Paracetamol (acetaminophen) as potential rst-choice analgesic in post-operative cognitive decline (POCD): Behavioral and molecular evidence in a middle-aged rat model

Post-operative cognitive dysfunction (POCD) is a debilitating clinical phenomenon in elderly patients. Management of pain in elderly is complicated because analgesic opiates elicit major side effects. In contrast, paracetamol (acetaminophen) has shown analgesic ecacy, no impact on cognition, and its side effects are well tolerated. We investigated the ecacy of paracetamol, compared to the opioid analgesic buprenorphine, in a model of POCD by investigating cognitive decline, allodynia, peripheral and hippocampal cytokines levels, and hippocampal microtubule dynamics as a key modulator of synaptic plasticity. A POCD model was developed in middle-aged (MA) rats by inducing a tibia fracture via orthopaedic surgery. Control MA rats did not undergo any surgery and only received isourane anaesthesia. We demonstrated that cognitive decline and increased allodynia following surgery was prevented in paracetamol-treated animals, but not in animals which were exposed to anesthesia alone or underwent the surgery and received buprenorphine. Behavioral alterations were associated with different peripheral cytokine changes between buprenorphine and paracetamol treated animals. Buprenorphine showed no central effects, while paracetamol showed modulatory effects on hippocampal cytokines and markers of microtubule dynamics which were suggestive of neuroprotection. Our data provide the rst experimental evidence corroborating the use of paracetamol as rst-choice analgesic in POCD. reported increased plasma and hippocampal levels of IL-6 and IL-1b at 6h and 24h after and a single injection of Bup (0.1 mg/kg, s.c.) as analgesic in YG adult mice 2,3 , but not at 2hr or 72hr after surgery 2 . Our study appears to conrm that plasma increase in IL-6 and IL-1b is also not a sustained event in an MA rat model of orthopaedic POCD, since it is not detectable at 8 days after surgery, with the exception of plasma IL-6 levels that were signicantly increased in animals receiving a daily high dose of Bup. This latter observation is consistent with work showing increased plasma IL-6 levels following isourane anaesthesia and Bup administration before mg/kg, s.c.) and after (0.05 mg/kg, s.c.) a scald-burn procedure 31 . Other authors observed increased plasma and synovial IL-6 levels after knee joint injury in mice injected with Bup (0.1 mg/kg, s.c.) followed by administration of the drug in drinking water for another 24hr 32 . The Bup-induced increase in plasma IL-6 levels might be linked to potential adverse effects of Bup as shown by our DNMTS task data.


Introduction
Post-operative cognitive dysfunction (POCD) is a recognized clinical phenomena de ned as a new cognitive impairment arising after anaesthesia and surgery with a higher prevalence in elderly patients 1 .
The available data from animal research suggest that POCD has a multifactorial pathogenesis resulting from a combination of anaesthesia and surgery effects on the systemic immune system 2 , neuroin ammation 3,4 , and synaptic plasticity 5,6 . Currently, no treatment is available to either prevent or rapidly treat POCD symptomatology, despite global population aging and extensive new developments in health care which both imply increasing incidences of surgery in older patients. The most common POCD symptomatology is memory impairment, with patients showing impaired performance on cognitive tasks 7 . POCD has been documented in 41.4% of elderly patients (aged ≥60) who had undergone any major surgery 8 . It is also shown that at 3 months after surgery, POCD can still be present in 12.7% of elderly patients 8 .
Orthopaedic surgery is highly associated with POCD 8 and this category of surgery is well known to be accompanied by signi cant acute post-operative pain 9 . Management of pain in aged patients with cognitive decline has to be approached with caution since the use of analgesic opiates, for example, elicits side effects on cognition and bowel function and may precipitate delirium 10 . In contrast, paracetamol (acetaminophen; Par) has shown analgesic e cacy, no impact on cognition and well tolerated side effects 10 . Par mechanism of action is still not fully understood, the drug lacks peripheral anti-in ammatory properties and it passes the blood-brain barrier to be homogenously distributed throughout the CNS even at low doses [11][12][13] . Experimental evidence suggest that the analgesic effects of Par might be due inhibition of Cyclooxygenase (COX) in the brain 14 ; and indirect activation of cannabinoid CB1 receptor 15,16 . Furthermore, growing evidence shows that Par have neuroprotective effects both in vitro 17 and in vivo 18 .
