The official journal of INTERNATIONAL FEDERATION OF PIGMENT CELL SOCIETIES · SOCIETY FOR MELANOMA RESEARCH
PIGMENT CELL & MELANOMA Research Detailed imaging and genetic analysis reveal a secondary BRAF L505H resistance mutation and extensive intrapatient heterogeneity in metastatic BRAF mutant melanoma patients treated with vemurafenib Marlous Hoogstraat, Christa G. Gadellaa-van Hooijdonk, Inge Ubink, Nicolle J. M. Besselink, Mark Pieterse, Wouter Veldhuis, Marijn van Stralen, Eelco F. J. Meijer, Stefan M. Willems, Michael A. Hadders, Thomas Kuilman, Oscar Krijgsman, Daniel S. Peeper, Marco J. Koudijs, Edwin Cuppen, Emile E. Voest and Martijn P. Lolkema
DOI: 10.1111/pcmr.12347 Volume 28, Issue 3, Pages 318–323 If you wish to order reprints of this article, please see the guidelines here
Supporting Information for this article is freely available here EMAIL ALERTS Receive free email alerts and stay up-to-date on what is published in Pigment Cell & Melanoma Research – click here
Submit your next paper to PCMR online at http://mc.manuscriptcentral.com/pcmr
Subscribe to PCMR and stay up-to-date with the only journal committed to publishing basic research in melanoma and pigment cell biology As a member of the IFPCS or the SMR you automatically get online access to PCMR. Sign up as a member today at www.ifpcs.org or at www.societymelanomaresarch.org
To take out a personal subscription, please click here More information about Pigment Cell & Melanoma Research at www.pigment.org
Pigment Cell Melanoma Res. 28; 318–323
SHORT COMMUNICATION
Detailed imaging and genetic analysis reveal a secondary BRAF L505H resistance mutation and extensive intrapatient heterogeneity in metastatic BRAF mutant melanoma patients treated with vemurafenib Marlous Hoogstraat1,2,*, Christa G. Gadellaa-van Hooijdonk1,2,*, Inge Ubink1,2, Nicolle J. M. Besselink1,2, Mark Pieterse1, Wouter Veldhuis3, Marijn van Stralen4, Eelco F. J. Meijer1, Stefan M. Willems2,5, Michael A. Hadders1, Thomas Kuilman6, Oscar Krijgsman6, Daniel S. Peeper6, Marco J. Koudijs1,2, Edwin Cuppen2,7, Emile E. Voest1,2 and Martijn P. Lolkema1,2 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht, The Netherlands 2 Netherlands Center for Personalized Cancer Treatment, Utrecht, The Netherlands 3 Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands 4 Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands 5 Department of Pathology, University Medical Center Utrecht, Utrecht, The Netherlands 6 Division of Molecular Oncology, Netherlands Cancer Institute, Amsterdam, The Netherlands 7 Department of Medical Genetics, University Medical Center Utrecht, Utrecht, The Netherlands
KEYWORDS BRAF/vemurafenib/intratumoral heterogeneity/therapy resistance/volumetric imaging analysis PUBLICATION DATA Received 25 August 2014, revised and accepted for publication 15 December 2014, published online 17 December 2014
CORRESPONDENCE Martijn P. Lolkema, e-mail: [email protected] doi: 10.1111/pcmr.12347 *Equal contribution.
Summary Resistance to treatment is the main problem of targeted treatment for cancer. We followed ten patients during treatment with vemurafenib, by three-dimensional imaging. In all patients, only a subset of lesions progressed. Next-generation DNA sequencing was performed on sequential biopsies in four patients to uncover mechanisms of resistance. In two patients, we identified mutations that explained resistance to vemurafenib; one of these patients had a secondary BRAF L505H mutation. This is the first observation of a secondary BRAF mutation in a vemurafenib-resistant patient-derived melanoma sample, which confirms the potential importance of the BRAF L505H mutation in the development of therapy resistance. Moreover, this study hints toward an important role for tumor heterogeneity in determining the outcome of targeted treatments.
