Human Cancer Biology
`
`Noninvasive Detection of EGFR T790M in Gefitinib or Erlotinib
`Resistant Non ^ Small Cell Lung Cancer
`Yanan Kuang,1Andrew Rogers,1Beow Y. Yeap,5 Lilin Wang,2 Mike Makrigiorgos,2 Kristi Vetrand,3,4
`Sara Thiede,3,4 Robert J. Distel,1and Pasi A. Ja« nne1,3,4,6
`
`Abstract Purpose: Tumors from 50% of epidermal growth factor receptor (EGFR) mutant non ^ small cell
`lung cancer patients that develop resistance to gefitinib or erlotinib will contain a secondary EGFR
`T790M mutation. As most patients do not undergo repeated tumor biopsies we evaluated wheth-
`er EGFR T790M could be detected using plasma DNA.
`Experimental Design: DNA from plasma of 54 patients with known clinical response to gefiti-
`nib or erlotinib was extracted and used to detect both EGFR-activating and EGFR T790M muta-
`tions. Forty-three (80%) of patients had tumor EGFR sequencing (EGFR mutant/wild type:
`30/13) and seven patients also had EGFR T790M gefitinib/erlotinib-resistant tumors. EGFR
`mutations were detected using two methods, the Scorpion Amplification Refractory Mutation
`System and the WAVE/Surveyor, combined with whole genome amplification.
`Results: Both EGFR-activating and EGFR T790M were identified in 70% of patients with known
`tumor EGFR-activating (21 of 30) or T790M (5 of 7) mutations. EGFR T790M was identified
`from plasma DNA in 54% (15 of 28) of patients with prior clinical response to gefitinib/erlotinib,
`29% (4 of 14) with prior stable disease, and in 0% (0 of 12) that had primary progressive disease
`or were untreated with gefitinib/erlotinib.
`Conclusions: EGFR T790M can be detected using plasma DNA from gefitinib- or erlotinib-
`resistant patients.This noninvasive method may aid in monitoring drug resistance and in directing
`the course of subsequent therapy.
`
`Epidermal growth factor receptor (EGFR) tyrosine kinase
`inhibitors (TKI) are effective therapies for non – small cell lung
`cancer (NSCLC) patients with activating EGFR mutations.
`Several prospective clinical trials treating chemotherapy-naBve
`patients with EGFR mutations with gefitinib or erlotinib have
`been reported to date (1 – 6). Cumulatively, these studies have
`prospectively identified and treated over 200 patients with
`EGFR mutations. Together they show radiographic response
`rates ranging from 60% to 82% and median times to
`
`Authors’ Affiliations: 1Translational Research Laboratory, Center for Clinical and
`Translational Research, 2Department of Radiation Oncology, 3Lowe Center for
`Thoracic Oncology, and 4Department of Medical Oncology, Dana-Farber Cancer
`Institute, 5Department of Medicine, Massachusetts General Hospital and Harvard
`Medical School, and 6Department of Medicine, Brigham and Women’s Hospital and
`Harvard Medical School, Boston, Massachusetts
`Received 10/7/08; revised 11/16/08; accepted 12/2/08; published OnlineFirst
`4/7/09.
`Grant support: NIH RO1CA114465-03 (P.A. Ja« nne, B.Y. Yeap), National Cancer
`Institute Lung SPORE P50CA090578 (P.A. Ja« nne, B.Y. Yeap), the Hazel and
`Samuel Bellin research fund (P.A. Ja« nne), and Department of Defense W81
`XOSHO6-1-0303 (P.A. Ja« nne).
`The costs of publication of this article were defrayed in part by the payment of page
`charges. This article must therefore be hereby marked advertisement in accordance
`with 18 U.S.C. Section 1734 solely to indicate this fact.
`Note: Y. Kuang and A. Rogers contributed equally to this article.
`Requests for reprints: Pasi A. Ja« nne, Lowe Center for Thoracic Oncology, Dana-
`Farber Cancer Institute, D820, 44 Binney Street, Boston, MA 02115. Phone: 617-
`632-6036; Fax: 617-582-7683; E-mail: pjanne@partners.org.
`F 2009 American Association for Cancer Research.
`doi:10.1158/1078-0432.CCR-08-2592
`
`progression of 9.4 to 13.3 months in the patients treated with
`gefitinib and erlotinib. These outcomes are 3- to 4-folder
`greater than that observed with platin-based chemotherapy (20-
`30% and 3-4 months, respectively) for advanced NSCLC (7).
