NEW YORK – A new study by researchers from Duke University and their collaborators has demonstrated the utility of high-resolution single-cell spatial analysis for investigating the molecular underpinnings of lethal pediatric brain tumors.
Posted as a preprint on BioRxiv last month, the paper also appears to be the first to showcase customer-generated data using the 10x Genomics Xenium in situ analyzer, providing insights into the real-world performance of the platform.
The study focused on diffuse midline gliomas (DMGs), a type of deleterious central nervous system tumor that can be aggressive and spread to other areas of the CNS through the cerebrospinal fluid.
"These are lethal brain tumors that typically occur in children or young adults," said Zach Reitman, a radiation oncologist at Duke and the lead author of the study. "They are just devastating."
Researchers have previously lacked effective tools to comprehensively study these tumors for treatment development, Reitman said. Single-cell RNA sequencing, for instance, cannot reveal the spatial information about the tumor's immune and vascular microenvironment, he pointed out. Meanwhile, traditional bulk RNA spatial analysis methods fail to offer insights into the transcriptomic dynamics at the single-cell level.
To overcome these limitations, the researchers deployed single-cell in situ spatial analysis to examine the molecular mechanisms driving the etiology and treatment response of DMG tumors.
"We need to take the next step towards understanding the molecular pathology of the tumors," said Simon Gregory, director of the molecular genomics core at the Duke Molecular Physiology Institute and the corresponding author of the study. "We thought it would be an ideal case for using the Xenium platform."
As part of the study, the researchers first developed mouse models that are genetically engineered to inactivate the tumor suppressor p53 while expressing oncohistone driver alteration H3.3K27M, mimicking mutations associated with DMGs in the human brain.
They then applied different treatments to the mice models, including radiation therapy, which is the standard of care for DMGs, and inhibition of ataxia-telangiectasia mutated (ATM), a master regulator of DNA damage response that is the current target for pharmacological clinical trials.
Overall, the study showed that the combination of radiation and ATM inhibition led to greater tumor regression and better outcomes for the mice, Reitman said, prompting the team to employ the Xenium platform to further investigate the molecular underpinnings of the response.
For their study, the Duke researchers analyzed formalin-fixed paraffin-embedded (FFPE) mouse brain tumor sections on the Xenium platform. The analysis looked at the transcription profile of a total of 298 genes using the Xenium Mouse Brain Gene Expression Panel, including 50 customized gene targets that were added to examine DMG-specific markers.
High-resolution spatially resolved transcriptomic profiling was especially critical in this study since vasculature and neoplastic compartments within the DMG tumor play distinct roles in therapeutic response, the authors noted.
In the end, the analysis identified overexpression of the cell cycle regulator Cdkn1a as a putative resistance factor in ATM-intact DMG. The study also showed that Cdkn1a, or transcriptional activity of p53 in general, is dispensable for DMG radiosensitization by ATM loss.
"The resolution is just phenomenal," Gregory said. "To be able to drill down to individual cells and look at morphological features like vasculature associated with the tumor is just absolutely unprecedented."
Having installed the Xenium platform in March, Gregory’s core lab was one of the first customers to receive the instrument. Overall, he said, he is pleased with its performance, which was "straightforward" to onboard and has run "relatively smoothly" so far.
"There are a couple of hiccups, but I think that is bound to be expected since it is one of the first machines to be placed in the field," Gregory said, adding that 10x resolved these issues, which were mostly machine or software glitches, "pretty much straight away."
In addition, he said the custom content design for the platform was fairly easy and the performance of those assays has been very good.
For the type of experiments his team normally performs, Gregory said, the turnaround time is usually three or four days. With respect to cost, he noted around $1,000 to $1,500 per sample, depending on the number of samples that can fit within one run.
"This is a good example of how one could utilize a new technology, such as Xenium, to understand new biology of cancers," said Kunal Rai, an associate professor of genomic medicine at MD Anderson Cancer Center who was not involved in the study. Rai’s lab, which focuses on exploring the epigenetics of cancer progression or resistance to certain therapies, is also an early adopter of the Xenium platform.
