The role of neutrophil gelatinase-associated lipocalin and iron homeostasis in object recognition impairment in aged sepsis-survivor rats

Older adult patients with sepsis frequently experience cognitive impairment. The roles of brain neutrophil gelatinase-associated lipocalin (NGAL) and iron in older sepsis patients remain unknown. We investigated the effects of lipopolysaccharide-induced sepsis on novel object recognition test, NGAL levels, an inflammatory mediator tumor necrosis factor-α (TNFα) levels, and iron ion levels in the hippocampus and cortex of young and aged rats. The effect of an iron chelator deferoxamine pretreatment on aged sepsis rats was also examined. Young sepsis-survivor rats did not show impaired novel object recognition, TNFα responses, or a Fe2+/Fe3+ imbalance. They showed hippocampal and cortical NGAL level elevations. Aged sepsis-survivor rats displayed a decreased object discrimination index, elevation of NGAL levels and Fe2+/Fe3+ ratio, and no TNFα responses. Pretreatment with deferoxamine prevented the reduction in the object recognition of aged sepsis-survivor rats. The elevation in hippocampal and cortical NGAL levels caused by lipopolysaccharide was not influenced by deferoxamine pretreatment. The lipopolysaccharide-induced Fe2+/Fe3+ ratio elevation was blocked by deferoxamine pretreatment. In conclusion, our findings suggest that iron homeostasis in the cortex and hippocampus contributes to the maintenance of object recognition ability in older sepsis survivors.

. Dual lipopolysaccharide (LPS) challenge induced lethal sepsis and object discrimination impairment in aged rats. (a) Schematic representation of the experimental protocol. Young and aged rats were assigned to control or 1 mg/kg LPS intraperitoneal injection groups after being subjected to the open field test (Y-CON group, n = 11; A-CON group, n = 11; Y-LPS group, n = 8; A-LPS group, n = 20). Y-LPS and A-LPS group rats were administered LPS after each novel object recognition test (NORT) habituation session. On Day 4, survivor rats were subjected to the NORT. (b) The Kaplan-Meier survival curves showed that the survival rate of the A-LPS group was significantly lower than that of the A-CON group. (c,d) In the NORT, the discrimination index and locomotor activity of the A-LPS group were significantly reduced compared with those of the A-CON group. There were no significant differences in the survival rate and the NORT performance between Y-CON and Y-LPS groups. The behavioral data are presented as the median ± IQR or mean ± SD. Y-CON, n = 11; Y-LPS, n = 6; A-CON, n = 11; A-LPS, n = 8. Log-rank test with Bonferroni correction, # P < 0.05, A-CON vs. A-LPS; permuted Brunner-Munzel test with Bonferroni correction, **P < 0.01, effect size p̂ ≥ 0.96; Welch's ANOVA [F 3,14.1 = 17.7, P < 0.001, ω 2 = 0.73] and post-hoc Bonferroni corrected Welch's t-test, ++ P < 0.01, Hedge's g = 2.08. www.nature.com/scientificreports/ A-LPS groups (Fig. 2a,b). The A-CON group displayed higher hippocampal and cortical NGAL levels than the Y-CON group [hippocampus, median (IQR): Y-CON, 6.2 (4.2-9.6) pg/mg tissue; A-CON, 11.0 (10.0-28.0) pg/ mg tissue; permutated Brunner-Munzel test, T = 4.60, P < 0.05, effect size p̂ = 0.87; cortex, median (IQR): Y-CON, 7.8 (6.5-12.1) pg/mg tissue; A-CON, 14.0 (12.5-30.3) pg/mg tissue; T = 6.08, P < 0.01, p̂ = 0.90]. No significant differences were observed in TNFα levels in the hippocampus or cortex (Fig. 2c,d). In the hippocampus (Fig. 2e), the concentration of Fe 2+ in the A-LPS group was significantly higher than those of the A-CON and Y-LPS The Y-LPS and A-LPS groups showed significant elevations of NGAL levels in the hippocampus and cortex. The hippocampal and cortical NGAL concentrations in the A-CON group were higher than those in the Y-CON group. (c,d) There were no significant changes in the hippocampal and cortical tumor necrosis factor α (TNFα) levels. (e) The A-LPS group displayed increases in the hippocampal Fe 2+ levels compared with the A-CON group. (f) The concentration of Fe 3+ , Fe 2+ , and total iron ion in the cortex in the A-LPS group was higher than those in the A-CON group. (g,h) The hippocampal and cortical Fe 2+ /Fe 3+ ratio of the A-LPS group was increased compared with those of the A-CON group. The data are presented as the median ± IQR. Y-CON, n = 11; Y-LPS, n = 6; A-CON, n = 11; A-LPS, n = 8. Permuted Brunner-Munzel test with Bonferroni correction, *P < 0.05, effect size p̂ ≥ 0.87, **P < 0.01, p̂ ≥ 0.90. www.nature.com/scientificreports/ groups. The cortical Fe 3+ , Fe 2+ , and total iron concentration in the A-LPS group were significantly increased compared with those in the A-CON group (Fig. 2f). The ratio of hippocampal Fe 2+ to Fe 3+ in A-LPS group rats was higher than those in A-CON and Y-LPS group rats (Fig. 2g). The A-LPS group showed a higher cortical Fe 2+ / Fe 3+ ratio than the A-CON groups (Fig. 2h).