Approximately 50% of all melanomas are driven by V600 mutations in BRAF. Vemurafenib is the first clinically approved drug that specifically inhibits activated BRAF both in vitro and in vivo (Joseph et al., 2010; Yang et al., 2010). This compound produces a response in approximately half the patients and significantly improves
progression-free and overall survival. Unfortunately, the majority of patients with metastatic melanoma inevitably develop resistance to vemurafenib after a median time of 6 months (Chapman et al., 2011; Sosman et al., 2012). Intratumoral heterogeneity and extensive genetic variation between tumors within a single patient (Campbell
Significance Improving the treatment with BRAF inhibitors for patients with metastatic melanoma now depends on treating or preventing therapy resistance. This study confirms the clinical relevance of a proposed secondary BRAF mutation in resistance to vemurafenib. This finding reinforces the current paradigm that reactivation of signaling downstream of BRAF is an important event in the development of resistance to BRAF inhibition. Moreover, this study hints toward extensive geographic and molecular heterogeneity of the treatment na€ıve melanoma. In designing our strategy to treat or prevent therapy resistance, we need take the full extent of this molecular heterogeneity into account.
318
ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Secondary BRAF mutation causes resistance to vemurafenib
et al., 2010; Gerlinger et al., 2012; Stoecklein and Klein, 2010; Swoboda et al., 2011; Vermaat et al., 2012) have been proposed as reasons for treatment failure. Clonal outgrowth of resistant populations of tumor cells is considered the cause of tumor progression. Other studies using sequential samples obtained before treatment and at time of progression show that selective pressure of treatment activates compensatory survival pathways (Ding et al., 2012; Trunzer et al., 2013). Based on these studies, we hypothesized that separate metastatic lesions within an individual patient can respond differently to vemurafenib due to intermetastatic genetic heterogeneity and clonal variations, and investigating these genetic differences can lead to an improved understanding of resistance and hence improved treatment. Indeed, heterogeneous response and progression were observed earlier in metastatic melanoma patients treated with dabrafenib + trametinib (Menzies et al., 2014), as well as heterogeneity in response mechanisms (Rizos et al., 2014; Shi et al., 2014; Van Allen et al., 2014). To test this hypothesis, we used a longitudinal imaging approach combined with genome analyses. We studied the radiological response of individual metastases and the mutation profiles of pretreatment samples and biopsy material upon progression from stage IV melanoma patients treated with vemurafenib (supplemental methods). Volumetric, three-dimensional assessments are time-consuming but have the benefit to identify growing lesions at an earlier stage, and their validity has been shown by previous studies (Schiavon et al., 2012; Tirumani et al., 2014). For the early detection of progression, two-dimensional RECIST criteria are not ideal as they induce a delay in the recognition of progression of individual sites. The geographic heterogeneity of response to BRAF inhibitors may become important as local progression is a common phenomenon (Menzies et al., 2014), and thus, early local treatment of growing lesions may be beneficial for patients. Ten patients were included in the volumetric response evaluation. Baseline characteristics of patients included in the volumetric response evaluation are listed in Table S1. Per patient, three to ten separate lesions were measured every eight to twelve weeks during treatment. At initial
evaluation, after two months of vemurafenib therapy, 45% of patients showed partial remission, 40% had stable disease, and 1 patient had progressive disease (Figure 1A). Progression-free survival (PFS) ranged from 1.8 to 21.9 months (median 7.8 months). Five patients developed progressive disease based on the growth of target lesions identified at baseline. The other patients developed new lesions or showed evident progression of non-target lesions. One patient, whose response could not be evaluated by volumetric measurement, progressed due to an evidently progressive bone lesion. In each patient, volumetric analysis revealed a heterogeneous pattern of disease progression through tumor growth combined with multiple metastases that continued to respond to treatment (Figure 1B). Moreover, by evaluating the volumetric changes per lesion over time, it was noticed that the growth rate of individual lesions could differ between lesions, as well as the moment progression becomes apparent (Figure 1C). These data show that a detailed intrapatient response evaluation reveals a fairly homogenous initial response after 2 months, but a heterogeneous growth pattern at progression. Based on the volumetric measurements, progressive lesions were identified and sequential biopsies were obtained in four patients (Figure 2A). In three of the four patients, the sequential biopsy was taken from the same lesion as the pretreatment biopsy (Table 1). All specimens contained 70–95% viable tumor cells based on histopathological examination. To identify the genetic defects in each tumor, the coding regions of 1977 cancerrelated genes were sequenced to a median coverage of 1509 from both pre- and post-treatment biopsies and a matched blood sample was analyzed as a reference (Table S2, Data S1). Between 8 and 100 coding somatic variants could be identified per tumor sample (Table S3). BRAF mutations were seen in all pretreatment and sequential biopsies, and mutations in DNA repair genes were found in three patients, including MSH3, RAD50, ERCC3, and ERCC4. Both ERCC3 and ERCC4 are associated with melanoma and other skin cancers, according to the Sanger Cancer Gene Census (Futreal et al., 2004). To investigate whether there were any
Table 1. Overview of sites of sequential melanoma biopsies
Biopsy location
Tumor percentage
Patient
Time between biopsies (weeks)
Baseline
Sequential
Baseline
Sequential
#1 #2 #3 #11
16 19 24 8
Subcutaneous lesion thorax Axillary lymph node Inguinal lymph node Subcutaneous lesion breast
Subcutaneous lesion thoraxa Axillary lymph nodea Inguinal lymph nodea Soft tissue mass Thoracic vertebra 10–11
95 80 90 90
70 80 90 90
a
Biopsy from same lesion as pretreatment biopsy.
ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
319
Hoogstraat et al.
A
B
Lesion-specific response at first evaluation
Lesion-specific response at time of progression
1400%
200%
1200%
% volume change
% volume change
100%
0%
600%
400%
200%
0%
PR
C
p5
p10
p6
p7
p1
p8
p2
p9
p4
p3
p10
p8
p9
SD
PD non-target lesions
PD
Volumetric measurements per lesion p4
Volumetric measurements per lesion p2 Lung, left lobe LN, right axilla (1) LN, right axilla (2) LN, right axilla (3)
200 Resection of 3 lymph nodes
100
0
100%
% volumetric change
300 % volumetric change
p2
p7
p4
p1
p6
p5
p3
−100%
LN, left inguinal LN, right inguinal LN, left psoas muscle LN, left para−iliac LN, mediastinal (1) LN, mediastinal (2)
50%
0%
−50%
−100%
−100 0
20
40
60
Weeks on treatment
80
0
5
10 15 20 Weeks on treatment
25
Figure 1. Topographic diversity in the development of resistance to vemurafenib. (A) Volumetric measurements of all different lesions in 10 patients treated with vemurafenib at the first evaluation (after approximately 8 weeks of treatment), compared to baseline volume. Grey box plots outline the dispersion of measurements per patient, while the black dots represent individual lesions. Data points highlighted in red indicate lesions undetectable on the CT-scans. Patients are indicated on the X-axis and sorted by initial response. (B) Volumetric measurements of the same patients at time of progression, compared to smallest volume measured for progressive lesions. All other lesions are compared to baseline. For each patient, multiple lesions were still smaller than baseline volume, indicating sensitivity of these lesions to vemurafenib. The lesion with increased volume shown for patient 1 was too small to account for progressive disease in itself, but progression was determined in combination with new lesions not shown in this graph. (C) Dynamic changes per lesion over time in 2 patients. Volume per lesion was measured every 8–12 weeks. In patient #2, a progressive lymph node was surgically removed after 32 weeks, together with two other lymph nodes nearby. After surgery, a remaining lesion in the lung started to grow. Three lymph node lesions in patient #4 continued to respond to vemurafenib until the end of treatment, while three other lesions showed progression at the fourth evaluation at 25 weeks.