Since receiving a Xenium in the spring, Rai’s lab has run over 150 samples on the platform, he said. These include various tumor types, such as melanoma, colorectal cancer, breast cancer, lung cancer, and human brain tumors.
Mirroring the Duke researchers’ experience, he said the instrument, which is "very easy to use," has been producing high-quality data that is "very sensitive" to low-abundance transcripts. The typical turnaround time for a Xenium run is about three to five days, he said, depending on the size and number of tissues analyzed.
When it comes to analysis, Rai said that currently, the Xenium Explorer software allows users to analyze one sample at a time. As such, a custom analytical pipeline will be necessary if a researcher plans to analyze multiple samples simultaneously, he pointed out.
In addition, 10x customer support has been "out of this world," he said. "The reason why we bought Xenium was because we worked with 10x for a while," he noted. "We really love their support system, and the data quality is always very good." Rai stressed that he was stating his personal opinion about the company and platform and was not speaking on behalf of his institution.
"The [Xenium] workflow has been seamless, definitely easier to onboard than some of our other spatial platforms," Olivia Koues, director of the advanced genomics core at the University of Michigan who was not involved in the Duke study, wrote in an email.
Koues said her team installed the Xenium platform in May and has already supported a wide range of samples — such as fresh frozen, FFPE, and fixed frozen — across human, mouse, and rat specimens.
She said customers "are happy with the resolution but almost always prefer to add custom content to the probe panels or design their own." In that regard, Koues noted that the customization process "is easy and not cost-prohibitive for [the most part]."
"Overall, this is a compelling study, and I'm excited to see more applications of cellular and subcellular data for cancer stratification," said Chris Mason, a genomics professor at Weill Cornell Medicine who was not involved in the study. "The company has presented data at conferences, and it looks very compelling, but it's always good to see [the platform] in the customers' hands."
Mason’s team is currently collaborating with NanoString Technologies on an initiative to create multiomic spatial maps of healthy human tissues from multiple organs at subcellular resolution. His lab is also an early customer for NanoString’s CosMx Spatial Molecular Imager, which also promises to enable high-plex single-cell spatial analysis at near 50 nm resolution — high enough for subcellular details.
While Mason said his lab is currently using the CosMx 1,000-plex RNA panel by default, it is also getting early access to the company’s newer panel that can accommodate 6,000 gene probes. Additionally, he said the lab has done protein and RNA analysis on the same samples using the CosMx platform.
In terms of turnaround time, Mason shared that a typical short run normally takes two days, while a larger run that covers a larger area and more samples could last four days.
While Xenium appears to be well received by its early adopters, it remains to be seen how much market share the instrument will garner for 10x given the increasing number of competitors — such as Resolve Biosciences, NanoString, Vizgen, and Akoya Biosciences — many of which are also entangled in IP ligation with the company.
"It may seem obvious, but the fact that all of this data was generated completely autonomously in their lab — with zero help from 10x — demonstrates the caliber of this technology and the commercial readiness of the platform," Nikhil Rao, VP of product management and Xenium business leader at 10x Genomics, wrote in an email. The company said it surpassed 100 Xenium shipments in August.
"We love Xenium for sure, but the limitation is the number of probes," MD Anderson’s Rai noted. "We are of course also exploring other [platforms] — mostly CosMx from NanoString."
Meanwhile, some researchers are seemingly prepared to stick with Xenium in the long run.
"They're definitely not at the plexity that CosMx is at the moment, but I expect their roadmap to take us to that level of panel complexity," said Duke’s Gregory, who also highlighted Xenium’s ability to analyze RNA isoforms, which are associated with cancer development, at the single-cell level.
"It will be great if it's not a monopoly, we'd like to see competition driving innovation," Gregory said. "Looking at things like throughput, cost, robustness of the platform, support from the company … makes me think that we've made the right decision."