Drivers of Persistent Inflammatory Memory Identified

Chronic inflammatory skin conditions like psoriasis remain mysterious because they can stay inactive for months or years before suddenly reactivating, often at identical sites. Elaine Fuchs' team at Rockefeller University first probed this in 2017, revealing that skin stem cells store "memories" of prior inflammation. Healthy memories accelerate injury repair, but dysfunctional ones heighten tissue sensitivity to triggers, fostering chronic inflammation. 

Fuchs' latest Science paper identifies key epigenetic drivers behind the long-term persistence of these memories. Through wet lab work and deep machine learning, they isolated genetic sequences that extend select critical memories over years, underpinning chronic diseases. “The findings could provide valuable inroads for developing new therapeutic strategies for chronic inflammation and possibly other human conditions such as cancer, pain, and weight regain, where the ability of our body’s cells and tissues to keep a record of past experiences may have deleterious consequences,” says Fuchs.

The 2017 findings established tissues—not just immune cells—as inflammatory memory holders. In 2021, they detailed how STAT3 and FOS-JUN open genome areas during inflammation, with host factors maintaining accessibility post-resolution. This primed state reactivates via FOS-JUN alone under new stresses, rationalizing diverse flare triggers. Around 1,000 chromatin "memory domains" opened by psoriasis-like flares in mice lingered a month later.

“In our initial understanding of how inflammatory memory worked, we couldn’t distinguish which, if any, of these regions might last long enough to have consequence for chronic human disease,” notes co-first author Christopher Cowley. Sairaj Sajjath, another co-first author, adds, “Considering that people can experience disease flares many months or even years apart, we wanted to find out how long we could recognize signs of prior acute inflammation in mice and what determines longevity.”

Inducing psoriasis in young mice, the team found 10–15% of one-month memories endured for the animal's lifespan (~2 years), unlike short-term ones fading by six months. Luis Soto-Ugaldi, third co-first author, crafted PersistNet—a deep learning tool—to compare DNA traits. Long-term domains featured elevated CpG dinucleotide density (cytosine-guanine pairs vital for regulation). “When we compared the DNA sequences of short and long-term memory domains, they looked very similar... We realized we needed to develop a new metric that specifically captures memory persistence across time,” Soto-Ugaldi explains. Higher CpG levels precisely predicted duration across all domains, per Sajjath: “Looking across all 1,000 memory domains, we discovered that these nucleotide densities alone... could distinguish how long each memory would linger.” 

Lab validation linked CpG density to epigenetic hallmarks: CpG-specific DNA demethylation, affinity of demethylation-favoring transcription factors, and H2A.Z histone binding. These reinforced open chromatin, heritable through cell divisions. “This really fills out the picture... At the front end, you still need that acute inflammatory experience to open up the chromatin... Then, depending on how much CpG you dial in, you either have a short-term or a long-term memory,” Cowley states. Sajjath highlights how this addresses dilution concerns: “Our study closes the gap between the mechanistic understanding of memory persistence and the physiological manifestations that we see in clinical settings and in the lab.”

Future work will contrast beneficial (healing) versus maladaptive (psoriasis-linked) memories. “We’ve spent more time looking at beneficial memories, so now we want to look more at the maladaptive ones... Identifying the unique characteristics of bad memories may help us to break the cycle of inflammatory disease,” Fuchs says.