Age-related Dysregulation of Autophagy Contributes to Microglial Dysfunction and Chronic Neurobehavioral De cits After Traumatic Brain Injury

Rodney Ritzel University of Maryland School of Medicine https://orcid.org/0000-0002-0160-2930 Yun Li University of Maryland School of Medicine Zhuofan Lei University of Maryland School of Medicine Jordan Carter University of Maryland School of Medicine Junyun He University of Maryland School of Medicine Harry Choi University of Maryland School of Medicine Niaz Khan University of Maryland School of Medicine Hui Li University of Maryland School of Medicine Samantha Allen University of Maryland School of Medicine Marta M Lipinski University of Maryland School of Medicine Alan I Faden University of Maryland School of Medicine Junfang Wu (  junfang.wu@som.umaryland.edu ) University of Maryland School of Medicine https://orcid.org/0000-0003-3338-7291


Background
Neurodegenerative disease in the elderly is a major healthcare challenge and growing economic burden. Traumatic brain injury (TBI), a leading cause of death among young people, is increasingly prevalent in the aging population. Millions of TBI survivors live with varying degrees of disability and have an increased risk for developing neurodegenerative disease such as Alzheimer's disease (AD), Parkinson's disease and chronic traumatic encephalopathy [1,2]. Elderly TBI patients have higher mortality and greater disability than younger individuals, and are more likely to develop a neurodegenerative disorder [3][4][5]. For these reasons, older subjects are often excluded from clinical trials and treated less aggressively Our knowledge of TBI pathophysiology in aged animals is limited, and our understanding of the secondary injury mechanisms that affect long-term outcome remains limited. Whereas direct targeting of grouping of each mouse was tasked with handling the subjects. The Catwalk apparatus records print position, gait postures, and weight distribution through its green illuminated walkway. A minimum of 3 valid runs, complete crossings with no turns or scaling of sidewalls, were obtained for each tested mouse. Runs that didn't comply to the preset standards were excluded from the nal analysis.
Rotarod test: The accelerating Rotarod was used to assess locomotor function and coordination at baseline and various timepoints post-injury [24]. The mouse was placed on a rotarod device (IITC Life Science, Inc.), and their latency to falling off the accelerating rotarod was recorded. The acceleration settings for the device were 4 to 40 rpm over 90 seconds (s), with each trial lasting for a maximum of 300s. Individual scores from three trials were averaged and evaluated relative to their baseline latencies.
Grip strength (GS) test: Grip strength was measured using a digital grip strength meter (Bioseb BP, In Vivo Research Instruments, France), as previously described [25]. Forelimb grip strength was measured from the mouse using both the ipsilateral and contralateral forepaws together. The mouse was held by its tail, the forelimbs were placed on the grasping metal wire grid and the mouse gripped the wire grid attached to the force transducer. Once the grip was secured, the animal was slowly pulled away from the bar. The maximal average force exerted on the grip strength meter by both forepaws was averaged from 10 trials per day for each mouse at each timepoint.
Cognitive function Y-maze (YM) test: YM was used to assess the hippocampus-dependent spatial working memory of mice as described previously [17,18]. The percentage of spontaneous alterations was calculated using the following equation: total alternations x 100/(total arm entries -2). If a mouse had a percentage of alternation above 50% (the chance level for randomly choosing the unfamiliar arm), it was indicative of a functional spatial memory.
Novel object recognition (NOR): For testing non-hippocampal mediated memory, mice underwent NOR according to procedures described in previous studies [17,18]. Mice were tested in an open eld apparatus after a 5 min habituation period on the rst day. The time spent with two identical objects was recorded on the second day of testing, and one of the familiar objects were switched out with a novel object on the third day. Testing stopped after each mouse went through a sum total of 30 s exploration time. Since mice would inherently prefer to explore novel objects, a preference for the novel object with an exploration time of more than 15 s was considered as having intact learning and memory skills.

Depressive-like behaviors
Novelty-suppressed feeding (NSF) test: The NSF was performed on subject mice to assess a rodent's aversion to eating in a novel environment. After fasting for 24 h, mice were placed in an open eld apparatus designed to model the introduction of a novel environment. Several food pellets were placed in the center of the apparatus and latency time for each mouse to reach the food and bite into it was recorded. If the mouse failed to eat within the maximum time of 10 min, a latency time of 600 s was recorded for this subject.
Social recognition (SR) test: This task is based on rodent's innate tendency to investigate a novel congener over a familiar one [17,18,26]. Using a three-chambered rectangular apparatus made of Plexiglas with each one at equal size (20 x 40 x 23 cm). An opening between the walls allows for free access to each chamber, which contains two identical wire mesh cup containers. Before testing, each mouse was single-housed overnight. On the rst day, the tested mice were placed in the apparatus with two empty cups for a 10 min habituation period. On the second day, a stranger mouse was introduced and randomly placed inside one of the empty cups in either the left-or right-side chamber while the other cup was left empty. The tested mouse started from the middle chamber and allowed to freely explore all three chambers for an exploration period of 10 min. Afterwards, a second unfamiliar, stranger was placed inside the previously empty cup. The test subject was once again allowed to freely explore all three chambers for a period of 10 min. Exploration time that the subject mice spent with each cup versus stranger mouse was recorded. Since a socially functional mouse would naturally seek out unfamiliar mice for interaction, the test subject was considered capable of social recognition if index for novel mouse scored higher than 50%.

Thermal stimulation
Hot plate (HP) test: To test pain sensitivity of the hindpaws, mice were placed on the contact probe of computerized thermal stimulator on an Incremental Hot/Cold Plate Analgesia Meter (PE34, IITC Life Science, Woodland Hills, CA). The temperature was increased from 30 C to 50 C with the incremental rate at 10 C per minute. When the tested mouse licked either one of its hindpaws, the test was stopped, and the threshold temperature was recorded. The test was conducted twice with the interval of 3 h [27].