The current study examined the e cacy of Par (75 mg/kg and 150 mg/kg; i.p.) in rescuing behavioural and molecular alterations induced by a model of orthopaedic surgery in middle-aged (MA) rats compared to iso urane anaesthesia and the opioid analgesic buprenorphine (Bup; 0.05 mg/kg and 0.1 mg/kg; s.c.). The delayed non-match-to-sample (DNMTS) operant task was used to measure shortterm working memory and exibly modulating behaviour in rats through time (i.e. pre-and post-surgery) following pharmacological treatments. Pain was measured using the cold plate test on day 1-5 post surgery in order to evaluate thermal allodynia. Additionally, a panel of cytokines was analysed in plasma and hippocampus in line with the neuroin ammatory hypothesis of POCD 3 . Further analysis was conducted to explore neuroprotective effects of paracetamol on microtubule dynamics, a key modulator of synaptic plasticity involved in learning and memory 19 . Thus, a-tubulin PTMs related to the C-terminal detyrosination/tyrosination cycle, such as tyrosination (Tyr-Tub: associated with dynamic microtubules), detyrosination (Glu-Tub: associated with stable microtubules), deglutamylation (D2-Tub: neuronal speci c) were analysed in the hippocampus.

Results
Validation and identi cation of age-related memory impairments in the DNMTS assay To corroborate previous ndings, we rst conducted a pilot study to demonstrate the natural delayinduced cognitive impairment in MA rats compared to young (YG) controls. MA rats were impaired in the DNMTS paradigm, compared to YG controls at each test point; 48hr, 72hr and Day 6 of testing (Fig. 1A, C and E respectively; p<0.05 for each). The delayed-induced group differences in the DNMTS task were analysed by two-way ANOVA for repeated measures. ANOVA analysis yielded a signi cant main effect of age [F (1,8) =7.6; p=0.02] and delay [F (5,40) =8.2, p<0.0001] and no signi cant interaction at 48hr test point.
Fisher's LSD analysis revealed that the MA group was signi cantly impaired at the following delay time bins; 6-10sec (p<0.05), 11-15sec (p<0.05) and 21-25sec (P<0.05) (Fig 1. B Paracetamol has a signi cant impact on the extent of POCD Following validation of in the DNMTS task in MA rats, we compared the effects of the post-operative analgesics treatment with Par or Bup on performance in the DNMTS task. Par was used at low (75mg/kg; i.p.) and high (150mg/kg; i.p.) doses and compared to Bup at low (0.05mg/kg; s.c.) and high (0.1mg/kg; s.c.) doses. First testing in the DNMTS task resumed 48hr following surgery/control procedures. All drugs were administered daily at 23hr before test. Animals exposed to anaesthetic alone or animals which underwent surgery and received either low or high Bup were impaired in total % correct response in the DNMTS task ( Fig. 2A, p<0.05, p<0.01, p<0.05, respectively). Further, surgery model groups treated with low or high Bup were signi cantly impaired in the DNMTS compared to the surgery group treated with the Par doses ( Fig. 2 A, p<0.01 for both). Analysis by two-way ANOVA for repeated measures revealed a signi cant effect of analgesia treatment [F (5,47) =3.71; p=0.006] and delay [F (5, 235) = 31.3; p<0.0001] with no signi cant interaction. Fisher's LSD analysis further showed that the surgical group treated with Bup at low or high dose were signi cantly impaired at both long and short time bins compared to MA controls (Fig. 2B). In contrast, the Par-treated surgery groups were not impaired, compared to the MA Controls, at any delay point (Fig. 2C). No group differences were observed at the 72hr test point (Fig. 2D, E and F). When animals were tested at 7 days post-surgery/control procedures, only the surgery groups treated with Bup were still impaired in overall total performance in the DNMTS task compared to MA Controls (Fig. 2G 2H) at 11-15sec delays (p<0.05 for both) and at 21-25sec delays (p<0.05, for both). The surgical group treated with low Par were impaired at earlier time bins such as at 1-5sec delays (p<0.05) and 11-16sec delays (p<0.05) compared to MA Controls (Fig. 2I).
