RAGE interacts with RIPK1 and enhances its phosphorylation
Given the crucial role of RAGE in hyperglycemia-induced neuroinflammation, we initially investigated the RAGE-interacting proteins in hippocampus of db/db mice through liquid chromatograph-mass spectrometer (LC-MS) analysis. We found that the abundance of proteins associated with the mitogen-activated protein kinase (MAPK) signaling pathway was significantly increased in db/db mice [21]. It was worth mentioning that the protein abundance of RIPK1, which can induce neurodegenerative diseases by enhancing the inflammatory response, was found to be significantly elevated in db/db mice (Fig.1 a1 and a2). Consistent with this finding, analysis by the ‘Correlation Analysis’ tool of the Gene Expression Profiling Interactive Analysis (GEPIA) database showed that both RIPK1 and its downstream protein NLRP3 inflammasome were positively correlated with protein expression of RAGE (r = 0.81 and 0.58 respectively, Fig. 1 b1 and b2). We logically next explored the main cell types of RAGE, RIPK1 and NLRP3 inflammasomes expressed in the brain through the single-cell database PanglaoDB and showed that RAGE is ubiquitously expressed in the central nervous system (CNS), while RIPK1 and NLRP3 are mostly expressed in microglia (Fig.1 c1-c4, d1-d4).
In addition, phosphorylated RIPK1 (p-RIPK1) is the form of RIPK1 activation, which play a key role in RIPK1-caused inflammation [22]. Therefore, it is necessary to determine the RIPK1 phosphorylation in high glucose condition. As expected, RIPK1 auto-phosphorylated (pS166) was increased after subjecting BV2 microglia to high-glucose conditions for 48 h (F (3, 12) = 86.22. HG 48 h vs NG, p < 0.001, Supplementary Fig. 1 a1 and a2), and we used 48 h of high-glucose conditions in the rest of the experiments. In addition, the hippocampus of db/db mice underwent RIPK1 auto-phosphorylation at S166 (t = 5.70, db/m vs db/db, p < 0.001, Supplementary Fig. 1 b1 and b2). Based on these observations, RIPK1 activates in high glucose environment may associated with RAGE.
RAGE regulates RIKP1 phosphorylation through RAGE-RIPK1 conjunction
Next, we validated the effect of RAGE on the RIPK1 phosphorylation in high glucose condition. In BV2 microglia, high glucose elevated p-RIPK1 level, while RAGE inhibitor FPS-ZM1 treatment down-regulated this increase, and FPS-ZM1 solvent DMSO had no reverse effect (F (3, 12) = 94.17, HG vs NG, p < 0.001, FPS vs HG, p < 0.001, DMSO vs FPS, p < 0.001, Fig. 2 a1 and a2). Because the cognitive disorder in db/db mice is gradual progress, the FPS-ZM1 was treated at 7-8 weeks and p-RIPK1 was observed at the age of 15-18 weeks. p-RIPK1 increased during disease progression and FPS-ZM1 was able to provide a better therapeutic effect (F (3, 12) = 165.30, db/db vs db/m, p < 0.001, FPS vs db/db, p < 0.001, DMSO vs FPS, p < 0.001,
Fig. 2 b1 and b2). We further tested the co-precipitation of RAGE with RIPK1, as shown in Fig. 2 c1 and c2, high glucose-induced co-precipitation levels of RAGE with RIPK1 were significantly increased in BV2 microglia, which could be abolished by the RAGE antagonist FPS-ZM1 (F (4, 10) = 176.30, HG vs NG, p < 0.001, FPS vs HG, p < 0.001, DMSO vs FPS p < 0.001, Fig. 2 c1 and c2). Similarly, RAGE inhibitor was also resistant to RAGE-RIPK1 interaction in hippocampus of db/db mice (F (3, 8) = 79.76, db/db vs db/m, p < 0.001, FPS vs db/db, p < 0.001, Fig. 2 d1 and d2). We also measured the co-location of RAGE and RIPK1 in high glucose-induced BV2 microglia by staining. The results showed that RAGE expression was elevated and strongly co-localized with RIPK1 in BV2 microglia under high glucose condition (F (4, 45) = 109.20, HG vs NG, p < 0.001, Fig. 2 e1 and e2). In contrast, cells receiving FPS-ZM1 exhibited lower level of RAGE and less overlapping values (F (4, 45) = 109.20, FPS vs HG, p < 0.001, Fig. 2 e1 and e2).