NanoString neuroin ammation panel analysis
Following completion of all behavioral tests, mice were euthanized, and RNA samples were obtained from the ipsilateral cerebral cortex surrounding the injury site and the ipsilateral hippocampus. Total RNA (20 ng/ul) was run on a NanoString nCounter system using the Mouse Neuroin ammation v1.0 panel (NanoString Technologies, Seattle, WA) to pro le RNA transcript counts for 757 genes and 13 housekeeping genes [28,29]. The transcription counts were normalized prior to downstream analysis and pairwise differential expression analysis with the NanoString nSolver software Version 4.0. All statistical analysis of NanoString data was performed in R language software RStudio Version 1.2.5033. Principle component analysis (PCA) was performed with the command 'prcomp' and a Euclidean distance measurement method was used for clustering. The four pairwise mentioned in this article are as follows: 1) Old Sham vs. Young Sham; 2) Young TBI vs. Young Sham; 3) Old TBI vs. Old Sham; 4) Old TBI vs. Young TBI. All pairwise comparisons of 'A vs. B' should be interpreted as 'A relative to B' in the text and gures.

RNA extraction and qPCR
Total RNA was extracted from ash frozen tissue samples with the miRNeasy Mini Kit (Qiagen, Cat# 74104). Complementary DNA (cDNA) was synthesized with the Verso cDNA RT kit (Thermo Scienti c, Cat# AB1453B). Both kits were used according to the manufacturer's instructions included in the kit box. Quantitative PCR for all target RNAs (see Table 1) was performed with the TaqMan gene expression assay kit (Applied Biosystems). Each sample was run in duplicates with 3 stages of 40 cycles, 2 min at 50 o C, 10 s at 95 o C (denaturing step) followed by a nal transcription step of 1 min at 60 o C. Gene expression was normalized by the transcription counts of GAPDH and nal relative expression levels were calculated with the 2 -ΔΔCt method [28,29].
Lesion volume, immunohistochemistry (IHC), and quanti cation A subset of tested mice was perfused intracardially with normal saline followed by 4% paraformaldehyde. The brain was extracted and embedded in Tissue-Tek OCT compound (Sakura, Cat# 4583). Serial sections of 20 mm and 60 mm thickness were placed on Superfrost Plus slides (ThermoFisher, Cat# 4951PLUS). Lesion volume was measured on 60 mm coronal sections that were stained with cresyl violet (FD NeuroTechnologies, Cat# PS102-02). Quanti cation of the lesion volume was performed with the Stereoinvestigator software (MBF Biosciences). By outlining the missing tissue on the injured hemisphere, the software was able to estimate lesion volume with the Cavalieri method at a grid spacing of 0.1 mm [18].
Immuno uorescence imaging was performed on 20 mm coronal brain sections at around -1.70 mm from bregma using standard immunostaining protocol, as described previously [22,28]. Brie y, sections were blocked with 5% goat or guineapig serum containing 0.3% Triton X-100. After incubation with primary and secondary antibodies (see Table 1), sections were counterstained with 4',6-diamidino-2-phenylinodole (DAPI, Sigma-Aldrich, Cat# MBD0015) and mounted with glass coverslips using an anti-fade Hydromount solution (National Diagnostics, Cat# HS106100ML). Images (n = 6 sections/location/mouse for 5-6 mice/group) were acquired using a uorescent Nikon Ti-E inverted microscope, at 20X (CFI Plan APO VC 20X NA 0.75 WD 1 mm) magni cation and the background of each image was subtracted using background ROI [30,31]. The number of NeuN + , LC3 + , p62 + , and Iba-1 + cells was normalized to the total imaged area (mm 2 ) using the NIH Image J software (1.43; NIH). Myelin images were acquired using a Leica TCS SP5 II Tunable Spectral Confocal microscope system (Leica Microsystems, Bannockburn, IL). All IHC staining experiments were performed with appropriate positive control tissue, as well as primary/secondary only negative controls.
The following commercially available dyes (see Table 1 Data were acquired on a BD LSRFortessa cytometer using FACSDiva 6.0 (BD Biosciences) and analyzed using FlowJo (Treestar Inc.). At least 5-10 million events were collected for each sample. Countbright Absolute Counting Beads (ThermoFisher, Cat# C36950) were used to estimate cell counts per the manufacturer's instructions. Data were expressed as counts/hemisphere. Leukocytes were rst gated using a splenocyte reference (SSC-A vs FSC-A). Singlets were gated (FSC-H vs FSC-W), and live cells were gated based on Zombie Aqua exclusion (SSC-A vs Zombie Aqua-Bv510). Resident microglia were identi ed as the CD45 int CD11b + Ly6C − population, whereas peripheral leukocytes were identi ed as CD45 hi CD11b + myeloid cells or CD45 hi CD11blymphocytes. Cell type-matched uorescence minus one (FMO) controls were used to determine the positivity of each antibody and indicator dye [32].

Statistical analysis
All quantitative data are plotted as mean + standard error of mean. Statistical analysis was performed with Sigmaplot Version 12 (Systat software) or Graphpad Prism Version 4 for Windows (Graphpad Software, Inc). Kaplan-Meier survival curves were analyzed using the log-rank Mantel-Cox test. When comparing between two individual samples/groups, statistical signi cance was evaluated with the Student's unpaired t-tests (detailed in gure legends). Comparisons within each surgery group were analyzed with 2-way ANOVA group analysis followed by multiple comparisons Dunnett's or Tukey's posthoc test. For non-parametric data, the Mann Whitney test was used. A p value of ≤0.05 was considered as statistically signi cant.

Results
Old age decreases long-term survival and gait performance after TBI To determine whether old age impacts long-term recovery after TBI we assessed young and old male C57BL/6 mice using a battery of behavioral tests over the course of a 12-week period (Fig. 1a). Old mice showed signi cantly higher mortality beginning at 9 weeks as determined by the Mantel-Cox test (Fig.   1b). Body weights gradually increased in young mice and decreased in old mice in a normal age-related fashion, however, TBI-induced weight loss was evident early during the rst week in the older group (Fig.  1c). To rule out any confounding effects in motor performance caused by differences in pain sensitivity we subjected mice to the hot plate test. Both age groups displayed similar increases in pain sensitivity to heat stimuli between 3-6 weeks after injury (Fig. 1d).
Motor function was measured using the open eld (OF) for spontaneous locomotor activity, Catwalk for gait analysis, rotarod for balance and stamina, and grip testing for forelimb strength. Baseline (i.e., pre-TBI) motor function was signi cantly reduced with age in all tests. A signi cant group effect of injury was seen in rotarod performance, especially in young mice which showed decreased latency to fall at every timepoint after TBI (Fig. 1e). Forelimb grip strength was signi cantly reduced during the rst week for both age groups but rebounded back to baseline or higher by 3 weeks (Fig. 1f). Together these data show that the effects of normal aging are more pronounced than TBI. Motor recovery was evident in both age groups for OF behavior and grip strength, while long-term de cits were seen in gait dynamics and rotarod performance.