`Unfortunately despite these benefits in EGFR -mutant
`NSCLC, all patients will ultimately develop progressive tumor
`growth while receiving gefitinib or erlotinib treatment. Two
`different mechanisms of acquired resistance in EGFR-mutant
`NSCLC patients have thus far been identified. These include a
`secondary mutation in EGFR (EGFR T790M) found in f50%
`of those with acquired resistance and MET amplification in
`f20% of patients (8 – 11). The therapeutic strategies for
`patients with these resistance mechanisms are also different.
`Irreversible EGFR inhibitors are effective in preclinical models
`at inhibiting the growth of EGFR T790M containing tumors
`in vitro and in vivo (12, 13). Several clinical trials involving
`irreversible EGFR inhibitors have now been initiated. However,
`whether these agents are effective clinically in gefitinib- and
`erlotinib-resistant NSCLC patients remains to be determined.
`Furthermore, if these agents are clinically effective, it will be
`important
`to determine the relationship to the presence/
`absence of EGFR T790M mutation. Unfortunately very few
`patients undergo repeated tumor biopsies at the time when
`resistance develops to help guide appropriate therapeutic
`choices. Thus, there is a need to develop noninvasive methods
`to identify these resistance mechanisms.
`A limited number of prior studies have evaluated the ability
`to detect EGFR-activating mutations from serum DNA of
`NSCLC patients treated with gefitinib (14, 15). The largest of
`
`Clin Cancer Res 2009;15(8) April 15, 2009
`
`2630
`
`www.aacrjournals.org
`
`00001
`
`EX1060
`
`

`

`Translational Relevance
`
`The epidermal growth factor receptor (EGFR) tyrosine
`kinase inhibitors (TKI) gefitinib and erlotinib are effective
`therapies for patients with non ^ small cell lung cancer
`(NSCLC) that harbor activating mutations in EGFR. How-
`ever, all patients ultimately develop progressive disease
`(acquired resistance) while receiving treatment with gefiti-
`nib or erlotinib. The cause of acquired resistance in 50% of
`patients is a secondary EGFR mutation (EGFR T790M).
`Second-generation EGFR TKIs are now entering clinical de-
`velopment that can inhibit the growth of cancers with
`EGFR T790M and may be clinically effective. Very few
`patients, however, undergo repeated tumor biopsies at the
`time of developing acquired resistance. In this study we
`identify both EGFR-activating and the EGFR T790M resis-
`tance mutation from plasma DNA derived from patients that
`have clinically developed resistance to gefitinib or erlotinib.
`This noninvasive method may help identify NSCLC patients
`who may benefit from second-generation EGFR kinase
`inhibitors.
`
`these to date examined 42 NSCLC patients treated with
`gefitinib. EGFR-activating mutations were detected in 8 tumor
`specimens and 6 of the 8 mutations were correctly identified
`from serum DNA (15). None of the studies to date have
`specifically examined for EGFR T790M. This may be even
`harder to detect than an EGFR-activating mutation as EGFR
`T790M can sometimes represent a minor allele which may be
`missed by direct DNA sequencing – based methods (16).
`In this study we examined the ability to detect EGFR T790M
`from plasma DNA from NSCLC patients that had clinically
`developed acquired resistance to gefitinib or erlotinib. We
`examine different methods of mutation detection and evaluate
`the benefits of whole genome amplification as a method to
`increase detection sensitivity.
`
`Materials and Methods
`
`Patients. From October 2006 to April 2008 patients with advanced
`NSCLC were identified using an institutional review board – approved
`protocol from the Thoracic Oncology clinic at the Dana Farber Cancer
`Institute. Only patients that had previously received single-agent
`gefitinib or erlotinib therapy and were at the time of the study off
`therapy were included in the study. In addition, patients were included
`if their clinical response, as defined by Response Evaluation Criteria in
`Solid Tumors, to gefitinib and erlotinib was known; they were willing
`to donate blood on one or more occasions; and they were receiving
`their treatment at Dana Farber Cancer Institute (17). Patients with
`known EGFR tumor genotype (mutant or wild type) were included only
`if they met the other criteria. Using these criteria we identified 50
`patients previously treated with gefitinib (n = 17) or erlotinib (n = 33);
`28 had a prior clinical partial response, 14 had prior stable disease, and
`8 had primary progressive disease. In addition we included four
`randomly selected advanced NSCLC patients as negative controls who
`fit the inclusion criteria but had not received any therapy with either
`gefitinib or erlotinib or with any other EGFR-directed agent. Thirty
`patients had known tumor EGFR-activating mutations. All patients
`provided written informed consent and the studies were approved by
`the Dana Farber Cancer Institute Institutional Review Board.
