This is Part 2 in a 3-part Q&A with a panel of liquid biopsy experts addressing many of the issues faced in developing and deploying NGS-based liquid biopsy assays for clinical applications in oncology.
At a 2020 liquid biopsy webinar, Professor Schuuring discussed the plethora of options available to detect low copy number mutations in plasma cfDNA of lung cancer patients. His research laboratory combines NGS, ddPCR, qPCR and mass spectrometry approaches to address three main applications: (1) primary diagnosis by detection of predictive mutations, (2) monitoring of treatment response based on changes in plasma mutant levels, and (3) detection of therapy resistance mechanisms.
At all stages of the patient treatment, stratification and management cycle, clinicians are faced with balancing between cost, sensitivity, breadth of mutation coverage and turn-around time of the different testing methods available. According to Professor Schuuring, there is no single golden bullet approach that would fit all clinical needs, while FDA approval of tests is also lagging behind research efforts. For example, when monitoring treatment response his lab often relies on ddPCR, which quickly and cheaply detects single mutations with excellent sensitivity. Using this technology, they have recently followed over 100 patients during immunotherapy treatment with survival correlating to levels of a single KRAS G12V mutation, detectable at VAF 0.1% sensitivity in a clinical setting.
However, a single mutation is not always predictive of treatment response and that is where NGS-based approaches, such as Roche Avenio ctDNA assay come into play. They enable monitoring a large number of mutations in multiple genes at once, but come at a much higher price, turn-around time and cost.
When detecting mutations associated with therapy resistance mechanisms, Dr Schuuring’s lab uses a multi-platform approach including qPCR COBAS ctEGFR v2, mass spectrometry-based Ageno UltraSEEK Lung ctDNA, as well as NGS-based Roche Avenio ctDNA assay. The lab has recently demonstrated good concordance between the Agena (MS) platform and other methods (qPCR and NGS), but was strongly dependent on the amount of input material, with 100% concordance in EGFR mutations reached when using at least 10 ng of input cfDNA1.
At the webinar event, Professor Schuuring mentioned that in typical clinical setting, solid tissue and liquid biopsy testing are not often performed in parallel so there is often no data on the direct concordance of the results for each patient. However, validation study showed a high level of concordance with mutations detected in ctDNA found in the tissue samples. Guardant Health also published data showing similar concordance. However, because around a third of solid tumors do not shed ctDNA, relying only on ctDNA testing means potential of missing these cases by liquid biopsy test.
Q&A with Professor Schurring:
What would be your advice for those looking to move from tissue to ctDNA testing for the first time?
Prof Schuuuring: Standardization and harmonization! When choosing a technology, there is always the decision between the cheap and quick versus comprehensive options that would be also appropriate for pre-screening. He recommended following the Cancer-ID and IQN-Path guidelines to ensure standardization of pre-analytical procedures in clinical CAP/CLIA labs.
Could you further expand on the idea of using a method with lower sensitivity as a screening tool?
Prof Schuuuring: For the screening to identify and/or classify cancer in high-risk patients, or for the early detection of cancer to predict a high risk of developing metastasis or recurrence, ctDNA methods other than the ones I presented are needed, often referred to as ultra-sensitive ctDNA assays. Originally, very-deep-sequencing was used, often restricted to 1 or few markers, but to be able to identify these low levels of cancer cells in all the patients at high risk, newer assays rely on ultra-sensitive approaches using larger panels, WGS or methylation profiles. Most recent data especially show the power of combining such assays. Examples would be PanSEEK, CancerSEEK, GRAIL, Guardant360/OMNI, FoundationOne Liquid, and TRACERx2.
For ddPCR at 0.1% sensitivity and mass input of 10ng cfDNA, does it imply that there are no false positive droplets in negative controls?
Prof Schuuuring: For each separate ddPCR assay we determine the number of FP droplets in numerous NTC samples in our lab (in super-clean room to set up the ddPCR tests to take major precautions to prevent cross-contamination!) and all runs typically show mostly no mutant droplets (<2) considered as FP droplets. Any run that shows more than 2 droplets in the NTC control samples is considered as invalid. In other words, a sample with >3 droplets is considered true positive. In addition, we need minimal in total 333 droplet/sample (sensitivity 3/333 = 1%) to be able to define a sample as ddPCR-negative3. Using reference material high sensitivities can be achieved and associated with DNA input. In clinical samples, we generally reach sensitivities of 0.01-0.1% with input of 20 ng and 0.1-1% at 2 ng. In my presentation, I also demonstrated that the %CV related to input DNA should be taken into account4.
What is median VAF in samples and how does it vary with different stages of lung cancer?