Paracetamol has a signi cant impact on the extent of POCD Following validation of in the DNMTS task in MA rats, we compared the effects of the post-operative analgesics treatment with Par or Bup on performance in the DNMTS task. Par was used at low (75mg/kg; i.p.) and high (150mg/kg; i.p.) doses and compared to Bup at low (0.05mg/kg; s.c.) and high (0.1mg/kg; s.c.) doses. First testing in the DNMTS task resumed 48hr following surgery/control procedures. All drugs were administered daily at 23hr before test. Animals exposed to anaesthetic alone or animals which underwent surgery and received either low or high Bup were impaired in total % correct response in the DNMTS task ( Fig. 2A, p<0.05, p<0.01, p<0.05, respectively). Further, surgery model groups treated with low or high Bup were signi cantly impaired in the DNMTS compared to the surgery group treated with the Par doses ( Fig. 2 (Fig. 2B). In contrast, the Par-treated surgery groups were not impaired, compared to the MA Controls, at any delay point (Fig. 2C). No group differences were observed at the 72hr test point (Fig. 2D, E and F). When animals were tested at 7 days post-surgery/control procedures, only the surgery groups treated with Bup were still impaired in overall total performance in the DNMTS task compared to MA Controls (Fig. 2G, p<0.05 for Bup high, p=0.058 for Bup low). The two-way ANOVA for repeated measures did not reveal a signi cant overall effect of analgesia treatments [F (5,47) (Fig. 2H) at 11-15sec delays (p<0.05 for both) and at 21-25sec delays (p<0.05, for both). The surgical group treated with low Par were impaired at earlier time bins such as at 1-5sec delays (p<0.05) and 11-16sec delays (p<0.05) compared to MA Controls (Fig. 2I).
Paracetamol has long-lasting analgesic effect in the cold plate test following surgery In addition to examination of cognition we also looked at the long-lasting effects of analgesia treatments in the cold plate test on days 1 to 5 post-surgery/control procedure. All drugs were administered daily 24hr before test in the cold plate assay. Interestingly, Par at both low and high doses was more effective at alleviating sensitivity in the cold plate test than Bup. Analysis by two-way ANOVA for repeated measures yielded signi cant main effect of test days [F (4,188) =18.17; p<0.0001]. Fisher's LSD revealed that the surgery group treated with the low Bup dose were signi cantly more sensitive on day 5 compared to the MA control group. In addition, Fisher's LSD pairwise comparisons revealed that the surgery group treated with the high Bup dose were signi cantly more sensitive on day 3 and day 5 compared to the MA control group (Fig. 4). Importantly, animals treated with Par, at both low and high dose, were not more sensitive in the cold plate test compared to MA controls.
Par and Bup have differential effects on peripheral cytokine levels following surgery.
Plasma cytokine levels were measured following sample collection at the end of the behavioural experiments, i.e. 8 days after exposure to iso urane anaesthesia or undergoing surgery procedure with analgesia treatments. Complete results are presented in Table 1. Brie y, Fisher's LSD test revealed that MA animals which underwent surgery and were treated with high or low dose Par had signi cantly higher levels of TNF-a compared to MA control (Table 1: p<0.05 for both). IL-10 levels were signi cantly increased in MA animals which underwent surgery and were treated with high dose Bup compared to MA animals exposed to anaesthesia only (Table 1: p<0.05). Moreover, IL-13 levels were signi cantly increased in MA animals which underwent surgery and were treated with high dose Bup compared to MA animals exposed to anaesthesia only ( Paracetamol increases hippocampal anti-in ammatory cytokines following surgery. Hippocampi were collected at the same time as plasma for the same cytokine analysis. Complete results are presented in Table 2. Brie y, Fisher's LSD test revealed that MA animals which underwent surgery and were treated with high dose Par had signi cantly higher levels of TNF-a compared to MA control or MA animals exposed to anaesthesia only ( Table 2: p<0.05). IL-13 levels were also signi cantly increased in MA animals which underwent surgery and were treated with low or high dose Par compared to MA control and compared to MA animals exposed to anaesthesia only (Table 2: p<0.01 for both). Moreover, IL-10 levels were signi cantly increased in MA animals which underwent surgery procedure and were treated with high dose Par compared to MA animals exposed to anaesthesia only (  Hippocampal a-tubulin PTMs are altered following exposure to anaesthesia or surgery. a-tubulin PTMs were measured in the same hippocampus samples used for cytokine analysis. Expression of D2-Tub was normalized on expression of total a-tubulin (TOT-Tub), while Tyr-Tub and Glu-Tub where analysed as Tyr-Tub/Glu-Tub ratio. Analysis by one-way ANOVA of Tyr-Tub/Glu-Tub ratio yielded no main signi cant group effects (F (5,45) =2.15, p=0.075). However, Fisher's LSD pairwise comparisons revealed a signi cant decrease in Tyr-Tub/Glu-Tub ratio with anaesthesia exposure (p<0.05) or surgery model procedure coupled with low dose Bup (p<0.05) or low dose Par (p<0.05) treatment compared to MA Control (Fig. 5A). The expression of D2-Tub/Tot-Tub ratio was also changed following anaesthesia or surgery model (one-way ANOVA; F (5,44) =7.6, p<0.0001). Fisher pairwise comparisons revealed a signi cant increase in D2-Tub/TOT-Tub ratio with anaesthesia exposure (p<0.01) or surgery model animals treated with low dose Bup (p<0.0001) or high dose Bup (p<0.0001) and surgery model animals treated with low dose Par (p<0.0001) or high dose Par (p<0.01), compared to MA Control (Fig.  5B). Finally, D2-Tub/Tot-Tub ratio was signi cantly increase in surgery model animals treated with high dose Bup compared to MA animals exposed to anaesthesia only ( Fig. 5B: p<0.01).

