Test, Treat, Repeat
Predicting cancer treatment response with liquid biopsy-based biomarkers
Alexandra Pender, Natacha Entz-Werle | | Longer Read
Oncology treatments must be tailored to individual patients – especially in childhood cancers. Blood-based circulating tumor DNA can be used for diagnosis, to track cancer progression, and to predict treatment response – but, for optimal treatment, we need sensitive tests to detect low levels of this DNA. When combined with new biomarkers, these tests can help inform patient prognosis and treatment selection.
Patients are increasingly diagnosed with cancer during the earlier stages of the disease. Blood-based methods to screen for and monitor cancer have been a major contributor to this breakthrough; now, doctors can select and administer treatment before widespread progression occurs. But is this all we can do? No – and the next major breakthrough may be the development of cancer-wide biomarker panels to monitor treatment response and predict outcomes. Such panels could allow doctors to determine an ongoing treatment’s effectiveness, alter dosage as needed, and even stop using therapies that are not working, avoiding unnecessary side effects.
Mutations in circulating tumor DNA (ctDNA) obtained through liquid biopsies can serve as biomarkers for such panels, enabling physicians to track disease progression and inform treatment strategy.
Genetic signatures in pediatric osteosarcoma
Pediatric osteosarcoma, which predominantly affects people between 10 and 30 years old, is the most common type of primary malignant bone tumor. Because there have been no significant advances in the treatment of this disease for 20 years, we have reached the limit of survival that current therapies can achieve for osteosarcoma patients – a five-year survival rate of 70 percent for all patients.
Standard therapy for pediatric osteosarcoma patients includes multi-agent neoadjuvant chemotherapy, followed 14 weeks later by tumor resection with the goal of complete tumor removal. Following primitive tumor surgery, a patient’s treatment response is classified based on the amount of cell death within the tumor – also known as Rosen grading. A patient with large (>90 percent) levels of tumor necrosis after initial chemotherapy is said to be a good responder, whereas one with little to no tumor necrosis is a poor responder. Poor responders to chemotherapy and tumor resection have a worse prognosis and require more aggressive second lines of chemotherapy postoperatively. But because it is difficult to sample and measure necrotic tumors, it is difficult to accurately classify how a patient’s tumor is responding to treatment. That’s why, at the moment, we need highly sensitive molecular residual disease evaluation to detect a low burden of cells in the tumor and/or in the plasma.
However, if we could identify DNA biomarkers that correlate to treatment response, physicians could use this information to choose a treatment strategy for a patient ahead of time based on whether they are likely to be a good or a poor responder to a particular therapy.
To identify new biomarkers for pediatric osteosarcoma, Natacha Entz-Werle and a team of researchers from CHRU Strasbourg’s Pediatric Oncology Department and the University of Strasbourg’s Laboratory of Bioimaging and Pathology retrospectively investigated the prognostic impact of three genes. Each of these genes – MET, TWIST, and APC – had been previously identified in the large tumor cohort of the French national protocol OS94 as biomarkers of bone dedifferentiation with a clear prognostic impact. Entz-Werle and her team carried out molecular assessments on the genes in an even larger cohort of patients in the OS2006 protocol, evaluating them by allelotyping and qPCR. They also used droplet digital PCR (ddPCR) to prospectively evaluate these metrics in a preliminary cohort of 20 patients with plasma available at diagnosis, before tumor surgery, and at the end of chemotherapy.
Because necrotic tumor cell DNA is difficult to analyze, the team used two separate methods for analysis. Complementary qPCR and allelotyping yielded accurate results in the diagnostic tumor, but not in plasma. These techniques were only sensitive enough to detect the three-gene signature in 65 percent of patients. The researchers therefore turned to ddPCR as an additional method to track DNA abnormalities at a level sensitive enough to reliably detect low DNA concentrations in the blood. Using ddPCR, they were able to track molecular abnormalities in both tumor tissue and cell-free DNA from liquid biopsy – making it a suitable method for monitoring treatment progression and checking for residual disease to quickly identify and treat relapse.
When correlating the results of these three methods to disease progression, the authors found that common rearrangements of ctDNA within the MET, TWIST, and APC genes could predict the therapeutic outcome of over 85 percent of their small patient cohort. They also found that persistence of this three-gene signature after neoadjuvant chemotherapy in the OS94 protocol indicated that the patient had not responded to initial chemotherapy treatment and was correlated with a higher likelihood of relapse after surgery. The researchers concluded that amplifying the three-gene signature with ddPCR could be used to track residual disease before and after tumor resection, giving physicians the ability to gauge the patient’s response throughout the course of treatment.