Prof Schuuuring: Recent NGS analysis of 340 plasma samples in advanced stage NSCLC showed variants with VAF varying from 0.01 to 60% (even variants within the same plasma). Most variants had VAF <1%. The number of mutant copies/ml plasma correlates with stage of disease (total tumor volume) as seen in other reported malignancies. In GIST we showed that we detected ctDNA only in metastasized disease (not in localized disease) independent of tumor size. Finally when monitoring treatment response, the success of treatment is also associated with ctDNA levels. Be aware that the total amount of cfDNA is influenced by collection, storage and processing as due to hemolysis the VAF can change due to technical factors. Standardisation is essential.
Can you comment on ctDNA analysis in metastatic prostate cancer?
Prof Schuuuring: This field is evolving rapidly recently with the detection of prognostic/predictive SNV markers, specifically for AV-variant and most recently HRD. For these purposes, several LDT NGS methods have been reported. And similar to other malignancies there is need to have liquid biopsy as not in all situations proper tissue biopsies can be obtained.
With CVs for ddPCR at 20%-30%-40 % around 0,1% AF, how can you still apply the technology accurately around small increases used to detect effectiveness of therapy?
Prof Schuuuring: In this study we illustrated that issue should be considered. We also showed that (when available) sufficient cfDNA should be used and anything below 10ng is not very reliable . Preliminary NGS data using cfDNA during treatment revealed that a >50% decrease in the variant-ctDNA levels when comparing cfDNA collected at first response evaluation with the pre-treatment sample, also patients with low VAF are associated with clinical outcome similar as those with VAF >1%.
Any idea if HRD testing can be performed by liquid biopsy?
Prof Schuuuring: Various assays are validated for this, such as cfDNA like FoundationOne Liquid Cdx (FDA approved for BRCA1/2), GUARDANT360 including BRCA1/2 and few others. HRD was developed for ctDNA testing and both these methods are offered as service. Other ctDNA detection options available are TSO500 (which covers all HRD genes) and Avenio-Expanded (BRCA1/2).
In your experience, do you recommend an UMI-based approach to pick up low mutant allele burden below 0.5%?
Prof Schuuuring: Yes, especially when using broader panels and UMIs are becoming common practice for all ctDNA NGS assays.
Is there a concordance assay between tissue and ccfDNA NGS for amplifications and fusions?
Prof Schuuuring: No. In my opinion, today ctDNA NGS does not reliably detect true CNV (amplification) in plasma with ctDNA <0.1% tumor burden. Data using low coverage whole genome sequencing seems to be more reliable to detect CNV in plasma. Regarding fusion detection, the most common rearrangements are included in genome-based NGS approaches but you should be aware that this will not cover all breaks as these are scattered over very large regions and often with many partners (>50). In tissue predictive testing, we use RNA-based NGS approaches (like Archer) that enable us to detect almost all fusions of the markers of interest. Using circulating RNA (either free, in EV or in platelets) is under investigation and far from implementation today.
Is lower sensitivity for detecting fusion genes other than SNV or indels a problem? What do you think would be the solution?
Prof Schuuuring: The answer is related to the clinically relevant sensitivity of your test. Do we need VAF<0.01% and how should we make treatment decisions for such low VAF (for predictive ctDNA testing)? Coverage is a challenge for optimal predictive testing while for monitoring (as for SNV) lower VAF might be useful. There is still some more work to be done here.
Did you see a rise in the number of requests for liquid biopsy in diagnostics over the past couple of years? Do you feel that the oncologists understand the power of this technology?
Prof Schuuuring: Yes, especially with the approval of osimertinib for TKI-resistant patients with T790M mutation. There is a need for many applications and enormous efforts in research setting are out there but the lack of clinical utility (using prospective studies) and the relatively high costs of ctDNA testing and lack of proper reimbursement by insurance companies, delay implementation in clinical practice.
Download the video to replay or watch the webinar The Promise of Liquid Biopsy for Cancer Diagnostics and Therapeutic Monitoring: Are We There Yet?
- Lamy, Pierre-Jean, et. al., “Mass Spectrometry as a Highly Sensitive Method for Specific Circulating Tumor DNA Analysis in NSCLC: A Comparison Study.” Cancers vol. 12, 10 3002. 16 Oct. 2020, https://doi.org/10.3390/cancers12103002
- Sheridan “Investors keep the faith in cancer liquid biopsies“, Nature Biotechnology 2019, doi: https://doi.org/10.1038/d41587-019-00022-7
- Boonstra, PA et al. “A single digital droplet PCR assay to detect multiple KIT exon 11 mutations in tumor and plasma from patients with gastrointestinal stromal tumors.” Oncotarget vol. 9,17 13870-13883. 14 Feb. 2018, https://doi.org/10.18632/oncotarget.24493
- Weber at al. “Technical Evaluation of Commercial Mutation Analysis Platforms and Reference Materials for Liquid Biopsy Profiling” Cancers 2020, 12, 1588; https://doi.org/10.3390/cancers12061588