None of this progress has come easily. The intuitive notion of harnessing the body’s own immune system to fight the enemy within is more than a century old—and seemingly, each time researchers have seemed close to turning that theory into reality, a new stumbling block appears.

It’s as if the immune system were a car that no longer functioned when a person got cancer. At first, scientists theorized that this vehicle simply ran out of gas. And so early efforts centered on making sure the gas tank was filled—priming the immune system with cytokines (such as the interleukins) or with immune-signaling proteins like interferon to rev up the tumor-fighting response. And though, in many cases, the RPM dial on the dashboard did spin higher, the car didn’t move much at all.

Then came the theory that, while the car needed gas, it also needed to be put in the right gear. So research teams began souping up various types of immune cells and even targeting them to specific protein markers on tumors. Sometimes they even made their own antigen-seeking antibodies to deliver the punch. Still, the cancers mostly grew unchecked.

Then, a Texan named Jim Allison came up with what seemed at the time to be a radical notion: that the car had its emergency brakes locked. And you had to unlock them if you were going to get that car to rush toward the target. From this theory came a promising class of drugs called checkpoint inhibitors, which effectively release those brakes. And while that has led to some seemingly miraculous cures of metastatic disease, many patients still don’t respond.

Others, meanwhile, have theorized that such treatment failures are due to the lack of a good GPS device: the car can go, they say, but it’s not heading to the right address. So researchers, with limited success, have tried to build such a GPS directly into a new type of biological assassin called a “CAR-T cell.”

Now, a new theory has emerged—and it’s a doozy: It’s not the gas, or the gear, or the brakes, or the GPS that’s holding us back; it’s the passengers! And those passengers aren’t human cells at all, but rather humble bacteria.

In recent weeks and months, several impressive studies have made the case that the intestinal flora of cancer patients play a significant role in whether those patients respond to immunotherapy or not.

In the most recent of these reports (in the latest issue of Science), Dr. Jennifer Wargo, a surgeon and research scientist at M.D. Anderson, along with several dozen colleagues at other institutions, reveal that the composition of a patient’s gut microbiota can significantly influence whether he or she responds to an immune checkpoint inhibitor—the type of cancer immunotherapy that releases the emergency brakes in the car analogy above.

Those with an abundant supply of Faecalibacterium prausnitzii, a common microbe in the Ruminococcaceae clan, fared much better than those with a limited supply, the study found. Alternatively, those with an enriched population of Bacteroidales tended to do poorly.

But what seemed to matter most wasn’t the level of a specific gut microbe, says Wargo in an interview this afternoon, but rather the overall diversity of the person’s gut microbiome. (The team also swabbed patients’ cheeks for bacteria, but the composition here seemed to have little effect on treatment outcomes.) “I don’t think it’s one bacteria per se that’s driving this entire response,” she says. “I think it’s probably a community of bacteria. And what we found is that, in patients who responded to the treatment, they actually had a much higher diversity of bacteria in their gut microbiomes compared to non-responders.”

The favorable composition of flora were associated with a “higher activation status” of their killer T cells within the tumor compared to those who had an abundance of Bacteroidales. The latter “really just had quiescent T-cells in their tumors,” says Wargo.

When the research team took fecal samples from responding patients and transplanted them into germ-free mice—“essentially reconstituting the mice’s gut microbiomes with a responding patient’s microbiome,” she says—they discovered that the mice had better immunity. “And then, when we put a melanoma tumor into those mice, they either were very slow to grow or were completely rejected,” says Wargo. “They also had phenomenal response to treatment with the anti–PD-1-based therapy [the checkpoint inhibitor]. Whereas when we transplanted the mice with a fecal sample from a non-responding patient, they had poor immune function after the transplant,” she says. (When the team implanted melanoma tumors into those mice, they grew rapidly and failed to respond to immunotherapy.)

As I’ve said, there have been a number of research papers of late, led in large part by the work of French scientist Laurence Zitvogel, that are building a strong case for a central role for the microbiome in cancer treatment response. Others have found that bacteria within the tumors themselves can influence the effectiveness of traditional chemotherapy as well.

“It also brings in the question of diet,” says Wargo, who is now working on new clinical studies on the microbiome with the Parker Institute for Cancer Immunotherapy. “Could it be that patients who have a fiber-rich diet with more whole grains—that is, with a more microbiome-friendly diet—might do better on cancer treatment?” she asks. “And could such a diet help facilitate and enhance the immune system such that you might be able to ultimately prevent cancer?”

“Perhaps,” she says, answering her own question—though she is quick to add that cancer patients shouldn’t alter their diets without talking to their doctors first. (They could do themselves harm.)

In the curious, serpentine way that science works, it would make perfect sense for the mystery of one world of cells (those of the human immune system) to be solved by those of another (the trillions of bacterial boarders that live within our bodies).

But hey, that’s just what my gut tells me.

This essay appears in today’s edition of the Fortune Brainstorm Health Daily. Get it delivered straight to your inbox.