A new metric for NSCLC
Pediatric osteosarcoma is not the only cancer in which biomarker discovery is driving better treatment strategy. Epidermal growth factor receptor (EGFR) mutant non-small cell lung cancer (NSCLC) is a subtype of lung cancer that occurs predominantly in people who have rarely or never smoked. These lung adenocarcinomas carry mutations in the EGFR gene that render tumors susceptible to EGFR tyrosine kinase inhibitors (TKIs) for a limited amount of time – until the tumor develops resistance. Doctors will then analyze ctDNA to plan subsequent treatment options. If a tumor tests positive for the EGFR T790M mutation, it is resistant to first- and second-generation TKIs and must be treated with the third-generation TKI osimertinib.
Researchers led by Alexandra Pender at the British Columbia Cancer Agency performed a retrospective study of EGFR ctDNA testing in 142 patients with EGFR-mutant NSCLC who were progressing on treatment with first- or second-line TKIs. Of these patients, 62 percent were receiving a first-line TKI inhibitor, 25 percent had received two lines of TKI inhibitors, and 13 percent had exhausted three or more lines of systemic treatment at the time of ctDNA testing.
The researchers tested whether two separate circulating biomarkers – a patient’s EGFR ctDNA mutational status and the concentration of circulating free DNA (cfDNA) in their bloodstream – could predict their overall survival. They first quantified the amount of cfDNA in each patient’s bloodstream with a circulating nucleic acid kit, then quantified the amount of EGFR ctDNA using ddPCR. With a limit of detection of <0.1 percent variant allele fraction, ddPCR offered researchers a sensitive method to measure the low concentrations of each patient’s mutant EGFR ctDNA and the means to multiplex the assay to examine more than one EGFR mutation simultaneously. Patients in the study were representative of an EGFR-mutant NSCLC population: a median age of 66 years, 64 percent female, 57 percent never having smoked, and 53 percent of Asian ethnicity.
Multivariate analysis showed a trend toward worse overall survival with a high cfDNA concentration, regardless of EGFR ctDNA result (p=0.086). The 12-month overall survival for patients with a cfDNA concentration above the median for the population was 50 percent, compared with 68 percent for those with a cfDNA concentration below the median. In addition, the researchers found that patients with detectable EGFR ctDNA had worse outcomes than those with undetectable EGFR ctDNA at the time of progression; the former population had a 12-month overall survival of 49 percent and the latter 68 percent (p=0.003). Patients with detectable EGFR ctDNA for the EGFR-activating mutation only (EGFR T790M negative, but EGFR L858R or exon 19 deletion positive) had a significantly worse 12-month overall survival of 11 percent, which remained a significant predictor of mortality on multivariate analysis.
The future looks brighter for patients with pediatric osteosarcoma; one day, physicians tracking the three-gene signature may be able to use it to tailor treatments to individual patients during the early stages of disease. If physicians can use biomarkers to predict whether patients will be good or poor responders to treatment, then – instead of administering the same neoadjuvant therapy to all patients – they will be able to gauge the strength of therapy each patient needs prior to surgery to remove metastatic cells. After surgery, they will also be able to measure the patient’s response and more decisively determine the degree of follow-up needed to optimize the outcome. For the first time in decades, we may be better equipped to improve survival rates for children suffering from this disease.
Similarly, tracking cfDNA in EGFR-mutant NSCLC patients could allow physicians to test patients for EGFR TKI resistance earlier and optimize treatment based on EGFR ctDNA testing results. Pender’s results indicate that patients with lower levels of cfDNA on EGFR TKIs were likely to have a longer overall survival. In contrast, patients with detectable EGFR ctDNA – but who test negative for EGFR T790M – should be considered for a switch to chemotherapy instead of awaiting repeat EGFR T790M testing. These patients have a poorer prognosis, likely due to an EGFR T790M-independent mechanism of resistance, and delaying the switch to chemotherapy may mean missing the window for effective treatment.
Overall, the ability to track cancer biomarkers via liquid biopsy may allow physicians to understand the unique nature of each patient’s disease. By using the strategies applied in the studies above in combination with other molecular diagnostics, we may eventually be able to generate a catalogue of biomarkers to improve outcomes for those with cancer.