All experimental procedures were conducted under protocols approved by the Institutional Animal Care and Use Committee at Duke University in accordance with the guidelines of the National Institutes of Health for animal care (Guide for the Care and Use of Laboratory Animals, Health and Human Services, National Institutes of Health Publication No. 86-23, revised 1996). Studies were conducted on C57BL6/J male adult mice (3 months; weight 20-26 g) and aged mice (18-22 months; 26-55 g) (Jackson Laboratories, Bar Harbor, ME). Nlrp3^-/- (19), Asc^-/- (19), and ASC-citrine mice (20) (3-4 months; weight 20-26 g) were bred in house. Mice were housed (3-5 animals/cage) in a 12:12-h light:dark cycle in a humidity- and temperature-controlled room with free access to food and water. The animals were acclimated for at least 1 week before starting any experiment. Note, in the aged cohort, 14 mice were excluded due to complications after surgery (tumor, stroke, death), while no mice from any other cohort were excluded.

Peptide preparation and treatment were performed as previously described (16). Briefly, ANXA1 biomimetic tripeptide (ANXA1sp, Ac-QAW, or QW-3 (15); Ac = acetyl, MW = 445.47 Da) was synthesized and purified (> 98% purity) by GenScript (Piscataway, NJ). The tripeptide was suspended in 100% DMSO and diluted in saline to a final dose of 1 mg/kg ANXA1sp in 0.5% DMSO. Mice were injected intraperitoneally with 1 mg/kg ANXA1sp in 0.5% DMSO and saline 30 min before orthopedic surgery.

We performed an open tibial fracture as previously described (21). Briefly, mice were injected with 0.1 mg/kg buprenorphine (s.c.) and they underwent orthopedic surgery with 2.1% isoflurane anesthesia. Mice not exposed to surgical manipulations served as naïve controls. Mice were euthanized under isoflurane anesthesia and were perfused with ice-cold PBS to flush blood from the brain vasculature. Hippocampi were rapidly (< 2 min) dissected on ice, frozen in liquid nitrogen for protein assays, and stored at -80°C until assay.

Mice destined for immunostaining were perfused with an ice-cold solution of 4% paraformaldehyde (PFA) PBS solution. Whole brains were incubated at 4°C in the 4% PFA solution overnight. The next day brains were rinsed in ice-cold PBS and incubated in 20% and 30% sucrose at 4°C for 24 h, respectively. The brains were placed into OCT solution and stored at -80°C for further use.

Mouse hippocampi were lysed in radioimmunoprecipitation assay (RIPA) buffer (R0278; Sigma) containing Halt Protease Inhibitors Cocktail (78430; Thermo Scientific). Proteins were separated on 4%-15% or 4%-20% Criterion TGX precast protein gels (64134751; Bio-Rad), and transferred to an Immun-Blot PVDF membrane (1620177; Bio-Rad). After blocking with 5% nonfat dry milk (170-6404; Bio-Rad), the blot was probed with the following antibodies (1:1000): NLRP3 (LS-C374964; LSBio), ASC (67824S; Cell Signaling), Caspase1 (Cleaved Asp210) (LS-C380449; LSBio), IL-1β (LS-C104813; LSBio), and Annexin-A1 (ANXA1) (3299S; Cell Signaling) or (71-3400; Invitrogen). The proteins were visualized using SuperSignal West Dura Extended Duration Substrate (34076; Thermo Scientific) on a ChemiDoc MP Imaging System (Bio-Rad). Protein loading was assessed using anti-GAPDH (2118S; Cell Signaling). Bands were quantified with Fiji software, and analyzed with Prism 8 (GraphPad Software). Hippocampal levels of ATP were determined using a commercially available assay kit (ab83355, Colorimetric/Fluorometric) following the manufacturer’s instructions.