OF results showed an injury effect in the total distance traveled and mean speed, marked by an early increase in each metric during the rst week ( Fig. 1g-i). This nding could not be attributed to increased anxiety-like behavior as no change in inner/outer zone preference was seen for either age (Fig. 1j). Gait analysis showed signi cant and lasting changes in all four limbs for all mice following injury ( Fig. S1) but given the unilateral nature of the injury we focused on the contralateral, or right-side limb dynamics. Our results revealed long-term de cits in step sequence regularity and pattern for both age groups at 12 weeks post-injury (Fig. 1k, l). The duration of contact with the walkway of the right forepaw, or stand time, was chronically increased in old, but not young TBI mice (Fig. 1m). Stride length was impeded early at 3 weeks in old mice and persisted up to 12 weeks (Fig. 1n). No injury effect was seen in the swing time for the right forelimb (Fig. 1o); however, swing speed was signi cantly increased at 12 weeks in old mice (Fig. 1p). Chronic alterations in body speed variation were evident in both age groups (Fig. 1q), whereas only young mice exhibited signi cantly decreased body speed late after TBI (Fig. 1r).
Old age exacerbates long-term cognitive dysfunction after TBI To investigate the role of age on cognitive decline following TBI we examined mice using Y-maze and novel object recognition (NOR) testing, to assess short-term spatial working memory and long-term recognition memory, respectively. Signi cant group effects of age and injury were seen in the percentage of spontaneous alternations in the Y-maze test (Fig. 2a); however, no difference was seen between pre-TBI and 12 weeks post-injury for either group. An injury-induced decrease in the number of arm entries was also seen in both age groups, with aged mice showing relatively fewer arm entries in general (Fig.  2b). NOR testing showed no clear preference in time spent between left and right objects during the sample phase, except for the young sham group (Fig. 2c). During the choice phase, a statistical preference was found for the novel object in all groups, except for the old TBI group which displayed late impairment in recognition memory (Fig. 2d). These ndings indicate that old age is associated with a more pronounced loss of recognition rather than spatial memory during the chronic phase of TBI.
To assess depressive-like phenotype we subjected mice to novelty suppressed feeding (NSF) and social interaction testing late after TBI. Given the high degree of mortality seen in the aged TBI group, we chose tests that were less stressful than those that typically measure behavioral despair. In the NSF test, a signi cant effect of injury, but not age, was seen in the ratio of time required for food-deprived mice to travel to the center zone of a novel versus familiar environment to obtain food (Fig. 2e). We then evaluated social behavior in chronically injured mice. No preference in time spent between left and right cups was seen in any group (Fig. 2f). However, when one cup was replaced with a stranger mouse, all groups spent signi cantly more time exploring the mouse (Fig. 2g). Moreover, when the other empty cup was replaced with a second stranger mouse, all groups showed a preference for the novel mouse (Fig.  2h). While an effect of age was found in time spent exploring the second stranger mouse (purple bars) in sham mice by Tukey's post-hoc test for multiple comparisons (p<0.05), no statistical differences were seen between the injury groups. These data suggest that despite modest differences at baseline, depression-like behavior is prevalent in all age groups following TBI.
Old age increases lesion volume, white matter loss, and microglial activation at 16 weeks following TBI We determined the effect of old age on chronic neurodegeneration after TBI. Quanti cation of cresyl violet stained tissue sections revealed signi cantly larger lesion volumes in old mice compared to their younger counterparts ( Fig. 3a-b). IHC con rmed there were fewer NeuN-positive neurons in the ipsilateral cortex of old mice at 16 weeks post-injury ( Fig. 3c-d). Concomitantly, the number of Iba1-positive microglia was found to be statistically increased in the old TBI cortex, but not young, as determined by two-way ANOVA with Tukey's multiple comparison's test ( Fig. 3e-f). Age-related pathological abnormalities were also seen in myelination of the medial corpus collosum as evidenced by Fluoromyelin stain intensity ( Fig. 3g-h). Together, we demonstrate a positive association between microglia/macrophage number, severity of neuronal degeneration, and white matter degradation after chronic TBI, which is exacerbated in older mice.
Gene pathways associated with microglia activation, phagocytosis, and autophagy are elevated with age in both the injured cortex and hippocampus To address age-related differences in the transcriptional response to TBI during the chronic phase, ipsilateral cortical and hippocampal tissues were sampled from young and old injured mice at 16 weeks post-injury. The Neuroin ammation panel tested a total of 757 genes within three themes of Immunity & In ammation, Neurobiology & Neuropathology and Metabolism & Stress [28,29]. PCA of all normalized gene counts in the cortex and hippocampus revealed a distinct separation of samples into individual groups across the rst two principal coordinates ( Fig. 4a-b). In the cortex, the rst component accounted for the majority of the variation (20.2%) across samples and separated the groups by injury, while the second principal coordinate (12.9%) separated the groups by age (Fig. 4a). To further analyze the connection between aging, TBI, and the behaviors re ective of cognitive decline and depression that we observed in tested mice, hippocampal tissue was also analyzed. In the hippocampus, injury accounted for 19.2% of the variation compared to 15.3% by age (Fig. 4b). In both regions, the old TBI samples clustered further away on the x-axis compared to the young TBI and two sham groups, perhaps re ecting the higher number of in ammatory genes that were increased in the brain of old mice at 16 weeks post-TBI.
Four pairwise comparisons of perilesional cortices and hippocampi were performed and outlined in Fig.   S2: (1) young sham vs. old sham -(Set 1); (2) young TBI vs. young sham -(Set 2); (3) old TBI vs. old sham -(Set 3); and (4) old TBI vs. young TBI -(Set 4). TBI resulted in increased expression of genes in both young and old mice (Set 2 and 3), while age-related differences were seen in both baseline (Set 1) and TBI conditions (Set 4). At baseline, 130 genes (17.1%) were differentially expressed (DE) between young and old sham groups, indicating that age affected the homeostatic regulation of genes related to the in ammatory process in a fundamental way (Fig. S2a). In the hippocampus, 121 genes were found to be DE with age (Fig. S2b). The number genes found to be down-regulated was greater in the cortex than hippocampus for both Sets 3 and 4. Venn diagrams illustrate the number of DE genes that overlap between Sets 1, 2, 3, and 4 ( Fig. S2c-d). With further analysis we were able to identify 83 genes in the cortex that were uniquely related to age and do not overlap with any genes from Set 2 and Set 3 (Fig.  S2c). In addition, a total of 130 genes were found to be related to TBI injury effects (Fig. S2c). In the hippocampus, a total of 74 genes were downregulated and 47 genes were upregulated when comparing samples in Old Sham to Young Sham (Fig. S2b). At 16 weeks after injury, Set 2 had as many as 126 genes within the in ammation panel downregulated while only 21 genes were upregulated. Similarly, the comparison between Old TBI versus Old Sham showed a total of 110 DE genes, with 83 being downregulated in the injury group while a mere 27 genes were upregulated. As demonstrated in the adjacent Venn diagrams, we identi ed 100 age-related genes and 101 injury-related genes (Fig. S2d).