Discussion
Paracetamol prevents cognitive decline and exerts long-lasting analgesic effects in experimental POCD Our pilot study validated the DNMTS task protocol and con rmed previous ndings, demonstrating that age-related working memory impairments in this task are evident in MA rats compared to YG controls 20 . The translational relevance of the DNMTS task is con rmed by the extensive use of delayed-response tasks in the clinic to identify age-related de cits in humans 21 . Here, we also showed that cognitive decline is evident in the DNMTS task at 48hr post-surgery in MA animals exposed to iso urane alone or animals which underwent a surgery procedure and were treated with low (0.05mg/kg, i.p) or high (0.1mg/kg, i.p) dose of Bup as analgesia. Interestingly, animals which were treated with low (75mg/kg, s.c.) or high (150mg/kg, s.c.) dose Par were protected from POCD.
Previous studies have demonstrated that exposure of aged rats or mice to iso urane alone was su cient to induce spatial memory impairments 22,23 . Moreover, tibia fracture surgery in YG mice (12-14 weeks) using Bup (0.1mg/kg, s.c.) as analgesia exhibited reduced freezing to context when compared with naive mice in a delay fear conditioning paradigm 2,3 . Aged mice (18mths) who underwent laparoscopy surgery with Bup (0.1mg/kg, ip) as analgesia treatment displayed reduced cognitive exibility when tested 24hr after surgery 24 . In one of the only other POCD studies carried out in rats, the authors report cognitive decline in aged rats (18-20mths) following tibia facture surgery with buprenorphine (0.3mg/kg, ip) analgesia in the contextual fear-conditioning test and the Y-maze when tested at 24hr, 72hr and 7 days post-surgery 5 . We report temporary cognitive decline at 48hr post-surgery which appears to have recovered by the test at 72hr post-surgery. However, at the test on day 7 post-surgery, the animals exposed to anaesthesia alone have a tendency towards decreased performance in the DNMTS task and the animals treated with the high dose of Bup are signi cantly impaired, whereas the animals treated with high dose Par are comparable to the MA control group not exposed to anaesthesia or surgery.
It is noteworthy that a large number of the animals treated with low or high dose of Bup performed signi cantly less trials in the DNMTS task compared to the control group or animals exposed to iso urane only. As a result, 7 animals were excluded from the low dose Bup group and 6 animals were excluded from the high dose Bup group. These side effects of buprenorphine treatment may be attributed to appetite suppression following Bup treatment 25 resulting in reduced motivation to perform the DNTMS task. These data suggest buprenorphine may have a range of effects impacting motivation, reward and memory ultimately having a global negative effect on cognition. It cannot be excluded that animals treated with Bup may have experienced pain at the time of DNMTS testing as drugs were always administered 23hr before testing to avoid possible acute effects on cognition. Thus, our data from the cold plate test, which was carried out after the DNMTS task at approx. 24hr after drug administration, showed that surgery animals treated with Bup had increased sensitivity in the cold plate test compared to MA controls.
