New therapies are sorely needed for patients infected with SARS-CoV-2, the virus that causes COVID-19, particularly the 15-20% whose acute infections progress to a cytokine storm, which are often deadly. Similarly, effective therapies are very much needed in the treatment of metastatic cancers, like that of the prostate.

Now, researchers at Fred Hutchinson Cancer Research Center have used machine learning, deep neural networks and other artificial intelligence tools to screen, identify and validate a handful of compounds, some Food and Drug Administration-approved, that may provide benefit to both these types of patients.

“Our goal was to determine if FDA-approved drugs available for clinical use might forestall drug development and deliver timelier solutions,” said Hutch systems biologist Dr. Taran Gujral, co-senior author of a proof-of-concept paper published in Molecular Systems Biology on COVID-19 cytokine storms.

Working with a team of Hutch infectious disease experts, Gujral and collaborators were “able to identify and validate FDA-approved, or clinical-grade, compounds, inhibiting inflammatory cytokine production in the context of SARS-CoV-2.”

One compound in particular, the multi-enzyme inhibitor ponatinib, which has been FDA-approved for use in certain leukemia patients, was identified “as a potent inhibitor of cytokine production in response to … SARS-CoV-2 and its emerging variants,” Gujral said.

Working with another team, headed by the Hutch’s Dr. Pete Nelson, Gujral’s subsequent research on cancer was just released in the Proceedings of the National Academy of Sciences. These findings showed that his AI-based drug discovery concept worked equally well in a type of treatment-resistant prostate cancer that has spread, or metastasized.

At least in mice and cells. Next step, they hope, will be testing out the concept in a clinical trial — with people — to prove it’s possible to calm COVID-19’s cytokine storms and setting the stage to perhaps save advanced prostate cancer patients’ lives with repurposed drugs.

Gujral co-authored the earlier paper with Drs. Julie McElrath and Eric Holland, respective directors of the Hutch’s Vaccine and Infectious Disease and Human Biology divisions, along with Hutch scientists Drs. Marina Chan, Siddharth Vijay and John McNevin.

And it is packed with findings.

First, the team discovered a previously unknown function of one end of the SARS-CoV-2 spike protein that promotes cytokine release in immune cells. This discovery was backed up by an international team of scientists who published data on this relationship in the journal Immunity while Gujral’s paper was being peer-reviewed.

“Their study showed that the SARS-CoV-2 virus induces cytokine production in human myeloid cells,” he said. (These immune cells snap into action when the body is injured or invaded by a pathogen.)

“We aren’t allowed to use the actual virus, we used the spike protein, just part of it,” he said. “But we showed that the spike protein does this and showed the cytokine release.”

Cytokine storms can occur with COVID-19 and with sepsis and have also been seen in cancer patients after certain cell-based immunotherapies. They’re caused by the rapid release of inflammatory molecules by the body in response to what it perceives as a threat.

Once the source of COVID-19's storms had been identified, the researchers turned to the sophisticated drug screening platform Gujral and his team developed over the last few years. The platform’s machine learning-based modeling methods allowed them to identify the pathways that the spike protein used to send out the storm signals as well as the drugs that could target those pathways.

Designed with rare diseases in mind — the Gujral Lab’s customary research focus — the scientists believe the platform will become an effective way to speed the identification and use of drugs that can help mediate and even cure overlooked, or “orphan,” diseases.

“My motivation has been to focus on areas that are underrepresented in cancers and other diseases,” Gujral said. “The idea is to find existing drugs approved for other indications that could be applied. But this was a perfect test scenario for COVID-19.”

The team decided to focus on kinases, a type of enzyme that helps orchestrate molecular pathways within cells that regulate everything from cytokine release to cell growth. Quickly narrowing down which kinases are involved in a specific cellular process can be tricky — there are hundreds of kinases, and genetic methods can take time.

That’s why Gujral developed the method he dubbed polypharmacology that combines math with kinase inhibitors.

Kinase inhibitors put the brakes on kinase-controlled molecular pathways, enabling researchers to use them to test which pathways choreograph a particular cellular process. But most kinase inhibitors have more than one target, making it difficult to interpret the results from individual inhibitors.

Gujral has turned this to his advantage: He uses large screens of inhibitors that each hit a different set of kinases, and untangles the results using complex computational methods. With this approach, he was able to narrow down the kinases that contribute most to spike-protein mediated cytokine release. He also discovered some unexpected kinases involved in the cytokine storm pathway that may hold potential for further areas of research.

“Our polypharmacology approach is useful because, in an unbiased way, we can identify all possible downstream kinases important for cytokine release,” he said. 

