NEW YORK (GenomeWeb) – As liquid biopsy products make further inroads in the clinic and in clinical research, rather than coalescing around a few dominant technologies or platforms, investigators and entrepreneurs are continuing to explore and advance a wide range of methods for examining genomic and other molecular biomarkers non-invasively.
This diversity was on full display at last week's annual meeting of the American Association for Cancer Research held in Washington, where studies stretched over five poster sessions and several symposia.
Multiple tests are now used clinically that employ next-gen sequencing to analyze circulating cell-free DNA across targeted regions of the genome, or digital or other PCR systems to search for known mutations.
Meanwhile, methods are also being advanced that extract and analyze the nucleic acids within circulating tumor cells or extracellular vesicles — isolated from blood samples with varying technologies. Still others have identified simple proteomic surface markers to quantify particular cells or vesicles of interest as an indicator of disease.
Most tests so far have been used mainly in later-stage cancer patients, but methods are also now being pushed toward earlier indications: for monitoring disease status or predicting the recurrence of early cancers after surgery or other first-line treatment, as well as for early cancer detection in individuals with no evidence of disease.
NGS and PCR
Conference presentations at AACR included validation data on several existing tests, like Guardant Health's newly expanded NGS assay Guardant360 and Personal Genome Diagnostics' recently launched PlasmaSELECT 64.
Though multiple firms have launched NGS tests, PCR is also being used widely for blood-based cancer testing, especially in lung cancer, offering clinicians an attractive option for patients who frequently cannot have a safe tissue biopsy but who can see dramatic changes in outcomes if treated with a particular targeted therapy.
Dana Farber Oncologist Geoffrey Oxnard spoke in one session at the meeting about how he has integrated Bio-Rad Laboratories' droplet digital PCR into clinical practice for the detection of EGFR and KRAS mutations in blood samples from non-small cell lung cancer patients.
Oxnard has at times been a voice of caution in regard to liquid biopsy technologies, especially in the context of cancer screening, but he said that the validation data his team had been able to collect for ddPCR in advanced NSCLC speaks strongly enough for clinical implementation, especially considering the real needs of NSCLC patients.
He raised the example of a newly diagnosed patient with stage IV lung cancer and brain metastases, who had no prior genotyping. "This is the kind of patient I want to offer advanced targeted therapies to, but without genotyping, I can offer chemotherapy and radiation," Oxnard said.
"I need DNA to be tested, but that takes time [if using a tissue biopsy], and this patient with stage-IV and brain metastases doesn't have time," he added.
Oxnard and colleagues published a study in JAMA Oncology last year in which they showed that their plasma ddPCR approach detected EGFR and KRAS mutations with the high specificity needed to select therapy and avoid repeat biopsies for patients.
Other researchers at the conference also reported on the use of ddPCR in clinical practice. Representatives from Biodesix and Bio-Rad shared a poster with collaborators from East Carolina University on the clinical use of Biodesix's ddPCR-based GeneStrat tests for EGFR, KRAS, and BRAF mutations.
Investigators reported that in 4,000 patient cases — more than 70 percent of which were from physician practices that self-identified as community-based — they were able to generate results for 98 percent of samples submitted.
Roche is another firm that has seen recent success in PCR-based liquid biopsy, receiving the first US Food and Drug Administration approval for a blood-based solid tumor mutation test with its Cobas EGFR Mutation Test v2. Initially approved for detection of EGFR exon 19 deletions or L858R mutations in patients considering treatment with Genentech's Tarceva (erlotinib), the test then gained an expanded label to also cover detection of EGFR T790M in the context of treatment with AstraZeneca's Tagrisso (osimertinib).
At AACR last week, the firm presented updated data from its intra- and inter-laboratory reproducibility experiments with the technology, conducted with ResearchDx. The company evaluated the sensitivity and specificity of both its Avenio panels across three lab sites using 48 contrived mutation and normal reference samples.
According to the investigators, the three sites achieved 100 percent sensitivity for SNVs and fusions at 0.5 percent and 1 percent allele frequency, respectively, using 30 ng of input DNA.
Specificity was 100 for the firm's 17-gene targeted kit — both by sample and by variant. The larger of the two Avenio panels had a specificity between 91 percent and 100 percent by sample and over 99 percent to 100 percent by variant.
Both kits had 100 percent specificity down to 1 percent mutant allele frequency for detecting gene fusions.
