The most common type of brain tumor, meningioma, grows from the membranes surrounding the brain and spinal cord.

Usually they are benign, which means they are not cancerous.

However, a subset of meningiomas behave aggressively, recur rapidly despite surgery and radiation, and often cause death.

To guess whether a tumor will be benign or malignant, pathologists typically examine a surgically removed sample of a tumor under the microscope and look for irregular cell shape, accelerated cell division, and other tell-tale signs of cancer.

But the eyeball test doesn’t always work.

Some stealthy tumors appear benign under the microscope but are just as aggressive and deadly as the malignant ones that look the part.

“The problem was they just kept coming back and people died of them,” said brain cancer surgeon Eric Holland, MD, PhD, who directs the Human Biology Division at Fred Hutch Cancer Center and holds the Endowed Chair in Cancer Biology. “There’s more going on than just the way it looks.”

Holland and his team at Fred Hutch have figured out a new way to classify tumors based on their underlying biology rather than their appearance under a microscope.

Holland’s team identified multiple subtypes that share similar genetics, including particularly aggressive clusters where stealthy tumors mislabeled “benign” fit right in with the deadly ones.

Their approach — described in a recent study that made the cover of the journal Cell Genomics —could improve meningioma diagnosis and could even guide treatment when applied to other solid-tumor diseases such as lung and breast cancer.

Stealth tumors that beat the eyeball test

The World Health Organization classifies meningiomas into three grades of severity based on how the tumor samples look under the microscope.

  • Grade 1 tumors are benign, which means they typically grow slowly in one place with defined borders and don’t pose an immediate threat. Some of them may not be discovered until the patient dies of something else.

  • Grade 2 tumors are atypical and require more scrutiny because they could become cancerous.

  • Grade 3 tumors are malignant, which means they can invade nearby tissues and spread to other parts of the body.

“All of those are just words to try to describe what you think is going to happen with a tumor once you take it out,” Holland said.

Will the tumor disappear or does it keep coming back, even after surgery and radiation?

Grade 3 tumors usually turn out to be as malignant as they look under the microscope. Cut them out and they come back. Blast them with radiation and they come back.

But some of the grade 1 and grade 2 tumors come back just as aggressively, and they don’t show any signs of having transformed into grade 3 tumors.

They still look benign under the microscope, even as they are killing the patient.

“There’s a bunch of grade 1s and 2s that also behave badly, and it wasn’t the least bit clear to anybody why,” Holland said.

Grabbing all the big data sets out there

Holland’s team, including the study’s lead author, Heshani “Nayanga” Thirimanne, PhD, a graduate research assistant at Fred Hutch, had a clue about where to start looking for answers.

A disorder called neurofibromatosis type 2 (NF2) attracted their attention because it is characterized by the loss of a copy of the NF2 gene, which plays a role in tumor-suppression.

When NF2 is missing, meningiomas proliferate, typically affecting the main nerve between the inner ear and the brain, leading to hearing loss and deafness.

Usually, NF2 goes missing because of the loss of chromosome 22, which harbors it, but it can also lose its function in other ways.

Because the majority of rapidly recurrent meningiomas are among those that show functional loss of NF2, the team wanted to understand the overall pattern of gene expression (which genes are turned on and off) in this aggressive subset so they could predict which tumors will fall into that category.

They started with 279 meningioma samples gathered in collaboration with Manuel Ferreira, MD, PhD, in the Department of Neurosurgery at the University of Washington, who also treats patients at Fred Hutch. The samples were graded 1, 2 or 3 and annotated with clinical histories about the kinds of treatment that patients received and whether that treatment worked.

“We actually know who ended up having to get another resection six months later,” Holland said.

They sequenced the UW tumors using a method that measures average gene expression across all samples, providing a comprehensive analysis of nearly 20,000 genes that provide the instructions for making proteins, the molecules that do most of the work in the cell.

Then they combined their UW data with a dozen publicly available meningioma datasets sequenced in the same way from nine institutions and five countries in North America, Europe and Asia.

This enabled them to amass the field’s largest meningioma dataset to date, comprising nearly 1,300 tumors sampled from patients all over the world combined with detailed clinical treatment histories for many of those cases.

“Basically, we grabbed all the big data sets out there,” Holland said. “The more you have, the better.”

Making a Map

Using computational tools invented at Fred Hutch, Holland’s team simplified that information, which comprises millions of data points, and represented it graphically on a digital reference map for meningioma tumors.

The map revealed a complex landscape of seven general regions of tumors further divided into several distinct subtypes based on similar genetics, tumor severity and treatment outcomes.

Think of it like a flood map that tells you which neighborhoods are unsafe based on whether the houses are built in a flood zone.

The malignant tumors clustered together like the houses built in flood zones, and the benign ones clustered together on higher ground.

The meningioma map also revealed how stealth tumors beat the eyeball test.

Although they appear benign under the microscope, their location on the meningioma map places them near clusters of malignant tumors because they share the same underlying biology.

They are like houses built so close to flood-zone houses they share the same risk.

“There’s two or three places where there’s a high concentration of people who do poorly,” Holland said. “It is true that some of those are grade 3s for sure, but there’s a bunch of grade 1s and 2s that are living right next to them, and they have just as bad an outcome as the grade 3s. What you’re really doing is you’re matching biology — forget the microscope.”

Finding a surprising connection

More than half of meningiomas in the dataset exhibit functional loss of the NF2 tumor-suppressing gene, which formed one big region of the map.

A subset of that region had an increased concentration of grade 2 and grade 3 tumors, but that grouping also contained stealthy grade 1s.

When the team delved deeper into what was going on with the deadliest tumors, they made a surprising discovery: The most aggressive tumors contained cells with gene expression characteristics of the earliest stages of life when embryos begin to form arms and legs.

 “What makes them different than the rest of the meningiomas is their similarity to embryonic development of limbs, muscles and nerves,” Holland said.

The tumors don’t look like embryos, but they express the same genes that are turned off and on during embryonic development and their cells resemble the cells that give rise to muscles and nerves.

Holland said more research is needed to understand the significance of this unexpected finding, but it’s intriguing because both malignant tumors and embryos undergo rapid cell growth.

It was a connection only their map location — not the microscope — could make visible.

“You don’t have any idea it’s there,” Holland said. “You need this volume of data.”

Locating new patients on the map

This study is the first to load all of its figures into an online, interactive data visualization tool called Oncoscape, also invented at Fred Hutch, which helps scientists navigate and analyze the complex molecular landscapes of tumors.

Interactivity is important because Holland’s team also devised an algorithm to overlay genetic data about new patients’ tumors onto the map.

Doctors can see whether their patients land closer to a neighborhood in a flood zone or closer to the houses on higher ground. They can also see how those with similar biology responded to various treatment options to get a better idea of what might work for their own patients.

“You can actually predict the outcome based on the nearest neighbors of where you land,” Holland said.

That feature isn’t immediately helpful for doctors treating meningioma because there are no targeted therapies yet that work on the subtypes of meningiomas the team revealed in their analysis.

“It’s possible that we’ll come up with something for a particular subtype now that we know they exist, but the reality is that just because you understand the problem doesn’t mean you can solve it,” Holland said.

However, the potential for improved diagnosis and treatment in other cancers has generated substantial interest in their work since the study was published.

“The ones we know the most about are lung and breast cancer,” Holland said. “I got invited to give a lot of talks around the world.”

This work was supported by funding from Seattle Translational Tumor Research and the Fred Hutchinson Cancer Center Genomics and Bioinformatics Core Facility.

This article was originally published July 15, 2024, by Fred Hutch News Service. It is republished with permission.