Many FDA-approved drugs also target kinases. Gujral and his team turned to these tested drugs to screen for potential COVID-19 treatments.

Using what they knew about the cytokine-regulating kinases, Vijay in the Gujral Lab developed algorithms to model how cells might respond to 428 individual kinase inhibitors. The team also predicted responses to 91,000 two-drug combinations that could influence the spike protein-mediated cytokine release.

Their computer models nominated several FDA-approved inhibitors as most likely to block the cytokine-agitating kinases and calm the storm.

“We found five or six FDA-approved compounds, all cancer drugs, out of dozens that we looked at,” he said. “We wanted to find drugs that are immediately useful in the clinic.”

The researchers tested their model’s most promising candidates with cells in petri dishes and in mice.

Ponatinib, a third-generation tyrosine kinase inhibitor approved for certain chronic myeloid leukemia patients, emerged as the best candidate as it was able to block all seven cytokines.

“More-specific inhibitors weren’t that useful, but the nonspecific inhibitors were more effective in inhibiting the cytokine release,” Gujral said. This is likely because the cytokine storm is caused by several molecules working together, he noted.

And several of the drugs identified in the team’s screen are generic.

“What's exciting is that is there are inexpensive, FDA-approved drugs that might be able to reduce the deadly impact of this virus in people who are already infected,” Holland said.

The team also tested and validated the effectiveness of some of the drugs with Delta and other emerging COVID variants.

If the drugs the researchers found are effective, they wouldn’t block infection, he said, but could address the dire need for drugs to help keep infected people alive and out of intensive care units.

Ponatinib is currently under patent protection so there is no generic available. But Gujral said it could be cheap to manufacture and patients can take it orally, as opposed to via infusion, saving time and money.

“All of the drugs we identified are daily pills,” Gujral said. “Our collaborator Rachel [Bender Ignacio, who leads the Hutch’s COVID-19 Clinical Research Center, or CCRC] was very excited about that. It could potentially decrease hospitalization.”

The CCRC opened in October of 2020 to address the need for COVID-19 therapies.

An epidemiologist, internist, infectious disease expert and the CCRC’s medical director, Bender Ignacio worked with the collaborators to create a mockup of a clinical trial to test this potential therapy in COVID-19 patients.

“It’s really fantastic that they’ve found something that seems to be more potent in this specific interaction than the drugs so far tested clinically,” Bender Ignacio said.

CCRC is currently running a handful of COVID-19 therapy clinical trials, but Bender Ignacio said more trials are absolutely needed, not just for therapies to alleviate the hyperinflammation of cytokine storms, but for those who contract long COVID-19.

While COVID-19 therapy trials are challenging for many reasons, therapies are coming.

“The pipeline is working,” she said, adding that Gujral’s drug-discovery platform could feed into that by identifying promising approved drugs for long-haul COVID-19 and other conditions that lack effective treatment.

Using FDA-approved drugs in a trial would also mean much faster results. Gujral said they hope to do a combined safety and efficacy trial of ponatinib in 100-150 patients at multiple sites. While the drug has side effects if used for a long period of time (cancer patients, on it for 25 to 30 months, have experienced blood clots), patients in a COVID-19 trial would take it for only five to seven days. And the trial could conceivably start as soon as patients are available.

There’s only one problem. They can’t access the drug, at least not yet.

Gujral said his team has had conversations with the manufacturer but they decided not to go forward with a clinical trial, provide any funding nor donate the drug for a trial funded entirely by donations.

“That was a setback,” he said. “But we are going to try knocking on their door again and also pursue other companies.”

Ponatinib wasn’t the only promising candidate from Gujral’s screen. But finding the interest and financing to fund new trials of generic drugs is a perennial problem, Holland noted.

Even so, trials must be conducted to know for sure if ponatinib or any other likely-looking compound has real lifesaving potential, he said.

“This could be a great place for philanthropy, or people who really want to make a difference to fund trials, to figure out if these drugs really could be used to save people’s lives,” he said.

Another promising aspect of the work is that this therapy might work in other storm scenarios.

“The [cytokine storm] phenomenon we studied is not specific to COVID,” Gujral said. “It’s how immune cells respond to a threat. It could be a viral assault, or bacterial, or cancer. What we’re studying is a response of the immune cell. It doesn’t matter where the threat comes from.”

Chan, lead author on the paper, said the versatility of the platform is also significant.

“We now have the ability to define a very specific assay that can measure the phenotype you want,” she said. “The phenotype is the problem you’re trying to solve for, in this case, it’s the cytokine release from immune cells. Then you couple that with this machine learning platform to screen for FDA-approved drugs that will block the phenotype you’re trying to measure. I think that’s very powerful.”

