Research Autopsy Demonstrates Polyclonal Acquired Resistance in a Patient With Metastatic GI Stromal Tumor

CASE DESCRIPTION

This molecular tumor board (MTB) discussion focuses on a 46-year-old man with metastatic GI stromal tumor (GIST) who underwent an institutional review board–approved research autopsy study at our institution for patients with advanced cancer. The patient first presented with severe abdominal pain in November 2012, and computed tomography scans revealed a large 6.5-cm intra-abdominal mass. A biopsy of this mass was obtained, and histology was consistent with GIST, a tumor type classically driven by activating genomic alterations in KIT. Shortly after diagnosis, he was started on first-line therapy with the tyrosine kinase inhibitor (TKI) imatinib until disease progression in April 2015. His therapy was subsequently switched to sunitinib, which was effective in controlling further growth of his cancer until January 2016. The patient went on to receive regorafenib and ponatinib as well as an experimental therapy BXQ-350 via a phase I clinical trial. His performance status gradually declined as a result of progressive cancer, and informed consent was obtained from the patient and his spouse to conduct a rapid research autopsy for tumor procurement after death. His entire clinical course is summarized in Figure 1A. Because the patient was not seen in our Precision Cancer Medicine Clinic until June 2016, no postprogression tumor biopsies were available until July 2016. From that point on, tumor biopsies and peripheral blood (for isolating circulating tumor DNA (ctDNA) from plasma) were obtained upon disease progression (Fig 1A) for analysis using next-generation sequencing.

FIG 1.

Genomic characterization of a patient with metastatic GI stromal tumor reveals polyclonal KIT resistance mutations. (A) Summary of clinical course, treatment history, and computed tomography scans depicting cancer progression. Red arrows and circles indicate the same two tumors demonstrating growth between June 2016 and January 2018. Yellow arrow and circle indicate tumors that appeared after August 2017. BXQ-350 was a clinical trial therapy for advanced solid tumors. (B) Location of metastatic tumor samples procured through research autopsy that were subjected to whole-exome sequencing. T4-T6 indicate multiregional sequencing of a large liver tumor. (C) Oncoplot of the top 25 most frequently mutated genes in biopsies and autopsy samples analyzed by WES. KIT was universally mutated in all tumor samples sequenced. All samples sequenced contained the K550_V559delinsI mutation, whereas secondary mutations were detected in different biopsies and autopsy samples. Vertical bar graphs (top) show the total number of mutations per corresponding tumor sample below. Type of mutations are indicated by colored box key. (D) The percentage of different clonal fractions that compose each biopsy and autopsy tumor sample are shown in bar graphs. Each color represents a unique clone. (E) A bioinformatics approach was used to determine which KIT mutations were present in each clone of tumor cells. No KIT mutation was detected in clone 1. (F) Detected KIT mutations and their sensitivity to four different tyrosine kinase inhibitors on the basis of literature review. In this patient, multiple mutations were detected in the activation loop of KIT, while a single mutation was detected in the ATP-binding pocket. (*) ΔKIT: K550_V559delinsI. PD, progressive disease; PR, partial response; R, resistant; Rego, regorafenib; S, sensitive; SD, stable disease; ?, unknown sensitivity.

In March 2018, our patient died, and a rapid research autopsy was performed within 8 hours postmortem.¹ Thirty metastatic tumors were procured from involved organ sites, including the liver, peritoneum, and bowel. A board-certified pathologist assessed the viability and tumor cell content of all tissue samples before selection for genomic analysis by either whole-exome sequencing (WES) or targeted sequencing (279 gene panel, including KIT). For WES, we selected four biopsy samples (biopsy 1-4, Fig 1A) and 9 unique autopsy tumors with tumor cell content > 80% (Fig 1B). We performed targeted sequencing of 19 additional autopsy tumors and 5 ctDNA samples (ctDNA1-5, Fig 1A). Libraries were prepared using the KAPA Hyper Prep Kit and captured with xGen Exome Research Panel (Integrated DNA Technologies, Coralville, IA). WES was performed on an Illumina HiSeq4000 (San Diego, CA) to an average depth of ×180. Targeted sequencing of tumor DNA and ctDNA was performed on an Illumina MiSeq to an average depth of ×2,400 and ×1,400, respectively. Data were analyzed with a custom DNA sequencing pipeline to identify single-nucleotide variants, insertions/deletions, and copy number variations, as well as to characterize clonal cancer cell populations as previously described.^2,3

