Significant advances in the past decade have dramatically improved the ability to use ctDNA to guide therapy. Advantages include ease of acquisition as acquiring a sample requires only a blood draw, and the potential that the pool of ctDNA is a better reflection of the relevant biology as it potentially reflects all metastatic tissues. Disadvantages are that sequencing attempts may not be productive if the sample is acquired at a time when the tumor is either quiescent or tumor burden is so low that only limited amounts of DNA are being shed. Performing ctDNA analysis when a tumor is not progressing is less likely to be productive for a number of tumor types.23,24 Sequencing ctDNA is also more susceptible than sequencing tumor biopsies to detection of alterations that are not from the tumor of interest but from clonal hematopoiesis of indeterminate potential (CHIP) or other clonal hematopoietic disorders (see Confounders section below).
Selecting the Tissue
Deciding on the tissue to analyze is a critical part of the decision process (Table 3). If the primary tumor tissue is old the likelihood of productive sequencing is lower, although age alone is not the only consideration and the methods of fixation may be just as relevant.
For prostate cancer in particular, the ability to successfully sequence primary tumor tissue decreases as the amount of tumor decreases in low-volume biopsies such as prostate needle biopsies. Generally, if tumor content is < 10% of the biopsy specimen, then sequencing is less likely to be productive.25 Also, if the alteration of interest is not known to be truncal, then a relevant target might be missed by sequencing tissue that does not reflect current biology. Metastasis biopsy may be the most appropriate tissue, particularly if this specimen has already been acquired. As above, a metastasis biopsy may have a higher tumor content, and it should reflect relevant biology if it is recent. However, bone biopsies have a relatively low yield for successful sequencing, so a soft tissue lesion (eg, liver or lymph node metastasis) is generally preferred.
The inability to safely access tissue is often a consideration. Proximity to vital structures such as large blood vessels or the potential for significant morbidity in the event of a complication (liver or lung biopsies, particularly in patients on anticoagulation medications) may make the risk/benefit ratio too high. The inability to conduct somatic testing has been reported to often be due to inadequate tissue sampling.26 ctDNA is an attractive alternative but should typically be drawn when a tumor is progressing with a reasonable tumor burden that is more likely to be shedding DNA. Performing ctDNA analysis in patients without obvious radiographic metastasis or in patients whose tumor is under good control is unlikely to produce interpretable results.
Interpreting the Results
The intent of sequencing tumor tissue is to identify alterations that are biologically important and may provide a point of therapeutic leverage. However, deciding which alterations are relevant is not always straightforward. For example, any normal individual genome contains around 10,000 missense variants, hundreds of insertion/deletion variants, and dozens of protein-truncating variants. Distinguishing these alterations, which are part of the individual, from those that are tumor-specific and have functional significance can be difficult in the absence of paired sequencing of both normal and tissue samples.
Specific Alterations
Although most commercial vendors provide important information in sequencing reports to assist oncology HCPs in deciding which alterations are relevant, the reports are not always clear. In many cases the report will specifically indicate whether the alteration has been reported previously as pathogenic or benign. However, some platforms will report alterations that are not known to be drivers of tumor biology. It is critical to be aware that if variants are not reported as pathogenic, they should not be assumed to be pathogenic simply because they are included in the report. Alterations more likely to be drivers of relevant biology are those that change gene and protein structure and include frameshift (fs*), nonsense (denoted by sequence ending in “X” or “*”), or specific fusions or insertions/deletions (indel) that occur in important domains of the gene.
For some genes, only specific alterations are targetable and not all alterations have the same effect on protein function. Although overexpression of certain genes and proteins are actionable (eg, HER2), amplification of a gene does not necessarily indicate that it is targetable. In NSCLC, specific alterations convey sensitivity to targeted therapies. For example, in EGFR-mutated NSCLC, the sensitizing mutations to EGFR TKIs are exon19 deletions and exon 21 L858R point mutations (the most common mutations), as well as less common mutations found in exon 18-21. Exon 20 mutations, however, are not responsive to EGFR TKIs with a few exceptions.27 Patients who have tumors that do not harbor a sensitizing EGFR mutation should not be treated with an EGFR TKI. In a variety of solid tumors, gene fusions of the NTRK 1/2/3, act as oncogenic drivers. The chromosomal fusion events involving the carboxy-terminal kinase domain of TRK and upstream amino-terminal partners lead to overexpression of the chimeric proteins tropomyosin receptor kinase (TRK) A/B/C, resulting in constitutively active, ligand-independent downstream signaling. In patients with NTRK 1/2/3 gene fusions, larotrectinib and entrectinib, small molecule inhibitors to TRK, have shown antitumor activity.28,29 No alterations beyond these fusions are known to be targetable.