Scientific Reports
DFO pretreatment prevented LPS-caused object recognition impairment in aged rats. We next investigated whether DFO pretreatment attenuated LPS-induced lethal sepsis and object recognition deficit of aged rats (Fig. 3a). Another 26 rats were randomly assigned to the 100 mg/kg DFO injection (A-DCON, n = 6) or 1 mg/kg LPS with 100 mg/kg DFO preinjection (A-DLPS, n = 20) groups. There were no significant differences in baseline activity in the OFT between groups [time spent in the center area, median (  Fig. 3b). DFO pretreatment blocked LPS-induced object discrimination impairment of the A-DLPS group in the NORT (Fig. 3c). Pretreatment with DFO did not influence LPS-induced decreases in locomotor activity (Fig. 3d).
Pretreatment with DFO blocked iron ion accumulation and imbalance in the hippocampus and cortex in aged sepsis-survivor rats. DFO pretreatment did not suppress LPS-induced elevation of NGAL levels in the hippocampus and cortex (Fig. 4a,b). TNFα levels in the hippocampus and cortex were comparable among the groups (Fig. 4c,d). DFO injection reduced the levels of iron ions in the hippocampus and cortex (Fig. 4e,f). The hippocampal and cortical iron ion levels in the A-DLPS group were significantly lower than those in the A-LPS group. In the hippocampus, there was no significant difference between A-LPS and A-DLPS groups in the ratio of Fe 2+ to Fe 3+ (Fig. 4g). The cortical Fe 2+ /Fe 3+ ratio in the A-DLPS group was significantly lower than that in the A-LPS group (Fig. 4h). There were no significant differences in each iron ion level and the Fe 2+ /Fe 3+ ratio in the hippocampus or cortex between the A-DCON and A-DLPS groups.
Hippocampal and cortical iron levels were negatively correlated with object recognition ability. The Spearman correlation and partial correlation analysis controlling for NGAL or Fe 2+ /Fe 3+ ratio were performed to estimate the relationship between object recognition ability and brain NGAL or iron levels. There www.nature.com/scientificreports/ were no significant correlations between the discrimination index and NGAL level in the hippocampus and cortex (Fig. 5a). The Spearman correlation and partial correlation analysis controlling for NGAL revealed that the Fe 2+ /Fe 3+ ratio in the hippocampus and cortex had a significant negative correlation with the discrimination index (Fig. 5b). www.nature.com/scientificreports/

Discussion
In this study, we investigated the effects of LPS-induced sepsis on object recognition and brain NGAL, TNFα, and iron ion levels in aged and young survivor rats. Young sepsis-survivor rats showed elevation of NGAL levels in the hippocampus and cortex. Aged sepsis-survivor rats displayed impairment of object recognition ability, elevations in the hippocampal and cortical NGAL levels, and dysregulation of iron ion levels. Pretreatment with the iron chelator DFO inhibited LPS-induced object recognition impairment and iron ion elevations. In addition, the Fe 2+ /Fe 3+ ratio in the hippocampus and cortex was negatively correlated with object recognition ability. These results suggest that LPS induces cognitive dysfunction associated with dysregulation of iron metabolism in aged sepsis-survivor rats. Furthermore, DFO may prevent these LPS-induced deleterious effects. In our study, dual LPS injection caused lethal sepsis but not induced hippocampal or cortical TNFα responses 48 h after the last injection in aged