A simple blood test can reveal the geographic relationships among healthy cells surrounding a cancerous tumour, researchers at Stanford Medicine and the Mayo Clinic have found.

The test is the first noninvasive way to study what’s called the tumour microenvironment, which plays a critical role in determining how different patients — even those with similar tumours — fare after diagnosis and treatment.

In addition to the blood test, the researchers identified nine cellular neighbourhoods, or spatial ecotypes, that cancers of all types share and some of which correlate with a tumour’s response to immunotherapy and a patient’s prognosis.

Because the blood test can be performed repeatedly, clinicians may soon have real-time access to information about which types of therapies are likely to be most successful.

“To date, cancer therapy has been very much like whack-a-mole,” said Aaron Newman, PhD, associate professor of biomedical data science.

“We have targeted therapies that are effective against cancers with certain mutations, but many patients with similar tumours respond differently to therapy and have widely variable clinical outcomes. What have we been overlooking? The ecosystem of nonmalignant cells around the tumour.”

Newman and Aadel Chaudhuri, MD, PhD, a professor of radiation oncology at the Mayo Clinic in Rochester, Minnesota, are the senior authors of the research, which will be published in Nature on May 6.

Postdoctoral scholar Wubing Zhang, PhD; graduate student Erin Brown; and Abul Usmani, PhD, a senior scientist from Washington University in St. Louis, are the lead authors of the study.

Cunning cancer cells

At first blush, it seems healthy cells that find themselves with a cancer in their midst would fight back valiantly.

And to some degree that is true.

But cancer cells are wily, exploiting normal cellular processes to build new blood vessels that provide nutrients and oxygen to the tumour or convincing would-be killers in the immune system to turn a blind eye to their presence — even to tone down other aspects of the immune response.

They convince cells called fibroblasts to churn out collagen to make scaffolds supporting tumour growth and to block immune cells’ access.

In short, their portfolio of trickery is ample.

And once a permissive microenvironment has been established, the cancer can spread more easily.

“Like plants that thrive in some soils and die in others, cancer cells exhibit vastly different growth patterns in different cellular environments,” Newman said.

“Our goal has been to understand on a large scale how these environments, or spatial ecotypes, affect cancer growth and response to therapies and how they vary from patient to patient.”

Until now, the ability to suss out who is rubbing membranes with whom, and what they might be saying to each another, has hinged on the close study of thin slices of tissue obtained via biopsy — the relationships frozen in place on a microscope slide — or by analysing the genetic messages humming along the internal wires of cells that make up a bulk tissue sample (also obtained by biopsy) from the tumour and its surroundings.

But biopsies are invasive and performed only occasionally if at all during a patient’s course of treatment.

Furthermore, because cellular interactions and relationships evolve as the disease progresses and in response to different treatments, a one-time look via biopsy has limited clinical usefulness: Imagine attempting to determine current social relationships among attendees of a 40-year class reunion by studying the couples and groups immortalised in the pages of a long-ago high school yearbook.

In this study, the researchers used machine learning tools developed in Newman’s lab to analyse the relationships between healthy and cancerous cells in two major solid cancer types: carcinomas and melanomas.

Carcinomas make up 80% to 90% of human cancers, including those of the breast, lung, colorectal tissue, liver and prostate.

One of these tools, called CytoSPACE, which Newman’s team designed in 2023, maps cells to precise locations in tumour tissue — forming a detailed map of the tumour and healthy cells.

Another, called Spatial EcoTyper, predicts the various types of cells in a tissue sample based on the relative abundance and patterns of genetic messages in the sample.

It then determines what the cells are up to (a condition called cell state) and which other cells they are interacting with.

The result is a collection of cellular neighbourhoods, or spatial ecotypes, with specific characteristics.

“For the first time, we can really look at cells and how their social networks and cell states differ depending on where they are localised in the tumour,” Newman said.

