NEW YORK (GenomeWeb) – A single-cell RNA sequencing analysis on thousands of cells from half a dozen oligodendroglioma tumors supports the notion that cancer stem cells can, indeed, contribute to proliferation of the brain and central nervous system cancer in humans.
Researchers from the Broad Institute, Massachusetts General Hospital, and elsewhere did single-cell RNA-seq on more than 4,300 individuals cells isolated from six oligodendroglioma tumors marked by either IDH1 or IDH2 driver mutations. Based on the cell types, cell stages, and genetic alterations that could be teased out of the single-cell transcriptomic data, they got a more refined look at the cell types involved in oligodendroglioma than has been possible in the past — directly detecting stem cell-like sub-populations in human forms of the disease. The study was published today in Nature.
"It's the first evidence that the cell types involved in the self-renewing of this tumor have a neural stem cell identity or a neural progenitor identity," co-corresponding author Mario Suvà, a pathology researcher affiliated with Mass General and the Broad, told GenomeWeb. Suvà, who co-led the study with the Broad Institute's Aviv Regev, noted that these and other results point to a role for cell types that are slightly less differentiated than those implicated in the disease in the past.
More generally, he added, the single-cell RNA-seq approach made it possible to learn about the development, differentiation, and genetic evolution more or less in parallel for the sub-clones present in a given tumor.
There is ongoing debate over the role that cancer stem cells play in human brain tumors, if any. While studies using human tumors that have been xenotransplanted into mice have bolstered the hypothesis that cancer stem cells are key contributors to the development of such diseases, Suvà explained, such assays don't necessarily reflect what is happening within humans due to inter-species differences, potentially selection, and other confounding factors.
"People still don't really know exactly how those kinds of observations apply to human tumors in situ," he said. "The biggest criticism of the cancer stem cell model has always been that maybe when we do these kinds of assays we're selecting for specific, aggressive [genetic] clones that really have not much to do with a stem cell program or stem cell identity."
Beyond those considerations, it has been tricky to study oligodendroglioma development because the rare, slow-growing brain tumor type is not amenable to mouse xenograft experiments or other functional assays.
To take a better look at the molecular and biological underpinnings of human oligodendroglioma, he and his colleagues turned to single-cell RNA-seq to get a glimpse at transcriptional signatures associated with cell type, cell function, and cell cycle stage, but also to profile genetic alterations present in these tumors.
With single-cell RNA-seq, "you're beginning to get a lot of information all of a sudden: you have genetic information, you have cell identity information, you have cell state information," Suvà noted. "You can reconstruct what's going on in patient samples in a pretty detailed way and without relying on any specific assay."
The researchers used the relatively well-established Smart-seq2 protocol to sequence the RNA in 4,347 individual cells that were isolated from freshly resected oligodendroglioma tumors with the help of fluorescence-activated cell sorting. For some of their follow-up analyses, including testing for specific mutations in some of the cells, they also developed more tailored methods such as single-cell qPCR assays or refined mutation-calling algorithms.
The oligodendroglioma cases selected for the study all involved grade II, untreated tumors that contained IDH1 or IDH2 mutations in combination with a co-deletion involving sections of chromosome 1 and 19, the team noted. Between 791 and 1,229 individual cells were analyzed for three of the tumors. Though the remaining three tumors were not profiled quite as deeply, the group still had information on 430 to 598 cells from each.
With these data, the researchers narrowed in on signatures for differentiated cells and cancer stem cells using unbiased approaches that did not rely on external classifiers, Suvà said.
"These are signatures that come from our own data. The reason we could define them is that we actually pushed the cell numbers very far," he explained. "When you get to that number of cells, you can get confident signatures about sub-populations of cells because even rare populations begin to be represented pretty well."
To subsequently interpret these signatures, the team then turned to known cell types from the human or mouse brain, demonstrating that the oligodendroglioma tumors contained small sub-populations of cells that not only resembled neural progenitor cell or neural stem cells, but were also more likely to express genes associated with proliferation.
In the differentiated tumor cells, on the other hand, the researchers identified transcriptomic signatures that lined up with known expression patterns in mature astrocytes or mature oligodendrocytes, but expression patterns did not point to proliferation.
"The thought in the field based on normal development is that stem cells should be quiescent, or cancer stem cells should be quiescent, and we showed that while not all stem cells are cycling — it's only a subset — nevertheless the very major source of proliferating cells comes from the stem cell compartment," Suvà said. "The cells that are really not cycling are the ones that are differentiated."
Though it remains difficult to pin down the minimum number of cells that need to be RNA sequenced per tumor to detect such patterns, Suvà noted that it was difficult to pick up the rarer stem cell populations in an unbiased manner in the samples with just 500 or so sequenced cells.
"With 1,000 cells, the signature became quite clear, so we feel that it's a good base line," he said, noting that "we may need to do even more to get to sub-populations of stem cells."
Scaling up to look at more tumors at the same depth or to profile greater numbers of cells per sample is still somewhat challenging, though.
At the recent American Society of Human Genetics conference in Vancouver, Canada, 10X Genomics presented preliminary, unpublished data demonstrating that it could combine its Chromium technology, analytical pipeline, and Illumina sequencing to analyze expression patterns in 1.3 million individual mouse embryonic brain cells. The company plans to release reads from that experiment onto its website once it has generated some 20,000 reads per cell.
Suvà said his team is using 10X Genomics' Chromium technology for some of its experiments, since the throughput and accuracy are very high. He explained that the approach is not completely compatible with all of the RNA-seq experiments the group is pursuing, though, since it generates 3' end transcript sequences and may lose some tumor-related genetic information from the other end of the transcript.
There are some logistic considerations as well. Since clinical samples from tumor resections may arrive at all hours of the day, for example, Suvà noted that it is advantageous to use a protocol that can isolate and freeze down individual cells in 96-well plates for later processing — something that is not always feasible with commercial sample preparation kits.
Based on their results so far, Suvà and his colleagues are getting ready to apply their single-cell RNA-seq analyses to other brain tumor types in the clinical settings. They are also interested in using the strategy to track oligodendroglioma response to treatment with standard-of-care therapy, experimental drugs, or immunotherapies.
Further down the road, the team is considering the possibility of developing its own immunotherapy approaches based on its more refined information on the cell types involved in oligodendroglioma and the ways these cells interact with cells in the immune system, which will also be profiled by single-cell RNA-seq.