survivor rats. Preconditioning of sublethal-dose endotoxin exposure induces tolerance to subsequent endotoxin exposure, reducing TNFα response and mortality 21 . This discrepancy might be because of differences in the experimental procedures (doses and timing of LPS injections). Although we could not further investigate endotoxin tolerance in this study, we found that dual LPS challenge induced severe sepsis, NGAL elevation, and cognitive impairment in aged rats. Recently, increasing evidence that NGAL is involved in both neuroinflammation and neuroprotection has been obtained 22 . Upon exposure to inflammatory conditions, NGAL is predominantly synthesized and secreted by astrocytes in the brain 14 . Secreted NGAL stimulates surrounding astrocytes and microglia to become reactive, resulting in an amplified inflammatory response 12,13 . On the other hand, NGAL absence exacerbates proinflammatory cytokine and chemokine responses in the mouse CNS after systemic LPS challenge or experimental autoimmune encephalomyelitis 15,16 . These studies indicate that NGAL might serve as a neuroprotective factor in response to neuroinflammation. However, this study provided no direct evidence that NGAL exerts inflammatory or protective effects or both.
In pathological conditions, iron accumulation and ferric/ferrous ion imbalance produce reactive oxygen species like highly reactive hydroxyl radicals by Fenton reaction, resulting in apoptosis, necrosis, autophagy, and ferroptosis 20,23 . Indeed, patients with neurodegenerative diseases, including Alzheimer's disease and Parkinson's disease, show iron accumulation in the cortex and hippocampus 24,25 . A recent study revealed that LPS induces cognitive deficits and iron accumulation in the hippocampi and cortices of aged mice by upregulating microglial heme oxygenase-1 (HO-1) expression 26 . Overexpression of HO-1 in older adults causes cognitive decline and aggregation of Aβ and Tau along with iron accumulation 27,28 . LPS increases iron accumulation with upregulation of divalent metal transporter 1 (DMT1) expression in neurons and microglia 29 . These reports support our present findings that increases in Fe 2+ levels and Fe 2+ /Fe 3+ imbalances in the hippocampus and cortex might contribute to cognitive disturbance in LPS-injected aged rats. www.nature.com/scientificreports/ In the present study, pretreatment with DFO blocked LPS-induced increases in Fe 2+ levels, iron dysmetabolism, and cognitive deficits in aged rats. Partly consistent with our findings, DFO treatment for six consecutive days can prevent neuroinflammation, iron overload, and sickness behaviors 26 . DFO has a high affinity for Fe 3+ and might chelate them in the brain following disruption of the BBB by LPS 30,31 . DFO-induced Fe 3+ chelation reduces the cytosolic labile iron pool (LIP), attenuates the expression of both DMT1 and the iron storage ferritin, and promotes the expression of the iron exporter ferroportin 1 after LPS treatment 26,32 . In addition, the reversible Fe 3+ to Fe 2+ transition might compensate for decreased Fe 3+ levels through the cytosolic LIP 33 . Indeed, compared with aged survivors, DFO-pretreated aged rats showed reductions in total iron levels and the Fe 2+ / Fe 3+ ratio in the hippocampus and cortex. Therefore, iron chelation might contribute to iron homeostasis in the hippocampus and cortex after endotoxin challenge to maintain cognitive function.