An initial analysis of the DE genes in each set showed several genes with large fold changes in the cortex, as illustrated in a volcano plot for each gene set ( Fig. 4c-j). To further investigate the effects of normal aging on transcriptional activation, we compared the percentage of genes modi ed by age (Set 1 and 4) for each pathway annotation. All pathways were affected with advanced age, but genes that facilitate oligodendrocyte function, neurons and neurotransmission, autophagy, astrocyte function, innate immune response, epigenetic regulation, and microglia function were far more likely to be activated in the cortex of old sham mice compared to young sham control. Alternatively, pathway analysis of genes related to injury (Set 2 and 3) showed that genes associated with oligodendrocyte function, angiogenesis, neurons and neurotransmission, matrix remodeling, astrocyte function, microglia function, and cellular stress were predominantly activated after injury ( Fig. 4c-d). Pathway analysis of DE genes in the cortex is shown (Fig.   5). A heatmap of DE genes implicated in each notated pathway generally show an age-related upregulation in the chronic phase of TBI ( Fig. 5a-g). The complement pathway, which has been previously implicated in age-related neurodegenerative disease and synaptic maintenance, was found to be increased both with age and injury (Fig. 5h). Genes associated with microglia function, the complement pathway, and autophagy regulation were also up-regulated with age and/or injury in the hippocampus ( Fig. 6a-c). Genes associated with epigenetic regulation were also increased with age and injury (Fig. 6d). A heatmap of DE genes implicated in astrocyte function, oligodendrocyte function, and neurons & neurotransmission showed a more mixed expression pattern after injury ( Fig. 6e-g). More speci cally, the age-and chronic injury-related upregulation of Axl, CD14, CD68, CD74, Ctss, Clec7a, Trem2, and Tyrobp genes in both the cortex and hippocampus is consistent with that seen in the multiple models of agerelated neurodegenerative diseases, and is representative of a disease-activated microglia (DAM) signature that re ects a pathological state of the CNS [33].
As an additional measure for examining transcriptional changes, we validated several genes that showed signi cant changes via real-time quantitative PCR ( Fig. S3 and S4). For example, the gene Clec7a that encodes a glycoprotein linked to the modulation of the natural killer gene complex region and connected to the innate immune response functions. Real-time qPCR validation of this gene revealed signi cantly higher expression levels of Clec7a in the cerebral cortex of Old TBI mice compared to the Old Sham group, but barely any change in the Young TBI versus Young Sham group. Another gene, ncf1 (neutrophil cytosolic factor 1), encodes a subunit of neutrophil NADPH oxidase, which is activated to produce superoxide anions. Taken together, our results show a diverse transcriptional response of in ammatory genes in old mice compared to young mice after TBI, with several genes showing differential expression between the Old TBI and Young TBI groups.
To better illustrate the relationship between neuropathology and expression of autophagic markers, we con rmed immuno-reactivity of LC3 and p62 at the protein level in the perilesional cortex ( Fig. 7a-d).
Signi cant effects of age-and TBI were seen in both the number of LC3-positive and p62-positive cells at 16 weeks post-injury, consistent with our Nanostring ndings. In sum, these data suggest autophagic changes in the aged and injured brain may precipitate chronic microglial activation and neuronal degeneration.
Old age increases in ltration of lymphocytes and exaggerates microglial responses to TBI Next, we investigated the cellular in ammatory response at early (48 h) and late (16 weeks) timepoints after TBI. Brain leukocyte identi cation and characterization was performed using ow cytometry. Old age was associated with fewer microglia (CD45 int CD11b + Ly6C -) and less robust microglial accumulation in terms of absolute number (Fig. 8a-b). Normal age-related reductions in microglia number have been reported previously by our group and others and likely re ects age-related dystrophy and proliferative senescence [34]. The number of in ltrating myeloid cells (CD45 hi CD11b + ), however, was comparable between young and old groups and showed two-way ANOVA group effects of injury and age, including a signi cant interaction between them (Fig. 8c). Delayed in ltration and accumulation of putative lymphocyte populations (CD45 hi CD11b -) was dramatically in old mice at all timepoints post-injury (Fig.   8d). These ndings highlight the profound effects of aging on the central and peripheral immune response to TBI and suggests that blood-brain barrier integrity may be uniquely compromised in the aged brain long after injury.
To address the rapidly shifting dynamics of microglia proliferation that occurs during the rst week of injury, we examined two commonly used nuclear protein markers of cellular proliferation, Ki67 and proliferating cell nuclear antigen (PCNA). Both markers were signi cantly upregulated in microglia at 48 h after TBI, but no statistical difference between injury groups was found ( Fig. 8e-f). Despite being fewer in number, microglia in old sham mice showed signi cantly higher basal expression of Ki67, indicating active cell cycling. The percentage of Ki67-positive microglia remained elevated late after TBI in young but not old microglia as determined by Dunnett's test (p=0.05). These data do not indicate age-related difference in the kinetic progression of microglial proliferation at 48 h, however, during the chronic stages of injury microglia from old TBI mice showed reduced proliferative capacity relative to sham compared to those from the young TBI group.