genetic changes between baseline and sequential biopsies, allele frequencies of variants in either sample were compared. In addition, copy number profiles were generated from the targeted sequencing data to be able to assess BRAF amplification and to aid in the interpretation of changes in allele frequencies of somatic variants (Figure S2). By exploring mutations unique for the progressive lesions, we sought to enrich for the resistance-associated mutations while the mutations with stable frequencies were deemed more likely to represent tumor driver mutations. Recently, a mutagenesis screen of BRAF was performed to detect potential secondary mutations that cause resistance. BRAF L505H was identified as secondary mutation able to induce resistance to vemurafenib in vitro (Wagenaar et al., 2014). Moreover, a second report 320
showed the presence of the same mutation in a patientderived cell line confirming the potential role for this specific mutation in BRAF inhibitor resistance (Choi et al., 2014). The affected amino acid is located in the vicinity of the ATP-binding pocket, which is where vemurafenib engages the BRAF protein. Molecular dynamics simulations show that vemurafenib binding indeed changes in V600E/L505H mutated BRAF compared to V600E/L505 wild-type BRAF (Choi et al., 2014), providing a plausible explanation for the therapy resistance. However, the fact that other larger studies on resistance causing mutations failed to detect this specific mutation implies that the frequency of secondary BRAF mutations is less prevalent than secondary mutations in receptor tyrosine kinases such as EGFR and KIT (Antonescu et al., 2005; Balak et al., 2006; Nishida et al., 2008). Apparently, resistance ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Pretreatment biopsy Sequential biopsy
S PD
19w
Patient 2
S1
Pre-T
26w
Patient 3 Pre-T Patient 11
8w
18w
S
43w S2 PD
PD
7w
Progressive disease according to RECIST Surgery: resection of progressive lymph nodes Radiotherapy of progressive lesions
PD
0w
Period of vemurafenib treatment
B
Putative founders Putative treatment induced Other somatic coding
PRKCH
80
S
60
Pre-T
BRAF 40
Pre-T
8w
BRAF
V600E
L505H
20
PD
17w
Patient 1
Post-treatment allele frequency
S
Pre-T
0
A
100
Secondary BRAF mutation causes resistance to vemurafenib
0
20
40
60
80
100
Pre-treatment allele frequency
Figure 2. Biopsies of progressive lesions within one patient revealed distinct mechanisms of therapy resistance. (A) Timing of biopsies and vemurafenib treatment in all 4 sequenced patients. Dotted lines indicate taking of biopsies. (B) Comparison of allele frequencies of variants in the pre- and post-treatment biopsy of patient #3. Putative founder mutations based on high allele frequency in both pretreatment and sequential biopsy are highlighted in red; de novo mutations and mutations with increased allele frequency in the sequential biopsy are highlighted in blue.
to BRAF inhibitors can be achieved through more diverse mechanisms. Other recent studies describing resistance against BRAF inhibition indicate several recurrent genetic mechanisms, including reactivation of the MAPK pathway through MAP2K1, MAP2K2, or NRAS mutations (Emery et al., 2009; Greger et al., 2012; Nazarian et al., 2010; Wagle et al., 2014), as well as activating mutations in PIK3CA (Mao et al., 2013; Romano et al., 2013) or other PI3K pathway alterations (Van Allen et al., 2014) and loss of function or aberrant splicing mutations in RB1, PTEN (Lito et al., 2013), or NF1 (Whittaker et al., 2013). In patient #3, the secondary BRAF L505H mutation described earlier was detected in the sequential biopsy (Figure 2B). Sanger sequencing of cDNA showed that both BRAF mutations were located on the same allele (Data S1, Figure S1): 53% (n = 33) of traces showed the V600E only, 23% (n = 14) showed both the V600E and the L505H mutation, in 24% (n = 15), BRAF was wild type, and in 0% (n = 0) of the traces, the L505H was found without the presence of the V600E. Our copy number analysis indicates that the pretreatment and sequential biopsies from this patient both contain four copies of chromosome 7 (Figure S2). Combined with the mutation allele frequencies and frequencies from the cDNA analysis, we conclude that both samples contain two copies of V600E-mutated BRAF from which one obtained the secondary L505H mutation. This suggests that only one double-mutant BRAF copy is sufficient to introduce vemurafenib resistance, consistent with a previous report (Choi et al., 2014). Furthermore, a missense mutation in PRKCH was not present in the pretreatment tumor sample but was detected in the post-treatment sample with an allele frequency of 95% (Figure 2B). However, as the copy number analysis also shows a deletion at this locus (chr14q23.1), mutation of a single allele is enough to account for the high allele frequency if the mutation occurred early in the outgrowth of this lesion or if it was already present in a subset of cells from the tumor before treatment. Furthermore, the functional implication of this missense mutation is yet ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
