The effect of propofol on hypoxia‐ and TNF‐α‐mediated BDNF/TrkB pathway dysregulation in primary rat hippocampal neurons

Abstract Aims Hypoxia and inflammation may lead to BDNF/TrkB dysregulation and neurological disorders. Propofol is an anesthetic with neuroprotective properties. We wondered whether and how propofol affected BDNF/TrkB pathway in hippocampal neurons and astrocytes. Methods Primary rat hippocampal neurons and astrocytes were cultured and exposed to propofol followed by hypoxia or TNF‐α treatment. The expression of BDNF and the expression/truncation/phosphorylation of TrkB were measured. The underlying mechanisms were investigated. Results Hypoxia and TNF‐α reduced the expression of BDNF, which was reversed by pretreatment of 25 μM propofol in hippocampal neurons. Furthermore, hypoxia and TNF‐α increased the phosphorylation of ERK and phosphorylation of CREB at Ser142, while reduced the phosphorylation of CREB at Ser133, which were all reversed by 25 μM propofol and 10 μM ERK inhibitor. In addition, hypoxia or TNF‐α did not affect TrkB expression, truncation, or phosphorylation in hippocampal neurons and astrocytes. However, in hippocampal neurons, 50 μM propofol induced TrkB phosphorylation, which may be mediated by p35 expression and Cdk5 activation, as suggested by the data showing that blockade of p35 or Cdk5 expression mitigated propofol‐induced TrkB phosphorylation. Conclusions Propofol modulated BDNF/TrkB pathway in hippocampal neurons via ERK/CREB and p35/Cdk5 under the condition of hypoxia or TNF‐α exposure.


| INTRODUC TI ON
Brain-derived neurotrophic factor (BDNF) is one of the most studied and well-characterized neurotrophic factors in central nervous system (CNS), and is mainly produced in the brain by hippocampal neurons and astrocytes. BDNF plays an extensive role in neuronal growth, differentiation, survival, synaptic plasticity, and neurotransmitter regulation. 1 As such, BDNF may have therapeutic potential for the treatment of neurological disorders, such as cerebral ischemia-reperfusion injury, 2 neuroinflammation-related brain injury, 3 age-related memory impairment, 4 Parkinson's disease, 5 Alzheimer's disease, 6 and postoperative cognitive dysfunction. 7 It was originally thought that BDNF exerts its biological functions through binding to two transmembrane receptors, the tropomyosin receptor tyrosine kinase B (TrkB) and p75 neurotrophin receptor (p75NTR). However, accumulating evidence suggested that mature BDNF has high affinity to TrkB, while its precursor proBDNF mainly activates p75NTR. 8 Numerous in vitro and animal studies have implied a role for BDNF/TrkB in CNS damages induced by multiple pathophysiological stimuli such as oxidative stress, inflammation, and ischemia/reperfusion injury. 9 Recently, BDNF/TrkB signaling has been identified as a potential therapeutic target for depression, 10 post-cerebral ischemic spatial cognitive dysfunction, 11 vascular dementia, 12 and postoperative cognitive dysfunction. 13 Propofol is (2, 6-diisopropyl phenol) an intravenous general anesthetic used extensively in the induction and maintenance of anesthetization and procedural sedation. Apart from its multiple anesthetic advantages, it has been reported to possess anti-oxidative and anti-inflammatory effect 14,15 as well as neuroprotective property. 16 A number of in vitro studies revealed that propofol may protect mouse hippocampal neurons from autophagy induced by inflammation 17 and from apoptosis induced by inflammation, hypoxia, or oxidative stress. [18][19][20] In addition, in vitro models indicated propofol may protect the integrity of blood-brain barrier (BBB) from impairment by hypoxia and inflammation. 17,[21][22][23] Although propofol has been reported to modulate the expression of BDNF and TrkB in the hippocampus of cerebral ischemia injury 24 in aged rats, the effects of propofol on BDNF/TrkB pathway in the neurons have not been thoroughly investigated.
Although multiple pathological factors may affect CNS, it is acknowledged that CNS is especially vulnerable to ischemia and inflammation. 25 Furthermore, in vitro models proved that hypoxia treatment and inflammation cytokine TNF-α treatment may induce neuron injury, and mimic in vivo ischemic or inflammatory status. 26,27 In the current study, we aimed to detect whether propofol could modulate hypoxia-and TNF-α-mediated BDNF/TrkB pathway dysregulation in primary rat hippocampal neurons and astrocytes, and further investigate the underlying mechanisms. Our findings may provide a novel therapeutic target for hypoxia-and inflammation-induced neuron injury.