In contrast, the paracetamol treated surgery animals were not statistically different to the MA controls, suggesting that Par is a more effective long-lasting analgesic compared to Bup. Previous studies corroborate our observations since it was shown that single Bup (0.05mg/kg, s.c.) administration was effective as post-operative analgesic in rats up to 4hr post administration 26 , while a single dose of Par (50-100mg/kg, p.o.) has been shown to have anti-nociceptive effects up to 6hr postadministration 27 .Taken together, our ndings are consistent with clinical studies in a young population (age from 26 to 41: mean = 36 years) showing that repeated administration of Bup at high dose (32mg, oral admin, 10 days) had negative effects on memory resulting in delay-induced verbal memory impairment 28 . In addition, repeated administration of Bup at low dose (7mg, oral admin, 18-28 weeks daily) was also associated with delayed recall of verbal information in opioid-addicted patients (mean age = 36 years) 29 . Remarkably, and again in line with our data, a single Par administration (2mg, oral admin) was shown to improve performance in an information sampling task and increase hippocampalbased spatial memory in a double-blind clinical trial 30 .

Paracetamol modulates in ammatory cytokines in the plasma and hippocampus
The current study analysed for the rst time a larger panel of pro-in ammatory and anti-in ammatory cytokines protein levels in both plasma and hippocampus at day 8 after orthopaedic surgery. Our data show that anaesthesia alone did not affect systemic and central levels of any of the analysed cytokines, which is in line with previous reports 3 . IL-6 levels were increased only in the plasma and in the high dose Bup surgery group, while IL-1b was unchanged in all groups in either plasma or hippocampus.
Previous studies reported increased plasma and hippocampal levels of IL-6 and IL-1b at 6h and 24h after receiving orthopaedic surgery and a single injection of Bup (0.1 mg/kg, s.c.) as analgesic in YG adult (3-4 months) mice 2,3 , but not at 2hr or 72hr after surgery 2 . Our study appears to con rm that plasma increase in IL-6 and IL-1b is also not a sustained event in an MA rat model of orthopaedic POCD, since it is not detectable at 8 days after surgery, with the exception of plasma IL-6 levels that were signi cantly increased in animals receiving a daily high dose of Bup. This latter observation is consistent with work showing increased plasma IL-6 levels following iso urane anaesthesia and Bup administration before (0.025 mg/kg, s.c.) and after (0.05 mg/kg, s.c.) a scald-burn procedure 31 . Other authors observed increased plasma and synovial IL-6 levels after knee joint injury in mice injected with Bup (0.1 mg/kg, s.c.) followed by administration of the drug in drinking water for another 24hr 32 . The Bup-induced increase in plasma IL-6 levels might be linked to potential adverse effects of Bup as shown by our DNMTS task data.
An increase in plasma TNF-a has been previously shown to be rapid and transient, since it appears at 30min after orthopaedic surgery following a single injection of Bup (0.1 mg/kg, s.c.) as analgesic, but not at 1h, 2h, 6h and 12h after surgery) in young adult (3-4 months of age) mice 3 . Additionally, protein levels of TNF-a have been reported to be increased in the prefrontal cortex of aged mice (20 months of age) at 6 days after surgery 4 . It is hypothesised that that circulating TNF-a plays an important role in POCD and that it reaches the brain, following surgery, via physiological penetration of the blood-brain barrier 3 and disruption of the blood-brain barrier associated both with cognitive impairment and in ammatory response has been reported following tibia fracture 33 . Additionally, TNF-a is produced and released in the brain predominantly by microglia, astrocytes and neurons 34 . It has been speculated that increased brain levels of TNF-a might be involved in cognitive decline in brain disorders via potentiation of glutamate excitotoxicity (reviewed in 35 ).
Intriguingly, our results show that 8 days after surgery animals that received Bup present cognitive de cits and no altered levels of hippocampal TNF-a, while animals treated with the high dose of Par have no cognitive de cit but increased hippocampal TNF-a. Therefore, TNF-a appears to play a different role in the long-term cognitive de cits observed in our orthopaedic surgery models of POCD. It is of note that TNF-a has been shown to physiologically modulate Hebbian synaptic plasticity and synaptic scaling in the hippocampus where it can exert both excitotoxic or neuroprotective effects. For example, pretreatment of hippocampal slices with TNF-a after hypoxia improved LTP in the DG 36 , while overexpression of TNF-a in transgenic mice results in potentiation of LTP in CA1 region 37 . In the central nervous system, Par can be converted into N-arachinodyl-phenolamine (AM404) 38 , which is an inhibitor of the anandamide membrane transporter (AMT) and therefore indirectly increases anandamide levels and stimulate CB1 receptors 39 . It has been shown that CB1 receptors activation reduces the TNF-amediated potentiation of striatal spontaneous glutamate-mediated excitatory postsynaptic currents 40 .
