A GWAS Gets Into Shape

Researchers have combed through our DNA to pinpoint gene versions that make some people more vulnerable to MS. Now, a team of scientists has performed a similar search to identify genetic variants that may alter the architecture of brain lesions in MS patients. The work, published online this week in Brain (Gourraud et al., 2013), combines two techniques that researchers have often used separately to investigate MS: MRIs of brain lesions and the genome-wide association (GWA) study, often called a GWAS.

“It’s a difficult topic that’s been addressed using very innovative tools,” says MS researcher Murali Ramanathan of the State University of New York, University at Buffalo, who wasn’t involved with the study. “It’s a very well done paper,” adds neuropsychologist Heather Wishart of Dartmouth Medical School in Lebanon, New Hampshire, who commends the researchers for taking great care to weed out false positive results—a potential pitfall for analyses of hefty data sets like those in the study. 

To probe the inheritance of disease susceptibility, researchers often turn to GWA studies. The method identifies one-letter alterations in the genetic code known as SNPs that are more prevalent in people who have a particular illness than in healthy individuals (see “Genetic Associations”). GWA investigations have helped boost to more than 50 the number of gene variants that hike the odds of developing MS (Gourraud et al., 2012). Other projects under way could double that total within the year, says neurologist Pierre-Antoine Gourraud of the University of California, San Francisco (UCSF), a co-author of the new study. “Thanks to GWAS, our knowledge about MS susceptibility is exponentially increasing,” he says.

But GWA studies could furnish insight into more than just MS vulnerability, notes Gourraud’s colleague and co-author, Sergio Baranzini, an MS geneticist at UCSF. Such research might reveal genetic factors that explain why the course of the disease varies from person to person, for instance, and help clarify why the location and shape of brain lesions are often dissimilar among patients. “We recognize that the disease is heterogeneous, and we are digging deeper into those differences,” Baranzini says.

In their new study, the scientists took their cue from investigations of other diseases that used MRI-derived measurements of brain structures in a GWA study. A recent paper on Alzheimer’s patients (Hibar et al., 2011), for instance, merged the two methods to hunt for genes that might influence the shrinkage or expansion of certain brain regions. These studies also follow a trend in the GWA field of including quantitative variables in the analysis instead of qualitative variables like the presence or absence of disease. “We want to see if genetic variation has an impact on a feature of the disease that can be measured,” Baranzini says.

For the MRI side of their study, Gourraud, Baranzini, and colleagues had access to a scanner with greater resolution than those typically available in hospitals, so they were able to chart lesions in the brains of 284 MS patients to an accuracy of 1 cubic millimeter (see “More Than Meets the Eye”). The team divvied up the MRI scans of each patient’s brain into more than 60,000 voxels, the three-dimensional equivalents of the pixels in a digital photograph, and ascertained whether a lesion intruded into each one. Using a program they devised that groups lesion-carrying voxels, the researchers determined the shape of each lesion and its distribution within the brain, a measure they refer to as the lesion’s topology.

For the genetic side of the work, the team gleaned data on more than 208,000 SNPs for each patient. The scientists then ran GWA analyses to determine if any of these SNPs are associated with the number of lesion clusters per patient, the average distance between them, or their average size. The results were not significant or barely significant, indicating at best a weak relationship between the SNPs and these lesion variables. However, the researchers reasoned, these three measures might not capture the morphology of a lesion.

So, to try to nab a clearer picture of lesions from the more than 60,000 voxel measurements for each patient, the investigators applied principal components analysis, a statistical technique that “squeezes” large data sets and makes it easier to tease out potentially important factors. The technique extracts combinations of variables known as principal components that account for the greatest amounts of variation in the data. The first principal component explains the most variation, the second principal component accounts for less, and so on. If you used principal components analysis to evaluate various measurements of the human body, for example, the first principal component might be a mixture of height, waist circumference, and head diameter.

The first principal component that emerged from the team’s analysis of the MRI data accounted for about 10% of the variability in lesion shape and distribution, and correlated with lesion volume, or lesion load—but not with other variables, such as patient’s age, length of illness, or treatment status. That the shape of a lesion is related to its size is no surprise, Baranzini says: “We knew that lesion load would show up first.” The principal components analysis allowed the researchers to factor out this expected feature and use a GWA study to test whether genetic differences correlated with the next nine principal components, which the team suspected might better reflect lesion configuration. The eighth principal component was the only one with a genetic connection; it explained about 2% of lesion heterogeneity and correlated with 31 SNPs. “We hypothesize that it is capturing lesion topology,” Baranzini says.

What those results imply, the researchers conclude, is that patients’ genetics affect the architecture of their brain lesions. However, principal components are composite variables, so the findings don’t identify specific lesion attributes, such as length or curvature, that the genes influence.

Several of the SNPs that the researchers flagged lie within genes already implicated in the formation and location of MS lesions, and many of those genes take part in processes that can go awry in MS patients, such as myelination. One example is SYK, the gene for spleen tyrosine kinase, an enzyme that adds phosphate groups to myelin basic protein, a component of myelin, and that may help rouse the nervous system immune cells called microglia. To dig deeper into the genes’ biological context, the team performed a network-based pathway analysis to uncover molecular circuits that include the SNP-containing genes. The first network they pinpointed was heavy on genes that are active in immune cells or during central nervous system development and regeneration, bolstering the idea that they are involved in lesion growth. “We don’t say these genes determine the distribution of lesions,” Baranzini says. “But we do say they are associated with important brain functions.”

Researchers not connected to the study see promise in this combination of GWA studies and MRI methods. “It puts us on the path to finding out about the problem more systematically,” Ramanathan says. “It’s a great approach for discovering genes that may be related to the progression of the illness,” Wishart says. But she would like to know more about why only the eighth principal component showed any genetic association—and what this component signifies about the shape and distribution of a lesion.

“We call this paper exploratory,” Gourraud says, adding that replication of the study is important to bolster the case that gene differences help shape lesions. He and Baranzini suggest that the combined technique is ready to branch out. “You have a new way to investigate complex data sets,” Gourraud says. As researchers perform more and more quantitative studies like this one, they might be able to nail down how genetic variants contribute to nervous system changes in MS, Baranzini says: “It’s scratching the surface of studies to come.”

Key open questions

• How do specific gene variants change the shape of a lesion?
• Do gene variants that influence MS susceptibility also affect lesion configuration?
• Are the SNPs that are linked to a certain lesion phenotype associated with progression of the disease?

Image credit

Thumbnail image on landing page. Adapted from “Monthly multiple sclerosis anim.gif” from Waglione, 2008. Released under Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license.