`
`Plasma Detection of EGFR T790M
`
`tumor specimens were
`Tumor mutation detection. Pretreatment
`analyzed for an EGFR mutation using either direct DNA sequencing
`(n = 43) or our previously described DNA endonuclease – based
`method (18). Seven patients had gefitinib or erlotinib posttreatment
`specimens that contained an EGFR T790M mutation and all were
`detected by direct sequencing. All detected mutations were indepen-
`dently confirmed.
`Blood sample collection and DNA extraction. Blood samples (average
`5 mL each) were collected in BD Vacutainer CPT Cell Preparation Tube
`with Sodium Heparin (BD). Plasma was isolated according to the
`manufacturer’s specifications and stored at -80jC until use. Plasma DNA
`was extracted using QIAamp DNA Micro Kit (Qiagen). DNA was eluted
`in 100 AL of Qiagen Buffer AE. In the DNA extraction optimization
`experiments two additional methods [Promega Wizard, (Promega) and
`NucleoSpin Plasma XS (Macherey-Nagel)] were also evaluated and used
`according to the manufacturer’s recommended specifications.
`Whole genome amplification. For whole genome amplification
`plasma DNA was processed either by a blunt-end ligation method
`described previously (19) or by an alternative method that favors the
`amplification of small,
`tumor-derived DNA (20). Whole genome
`amplification was carried out using GenomiPhi V2 DNA Amplification
`Kit (GE Healthcare).
`Plasma DNA quantification using Alu qPCR. Primer sequences for
`Alu 115bp and Alu 247 bp fragments were previously described (21).
`Standard curve was constructed using serial diluted female genomic DNA
`(Promega; 0.01-100 pg DNA). Male genomic DNA (Promega) was used
`as a calibrator in the assay. The cycling conditions were 95jC for 10 min,
`followed by 40 cycles of 95jC for 30 sec, 64jC for 30 sec, and 72jC for
`30 sec. Reactions were run on an ABI 7500Fast real-time PCR instrument.
`EGFR mutation analysis by Scorpion Amplification Refractory Mutation
`system Real-time PCR. EGFR mutation detection of the common
`EGFR-activating mutations (del E746_A750 and L858R) or the EGFR
`T790M resistance mutation were done using the EGFR Scorpion
`Amplification Refractory Mutation system (SARMS) technology (DxS
`Ltd.) as previously described (15). One microliter of plasma-derived or
`whole genome amplified DNA was added to 24 AL of master mix
`prepared according to manufacturer’s instructions. The real-time PCR
`reactions were run on an ABI 7500Fast System and according to the
`manufacturer’s recommended conditions. Comparative threshold
`values were calculated using 7500Fast System SDS Software. Positive
`samples fell into the window between the comparative threshold of the
`control assay, and the background comparative threshold and cutoff
`values were determined according to the manufacturer’s instructions.
`EGFR mutation analysis by WAVE/SURVEYOR. EGFR exons 18
`through 21 were PCR-amplified using primers that flank the exonic
`regions. For the detection of
`insertion/deletion mutations, PCR
`products were loaded on to the WAVE system (Transgenomic Inc.)
`and resolved at 50jC. For the detection of point mutations, PCR
`products were subjected to enzymatic digestion using the SURVEYOR
`enzyme at 42jC, and the resulting products resolved on the WAVE at
`50jC. Detailed protocols for exon-specific PCR and WAVE analysis were
`described previously (18).
`Statistical analysis. Fisher’s exact test was used to compare the effect
`of whole genome amplification on detection of EGFR mutations and to
`assess the association between EGFR mutation status and clinical
`response. Data were analyzed on a per patient basis. The Wilcoxon
`rank-sum test was used to compare the differences in time between the
`development of resistance and collection of plasma DNA in patients
`with and without EGFR T790M. All the exact P values were based on a
`two-sided hypothesis test and were computed using StatXact verson 6.1
`(Cytel Software Corp.).