Since the behavioural data of this study shows that Par prevent cognitive decline in POCD, it is possible to speculate that the observed sustained central increase in TNF-a might keep the correct synaptic plasticity homeostasis, and that its potential glutamate-induced excitotoxicity might be reduced by the indirect activation of the CB1 receptors induced by AM404. Additional experiments are required to investigate this speculative hypothesis in the future. On the other hand, TNF-a promotes fracture repair in both rodent models and in clinical settings 41 . Thus, the increase in circulating levels of TNF-a in the Par treated group can be bene cial in promoting a rapid repair of the tibial damage caused by the orthopaedic surgery employed in our POCD model. Importantly, this hypothesis appears to be corroborated by the long-lasting analgesic effects we have observed in the Par treated animals compared to Bup.
Increased circulating levels of the pro-in ammatory IL-5 and the anti-in ammatory IL-13 have been observed in our study 8 days after surgery in MA rats receiving Bup in a dose-dependent pattern, but not in animals that received Par. Both cytokines are secreted peripherally and their increase is associated with lung allergic reaction 42,43 and drug hypersensitivity 44 . Thus, our results may be linked to an adverse systemic reaction to the repeated Bup treatment. In contrast with the plasma data, we observed a signi cant increase of IL-13 in the hippocampus following Par administration (at both doses), but not following Bup. There is no evidence that IL-13 can pass the BBB, but some experimental studies showed its local production in the CNS by microglia and neurones and a potential neuroprotective role (reviewed in Mori, 2016 45 ). Thus, IL-13 can be produced by neuronal cells in the hippocampus and the cortex in models of ischemic insult where it induced an alternative activation of microglia, exerting a protective effect against neuronal damage 46 . Hence, it is possible to speculate that the observed increase in hippocampal IL-13 levels may have neuroprotective effect.
IL-10 is systemically produced and plays a critical role in preventing in ammatory and autoimmune pathologies by limiting the release of pro-in ammatory cytokines 47 . As for IL-13, we showed that IL-10 is increased in the plasma of animals receiving high dose Bup, but not Par; while the opposite is observed in the hippocampus. Our results are consistent with a previous study showing increased IL-10 serum levels following repeated Bup (0.075 mg/kg, s.c.) in a mouse model of arthritis 48 . Experimental models have shown that IL-10 is a "brain active" cytokine potentially produced in situ by microglial cells (for a review see 49 ). IL-10 protects astrocyte from excessive in ammation by inhibiting the microglia production of pro-in ammatory cytokines 50,51 and IL-10 receptor signalling has been associated with increased cellular survival and neurogenesis [52][53][54] . Therefore, the increase in hippocampal IL-10 observed following administration of the high dose of Par may have neuroprotective role, which is in line with the parallel increase of IL-13 and possibly that of TNF-a.

Paracetamol modulates microtubule dynamics in the hippocampus
Previous studies have shown persistent alterations in hippocampal synaptic plasticity in experimental rodent models of POCD 5,6 , but microtubule dynamics has never been investigated. Here we have analysed hippocampal a-tubulin PTMs resulting from the cycle of detyrosination/tyrosination and associated with microtubule dynamics (i.e. Tyr-Tub, Glu-tub and D2-Tub) at 8 days after orthopaedic surgery. Speci cally, the detyrosination/tyrosination cycle of a-tubulin consists of the enzymatic removal of the C-terminal tyrosine the re-addition of the tyrosine residue 55,56 resulting in Glu-Tub (detyrosinated atubulin) and Tyr-Tub (tyrosinated a-tubulin, respectively), which are here analysed as a Tyr-Tub/Glu-Tub ratio 57,58 . High levels of Glu-Tub are found in stable microtubules, while dynamic microtubules express more Tyr-Tub 59-61 . Additionally, Glu-Tub can be converted into a stable, entity which cannot re-enter the cycle, named D2-Tub by removal of the last glutamate residue 62 . In the brain, D2-Tub is principally expressed in neuronal cells where it appears restricted to very stable microtubules 62 .
