PCR-HRM analysis provides an approximate three- starting amounts (30 ng, 10 ng, and 3 ng). As this approach fold improvement in detection sensitivity over conventional- is focused on the amplification of small amounts of DNA PCR-HRM (see Supplemental Figure S3 at http://jmd. from precious samples, we proceeded with 3 ng of DNA in amjpathol.org). In Sanger sequence chromatograms, a mu- all subsequent analyses. Our results for the serial dilution tant concentration of 5% is not discernable; however, after analysis of 3 ng of DNA are presented below. COLD-PCR amplification and sequencing, the 5% MD- HRM analysis of 3 ng serial dilutions of mutation-con- MBA-435 mutation abundance can be clearly identified taining human cancer cell lines diluted into wild-type (see Supplemental Figure S3 at http://jmd.amjpathol.org). human genomic DNA revealed increased mutation-de- Analysis of the exon 8Ϫ155 bp amplicon reveals that tection sensitivity in products of COLD-PCR amplification HRM screening of the cell-line HCC2218 mutation over conventional PCR amplification. After PCR amplifi- (c.847CϾT, p.R283C) after conventional-PCR can detect cation and HRM analysis of human cancer cell-line T47D a mutation abundance of 5%. However, after COLD-PCR serial dilutions in wild-type DNA, fluorescence difference amplification, a 1% mutation abundance of HCC2218 curve plots relative to wild-type DNA were generated diluted in wild-type DNA can be clearly differentiated using the LightScanner Call-IT 2.0 software (Idaho Tech- from wild-type DNA melting curves. Thus, for the 155-bp nologies Inc.). Conventional PCR-HRM was performed in exon 8 amplicon, COLD-PCR-HRM analysis results in an parallel for comparison with COLD-PCR-HRM. After con- approximate fivefold improvement in detection sensitivity ventional-PCR, the exon 6 (129-bp amplicon) T47D cell- over HRM analysis of conventional PCR amplicons (see line mutation (c.580CϾT, p.L194F) remains clearly differ- Supplemental Figure S4 at http://jmd.amjpathol.org). In entiated from the wild-type amplicons down to 2% to 3%. Sanger sequence chromatograms, a mutant concentra- In comparison, after COLD-PCR, a 0.5% mutation-abun- tion of 5% is not reliably detected; however, after COLD- dance of T47D diluted in wild-type DNA can be differen- PCR amplification and sequencing, the 5% HCC2218 tiated from the wild-type DNA melting curves. Thus, for the mutation-abundance can be clearly identified (see Sup- 129-bp exon 6 amplicon, HRM analysis of COLD-PCR prod- plemental Figure S4 at http://jmd.amjpathol.org). ucts provides an approximate fourfold to sixfold improvement in mutation detection over conventional-PCR-HRM Tumor Mutation Scanning (see Supplemental Figure S2 at http://jmd.amjpathol.org). To determine the degree of mutant enrichment and the ability Mutations and single-nucleotide polymorphisms (SNPs) to identify the mutation through downstream applications, were detected by HRM analysis of both conventional and Sanger sequencing of the low-abundance mutation was COLD-PCR amplicons (Table 4) in all coding exons ex- performed. Despite the inherent sensitivity of HRM screen- cluding exons 3, 10, and 11. Additionally, mutations and ing, a mutant concentration of 10% cannot be identified in SNPs were detected in the intronic splice sites between the sequence chromatograms of the conventional PCR. In exons 7-8, exons 8-9, and exons 9-10. Mutations were comparison, after COLD-PCR amplification and sequenc- detected in 75% and 42% of the colorectal and glioblas- ing, the 5% T47D mutation-abundance can be identified toma specimens examined, respectively; 25% and 33% (see Supplemental Figure S2 at http://jmd.amjpathol.org), of the colorectal and glioblastoma specimens examined and results in an approximate ϳ35% mutation abundance contained more than one mutation in TP53. Twenty-four by examining the sequencing chromatograms. mutations were identified (Table 4): 14 (58%) were mis- Similarly, for the exon 7 (185 bp) amplicon, after con- sense mutations, three (12.5%) were deletions resulting ventional-PCR, the human cell-line HCC2157 mutation in frameshifts, two (8.3%) were found in splice regions, (c.742CϾT, p.R248W) can be differentiated from the and five were intronic (21%). wild-type amplicons down to 3%. In comparison, after Seven and eight moderate abundance mutations were COLD-PCR, a 0.5% mutation-abundance of HCC2157 detected in the conventional amplicons of the glioblas- diluted in wild-type DNA remains clearly differentiated toma and colorectal tumor specimens, respectively. Al- from the wild-type DNA melting curves. Thus, for the 185 though no low-abundance mutations were detected after bp exon 7 amplicon, COLD-PCR-HRM results in an ap- COLD-PCR amplification of glioblastoma specimens, four proximate sixfold improvement in mutation detection over low-abundance mutations were detected after COLD- conventional-PCR-HRM (Figure 1). In Sanger sequence PCR amplification of the colorectal specimens. COLD- chromatograms, a mutant concentration of 5% is not dis- PCR increased the mutation detection of HRM by 33% in cernable; however, after COLD-PCR amplification and the colorectal specimens. Low-abundance mutations 226 Milbury et al JMD March 2011, Vol. 13, No. 2