Interaction between RAGE and RIPK1 accelerates RIPK1-related signaling pathway activation in BV2 microglia
Since RAGE has been shown to adjust the RIPK1 phosphorylation, we further elucidated its role in regulating RIPK1-mediated inflammatory signaling pathways. NLRP3 inflammasome is considered to be the critical mediator of RIPK1-related inflammation response in microglia [19], so we first figured out whether RAGE affected the NLRP3 expression. The results showed that high glucose triggered higher NLRP3 in BV2 microglia, while FPS-ZM1 significantly attenuated this augment (F (4, 15) = 55.27, HG vs NG, p < 0.001, FPS vs HG, p < 0.001, DMSO vs FPS, p < 0.001, Fig. 3 a1 and a2). Indeed, hyperglycemia promoted the activation of the NLRP3 inflammasome in microglia, leading to the oligomerization of ASC and to the formation of caspase-1-dependent inflammatory bodies and the maturation and secretion of proinflammatory cytokines, such as IL-1β and IL-18. For these reasons, we assessed the activity of NLRP3 inflammasomes during microglial was subjected to high glucose. The results demonstrated the cleaved active form of caspase-1, the protein abundance of mature IL-1β and the expression of IL-18 were markedly increased in high glucose-treated BV2 microglia. Notably, FPS-ZM1 was able to dramatically diminish the effects of RAGE overexpression on activating NLRP3 inflammasomes (F (4, 15) = 139.20 (b2), 99.33 (c2) and 68.71 (d2), HG vs NG, p < 0.001, FPS vs HG, p < 0.001, DMSO vs FPS, p < 0.001 in b2, c2 and d2, Fig. 3 b1-d2).
To further validate these observations, we transfected LV-RAGE-shRNA into BV2 microglia to knock down RAGE (t = 11.50, RAGE-KD vs NC, p < 0.001, Supplementary Fig. 2 a1 and a2) and then measured the levels of p-RIPK1, NLRP3, cleaved caspase-1, IL-1β and IL-18 under high-glucose conditions (Supplementary Fig. 2 b1 and f2). The reduction of RAGE downregulated the expression of p-RIPK1 and NLRP3, and inhibited the secretion of inflammatory chemokines ( F (3, 12) = 98.56 (b2), 137.20 (c2), 125,80 (d2), 158.20 (e2) and 77.77 (f2), HG vs NG, p < 0.001, HG + RAGE-KD vs HG, p < 0.001, HG + NC vs HG + RAGE-KD, p < 0.001 in b2, c2, d2, e2 and f2, Supplementary Fig. 2 b1-f2). Furthermore, BV2 microglia were transfected with LV-RAGE to overexpress RAGE (t = 12.89, LV-RAGE vs LV-NC, p < 0.001, Supplementary Fig. 2 g1 and g2), and then subjected to high glucose stimulation. RAGE expression, RIPK1 and its associated downstream signaling pathway were assessed by WB. The results showed that overexpression of RAGE triggered the activation of RIPK1 and NLRP3 inflammasome in high glucose-treated BV2 microglia (F (3, 12) = 164.70 (h2), 121.00 (i2), 123.40 (j2), 136.40 (k2) and 128.80 (i2), HG vs NG, p < 0.001, HG + RAGE-KD vs HG, p < 0.001, HG + RAGE-NC vs HG + RAGE-KD, p < 0.001 in h2, i2, j2, k2 and i2, Supplementary Fig. 2 h1-l2). Together, these findings demonstrate that RAGE activated RIPK1 and subsequent excited NLRP3 inflammasome, leading to the production of inflammatory factors in microglia.
RAGE directly binds RIPK1 through its intracellular 362-367 domain
To further investigate whether RAGE regulated RIPK1 phosphorylation via direct binding to its regulatory regions, we next tested the direct interaction through GST pull-down assay. We generated recombinant His-tagged pIRES2-RIPK1 plasmids and tranfected them into BV2 microglia, then purified the extracted proteins using nickel columns (Fig. 4 a1). GST-pGEX-4T-1-RAGE recombinant plasmid was also overexpressed and purified as bait protein, and His-RIPK1 was a prey protein. The data indicated that His-RIPK1 combined with GST-RAGE directly (Fig. 4 a2). RAGE mediate the effects of ligands or endogenous stimulators through its intracellular C-terminal domain, thereby activating downstream signaling pathways that lead to cellular injury [23]. Specifically, we next focused on clarifying the precise interacting motifs for RAGE binding to RIPK1. On account of the reported motifs for RAGE binding to protein and small molecule compounds, we generated two recombinant mutations in the C-terminal RAGE of mice (Mut1: R362A-K363A-R364A- Q365A-P366A-R367A, Mut2: R383A-E385A) (Fig. 4 b). GST pull-down results demonstrated that RAGE with AAs 362-367 mutation diminished the binding of RAGE and RIPK1, but the AAs 383-385 mutation did not have this effect. (Fig. 4 c). According to this analysis, the AAs 362-367 motif in RAGE is crucial for the RAGE-RIPK1 interaction, therefore RAGE with mutations on AAs 362-367 was selected for the subsequent experiments.
