Monitoring resistance through liquid biopsy

The clonal evolution of tumors and the development of drug resistance is arguably the most substantial hindrance in the treatment of patients with targeted therapy. In their prospective study ‘Dynamic molecular analysis and clinical correlates of tumor evolution within a phase 2 trial of panitumumab-based therapy in metastatic colorectal cancer’ [1], Siena et al. illustrate the value of cell-free DNA (cfDNA) in understanding the clonal evolution of colorectal cancer during anti-EGFR therapy. This study demonstrates how cfDNA may be useful in understanding mechanisms of acquired resistance to anti-EGFR therapy and how the detection of resistance mechanisms in cfDNA may correlate with clinical outcomes. In this phase II study, patients with KRAS exon 2 wild-type metastatic colorectal cancer were treated panitumumab and irinotecan. The original design of the trial was to explore the role of emergent KRAS exon 2 mutations in acquired resistance to anti-EGFR therapy. However, as the clinical importance of additional RAS family mutations became apparent, this study was expanded to explore the role of these additional RAS mutations.

The mainstay of understanding the clonal evolution of tumors has been through sequencing of standard tissue-based tumor biopsies, but there is growing evidence that the use of a plasma-based assays and the use of cfDNA may allow for a better representation of clonal evolution than a single biopsy [2]. Several studies have described the inter- and intra-lesional heterogeneity of tumors, demonstrating genomic differences not only between different metastatic lesions in the same patient, but also within the same lesion [3–5]. This has subsequently been demonstrated in a variety of tumor types including colorectal cancers [6]. With the potential for profound tumor heterogeneity, a single biopsy at a single time point may vastly underrepresent the diversity of the tumor genomic landscape [7, 8]. Conversely, since cfDNA is shed from cancer cells throughout the body, this approach allows the detection of heterogeneous genomic alterations present in distinct tumor subclones or different tumor lesions [9–11]. However, the ability of serial tumor biopsies and serial liquid biopsies to detect emergent mutations during anti-EGFR therapy in colorectal cancer has not been compared in a prospective study.

To compare the effectiveness of tissue and liquid biopsies prospectively, the authors obtained serial tumor biopsies in parallel with plasma collection, both before treatment and upon disease progression. Plasma was also collected at set intervals during therapy. A striking difference in the detection of emergent RAS mutations was observed between tissue and plasma, with emergent RAS mutations detected in 37% (11 of 30) of post-progression plasma samples, compared with only 9% (2 of 21) of post-progression tumor biopsies (Figure 1). In a smaller cohort of 14 patients who had both plasma and tumor biopsies collected in parallel, emergent RAS mutations were detected in 57% (8/14) of post-progression plasma samples, compared with only 7% (1/14) of post-progression tumor biopsies, confirming that the marked differences in detection of RAS mutations between plasma and tumor were not simply due to differences between the patients evaluated. In this study, plasma was analyzed only for KRAS exons 2, 3 and 4; NRAS exons 2 and 3; and HRAS exons 2, 3 and 4. However, many additional resistance mechanisms to anti-EGFR therapies have been reported previously, including amplifications of KRAS and NRAS, EGFR extracellular domain (ECD) mutations, MET or ERBB2 amplifications, and BRAF and MEK mutations [6, 12–14]. Furthermore, studies have illustrated that individual patients may harbor multiple concurrent resistance mechanisms detectable in cfDNA [15, 16]. It is therefore possible that if this study had analyzed the broader set of resistance mechanisms, as opposed to RAS mutations only, the difference in alterations detected in plasma versus tissue may have been higher still. Overall, these results highlight how tumor biopsies may fail to capture the presence of subclonal resistance alterations, whereas liquid biopsy may provide a more effective means of detecting emergent resistance alterations.

Figure 1.

RAS mutations detected at progression in tumor versus plasma. Percentage of patients with emergent RAS mutations detected after progression in panitumumab in tumor biopsies (blue) or plasma cfDNA (red). Values are shown for all patients (left), and for patients with paired tumor and plasma specimens obtained at progression.

Interestingly, differences in the rate of detection of RAS mutation in pretreatment plasma and tumor biopsies were also noted. Of 39 patients with pretreatment plasma, 8 had detectable RAS mutations in plasma. However, in matched pretreatment tumor biopsies from these same eight patients, RAS mutations were only detected in three patients. The presence of RAS mutations in baseline tumor biopsies is not surprising since this study only required patients to be wild-type for KRAS exon 2 mutations for enrollment. However, the discrepancy between RAS status in tumor and plasma is unexpected, and could indicate that subclonal RAS alterations may be present in some patients before anti-EGFR therapy. Of the five patients with baseline RAS mutations detected in plasma only, two patients showed the same RAS mutations in plasma at the time of disease progression. However, the remaining three patients no longer had detectable RAS mutations in plasma after progression on anti-EGFR therapy. Of note, these three patients had very low levels of RAS mutations detectable in baseline plasma, with mutant allele frequencies (MAF) ranging from 0.03% to 0.05%. As these levels are near the limit of detection of the assay, it is possible that these RAS mutations may represent false positives. However, it is also possible that these subclonal RAS mutations were ‘lost’ during anti-EGFR therapy. Indeed, patients who had detectable RAS mutations in baseline plasma, but not at progression, were also noted in the ASPECCT study, which compared cetuximab with panitumumab as monotherapies in refractory metastatic colorectal cancer in patients who were KRAS exon 2 WT [17–19]. Thus, these observations raise the possibility that there may be a certain threshold for MAF (or perhaps more accurately Cancer Cell Fraction) at which a baseline subclonal RAS alteration may have clinical relevance and impact efficacy [20]. This question merits further clinical investigation.