Figure 1. A 185-bp amplicon of TP53 exon 7 was analyzed via HRM after conventional PCR (A) and COLD-PCR (C). Amplicons were subsequently sequenced (Sanger, sense strand). While a 5% mutant abundance of the HCC2157 cell line (CϾT) is unreliable in the chromatogram of conventional PCR amplicons (B), a 3% mutant abundance is easily visible in the chromatogram of COLD-PCR (D). Wild-type (WT) sequences are presented for comparison. were confirmed in COLD-PCR amplifications from the However, this mutation is unique in that it is detected in genomic DNA; evaluation of the matched normal the matched normal sample when amplified by COLD- genomic DNA was performed simultaneously. PCR; the mutation was not observed in sequence chro- Four low-abundance mutations were detected in exons matograms from the genomic male control sample ampli- 5, 8, and 9 of colorectal tumor specimens. Colorectal fied in parallel (Figure 4). Although located close to a tumor specimen CT20 possessed two low-abundance mutational hotspot associated with germline Li-Fraumeni mutations (exons 8 and 9) in addition to another heterozy- syndrome,38 this mutation (c.527GϾT, missense p.C176F) gous mutation in exon 5. Exon 8 exhibited a GϾA muta- is not documented as a common germline mutation in the tion at the commonly mutated codon 273; and results in a IARC TP53 database (http://www-p53.iarc.fr/, December missense protein change from arginine to histidine. This 2009). A potential explanation for the occurrence of this low-abundance mutation was detectable only in COLD- mutation is that the matched normal tissue may not be PCR amplicons (Figure 2). Sequence chromatograms of purely normal (due to the macrodissection); this tissue the COLD-PCR amplicons from both the multiplex-PCR may contain tissue from the tumor margin. product as well as the genomic DNA were used to iden- A low-abundance mutation was detected by HRM for tify and confirm the mutation; the mutation was not pres- CT2 in the COLD-PCR amplicons of exon 5 (148 bp ent in the matched normal amplicon from genomic DNA. amplicon) (see Supplemental Figure S5 at http://jmd. Similarly a second low-abundance mutation was de- amjpathol.org). This CT2 mutation (c.523CϾA, p.R175S) tected by HRM for CT20 in exon 9 in the COLD-PCR results in a missense protein change from arginine to amplicons only (Figure 3, sequence of antisense strand serine; the mutation was not present in the matched nor- presented). This CT20 mutation is a CϾT mutation at mal amplicon from genomic DNA. CT2 contains another codon 309 and results in a missense protein change from mutation in exon 7 which is a heterozygous AϾT mutation proline to serine; the mutation was not present in the (c.739AϾT, p.N247I missense mutation), and can be de- matched normal amplicon from genomic DNA. The exon tected in both conventional and COLD-PCR amplicons 5 mutation for CT20 is, in contrast, a heterozygous GϾT (see Supplemental Figure S5 at http://jmd.amjpathol.org). mutation (missense p.C176F), and can be detected in The last low-abundance mutation was detected in both conventional and COLD-PCR amplicons (Figure 4). the intronic spice region between exons 7 and 8 in COLD-PCR and HRM Mutation Detection 227 JMD March 2011, Vol. 13, No. 2

Table 4. Mutational Status of CT and CMK-T Specimens in Exons 2–11 of TP53

MPLX ϩ MPLX ϩ HRM- Matched Chromatogram Chromatogram Tumor TP53 Protein HRM- conv. COLD- Chromatogram Chromatogram normal conv./ genomic DNA genomic DNA specimen exon Variant (nt) change Result PCR PCR conv. PCR COLD-PCR COLD-PCR conv. PCR COLD-PCR

CMK-T7 2 c.40del2 p.L14del2 Frameshift (novel) Y Y Y Y CT12 4 c.108GϾA p.P36P SNP- synonymous Y Y Y Y Y CMK-T3 4 c.108GϾA p.P36P SNP- synonymous Y Y Y Y CMK-T12 4 c.108GϾA p.P36P SNP- synonymous Y Y Y Y CMK-T13 4 c.108GϾA p.P36P SNP- synonymous Y Y Y Y CMK-T14 4 c.108GϾA p.P36P SNP- synonymous Y Y Y Y CMK-T1 5 c.403del3 p.C135del3 Frameshift (novel) Y Y Y Y CT6 5 c.476CϾT p.A159V Missense YY Y CT2 5 c.523CϾA p.R175S Missense Undetectable Y Undetectable Y WT/WT Undetectable Y CT10 5 c.524GϾT p.R175K Missense YY Y CT17 5 c.524GϾA p.R175H Missense YY Y CT20 5 c.527GϾT p.C176F Missense YY Y Y WT/MUTANT Y Y CMK-T3 6 c.568CϾT p.P190L Missense YY Y CMK-T5 6 c.610del3 p.E204del3 Frameshift YY Y CMK-T12 6 c.639AϾG p.R213R SNP- synonymous Y Y Y CMK-T16 7 c.712TϾG p.C238W Missense YY Y CT2 7 c.739AϾT p.N247I Missense YY Y CT11 7-intronic c.782 ϩ 1GϾT Splice Undetectable Y Undetectable Y WT/WT Undetectable Y CT10 8 c.817CϾT p.R273C Missense YY Y CMK-T7 8 c.817CϾT p.R273C Missense YY Y CT20 8 c.818GϾA p.R273H Missense Undetectable Y Undetectable Y WT/WT Undetectable Y CT5 8 c.853GϾA p.E285K Missense YY Y CMK-T16 8 c.916CϾT p.R306X Nonsense YY Y CT20 9 c.925CϾT p.P309S Missense Undetectable Y Undetectable Y WT/WT Undetectable Y CT12 9-intronic c.993 ϩ 12TϾC Intronic YY Y Y CMK-T3 9-intronic c.993 ϩ 12TϾC Intronic YY Y Y CMK-T12 9-intronic c.993 ϩ 12TϾC Intronic YY Y Y CMK-T13 9-intronic c.993 ϩ 12TϾC Intronic YY Y Y CMK-T14 9-intronic c.993 ϩ 12TϾC Intronic YY Y Y CT4 9-intronic c.994-1GϾC Splice YY Y

Mid- to high-abundance mutants and SNPs could be detected in HRM analysis of both conventional (conv.) and COLD-PCR amplicons (denoted “Y”) amplified from the multiplex-PCR product; however, four low-abundance mutations (in bold) were not detected by HRM analysis of conventional PCR (denoted “undetectable”), though could be detected in COLD-PCR amplicons (denoted “Y”). Sanger sequence analysis was used to identify the mutations in the product amplified from the multiplex-PCR. Furthermore, low-abundance mutations, and additional mutations of interest, were secondarily sequenced from amplified genomic DNA. CT, colorectal tumor; CMK-T, glioblastoma tumor. colorectal tumor specimen CT11. After COLD-PCR am- They were detected in both the conventional and plification, a GϾT mutation (c.782 ϩ 1GϾT) was de- COLD-PCR amplicons by HRM, and identified via tected by HRM via the Call-IT 2.0 software (in triplicate) Sanger sequencing. CMK-T7 exhibits a 2-bp deletion and