Neuroinflammation and Aβ deposition are two hallmarks of AD, and now, a study published September 3 in Nature is forging an alluring mechanistic link between them. Researchers led by Yueming Li at Memorial Sloan Kettering Cancer Center in New York reported that interferon-induced transmembrane protein 3 (IFITM3) binds to γ-secretase, revving up the enzyme’s production of Aβ peptides. Knocking down IFITM3 squelched Aβ production in human cells and in a mouse model of amyloidosis. In the human brain, levels of the protein rose with age and in people with AD, and correlated with the amount of inflammatory cytokines and viral proteins present in the brain, as well. In all, the findings link age-related neuroinflammation with increased production of Aβ, and highlight IFITM3 as a potential therapeutic target.
- IFITM3 binds γ-secretase, ramping up Aβ production.
- Knocking out IFITM3 reduces Aβ deposition in 5xFAD mice.
- In human brain, IFITM3 increases with age and in AD.
“The paper is very convincing and springs a new surprise in the study of the γ-secretases,” wrote Bart De Strooper of KU Leuven in Belgium. “It actually turns the classical view that inflammation is a consequence of amyloid plaque accumulation upside-down, providing mechanistic support for the hypothesis that inflammation causes increased Aβ generation.”
Beyond its four essential subunits of presenilin, nicastrin, Aph1, and Pen-2, the γ-secretase complex hosts some hangers-on that modulate its activity in different ways. Researchers are keen to find ways to selectively block the enzyme’s processing of amyloid-β precursor protein (APP), while sparing the cleavage of other important substrates such as Notch. To that end, scientists have developed small molecule γ-secretase modulators (GSMs) that, besides having clinical potential, are able to identify proteins that associate with active γ-secretase complexes.
By attaching a photo-affinity probe to GSMs and then flashing them with UV light to trigger cross-linking with adjacent proteins, researchers hope to capture proteins likely to play a physiological role in the enzyme’s function. While at Merck, Li used this approach, albeit with γ-secretase inhibitors, to help peg PS1 as the catalytic subunit of γ-secretase (Li et al., 2000; Xu et al., 2002).
For the current study, co-first authors Ji-Yeun Hur and Georgia Frost used a photo-affinity labeled version of E2012—a GSM created by Eisai but no longer in clinical development—to fish out endogenous protein modulators of γ-secretase (Apr 2011 conference news; Aug 2013 news). Even when conjugated with the photoaffinity probe BPyne, E2012 binds to PS1 N terminal fragments within active γ-secretase complexes (Pozdnyakov et al., 2013). The researchers mixed neuronal cell membranes with E2012-BPyne, induced cross-linking with UV light, purified the BPyne-labeled complexes, and identified two proteins via western blot: the N-terminal fragment of PS1, as predicted, and IFITM3.
An interferon-inducible protein involved in antiviral responses, IFITM3 had previously been spotted among PS1’s interactors in a proteomics screen led by De Strooper (Oct 2009 news on Wakabayashi et al., 2009). Li’s group used several complementary approaches, including co-immunoprecipitation with various PS1 antibodies, to confirm that indeed, IFITM3 bound to PS1 within active γ-secretase complexes in mouse primary neurons. IFITM3 also interacted with PS2.
This interferon-induced protein is expressed in many cell types, including mouse hippocampal neurons, astrocytes, and microglia. Human induced pluripotent stem cell-derived neurons and primary human astrocytes make the protein (see image above). In cells lacking both PS1 and PS2, IFITM3 protein levels plunged, suggesting that presenilins stabilize it.
Does IFITM3 influence γ-secretase activity? To investigate, the researchers knocked down IFITM3 in human embryonic kidney (HEK) cells. Without it, production of Aβ40 and Aβ42 peptides dropped by 17 and 24 percent, respectively. In an astrocytoma cell line lacking IFITM3, the amyloid peptides dropped by 36 and 27 percent, respectively. The opposite was true for γ-secretase cleavage products of Notch, which ramped up without IFITM3. The findings suggest that IFITM3 enhances processing of APP, but suppresses processing of Notch.