The authors also examined whether progression-free survival (PFS) was different between those patients who exhibited emergence of RAS mutations at progression compared with patients who remained RAS wild-type at progression. Interestingly, no difference in PFS was observed suggesting no difference in the kinetics of acquired resistance driven by emergent RAS mutations or by RAS-independent mechanisms. These findings are consistent with work done in the ASPECCT study where investigators looked at emergence of RAS mutations in cfDNA in patients who were RAS WT in the blood at baseline and treated with anti-EGFR antibodies and looked at clinical outcomes of those who developed RAS mutations in the blood including KRAS exon 2, 3 and NRAS, exons 2, 3 and 4. In the study of 164 assessable patients who were RAS WT treated with panitumumab that had matched baseline and post-treatment blood, emergence of RAS mutations was observed in 32% of patients, with more patients having emergence of KRAS mutations than NRAS mutations. There was no difference in PFS, response rate, or overall survival between the two groups [17].

The lack of difference seen in the clinical outcomes seen by Siena et al. and in the ASPECCT study is in contrast to work done by Emburgh et al. Emburgh et al. compared PFS on anti-EGFR therapy for patients who developed emergent RAS mutations in the blood upon disease progression versus those that developed EGFR ECD mutations that block anti-EGFR antibody binding. They found that PFS on anti-EGFR therapy was longer for patients who developed ECD mutations than for those who developed RAS mutations, suggesting a clinically relevant difference between these two modes of resistance [21]. However, the seemingly contradictory observations in these studies may be reconciled by the fact that patients who did not develop RAS mutations upon progression in the current study may have developed a number of potential resistance mechanisms other than EGFRECD mutations, such as MET mutations, ERBB2 amplifications or BRAF mutations. Thus, while it is possible that EGFRECD mutations may emerge with longer latency, other RAS-independent alterations common in anti-EGFR resistance may emerge with similar kinetics to RAS mutations.

Finally, Siena et al. also evaluated the lead-time of RAS detection in the blood to clinical and/or radiographic progression. They observed that RAS alterations were detectable in the blood at a median of 3.6 months before disease progression. Thus, real-time monitoring of cfDNA during therapy may help to identify impending resistance before it is evident through standard clinical or radiographic monitoring, which may aide in the selection of subsequent therapies. Other potential advantages of real-time plasma monitoring during therapy in colorectal cancer patients were illustrated by Siravegna et al. who found that the levels of RAS mutations that emerge during anti-EGFR therapy can decline after cessation of anti-EGFR therapy, creating an opportunity for patients to respond again if challenged with anti-EGFR therapy after a brief interval off of anti-EGFR antibodies. Currently, several anti-EGFR challenge studies are in development given these findings [21,22].

This study contributes to the existing literature that liquid biopsies may provide a noninvasive means to capture the inter- and intra-lesional heterogeneity of tumors that is underrepresented in tissue-based approaches. Large genomic profiling efforts have shown that plasma analyses are concordant with tissue-based sequencing databases but that cfDNA may capture clinically relevant as well as novel insights into the clonal evolution of tumors and acquired resistance. This study is an important prospective example of how liquid biopsies may provide a more effective means of detecting acquired resistance alterations compared with tumor biopsies. This study also highlights the potential utility of real-time blood-based monitoring of patients during therapy to track the dynamic process by which resistant sub clones emerge, which may provide clinicians the opportunity to tailor treatment strategies accordingly.

Funding

Damon Runyon Clinical Investigator Award, NIH/NCI R01CA208437, K08CA166510, a Stand Up To Cancer Colorectal Dream Team Award (no grant number applies) (to RBC), and NIH/NCI Gastrointestinal Cancer SPORE P50CA127003 (to RBC and ARP) and American Cancer Society (to ARP).

Disclosure

RBC is a consultant/advisory board member for Amgen, Astex Pharmaceuticals, Avidity Biosciences, BMS, Genentech, LOXO, Merrimack, N-of-one, Roche, Shire, and Taiho, and has received research funding from AstraZeneca and Sanofi. ARP has been a past employee of Genentech.