MTB EVIDENCE-BASED DISCUSSION

GISTs are cancers that arise from mesenchymal cells of the GI tract. More than 75% to 80% of GIST cases harbor activating KIT genomic alterations, which render GIST sensitive to TKI therapy (eg, imatinib).⁴ Unfortunately, most patients with GIST develop TKI resistance as a result of the emergence of secondary KIT mutations that interfere with the mechanism of action of imatinib, which is a US Food and Drug Administration–approved first-line therapy for metastatic or inoperable GIST. Second-line and subsequent treatments for GIST include various TKIs that are differentially effective, depending on the type of acquired KIT mutations present, and that can themselves induce further resistance mutations in KIT. In this MTB discussion, we highlight the polyclonal, heterogeneous nature of acquired KIT mutations in advanced, TKI-refractory GIST through genomic profiling of numerous spatially distributed tumors procured from research autopsy. Furthermore, genomic profiling of tumor biopsies and ctDNA from earlier in this patient’s clinical course revealed that the analysis of either specimen alone was insufficient to capture resistance heterogeneity, even though there was reasonable concordance between KIT mutations detected in the biopsy tissue and ctDNA. The latter finding demonstrates the clinical utility of integrating ctDNA profiling to monitor and detect acquired resistance in patients who receive molecularly targeted therapies.

Heterogeneity of Secondary KIT Mutations That Confer TKI Resistance

To fully characterize the heterogeneity of acquired KIT mutations in this patient, we performed genomic profiling of 28 different metastatic tumors from multiple organs procured via rapid research autopsy. Of the 28 tumors, nine were profiled using WES, with the remaining 19 profiled by a targeted gene panel that included KIT and other cancer-relevant genes (Table 1). In addition, we profiled four serial tumor biopsies using WES and matching ctDNA samples by targeted sequencing (Table 1). A fifth ctDNA sample from shortly before death was also sequenced. Altogether, genomic profiling of biopsies, ctDNA, and autopsy samples revealed seven acquired KIT alterations among tissue and blood samples.

TABLE 1.

Whole-Exome and Targeted Sequencing Summary of Biopsy, Autopsy Tumor, and Circulating Tumor DNA Samples

GIST is a tumor type characterized by low tumor mutational burden, consistent with the average tumor mutational burden of only 1.5 mut/Mb in this patient’s tumors (Data Supplement). The twenty-five most frequently mutated genes across 4 biopsies and 9 autopsy tumors are presented in Figure 1C. Of the 420 total unique mutations detected in the entire exome across all biopsies and autopsy tumors, 122 (39%) were present in all 13 samples (ubiquitous), 250 (59%) in two or more samples (shared), and 48 (11%) in a single sample (private). The finding that 70% of mutations are either shared or private suggests that biopsy of a single tumor at diagnosis or upon disease progression to determine next-line treatment may miss the identification of potentially actionable alterations. As we will discuss later, the limitations of sampling a single tumor may be overcome, at least in part, by profiling ctDNA from the same patient at appropriate time points, recognizing the caveat that such analyses may not be feasible in every cancer type given multiple factors that can impact the duration or quantity of ctDNA in the blood.⁵ Nonetheless, as expected for a driver alteration, KIT K550_V559delinsI was ubiquitously detected, whereas secondary KIT mutations that developed in response to TKI therapies were either shared or private. The secondary KIT mutations identified through autopsy tumor tissue profiling include D820V, D820A, N822K, Y823D, and F681L (Table 1). All point mutations in KIT were located within exons 14 and 17 encoding the tyrosine kinase domain.