Our results showed for the rst time decreased Tyr-Tub/Glu-Tub in MA rats exposed to iso urane alone or in animals which underwent orthopaedic surgery procedure and were treated with low or high dose Bup and low dose of Par. Previous studies have shown that cognitive de cits induce by a rat model of social isolation are paralleled by decreased Tyr-Tub/Glu-Tub in the hippocampus 63 and rescued by drugs having pro-cognitive e cacy 57 . Recently, the clinical link between Tyr-Tub/Glu-Tub and cognitive decline has been proposed based on post-mortem studies carried out in the hippocampus of Alzheimer disease patients showing alterations in the detyrosination/tyrosination cycle of a-tubulin 64 . Furthermore, our results show that the observed increased expression of Glu-Tub (i.e. decreased Tyr-Tub/Glu-Tub ratio) is accompanied by signi cant increase in D2-Tub production in all experimental groups. Interestingly, such an increase in D2-Tub is more pronounced in animals that received Bup following orthopaedic surgery which is suggestive of neuronal accumulation of this irreversible a-tubulin PTM due to overexpression of Glu-Tub induced by Bup over time.
Lack of Tubulin Tyrosine Ligase (TTL; the enzyme that produces Tyr-Tub) in mice leads to brain accumulation of Glu-Tub and D2-Tub, impairment of the cortico-thalamic loop caused by abnormal neuronal projections, and alterations of neurite, dendrite and axon formation in primary neuronal cell culture 65 . It has been proposed that neuronal abundance of Glu-Tub and D2-Tub may lead to hyperstabilization and altered interaction with MAPs eventually resulting in major impairments in axonal and dendritic formation (for a review see Janke, 2010 66 ). Consistently, sustained changes in markers of synaptic plasticity have been reported in experimental models of POCD. Speci cally, aged mice (16 months of age) that underwent laparotomy exhibited long-term cognitive decline paralleled by increases in neuroapoptotic markers (i.e. caspase-3 and iNOS) and decreased neuronal plasticity markers (i.e. BDNF, PSD-95 and synapsin-1) 7 days after surgery 6 . In the only other POCD studies carried out in rats and using a similar orthopaedic surgery, aged rats (18-20 months of age) showed long-term cognitive de cits accompanied by increased apoptosis and AMPAR GluA2 internalization 7 days after surgery 5 . It was also observed that 2h inhaling exposure to the anaesthetic sevo urane, compared to infusion of propofol, precipitated the reported surgery-induced synaptic changes 5 . Therefore, our results on the C-

Materials And Methods
Animals Middle aged (14-16 months old) Male Sprague Dawley rats sourced from Envigo UK, were used in these experiments. The animals were pair-housed in a controlled environment (temperature: 20-22°C, 12/12 hr light/dark cycle (lights on at 8am)), with water ad libitum. Animals were maintained at 85% free-feeding weight by controlled access to food. Animals were acclimatized to the facility environment for 2 week additional control treatment of vehicle 2. All treatments were administered approx. 23hr before testing in the DNMTS task. Therefore, the acute effects of the compounds did not interfere with performance in the DNMTS task. Treatments were administered daily from the day of surgery and every day until animals were euthanized.
Delayed non-match to sample (DNMTS) training protocol DNMTS was performed as previously described 20 . Speci cally, the rats were initially habituated to the operant conditioning chambers with the three levers extended. The animals were trained for 2 days to lever press for food reward on a continuous reinforcement schedule (i.e. pressing of any lever would result in the delivery of a sucrose pellet to the hopper). On the subsequent 2 days the levers were programmed to retract once pressed, delivering a pellet and then extending again. This was also on a continuous reinforcement schedule aimed to habituate the animals to the retraction and extension of the levers. On day 5, the same program was used with the exception that one speci c lever could not be reinforced more than 3 consecutive times. This modi cation was aimed to force the animals to perform alternate lever pressing, thereby suppressing lever preferences to obtain reward.
The next phase of training involved randomised presentation of the front lever (left or right) and once pressed the extension of the back-lever was triggered. The reward was delivered only after the back lever was pressed. These lever combinations were repeated 60 times (30 left/center and 30 right/center) at 10s intervals; this procedure was repeated for 2 days.