We next cotransfected Flag-RAGE-Wt/Mut and His-RIPK1 in BV2 microglia. By co-immunoprecipitation using anti-Flag and anti-His bodies respectively, we found that Flag carried RAGE-Wt, but not RAGE-Mut could co-immunoprecipitated with His, suggesting that binding between RAGE and RIPK1 in BV2 microglia was intercepted by mutated C-terminal RAGE (F (2, 9) = 77.87 (d2) and 45.64 (d4), His-RIPK1 + Flag-RAGE-Mut vs His-RIPK1 + Flag-RAGE-Wt, p < 0.001 in d2 and d4, Fig. 4 d1-d4). Additionally, the results from immunofuorescence presented that the co-localization of RIPK1 with mutant RAGE was lower than that of wild-type RAGE under high glucose conditions (F (2, 21) = 44.99, His-RIPK1 + Flag-RAGE-Mut vs His-RIPK1 + Flag-RAGE-Wt, p < 0.001, Supplementary Fig. 3 a1 and a2). From these results, we fully verified the AAs 362-367 of RAGE mediated the interaction between RAGE and RIPK1.
The death domain (DD AAs 568-654) plays a key role in RIPK1 binding to membrane receptors (Fig. 4 e), we then employed molecular docking to predicte the spatial characteristics of RAGE and RIPK1 interaction. Notably, the docking results showed that C-terminal RAGE AAs 362-367 and RIPK1 DD were critical in the formation of the RAGE-RIPK1 interaction as they got involved in the polar contacts (Fig. 4 f1). The vacuum electrostatic map suggested RIPK1 DD formed a positively-charged “semi-pocket” that largely matched the spatial structure of RKRQPR motif in C-terminal RAGE domain (Fig. 4 f2). In particular, 559E-363K, 603D-363R, 607D/611E-366P and 603D-367R in RIPK1 DD and C-terminal RAGE were crucial binding sits in the formation of the RAGE-RIPK1 complex, as they were involved in the polar contacts (Fig. 4 f3). Taken together, these findings indicate that RAGE directly bind to RIPK1 through the interaction of RAGE AA 362-367 and RIPK1 DD.
RAGE AAs 362-367 mutation inhibits activation of the RIPK1 signaling pathway by restricting the RAGE-RIPK1 combination
T2DM is characterized by hyperglycemia with chronic low-grade inflammation, leading to severe complications [24, 25]. Hippocampus is the earliest brain function area, which were impacted by the inflammation through secreting large amounts of inflammatory factors from microglia [5]. Activation of the RIPK1 signaling pathway in microglia drives neuroinflammation through mediating RIPK1-dependent downstream proteins [26]. Therefore, we investigated the effect of mutated RAGE on the activation of the RIPK1 signaling pathway in microglia of the hippocampus, and the experimental schedule was shown in Fig. 5 a. To avoid interference of endogenous RAGE, LV-RAGE-shRNA was microinjected bilaterally into the hippocampal CA1 regions to suppress endogenous RAGE. Two weeks later, GFP tagged AAV-RAGE-Wt or Mut with CD68 promoter targeted microglia were injected bilaterally into the hippocampus to overexpress mutant RAGE in microglia. The results displayed unambiguously that LV-RAGE-shRNA decreased RAGE level in db/db mice, and the wild type and mutation equally elevated RAGE expression in hippocampus (F (5, 18) = 71.71, db/db vs db/m, p < 0.001, db/db + RAGE-KD vs db/db, p < 0.001, db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001, db/db + RAGE-KD + Mut vs db/db + RAGE-KD +Wt, p < 0.001, Fig. 5 b1 and b2). Immunofluorescence assay also presented that RFP-tagged AAV-RAGE-Wt/Mut were overexpressed in the hippocampal CA1 region despite the early knockdown of RAGE (Supplementary Fig. 4 a).