`
`Results
`
`Alu real-time PCR and optimization of plasma DNA
`extraction. We first established a DNA extraction procedure
`
`www.aacrjournals.org
`
`2631
`
`Clin Cancer Res 2009;15(8) April 15, 2009
`
`00002
`
`

`

`Human Cancer Biology
`
`that resulted in the greatest yield of tumor-derived circulating
`DNA. Prior studies suggest that most malignant tumor-derived
`DNA varies in size (median, 544 bp; range, 185-926 bp)
`whereas apoptotic DNA from normal cells is more uniformly
`sized as 185 to 200 bp fragments (22). We thus adopted a
`previously developed real-time PCR assay to determine the ratio
`between Alu sequences of 115 bp (Alu115) and 247 bp
`(Alu247), and used it as the indication of DNA integrity (21).
`Alu115 represents both short, apoptotic DNA fragments and
`tumor-derived fragments (total circulating DNA), whereas
`Alu247 represents tumor-derived DNA alone. The ratio of
`Alu247/Alu115 was used to calculate the percentage of tumor
`DNA in total circulating DNA.
`For these initial studies we evaluated plasma DNA from
`seven patients. These seven samples were not included in the
`larger study, nor were they subjected to whole genome
`amplification. Plasma DNA was extracted in parallel from each
`sample using three independent protocols: Qiagen, Promega
`Wizard, and NucleoSpin Plasma XS. The Alu 247/Alu115 ratio
`was determined for each sample prior to EGFR mutation
`analysis, and DNA input for mutational analyses was normal-
`ized to total circulating DNA (Alu115). The median total circu-
`lating DNA (Alu115) yields of the Qiagen, Promega Wizard,
`and NucleoSpin methods were 0.064 ng/AL, 0.021 ng/AL, and
`0.086 ng/AL, respectively. The median Alu247/Alu115 ratio
`obtained using Qiagen, Promega Wizard, and NucleoSpin
`methods were 50.9%, 59.4%, and 10.9%, respectively. The
`DNA derived using the Qiagen extraction method was success-
`fully amplified 100% of the time using both the SARMS and the
`WAVE/Surveyor methods. In contrast, DNA derived using the
`Promega Wizard or Nucleospin methods successfully amplified
`in only 75% or 67% of the reactions, respectively. Based on
`the high Alu247/Alu115 ratio and the ability to successfully
`amplify the DNA we used the Qiagen DNA extraction method
`for all subsequent studies.
`Patient characteristics. Fifty-four patients were enrolled in
`this study (Table 1). Fifty of the 54 patients (93%) had received
`prior treatment with either gefitinib (n = 17; 31.4%) or erlotinib
`(n = 33; 61.1%) and all had developed disease progression at
`
`Table 1. Patient characteristics
`
`No. of patients
`
`Gender
`Male
`Female
`EGFR TKI treatment
`Gefitinib
`Erlotinib
`None
`Response to prior EGFR TKI treatment
`Partial response
`Stable disease
`Progressive disease
`Not treated
`Tumor EGFR mutation
`Exon 19 deletion
`L858R
`L861Q
`Exon 20 insertion
`Wild type
`Unknown
`
`n = 54
`
`10 (18.5%)
`44 (81.5%)
`
`17 (31.4%)
`33 (61.1%)
`4 (7.5%)
`
`28 (56%)
`14 (28%)
`8 (16%)
`4 (7.5%)
`
`20 (37.0%)
`7 (12.9%)
`1 (1.9%)
`2 (3.7%)
`13 (24.1%)
`11 (20.4%)
`
`the time blood specimens were obtained. Four patients (7.5%)
`were never treated with either gefitinib or erlotinib and served as
`negative controls. The best response to prior therapy was partial
`response with 28 patients (56 %), followed by stable disease
`with 14 patients (28 %) and progressive disease with 8 patients
`(16 %). Tumor EGFR mutation status, obtained from baseline
`pre-gefitinib or -erlotinib treatment specimens, was available in
`43 of 54 (80%) of patients (Table 1).
`Seventy-six plasma specimens were obtained from the 54
`patients (median number per patient, 1; range, 1-5) and were
`used for DNA extraction. All DNA specimens were subjected to
`whole genome amplification, with the DNA quantified before
`and after whole genome amplification. The median concen-
`trations were 0.252 ng/AL (range, 0.023-100.1 ng/AL) for
`unamplified plasma DNA samples and 52.3 ng/AL (range, 9.9-
`162.7 ng/AL) for whole genome – amplified DNA samples. Both
`unamplified plasma DNA and whole genome – amplified DNA
`specimens were used for subsequent genotyping studies.