We then examined the activation state of microglia based on light scatter properties and ex vivo functional analysis. Signi cant effects of age and injury were seen microglia size and granularity at all timepoints, as measured by forward and side scatter intensity, respectively ( Fig. S5a-b). Old microglia became markedly more granular after 48 h post-injury compared to young microglia, suggesting increased in ammatory activation. Reactive oxygen species (ROS) production is associated with oxidative stress which can exacerbate neurodegeneration [18,35]. Old microglia had dramatically higher ROS production at baseline and at all timepoints post-injury, whereas young microglia exhibited a more robust increase in ROS relative to its respective sham control (Fig. 8g). The percentage of microglia producing detectable ROS remained elevated at 16 weeks in young mice as determined by Dunnett's test. Lastly, we measured cytokine production in microglia which showed signi cant effects of both age and injury in expression level of IL1β and TNF protein (Fig. 8h-i). Both cytokines showed an age-related increase during the acute phase of TBI, which largely abated by 16 weeks (p<0.05). Taken together, our data imply that old age primes microglia to become more reactive to acute TBI, consistent with previous work. Independent of any change in cell number, basic features of microglial activation during the chronic phase of TBI appear to be more subtle on a per cell basis.
Microglial phagocytosis of neurons, expression of autophagy markers, and lipofuscin content are increased with age and injury The emergence of a DAM signature led us to surmise that age-related de cits in autophagy function underlie chronic microglial activation and dysfunction following TBI. To better understand the temporal effects of TBI on microglia function, we performed an extensive characterization of phagocytosis and autophagy informed largely by our NanoString results indicating age-related increases in these speci c pathways late after injury. Phagocytosis was measured by intracellular detection of neuronal antigens, myelin content, and the lysosomal/endosomal membrane marker CD68. Signi cant group effects of both age and injury were seen in the percentage of NeuN-positive microglia (Fig. 9a). The frequency of NeuNpositive microglia increased sharply at 48 h in both age groups and remained high at 16 weeks in old mice. Similar changes were seen in intracellular myelin content; however, while levels trended higher in old mice at 16 weeks (p=0.06) it did not meet statistical signi cance in the two-way ANOVA multiple comparisons test (Fig. 9b). Intracellular detection of vesicular glutamate transporter 1 (Vglut1), a synaptic protein in neurons, was also increased acutely in microglia after TBI (Fig. 9c). A group effect of age was found, indicating a higher presence of this synaptic marker in old microglia. The phagocytosis marker CD68 was signi cantly higher in old microglia both at baseline and at 16 weeks but was elevated to a similar extent at 48 h in both age groups (Fig. 9d). These ndings suggest that microglial phagocytosis of dead or dying neurons is robust in the early stages of TBI, and modestly increased with both age and time post-injury, consistent with our gene expression data.
Lysosome and autophagosome content in microglia were signi cantly increased with both age and injury and remained elevated for up to 16 weeks ( Fig. 9e-g). Protein expression of Sqstm1/p62 and Lamp2 were increased at 48 h, but no effect of age was seen (Fig. 9h-i). The ubiquitin-proteasome system degrades intracellular proteins into peptide fragments that can be presented by major histocompatibility complex (MHC) class I molecules. Signi cant group effects of age and injury were found in surface expression of MHCI with the highest expression seen at 48 h (Fig. 9j). Defects in the autophagy pathway can result in the cytosolic and lysosomal accumulation of an auto uorescent amalgam of oxidized lipids, proteins, and metals known as lipofuscin. Group effects of age and injury were seen in intracellular lipid content, iron levels, peptide/protein aggregation, and auto uorescence (Fig. S5c-f). Neutral lipid and iron content increased sharply at 48 h, consistent with the increase in phagocytic activity. Protein aggregation and auto uorescence were largely unperturbed at 48 h, however, both were found to be elevated above sham levels at 16 weeks in young microglia as determined by Dunnett's test (p<0.05 and p<0.05, respectively). Collectively, these data imply that phagocytosis and autophagic function are exacerbated with age and TBI, consistent with our DAM gene signature. Although these processes are more active during the acute stages of injury, many markers, including the accumulation of lipofuscin material, show a sustained increase during the chronic phase.
Age-related changes in microglial histone acetylation patterns, metabolic activity levels, and expression of senescence markers are altered after TBI Given the widespread effects of aging on microglial function, we also evaluated changes in epigenetic regulation, metabolic activity, and senescence-like phenotype following TBI. Microglia activation is associated with increased histone deacetylase (HDAC) and sirtuin (SIRT) activity resulting in hypoacetylation of histones and non-histone proteins and increased expression of in ammatory markers [36]. Our results show an age-related loss of histone 3 (H3) within the general microglia population, however, H3 expression levels were dramatically increased at 48 h in both age groups (Fig. 10a). The percentage of microglia expressing acetylated lysine residues was signi cantly reduced with age and injury indirectly suggesting increased HDAC and SIRT activity (Fig. 10b). To further probe the impact of TBI on histone modi cation we examined acetylation patterns on H3. Signi cant effects of age and injury, or injury alone, were seen in H3 hypoacetylation of the lysine residue at N-terminal positions 9, 18, 27, and 36 ( Fig. 10c-f). These data suggest that aging and TBI result in the removal of acetyl groups from H3 which is known to increase chromatin condensation and decrease gene transcription.
TBI acutely increased mitochondrial activity, glucose uptake, and ATP production in microglia (Fig. 11ac). At 16 weeks post-injury, the cellular demand for glucose remained higher than sham control in young mice as determined by Dunnett's test (p<0.01). These ndings suggest that TBI causes a metabolic crisis in microglia that is associated with a chronic increase in glucose uptake in young mice.
We have previously demonstrated that a small population of microglia upregulate markers of cellular senescence in an age-and TBI-dependent manner [37]. However, the sustained presence of a senescentlike phenotype has yet to be reported in microglia at chronic timepoints after injury. Expression of the DNA damage sensor, p-H2AX, showed a basal elevation with age (Fig. 11d). Interestingly, p-H2AX levels were signi cantly decreased at 48 h, likely due to a stress-mediated increase in DNA repair mechanisms.
Although expression was further increased from 48 h to 16 weeks, levels did not surpass that seen in sham controls. Expression of the cell cycle arrest markers p16 and p21 were increased with normal aging ( Fig. 11e-f). The percent of p16-positive microglia trended higher in young TBI mice at 16 weeks compared to sham, but signi cance was not achieved by two-way ANOVA multiple comparison test. It should be noted that the absolute number of p16-positive microglia may be signi cantly increased in the absence of a statistical change in percentage. In contrast, a robust increase in the relative expression of p21 was seen at 48 h in both age groups, however, levels were higher in old microglia at both timepoints post-injury (Fig. 11f). Together these results indicate that aging interacts with TBI to modify key aspects of microglial homeostasis, including post-translational epigenetic regulation, metabolic function, and development of a senescence-like phenotype.