Immunostaining and analyses of microglial morphology were performed on the PFA-perfused slices using ionizing calcium-binding adaptor molecule 1 (Iba1) rabbit antibody (019-19741; Wako). Briefly, brain slices (30 μm thick) were incubated in 10% normal donkey or goat serum in PBST [1 x PBS (11189, Gibco) + 0.5% Tween 20 (P1379, Sigma)] containing 0.3% Triton X 100 (T8787, Sigma) for antigen retrieval and to block any background for 2 h at RT. Slices were incubated with the primary antibody (1:500) at 4°C overnight. After 3 washes with PBST, slices were incubated with the secondary antibody conjugated with Alexa Fluor 488 or Fluor 647 (1:500, Invitrogen) for 2 h at RT. After 5 washes with PBST, the sections were mounted, dehydrated, and cover-slipped. Images were acquired using Zeiss 880 inverted confocal Airy scan.

Imaged tile scans were stitched together and presented as high-resolution images. Microglial morphology was quantified as described in Schwarz et al. (22), based on 4 morphologic subtypes: round/amoeboid microglia (Round), stout microglia (Stout), microglia with thick long processes (Thick), and microglia with thin ramified processes (Thin). The number of Iba1-positive cells was determined in three representative areas of the hippocampal CA1 region per animal. To determine the state of microglial activation, we manually traced their branch lengths using Imaris Filament Tracer to compare branch lengths from soma and the complexity of individual microglia in the CA1 region before and after orthopedic surgery.

Hippocampal-dependent memory was assessed with two mnemonic tasks (23, 24). The first task assessed memory for objects (“what”), their locations (“where”), and the order in which they were presented (“when”). A second task examined memory load with seven unique objects. For the What-Where-When (W-W-W) task, training/testing were conducted in a 65 x 52 x 25 cm arena with complex visual cues on the walls of the room. Briefly, mice were acclimated to the arena for 5 min and then were subjected to two 5-min training and a single test trial, each separated by 55-60 min. In the first training trial, mice were presented with 4 identical objects (Set A). Three of the objects were spaced equally along the long axis of one wall with the fourth object centered along the opposite wall in a triangle formation. In the second training trial (Set B), a new set of 4 identical objects was presented. Objects 1 and 2 of Set B overlapped in location with objects 1 and 3 of Set A, while the two remaining objects of Set B were positioned in corners 3 and 4. At testing two objects each from Sets A and B were used, for a total of 4 objects. One of the Set A objects was placed in corner 1 (stationary object A). The second Set A object was placed in corner 3 (displaced Set B object). Set B objects remained in corners 2 and 4 (“recent” objects). Behavior was filmed for all trials and scored by blinded researchers subsequently with Noldus Ethovision 11 (Noldus Information Technologies, Asheville, NC) using nose-point tracking. An object interaction was scored when the nose point was within 2 cm of the object center with the body axis of the mouse oriented toward the object. All preference scores were calculated from the test trial. “What” preference was calculated as the time exploring the two A objects subtracted from the exploration times of the two B objects (recent objects in corners 2 and 4), divided by the total times for all of the A and B objects. “When” memory preference was calculated as the exploration time for stationary A object (in corner 1) minus the exploration times of the two B objects (corners 2 and 4), divided by the total times with these three objects. “Where” preference score was calculated as the difference between the exploration time for the A object (which displaced object B in corner 3) minus the time with stationary A object (corner 1) divided by the total times with these two objects.

Twenty-four hours after completion of the W-W-W task, the mice were subjected to a seven-trial memory load test (23, 24). This test was conducted in the same arena as the W-W-W task, but without visual cues on the walls. Testing was performed over 7 consecutive test trials, with an approximate 60-90 sec inter-trial interval. To accommodate the increasing number of objects on each trial, the first three trials were 3 min, trials 4-5 were 4 min, and trials 6 -7 were 5 min in length. On trial 1, one object was presented. With each subsequent trial a novel object (i.e., differing in configuration, color, and texture) was added in a random location to prevent the animal from predicting the new object based on the previous object’s placement. By the end of testing, all objects were equally spaced with no object being within 8-10 cm from the nearest object or wall of the test arena. Once the exploration time for a given trial is over, the mouse was placed into an empty mouse cage while the next object was placed in the arena before the mouse was reintroduced into the arena in the same location as before. Testing was filmed and scored using Noldus Ethovision 11 as in the W-W-W test. Exploration time with test objects was expressed as time spent in contact with the object/min to adjust for the varying lengths of the test-trials. Preference for the new object in each trial was expressed as the time spent with the new object minus the time spent with the remaining objects, divided by the sum of these times. The preference ratio ranged from -1 to +1, where positive scores indicated a preference for the new object, negative scores indicated a preference for the previously introduced objects, and a score approaching “0” indicated no preference.