`EGFR mutation detection. We used two different methods,
`SARMS and WAVE/Surveyor, to detect EGFR activation and
`resistance mutations from plasma DNA. Both SARMS and
`WAVE/Surveyor technologies are PCR-based methods for
`mutation detection. SARMS uses a Scorpions primer/probe in a
`real-time PCR setting. Short probes allow greater allelic
`specificity and a lower background. The WAVE/Surveyor method
`combines standard PCR followed by an endonuclease digestion
`(Surveyor) that targets wild-type/mutant heteroduplexes. The
`resulting products are resolved on the WAVE HS system (18).
`We first tested the sensitivity and specificity of detecting
`EGFR T790M with the SARMS assay using NSCLC cell lines
`with known EGFR T790M mutation status (H1975, H820, and
`H3255 GR, all known to contain an EGFR T790M mutation,
`and A549 that does not contain an EGFR T790M mutation).
`Using this assay, we determined the EGFR T790M allele
`frequencies for each of the cell lines: H1975 at 55%, H820 at
`7%, H3255 GR at 2%, and A549 at 0%. These results were
`consistent with our own previous genotyping results using
`WAVE/Surveyor and published data (16, 23).
`Next, we determined whether we could detect EGFR-
`activating mutations and the T790M resistance mutation in
`patient-derived plasma DNA. Based on previous reports
`(14, 15) and our determination of median patient plasma
`DNA concentration (0.252 ng/AL, which is equivalent to a
`median of 43 genome copies) in our sample cohort, we used
`1 AL of patient plasma DNA in both the SARMS and WAVE/
`Surveyor assays. Figure 1 depicts the detection of the EGFR
`T790M mutation in a representative patient plasma DNA
`sample using both the WAVE/Surveyor and the SARMS
`methods. Using the SARMS assay we detected 12 patients with
`EGFR del E746_A750, 7 patients with L858R, and 8 patients
`with EGFR T790M mutations. All plasma DNA samples were
`also independently PCR-amplified and screened for EGFR exon
`19 to 21 mutations using WAVE/Surveyor as previously
`described (18). At the time of the study, the Scorpions assays
`were only available to detect two EGFR-activating mutations
`(del E746_A750 and L858R) and the EGFR T790M resistance
`mutation. Thus, we used the WAVE/Surveyor method to
`evaluate for the remaining EGFR mutations and also as a
`complementary approach to the SARMS assays. Using the
`WAVE/Surveyor method we detected EGFR exon 19 deletion
`mutations in 25 patients, no exon 20 insertion mutations,
`
`Clin Cancer Res 2009;15(8) April 15, 2009
`
`2632
`
`www.aacrjournals.org
`
`00003
`
`

`

`Plasma Detection of EGFR T790M
`
`Figure 1. Detection of EGFR T790M using
`WAVE/Surveyor and SARMS. A, detection of EGFR
`T790M from the H1975 (EGFR L858R/T790M) cell
`line (top) and plasma DNA from patient 35 (bottom).
`Exon 20 of EGFR was amplified by PCR, the resulting
`product digested with Surveyor and analyzed using the
`WAVE-HS system (Materials and Methods). In the
`presence of EGFR T790M, two fragments (asterisk) are
`generated by Surveyor digestion (solid lines) from the
`positive control (H1975) and patient 35. The wild-type
`control (A549; dashed line) is uncut. B, SARMS
`analysis of EGFR T790M. Included are positive and
`negative control DNA samples and plasma DNA from
`patients 35 and 37. The horizontal dotted line represents
`the threshold. DNA from the negative control and patient
`37 do not amplify above the threshold whereas DNA
`from the positive control and patient 35 both cross the
`threshold in the linear portion of the assay. Fluorescence
`was measured quantitatively in relative fluorescence
`units.
`
`EGFR L858R mutations in 2 patients, and EGFR T790M
`mutations in 4 patients with. Of the 25 patients with EGFR
`exon 19 deletion mutations detected by WAVE/Surveyor, 11
`were exon 19 deletions other
`than the del E746_A750
`mutation. Such deletions were not a part of the SARMS assay.