In another poster, researchers from Roche discussed a study of more than 200 samples — including cell line blends and clinical samples — using the same two previously announced Avenio panels, plus a third larger panel covering 197 genes.
According to Roche, all three panels showed more than 95 percent sensitivity for SNVs at 0.5 percent mutant allele frequency, and for indels and structural variants at more than 1 percent MAF.
Firms making strides with tests that assay circulating tumor cells also shared updates and new data at the meeting.
Epic Sciences, which signed a distribution deal with Genomic Health last year for its CTC-based AR-V7 prostate cancer test, shared data with collaborators at Memorial Sloan Kettering on a study of 221 metastatic castration-resistant prostate cancer patient samples. Epic has positioned its unbiased, imaging-based CTC platform apart from other existing methods that use immunohistochemistry staining or other techniques to isolate cells that express certain proteins.
The firm's study presented at AACR found that only about 75 percent of AR-positive cells from prostate cancers expressed cytokeratin, one of several markers used to extract CTCs from the overall cellular component of a blood sample.
AR-positive, CK-negative CTCs had a lower overall AR signal, less nucleoli, and more nuclear speckles by digital pathology, but had gross genomic alterations consistent with malignant prostate origin and often shared the same genomic alterations with CK-positive CTCs in the same patients, the researchers reported.
The presence of these CK-negative CTCs was also associated with worse overall survival independent of treatment line, pre-therapy PSA, and therapy type in multivariate models, they said.
Investigators also presented and discussed a number of newer platforms and methods at the meeting. In one poster, a team from Genentech shared data from an evaluation of a low-pass whole-genome sequencing method for determining copy number variations in circulating tumor DNA as a way of classifying patients into different subgroups, monitoring tumor progression, and identifying drug resistance mechanisms.
Investigators studied patients with metastatic breast cancer and non-small cell lung cancer from clinical trials of the PI3K inhibitor pictilisib using Qiagen's QIAamp Circulating Nucleic Acid Kit and Rubicon Genomics' ThruPLEX Plasma-Seq Kit.
The results showed that WGS of ctDNA at 0.5x coverage efficiently identified CNVs, which showed up in about half of the patients analyzed.
Genentech also reported on liquid biopsy experiments using a microfluidic multiplex-PCR-based targeted sequencing method it developed several years ago called MMP-Seq, to which it added an additional PCR pre-amplification step for increased sensitivity in blood-based mutation detection.
In the company's hands, this PreAmp MMP-Seq allowed quantitative and reproducible mutation detection down to 0.6 percent allele frequency from as low as 1.6 ng DNA input. In 149 longitudinal plasma samples from 12 ovarian and 10 pancreatic cancer patients, 79 percent of driver mutations found in tissue samples were also identified in matching baseline plasma samples, study authors reported. Moreover, ctDNA concentration changes in post-treatment plasma samples correlated with changes in tumor burden.
A team from Dana-Farber shared a report on a new method for determining the tissue of origin of circulating DNA fragments — a goal targeted by numerous groups recently who have described methods based on both nucleosome positioning and methylation patterns.
The Dana-Farber researchers developed a method that doesn’t require actual methylation sequencing, but uses machine learning to infer methylation at a base pair resolution from a simpler data source — the size of DNA fragments as measured by whole-genome sequencing.
In early experiments, the group's prediction of DNA methylation patterns showed high concordance with the ground truth DNA methylation level seen in bisulfite sequencing of the same cancer patients. Then, by using whole-genome bisulfite sequencing datasets from different tumor and normal tissues and cells as a reference, the researchers showed that they were able to deconvolute the tissue-of-origin status of circulating DNA in a cohort of thousands of breast and prostate cancer samples and in healthy individuals.
CTC Isolation Tools
In the same vein as Epic Sciences, other groups also appear to be eager to validate CTC isolation methods that corral cells in an unbiased manner rather than targeting a particular expression marker or surface protein.
Researchers from a group of European institutions, for example, reported on an open-source method they developed to serve as a standardized tool for classifying cellular objects as CTCs within the efforts of the CANCER-ID consortium. Called ACCEPT (Automated CTC Classification, Enumeration and Phenotyping), the tool can process images generated by various CTC isolation technologies, the researchers said, including quantifying antigens expressed by selected CTCs, and classifying cellular objects across an image either manually or automatically.
A team from Massachusetts General Hospital also reported on an assay they have developed using a microfluidic CTC enrichment platform they created called iChip coupled with a highly sensitive and specific multi-gene RNA biomarker panel.
