SPRING 2022 | THE SCIENTIST 43LEFT: PICTURE BY LARRY LUXNER; RIGHT: PICTURE BY LEORE GELLER FROM THE STRAUSSMAN LABentist agree that the evidence all points to the idea that tumors
are indeed not sterile, but typically harbor a microbiome. “The
bottom line is, yes,” says Straussman, “there’s actually live bac-
teria in tumors.”Microbial diagnostics
Straussman’s 2020 Science paper came on the heels of some-
what similar study, published in Nature by Knight’s group. Like
Straussman’s team, Knight and his colleagues had surveyed a
range of tumor types and discarded up to nearly 93 percent of
the sequence data from their analyses to guard against contam-
ination. And like Straussman’s findings, Knight’s results had
showcased microbes as common features of cancer, with micro-
bial makeup varying by cancer type. But the UCSD team had
taken an entirely different approach—clean and analyze existing
sequence data rather than sequence samples anew—and they
had come to the idea for very personal reasons.
Sepich-Poore was a freshman in college when his grand-
mother was diagnosed with pancreatic cancer the day after
Christmas 2012. By the end of January, she had died from the
disease. When the doctors were unable to explain why the cancer
hadn’t been detected sooner, why it had progressed so quickly,
and why it hadn’t responded to therapy, Sepich-Poore decided
to go looking for answers himself. He taught himself bioinfor-
matics and machine learning to interrogate how cancers evolve
over time. He later began reading the literature on blood-based
diagnostics, searching for new strategies in cancer screening and
testing. And in 2016, he enrolled in medical school at UCSD,
eager to learn about new advances in pancreatic cancer diag-
nosis and treatment.
Unfortunately, he found that despite the massive genomics
revolution that had occurred in recent years, not much prog-
ress had been made for patients with pancreatic cancer. The
five-year survival rate for the disease has lingered at around10 or 11 percent for decades. “It made me wonder [if ] perhaps
there are markers or entities that are not yet being accounted
for by those focused on host DNA or host RNA or host pro-
teins,” but which could help catch cancer earlier, notes Sepich-
Poore. Then he came across Straussman’s 2017 paper showing
that not only did bacteria exist in the tumor microenviron-
ment, they could influence the cancer’s response to therapy.
He recalls being stunned.
Sepich-Poore suspected that no tumors were actually sterile,
and that the microbes present could provide critical information.
He had spent some time in Knight’s lab the summer before start-
ing med school, and he landed back there for his PhD, spend-
ing his first six months interrogating whole genome sequences
and transcriptomic data that was gathered as part of The Cancer
Genome Atlas. His analysis covered the tissue and blood samples
of more than 10,000 patients with more than 30 cancer types.^5
At first blush, the data were disappointing, Knight
explains, as most of the variation the researchers found in
microbial DNA could be attributed to where the study had
been done. But Sepich-Poore applied statistical approaches
he’d used to account for batch effects when analyzing human
cancer data and “was able to get rid of all that noise and see
a signal where the bacteria in a particular sample can give
a readout of what kind of tumor you have and what stage,”
Knight explains. Then, to scrub the data of potentially con-
taminating bacteria, Sepich-Poore used a series of existing
statistical techniques that take into account, for example, the
microbial sequences’ relative abundance across samples of
different concentrations.A CLOSER LOOK: Ravid Straussman of Weizmann Institute of Science
has found evidence of bacteria inside tumor cells, such as these human
pancreatic cells, which harbor GFP-positive bacteria.ADAPTED FROM
SCIENCE
, 371:EABC4552, 2021; © NICOLLE FULLER, NATASHA MUTCH, SAYOSTUDIO