Driver-Positive Tumor Frequency Decreases With Increasing Copy Number Gain in NSCLC

Tissue next-generation sequencing (NGS)-based HER2, KRAS, and MET copy number gain thresholds that were set by oncogene overlap identified potential clinically relevant amplified subgroups of non–small cell lung cancer (NSCLC) tumors with altered genetic profiles and decreased survival, according to findings published in Clinical Lung Cancer by Alexander Watson, MD, DPhil, FRCPC, and colleagues.1

“The thesis of our paper was that we could use this concept of oncogene overlap and mutual exclusivity to define a copy number gain threshold, a number of copies of a gene via NGS sequencing that would have clinical impacts on the outcomes in that cancer group, [that] could then be used to examine what genes co-occur with different amplifications of a high significance. Which pathways seem to rely more on amplification as a mechanism of resistance compared with other pathways?” Watson said in an interview with OncLive®. “We collaborated with Caris Life Sciences and had a cohort of [more than] 13,000 existing sequenced NSCLC tumors with limited annotation but full genetic profiles on them, and that’s what we used as our cohort for this group.”

Findings from the study revealed that the frequency of driver-positive tumors decreased with increasing copy number gain. When setting copy number gain thresholds by oncogene overlap and dataset size, NSCLC tumors considered relevantly amplified for MET, HER2 and KRAS were significantly less likely to be driver-positive (P < .001). The thresholds that were set during the study by oncogene overlap were at least 6 for HER2 and KRAS genes, and at least 4 for MET. Additionally, when driver-positivity and amplification status overlapped, same-gene alterations—mutation and copy number gain—were significantly enriched for all 3 genes.

In the interview, Watson, an advanced fellow in Thoracic Oncology and Investigational Cancer Therapeutics at University of Colorado Anschutz Medical Campus in Denver, detailed the rationale for the study and how it was conducted. He also highlighted findings showing oncogene overlap is adequately applied to NGS-based tissue sampling by decreasing frequency and increasing copy number gain. In a concurrent interview, Watson examined further findings including associations with high copy numbers and overall survival as well as the next steps with this research.

OncLive: What led you to examine oncogene overlap by tissue-based NGS?

Watson: The background gets complex. Whether or not a tumor carries an EGFR mutation is a yes/no question and that’s nice and simple—you get a result on a NGS report, and it says EGFR-mutated with a certain mutation that we know is targetable/not targetable or not [mutated]. Amplification is messier because it’s a continuous variable. Amplified can mean 2 copies of a gene that is just part of genomic instability, or it could mean highly amplified where there is [a] meaningful increase in protein, reliance on that pathway, and it’s driving the cancer. As of yet, a lot of mutation reports don’t give you the level of amplification foundation. The level of amplification isn’t well reported because the implications of what is a meaningful amplification [vs] what is not [meaningful] is not yet well understood.

Therefore, for a clinician getting a report on a patient’s tumor genetic profile, the meaning of an amplification is harder. Is that truly something I should focus on for this patient’s care? That’s why not just for therapeutic investigations, but for clinicians, knowing the implications of seeing an amplification [in] genetics [and asking] ‘is this important for me or not?’ is something we all need to work upon as a field. As of yet, and this work is a small step along the pathway—it’s by no means definitive—we need to better define the yes/no relationship or at least what variables we need to consider when interpreting an amplification in a way that properly captures the complexity of what an amplification can mean [in] the tumor; [it’s key to do this] in the way that a yes/no answer for a mutation can be interpreted by a clinician getting a mutation report. The challenge of a continuous variable compared with a discrete yes/no answer on a mutation is part of the challenge of amplification.

What was the rationale for conducting this study?

In NSCLC, there is a clear divide between cancers that are driven by genetic mechanisms that are identifiable and actionable—which [is] variously called driver-positive NSCLC or some will call it actionable, genomic mutation-positive—and those where we can’t identify a mechanism that’s targetable, and chemoimmunotherapy remains the standard of care [in that setting]. There are driver mutations which we are familiar with the implications of—these tend to be either fusions, single nucleotide [variants], or insertion [and] deletion mutations in a set of known genes which are oncogenes and drive a pathway of growth and division and metastasis.

The implications of amplification or copy number gain in those same genes are not as well understood. You can imagine more copies of a gene could, in theory, lead to more transcripts, more protein, and if that pathway was then a growth promoting pathway, you’d get potential for oncogenic transformation there. But these events can also happen as part of a cancers evolution. Cancers naturally gain ploidy. There can be multiple copies of different chromosomes in cancers, genetic instability is [one] of the hallmarks of cancer. Identifying a threshold where a copy number gain in an oncogene is of clinical importance and oncogenic importance is not very well defined in the literature. As such, cancers with these copy number gain in oncogenes are not yet traditionally targetable by the same therapies that we use for traditional driver mutations such as fusions, insertions, deletions, or point mutations.

How were the 3 genes selected to be examined in this study?

In this study, we focused on 3 genes—HER2, KRAS, and MET. We are developing therapeutic pathways for driver mutations in these genes. MET has approved therapies [and] there are ongoing clinical trials [surrounding], and agents approved [that are], HER2 TKIs. For KRAS, we have KRAS G12C inhibitors, but we don’t have inhibitors against the other mutations in KRAS. Not only do all 3 of these genes have evolving therapeutic pathways, they are implicated in off-target resistance of other oncogenic targeted pathways.

For example, EGFR is probably our longest targeted and best understood driver mutation pathway in lung cancer. If you give [patients with] cancer and an EGFR traditional mutation an EGFR inhibitor, over time, they’ll either develop on-target resistance which is less common with newer generations of EGFR TKIs, or they can develop off-target resistance—that is development of another pathway that drives oncogenesis—even in the presence of an effective inhibitor. All 3 of these genes—MET, HER2, and KRAS—are implicated in that resistance mechanism in various ways.