Trehalose treatment enhances functional recovery in motor and cognitive tasks and decreases depressive-like behavior Old mice were administered trehalose or sucrose control in their drinking water beginning on d1 for up to 8 weeks after surgery. A battery of behavioral tests was performed to monitor changes in motor performance, cognitive function, and anxiety/depression (Fig. 12a). Trehalose treatment resulted in a delayed decrease in body weight (Fig. 12b). No statistical de cits were seen in rotarod performance at any timepoint in mice treated with trehalose, whereas mice treated with sucrose control exhibited lasting impairment for up to 8 weeks after TBI (Fig. 12c). OF activity was then assessed (Fig. 12d). Interestingly, sucrose-treated mice traveled signi cant shorter distances in the OF during the rst week of injury (Fig.   12e). This was also accompanied by a decrease in speed, not seen in trehalose-treated mice (Fig. 12f). A statistical increase in time spent in the center zone of the OF late after injury was recorded in mice administered trehalose, suggesting that enhancing autophagy during TBI may result in delayed anxiolytic effects (Fig. 12g). These data suggest continued administration of trehalose may attenuate motor de cits in aged mice after TBI.
To address the effect of trehalose on long-term cognitive outcomes in old mice, we evaluated mice using the Y-maze and NOR tests. While both treatment groups showed a temporary reduction in the number of arm entries early after TBI, no therapeutic effects were seen (Fig. 12h-i). However, trehalose-treated mice did show improvement in recognition memory at 8 weeks post-TBI compared to sucrose control ( Fig. 12jk). Lastly, NSF and social interaction tests were employed to evaluate anxiety/depressive-like phenotype.
Trehalose-treated mice displayed shorter latency times to reach food at the center of the OF compared to sucrose control, suggesting less depression-related behavior, consistent with our OF ndings (Fig. 12l).
Finally, social interaction testing showed that trehalose increased the amount of time spent interacting with a novel stranger mouse in the social discrimination phase (i.e., sequence 3) of the test (Fig. 12m-o).
Together these ndings show that continued trehalose treatment can preserve recognition memory and reduce anxiety/depression in old mice late after injury.
Trehalose treatment reduces chronic brain in ammation and phagocytosis of neuronal synapses by modulating multiple steps of the autophagy pathway in microglia To better understand the impact of trehalose on chronic microglia activation and neuroin ammation in old mice after TBI we examined autophagic function at 9 weeks post-injury. Trehalose treatment caused a clear reduction in the number of proliferated microglia (CD45 int CD11b + Ly6C -) present in the ipsilateral hemisphere and signi cantly prevented the delayed recruitment or accumulation of peripherally-derived myeloid (CD45 hi CD11b + ) and lymphocytes (CD45 hi CD11b -) (Fig. 13a-d). Next, we examined the effect of trehalose on phagocytosis. A signi cant injury-related increase was found in microglial expression of CD68, but no effect of treatment was seen (Fig. 13e). Lipid content trended higher in sucrose-treated mice (p=0.05), while trehalose treatment caused no change (Fig. 13f). The percentage of microglia that engulfed neuronal Vglut1 and the relative level of intracellular NeuN protein were both signi cantly lower in trehalose-treated mice (Fig. 13g-h).
A comprehensive analysis of autophagy-related markers showed no effect of treatment or injury at 8 weeks as measured by LysoTracker dye, lamp1, lamp2, and Sqstm1/p62 (Fig. S6a-d). Although increased formation of LC3II-positive autophagosomes was found in sucrose-but not trehalose-treated mice, this difference was not statistically signi cant (p=0.05; Fig. 14a). A group effect of TBI was seen in ATG5 and ATG7 expression in microglia; trehalose treatment prevented a signi cant reduction in ATG7 (Fig. 14b-c).
Examination of lysosomal enzymatic activity in microglia revealed a robust increase in cathepsin D and lysosomal lipase activity in mice treated with trehalose ( Fig. 14d-e). There was a trend towards increased ubiquitin expression in microglia in sucrose-treated mice, but changes did not reach signi cance (p=0.05; Fig. 14f). Whereas antigen presentation by MHCI was more augmented after trehalose treatment, it did not alter injury-induced increases in intracellular protein aggregation (Fig. 14g-h). Lastly, trehalosemediated protection was associated with attenuated lysine acetylation in microglia (Fig. S6e). Taken together, our results indicate that long-term trehalose treatment reduces chronic brain in ammation via modulation of several key proteins involved in the autophagy pathway, including the upstream regulation of phagocytosis.

Discussion
We show that old age augments the expression of the disease-associated microglia signature during the chronic phase of TBI. This transcriptional and functional response is characterized by the downregulation of homeostatic pathways and upregulation of genes involved in lysosomal and phagocytic pathways.
Neuronal/myelin engulfment, autophagic processes, and in ammatory activation were exacerbated with advanced age in microglia at both acute and chronic times after injury. Continuous treatment with orally administered trehalose following brain injury improved long-term neurological recovery in old mice, which was associated with chronically reduced phagocytic activity and increased lysosomal function. Our ndings suggest that age-related dysregulation of autophagy plays a role in chronic microglial activation, neuroin ammation, and recovery following brain injury.
An extensive longitudinal assessment of functional recovery in aged mice following experimental TBI showed that older mice have similar recovery patterns for many, but not all behavioral tests compared to younger mice. Indeed, overall performance was more profoundly affected by normal aging than TBI.
Thus, it can be problematic to differentiate effects of aging from those of injury due to differences in baseline performance. Thus, we have expressed the behavioral data as absolute changes rather than as percent changes from baseline to better illustrate recovery dynamics within groups. One conclusion from our ndings is that older mice show some behavioral restoration late after TBI; thus, the same reparative processes that may be therapeutically modulated in young mice may also be targeted in older animals.
Using the NanoString Neuroin ammation Panel, TBI caused transcriptional activation of more than 100 additional in ammation-related genes in the cortex, and 40 fewer genes in the hippocampus of injured aged mice versus sham control as compared to changes in the injured young group. Based on the incremental increases in expression of DAM genes observed with both aging and chronic injury, these ndings are consistent with the conclusion that TBI accelerates normal age-related expression of this in ammatory signature which may contribute to accelerated neurological impairment. Moreover, this work suggests that many prognostic biomarkers of TBI outcome, including the activation of speci c gene networks, may need to be reassessed in older patients, given the normal age-related increases in gene expression levels and protein concentrations.
