Study Uses Open Data to Analyze “Normal” Tissue Near Tumors

October 23, 2017, by NCI Staff

Credit: Dvir Aran, Ph.D., University of California, San Francisco

Findings from a new study suggest that the tissue immediately surrounding a tumor may not be normal, even if it appears normal under the microscope.

Rather, tissue adjacent to a tumor—which is commonly used in research studies to represent a patient’s normal cells—has characteristics that distinguish it from both healthy and tumor tissue.

At the molecular level, the tissue appears to be in “a unique intermediate state” between healthy and tumor tissue, researchers at the University of California, San Francisco (UCSF), reported October 20 in Nature Communications.

Comparing healthy cells and tumor cells from the same patient can reveal molecular changes associated with cancer. For decades, doctors have obtained healthy cells by collecting tissue from the areas surrounding a tumor.

But little is known about the genomic features of this tissue. To learn more, the UCSF researchers combined and analyzed results from open-access databases for two projects supported by NCI and the National Institutes of Health (NIH): The Cancer Genome Atlas(TCGA) and the Genotype-Tissue Expression (GTEx) project.

Specifically, they compared the molecular profiles of 6,500 tissue samples: cancerous and adjacent tissue from TCGA and nontumor tissue from GTEx. The samples represented eight cancer sites in the body, including the breast, liver, colon, and prostate.

This analysis revealed a pattern across all eight sites: certain signaling pathways were activated in tissue adjacent to tumors but not in the tumors themselves or in non-tumor tissue from individuals without cancer. The researchers also identified a set of 18 genes that was expressed specifically in tissue adjacent to a tumor in at least four sites.

“Our findings suggest that tumors are interacting with the host tissues,” said study investigator Dvir Aran, Ph.D. “We think that probably all tumors have these types of interactions with adjacent tissues.”

Researchers have known that cancer cells interact extensively “with the surrounding microenvironment of the tumor,” added Dinah Singer, Ph.D., director of NCI’s Division of Cancer Biology. “What’s novel here is the suggestion that the interactions extend beyond that to tissue that appears to be pathologically normal.”

Integrating Results from Different Databases

Because healthy individuals do not normally donate organs for research, samples of normal tissue from living individuals without cancer are extremely rare.

The current study was made possible by the existence of the two open-access genomic databases, as well as by funding from NCI through the Oncology Models Forum , noted Atul Butte, M.D., Ph.D., of the Institute for Computational Health Sciences at UCSF, who led the study.

“We realized we could put these two major NIH-funded data sets together and ask: What happens in the tissue around cancers?” Dr. Butte said. “I am a big fan of using open data to ask questions that might not otherwise be addressed.”

The GTEx project includes genomic profiles of thousands of tissue samples from deceased donors, including many individuals who did not have cancer. And TCGA has genomic profiles of thousands of tumor samples and hundreds of tumor-adjacent samples from multiple tissues.

Before comparing the tissue samples, the researchers first showed that the results from two different databases could be analyzed together. They found, for example, a “high level of agreement” among the expression levels of certain genes within the entire data set.

The researchers acknowledged that a “a perfect study” would use freshly collected normal and cancer samples from the same individuals, but said that their results demonstrate that GTEx and TCGA data can be analyzed together.

“This is a very interesting study,” said Cheryl Marks, Ph.D., also of NCI’s Division of Cancer Biology. “It shows that you can combine data on normal tissue with data on cancer tissue and then use the results to tease out new insights.”

Scientists have always used normal tissue adjacent to tumors, Dr. Marks added. "But I’m sure this study will spawn a debate about what we mean by ‘normal’ tissue.”

A small number of studies over the past decade have compared tumor-adjacent tissue with tumor and healthy tissues for individual types of tissues. But this new study is the first to compare the molecular profiles, including the transcriptomes, of tissue samples from multiple types of tissues and organs.

“There’s some effect of the tumors on adjacent tissue,” said Dr. Butte. “We were surprised that there were common effects across the different types of cancer that we could study.”

Results Raise New Research Questions

The researchers also conducted experiments in mice to explore the idea of blocking certain signaling pathways associated with cancer that are active in the tumor-adjacent tissues. The results, they reported, suggest that disrupting the interplay between tumors and nearby tissues might be a therapeutic strategy.

“We plan to continue with mouse studies to see whether blocking the signals we uncovered will affect the tumor,” said Dr. Aran. “There are probably a mix of many factors involved, and more research is needed to better understand what we would be blocking.”

As with other computationally based studies, the results need to be confirmed independently before any firm conclusions can be drawn, the study authors cautioned. They cited limitations of the study, including that, because GTEx and TCGA used different protocols to collect and analyze the data, there may be disparities between sample sets.

Another limitation was the lack of detailed information about distances between tumor-adjacent tissues and tumors. Without this, the researchers could not determine whether the characteristics they observed in tumor-adjacent tissues were unique to these cells or part of the disease process of entire organs affected by the development of cancer, they noted.

“This study raises a number of questions, such as: If you’re trying to analyze a tumor, what’s the correct control?” asked Dr. Marks. “Is the adjacent normal tissue that scientists have used for decades the appropriate control? Or is it actually garden-variety normal tissue from another donor that someone collected during surgery?”

The answer, she added, may come from the research that builds on the current study. She and Dr. Butte both agree that the existence of large, publicly available databases has created an opportunity for researchers to explore new questions and develop hypotheses.

“The data repositories supported by NIH have incredible data, and we encourage more scientists to use them in their experiments,” said Dr. Butte. “The data are just sitting there—waiting to be used.”