In our analysis, we used the full spectrum of somatic mutations, including in KIT, identified using WES to characterize clonal heterogeneity and evolution in the context of this patient’s therapeutic course. Application of the clonal inference algorithm Canopy⁶ to filtered WES data revealed seven distinct clonal cell populations, which were present in varying proportions across tumor samples (Fig 1D and Data Supplement). In T4 to T6, which represent three separate regions of a large liver tumor, clonal compositions were overall similar, with the exception that Clone 3 was more prevalent in T5 and Clone 7 was more prevalent in T6 (T4 had both Clones 3 and 7). Furthermore, our bioinformatics approach predicted that each clone of tumor cells contained one or more secondary KIT mutations (Fig 1E). The secondary subclonal mutation KIT D820V was detected in Clones 3 to 7, which were present in all biopsy and tumor samples analyzed, and was shown to confer resistance to imatinib, sunitinib, regorafenib, and ponatinib (Fig 1F). Clones 5 and 6 also contained Y823D and F681L, respectively. As copy number variations were also analyzed (Appendix Fig A1 and Data Supplement),⁷ we detected loss of heterozygosity of chromosome 4q containing KIT in tumor cell Clones 4 to 7. Of interest, we noticed that Clone 5 containing Y823D was overall more prevalent in biopsy specimens B1 to B4 than in autopsy tumor samples T1 to T9, which suggested the elimination of Clone 5 by ponatinib therapy and was consistent with in vitro studies demonstrating Y823D sensitivity to ponatinib.⁸ Finally, in contrast to Clone 5, Clone 6 was detected at low proportions in all biopsy samples but at a high proportion in a small subset of autopsy samples (T2, T3, and T9). We noticed that Clone 6 had F681L, for which a thorough literature search revealed a lack of any TKI sensitivity or resistance data. However, on the basis of the previous observation of Clone 6 being detected at lower proportions in biopsy specimens, we surmised that it likely conferred resistance to one or more TKI therapies. This reasoning is further bolstered by the observation that Clone 2, having a similar distribution as Clone 6, contained the secondary mutation N822K, which has demonstrated resistance to imatinib, sunitinib, and regorafenib.⁸

In summary, we show that there is significant heterogeneity in both the spatial (comparing multiple autopsy tumors) and temporal distribution (comparing biopsies with autopsy tumors) of subclonal secondary KIT mutations induced by the selective pressure of sequential TKI therapy in our patient with GIST. This heterogeneity further illustrates the limitation of a single tumor biopsy for genomic evaluation of secondary resistance mutations upon disease progression in advanced GIST and multiple other solid tumor types.

ctDNA Analysis: A Complement to Tissue Genomic Profiling

Liquid biopsy, or the detection of somatic mutations through profiling of ctDNA isolated from blood, has emerged as a promising tool for the assessment of tumor heterogeneity in patients with metastatic solid cancer.⁵ In the clinical setting, obtaining multiple tumor biopsies is both ethically questionable and procedurally impractical because of the potential risks to patients. However, comprehensive profiling of multiple tumors is ultimately expected to benefit the patient by unveiling more actionable targets or potential therapeutic vulnerabilities, particularly when one has already exhausted standard-of-care therapies. ctDNA analysis has been applied to study tumor heterogeneity and early detection of disease progression in colorectal and lung cancer.^9-11 Although insightful, these studies compared somatic variants identified in ctDNA with those identified often in just a single or limited number of tissue samples. Therefore, it has been difficult to determine whether ctDNA analysis is a true surrogate for genomic profiling of multiple genetically heterogeneous tumors in advanced metastatic cancer.

In our patient with GIST, who underwent research autopsy in addition to genomic evaluation of his biopsy and autopsy tumor tissues, we evaluated five ctDNA samples using targeted sequencing (Fig 1A). Our first analysis was to determine concordance of KIT variants between temporally corresponding biopsy and ctDNA samples. As shown in Figure 2A, Biopsy 1 contained two KIT mutations, neither of which were detected in matching ctDNA 1. This could be explained by lower overall tumor burden and less shedding of ctDNA into the circulation at this time point, as well as by differences in the detection limits of WES versus targeted sequencing. Given that KIT Y823D was present in Biopsy 1 and has shown sensitivity to ponatinib in vitro,⁸ and on the basis of the results of clinical studies of ponatinib in patients with advanced GIST,^12,13 the patient was started on ponatinib therapy, to which he responded by having stable disease for nearly 8 months (Fig 1A). At the next time point and after the cessation of ponatinib as a result of cancer progression, Biopsy 2 and ctDNA 2 demonstrated 100% concordance of KIT mutations. Similarly, the K550_V559delinsI and D820V mutations were concordant between Biopsy 3 and ctDNA 3. An additional KIT mutation, D820Y, was detected in ctDNA 3 only. Finally, at the February 2018 time point, two different acquired KIT mutations were detected via sequencing of ctDNA 4 (N822K) and corresponding Biopsy 4 (Y823D). This is approximately 1 month before the patient’s death, and the discordance between tissue and ctDNA likely reflects the growing heterogeneity of his disease and how tissue sampling in the advanced metastatic setting may not generate all the information needed to guide therapy.