`We compared the findings between these two mutation
`detection methods. The SARMS and WAVE/Surveyor detected
`EGFR del E746_A750 in a combined 15 patients, L858R in a
`combined 7 patients, and T790M in a combined 9 patients,
`with concordance rates of 73% (11 of 15), 28% (2 of 7), and
`33% (3 of 9), respectively (Table 2).
`Impact of whole genome amplification on plasma DNA-based
`mutation detection. We further investigated whether whole
`genome amplification facilitated the detection of additional
`EGFR mutations from plasma DNA. Whole genome – amplified
`DNA samples were screened for mutations in EGFR exons
`19, 20, and 21 in an identical fashion to non-whole genome –
`amplified samples using both SARMS and WAVE/Surveyor
`assays. Using the SARMS assay we detected 13 additional
`EGFR mutations: 2 patients with EGFR del E746_A750, 1 with
`L858R, and 10 patients with EGFR T790M not detected from
`the plasma DNA. The WAVE/Surveyor method detected 7
`additional patients with EGFR mutations: 3 with EGFR exon 19
`deletions, 1 patient with an EGFR L858R mutation, and 3
`patients EGFR T790M mutations not detected from the plasma
`DNA (Table 2). One of the additional EGFR exon 19 deletion
`
`mutations detected by the WAVE/Surveyor method one was a
`non – exon 19 del E746_A750 mutation, thus not assayed by
`SARMS method. Nine of 10 (90%) of the patients in which we
`detected an EGFR T790M using the SARMS assay also
`contained a concurrent EGFR-activating mutation whereas this
`occurred in 67% (2 of 3) of EGFR T790M containing patient
`specimens using the WAVE/Surveyor method.
`
`Table 2. Comparison of SARMS and WAVE/
`Surveyor methods in detecting EGFR exon 19
`(del E746_A750), L858R, and T790M mutations
`from gefitinib/erlotinib-treated NSCLC patients
`
`EGFR mutation
`
`Del E746_A750 L858R T790M
`
`Plasma DNA alone
`15
`Total positive patients
`12
`SARMS-positive
`14
`WAVE/Surveyor-positive
`11/15
`Concordance
`Plasma DNA and whole genome amplified
`Total positive patients
`18
`SARMS-positive
`14
`WAVE/Surveyor-positive
`16
`Concordance
`12/18
`
`7
`7
`2
`2/7
`
`8
`8
`3
`3/8
`
`9
`8
`4
`3/9
`
`19
`18
`7
`6/19
`
`NOTE: The data are displayed on a per patient basis.
`
`www.aacrjournals.org
`
`2633
`
`Clin Cancer Res 2009;15(8) April 15, 2009
`
`00004
`
`

`

`Human Cancer Biology
`
`We next combined the results obtained from non-whole
`genome – amplified and whole genome – amplified samples and
`compared the findings between SARMS and WAVE/Surveyor
`detection methods. The SARMS and WAVE/Surveyor detected
`EGFR del E746_A750 in a combined 18 patients, L858R in a
`combined 8 patients, and T790M in a combined 19 patients,
`with concordance rates of 67% (12 of 18), 38% (3 of 8), and
`32% (6 of 19), respectively (Table 2).
`Overall the effect of whole genome amplification seemed to
`have the greatest effect on the detection of EGFR T790M
`(Table 2). For EGFR del E746_A750 and L858R, whole
`genome amplification identified only 4 additional patients
`with mutations whereas for EGFR T790M whole genome
`amplification resulted in the identification of 10 additional
`patients (P = 0.011).
`Concordance of primary tumor sequencing and clinical response
`with detection of plasma EGFR mutations. We compared the
`EGFR-activating mutation detected in plasma DNA with the
`tumor EGFR-activating mutation. For these studies we com-
`bined the findings from the SARMS and WAVE/Surveyor
`methods and included findings from the whole genome –
`amplified specimens. Tumor EGFR mutation status was known
`in 43 of 54 (80%) and not available in 11 of 54 (20%) of
`patients. Thirteen (13) of the 43 patients were EGFR wild-type
`(30%) whereas 30 (70%) had an EGFR-activating mutation in
`exons 19 to 21 (Table 1). Collectively the plasma-based
`detection methods identified 29 of 54 (54%) of patients as
`having an EGFR-activating mutation whereas 25 of 54 (46%)
`were EGFR wild-type. In the 43 patients whose tumor EGFR
`mutation status was known, we identified EGFR mutations
`from plasma DNA in 21 of 30 patients (70%). The overall
`concordance of tumor EGFR mutation with plasma EGFR
`mutation was 74% (32 of 43; Table 3). We also examined
`concordance as a function of the specific type of mutation
`(exon 19 deletion versus L858R). In the patients with a known
`tumor exon 19 deletion mutation there was an 85% (17 of 20)
`concordance with the plasma EGFR mutation whereas in those
`with a tumor L858R mutation the concordance rate was only
`29% (2 of 7) with the plasma EGFR mutation (P = 0.011).