The behavioral data were analyzed with SPSS 25 (IBM SPSS, Chicago, IL), while all other data were analyzed with Prism 8 (GraphPad Software). The preference scores and time interacting with objects were analyzed with multivariate ANOVA (MANOVA). In the W-W-W test, the dependent variables were for the 3 different preference scores that were assessed simultaneously at testing. In the memory load test, the six different preference scores for trials 2-7 were the dependent variables. In all analyses the treatment condition was the fixed factor. Post-hoc tests were with Tukey corrected pair-wise comparisons. The statistics for the behavioral studies are in Tables 1–4. All non-behavioral data were analyzed by t-tests or ANOVA followed by Tukey’s multiple comparisons post-hoc tests. These data are presented as mean ± SEM. In all cases, a P<0.05 was considered statistically significant.

Failed resolution of CNS inflammation has been proposed as a driver of this condition in PNDs (25), yet the role of ANXA1 in postoperative inflammation remains unknown. After orthopedic surgery, hippocampal expression of the endogenous pro-resolving mediator ANXA1 was significantly reduced, peaking at 24 h (0.63 ± 0.06) compared to naïve (1.7 ± 0.13, P < 0.001) and almost fully resolving by 72 h (1.13 ± 0.11, P < 0.01 vs. naïve; Figure 1A). Given the acute reduction in endogenous ANXA1, we next treated mice with 1 mg/kg ANXA1sp (i.p.) or vehicle (0.5% DMSO in saline, i.p.) 30 min before surgery. After orthopedic surgery, the NLRP3 inflammasome was found to be activated in the CNS of vehicle-treated animals when compared to naïve controls (1.35 ± 0.05 vs. 0.07 ± 0.01, P < 0.001; Figures 1B, C), including increased ASC expression (0.89 ± 0.07 vs. 0.08 ± 0.01, P < 0.001; Figures 1B, D), activation of caspase-1 (1.43 ± 0.06 vs. 0.09 ± 0.02, P < 0.001; Figures 1B, E), and cleavage of IL-1β (1.33 ± 0.13 vs. 0.06 ± 0.02, P < 0.001; Figures 1B, F). Notably, NLRP3 inflammasome activation was effectively reduced in ANXA1sp-treated mice (0.71 ± 0.05) compared to vehicle-treated mice after surgery (P < 0.001; Figures 1B, C). Hippocampal expression of ASC (0.461 ± 0.03; Figures 1B, D), caspase-1 (0.89 ± 0.04; Figures 1B, E), and IL-1β (0.81 ± 0.11; Figures 1B, F) were all significantly inhibited in ANXA1sp-treated mice. Hippocampal levels of ATP were also upregulated after surgery in vehicle-treated animals compared to naïve controls (17.81 ± 0.86 vs. 4.14 ± 0.29, P < 0.001; Figure 1G). However, ATP was not induced in ANXA1sp-treated mice (6.26 ± 1.44) compared to naïve animals, while it was significantly reduced relative to vehicle-treated mice (P < 0.001; Figure 1G).