observed in sequence chromatograms of both the in exon 2 (p.L14del2, c.40del2) and the CMK-T1 3-bp multiplex-PCR product and the independent analysis deletion is in exon 5 (p.C135del3, c.403del3). Se- of the genomic DNA (see Supplemental Figure S6 at quence chromatograms are presented in Supplemen- http://jmd.amjpathol.org); the mutation was not ob- tal Figure S8 (http://jmd.amjpathol.org) for products served in either the conventional PCR amplicons or the amplified from both the multiplex-PCR and the original matched normal sample. However, due to the very low genomic DNA. These mutations are not documented in abundance of the mutation observed via both HRM and either the IARC TP53 database (http://www-p53.iarc. Sanger sequencing, the CT11 mutation is close to the fr/, December 2009), the TP53–free database (http:// sensitivity limits of this method. The HRM differentiation p53.free.fr/Database/p53_database.html, November was quite small, and the enriched mutant fraction via 2008), or the COSMIC database (http://www.sanger.ac. COLD-PCR was quite low. We performed an indepen- uk/genetics/CGP/cosmic/, May 2010). dent verification of this approach using a restriction Lastly, a panel of 12 FFPE glioblastoma tissue sec- enzyme (PspGI) to digest the wild-type amplicons and tions were evaluated and compared with the matched effectively enrich the mutant fraction following a first panel of DNA specimens prepared from frozen speci- conventional PCR (PCR-RFLP). A second, nested PCR mens. For each of the 14 amplicons, amplified by both and sequence analysis of the digested product was conventional PCR and COLD-PCR, there was 100% then performed in parallel on both the multiplex-ampli- concordance in the HRM mutation detection analysis. fied CT11 DNA and the genomic CT11 DNA, as well as A representative result demonstrating this concor- both multiplex-amplified and genomic wild-type DNA. dance is represented in Supplemental Figure S9 (see The analysis revealed clearly the GϾT mutation (c.782 http://jmd.amjpathol.org). ϩ 1GϾT) at an appreciably enriched level in CT11, while the wild-type sequence remained without evi- dence of a mutation (see Supplemental Figure S7 at Discussion http://jmd.amjpathol.org). Two novel mutations, both heterozygous deletions, Many molecular applications and their respective were detected in two glioblastoma tumor specimens. depth of analysis can be limited by low-yield specimen 228 Milbury et al JMD March 2011, Vol. 13, No. 2

Figure 2. In the exon 8 (128 bp) amplicon, a low-abundance mutation in a colorectal tumor sample (CT20) was detected after analyzing COLD-PCR amplicons with HRM (A); however, this mutation was not detected in conventional PCR amplicons. An additional three samples possessed mid- to high-abundance mutations detectable by HRM screening of both conventional and COLD-PCR amplicons. For the exon 8 (128 bp) amplicon, Sanger sequencing analysis (sense strand) of the multiplexed-PCR product for CT20 revealed a GϾA mutation in the sequence chromatograms of COLD-PCR (B); however, this mutation was not detected in conventional PCR amplicons. The mutation was confirmed in subsequent analysis of the COLD-PCR-amplified genomic DNA (C); however, the mutation was not observed in either the conventional PCR amplicon or the matched normal sample (D). samples. The application of a targeted multiplex-PCR cycles that are performed to exclude the possibility for poly- on low-yield precious specimen samples generates merase-introduced errors. Though PCR errors were not ob- high quantities of template that can be used in down- served in this investigation, such errors in principle can stream applications, allowing for an in-depth analysis occur if an excessive number of PCR cycles are performed. of mutation spectra. Primers used in multiplex-PCR can To reduce the probability of errors, we use the highest- be selected for particular regions of interest. For ex- fidelity polymerase commercially available (Phusion DNA ample, Fredriksson et al37 amplified the entire coding Polymerase; Thermo Scientific Inc.), and we analyze multi- sequence of ten human cancer genes in one assay, ple wild-type control samples within all evaluations. generating an extensive template panel for biomarker For this particular study and method development, we screening. Here we demonstrate a multiplex-PCR tailored elected to use fast-COLD-PCR29,30,39,40 to enrich muta- for the amplification of exons 2-11 of TP53 that can be tions. Fast-COLD PCR will enrich only those mutations performed on as little as 3 ng of DNA. The merging of that result in decreasing the melting temperature (Tm)of multiplex-PCR with downstream COLD-PCR reactions not a particular amplicon. We chose to use the fast-COLD only increases the number of downstream assays that can approach for several reasons. The majority of human 41 be performed, but simultaneously allows for mutation and TP53 mutations are Tm-reducing, a bias which puta- minority allele enrichment from as little as 1000 genome tively reflects the methylation of 5=-CpG-3= dinucle- equivalents. Subsequent amplicon analysis with HRM cre- otides.42,43 Fast-COLD-PCR assays can be run in less time, ates a high-throughput, efficient, and inexpensive screen- and products yield robust enrichment. Full-COLD PCR is ing method to detect mutant-containing amplicons. another COLD-PCR format that can be used in place of The use of multiplex-PCR as a pre-amplification step, fast-COLD-PCR to enrich all mutation types; however, in its as opposed to whole-genome pre-amplification ap- original format, full-COLD-PCR requires much longer hy- proaches,1–6 enables