unknown. Patient #3 underwent surgical resection, where not only the resistant, BRAF V600E+L505H mutation containing lymph node was removed, but also two other lymph nodes that were still sensitive to vemurafenib. Samples from these lymph nodes and from the original baseline sample were interrogated using ultradeep sequencing, and neither PRKCH nor the secondary BRAF L505H mutation could be detected while sequence coverage on their positions reached 17 000 and 40 0009, respectively (Tables S3 and S4). Although these data do not present definitive proof of absence of the BRAF L505H mutation in the pretreatment sample, it supports the model of the secondary BRAF mutation as a mechanism of resistance, either acquired or pre-existent. Unfortunately, tissue from the primary tumor was not available to screen for the presence of this mutation. We cannot exclude the possibility that these mutations were present in a subclone of the tumor that was not sampled in the pretreatment biopsy or that the mutations were present in
PIGMENT CELL & MELANOMA Research Detailed imaging and genetic analysis reveal a secondary BRAF L505H resistance mutation and extensive intrapatient heterogeneity in metastatic BRAF mutant melanoma patients treated with vemurafenib Marlous Hoogstraat, Christa G. Gadellaa-van Hooijdonk, Inge Ubink, Nicolle J. M. Besselink, Mark Pieterse, Wouter Veldhuis, Marijn van Stralen, Eelco F. J. Meijer, Stefan M. Willems, Michael A. Hadders, Thomas Kuilman, Oscar Krijgsman, Daniel S. Peeper, Marco J. Koudijs, Edwin Cuppen, Emile E. Voest and Martijn P. Lolkema
DOI: 10.1111/pcmr.12347 Volume 28, Issue 3, Pages 318–323 If you wish to order reprints of this article, please see the guidelines here
Supporting Information for this article is freely available here EMAIL ALERTS Receive free email alerts and stay up-to-date on what is published in Pigment Cell & Melanoma Research – click here
Submit your next paper to PCMR online at http://mc.manuscriptcentral.com/pcmr
Subscribe to PCMR and stay up-to-date with the only journal committed to publishing basic research in melanoma and pigment cell biology As a member of the IFPCS or the SMR you automatically get online access to PCMR. Sign up as a member today at www.ifpcs.org or at www.societymelanomaresarch.org
To take out a personal subscription, please click here More information about Pigment Cell & Melanoma Research at www.pigment.org
Pigment Cell Melanoma Res. 28; 318–323
SHORT COMMUNICATION
Detailed imaging and genetic analysis reveal a secondary BRAF L505H resistance mutation and extensive intrapatient heterogeneity in metastatic BRAF mutant melanoma patients treated with vemurafenib Marlous Hoogstraat1,2,*, Christa G. Gadellaa-van Hooijdonk1,2,*, Inge Ubink1,2, Nicolle J. M. Besselink1,2, Mark Pieterse1, Wouter Veldhuis3, Marijn van Stralen4, Eelco F. J. Meijer1, Stefan M. Willems2,5, Michael A. Hadders1, Thomas Kuilman6, Oscar Krijgsman6, Daniel S. Peeper6, Marco J. Koudijs1,2, Edwin Cuppen2,7, Emile E. Voest1,2 and Martijn P. Lolkema1,2 1 Department of Medical Oncology, University Medical Center Utrecht, Utrecht, The Netherlands 2 Netherlands Center for Personalized Cancer Treatment, Utrecht, The Netherlands 3 Department of Radiology, University Medical Center Utrecht, Utrecht, The Netherlands 4 Image Sciences Institute, University Medical Center Utrecht, Utrecht, The Netherlands 5 Department of Pathology, University Medical Center Utrecht, Utrecht, The Netherlands 6 Division of Molecular Oncology, Netherlands Cancer Institute, Amsterdam, The Netherlands 7 Department of Medical Genetics, University Medical Center Utrecht, Utrecht, The Netherlands
KEYWORDS BRAF/vemurafenib/intratumoral heterogeneity/therapy resistance/volumetric imaging analysis PUBLICATION DATA Received 25 August 2014, revised and accepted for publication 15 December 2014, published online 17 December 2014
CORRESPONDENCE Martijn P. Lolkema, e-mail: [email protected] doi: 10.1111/pcmr.12347 *Equal contribution.
Summary Resistance to treatment is the main problem of targeted treatment for cancer. We followed ten patients during treatment with vemurafenib, by three-dimensional imaging. In all patients, only a subset of lesions progressed. Next-generation DNA sequencing was performed on sequential biopsies in four patients to uncover mechanisms of resistance. In two patients, we identified mutations that explained resistance to vemurafenib; one of these patients had a secondary BRAF L505H mutation. This is the first observation of a secondary BRAF mutation in a vemurafenib-resistant patient-derived melanoma sample, which confirms the potential importance of the BRAF L505H mutation in the development of therapy resistance. Moreover, this study hints toward an important role for tumor heterogeneity in determining the outcome of targeted treatments.
