CLICK ON weeks 0 - 40 and follow along every 2 weeks of fetal development
Uncovering origins for brain cancer
When Leland Dunwoodie, an undergraduate student in biochemistry, approached his lab director about wanting to start research on "some human stuff" in the spring of 2016, he didn't imagine it would lead to his discovery of 22 genes now implicated in glioblastoma, the most aggressive form of brain cancer.
"I definitely didn't come to Clemson University thinking about brain cancer research," Dunwoodie says. "I was working on a project with grapes and other plants. I told Dr. [Alex] Feltus that I wanted to do some human stuff, and he said, 'That's cool - pick an organ.'" But after consulting with his family, should he study the brain or the heart?, Dunwoodie decided on the brain, and specifically on brain cancer. A prior summer internship at the Van Andel Institute had spurred his interest in cancer research.
Fast-forward two years to a January 2018 publication in the journal Oncotarget, Dunwoodie is the first to describe glioblastoma specific gene co-expression between a group of 22 specific genes.
In the news as the disease afflicting Senator John McCain and Beau Biden, the late son of U.S. Vice President Joe Biden, glioblastoma is highly malignant and known for its lethality. Patients with glioblastoma survive on average only 14.6 months after diagnosis.
"Like many other tumors, diseases, and complex traits, glioblastoma is controlled by a variety of genetic and epigenetic factors," Dunwoodie explains. "If there was one master-regulator of these cancers, we'd say, 'We're going to drug that, and save millions of lives every year,' but there are more things going on in glioblastoma than we can presently identify."
However, the complexity of glioblastoma fits the Sytems Genetics approach of professor Alex Feltus's Laboratory in the department of genetics and biochemistry, where Dunwoodie is a student. Systems genetics uses computer and math based approaches to analyze biological systems, such as genes and their regulatory pathways.
To make the discovery, Dunwoodie compiled data from two online public databases for genomic information: The Cancer Genome Atlas (TCGA) and the National Center for Biotechnology Information (NCBI). From TCGA, more than 2,000 tumor expression datasets were downloaded, each one detailing on a genetic level how tumor cells differ from normal cells. Five different types of tumors, including those from bladder, ovarian, thyroid, lower-grade glioma and glioblastoma cancers, were included to achieve a well-rounded case study.
The 2,000-plus datasets, each showing approximately 75,000 genes, were next organized into a gene expression matrix (GEM) table to quantify the expression level of each gene across every sample. For example, one of the genes pulled from TCGA, called LAPTM5, encodes a protein involved in the formation of blood cells. So, the LAPTM5 expression matrix was assessed across each tumor type to gauge whether it is overly active (overexpressed) or underactive (underexpressed), and given a number ranking. The same process was then used on the 74,999 remaining genes across five tumor types in TCGA.
A separate GEM, encompassing 210,000 genes from 204 datasets from the NCBI database - including normal brain samples, glioblastoma brain samples and brain samples from patients with Parkinson's disease - was created independently for comparison. Will Poehlman, a graduate student in the Systems Genetics Lab, assisted Dunwoodie in preparing these GEMs.
Using novel computer software developed by Feltus and former graduate student Stephen Ficklin, now assistant professor at Washington State University, Dunwoodie was able to translate the GEMs into 2 different gene co-expression networks (GCNs), in a visual representation of how genes interact.
The software package, known as Knowledge Independent Network Construction (KINC), is novel in that it finds expression relationships between genes without the researchers having to conduct any prior analyses. This knowledge-independent method reduces the amount of "noise," from laboratory protocols or from natural variation between cells, that can prevent genetic interactions being discovered.
"Through the two GCNs, we found a group of 22 genes that were co-expressed in a single module both in The Cancer Genome Atlas network and in the NCBI brain network," Dunwoodie adds. "Only about 70 genes overlapped between the two networks, and 22 of them were in the same module - the same group of co-expressed genes. The overlap was really easy to spot."
While it's tempting to think that the genes - many of which function in the immune system - are feeding off one another to affect glioblastoma, Dunwoodie says this isn't exactly the case.
Dunwoodie: "It's hard to say that they're working together, as these are correlations. If person A runs eight miles on the same day that person B runs eight miles, it doesn't necessarily mean they're running together. It's more likely these genes are regulated in the same way, and the several things regulating them we can't currently identify."
What's more, these 22 genes compared between glioblastoma and noncancerous brains, were found to have much stronger co-expression in glioblastoma, suggesting a disease-specific regulatory mechanism. The same finding was uncovered after comparing glioblastoma to lower-grade glioma, a less aggressive type of brain cancer, also indicating a glioblastoma-specific activity in the 22 genes. The other notable finding is that these 22 genes are more associated with mesenchymal glioblastoma, a distinct subtype. And when those same 22 genes are highly expressed, they decrease survival time for patients in the mesenchymal group.
As is the case in research, answering one question results in a plethora of new questions. The work is just one step toward understanding glioblastoma pathogenesis. Dunwoodie: "It would be nice to find out what the 22 genes are specifically doing. Are they expressed in the surrounding immune cells? Are they a cause of cancer, or are they an effect of cancer? Does cancer propagate their expression? Why these genes are co-expressed there and what they're doing are questions that haven't been answered."
Dunwoodie - who plans to attend medical school to become a physician informaticist - says the tools and methods he learned in the Systems Genetics Lab will stick with him long into his career. Dunwoodie: "Cancer research is interesting because there are so many amazing people doing so many amazing things - but this is just one drop in the bucket. For me, the real purpose is patients being cured. Getting a paper published is great, but no one was immediately cured because of this, and that's the ultimate goal."
Tumors exhibit complex patterns of aberrant gene expression. Using a knowledge-independent, noise-reducing gene co-expression network construction software called KINC, we created multiple RNAseq-based gene co-expression networks relevant to brain and glioblastoma biology. In this report, we describe the discovery and validation of a glioblastoma-specific gene module that contains 22 co-expressed genes. The genes are upregulated in glioblastoma relative to normal brain and lower grade glioma samples; they are also hypo-methylated in glioblastoma relative to lower grade glioma tumors. Among the proneural, neural, mesenchymal, and classical glioblastoma subtypes, these genes are most-highly expressed in the mesenchymal subtype. Furthermore, high expression of these genes is associated with decreased survival across each glioblastoma subtype. These genes are of interest to glioblastoma biology and our gene interaction discovery and validation workflow can be used to discover and validate co-expressed gene modules derived from any co-expression network.
Authors: Leland J. Dunwoodie, William L. Poehlman, Stephen P. Ficklin and Frank Alexander Feltus
The authors declare no conflicts of interest.
This work used Clemson University’s Palmetto Cluster, Washington State University’s Kamiak Cluster (both high performance compute clusters) and the Open Science Grid (OSG). The OSG is supported by the National Science Foundation and the U.S. Department of Energy’s Office of Science. We acknowledge the assistance of M. Rynge, D. Balamurugan, and the OSG support staff for technical support and assistance. We acknowledge the assistance of J. Schipper from Van Andel Institute for reviewing the manuscript.
Return to top of page
Points represent transcripts (nodes in the brain) and lines represent significant
expression correlations or edges between nodes. Image credit: Oncotarget