better mutation enrichment during bridization times and the achievable enrichment is typically subsequent COLD-PCR amplification. Thus, enriching lower than fast-COLD-PCR. Nevertheless, improved full- the target(s) of interest during the multiplex-PCR pre- COLD-PCR protocols that generate better enrichment are amplification simplifies the target and allows COLD-PCR currently under development in our laboratory. These will to be initiated after only five cycles of the conventional allow faster full-COLD-PCR reactions that can be used in the PCR mode, thereby enabling more cycles in COLD-PCR present approach to identify all possible mutation types. mode, and subsequently increasing the enrichment po- Twenty-four clinical tumor specimens were evaluated tential. However, it is important to limit the number of PCR in this study; 12 colorectal tumor and 12 glioblastoma COLD-PCR and HRM Mutation Detection 229 JMD March 2011, Vol. 13, No. 2

Figure 3. In the exon 9 amplicon, a low abundance mutation in a colorectal tumor sample (CT20) was detected after analyzing COLD-PCR amplicons with HRM (A); however, this mutation was not detectable in conventional PCR amplicons. An additional six samples possessed mid- to high-abundance SNPs detectable by HRM screening of both conventional and COLD-PCR amplicons. For the exon 9 amplicon, Sanger sequencing analysis of the multiplexed-PCR product for CT20 revealed a CϾT mutation in the sequence chromatograms of COLD-PCR (B, anti-sense strand sequencing presented); however, this mutation was not detectable in conventional PCR amplicons or the matched normal sample (D). The mutation was confirmed in subsequent analysis of the COLD-PCR amplified genomic DNA (C). tumor specimens. The 12 glioblastoma samples were COLD-PCR. Although still unresolved,44 it was reported evaluated using matched DNA isolated from frozen tissue that TP53 mutation is an early event in the progression of as well as FFPE histological slide sections. FFPE pre- gliomas.45–48 Similarly, it has been observed that the total served tumor tissues are widely available and used for number and the frequency of nucleic mutations and vari- large-scale clinical investigations; however, it is well ants are substantially reduced in gliomas when compared known that the FFPE preservation method may result in against other cancers such as colorectal, breast, and pan- high levels of DNA damage and induced errors. As such, creatic.49 In reference to an early study,50 Parsons et al49 the analysis of low-abundance mutations in such low- attribute the fact that fewer cell generations occur in glial quality DNA has the potential to yield false-positive re- cells before neoplasia inception as a potential explanation sults. In this evaluation, we compared mutation detection for a relatively small number of genetic aberrations in glio- in frozen and FFPE glioblastoma sections, and recovered mas. As such, one might anticipate finding fewer low-abun- 100% concordance across all 14 targets amplified by dance mutations in the glioma tumor specimens compared conventional and COLD-PCR. to the colorectal tumor specimens. Although these FFPE-based results are promising, one Eight mutations were detected in the conventional must keep in mind that COLD-PCR is designed to enrich PCR-HRM screening of the colorectal tumor specimens low-abundance variants. As such, an error induced by FFPE preservation may be enriched in the same manner compared to 12 mutations via COLD-PCR-HRM screen- as a true mutation. Therefore, the methodology presented ing, allowing us to detect an additional four mutations and herein provides an efficient and sensitive approach for demonstrating a 33% increase in the mutation detection large-scale screening. However, after an aberrant HRM of colorectal tumor mutations. For each of the four low- melting curve is detected, it is important to analyze FFPE- abundance mutations detected in COLD-PCR products preserved specimens with a secondary independent amplified from the multiplex-PCR-template, we confirmed method to prevent false-positive results. the presence of the mutation in the genomic DNA. Previ- Within the glioblastoma specimens analyzed, seven ous analysis30 of TP53 exon 8 in the specimen CT20 has mutations and nine SNPs were detected via conventional validated the low-abundance mutation by independent PCR-HRM screening; however, no additional low-abun- diagnostic approaches such as restriction fragment dance mutations were detected after amplification with length polymorphism analysis. 230 Milbury et al JMD March 2011, Vol. 13, No. 2

Figure 4. In the exon 5 (148bp) amplicon, several mid- to high- abundance mutations were detected in the tumor samples, including colorectal tumor sample CT20. Additionally, after analyzing COLD-PCR amplicons with HRM (A), a low-abundance mutation was detected in CT2; however, this mutation was not detectable in conventional PCR amplicons. A mid- to high-abundance mutation was detected in CT20 by HRM for both PCR amplicons of the exon 5 148bp amplicon (in addition to the previously discussed low-abundance mutations in exons 8 and 9). Sanger sequencing analysis (sense strand) of the multiplexed-PCR product (B) and genomic DNA (C) for CT20 revealed a heterozygous GϾT mutation in conventional PCR amplicons; the T allele was enriched during COLD-PCR. The mutation was also detected in the COLD-PCR amplification of the matched normal tissue (D), suggesting that the putatively normal sample may contain tissue from the tumor margin. This GϾT is unlikely to be artifact of COLD-PCR as it is not observed in the wild-type (genomic human male) amplicons (E).