Approximately 50% of all melanomas are driven by V600 mutations in BRAF. Vemurafenib is the first clinically approved drug that specifically inhibits activated BRAF both in vitro and in vivo (Joseph et al., 2010; Yang et al., 2010). This compound produces a response in approximately half the patients and significantly improves
progression-free and overall survival. Unfortunately, the majority of patients with metastatic melanoma inevitably develop resistance to vemurafenib after a median time of 6 months (Chapman et al., 2011; Sosman et al., 2012). Intratumoral heterogeneity and extensive genetic variation between tumors within a single patient (Campbell
Significance Improving the treatment with BRAF inhibitors for patients with metastatic melanoma now depends on treating or preventing therapy resistance. This study confirms the clinical relevance of a proposed secondary BRAF mutation in resistance to vemurafenib. This finding reinforces the current paradigm that reactivation of signaling downstream of BRAF is an important event in the development of resistance to BRAF inhibition. Moreover, this study hints toward extensive geographic and molecular heterogeneity of the treatment na€ıve melanoma. In designing our strategy to treat or prevent therapy resistance, we need take the full extent of this molecular heterogeneity into account.
318
ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Secondary BRAF mutation causes resistance to vemurafenib
et al., 2010; Gerlinger et al., 2012; Stoecklein and Klein, 2010; Swoboda et al., 2011; Vermaat et al., 2012) have been proposed as reasons for treatment failure. Clonal outgrowth of resistant populations of tumor cells is considered the cause of tumor progression. Other studies using sequential samples obtained before treatment and at time of progression show that selective pressure of treatment activates compensatory survival pathways (Ding et al., 2012; Trunzer et al., 2013). Based on these studies, we hypothesized that separate metastatic lesions within an individual patient can respond differently to vemurafenib due to intermetastatic genetic heterogeneity and clonal variations, and investigating these genetic differences can lead to an improved understanding of resistance and hence improved treatment. Indeed, heterogeneous response and progression were observed earlier in metastatic melanoma patients treated with dabrafenib + trametinib (Menzies et al., 2014), as well as heterogeneity in response mechanisms (Rizos et al., 2014; Shi et al., 2014; Van Allen et al., 2014). To test this hypothesis, we used a longitudinal imaging approach combined with genome analyses. We studied the radiological response of individual metastases and the mutation profiles of pretreatment samples and biopsy material upon progression from stage IV melanoma patients treated with vemurafenib (supplemental methods). Volumetric, three-dimensional assessments are time-consuming but have the benefit to identify growing lesions at an earlier stage, and their validity has been shown by previous studies (Schiavon et al., 2012; Tirumani et al., 2014). For the early detection of progression, two-dimensional RECIST criteria are not ideal as they induce a delay in the recognition of progression of individual sites. The geographic heterogeneity of response to BRAF inhibitors may become important as local progression is a common phenomenon (Menzies et al., 2014), and thus, early local treatment of growing lesions may be beneficial for patients. Ten patients were included in the volumetric response evaluation. Baseline characteristics of patients included in the volumetric response evaluation are listed in Table S1. Per patient, three to ten separate lesions were measured every eight to twelve weeks during treatment. At initial
evaluation, after two months of vemurafenib therapy, 45% of patients showed partial remission, 40% had stable disease, and 1 patient had progressive disease (Figure 1A). Progression-free survival (PFS) ranged from 1.8 to 21.9 months (median 7.8 months). Five patients developed progressive disease based on the growth of target lesions identified at baseline. The other patients developed new lesions or showed evident progression of non-target lesions. One patient, whose response could not be evaluated by volumetric measurement, progressed due to an evidently progressive bone lesion. In each patient, volumetric analysis revealed a heterogeneous pattern of disease progression through tumor growth combined with multiple metastases that continued to respond to treatment (Figure 1B). Moreover, by evaluating the volumetric changes per lesion over time, it was noticed that the growth rate of individual lesions could differ between lesions, as well as the moment progression becomes apparent (Figure 1C). These data show that a detailed intrapatient response evaluation reveals a fairly homogenous initial response after 2 months, but a heterogeneous growth pattern at progression. Based on the volumetric measurements, progressive lesions were identified and sequential biopsies were obtained in four patients (Figure 2A). In three of the four patients, the sequential biopsy was taken from the same lesion as the pretreatment biopsy (Table 1). All specimens contained 70–95% viable tumor cells based on histopathological examination. To identify the genetic defects in each tumor, the coding regions of 1977 cancerrelated genes were sequenced to a median coverage of 1509 from both pre- and post-treatment biopsies and a matched blood sample was analyzed as a reference (Table S2, Data S1). Between 8 and 100 coding somatic variants could be identified per tumor sample (Table S3). BRAF mutations were seen in all pretreatment and sequential biopsies, and mutations in DNA repair genes were found in three patients, including MSH3, RAD50, ERCC3, and ERCC4. Both ERCC3 and ERCC4 are associated with melanoma and other skin cancers, according to the Sanger Cancer Gene Census (Futreal et al., 2004). To investigate whether there were any
Table 1. Overview of sites of sequential melanoma biopsies
Biopsy location
Tumor percentage
Patient
Time between biopsies (weeks)
Baseline
Sequential
Baseline
Sequential
#1 #2 #3 #11
16 19 24 8
Subcutaneous lesion thorax Axillary lymph node Inguinal lymph node Subcutaneous lesion breast
Subcutaneous lesion thoraxa Axillary lymph nodea Inguinal lymph nodea Soft tissue mass Thoracic vertebra 10–11
95 80 90 90
70 80 90 90
a
Biopsy from same lesion as pretreatment biopsy.
ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
319
Hoogstraat et al.
A
B
Lesion-specific response at first evaluation
Lesion-specific response at time of progression
1400%
200%
1200%
% volume change
% volume change
100%
0%
600%
400%
200%
0%
PR
C
p5
p10
p6
p7
p1
p8
p2
p9
p4
p3
p10
p8
p9
SD
PD non-target lesions
PD
Volumetric measurements per lesion p4
Volumetric measurements per lesion p2 Lung, left lobe LN, right axilla (1) LN, right axilla (2) LN, right axilla (3)
200 Resection of 3 lymph nodes
100
0
100%
% volumetric change
300 % volumetric change
p2
p7
p4
p1
p6
p5
p3
−100%
LN, left inguinal LN, right inguinal LN, left psoas muscle LN, left para−iliac LN, mediastinal (1) LN, mediastinal (2)
50%
0%
−50%
−100%
−100 0
20
40
60
Weeks on treatment
80
0
5
10 15 20 Weeks on treatment
25
Figure 1. Topographic diversity in the development of resistance to vemurafenib. (A) Volumetric measurements of all different lesions in 10 patients treated with vemurafenib at the first evaluation (after approximately 8 weeks of treatment), compared to baseline volume. Grey box plots outline the dispersion of measurements per patient, while the black dots represent individual lesions. Data points highlighted in red indicate lesions undetectable on the CT-scans. Patients are indicated on the X-axis and sorted by initial response. (B) Volumetric measurements of the same patients at time of progression, compared to smallest volume measured for progressive lesions. All other lesions are compared to baseline. For each patient, multiple lesions were still smaller than baseline volume, indicating sensitivity of these lesions to vemurafenib. The lesion with increased volume shown for patient 1 was too small to account for progressive disease in itself, but progression was determined in combination with new lesions not shown in this graph. (C) Dynamic changes per lesion over time in 2 patients. Volume per lesion was measured every 8–12 weeks. In patient #2, a progressive lymph node was surgically removed after 32 weeks, together with two other lymph nodes nearby. After surgery, a remaining lesion in the lung started to grow. Three lymph node lesions in patient #4 continued to respond to vemurafenib until the end of treatment, while three other lesions showed progression at the fourth evaluation at 25 weeks.