A broad look at cancer cell environments

The researchers studied more than 100 tumour specimens from 10 distinct types of cancer using the tools they had developed to map patterns of gene expression in nine cell types at varying locations throughout the tumour.

They identified nine distinct spatial ecotypes, or neighbourhoods, each roughly the diameter of a human hair.

They found that patterns of spatial ecotypes were conserved among all the tumours they studied; some ecotypes were more likely to occur at the border of the tumour and healthy tissue, while others were more likely found deeper inside the tumour, for example.

Several of the newly identified ecotypes correlated with whether a tumour would respond to immunotherapy — suggesting they could help guide clinical decision-making.

“Now we can begin to understand cancer heterogeneity beyond just the cancer cells,” Newman said.

“Each spatial ecotype has its own internal social network: The cellular behaviours, or genetic programmes it’s carrying out, are influenced by the cells around it. So, an immune cell like a CD8 T cell might be found in several different neighbourhoods, or locations within the tumour, but it will be expressing different genes based on the nearby cells.”

In other words, what you confide to your closest friends over coffee is likely different from the conversations you might have with acquaintances at a party.

But there are some commonalities, the researchers found.

“Each ecotype has its own set of consensus programmes independent of cell type, which was a really interesting finding we didn’t expect,” Newman said.

“We thought that every cell type would express a unique genetic programme, and they do. But there are also programmes that are shared within ecotypes.”

To extend the previous example, party conversations tend to swirl around common topics: What do you do for work?

Where do you live?

Do you have plans for the weekend?

In contrast, a meeting of condominium shareholders might focus on cleaning common areas or hiring a new doorsman, while a group of close friends might share relationship advice and parenting tips.

Newman and his colleagues wondered whether neighbourhood-specific conversation types, represented by pooled gene expression patterns, could be captured in a blood test.

DNA circulates in the blood when cancer or healthy cells die and release their contents into the circulation.

Fortunately, telltale genetic signposts called methyl groups indicate which genes along a stretch of DNA were actively expressed in their cell of origin.

“We had this epiphany that we could use tools that determine these methylation patterns to determine the gene expression patterns in the cells from which they were derived,” Newman said.

“And that could serve as a readout for cell state and type in the tumour microenvironment.”

To test this, they developed another tool, called Liquid EcoTyper, that harnesses artificial intelligence to reconstruct the tumour microenvironment from methylation patterns.

They then validated their idea through multiple subsequent experiments to compare the predicted ecotypes from biopsy and surgical samples with those derived from blood samples from the same patient.

“Now we can infer these clinically vital, spatial landscapes in a tumour without having to do any tissue biopsy at all,” Newman said.

Although more studies need to be completed before the blood test is approved for routine use in the clinic, Newman and his colleagues envision a time when it can help a clinician decide which treatment to use first, then follow the evolution of a tumour microenvironment to determine when it might be necessary to switch therapies.

“This technique has the potential to be much more holistic and powerful than any current method of tracking the tumour microenvironment,” Newman said.

“The clinical possibilities are exciting.”

Researchers from the Mayo Clinic, Oslo University Hospital and the University of Oslo, Washington University in St. Louis, the Medical College of Wisconsin, the Clement J.

Zablocki VA Medical Centre, Columbia University, Yale University, the University of Rochester School of Medicine, and the Chan Zuckerberg Biohub contributed to the research.

The study was funded by the National Institutes of Health (grants K08CA237727, P50CA121974 and R01CA283317), the Research Council of Norway, the National Science Foundation, the Cancer Research Foundation, the V Foundation for Cancer Research, Alvin Siteman Cancer Research, the Melanoma Research Alliance, the Virginia and D.K. Ludwig Fund, Stanford Bio-X, and the Chan Zuckerberg Biohub Newman is a Chan Zuckerberg Biohub investigator and a member of the Stanford Cancer Institute.

He has patent filings related to digital cytometry, liquid biopsy and cancer biomarkers.

He has served as a consultant to, and has ownership interests in, CiberMed and LiquidCell Dx.

Source: Stanford Medicine