Unlike the glioma tumor specimens, the percent tumor that the mutation does not possess a strong selective ad- content of the colorectal tumor specimens was not con- vantage and, therefore, it is present in only a low-abun- firmed by histopathology before genomic DNA extraction, dance relative to the other alleles. Recent investiga- so we cannot rule out the possibility that stromal contami- tions51–54 of colorectal adenocarcinoma specimens have nation may reduce the relative mutation concentration. reported intratumoral heterogeneity in mutations of KRAS, However, the data indicate that stromal contamination is at BRAF, P1K3CA, transforming growth factor b type II recep- most a minor contribution in these specimens as other high- tor (TGFBRII), insulin-like growth factor II receptor (IGFIIR), abundance mutations were detected in these same speci- BAX, MSH3, and MSH6, as well as TP53. Supporting our mens that contain also the low-level mutations. For example, observations, Losi et al52 and Giaretti et al54 observed TP53 in two of the samples within which low-abundance muta- intratumoral heterogeneity in up to 59% of the colorectal tions were detected (CT20 and CT2), a high-abundance adenocarcinoma specimens they evaluated, and they re- mutation was also detected in exons 5 and 7, respectively. port that TP53 were late events in the progression of colo- A likely explanation for the observed variability in the colorectal adenocarcinoma. rectal tumor specimens is that TP53 mutational events oc- Using this approach, we demonstrate that low-abun- cur throughout the course of cancer progression, resulting dance mutations that may be missed by conventional in intratumoral heterogeneity. Another explanation would be assays can be detected for several genetic regions of COLD-PCR and HRM Mutation Detection 231 JMD March 2011, Vol. 13, No. 2

interest without the normal limitations presented by small 2. Dean FB, Hosono S, Fang L, Wu X, Faruqi AF, Bray-Ward P, Sun Z, sample size or low-concentration specimens. The clinical Zong Q, Du Y, Du J, Driscoll M, Song W, Kingsmore SF, Egholm M, significance of low-abundance mutations remains largely Lasken RS: Comprehensive human genome amplification using multiple displacement amplification. Proc Natl Acad Sci USA 2002, 99: undetermined at this time, and will be highly variable 5261–5266 depending on tumor type or stage. It is likely that some 3. Hughes S, Arneson N, Done S, Squire J: The use of whole genome mutations will remain insignificant for tumor development amplification in the study of human disease. Prog Biophys Mol Biol and treatment (passenger mutations), whereas others will 2005, 88:173–189 demonstrate a strong selective advantage throughout the 4. Wang G, Maher E, Brennan C, Chin L, Leo C, Kaur M, Zhu P, Rook M, Wolfe JL, Makrigiorgos GM: DNA amplification method tolerant to course of disease progression and/or treatment (driver sample degradation. Genome Res 2004, 14:2357–66 55–59 mutation). We are particularly interested in evaluating 5. Li J, Harris L, Mamon H, Kulke M, Liu W, Zhu P, Makrigiorgos GM: low-abundance mutations that fall into the driver mutation Whole genome amplification of plasma-circulating DNA enables ex- category because they hold the potential to influence and panded screening for allelic imbalance in plasma. J Mol Diagn 2006, determine patient response to treatment, disease progres- 8:22–30 6. Wang G, Brennan C, Rook M, Wolfe JL, Leo C, Chin L, Pan H, Liu WH, sion and evolution, and ultimately survival. Price B, Makrigiorgos GM: Balanced-PCR amplification allows unbi- We elected to evaluate TP53 because the mutational ased identification of genomic copy changes in minute cell and tissue type and position often define their prognostic value.60 Cer- samples. Nucleic Acids Res 2004, 32:e76 tain TP53 mutations have been associated with poor prog- 7. Lim EH, Zhang S-L, Li J-L, Yap W-S, Howe T-C, Tan B-P, Lee Y-S, nosis for several cancers60,61 and loss of TP53-function Wong D, Khoo K-L, Seto K-Y, Tan L, Agasthian T, Koong H-N, Tam J, 62 Tan C, Caleb M, Chang A, Ng A, Tan P: Using whole genome ampli- correlates with the onset of multidrug resistance. By using fication (WGA) of low-volume biopsies to assess the prognostic role of multiplex PCR as a preliminary template construction tool, EGFR, KRAS, p53, and CMET mutations in advanced-stage non-small COLD-PCR for mutation enrichment, and HRM as a scan- cell