NEW YORK — Carbon monoxide is a poison. Could it also be a medicine?

Dr. Augustine Choi has spent two decades trying to figure that out. And he’s inching toward a surprising answer: yes.

Most of the research has been conducted in animals; a few early-stage clinical trials are underway in patients. But the data look promising enough that several companies are jumping into the field, developing ways to deliver small amounts of carbon monoxide to damaged tissues through pills, lotions, beverages, and inhalable gas.

It’s not an entirely new idea: Scientific papers dated as early as 1932 hinted that small doses of carbon monoxide — a tasteless, odorless, and colorless gas — could have beneficial effects. But no one had actually done experiments. Not until Choi, who chairs the department of medicine at Weill Cornell Medical College, started thinking about why our lungs work so well despite daily bombardment from pollutants.

More specifically, Choi started researching a protein called heme oxygenase, or HO-1. It’s an enzyme that converts heme, a central element of hemoglobin, into a slew of chemicals: iron, a pigment called biliverdin that’s responsible for the greenish color of bruises — and carbon monoxide. Every day, our bodies naturally generate about 10 milliliters of carbon monoxide. It’s critically important: Mice that have been genetically engineered without HO-1 die young from multi-organ failure.

What’s more, HO-1 production increases every time an organ responds to stressors. It seems to be part of some kind of protective feedback loop.

Choi was eager to understand how HO-1 worked — and, in particular, how it seemed to protect the lungs from the abuses and pollution we throw at them daily. Was the release of iron protective in some way? Could the pigment be therapeutic? Or, as he came to believe, could HO-1’s effects be explained by carbon monoxide?

It was a somewhat crazy idea.

The noxious gas exerts its toxic effects by binding tightly to hemoglobin, the molecule that delivers oxygen throughout the body. Carbon monoxide also inhibits respiration in the powerhouses of the cells, the mitochondria. It becomes toxic when people are exposed to it at a concentration above 10,000 parts per million. Choi had his eye on much lower doses, around 250 parts per million.

Either Choi or his graduate student at the time, Leo Otterbein, had the idea of giving animals carbon monoxide to test its therapeutic value. (Exactly who came up with this idea remains a contentious point between the two scientists today.)

Otterbein recruited the help of an engineer and designed an apparatus for the experiment. In one chamber, six rats were exposed to lethal doses of oxygen. He placed a second group of rats in an identical chamber, exposing them to the same concentration of oxygen, but with a small amount of carbon monoxide mixed in.

Three days later: “The animals who received carbon monoxide were still alive, and those that did not were all dead,’’ Otterbein said.

It took another three years before these first results were published in 1999.

“We must have tested more than 200 animals just to confirm our results, to make sure that [Otterbein’s] career and my career didn’t go down the wrong path if we were wrong,’’ Choi told STAT. “Because of the stigma of CO, there’s a higher risk of more pushback from the scientific community.’’

And indeed, Choi said, the paper was met with widespread disbelief, rejected by journal after journal before being accepted by the American Journal of Physiology.

“The reviewers didn’t say that our quality of data was bad, or that the experiments were poorly designed,’’ he said. “They just said they don’t believe that CO can be protective.’’

Even Dr. Claude Piantadosi, Choi’s senior attending physician from his residency at Duke University, had serious doubts. Piantadosi did his own study and found that carbon monoxide had no protective effect on rats with lung injury. He wrote a paper rebutting Choi’s 1999 findings.

But the question continued to intrigue Piantadosi, and he took a closer look at how carbon monoxide interacted with cells. He found that while exposure to low doses of the gas stressed the mitochondria, the overall cell did seem to respond positively.

Now, Piantadosi and Choi are collaborating on a Phase I clinical trial to investigate the safety of low-dose carbon monoxide on approximately 48 patients with acute respiratory distress syndrome, a life-threatening lung condition.

Choi has received more than $21 million in federal funding from the National Institutes of Health to pursue clinical trials.

One such study, which wrapped up in May 2015, assessed the safety of low-dose carbon monoxide in patients with the terminal lung condition idiopathic pulmonary fibrosis. The results suggested that low doses of the gas are indeed safe for these patients. This study will progress later this year to a Phase II trial, meant to determine whether exposure to carbon monoxide actually helps them.

“I think in about three to five years we’ll know what the landscape will be like for carbon monoxide in the clinic,’’ Choi said.

Outside researchers agree more data are needed.

“Carbon monoxide could be useful in small doses, but we have to know more about exactly how to use it,’’ said Dr. Phyllis Dennery, who chairs the pediatrics department at Brown University’s Warren Alpert School of Medicine and has studied heme oxygenase for more than 27 years.

But entrepreneurs are moving ahead. Choi and cofounder Jeff Wager of Apeiron Partners LLC have launched Proterris, a private company that aims to develop carbon monoxide therapeutically.

Prolong Pharmaceuticals, a biotechnology company in New Jersey, is investigating whether molecules of hemoglobin infused with carbon monoxide can be used to prevent organ rejection in patients undergoing kidney transplantation.

And Hillhurst Biopharmaceuticals, based in Southern California, is developing a carbon monoxide drink for patients with sickle-cell anemia, as well as those who have undergone kidney transplants or suffered traumatic brain injuries.

One potential use for CO therapy is in organ transplants. In the late 1990s and early 2000s, researchers in the lab of the late Dr. Fritz Bach exposed mice to carbon monoxide for two days before removing their hearts and transplanting them into rats. For two weeks following the surgery, the rats were then exposed to carbon monoxide.

Otterbein was in regular touch with Miguel Soares, a scientist in Bach’s group.

“Every day, I’d ask [Soares], how are the animals? And every day, his e-mails would just say, ‘THE HEARTS ARE STILL BEATING!!!’’’ Otterbein recalled.

Otterbein knew that typically, hearts transplanted in this way would give out within a week or so. But the e-mails from Soares kept coming — on days 8, 12, and 60.

“That to me, is still some of the most powerful data out there,’’ Otterbein said, “because someone else, independently of me, had proven it would work.’’