There were several limitations of this study. First, we found that aging increased approximately twofold NGAL levels in the cortex and hippocampus. A recent human and mouse plasma proteome study reveals that plasma NGAL levels show an aging-related exponential increase 34 , suggesting that NGAL might serve as an aging predictor of the CNS; however, the mechanisms remain unclear. Second, based on our previous study 18 , the present study assumed that DFO pretreatment would attenuate LPS-induced NGAL elevation and cognitive deficits. However, our present results have supported the finding that NGAL elevation in the CNS during LPS-based sepsis did not affect cognitive impairments 22 . Unexpectedly, LPS injections following DFO pretreatments caused a greater brain NGAL level elevation than LPS injections alone. Although iron-chelated DFO has anti-microbial potential, some gram-negative and -positive bacteria can use DFO for efficient iron uptake and growth 35,36 . These findings suggest that DFO pretreatment had promoted pathogenic bacterial growth in aged rat body cavity even if under a specific pathogen-free environment, probably leading to enhance immune response and sepsis. Thus, DFO-induced iron depletion might prime NGAL responsiveness as an anti-bacterial effector or iron donor; however, whether LPS-induced NGAL elevation exerts pathogenic, protective, or no roles in the CNS via iron-endocytosis or -exocytosis remains unknown. Furthermore, the dose of DFO 32 mg/kg/day, equivalent to rat dose 200 mg/kg 37 , improved outcomes of patients with intracerebral hemorrhage at day-180 38 .
In the aged rat model of intracerebral hemorrhage, DFO 100 mg/kg (human-equivalent dose, 16 mg/kg/day 37 ), the same amount used in this study, improved neurological and functional recovery 39 . These clinical trials and experimental studies indicate that DFO might have protective effects on cognitive function; however, additional studies are needed to validate the safety and effectiveness of DFO to sepsis-associated cognitive dysfunction. Finally, we could not investigate the effects of DFO on brain HO-1 expression and other cognitive-behavioral tests such as the object location test, contextual fear conditioning, and various operant tasks in this study. Further behavioral and biochemical studies are required to provide a comprehensive mechanism of sepsis-induced cognitive dysfunction, including NGAL, HO-1, and other iron regulator proteins.
In conclusion, we have demonstrated that iron homeostasis in the cortex and hippocampus contributes to maintenance of object recognition after LPS-induced sepsis in aged rats. Although further study is required to improve the survival rate, our study provides new insights into the potential of pharmacological iron chelators as therapeutic targets for sepsis-associated cognitive deficits, including SAE.

Methods
Animals. Male Sprague-Dawley rats aged 1.5-3 months (n = 19, 180-440 g) and 12-18 months (n = 57, 580-900 g) were used for behavioral and biochemical analysis. The animals were group-housed at 24 ± 2 °C under a 12-h light-dark cycle (lights on at 8:00 a.m.), and food and water were available ad libitum. All experiments were carried out in accordance with the ARRIVE Guideline and the guidelines for animal research issued by the Hirosaki University Graduate School of Medicine. This study was approved by the Animal Research Committee of Hirosaki University (approval number M19019). All experimental analyses were performed by the observers blinded and automatic analysis software. All efforts were made to minimize the number of animals used and their suffering.

Drugs
Rats in the LPS groups were intraperitoneally (i.p.) injected with 1 mg/kg of LPS (E. coli O111:B4; Wako Pure Chemical Industries, Osaka, Japan) after two consecutive daily habituation sessions for the novel object recognition test. Another groups of aged rats were i.p. treated with 100 mg/kg of DFO (Novartis Pharmaceutical Corporation, Tokyo, Japan) 1 h before LPS administration or treated with DFO 39 .
Behavioral experiment protocol. An experimental protocol based on and modified from previous studies was used to investigate the effects of LPS on cognitive function 40,41 . Rats were randomly assigned to four groups: the aged control (A-CON); aged 1 mg/kg LPS injection (A-LPS); young control (Y-CON); and young 1 mg/kg LPS (Y-LPS) groups (Fig. 1a). Other aged rats were randomly assigned to the 100 mg/kg DFO injection (A-DCON) or 1 mg/kg LPS with 100 mg/kg DFO preinjection (A-DLPS) groups (Fig. 3a).
Open field test (OFT). We performed the OFT to confirm the effects of age on locomotor activity and anxiety. The behavior of the rats in an OFT arena (60 × 60 × 60 cm, 50 lx) was recorded for 5 min using the CaptureStar video tracking system (CleverSys, Reston, VA, USA). Horizontal locomotor activity [total distance traveled (m)] and the time spent in the center area (s) were automatically analyzed with TopScan behavioral analysis software (CleverSys).