genetic changes between baseline and sequential biopsies, allele frequencies of variants in either sample were compared. In addition, copy number profiles were generated from the targeted sequencing data to be able to assess BRAF amplification and to aid in the interpretation of changes in allele frequencies of somatic variants (Figure S2). By exploring mutations unique for the progressive lesions, we sought to enrich for the resistance-associated mutations while the mutations with stable frequencies were deemed more likely to represent tumor driver mutations. Recently, a mutagenesis screen of BRAF was performed to detect potential secondary mutations that cause resistance. BRAF L505H was identified as secondary mutation able to induce resistance to vemurafenib in vitro (Wagenaar et al., 2014). Moreover, a second report 320
showed the presence of the same mutation in a patientderived cell line confirming the potential role for this specific mutation in BRAF inhibitor resistance (Choi et al., 2014). The affected amino acid is located in the vicinity of the ATP-binding pocket, which is where vemurafenib engages the BRAF protein. Molecular dynamics simulations show that vemurafenib binding indeed changes in V600E/L505H mutated BRAF compared to V600E/L505 wild-type BRAF (Choi et al., 2014), providing a plausible explanation for the therapy resistance. However, the fact that other larger studies on resistance causing mutations failed to detect this specific mutation implies that the frequency of secondary BRAF mutations is less prevalent than secondary mutations in receptor tyrosine kinases such as EGFR and KIT (Antonescu et al., 2005; Balak et al., 2006; Nishida et al., 2008). Apparently, resistance ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Pretreatment biopsy Sequential biopsy
S PD
19w
Patient 2
S1
Pre-T
26w
Patient 3 Pre-T Patient 11
8w
18w
S
43w S2 PD
PD
7w
Progressive disease according to RECIST Surgery: resection of progressive lymph nodes Radiotherapy of progressive lesions
PD
0w
Period of vemurafenib treatment
B
Putative founders Putative treatment induced Other somatic coding
PRKCH
80
S
60
Pre-T
BRAF 40
Pre-T
8w
BRAF
V600E
L505H
20
PD
17w
Patient 1
Post-treatment allele frequency
S
Pre-T
0
A
100
Secondary BRAF mutation causes resistance to vemurafenib
0
20
40
60
80
100
Pre-treatment allele frequency
Figure 2. Biopsies of progressive lesions within one patient revealed distinct mechanisms of therapy resistance. (A) Timing of biopsies and vemurafenib treatment in all 4 sequenced patients. Dotted lines indicate taking of biopsies. (B) Comparison of allele frequencies of variants in the pre- and post-treatment biopsy of patient #3. Putative founder mutations based on high allele frequency in both pretreatment and sequential biopsy are highlighted in red; de novo mutations and mutations with increased allele frequency in the sequential biopsy are highlighted in blue.
to BRAF inhibitors can be achieved through more diverse mechanisms. Other recent studies describing resistance against BRAF inhibition indicate several recurrent genetic mechanisms, including reactivation of the MAPK pathway through MAP2K1, MAP2K2, or NRAS mutations (Emery et al., 2009; Greger et al., 2012; Nazarian et al., 2010; Wagle et al., 2014), as well as activating mutations in PIK3CA (Mao et al., 2013; Romano et al., 2013) or other PI3K pathway alterations (Van Allen et al., 2014) and loss of function or aberrant splicing mutations in RB1, PTEN (Lito et al., 2013), or NF1 (Whittaker et al., 2013). In patient #3, the secondary BRAF L505H mutation described earlier was detected in the sequential biopsy (Figure 2B). Sanger sequencing of cDNA showed that both BRAF mutations were located on the same allele (Data S1, Figure S1): 53% (n = 33) of traces showed the V600E only, 23% (n = 14) showed both the V600E and the L505H mutation, in 24% (n = 15), BRAF was wild type, and in 0% (n = 0) of the traces, the L505H was found without the presence of the V600E. Our copy number analysis indicates that the pretreatment and sequential biopsies from this patient both contain four copies of chromosome 7 (Figure S2). Combined with the mutation allele frequencies and frequencies from the cDNA analysis, we conclude that both samples contain two copies of V600E-mutated BRAF from which one obtained the secondary L505H mutation. This suggests that only one double-mutant BRAF copy is sufficient to introduce vemurafenib resistance, consistent with a previous report (Choi et al., 2014). Furthermore, a missense mutation in PRKCH was not present in the pretreatment tumor sample but was detected in the post-treatment sample with an allele frequency of 95% (Figure 2B). However, as the copy number analysis also shows a deletion at this locus (chr14q23.1), mutation of a single allele is enough to account for the high allele frequency if the mutation occurred early in the outgrowth of this lesion or if it was already present in a subset of cells from the tumor before treatment. Furthermore, the functional implication of this missense mutation is yet ª 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
unknown. Patient #3 underwent surgical resection, where not only the resistant, BRAF V600E+L505H mutation containing lymph node was removed, but also two other lymph nodes that were still sensitive to vemurafenib. Samples from these lymph nodes and from the original baseline sample were interrogated using ultradeep sequencing, and neither PRKCH nor the secondary BRAF L505H mutation could be detected while sequence coverage on their positions reached 17 000 and 40 0009, respectively (Tables S3 and S4). Although these data do not present definitive proof of absence of the BRAF L505H mutation in the pretreatment sample, it supports the model of the secondary BRAF mutation as a mechanism of resistance, either acquired or pre-existent. Unfortunately, tissue from the primary tumor was not available to screen for the presence of this mutation. We cannot exclude the possibility that these mutations were present in a subclone of the tumor that was not sampled in the pretreatment biopsy or that the mutations were present in
