Molecular pathology selected abstracts

Create PDF

Editors: Donna E. Hansel, MD, PhD, chief, Division of Anatomic Pathology, and professor, Department of Pathology, University of California, San Diego; James Solomon, MD, PhD, resident, Department of Pathology, UCSD; Richard Wong, MD, PhD, molecular pathology fellow, Department of Pathology, UCSD; and Sounak Gupta, MBBS, PhD, molecular pathology fellow, Memorial Sloan Kettering Cancer Center, New York.

Analyzing fragment sizes of cell-free DNA to improve sensitivity of cancer detection

July 2018—The use of liquid biopsies to non-invasively detect and monitor cancer is rapidly expanding. In these assays, fragments of cell-free DNA (cfDNA), which are released when cells undergo necrosis or apoptosis, are isolated from the patient’s plasma and sequenced. Because cfDNA can originate from cancer cells and normal cells, the variant allele frequency of cancer-specific somatic mutations can often be very low. Therefore, the clinical utility of liquid biopsies is greatest when the patient’s alterations are already known, which is typically the case when monitoring residual disease, examining clonal evolution, and assessing therapy response and resistance. However, the sensitivity is lower for early detection and characterization of cancer. DNA in the bloodstream will rapidly degrade, except where the DNA is tightly coiled around histone proteins to form nucleosomes. The DNA that wraps around the nucleosome measures 166 base pairs, so many of the DNA fragments that are isolated are close to this size. Recent studies have analyzed the size distribution of cfDNA fragments in cancer patients, but the results have provided conflicting information about fragment size. Some recent studies have also shown that selecting for small fragments of cfDNA can enrich for cfDNA originating from cancer cells. The authors of this study developed a method for profiling cfDNA fragment sizes, called DNA evaluation of fragments for early interception, or DELFI. In this method, whole genome sequencing of cfDNA is performed at low coverage, and the fragment size distribution is examined across the genome in windows of five megabases. Within each window, the authors compared the fragment size distribution of cfDNA from cancer patients with that from healthy controls. The healthy controls had very similar, concordant fragmentation profiles, whereas the cfDNA from cancer patients was highly variable in length and did not correlate with the median healthy patient profile. The authors hypothesized that the cfDNA from healthy people derives mostly from circulating lymphocytes. To confirm this hypothesis, they tested lymphocyte nucleosomal DNA created by isolating nuclei from lymphocytes and treating them with nuclease, which showed that the fragment size distribution across the genome was very similar to the cfDNA from healthy people. To test the clinical sensitivity of DELFI, the authors studied cfDNA from 208 patients with various cancer types and compared the fragmentation profiles to those of 215 healthy individuals. The authors used a machine-learning model to report a predictive score indicating whether a patient’s cfDNA had the fragment size variability more characteristic of a cancer patient or a healthy person. The model correctly classified 152 of 208 cancer patients and misclassified only four of 215 healthy patients (73 percent sensitivity and 98 percent specificity; area under the curve, .94). DELFI was more sensitive than mutational analysis of cfDNA by targeted next-generation sequencing (82 verus 66 percent sensitivity for cases examined by both modalities). When targeted sequencing and DELFI were used together, the combined sensitivity was 91 percent and the specificity was 98 percent. Moreover, the authors showed that fragmentation profiling could potentially offer additional information about site of origin. They concluded that adding cfDNA fragmentation profiling to somatic mutation analysis could enhance the sensitivity of liquid biopsies and become a viable strategy for cancer screening.

Cristiano S, Leal A, Phallen J, et al. Genome-wide cell free DNA fragmentation in patients with cancer. Nature. 2019. https://doi.org/10.1038/s41586-019-1272-6.

Correspondence: Dr. Robert B. Scharpf at rscharpf@jhu.edu or Dr. Victor E. Velculescu at velculescu@​jhmi.edu

Identification of noncoding mutations that contribute to risk of autism

A growing body of research is clarifying the genetic underpinnings of autism spectrum disorder, a neurodevelopmental disorder characterized by impaired social interaction and communication. A significant recent finding is that de novo mutations and copy number alterations contribute to the genetic cause of the disorder. In simplex autism spectrum disorder (ASD) cases, defined as cases in which only one family member has ASD, it is thought that de novo alterations affecting protein-coding genes contribute to up to 30 percent of cases. However, many de novo mutations affect intronic and intergenic regions, and little is known about how these regions contribute to ASD. To assess the contribution of noncoding mutations, the authors tapped into the Simons Simplex Collection, in which whole genome sequencing data were collected from 1,790 simplex families. The advantage of using this collection is that genomic data are available not only from the patient and both parents but also from unaffected siblings. This means that a patient’s de novo mutational profile, obtained by comparing his or her DNA to those of both parents, can be compared to the de novo mutational profile of unaffected siblings. Evaluating the biological and functional effects of noncoding mutations and their contribution to disease phenotype is notoriously difficult. To attempt to do so, the authors trained a deep neural network-based framework to interpret each mutation. The model was trained at both the transcriptional level, for which it examined transcription-factor binding, chromatin marks, and chromatin assembly, as well as at the post-transcriptional level, for which it examined interactions between the transcribed RNA and RNA-binding proteins. Through this training, the model could predict the quantitative impact of noncoding mutations to single-nucleotide resolution. Using the model, the authors found that the functional impact of the de novo mutations was significantly higher in probands than in unaffected siblings. When tissue-specific expression profiles were used to analyze the effects of the variants in various tissues, they discovered that the highest functional impact of the mutations was in brain tissue-specific genes. Furthermore, the authors identified the two pathways and processes most significantly affected by the de novo mutations as the synaptic cluster, which consists of genes involved in synapse organization, synaptic signaling, and neurotransmission, and the chromatin cluster, which includes genes involved in chromatin regulation and neurogenesis. Both pathways contain numerous genes shown to be associated with ASD in previous studies. Finally, the authors performed a cell-based experimental assay to examine the effect of some of the highest impact noncoding mutations affecting transcription. They confirmed differential expression in 96 percent of the mutations predicted by the model. This work further elucidates the mutational spectrum of ASD and highlights the role of noncoding mutations in human disease in general. Interestingly, the de novo noncoding alterations converge on genes and pathways known to be associated with ASD, suggesting that they contribute to ASD phenotype formation.