Research

Rational clinical decisions on the management and treatment of cancer rely on accurate diagnostic information. Molecular analysis of tumour samples has been used to predict prognosis or response to treatment, but should be complemented by non-invasive methods for monitoring disease progression or dynamics. Circulating DNA in plasma and serum include tumour-specific sequences that are a promising source of diagnostic information.

The mechanisms through which tumour DNA reaches blood circulation are unclear, although fragmentation patterns of DNA in the plasma of cancer patients suggest it may originate from cell death. Overall levels of circulating DNA are higher in cancer patients compared with healthy controls, but these differences are not consistent enough for robust diagnostic tools. The maturation of genomic technologies allows circulating tumour-specific DNA to be used as personalised biomarkers (Figure 1).

Circulating tumour DNA (ctDNA) can be measured by tying together genomic and molecular techniques. First, tumour-specific somatic alterations must be identified on a case-by-case basis. Second, sequence-specific molecular assays must be designed that can precisely detect and measure tumour-specific sequences in the background of circulating genomic DNA. Finally, these assays must be applied to body fluid samples such as blood plasma that have been carefully collected and processed to extract circulating DNA.

Circulating tumour DNA may be useful for identifying the presence of cancer mutations, for detecting systemic or residual tumour burden, or for non-invasive monitoring of tumour changes. Preliminary studies suggest that ctDNA compares favourably to imaging or to currently used protein markers. Our goal is to translate this potential into diagnostic applications, by integrating new quantification methods and computational insights with clinical research.

Measurement and noise in molecular biology

Quantitative measurements in molecular biology are challenging; objects of study are highly sensitive biochemical systems and repeated sampling is limited since living organisms are highly variable and dynamic. Reliability depends on our ability to take into account biological variation, measurement noise and biases. In earlier studies (at the Weizmann Institute of Science), time-lapse microscopy and fluorescent reporter fusions were used to study gene regulation circuits. These studies demonstrate one approach to overcoming biological variation, by performing measurements in individual living cells.

Medical diagnostics poses different challenges. Clinical samples are often limited and heterogeneous, and can vary in collection conditions or contain a mixture of tumour and other material. Molecular quantification methods introduce additional noise and bias. We need to understand these effects and consider their impact on the design of diagnostic tests.

We are studying collection and processing protocols for peripheral blood samples, to optimise these for measurement of ctDNA and adapt them for simplified clinical use. We quantify DNA using parallel or 'digital' PCR, arguably the most accurate method for quantification of nucleic acid sequences. Template molecules are distributed into multiple independent reactions, reducing background interference. Quantification is obtained through counting of positive amplifications, and does not rely on calibration standards or curves.

Figure 1
Workflow for studies on circulating tumour-specific DNA. DNA obtained from a patient’s tumour or biopsy sample is used to identify tumour-specific genomic alterations. Assays are designed to specifically measure these tumour-specific DNA sequences. Assays are validated using tumour DNA as positive control and DNA from other subjects (and normal) as negative controls. The assays are used to measure ctDNA levels in blood samples from the same patient. These data are compared to clinical information to study ctDNA dynamics and diagnostic potential.

Diagnostic algorithms

A major challenge in designing diagnostic tests is in defining categories that are clinically informative and can also be robustly identified. Tumours can be classified, for example, as positive or negative for hormone receptors, indicating the suitability of hormonal treatment. To be effective, diagnostic algorithms need to take into account both measurement limitations and clinical considerations.

In previous projects (at Rosetta Genomics Ltd.), microRNA expression levels were used to classify tumour histological types and sites of origin. Classification was based on strong biomarkers and intuitive, 'logical' decision criteria. The robust design of these algorithms enabled their rapid translation into clinical tests. This practical approach to molecular classification is likely to be effective in translating other types of diagnostic assays into clinical practice.

We use state of the art genomic tools to identify somatic changes in the DNA from a tumour or biopsy sample, and design tumour-specific molecular assays. The complex analysis of tumour material shifts the burden of proof and makes the measurement of ctDNA in blood samples direct and unequivocal. We believe that these personalised biomarkers will prove to be highly informative and clinically effective.

Non-invasive diagnostics using ctDNA

The study of ctDNA requires carefully collected samples from clinical studies that include tumour or biopsy material and matched collections of blood samples. It is possible to accurately measure ctDNA in these samples using the methods that we are developing and utilising. These data must then be compared to clinical follow-up data to identify associations and potential diagnostic roles of ctDNA (Figure 1).

We work in close collaboration with clinical groups (such as the Brenton, Caldas and Neal laboratories at the CRI) to study the dynamics and utility of ctDNA in epithelial cell cancers, with the aim of developing findings into robust diagnostic assays. In 2010, we recruited the lab's first postdoctoral scientists, integrated technologies, and produced our first proofs of concept. These show early evidence for the informative content of ctDNA as a personalised biomarker for cancer monitoring, as well as its analytical robustness.