In the brains of wild-type mice, IFITM3 levels rose with age, more than doubling between the ages of 4 and 28 months. What’s more, IFITM3 associated with γ-secretase only in the older mice. The researchers generated IFITM3 knockout mice, and found that γ-secretase production of Aβ40 and Aβ42 dropped by 15 and 24 percent, respectively, in brain tissue extracted from 18-month-old knockouts relative to wild-type.
IFITM3 levels also ramped up with age in the 5xFAD mouse model of amyloidosis. At 12 months of age, IFITM3 levels in 5xFAD mice were nearly double those in wild-type. These transgenic mice overexpress human APP and PS1, which likely also influences IFITM3 expression levels. While IFITM3 primarily localized to blood vessels and meninges in wild-type mice, in plaque-ridden 5xFAD mice the protein also comingled with astrocytes and microglia. The researchers crossed 5xFAD mice to IFITM3 knockouts, and found that loss of IFITM3 not only reduced the production of Aβ peptides by γ-secretase, but also dramatically lessened Aβ plaque burden. Compared with 5xFAD controls, 5xFAD lacking IFITM3 had 54 and 81 percent fewer plaques in the hippocampus and cortex, respectively, at 4 months (see image above).
What about in people? The researchers assessed IFITM3 expression in postmortem brain samples from multiple cohorts. In the Genotype-Tissue Expression (GTEX) cohort, they measured more IFITM3 transcripts in the hippocampus and cortex with age: IFITM3 expression approximately doubled between the ages of 20 and 70. In a Mayo Clinic cohort, IFITM3 transcripts were more abundant in the temporal cortices of 80 people with AD than in 76 controls (see figure below). In a separate cohort of frontal cortex samples from the UCSD AD Research Center (ADRC), 18 people with AD had more IFITM3 mRNA and protein than did 10 controls.
Although IFITM3 protein was more abundant on average, there was high variability in IFITM3 among people with LOAD in the UCSD cohort. To see if IFITM3 levels correlated with γ-secretase activity, Hur and colleagues divided the samples into high and low IFITM3 bins, then measured APP processing in cell membranes extracted from each. Lo and behold, the eight LOAD samples with high IFITM3 protein cranked out 127 and 130 percent more Aβ40 and Aβ42, respectively, than did controls, or LOAD samples with low IFITM3. The findings suggest that IFITM3 correlates with higher Aβ production. In support of this idea, IFITM3 expression levels correlated with Aβ load in multiple brain regions among postmortem samples from the Mount Sinai Brain Bank Dataset.
What dictates IFITM3 expression in the brain? Treating cell cultures with either type I or type II interferons boosted IFITM3 expression and ramped up production of Aβ. Dovetailing with this, IFITM3 expression correlated with pro-inflammatory cytokine levels in brain samples from Harvard Brain Tissue Resource Center, and with herpesvirus proteins and hepatitis C infection in the Mount Sinai brain samples. Together, the findings suggest that inflammatory stimuli—such as viral infections or age-related cellular stress—might enhance IFITM3 expression, leading to stepped-up production of Aβ.
The findings jibe with the notion that Aβ peptides are rallied to contain microbes (May 2016 news; Jun 2018 news). IFITM3 has been implicated in defense against numerous viruses, including influenza, Zika, and most recently SARS-CoV-2.
“This work directly links Aβ production with innate immunity and neuroinflammation and provides a novel mechanism as to how Aβ secretion is stimulated in response to an invading pathogen,” wrote Huaxi Xu, co-editor in chief at Molecular Neurodegeneration. “This not only provides insight into Aβ’s function as an antimicrobial peptide, but also establishes IFITM3 as a potential therapeutic target to reduce Aβ production.”
Feel the Burn. Glial cells secrete pro-inflammatory cytokines in response to inflammatory stimuli such as viral infections (left). This enhances IFITM3 expression in neurons and other cells, upping Aβ (middle). Aβ peptides corral microbes, and can accumulate, leading to AD (right). [Courtesy of Hur et al., Nature, 2020.]