Consistent with the results of experiments in BV2 in microglia, we found that the binding of mutated RAGE with RIPK1 was dramatically reduced compared with that of wild type RAGE with RIPK1 in hippocampus of db/db mice (F (3, 12) = 45.98 (c2) and 254.20 (c4), db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001, db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p < 0.001 in c2 and c2, Fig. 5 c1-c4). We next analyzed the contribution of RAGE AAs 362-367 mutation in RIPK1 phosphorylation and found that mutant AAs 362-367 were highly resistant to hyperglycemia-induced RIPK1 phosphorylation compared to wild-type RAGE (F (5, 18) = 161.70, db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001, db/db + RAGE-KD + Mut vs db/db + RAGE-KD +Wt, p < 0.001, Fig. 5 d1 and d2). Because activation of the NLRP3 inflammasome is mainly involved in RIPK1-mediated neuroinflammation in microglia [19, 20], we also analyzed the expression of NLRP3 and its related inflammatory factors to determine whether similar mechanisms regulate inflammatory program in a hyperglycemic environment. As expected, the levels of NLRP3, cleaved caspase-1, IL-1β and IL-18 showed strong elevations in db/db mice, but these elevations were also significantly decreased by inhibition of RAGE. Correspondingly, the activation of NLRP3 inflammasome in RAGE-reduced db/db mice was induced by AAV-RAGE-Wt, but not AAV-RAGE-Mut (F (5, 18) = 51.61 (e2), 59.98 (f2), 132.50 (g2) and 143.50 (h2), db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001, db/db + RAGE-KD + Mut vs db/db + RAGE-KD +Wt, p < 0.001 in e2, f2, g2 and h2, Fig. 5 e1-h2). Iba1 immunofluorescence assay demonstrated that wild type RAGE overexpression activated microglia, while the activity of microglial was also inhibited in response to RAGE mutation in hippocampus (F (5, 42) = 22.42, db/db vs db/m, db/db + RAGE-KD vs db/db, db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001, db/db + RAGE-KD + Mut vs db/db + RAGE-KD +Wt, p < 0.001, Fig. 5 i1 and i2). Consequently, in db/db mice, mutation of RAGE (AAs 362-367) is critical for suppressing neuroinflammation via blocking RAGE-RIPK1 interaction, ultimately reducing activation of the RIPK1 signaling pathway and decreasing excitation of microglia in hippocampus.
RAGE mutation ameliorates cognitive impairment in db/db mice
Finally, we investigated the potential effect of RAGE specific mutation on cognitive impairment of db/db mice using MWM, fear conditioning test and NOR paradigms. There was no significant difference in escape latency during the first three days among the six groups. However, on fourth and fifth days, the escape latency of db/db mice was significantly prolonged relative to db/m mice and db/db mice combined with RAGE inhibition. The db/db mice with mutant RAGE had shorter escape latencies than db/db mice with wild type RAGE (On fourth day, q (8, 120) = 6.63 (db/db vs db/m, p < 0.001), = 5.51 (db/db + RAGE-KD vs db/db, p = 0.002), = 6.27 (db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001), = 4.74 (db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p = 0.02, On fifth day, q (8, 120) = 13.30 (db/db vs db/m, p < 0.001), 9.72 (db/db + RAGE-KD vs db/db, p < 0.001), 10.85 (db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p < 0.001) , 9.60 (Wt vs db/db + RAGE-KD + Mut vs Wt vs db/db + RAGE-KD + Wt, p < 0.001), Fig. 6 a1). In the probe trial, db/db mice and db/db mice received wild type RAGE performed worse than the db/m and db/db mice treated with mutation (For the time, F(5, 42) = 31.89, db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p < 0.001, For the distance, F(5, 42) = 10.52, db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p = 0.02, Fig. 6 a2-a4). Both db/db mice and knock-down db/db mice treated with wild type RAGE spent significantly less time and distance in the target quadrant than the db/m mice and RAGE knock-down db/db mice, as well as knock-down db/db mice combined with mutated RAGE (For the time, F(5, 42) = 31.89, db/db vs db/m and db/db + RAGE-KD vs db/db, p < 0.001, For the distance, F(5, 42) = 10.52, db/db vs db/m, p < 0.001 and db/db + RAGE-KD vs db/db, p = 0.002, Fig. 6 a2-a4).
Meanwhile, db/db mice presented less freezing time than db/m mice in contextual and cued fear conditioning trials, and the freezing time was significantly greater in mice with mutant RAGE than in mice receiving wild-type RAGE (In contextual test: F(5, 42) = 10.41, db/db vs db/m, p = 0.01, db/db + RAGE-KD vs db/db, p = 0.001, db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p = 0.005, db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p = 0.003, Fig. 6 b). However, in cued fear conditioning test, db/db mice and db/db mice receiving different treatments were not significantly altered (In cued test, F(5, 42) = 4.23, db/db vs db/m, p = 0.02, Fig. 6 b). Similarly, hyperglycemia induced cognitive decline, as measured by a decrease in DI, but not in mutant RAGE treated db/db mice (F(5, 42) = 7.98, db/db vs db/m, p = 0.002, db/db + RAGE-KD vs db/db, p = 0.006, db/db + RAGE-KD + Wt vs db/db + RAGE-KD, p = 0.008, db/db + RAGE-KD + Mut vs db/db + RAGE-KD + Wt, p = 0.02, Fig. 6 c). Taken together, these findings suggest that RAGE mutation attenuates cognitive deficit in db/db mice due to disruption of RAGE binding to RIPK1.