Because of that, we focused on these 3 genes and used the oncogene overlap concept to better define what is a meaningful copy number gain in these 3 genes. [Our] hypothesis [was] that we’d be able to use a better definition of a copy number gain that’s clinically significant in future trials of therapeutic agents either in the primary setting or in the resistance setting.

What methods were used to conduct the study?

The 13,072 [tumor] cohort [was] either sequenced by whole exome sequencing or a preselected panel of 592 meaningful genes. These 2 different platforms will look at the important genes in lung cancer, and using that cohort, sequence 1 of those 2 ways. We defined the DNA-based mutations, then [had] an RNA transcriptome which is used for the fusion mutations that are driver-positive. Using that, we set out to define driver mutation [positivity], which [was] done with standard methods of what a meaningful, pathogenic, or likely pathogenic driver mutation [is] in lung cancer, and we looked concurrently in those same group of tumors at the copy numbers of these 3 key genes.

Copy numbers by NGS [aren’t] as clear and linear as it would in looking under a microscope, but to look at copy number gain, you look at the copies of a given gene of interest, and you calibrate that to an estimate of disomy in the overall sample size. You have to decide what would be a normal copy number for a gene based on the number of reads, the number of times that sequence of DNA has come up in your NGS, and then you compare that to your gene of interest: How many times did that gene get sequenced in your meshed up sample?

Having done that, we then divided our cohort into those with or without driver mutations and those with or without amplification of 1 of these 3 genes of interest—HER2, MET, or KRAS. You could build that as a quadrant then; either something has a driver mutation or not [and] either something has an amplification in your gene of interest or not. This resulted in 4 groups for every gene of interest—every copy number gain gene being generated of the total cohort of tumors. We then looked at the genetic profiles of those tumors in those 4 groups, and then looked at survival implications.

What were key findings surrounding oncogene overlap?

We first were looking at oncogene overlap and trying to examine how the relationship between the number of copies of a gene influenced the driver mutation frequency in that segment of cancers with that number of copies of a gene, so that the copies of [a] gene would be expected to be 2 for a disomy cell.

As the copy number of that gene of interest increased in a tumor, we found that the frequency of [co-occurring] driver [oncogenes] decreased for all 3 and that was expected because that gets back to the principle of mutual exclusivity; you’d expect [that] as the copy numbers of a gene increased that it’s more likely to be influencing the growth of a tumor in a driver-positive way, acting as a driver mutation in that cancer, and the cancer doesn’t need multiple driver mutations in general, although there’s a resistance setting. [Therefore], you’d expect that as the copy numbers of a gene of interest increased, it would be more likely to have a growth promoting influence.

Some tumors we haven’t been able to identify in this cohort [as] we don’t know whether they’re untreated or not, [and] we’d expect that a certain percentage of our cohort will have overlap because that would be more setting resistance where cancer had 1 driver mutation it has acquired, a copy number gain, for example. But we would expect overall in the whole cohort, a decrease in the frequency of co-existing driver mutations as the copy number for each of these 3 genes increased. We did find that.

We considered every identified mutation that can drive a lung cancer and looked at that frequency vs the frequency of the copy number gain [for KRAS, MET, and HER2]. We also looked at that same plot but [excluded] same gene mutations; that means we excluded, for example, KRAS mutations if we were looking at KRAS copy number gain. That’s important because the implications of an already mutated gene being amplified are very different than [that of] a different driver pathway co-existing with amplified wild-type KRAS.

Dividing those plots with all driver mutations and ignoring same gene co-existing amplification and driver mutations, we saw that as copy number gain increased the frequency of driver mutations decreased. For MET, that started at a lower copy number gain than for HER2 or KRAS; based on our principle of mutual exclusivity, you might ask, ‘Does that imply that MET at a lower gene copy number can influence a cancer whereas the other copy number gains don’t seem to be clinically important because they’re co-existing with other driver mutations until they get to a much higher threshold?’ It’s possible, that would need more basic science [research], but it is hypothesis generating in this context.

What was the significance of these findings from the first portion of the study prior to and during defining the thresholds?

We proved that the principal oncogene overlap is adequately applied to NGS-based tissue sampling using this method of decreasing frequency and increasing copy number gain. We then defined a threshold, what we call meaningfully amplified for each of these 3 genes. For MET, it was simplest because the frequency of co-existing driver mutations plateaued, meaning that past a certain threshold the cancer didn’t seem to want or need more MET—at least that didn’t further decrease the frequency of co-existing mutations—and that’s a much easier line to draw. For HER2 and KRAS, the frequency continued to decrease to the highest threshold we could examine in our sample, so we set it at the highest meaningful copy number gain because it seemed like more copies still led to further decrease in the frequency of co-existing driver mutations. This is more of a hypothesis-generating [finding]. We don’t have a good annotation to look at what treatments these cancers have received before, which is a big limit to our data set.

[Based on the] first part of the results, using NGS and oncogene overlap, we can define a threshold that isn’t just based on us picking a number of what copy number gain we think is clinically meaningful for a cancer. But based on biological mutual exclusivity this copy number gain seems to define a subgroup of meaningfully HER2-, KRAS-, and MET-amplified cancers that then could be studied in targeted therapy trials.

Reference

1. Watson AS, Krause HB, Elliott A, et al. Use of oncogene overlap by tissue-based next-generation sequencing to explore the mutational landscape and survival impact of HER2, KRAS and MET copy-number gain in nonsmall cell lung cancer. Clin Lung Cancer. 2024;25(8):712-722.e1. doi:10.1016/j.cllc.2024.09.001