FIG 2.

Concordance of KIT variants detected between tissue and circulating tumor DNA (ctDNA). (A) KIT variants detected in paired biopsy and ctDNA samples at four different time points. (B) KIT variants detected in autopsy samples and corresponding ctDNA. The primary K550_V559delinsI mutation was detected in all 28 autopsy samples sequenced by whole-exome sequencing and targeted sequencing (28 of 28). Acquired KIT mutations D820V and N822K were detected in 17 of 28 and 8 of 28 tumor samples, respectively. (*) ΔKIT: K550_V559delinsI.

We next compared KIT mutations in 28 autopsy tumor samples (with both WES and targeted sequencing) against those detected in corresponding ctDNA (Fig 2B). Altogether, six unique KIT alterations were detected across all autopsy tumors, with the primary mutation K550_V559delinsI present in 28 of 28 tumors (ubiquitous). The next most prevalent mutation D820V was present in 17 of 28 tumors (shared), followed by N822K detected in 8 of 28 tumors (shared). Two of the six KIT mutations were detected in only 1 of 28 tumors (private). In ctDNA 5, two KIT alterations were detected: K550_V559delinsI and D820V. Overall, we observed that the KIT mutations that were more prevalent in autopsy tumors were also detected in ctDNA, whereas mutations with lower prevalence (eg private) were not. These results are consistent with the expectation that tumor burden can affect the ability to detect mutations of interest in ctDNA.¹⁴ In summary, using multiple tumor specimens from autopsy as a gold standard for tumor burden, we observed that ctDNA could effectively detect ubiquitous and shared secondary KIT mutations among tumor tissues.

In conclusion, this MTB case illustrates the value and limitations of tumor biopsies for assessing acquired resistance mechanisms to molecularly targeted therapies. Tumor biopsies are critical for tissue-based research studies to assess genomic, transcriptomic, and proteomic alterations that can lead to acquired resistance. However, polyclonal mechanisms of resistance may be inadequately revealed through the analysis of a single tumor sample as distinct and/or redundant mechanisms of resistance have been shown to coexist in the same patient. The comparison of tumor samples from rapid research autopsy and biopsies with corresponding ctDNA samples showed that the assessment of ctDNA is an approach that may capture subclonal mutations among tumor sites, recognizing the caveat that tumor burden ultimately may limit the detection of these mutations. Moving forward, we recommend combined analysis of both tumor biopsies and ctDNA, particularly at times of cancer progression, to capture dominant or emerging tumor subclones that underlie acquired resistance.

Appendix

FIG A1.

Curated copy number variations (CNVs) per tumor sample. CNVs were called with FALCON⁶ and manually curated on a per-sample basis. The copy number of the major allele (red) and minor allele (blue) is depicted for each tumor sample. Areas without a listed allele are copy neutral. Chr, chromosome.

AUTHOR CONTRIBUTIONS

Conception and design: Anoosha Paruchuri, Hui-Zi Chen, Russell Bonneville, Julie W. Reeser, Michele R. Wing, Melanie A. Krook, Eric Samorodnitsky, Sameek Roychowdhury

Financial support: Sameek Roychowdhury

Administrative support: Sameek Roychowdhury

Provision of study materials or patients: Patricia Allenby, Sameek Roychowdhury

Collection and assembly of data: Anoosha Paruchuri, Hui-Zi Chen, Russell Bonneville, Michele R. Wing, Eric Samorodnitsky, Jharna Miya, Thuy Dao, Amy Smith, Aharon G. Freud, Patricia Allenby, Sameek Roychowdhury

Data analysis and interpretation: Anoosha Paruchuri, Hui-Zi Chen, Russell Bonneville, Julie W. Reeser, Michele R. Wing, Melanie A. Krook, Eric Samorodnitsky, Jharna Miya, Sameek Roychowdhury

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).

Hui-Zi Chen

Honoraria: QED Therapeutics

Travel, Accommodations, Expenses: QED Therapeutics

Sameek Roychowdhury

Stock and Other Ownership Interests: Johnson & Johnson (I)

Consulting or Advisory Role: Incyte Pharmaceuticals, AbbVie, QED Therapeutics, Merck

Research Funding: Takeda, Ignyta, Incyte Pharmaceuticals, QED Therapeutics

No other potential conflicts of interest were reported.