`We also analyzed the findings based on prior response to
`therapy (Tables 1 and 3). EGFR-activating mutations were
`
`Table 3. Summary of detecting EGFR-activating
`mutations from plasma DNA
`
`Plasma DNA
`
`Mutation
`
`No Mutation
`
`Tumor tissue
`30
`Mutation
`13
`No mutation
`11
`Not available
`Response to prior EGFR TKI therapy
`Partial response
`28
`Stable disease
`14
`Progressive disease
`8
`Untreated
`4
`
`21
`2
`6
`
`23
`4
`2
`0
`
`9
`11
`5
`
`5
`10
`6
`4
`
`Table 4. Comparison of NSCLC patient clinical
`response to prior EGFR TKI therapy and known
`EGFR T790M– containing tumors with detection of
`EGFR T790M using plasma DNA
`
`Plasma EGFR T790M
`
`Response to prior EGFR TKI Therapy
`Partial response
`28
`Stable disease
`14
`Progressive disease
`8
`Untreated
`4
`Tumor EGFR T790M
`7
`
`Yes
`
`15
`4
`0
`0
`5
`
`No
`
`13
`10
`8
`4
`2
`
`detected in plasma DNA from 23 of 28 (82%) patients with a
`complete response (CR)/partial response, 4 of 14 (28.5%)
`patients with stable disease, and 2 of 8 (25%) patients with
`progressive disease. EGFR-activating mutations detected in
`plasma DNA are associated strongly with a clinical response
`among the patients treated with gefitinib or erlotinib (P < 0.001).
`Correlation of EGFR T790M detected in plasma DNA with
`prior drug response and tumor EGFR T790M. We evaluated the
`relationship with prior clinical response to gefitinib or erlotinib
`in patients in which EGFR T790M was detected in plasma DNA.
`Prior tumor-based studies suggest that EGFR T790M can be
`detected in 50% of NSCLC patients with a prior response
`(CR or partial response) to gefitinib or erlotinib therapy
`(10, 11). EGFR T790M was detected from plasma DNA in 35%
`(19 of 54) patients in this study. In the 28 patients that had a
`prior partial response to either gefitinib or erlotinib EGFR
`T790M was detected in the plasma DNA in 15 of 28 (54%)
`patients (Table 4). EGFR T790M was detected in 5 of 7 patients
`(71%) for whom posttreatment biopsy specimens were
`available and had been confirmed to contain an EGFR
`T790M by direct sequencing (Table 4). EGFR T790M was also
`detected in 4 of 14 (29%) of patients with stable disease. One
`of the four patients had a concurrent EGFR-activating mutation
`detected from plasma DNA. EGFR T790M was detected in none
`(0 of 8; 0%) of the patients with progressive disease to gefitinib
`or erlotinib or in patients who had never been treated with
`these agents (0 of 4; 0%). Collectively, these findings show that
`the EGFR T790M mutation detected in plasma DNA is
`associated strongly with a prior clinical response to gefitinib
`or erlotinib (P = 0.004). We further evaluated the time between
`the clinical development of resistance and plasma collection and
`the presence or absence of EGFR T790M in patients with a prior
`clinical response to gefitinib or erlotinib. The time between the
`development of clinical resistance and plasma collection was
`numerically longer but not significantly different (P = 0.829;
`Wilcoxon rank-sum test) in patients in whom we did not
`identify an EGFR T790M (median, 68 days; range, 1-940 days)
`compared with those in which an EGFR T790M was identified
`from plasma DNA (median, 38 days; range, 1-817 days).
`
`Discussion
`
`NOTE: NSCLC patients broken down based on known tumor EGFR
`mutations and based on clinical outcome with prior EGFR TKI
`therapy.