Gujral, Chan and team put the drug’s cytokine storm-dampening effect to the test with a mouse that developed sepsis, another type of storm, following a lung infection.

“It cleared the infection,” Gujral said.

But the researchers cautioned that “there’s a fine balance” between suppressing a cytokine storm and suppressing necessary immune function.

“You want the immune system to fight the virus or whatever the threat is,” Gujral said. “You want to show that you’re not suppressing it indefinitely, that you’re not suppressing all the immune cells.”

To that end, in collaboration with Hutch virology experts McNevin and McElrath, who holds the Joel D. Meyers Endowed Chair, the team tested how ponatinib worked against cytokine release by two different types of immune cells: monocytes and T cells.

“We found the drugs affected the ability of the monocytes to secrete cytokines,” Gujral said. “But they had little effect on T cells.”

Bender Ignacio said it was important to inhibit the hyperinflammatory response triggered by a subset of immune cells while theoretically allowing preservation of the parts of the immune system needed to fight the virus. The steroids currently being used to suppress the immune response are a “blunt tool,” she said.

“Looking at these precision tools is very important and may allow them to be used earlier in some cases,” she said. “There’s also potential for using them in people who never develop severe disease, to either prevent or treat long COVID.”

The Hutch team is thrilled by the success; Gujral called it a “a nice validation of the drug-discovery approach.”

Chan added that coming up with fast answers mid-pandemic felt incredibly rewarding.

“Whether this can have a translational impact depends on many stakeholders,” she said. “But this is an unprecedented time. As a scientist, I’m very excited that we can apply our platform to this crisis and find a solution. It feels very rewarding and also shows how versatile and flexible research can be.”

Best of all, it’s just the start.

The team’s latest proof-of-concept paper in the Proceedings of the National Academy of Sciences shows the approach’s promise for late-stage prostate cancer.

“One of the challenges of prostate cancer is when it metastasizes to different organs,” Gujral said. “It’s very painful and very difficult to treat.”

He, Nelson and their team used the same polypharmacology-based approach to identify the kinase inhibitors that curtail prostate cancer cell growth. Their goal was to find inhibitors that block several kinases involved in prostate cancer growth and progression but aren’t likely to cause widespread side effects.

Two compounds, PP121 and SC-1, met these criteria and suppressed prostate tumor growth in lab dishes and in preclinical models in which human tumors are allowed to grow in mice.

The team hoped that the new compounds could also work against tumor spread, including tumors that spread to the bone.

“Bone metastases are very painful for patients, and very hard to treat because of the complexity of the bone microenvironment,” Gujral said.

He and Nelson teamed up with Dr. Eleonora Dondossola at the University of Texas MD Anderson Cancer Center to test their compounds’ potential against bone metastases. Dondossola had developed a new approach to model the molecular and cellular environment that tumor cells encounter in bone by studying both a lab dish-based system of scientist-engineered bone tissue and tumor cells implanted into bone.

“The most innovative aspect of our work lies in the ability to predict the activity of a therapeutic agent to influence a certain parameter (in our case tumor growth) through a computational system,” Dondossola said, noting that the computational approach saved the team time and resources when identifying and testing PP121 and SC-1.

The initial tests of the compounds against tumor cells growing in Dondossola’s engineered bone tissue looked promising. But when she and Gujral first tested their compounds against prostate tumor cells implanted into and allowed to grow in bone, the results were disappointing: Tumors grew at a normal rate.

But when they tested their kinase inhibitors combined with docetaxel, a standard chemotherapy for metastatic prostate cancer, they found that the drug combo slowed bone metastasis growth by six times compared to docetaxel alone. The combination also prolonged mouse survival, from less than 50 days to 120 days.

Dondossola and Gujral are now working to understand how PP121 and SC-1 may help make bone metastases more sensitive to docetaxel.

“If we can understand that part of the biology better, we could probably come up with even better [drug] combinations,” Gujral said.

PP121 and SC-1 have yet to be approved by the FDA for use as therapeutics, but Gujral hopes that will change soon. The Hutch and MD Anderson recently filed a patent application for the use of these compounds with docetaxel as a treatment against metastatic prostate cancer. More preclinical testing is in the works.

“This really opens up a new area for biologists to look at the role of kinases in various diseases,” he said.

This work was funded by the Fred Hutch COVID-19 Pilot Fund, the National Cancer Institute, the Pacific Northwest Prostate Cancer and MD Anderson Cancer Center SPOREs, the Department of Defense and an AACR-Bayer Innovation and Discovery Grant.

Note: Scientists at Fred Hutch played a role in developing these discoveries, and Fred Hutch and certain of its scientists may benefit financially from this work in the future.