lung cancer (NSCLC). J Thorac Oncol 2009, 4:12–21 ning method, only those amplicons with aberrant HRM pro- 8. Wittwer CT, Reed GH, Gundry CN, Vandersteen JG, Pryor RJ: High- files require downstream processing and analysis (such as resolution genotyping by amplicon melting analysis using LCGreen. Clin Chem 2003, 49:853–860 Sanger sequencing). The downstream analysis load is re- 9. Reed GH, Wittwer CT: Sensitivity and specificity of single-nucleotide duced appreciably while a great number of targets can be polymorphism scanning by high-resolution melting analysis. Clin rapidly and efficiently analyzed. The present multiplex Chem 2004, 50:1748–1754 COLD-PCR-HRM approach provides a convenient and low- 10. Vaughn CP, Elenitoba-Johnson KSJ: High-resolution melting analysis cost method for the detection and identification of low-level for detection of internal tandem duplications. J Mol Diagn 2004, 6:211–216 somatic mutations in tumors. 11. Zhou L, Wang L, Palais R, Pryor R, Wittwer CT: High-resolution DNA melting analysis for simultaneous mutation scanning and genotyping in solution. Clin Chem 2005, 51:1770–1777 Conclusions 12. Herrmann MG, Durtschi JD, Bromley LK, Wittwer CT, Voelkerding KV: Amplicon DNA melting analysis for mutation scanning and Using a multiplex-PCR approach37 to amplify from small genotyping: cross-platform comparison of instruments and dyes. Clin amounts (eg, 3 ng, 1000 genome equivalents) of DNA pro- Chem 2006, 52:494–503 13. De Leeneer K, Coene I, Poppe B, De Paepe A, Claes K: Rapid and vides a large quantity of DNA template for downstream sensitive detection of BRCA1/2 mutations in a diagnostic setting: applications. The multiplex PCR can be tailored to specific comparison of two high-resolution melting platforms. Clin Chem 2008, regions of interest (ie, tumor suppressor and oncogenic 54:982–989 gene panels) for the analysis of precious DNA such as fine 14. Polakova KM, Lopotova T, Klamova H, Moravcova J: High-resolution needle aspirations, circulating-DNA, or tumor cells where melt curve analysis: initial screening for mutations in BCR-ABL kinase domain. Leuk Res 2008, 32:1236–1243 extensive molecular testing is difficult due to the small 15. Seipp MT, Pattison D, Durtschi JD, Jama M, Voelkerding KV, Wittwer amount of starting material. The use of downstream COLD- CT: Quadruplex genotyping of F5. F2, and MTHFR variants in a single PCR before HRM increases the mutation scanning sensitiv- closed tube by high-resolution amplicon melting. Clin Chem 2008, ity and allows low-abundance mutations to be enriched to 54:108–115 levels that can be easily identified in Sanger sequence 16. Smith GD, Chadwick BE, Willmore-Payne C, Bentz JS: Detection of epidermal growth factor receptor gene mutations in cytology speci- chromatograms. We demonstrate that the use of COLD- mens from patients with non-small cell lung cancer utilising high- PCR increased the mutations detected in colorectal adeno- resolution melting amplicon analysis. J Clin Path 2008, 61:487–493 carcinoma specimens by 33% and increased the sensitivity 17. Vandersteen JG, Bayrak-Toydemir P, Palais RA, Wittwer CT: Identi- of HRM by approximately threefold to sixfold. fying common genetic variants by high-resolution melting. Clin Chem 2007, 53:1191–1198 18. Takano T, Ohe Y, Tsuta K, Fukui T, Sakamoto H, Yoshida T, Tateishi U, Nokihara H, Yamamoto N, Sekine I, Kunitoh H, Matsuno Y, Furuta Acknowledgments K, Tamura T: Epidermal growth factor receptor mutation detection using high-resolution melting analysis predicts outcomes in patients We thank Dr. Jin Li for his insight and critique throughout with advanced non small cell lung cancer treated with gefitinib. Clin the course of this project. Cancer Res 2007, 13:5385–5390 19. Monis PT, Giglio S, Saint CP: Comparison of SYTO9 and SYBR Green I for real-time polymerase chain reaction and investigation of the effect of dye concentration on amplification and DNA melting curve References analysis. Anal Biochem 2005, 340:24–34 20. Gudnason H, Dufva M, Bang DD, Wolff A: Comparison of multiple 1. Hosono S, Faruqi AF, Dean FB, Du Y, Sun Z, Wu X, Du J, Kingsmore DNA dyes for real-time PCR: effects of dye concentration and se- SF, Egholm M, Lasken RS: Unbiased whole-genome amplification quence composition on DNA amplification and melting temperature. directly from clinical samples. Genome Res 2003, 13:954–64 Nucl Acids Res 2007, 35:e127 232 Milbury et al JMD March 2011, Vol. 13, No. 2