Age and injury-related increases in the expression of genes associated with the complement pathway is consistent with increased innate immune activation and supports a role for phagocytosis in the chronic brain injury setting. Complement activation products opsonize synaptic material on neurons for phagocytic removal, which during development, functions to prune excess synapses to re ne neural circuitry [38]. This process is also important for synaptic plasticity throughout life. However, an increasing number of studies have recently demonstrated that excessive complement-mediated synaptic engulfment is a pathological process in the context of neurodegenerative disease, including TBI [39]. This would be consistent with our ex vivo phagocytosis data showing increased microglial engulfment of neuronal/synaptic antigens and decreased numbers of viable neurons and synaptic plasticity late after TBI [17]. Genes associated with astrocyte function were also relatively more upregulated in the aged brain late after TBI. Astrocytes can induce the expression of C1q in neurons via TGF-β [40]. C1q is involved in the pathological pathway of several neurodegenerative diseases due to its role in abnormal protein aggregate clearance, astrocyte reactivation, binding and activation of microglia, or in ammatory responses [41]. It remains to be seen whether complement gene expression is elevated in astrocytes during the chronic phase of TBI, and if C1q, C3, and C4a factors are localized at synapses and neurons in affected brain regions.
Brain aging leads to decline in autophagy e ciency, accumulation of lipofuscin, and increases in in ammatory activity including ROS production. Moreover, previous work has found that cellular and molecular markers of in ammation in the brain are acutely altered with age after TBI [7,8,37,42]. The present study has now captured similar changes in aged mice during the chronic phase of TBI using a combination of ex vivo cellular assays, transcriptional gene pro ling, and histological approaches. At the cellular level, the acute phase of injury was marked by an increase in in ltrating myeloid cells and microglial proliferation, whereas the chronic phase is highlighted by delayed lymphocyte recruitment. The leukocyte dynamics, however, were altered with age. Older mice showed signi cantly fewer microglia in general, either due cell senescence and/or death or the cell isolation procedure. Nonetheless, similar changes have been seen in other CNS cell types with age, including microglia [43]. Lymphocyte in ltration was exacerbated in old mice at all timepoints. Similar changes have been reported in aged models of ischemic stroke [44,45], suggesting this may be a conserved age-related hallmark of brain injury. However, lymphocyte numbers were far exceeded by the resident microglial population. Although we made efforts to characterize the age-related temporal kinetics of microglial proliferation after TBI, two timepoints alone cannot accurately capture these dynamics, especially during the fast-evolving acute period. For example, in a previous study we noted relatively greater microglial proliferation at 72h postinjury in aged mice compared to young controls [37]. The fact that we did not observe this response at 48h post-injury in our study might suggest that microglial proliferation is delayed with age. This would be in line with other reports, and our own data showing similar or higher expression of cell proliferation markers at that point in time for old microglia [37]. Although numbers of brain-in ltrating myeloid cells were similar between groups at 48h, we had previously reported that the composition of these cells skewed more neutrophilic in older mice [37]. Thus, the cellular immune dynamics of the neuroin ammatory response to TBI are different with older animals. Changes in leukocyte number and composition, notwithstanding the effects of aging alone, suggest the quality and severity of immune function and in ammatory activity may be markedly different in older mice.
To ascertain whether microglial reactivity at acute and chronic stages of TBI was affected in old age we performed an extensive pro ling of cellular functions. In most measures, pro-in ammatory activity was relatively greater in older microglia after acute TBI. Of note, however, were the signi cant effects of age on phagocytosis, autophagy, and lipofuscin accumulation in microglia following TBI. In general, older microglia were more likely to engulf neurons and/or neuronal debris than their younger counterparts. Although the effects of TBI appear subtle at 16 weeks due to scaling and higher statistical stringency required for multiple comparison's testing, we have previously reported a chronic increase in phagocytosis of apoptotic neurons and myelin debris as late as 8 months post-TBI [17]. The acute phase was also characterized by a robust age-related increase in lysosome and autophagosome formation. Lipofuscin accumulation was elevated in both age groups after injury. Interestingly, intracellular protein aggregation was reduced at 48h in microglia; however, by 16 weeks post-injury protein aggregation recrudesced in both age groups, implying late impairment of autophagic processes in microglia. These age-related alterations in phagocytosis, autophagy, and in ammatory activity are consistent with the agerelated increase in the DAM gene signature, and its potential role in microglial activation changes during chronic brain injury.
We also identi ed novel age-and TBI-driven changes in epigenetic, metabolic, and senescence markers. The microglial activation, which often persists chronically following more severe head injury is consistent with an enduring change in cellular reprogramming mediated by epigenetic mechanisms, as we have suggested previously [46]. Earlier studies have reported imbalances in epigenetic marks of DNA methylation and histone acetylation after TBI [47]. Thus, it would appear histone deacetylation is a pathological feature of TBI and potential driver of microglial-mediated in ammation. The present study is among the rst to examine these changes at the cellular level in microglia. We found that TBI not only reduced global lysine acetylation levels, but also histone acetylation. Moreover, these changes were lasting, exacerbated with age, and reversed with trehalose treatment. Thus, our data are consistent with the view that epigenetic changes of histones occur early after TBI and are sustained in microglia. Microglia adopted a senescence phenotype as a function of age and injury. The interaction between age and injury found in our study highlights a potential mechanism for the exaggerated in ammation and neurological dysfunction seen in older animals [55-58].
Using aged animals, we employed a highly studied neuroprotectant and autophagic inducer as a potential therapeutic agent. Trehalose is currently in clinical trials for AD and neuronal ceroidlipofuscinoses. Importantly, it is well-tolerated, widely available, low-cost and can be orally administered without additives [59]. To assess longitudinal recovery in aged animals treated with trehalose, we performed a similar behavioral battery but with shorter injury duration to minimize long-term mortality.
Sucrose was used as the control group because it shares a similar molecular structure with trehalose. Like many control solutions, sucrose is not an inert substance and may exert its own effects on metabolism and neurological recovery which have not been taken into consideration here. Nevertheless, trehalose was found to be more protective in aged mice after TBI. Previous work has shown that young mice treated with trehalose perform better in Morris water maze, OF testing, and Y-maze [20].
Furthermore, trehalose is known to also act as a chemical chaperone [60] and may therefore have protective actions that are independent of autophagy. For example it has been reported that trehalose treatment is associated with upregulation of synaptophysin, doublecortin, and BDNF [20].