Novel object recognition test (NORT). Rats were habituated to a NORT arena (60 × 60 × 60 cm, 50 lx) not containing objects for 5 min for two consecutive days. Twenty-four hours after the last habituation session, www.nature.com/scientificreports/ the animals were subjected to the NORT. Each rat was introduced to the test arena, which contained two reference objects, and allowed to explore for 5 min freely (familiarization session). After a 1-h retention interval, the animals were returned to the same test arena containing a familiar object and novel one and allowed to explore for 5 min (probe test session). The objects were colored brick towers (5 × 5 × 10 cm), white Styrofoam balls (10 cm diameter), or grey steel boxes (3 × 8 × 8 cm). These objects were randomly used as familiar or novel without duplication. During each session, exploratory behavior was recorded with the CaptureStar system (Clever-Sys). The total distance traveled (m) and time spent exploring each object (s) in the probe test were automatically analyzed by TopScan software (CleverSys). The following formula was used to assess novel object recognition ability during the NORT probe test session: discrimination index = [novel object contact (s) − familiar object contact (s)]/total object contact time (s).
Biochemical assay for NGAL and TNFα. Rats were sacrificed by rapid decapitation 24 h after the NORT, and the hippocampi and cortices were collected. Each brain sample was homogenized in ice-cold phosphatebuffered saline. The centrifuged supernatants were stored at − 20 or -80 °C until further analysis. NGAL levels in the hippocampus and cortex were measured using an enzyme-linked immunosorbent assay kit (NGAL ELISA Kit, KIT046, BioPort Diagnostics, Hellerup, Denmark). The assay detection limit was 0.5 pg/ml, and the intraand interassay coefficients of variance were 4% and 12%, respectively. The hippocampal and cortical TNFα levels were also determined by an ELISA kit (TNF-α ELISA kit CSB-E11987r, Cusabio Technology, Wuhan, China). The detection limit of the assay was 1.56 pg/ml, and the intra-and interassay coefficients of variance were < 8% and < 10%, respectively. Data were obtained using the Bio-Rad iMark™ Microplate Reader (Bio-Rad Laboratories, Hercules, CA, USA).
Iron ion level assay. Total iron and Fe 2+ levels in the hippocampus and cortex were measured using an iron assay kit (Iron Assay Kit ab83366, Abcam, Tokyo, Japan). According to the instructions, the brain samples were homogenized in assay buffer and centrifuged. For Fe 2+ level assay, Fe 2+ reacted with 3-(2-pyridyl)-5,6-bis(2-(5-furylsulfonic acid))-1,2,4-triazine (Ferene-S) was measured at 593 nm. Total iron ions were detected at 593 nm following the reduction of Fe 3+ to Fe 2+ using an iron reducer provided by the assay kit 42 . Colorimetric data were acquired using the microplate reader (Bio-Rad Laboratories). The assay detection limit was 8 μM. The Fe 3+ level was calculated by subtracting the Fe 2+ level from the total iron ion level.
Statistical analysis. The sample size (estimated minimal N/group = 5) to achieve a power (1-β error probability) of 0.8, α error probability of 0.05, and effect size of 0.94 was determined by G*Power 3.1.9.7 (Faul F, University of Kiel, Kiel, Germany) based on the NORT study by Kawano et al. 40,43 . The Shapiro-Wilk test was used to estimate the normality of the data 44 . Based on the normality, data are presented as the mean (standard deviation, SD) or median (inter-quartile range, IQR) with individual data points 45 . The log-rank test for Kaplan-Meier analysis with Bonferroni correction was used to assess the survival data between groups 46 . The two-tailed Welch's t-test, Welch's analysis of variance (ANOVA) followed by the post hoc t-test, Brunner-Munzel test, or Brunner-Munzel permutation test was used to determine the significance of differences between groups 47 . P < 0.05 was considered significant. For multiple comparisons, each P-value was adjusted to < 0.05 by the Bonferroni correction. The effect size was estimated using Hedge's g, ω 2 , or p̂4 7,48 . The Spearman correlation analysis and partial correlation analysis controlling for NGAL level or Fe 2+ /Fe 3+ ratio in each brain region were performed to evaluate the relationships between the discrimination index and NGAL level or Fe 2+ /Fe 3+ ratio. Statistical analyses were performed with R version 4.0.5 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and MATLAB 2020b (The MathWorks, Natick, MA, USA).