In recent years, a flurry of studies has discovered that tumors harbor an extraordinary variety of bacteria, fungi and viruses. These surprising discoveries have led many scientists to rethink the nature of cancer.
The medical possibilities were exciting: If tumors released their characteristic microbes into the bloodstream, could these serve as an early marker of disease? Or could antibiotics even shrink tumors?
In 2019, a startup investigated these findings to develop microbe-based cancer tests. This year, regulators agreed to prioritize the company’s upcoming trial because of its promise to save lives.
But now several research teams have cast doubt on three of the most prominent studies in the field, reporting that they were unable to reproduce the results. The supposed tumor microbes, critics said, were likely mirages or the result of contamination.
“They just found things that weren’t there,” said Steven Salzberg, an expert in DNA sequence analysis at Johns Hopkins University who published one of the recent critiques.
Biologists have known for decades that at least some microbes play a role in cancer. The most striking example is a virus known as HPV, which causes cervical cancer by infecting cells. And certain strains of bacteria cause other types of cancer in organs like the intestine and stomach.
For decades, these links slowly came to light because scientists lacked much of the technology available today. The search accelerated dramatically when researchers learned how to extract fragments of DNA from tumors. Then they used computers to find out whether the genetic material came from human cells or cells from other species.
In 2019, a team of scientists at New York University School of Medicine used these techniques in a study of pancreatic cancer published in the journal Nature. In many tumors, they found fragments of DNA from a few different species of fungi. Further research led them to conclude that fungi were driving the growth of tumors.
These surprising results attracted the attention of Dr. Peter Allen, a surgeon at the Duke University School of Medicine, who began looking for microbes in his patients’ pancreatic tumors.
After searching 140 tumors, Allen and his colleagues were unable to find a significant amount of DNA from any microbes, including fungi. “We didn’t find any real signatures,” he said.
They then closely examined the original study, whose genetic data had been included in a public database. Allen’s team also failed to find a notable amount of fungal DNA in this data. They published their findings in Nature on August 2.
The NYU researchers defended their work. “My group still confirms what we found,” said Deepak Saxena, one of the authors of the original study. He pointed to other data in line with his results.
In August, for example, researchers at Tokyo Medical and Dental University reported the discovery of fungi in pancreatic tumors from 78 of 180 patients. And patients with fungus-containing tumors were at greater risk of dying in the three years after surgery, the study found.
Other researchers are questioning a 2020 Science report by a team at the Weizmann Institute of Science in Israel. Examining 1,500 tumors from seven cancer types, the study found that each tumor type had a different set of bacteria, with breast cancer harboring a particularly rich variety.
But Jacques Neefjes, a microbiologist at Leiden University in the Netherlands, and his colleagues were unable to detect bacteria inside cancer cells using some of the Weizmann team’s methods on their own collection of 129 breast cancer samples. “We haven’t found a single case,” he said.
In January, Neefjes’ group published a summary of their findings, which Science attached to the Israeli paper. They argued that the bacteria found by Weizmann’s team were byproducts of infections and are not, in fact, a normal part of breast cancer tumors.
Ravid Straussman, Weizmann’s research leader, said his group has carried out further studies and that “the results clearly confirm the presence of bacteria in cancer cells.” He also said it was impossible to evaluate the Neefjes team’s claims because they provided few details about their experiment.
In a third study, published in Nature in 2020, researchers at the University of California, San Diego analyzed a government tumor DNA database, called the Cancer Genome Atlas, and trained a computer to identify microbial DNA sequences from 18,000 tumors. The computer learned to recognize 33 different types of cancer based on their distinct combinations of microbes.
“It seemed like an incredible proof of concept,” said Abraham Gihawi, a postdoctoral researcher at the University of East Anglia in the United Kingdom.
But Gihawi and his colleagues changed their minds when they took a closer look at the microbes that are thought to favor certain types of cancer. They looked totally out of place. The adrenal gland tumors appeared to host a virus that was previously only known to infect shrimp in the Gulf of Mexico. Bacteria known to grow only in seaweed seemed to prefer bladder cancer.
“This is a clear sign that something is wrong,” Gihawi and his colleagues wrote in a letter published Aug. 9 in the journal Microbial Genomics. They dismissed bacteria from seaweed and other displaced species as “absurd.”
In a subsequent study with Salzberg, the researchers reanalyzed the data themselves. “We showed that the magazine is wrong,” Salzberg said. The second analysis was accepted by the journal mBio, he said.
Salzberg and his colleagues have pointed to several possible reasons for the seemingly inexplicable results. To identify microbial DNA from tumors, for example, it is first necessary to remove as many human sequences as possible. Critics say the San Diego team left some sequels behind.
Critics also argue that errors can arise when scientists compare tumor sequences with microbial DNA to look for matches, because some of that data is contaminated with human DNA. This is how similar the DNA of a human cancer cell can look to the DNA of a seaweed microbe.
The San Diego team, led by Rob Knight, responded to these criticisms at length. Knight said he and his colleagues used the best resources they could for their 2020 paper and improved their methods for a paper published last year in the journal Cell with Straussman’s group.
In this study, they used new techniques to remove more human DNA from their analysis. To predict different types of cancer, they only considered bacteria with DNA that passed very rigorous inspection. “You even get tumor type-specific signatures,” Knight said.
In 2019, Knight co-founded a company called Micronoma to develop cancer tests based on his microbial discoveries. (Straussman sits on its scientific advisory board.) So far, the company has raised $17.5 million from private investors.
In January, Micronoma received “Breakthrough Device” designation from the Food and Drug Administration for a lung cancer test, which will accelerate its development into a clinical trial. Sandrine Miller-Montgomery, CEO of Micronoma, said the trial would begin in 2024.
“These criticisms have not led to any changes to our company’s plans,” Miller-Montgomery said.
Dr. Sven Borchmann, a physician-scientist at the University of Cologne in Germany, questioned whether the San Diego team was trying to turn their findings into a medical test that was too quick, rather than doing more experiments to find out what the results really meant. “I think they focused too quickly on application rather than understanding,” he said.
Still, Borchmann suspected that Knight’s team had indeed found a number of species that would withstand close scrutiny, despite the recent outcry. “It doesn’t ruin the whole thesis,” he said.