21. Erali M, Voelkerding KV, Wittwer CT: High resolution melting applications Papadopoulos N, Vogelstein B, Kinzler KW, Velculescu VE: The for clinical laboratory medicine. Exp Mol Pathol 2008, 85:50–58 consensus coding sequences of human breast and colorectal can- 22. Vossen R, H. A. M., Aten E, Roos A, den Dunnen JT: High-Resolution cers. Science 2006, 314:268–274 Melting Analysis (HRMA) - more than just sequence variant screen- 43. Soussi T, Lozano G: p53 mutation heterogeneity in cancer. Biochem ing. Hum Mutat 2009, 30:860–866 Biophys Res Comm 2005, 331:834–842 23. Bastien R, Lewis TB, Hawkes JE, Quackenbush JF, Robbins TC, 44. Mawrin C: Molecular genetic alterations in gliomatosis cerebri: what Palazzo J, Perou CM, Bernard PS: High-throughput amplicon scan- can we learn about the origin and course of the disease? Acta ning of the TP53 gene in breast cancer using high-resolution fluores- Neuropathol 2005, 110:527–536 cent melting curve analyses and automatic mutation calling. Hum 45. Ohgaki H, Dessen P, Jourde B, Horstmann S, Nishikawa T, Di Patre Mutat 2008, 29:757–764 P-L, Burkhard C, Schuler D, Probst-Hensch NM, Maiorka PC, Baeza 24. Nomoto K, Tsuta K, Takano T, Fukui T, Yokozawa K, Sakamoto H, N, Pisani P, Yonekawa Y, Yasargil MG, Lutolf UM, Kleihues P: Genetic Yoshida T, Maeshima AM, Shibata T, Furuta K, Ohe Y, Matsuno Y: pathways to glioblastoma: a population-based study. Cancer Res Detection of EGFR mutations in archived cytologic specimens of 2004, 64:6892–6899 non-small cell lung cancer using high-resolution melting analysis. 46. Ohgaki H: Genetic pathways to glioblastomas. Neuropathology 2005, Am J Clin Pathol 2006, 126:608–615 25:1–7 25. Krypuy M, Newnham GM, Thomas DM, Conron M, Dobrovic A: High 47. Ohgaki H, Kleihues P: Genetic pathways to primary and secondary resolution melting analysis for the rapid and sensitive detection of glioblastoma. Am J Pathol 2007, 170:1445–1453 mutations in clinical samples: KRAS codon 12 and 13 mutations in 48. Watanabe K, Sato K, Biernat W, Tachibana O, von Ammon K, Ogata non-small cell lung cancer. BMC Cancer 2006, 6:295 N, Yonekawa Y, Kleihues P, Ohgaki H: Incidence and timing of p53 26. Do H, Krypuy M, Mitchell PL, Fox SB, Dobrovic A: High resolution mutations during astrocytoma progression in patients with multiple melting analysis for rapid and sensitive EGFR and KRAS mutation biopsies. Clin Cancer Res 1997, 3:523–530 detection in formalin fixed paraffin embedded biopsies. BMC Cancer 49. Parsons DW, Jones S, Zhang X, Lin JC-H, Leary RJ, Angenendt P, 2008, 8:142 Mankoo P, Carter H, Siu IM, Gallia GL, Olivi A, McLendon R, Rasheed 27. Krypuy M, Ahmed AA, Etemadmoghadam D, Hyland SJ, deFazio A, BA, Keir S, Nikolskaya T, Nikolsky Y, Busam DA, Tekleab H, Diaz LA, Fox SB, Brenton JD, Bowtell DD, Dobrovic A: High resolution melting Jr., Hartigan J, Smith DR, Strausberg RL, Marie SKN, Shinjo SMO, for mutation scanning of TP53 exons 5-8. BMC Cancer 2007, 7:168 Yan H, Riggins GJ, Bigner DD, Karchin R, Papadopoulos N, Parmi- 28. Fukui T, Ohe Y, Tsuta K, Furuta K, Sakamoto H, Takano T, Nokihara giani G, Vogelstein B, Velculescu VE, Kinzler KW: An integrated H, Yamamoto N, Sekine I, Kunitoh H, Asamura H, Tsuchida T, Kaneko genomic analysis of human glioblastoma multiforme. Science 2008, M, Kusumoto M, Yamamoto S, Yoshida T, Tamura T: Prospective 321:1807–1812 study of the accuracy of EGFR mutational analysis by high-resolution 50. Kraus-Ruppert R, Laissue J, Bürki H, Odartchenko N: Proliferation melting analysis in small samples obtained from patients with non- and turnover of glial cells in the forebrain of young adult mice as small cell lung cancer. Clin Cancer Res 2008, 14:4751–4757 studied by repeated injections of H-thymidine over a prolonged pe- 29. Milbury CA, Li J, Makrigiorgos GM: COLD-PCR-enhanced high-res- riod of time. J Compar Neurol 1973, 148:211–216 olution melting enables rapid and selective identification of low-level 51. Baldus SE, Schaefer K-L, Engers R, Hartleb D, Stoecklein NH, Gab- unknown mutations. Clin Chem 2009, 55:2130–2143 bert HE: Prevalence and heterogeneity of KRAS. BRAF, and PIK3CA 30. Li J, Wang L, Mamon H, Kulke MH, Berbeco R, Makrigiorgos GM: mutations in primary colorectal adenocarcinomas and their corre- Replacing PCR with COLD-PCR enriches variant DNA sequences and sponding metastases. Clin Cancer Res 2010, 16:790–799 redefines the sensitivity of genetic testing. Nat