Morphologic changes in microglia have been reported in several models of surgery-induced cognitive impairments, including orthopedic surgery (26). Indeed, in our model of orthopedic surgery, we observed postoperative microglial morphologic changes in the hippocampus, with more stout (13.00 ± 2.1 vs. 1.00 ± 0.45, P < 0.001; Figures 2A, B) and fewer thin Iba-1-postive cells at 24 h (4.80 ± 0.49 vs. 21.60 ± 2.32, P < 0.001; Figures 2A, B) indicating a pathological state compared to naïve controls. Tracing of these individual microglia further showed this stout morphology by decreased branch length with increased soma size after orthopedic surgery compared to thinner cells in naïve controls. Notably, postoperative microgliosis was improved in adult mice pretreated with ANXA1sp, with Iba-1 immunoreactivity shifted toward a more resting and ramified phenotype.

To confirm the role of the NLRP3 inflammasome in surgery-induced neuroinflammation, Nlrp3^-/- and Asc^-/- adult mouse lines were used. Nlrp3^-/- mice remained protected from surgery-induced neuroinflammation. In fact, compared to wild-type mice (presented above), surgery in Nlrp3^-/- mice failed to upregulate expression of ASC (0.59 ± 0.14 vs. 0.92 ± 0.05, P = 0.06) as shown by Western blot analysis (Figures 3A, B). Caspase-1 and cleavage of pro-IL-1β are non-detectable in Nlrp3^-/- mice. Further, Nlrp3^-/- displayed no signs of reactive microgliosis at 24 h after surgery, retaining thin and ramified morphologies (Figures 3C, D).

Next, we evaluated Asc^-/- mice and found protection from surgical trauma that was similar to our findings in Nlrp3^-/- mice, with minimal effects on NLRP3 (1.28 ± 0.07 vs. 1.22 ± 0.09, P = 0.56) and cleavage of pro-IL-1β (0.83 ± 0.01 vs. 0.80 ± 0.01, P = 0.13; Figures 4A–C). To further evaluate the role of NLRP3 inflammasome activation in surgery-induced inflammation, we used ASC-citrine (ASC-cit) reporter mice. After surgery, ASC-cit mice displayed a significant increase in ASC speck formation, indicating the activated NLRP3 inflammasome complex, at 24 h (42.80 ± 5.46 vs. 21.60 ± 2.4, P < 0.01; Figures 4D–F). These mice also showed signs of hippocampal microgliosis similar to C57BL6/J mice, with greater Iba-1 immunoreactivity and more stout cell bodies 24 h after surgery. Notably, some of these ASC specks were co-localized with Iba-1-postive microglia (15.00 ± 0.4 vs. 9.60 ± 0.51, P < 0.001; Figures 4D, F).

The mice were 3 months of age at the time of behavioral testing. To study the functional effects of ANXA1sp, we evaluated cognitive responses that are affected in our PND-like model (Figures 5A, B). In the “What-Where-When” task (27), mice were given free access to two sets of four objects. At testing, “What” memory was examined to determine whether the animal remembered the replacement of an object with another; “Where” memory analyzed if the animals recalled where an object had been previously located; and “When” memory reflected whether the mouse remembered the objects that were last investigated. For “What” memory, responses among the three groups (naïve, surgery plus vehicle injection, and surgery plus ANXA1sp treatment) were not statistically distinguished (Figure 5C; Table 1). Nevertheless, marked statistical differences in object preferences for the “Where” (Figure 5D) and “When” (Figure 5E) memory tests were evident (Table 1). For “Where” memory the vehicle-treated mice with surgery showed a significant reduction in their abilities to identify the displaced object compared to the naïve control (P<0.011) and mice given ANXA1sp prior to surgery (P<0.038); these two latter groups had similar preferences for the displaced object. The “When” comparison was more difficult for the animals. Both the naïve and vehicle + surgery mice had similar moderate preference scores. By contrast, mice treated with ANXA1sp prior to surgery had significantly higher preference scores than the other two groups (P-values<0.052). The higher preference scores for the ANXA1sp + surgery mice cannot be attributed to longer times spent with the objects or differential locomotor activities since they had some of the lowest object interaction times across training and testing than any other group (Table 2). Indeed, no significant group differences in object interaction times were discerned at training with the set A objects (Figure 5F). At training with set B objects, mice treated with the vehicle prior to surgery spent more time interacting with the objects compared to the naïve mice and those given ANXA1sp prior to surgery (P-values<0.051). Interestingly, at testing the vehicle + surgery and ANXA1sp + surgery mice spent less time exploring objects compared to the naïve animals (P-values<0.001).