`
`EGFR inhibitors are effective therapies against EGFR-mutant
`NSCLC (1 – 6). Given that only 10% to 15% of Caucasian
`NSCLC patients harbor EGFR -activating mutations,
`it
`is
`
`Clin Cancer Res 2009;15(8) April 15, 2009
`
`2634
`
`www.aacrjournals.org
`
`00005
`
`

`

`important to identify this subgroup of lung cancer patients
`(24). However, not all lung cancer diagnostic specimens are
`amenable to genotyping because they often contain only a
`limited number of cancer cells. Potential solutions to this
`barrier include developing more sensitive diagnostic methods
`and/or developing noninvasive diagnostic methods. The latter
`can also potentially be used to study secondary resistance
`alterations, such as EGFR T790M, that occur during the course
`of treatment.
`Using both SARMS and WAVE/Surveyor we detected both the
`EGFR-activating and the EGFR T790M resistance mutation
`from plasma DNA in 70% of patients in which these mutations
`were known to occur in the patients’ tumor specimens (Tables 3
`and 4). These findings are similar to two smaller studies of
`plasma-based DNA analyses of EGFR-activating mutations and
`show a 70% concordance with the primary tumor mutation
`(14, 15). However, they differ from a more recent larger study
`which used the SARMS technique which detected only 7 (39%)
`EGFR mutations in plasma DNA from 18 known EGFR-mutant
`patients (25). The higher plasma EGFR mutation detection rate
`in the current study may be a result of using two mutation
`detection methods (SARMS and WAVE/Surveyor) and combin-
`ing these with whole genome amplification. In fact, in our
`study SARMS alone detected only 14 (47 %) plasma EGFR
`mutations from 30 patients with tumor EGFR mutations (data
`not shown). Improvements in the sensitivity of genotyping,
`such as by using single-molecule sequencing or digital PCR,
`may further improve the ability to detect EGFR-activating
`mutations from plasma DNA using just a single technique as
`opposed to a combination of methods as in the current study.
`Our study also included a large number of known EGFR-
`mutant patients (30) compared with prior studies(14, 15, 25).
`Given that, we are able to compare the ability to detect the
`common EGFR mutations (exon 19 deletion and L858R) from
`plasma DNA. Intriguingly, our techniques were significantly
`more likely to detect an exon 19 deletion mutation (85%; 17 of
`20) than L858R (29%; 2 of 7) from plasma DNA. Although we
`cannot completely exclude the possibility, based on our prior
`studies suggesting a similar sensitivity for detecting EGFR exon
`19 deletions and L858R using DNA derived from tumor tissue,
`this observation is unlikely due to differences in the sensitivity
`assays in detecting exon 19 deletion and L858R mutations (18).
`It is possible that this reflects a biological difference between
`these two different EGFR mutant cancers and needs to be
`evaluated in future noninvasive DNA-based studies.
`In two patients EGFR-activating mutations were identified
`from plasma DNA but were not detected in the
`corresponding tumor specimen (Table 3). Although these
`findings may represent false positive results of our current
`technique, a further look into the details of these patients
`may suggest an alternative hypothesis. In one of the patients,
`the EGFR wild-type tumor specimen was obtained from the
`time of
`surgery following chemotherapy and radiation.
`Following tumor relapse, the patient was subsequently treated
`with gefitinib and after >24 months developed disease
`progression with new brain metastases, at which time the
`plasma sample was obtained. Analysis of DNA from the
`plasma sample showed both EGFR del E746_A750 and
`T790M mutations. EGFR sequencing of the brain metastasis
`also showed both EGFR del E746_A750 and T790M
`mutations. These findings suggest either clonal evolution of
`
`Plasma Detection of EGFR T790M
`
`for EGFR
`the patient’s NSCLC or a false negative result
`delE746_A750 from the original
`tumor
`sequencing. The
`second patient with an EGFR wild-type pretreatment tumor
`specimen was treated with erlotinib for 22 months with
`stable disease prior to developing disease progression, at
`which time an L858R mutation alone was detected from
`plasma DNA. In the pathologist’s description of her tumor it
`is noted that her
`tumor specimen contains a significant
`portion of
`inflammatory cells
`that could interfere with
`genotyping by direct sequencing. Thus,
`it
`is possible that
`the tumor EGFR sequencing represents a false negative.
`Prolonged stable disease with erlotinib treatment has
`previously been observed in patients with EGFR-activating
`mutations (26).
`Our studies detected EGFR T790M from plasma DN