Training in the non-match-to-sample task was comprised of 90 trials in a maximum 90min session daily. At the start of each session the house light is on with the levers in the retracted position. The animals were initially trained on the task contingencies with no enforced delay between the sample and the choice component (0-delay condition). At the start of each trial one response lever was randomly selected and inserted into the chamber (the "sample"). As soon as the lever press response was registered the lever was retracted and the rear lever on the opposite wall extended. Once the response on the back lever was registered the two front levers were extended into the chamber together (the "choice"). If a correct response was registered (i.e. a response on the non-matching to sample lever) the levers retract and a pellet delivered to the hopper, the house light remained on and an inter-trial interval of 10 s was initiated before the next trial began. If an incorrect response was registered (i.e. a response on the initial sample lever) no pellet was delivered, the house light extinguished and the 10s interval initiated before the next trial started. Rats were required to meet a criterion of 85% for 3 consecutive days on this program before introduction of the delay. In the next stage of training a randomised 1 to 5s delay was introduced between the response on the sample lever and the extension of the rear lever. This phase lasted for 3 days.
In the nal stage of training, a random delay of 1-30s was introduced requiring the rat to wait for the extension of the rear lever before moving to the choice phase. Training continued on this phase of the task until the animals' performance reached a plateau (40 sessions).
Testing: Following surgery, animals were allowed to recover for 48hr before resuming testing in the DNMTS task. Animals were treated with Vehicle or Analgesia 23hr before test in the DNMTS task to avoid potential acute drug effects interfering with task performance. Animals were tested each day from 48hr post-surgery/control procedure up to and including day 8 post-surgery. Each daily session was composed of 90 trials of different delay lengths. Completion of the 90 trials each day was used as a control for any potential drug-induced interference in task performance. As such any animal which did not complete the This procedure was performed as described 69,70 and adapted for rats 5 . The procedure was performed by an experienced surgeon under supervision of the Designated Veterinarian (DV) at Trinity College Dublin. Induction and maintenance of anesthesia monitoring was carried out. Rats were placed in an induction box with 5% iso urane. The left hind paw was shaved and sterilized with surgical scrub. Rats were placed in a facemask and on a heat pad and maintained under iso urane anesthesia at 2-3%. Rats received one dose of their respective drug treatment (depending on treatment group) prior to surgery following anaesthesia with iso ourane. An incision was made in the surgical area and an appropriately sized pin (width 0.25-mm) inserted into the medullary canal. The wound was sutured and the rat was placed in a recovery cage on a heating pad before being returned to the home cage. Rats were treated with analgesia as indicated per group (Bup (0.05mg/kg or 0.1mg/kg) or Par (75mg/kg or 150mg/kg)) following surgery and once daily for the remaining duration of the study. All animals which underwent the surgery model procedure received analgesia as it would be unethical to perform this procedure without analgesia.
Plasma and tissue collection Animals were sacri ced following completion of the last test in the DNMTS task and brain tissue and blood was collected. Trunk blood was collected immediately in lithium heparin tubes and centrifuged at 200G for 15 minutes at room temperature. The platelet-rich plasma was then removed and placed into an Eppendorf and spun again at 2100G, 4°C for 10 minutes, the plasma is then transferred to another Eppendorf, with 2% protease inhibitor cocktail (P8340, Sigma) and frozen at -80°C. Brains were immediately extracted and the hippocampi were removed and placed into Eppendorfs, snap frozen on dry ice and stored at -80ºC until use.

Multiplex Cytokine Analysis
Hippocampal samples were removed from -80 °C and homogenised by sonication in lysis buffer (490 µl and 1:50 protease inhibitor, P8340, Sigma). Bradford assay was performed to prepare all samples to a common protein concentration. Prior to multiplex analysis plasma samples were defrosted and diluted (1:2) and standardized hippocampal samples by dilution (1:8). Cytokine levels for rat plasma/hippocampal samples were determined in duplicate using a V-Plex Multi-Spot assay system (proin ammatory panel 2 (Rat)) by Meso Scale Diagnostics. Plates were read using the MESO QuickPlex SQ 120 instrument and analyzed by Discovery Workbench 4.0 software. Statistical Analysis DNMTS trials were sorted by performance according to length of delay on individual trials and were grouped according to 5-s interval time bins (1-5, 6-10, 11-15, 16-20, 21-25, and 26-30). The data are presented graphically by total correct responses or percentage of correct responses at each 5-s delay time bin. All DNMTS data were statistically analyzed with SPSS, using t-test or RM ANOVAs and Fisher's pairwise comparisons. Molecular data were analysed by one-way ANOVA followed by Fisher's pairwise comparisons. GraphPad prism was used for graphical representations.