Network-based analyses uncover how neuroinflammation-causing microglia in Alzheimer’s disease form

Cleveland Clinic Genome Center researchers have unraveled how immune cells called microglia can transform and drive harmful processes like neuroinflammation in Alzheimer’s disease. The study, published in Alzheimer’s & Dementia: The Journal of The Alzheimer’s Association, also integrates drug databases with real-world patient data to identify FDA-approved drugs that may be repurposed to target disease-associated microglia in Alzheimer’s disease without affecting the healthy type.  

The researchers, led by study corresponding author Feixiong Cheng, PhD, hope their unique approach of integrating genetic, chemical and human health data to identify drug targets and corresponding drugs will inspire other scientists to take similar approaches in their own research.  

Microglia are specialized immune cells that patrol our brains, seeking and responding to tissue damage and external threats like bacteria and viruses. Different types of microglial cells use different methods to keep the brain safe. Some may cause neuroinflammation – inflammation in the brain – to fight invaders or kickstart the repair process in damaged cells. Others may work to “eat” dangerous substances in the brain, and clean up damage and debris. However, during Alzheimer’s disease, new types of microglia can form that promote disease progression. 

“Microglia have been implicated in Alzheimer’s disease for over a century. So far, attempts to stop disease progression with broad spectrum anti-inflammatory drugs and ‘harmful’ microglial blockers have been ineffective,” says Dr. Cheng, director of Cleveland Clinic’s Genome Center. “We need to selectively block harmful microglia subtypes while leaving normal, healthy microglia intact.” 

The challenge, Dr. Cheng says, is that both the factors that cause these different subtypes of harmful microglia and the specific ways some of these subtypes function is unknown. 

To develop a more specific drug that targets harmful microglia, Dr. Cheng and his laboratory asked: 

1. What made harmful microglia different from their normal, helpful counterparts on a molecular level?
2. What drugs could target these differences specifically, to block or even reverse the process that causes harmful microglia to form?
3. If they identified more than one potential drug, which was the most promising? Was there any evidence to suggest that any drugs they identified could be helpful in humans?   

Each of these questions required different types of data to answer. To quickly and efficiently integrate the large amounts of data for computational analysis, Dr. Cheng assembled a team to take an integrative, “network-based” approach.   

The team received collaboration and help interpreting their data from collaborators from IBM, Weill Cornell Medicine, Case Western Reserve University, the Cleveland Clinic Lou Ruvo Center for Brain Health and the University of Nevada Las Vegas. 

Led by first author Jielin Xu, PhD, the team created an algorithm to combine and analyze: 

• Publicly available RNA-sequencing datasets obtained from over 700,000 individual Alzheimer’s-affected single cells, to identify unique signatures of harmful microglia by determining which genes were turned “on” or “off” in different subtypes, termed molecular ‘drivers.’
• Protein-protein interaction data from 18 publicly available datasets, to predict how genes unique to harmful microglia impact cellular processes.
• Chemical and drug databases to determine which FDA-approved drugs, if any, could block disease-specific protein-protein interactions to treat harmful processes caused by the gene activity of disease-associated microglia.
• Real-world patient databases from millions of insured individuals to determine whether any drugs are associated with lower instances of Alzheimer’s disease diagnoses. 

“Our study offers a powerful deep generative model to identify repurposable drugs from many types of Alzheimer’s disease findings, but the overall methods can be broadly applied to other diseases as well,” Dr. Cheng says. 

The team’s network-based analyses identified three unique subtypes of harmful microglia that promoted disease progression. Each of these subtypes had their own genetic signatures that drove unique behaviors to support Alzheimer’s disease. For example, one microglial subtype causes harmful neuroinflammation, while another supports the buildup of proteins in our brains that cause Alzheimer's, like tau.  

Each subtype also had unique genetic signatures that caused them to change from helpful to harmful. Further study into the different harmful microglia subtypes and their genetic signatures has the potential to reveal more drug targets and advance Alzheimer’s disease treatments. 

The analyses also revealed that there were already FDA-approved drugs on the market designed to block many of these harmful transitions. Repurposing an FDA-approved drug to treat Alzheimer’s disease is safer and faster than designing a drug from scratch, Dr. Xu says. 

The team’s algorithms also showed that patients who took one of the potentially repurposable drugs, an NSAID called Ketorolac used to treat mild-to-moderate pain, were diagnosed with Alzheimer’s less than individuals who did not take the drug. The team validated their computational predictions in dish experiments on microglia derived from patients affected by Alzheimer’s disease, where they showed that Ketorolac blocked an immune process called type-I interferon (IFN) signaling. The next step is designing further experimental and clinical validation to evaluate the effects of Ketorolac on Alzheimer’s disease.

Dr. Cheng adds that even though his team’s analyses focused primarily on Alzheimer’s disease, their overall findings have wide-reaching implications in many other neurogenerative diseases and age-related complex diseases. 

“In the past, each of these discoveries would have needed to be made with their own extensive research project,” Dr. Cheng says. “Our advanced computing techniques allow us to make biological, chemical and patient-based discoveries with one study. We believe these types of artificial intelligence (AI)-assistant network-based analyses represent the future of biomedical research.”  

This research was supported by grants from the National Institute on Aging (NIA).