“Taken together, this broad and deep study suggests that IFITM3 may be a safe target for the treatment of AD, lowering Aβ production without inhibiting Notch signaling, the latter a well-known liability of γ-secretase inhibitors,” commented Michael Wolfe of the University of Kansas in Lawrence.
Li told Alzforum that efforts are already underway to develop IFITM3 inhibitors. In a comment to Alzforum, Weiming Xia of Boston University School of Medicine agreed that this might be a promising therapeutic approach. “While efforts have been made to explore other γ-secretase interacting proteins as therapeutic targets for AD in the past two decades, IFITM3 will likely outperform those previous targets for a number of reasons, mainly anti-inflammation/Aβ dual efficacies, existing potent compounds as candidate IFITM3 inhibitors, and anti-aging potential,” he wrote. Tamping down IFITM3 expression with brain-penetrant, anti-inflammatory drugs is one potential strategy, Xia noted. Another is to modify existing small molecule GSMs to thwart IFITM3’s binding to PS1.
“The identification of a subpopulation of LOAD patients in whom IFITM3 expression strongly correlates with γ-secretase activity suggests that IFITM3 may be used as a biomarker to stratify AD patients,” added Xu. “As LOAD is a multifactorial disease, identification of biomarkers for subpopulations of AD is invaluable to studying underlying mechanisms and developing targeted therapeutics.”—Jessica Shugart
- Barcelona: Allosteric γ Modulation Moves Toward Clinic
- More Evidence that γ-Secretase Modulators Spare Essential Substrates
- New Spin on γ-Secretase—Tangled in Tetraspanin Web?
- Like a Tiny Spider-Man, Aβ May Fight Infection by Cocooning Microbes
- Herpes Triggers Amyloid—Could This Virus Fuel Alzheimer’s?
- Li YM, Xu M, Lai MT, Huang Q, Castro JL, DiMuzio-Mower J, Harrison T, Lellis C, Nadin A, Neduvelil JG, Register RB, Sardana MK, Shearman MS, Smith AL, Shi XP, Yin KC, Shafer JA, Gardell SJ. Photoactivated gamma-secretase inhibitors directed to the active site covalently label presenilin 1. Nature. 2000 Jun 8;405(6787):689-94. PubMed.
- Xu M, Lai MT, Huang Q, DiMuzio-Mower J, Castro JL, Harrison T, Nadin A, Neduvelil JG, Shearman MS, Shafer JA, Gardell SJ, Li YM. gamma-Secretase: characterization and implication for Alzheimer disease therapy. Neurobiol Aging. 2002 Nov-Dec;23(6):1023-30. PubMed.
- Pozdnyakov N, Murrey HE, Crump CJ, Pettersson M, Ballard TE, Am Ende CW, Ahn K, Li YM, Bales KR, Johnson DS. γ-Secretase Modulator (GSM) Photoaffinity Probes Reveal Distinct Allosteric Binding Sites on Presenilin. J Biol Chem. 2013 Apr 5;288(14):9710-20. PubMed.
- Wakabayashi T, Craessaerts K, Bammens L, Bentahir M, Borgions F, Herdewijn P, Staes A, Timmerman E, Vandekerckhove J, Rubinstein E, Boucheix C, Gevaert K, De Strooper B. Analysis of the gamma-secretase interactome and validation of its association with tetraspanin-enriched microdomains. Nat Cell Biol. 2009 Nov;11(11):1340-6. PubMed.
- Hur JY, Frost GR, Wu X, Crump C, Pan SJ, Wong E, Barros M, Li T, Nie P, Zhai Y, Wang JC, Tcw J, Guo L, McKenzie A, Ming C, Zhou X, Wang M, Sagi Y, Renton AE, Esposito BT, Kim Y, Sadleir KR, Trinh I, Rissman RA, Vassar R, Zhang B, Johnson DS, Masliah E, Greengard P, Goate A, Li YM. The innate immunity protein IFITM3 modulates γ-secretase in Alzheimer's disease. Nature. 2020 Oct;586(7831):735-740. Epub 2020 Sep 2 PubMed.