Given its ability to stabilize biomolecules and increase autophagy, trehalose has received growing attention as a potential neuroprotective agent [59,61]. However, the precise mechanisms through which trehalose acts on the autophagy pathway in microglia have not been explored. Most importantly, we found that trehalose reduces overall numbers of in ammatory immune cells in the injured brain. Whether trehalose acts directly on the brain or in the periphery to achieve this effect is not clear. Our results also showed unexpected changes in phagocytic activity late after TBI. Trehalose reduced continued phagocytosis of neurons and neuronal synapses, and attenuated lipid accumulation. This suggests that trehalose can act upstream of the autophagy pathway to potentially prevent the ingestion and accumulation of phagocytized material. Perhaps surprisingly, we did not see changes in lysosome formation or Sqstm1/p62 expression in microglia following trehalose treatment. Rather, the effects of trehalose on autophagy are likely more relevant during the acute phase of injury when phagocytosis and lysosomal degradation processes are sharply elevated [62]. Nonetheless, chronic increases in lysosomal enzyme activity were seen in trehalose-treated mice indicative of improved autophagy function. Further, trehalose prevented injury-induced increases in autophagosome formation and decreases in ATG7 expression late after TBI. Taken together, we identi ed several potential steps in the autophagy pathway where trehalose may exert important effects on microglia that are associated with better functional outcome following TBI

Conclusion
Our histological, cellular, and molecular ndings provided complementary evidence for age-related, chronic dysregulation of microglial function and autophagic processes. We show that therapeutically modulating autophagy using trehalose can reduce chronic brain in ammation in aged mice and enhance long term recovery.      ****p < 0.0001, ***p < 0.001, ** p < 0.01, * p < 0.05.

Figure 2
Effects of age on long-term cognition, depression, and social behavior. sample phase of the NOR task, no differences between groups were seen in exploration time between left and right-side objects.  demonstrating speci c genes with log2 (fold change) larger than 1 and statistical signi cance of log10(P) higher than 1. n=6 mice/group.

Figure 5
Age signi cantly alters neuroin ammatory pro le of the injured cerebral cortex at the transcriptome level.
(a-g) Heatmap of genes related to microglia function (a), autophagy (b), lipid metabolism (c), oligodendrocyte function (d), astrocyte function (e), epigenetic regulation (f), and neurons and neurotransmission (g) that were altered in differential expression analysis between Old TBI and Young Page 37/46 TBI, i.e., genes only altered in Set 4. Color coding was based on z-score scaling. (h) Bar graph of genes within the complement pathway that showed differential expression. n=6 mice/group.

Figure 6
Age-related transcriptomic changes in neuroin ammatory genes of the hippocampus at 16w post-injury.
(a-c) Bar graph of genes related to microglia function (a), the complement pathway (b), and autophagy (c) that showed differential expression in pairwise comparisons between Old TBI and Young TBI. (d-g) Heatmap of differentially expressed genes related to epigenetic regulation (d), astrocyte function (e), oligodendrocyte function (f), and neurons and neurotransmission (g) when comparing between Old TBI and Young TBI, i.e., genes only altered in Set 4. Color coding was based on z-score scaling. n=6 mice/group.  Microglial proliferation as measured using cell cycle markers. A group effect of injury was seen in Ki67 (e) and PCNA (f) expression, driven largely by the increase at 48h after TBI. Reactive oxygen species (ROS) and oxidative stress were measured using DCF. Group effects of age and injury were seen in microglial ROS levels (g). Pro-in ammatory cytokine production was assessed. The relative expression level of IL-1β (h) and TNF (i) in microglia is shown. For all histograms, gray = FMO control, blue = young, red = old, Sham = no outline/no ll, 48h TBI = bold outline/no ll, and 16w TBI = bold outline/bold ll. N=5-7/group. Data were analyzed using 2-way ANOVA group analysis with Tukey's test for multiple comparisons. ***p < 0.001, **p < 0.01, *p < 0.05.

Figure 9
Microglial phagocytosis of neurons, expression of autophagy markers, and lipofuscin content is chronically increased with age and injury. Phagocytosis was assessed by intracellular detection of neuronal and myelin antigens. (a) The percentage of NeuN-positive microglia was signi cantly increased with age and injury. The mean uorescence intensity of Fluoromyelin Red staining was acutely increased in microglia from both age groups (b). The presence of vesicular glutamate transporter 1 (vGlut1) (c) and the phagosome marker CD68 (d) were increased in microglia with both old age and TBI. Representative histograms illustrate the relative abundance of lysosomes and LC3II-positive autophagosomes in microglia as measured by LysoTracker (e) and Cyto-ID Autophagosome dyes (f). For all histograms, gray = FMO control, blue = young, red = old, Sham = no outline/no ll, 48h TBI = bold outline/no ll, and 12w TBI = bold outline/bold ll. The mean uorescence intensity of Lamp1 (g), Sqstm1/p62 (h), Lamp2 (i), and MHCI (j) for microglia are shown. N=5-7/group. Data were analyzed using 2-way ANOVA group analysis with Tukey's test for multiple comparisons. ****p < 0.0001, **p < 0.01, *p < 0.05.  For all histograms, gray = FMO control, blue = young, red = old, Sham = no outline/no ll, 48h TBI = bold outline/no ll, and 16w TBI = bold outline/bold ll. N=5-7/group. Data were analyzed using 2-way ANOVA group analysis with Tukey's test for multiple comparisons. ****p < 0.0001, ***p < 0.001, *p < 0.05. be observed in the TBI Sucrose group during the choice phase of NOR experiment. (i) The latency to reach food in the center of a novel arena was recorded for the NSF test. The Trehalose TBI group showed signi cantly lower Open Field versus Home cage ratio when compared to the Sucrose TBI group. No difference was found in latency time in the home cage, con rming that the novel environment was a source of anxiety for TBI mice. (m-o) Graphs depicting the preference index of each sequence for the social recognition experiment. In sequence 3, The TBI Sucrose group failed to show preference for the second, novel stranger mouse, which suggests neurological de cits in social interaction (o). N=8 (Sham Sucrose), 8 (Sham Trehalose), 11 (TBI Sucrose), and 12 (TBI Trehalose). 2-way ANOVA followed by Tukey's post-hoc test. ****p < 0.0001, ** p < 0.01, * p < 0.05.