Allele Fraction
Knowing the fraction (or proportion) of the alteration of interest in the sequenced tissue relative to the estimated tumor content can assist in decision making. Not all platforms will provide this information, which is referred to as mutation allele fraction (MAF) or variant allele fraction (VAF), but sometimes will provide it on request. Platforms will usually provide an estimate of the percent tumor in the tissue being sampled if it is from a biopsy. If the MAF is around 50% in the sequenced tissue (including ctDNA), then there is a reasonable chance that it is a germline variant. However, there are nuances as germline alterations in some genes, such as BRCA1/2, can be accompanied by loss of the other allele of the gene (LOH). In that case, if most of the circulating DNA is from tumor, then the MAF can be > 50%.
If there are 2 alterations of the same gene with MAF percentages that are each half of the total percent tumor, there is a high likelihood of biallelic alteration. These sorts of paired alterations or one mutation with apparent LOH or copy loss would again indicate a high likelihood that the alteration is in fact pathogenic and a relevant driver. Not all pathogenic alterations have to be biallelic to be driver mutations but in BRCA1/2, or mismatch repair deficiency genes, the presence of biallelic alterations increases the likelihood of their being pathogenic.
Tumors that are hypermutated—containing sometimes hundreds of mutations per megabyte of DNA—can be particularly complicated to interpret, because the likelihood increases that many of the alterations are a function of the hypermutation and not a driver mutation. This is particularly important when there are concurrent mutations in mismatch repair genes and genes, such as BRCA1/2. If the tumor is microsatellite instability high or hypermutated, concurrent BRCA1/2 alterations are often passengers as the tumors rarely have coexisting “signatures” suggesting that they have a true deficiency in homologous recombination.30 Large genes such as BRCA1/2 have microsatellite tracts that are prone to frameshift mutations as a result of microsatellite instability, and such mutations in this context are typically subclonal and not drivers. In hypermutated tumors, the likelihood is significantly decreased that any of the mutations other than mismatch repair deficiency or polymerase genes are targetable drivers.
Confounders
In some situations, interpretation can be particularly challenging. For example, several alterations for which there are FDA on-label indications (such as ATM or BRCA2) can be detected in ctDNA that may not be due to the tumor but to CHIP. CHIP represents hematopoietic clones that are dysplastic as a result of exposure to DNA-damaging agents (eg, platinum chemotherapy) or as a result of aging and arise when mutations in hematopoietic stem cells provide a competitive advantage.31 The most common CHIP clones that can be detected are DNMT3A, ASXL1, or TET2; because these alterations are not targetable, their importance lies primarily in whether patients have evidence of hematologic abnormality, which might represent an evolving hematopoietic disorder. Because CHIP alterations can overlap with somatic alterations for which FDA-approved drugs exist, such as ATM or CHEK2 (olaparib for prostate cancer) and BRCA2 (poly-ADP-ribose polymerase inhibitors in a range of indications) there is concern that CHIP might result in patient harm from inappropriate treatment of CHIP rather than the tumor, with no likelihood that the treatment would affect the tumor, causing treatment delays.32 General considerations for deciding whether an alteration represents CHIP include excluding alteration in which the VAF is < 1% and when the VAF in the alteration of interest is < 20% of the estimated tumor fraction in the sample. Exceptions to this are found in patients with true myelodysplasia or chronic lymphocytic leukemia, in whom the VAF can be well over 50% because of circulating tumor burden. The only way to be certain that an alteration detected on ctDNA reflects tumor rather than CHIP is to utilize an assay with matched tumor-normal sequencing.
Resources for Assistance
For oncology HCPs, perhaps the best resource to help in selecting and interpreting the appropriate testing is through a dedicated molecular oncology tumor board and subject matter experts who contribute to those tumor boards. In the US Department of Veterans Affairs, the national precision oncology program and its affiliated clinical services, such as the option to order a national consultation and molecular tumor board education, are easily accessible to all HCPs (www.cancer.va.gov). Many commercial vendors provide support to assist with questions of interpretation and to inform clinical decision-making. Other resources that can assist with deciding whether an alteration is pathogenic include extensive curated databases such as ClinVar (www.ncbi.nlm.nih.gov/clinvar) and the Human Genetic Mutation Database (www.hgmd.cf.ac.uk/ac/index.php) for germline alterations or COSMIC (cancer.sanger.ac.uk/cosmic) for somatic alterations. OncoKB (www.oncokb.org) is a resource for assistance in defining levels of evidence for the use of agents to target specific alterations and to assist in assigning pathogenicity to specific alterations. Additional educational resources for training in genomics and genetics are also included in the Appendix.
The rapid growth in technology and ability to enhance understanding of relevant tumor biology continues to improve the therapeutic landscape for men and women dealing with malignancy and our ability to find targetable genetic alterations with the potential for meaningful clinical benefit.
Acknowledgments
Dedicated to Neil Spector.