In the memory load test naïve mice and those treated with ANXA1sp prior to surgery demonstrated an ability to identify the novel object even as the memory load increased with up to six objects (Figure 5G; Table 3). By comparison, the mice treated with the vehicle prior to surgery were impaired. Here, the vehicle + surgery mice had significantly reduced preferences for the novel object compared to naïve group when 2 (P<0.052), 4 (P<0.051), 5 (P<0.001), and 6 (P<0.001) objects were present. Although performance by the mice in the naïve and ANXA1sp + surgery groups were similar, the ANXA1sp-treated group performed even better than the vehicle + surgery mice when 2 (P<0.057), 3 (P<0.018), 4 (P<0.001), 5 (P<0.001), and 6 (P<0.001) objects were presented. When time spent investigating objects across trials was examined, few group differences were observed (Figure 5H; Table 4). However, on the first trial both treatment groups with surgery engaged in higher object exploration times than the naïve control (P-values<0.022). When 4 objects were present, mice in the ANXA1sp + surgery group spent more time in object exploration than naïve mice (P<0.032). With 5 objects the naïve animals interacted with the objects for a longer time than the vehicle + surgery mice (P<0.001). Together these results demonstrate that vehicle-treated tibia-fracture mice are markedly impaired in cognitive tasks compared to naïve mice, and that treatment with ANXA1sp prior to surgery prevents these cognitive impairments.

Aging is a critical risk factor for many neurologic conditions, including PND (28). Aged mice showed a heightened, but similar, curve for NLRP3 activation after surgery, which peaked at 6 h (2.48 ± 0.06, P < 0.0001), remained elevated at 24 h (2.00 ± 0.14, P < 0.0001), and resolved to baseline by 72 h (0.98 ± 0.05, P = 0.28) compared to adult mice (0.70 ± 0.09). Notably, more NLRP3 activation was found in aged mice compared to adult mice after surgery [R² = 0.98; F(1,3) = 130.5]; P = 0.0014; Figure 6A].

We next administered the same therapeutic regimen of ANXA1sp in 18- to 22-month-old C57BL6/J mice. Western blot analysis showed increased levels of NLRP3 in the hippocampus (1.95 ± 0.16 vs. 0.32 ± 0.11, P < 0.001; Figures 6B–F). This exacerbated the NLRP3 inflammasome activation which was significantly blunted in aged mice pretreated with ANXA1sp (1.39 ± 0.08 vs. 1.95 ± 0.16, P < 0.001; Figure 6C), affecting the expression of ASC (ANXA1sp vs. vehicle: 0.85 ± 0.23 vs. 1.79 ± 0.34, P < 0.001; Figure 6D), activation of caspase-1 (ANXA1sp vs. vehicle: 1.51 ± 0.07 vs. 2.18 ± 0.20, P < 0.001; Figure 6E), and levels of IL-1β (ANXA1sp vs. vehicle: 1.77 ± 0.15 vs. 2.50 ± 0.36, P < 0.001; Figure 6F). We next evaluated microglial activation at 24 h after surgery. Surgery triggered more stout (16.4 ± 1.78 vs. 2.4 ± 1.03, P < 0.001) and round (4.20 ± 0.37 vs. 0.60 ± 0.24, P < 0.001) Iba-1-postive cells in the hippocampus with reduced branch length compared to naïve controls. The round microglia were hardly visible in adult mice at the same time point (Figures 6G, H). Further, hippocampal ATP was retained at baseline levels in aged mice pretreated with ANXA1sp vs. vehicle (17.56 ± 103 vs. 29.56 ± 2.02, P < 0.001; Figure 6I). Overall, aged mice pretreated with ANXA1sp showed significantly improved microglial pathology, especially soma enlargement and swelling, 2 key features of CNS dyshomeostasis (29).