Med 2008, 14:579–584 52. Losi L, Baisse B, Bouzourene H, Benhattar J: Evolution of intratumoral 31. Li J, Milbury CA, Li C, Makrigiorgos GM: Two-round coamplification at genetic heterogeneity during colorectal cancer progression. Carci- lower denaturation temperature-PCR (COLD-PCR)-based sanger se- nogenesis 2005, 26:916–922 quencing identifies a novel spectrum of low-level mutations in lung 53. Barnetson R, Jass J, Tse R, Eckstein R, Robinson B, Schnitzler M: adenocarcinoma. Hum Mutat 2009, 30:1583–1590 Mutations associated with microsatellite unstable colorectal carcino- 32. Zuo Z, Soape M, Doan S, Chandra P, Galbincea J, Barkoh B, Luthra mas exhibit widespread intratumoral heterogeneity. Genes Chromo- R: Application of COLD-PCR for improved detection of KRAS muta- somes Cancer 2000, 29:130–136 tions in clinical samples. Mod Pathol 2009, 22:1773 54. Giaretti W, Macciocu B, Geido E, Hermsen MAJA, Postma C, Baak 33. Li J, Wang L, Janne PA, Makrigiorgos GM: Coamplification at lower JPA, Williams RA, Meijer GA: Intratumor heterogeneity of k-ras and denaturation temperature-PCR increases mutation-detection selectiv- p53 mutations among human colorectal adenomas containing early ity of TaqMan-based real-time PCR. Clin Chem 2009, 55:748–756 cancer. Analyt Cellular Pathol 2000, 21:49–57 34. Maekawa M, Taniguchi T, Hamada E, Takeshita A: Efficiency of 55. Greenman C, Stephens P, Smith R, Dalgliesh GL, Hunter C, Bignell G, COLD-PCR for enrichment of K-ras mutation; a proof by use of SSCP Davies H, Teague J, Butler A, Stevens C, Edkins S, Meara S, Vastrik analysis. Clin Chem 2009, 55:A219-A219 I, Schmidt EE, Avis T, Barthorpe S, Bhamra G, Buck G, Choudhury B, 35. Delaney D, Diss TC, Presneau N, Hing S, Berisha F, Idowu BD, Clements J: Patterns of somatic mutation in human cancer genomes. O’Donnell P, Skinner JA, Tirabosco R, Flanagan AM: GNAS1 muta- Nature 2007, 446:153–158 tions occur more commonly than previously thought in intramuscular 56. Campbell PJ, Pleasance ED, Stephens PJ, Dicks E, Rance R, Good- myxoma. Mod Pathol 2009, 22:718–724 head I, Follows GA, Green AR, Futreal PA, Stratton MR: Subclonal 36. Luthra R, Zuo Z: COLD-PCR finds hot application in mutation analy- phylogenetic structures in cancer revealed by ultra-deep sequenc- sis. Clin Chem 2009, 55:2077–2078 ing. Proc Natl Acad Sci 2008, 105:13081–13086 37. Fredriksson S, Baner J, Dahl F, Chu A, Ji H, Welch K, Davis RW: 57. Jones S, Chen W-d, Parmigiani G, Diehl F, Beerenwinkel N, Antal T, Multiplex amplification of all coding sequences within 10 cancer Traulsen A, Nowak MA, Siegel C, Velculescu VE, Kinzler KW, Vogelstein genes by Gene-Collector. Nucl Acids Res 2007, 35:e47 B, Willis J, Markowitz SD: Comparative lesion sequencing provides 38. Varley JM: Germline TP53 mutations and Li-Fraumeni syndrome. Hum insights into tumor evolution. Proc Natl Acad Sci 2008, 105:4283–4288 Mutat 2003, 21:313–320 58. Loeb LA, Bielas JH, Beckman RA, Bodmer IW: Cancers exhibit a mutator 39. Milbury CA, Li J, Makrigiorgos GM: PCR-based methods for the enrich- phenotype: clinical implications. Cancer Res 2008, 68:3551–3557 ment of minority alleles and mutations. Clin Chem 2009, 55:632–640 59. Torkamani A, Verkhivker G, Schork NJ: Cancer driver mutations in 40. Li J, Makrigiorgos GM: COLD-PCR: a new platform for highly im- protein kinase genes. Cancer Lett 2009, 281:117–127 proved mutation detection in cancer and genetic testing. Biochemical 60. Huang C, Taki T, Adachi M, Konishi T, Higashiyama M, Miyake M: Society Transactions 2009, 37:427–432 Mutations in exon 7 and 8 of p53 as poor prognostic factors in patients 41. Olivier M, Eeles R, Hollstein M, Khan MA, Harris CC, Hainaut P: The with non-small cell lung cancer. Oncogene 1998, 16:2469–2477 IARC TP53 database: new online mutation analysis and recommen- 61. Ahrendt SA, Hu Y, Buta M, McDermott MP, Benoit N, Yang SC, Wu L, dations to users. Hum Mutat 2002, 19:607–614 Sidransky D: p53 mutations and survival in stage I non-small-cell lung 42. Sjoblom T, Jones S, Wood LD, Parsons DW, Lin J, Barber TD, Man- cancer: results of a prospective study. J Natl Cancer Inst 2003, delker D, Leary RJ, Ptak J, Silliman N, Szabo S, Buckhaults P, Farrell 95:961–970 C, Meeh P, Markowitz SD, Willis J, Dawson D, Willson JKV, Gazdar 62. Wallace-Brodeur RR, Lowe SW: Clinical implications of p53 muta- AF, Hartigan J, Wu L, Liu C, Parmigiani G, Park BH, Bachman KE, tions. Cell Mol Life Sci 1999, 55:64–75