In experiments in breast cancer patients, the assay successfully identified CTC signal in 50-70 percent of metastatic and 20-40 percent of localized, pretreatment patient samples, the researchers reported at AACR. Using a multi-marker prediction algorithm, CTC scores also correlated with cancer stage and grade.
When the investigators tested patients monthly over time, they could see that CTC scores as early as one month after initiation of therapy were predictive of progression-free survival and other endpoints. In addition, a subset of high-risk genes showed expression patterns over the course of treatment that highly correlated with disease recurrence.
Posters at the conference also tackled comparisons of some of these different technologies and approaches.
As multiple tests have entered the clinic, companies aim to distinguish themselves from competitors based on what can often be minute advantages in sensitivity.
In general, digital and other PCR methods are expected to be more sensitive than sequencing approaches for circulating cell-free DNA, though they can't address nearly as many targets. But that makes PCR a good benchmark for demonstrating that an NGS test has high enough sensitivity, at least for specific alterations of interest.
Inivata, for example, presented a poster at AACR with collaborators at London's Institute for Cancer Research reporting on a study of 35 women with advanced breast cancer in which they compared results using BioRad's droplet-digital PCR assays to their own multi-gene NGS panel.
According to the authors, there was a 96 percent concordance between Inivata's InVision test and ddPCR for PIK3CA mutations, and 100 percent concordance for ESR1 mutations. Allele frequencies measured by the two technologies were also highly correlated.
Guardant Health also shared data from a ddPCR comparison study conducted with a team from Dana-Farber. Investigators recruited 220 advanced cancer patients, in which they compared Guardant360 results to ddPCR assays for EGFR L858R, exon19 del, and T790M mutations, as well as BRAF, KRAS, ESR1, and PIK3CA alterations. Overall, the concordance between the assays was 98.3 percent. Only one sample with a mutated allele frequency above 0.4 percent was discordant, along with nine others that had MAF below 0.4 percent.
In another comparison study, researchers from Chinese liquid biopsy firm AccuraGen compared four technology platforms for EGFR mutation detection in circulating cell-free DNA — PCR methods from Roche, AmoyDx, Bio-Rad's ddPCR, and NGS.
According to the authors, the sensitivity of Cobas, ddPCR, and NGS were comparable and superior to the ADx-ARMS PCR method, which could only detect EGFR mutations reliably at allele frequencies greater than one percent.
The team also said that concordance between plasma and tissue EGFR status was high overall, and that discrepancies between plasma and tissue were often attributable to spatial or temporal tumor heterogeneity.
Monitoring and early detection
Apart from examinations of technical sensitivity and validity, meeting participants also shared data supporting a shift from the late-stage cancer setting to other clinical indications like monitoring or relapse prediction in earlier-stage disease, or eventually detection of cancer in healthy people.
In collaboration with the Mayo Clinic, Exact Sciences shared data from a study of nearly 400 patients (about 300 controls and nearly 90 cancer cases) showing that a panel of four novel methylated DNA markers could distinguish lung cancer at various stages from healthy individuals without cancer.
Exact has said it is working with Mayo on early-detection tests for other cancer types as well, including liver and pancreatic tumors.
A team from Denmark and the Netherlands reported on development of ddPCR assays to aid disease surveillance in patients with bladder cancer by picking up progression and metastasis earlier and more sensitively.
Researchers tested urine and plasma samples from two cohorts representing more than 700 patients in total using ddPCR for FGFR2 and PIK3CA hotspot mutations. Samples were taken over long time periods, with patient outcomes followed for up to 13.5 years. According to the authors, the results suggest that increased levels of FGFR3 and PIK3CA mutations were indicative of later progression and metastasis in bladder cancer patients.
Resolution Biosciences, which has offered an NGS-based ctDNA test for non-small cell lung cancer for the last two years, also shared data from a study of a new assay for small-cell lung cancer, aimed at monitoring the disease burden of patients in treatment as a way of assessing therapeutic effectiveness.
The firm's SCLC assay, which targets a panel of 14 genes, was tested on 141 plasma samples from a cohort of 27 patients, 11 of which had earlier-stage disease. Researchers analyzed samples before, during, and after treatment and were able to detect disease-associated mutations in the cfDNA of 78 percent of patients.
Observed mutant allele frequencies in longitudinal samples tracked closely with treatment responses, the authors reported, and the assay showed an ability to detect the re-emergence of ctDNA mutations weeks before patients showed clinical signs of relapse.