| Experimental design
Primary rat hippocampal neurons and astrocytes were cultured in normoxic condition (95% humidified air and 5% CO 2 ) until ready for experiments. To mimic hypoxic condition, cells were maintained in a hypoxic chamber flushed with a humidified gas mixture (90% humidified N 2 , 5% O 2 , and 5% CO 2 ) for different duration (0, 1, 2, 3, 6, and 12h). To mimic inflammation condition, cells were treated with 40ng/ mL TNF-α for different duration (0, 1, 2, 3, 6, and 12h). To examine the effect of propofol, cells were treated with different concentrations (1,5,10,25,50, and 100 μM) of propofol (Sigma-Aldrich, St. Louis, MO, USA) or its solvent, 0.1% dimethyl sulfoxide (DMSO) for 1h, and then exposed to hypoxic or inflammation condition. These treatment conditions have been repeatedly used to induce vicious effects in hippocampal neurons. 17,21 We intended to identify the effect of hypoxia, inflammation, and propofol on the production of BDNF and the expression/truncation/phosphorylation of TrkB in hippocampal neurons and astrocytes. More importantly, we aimed to investigate the mechanisms, including extracellular regulated protein kinase (ERK), cAMP response element-binding protein (CREB), p35, and cyclin-dependent kinase 5 (Cdk5). To confirm their roles, specific inhibitors and short-interference RNAs (siRNAs) were applied.

| Cell culture
Primary rat hippocampal neurons and astrocytes were purchased from ScienCell Research Laboratories (Carlsbad, CA, USA). The cryopreserved primary rat hippocampal neurons were thawed and seeded into tissue culture flasks containing 5 mL neuronal medium, which was supplemented with neuronal growth supplement and 1% penicillin/streptomycin. The culture media were replaced every 2-3 days.
Neurons were incubated at 37°C in a humidified atmosphere with 5% CO 2 , and were ready for experiments without sub-culturing.
The cryopreserved rat astrocytes were thawed and seeded into tissue culture flasks containing 5ml astrocyte medium-animal, which was supplemented with 2% fetal bovine serum (FBS), astrocyte growth supplement-animal, and 1% penicillin/streptomycin.
Astrocytes were incubated at 37°C in a humidified atmosphere with 5% CO 2 , and culture media were replaced every 2-3 days. The cells were sub-cultured when reaching 80-90% confluency, and the fourth passage of astrocytes was used in the present study.

| Cell viability assay
To check the effect of hypoxia, TNF-α, and propofol on the viability of hippocampal neurons and astrocytes, 3-(4,5)-dimethylthiahiazo (-z-y1)-3,5-di-phenytetrazoliumromide (MTT) protocol was used. In brief, cells were cultured in six-well tissue culture plates until reaching 80% confluency. After respective treatment, the media were removed and the cells were carefully rinsed with phosphate-buffered saline (PBS). Then, 5 mg/mL MTT was added to each well and incubated for 4h. Finally, the resulting formazan crystals were dissolved in DMSO, and the absorbance was measured by a spectrophotometer at 490 nm. Cell viability was expressed as the percentage of absorbance of treated cells compared with that of untreated control cells.

| mRNA analysis by RT-PCR
After treatment, total RNA was isolated from cells with the use of

| Transient transfection of siRNA
In this in vitro study, we used siRNA technology to knockdown spe-

| Statistical analysis
Data were presented as mean ± standard deviation. All experiments were conducted with five independent repeats, which were performed with different cultures. The Shapiro-Wilk test was used to analyze the normality of the data and we found that all data were normally distributed. Differences between groups were assessed with paired, two-tailed Student's t-test or one-way ANOVA, followed by post hoc Tukey testing. All statistical analyses were performed with SPSS software 11.5, and a significant difference was set at P < 0.05.

| Hypoxia and TNF-α reduced BDNF expression and cell viability in rat hippocampal neurons and astrocytes
Rat hippocampal neurons and astrocytes were cultured and exposed to hypoxia (5% O 2 ) or TNF-α (40 ng/mL) treatment for different duration (0, 1, 2, 3, 6, and 12h), and the expression of BDNF as well as cell viability was measured. As shown in Figure 1, we reported that in hippocampal neurons and astrocytes, hypoxia reduced BDNF protein expression in a time-dependent manner, with the significant effects appearing at 3h in hippocampal neurons ( Figure 1A, P < 0.01 vs. control) and at 6h in astrocytes ( Figure 1B, P < 0.01 vs. control). In addition, we found that TNF-α also reduced BDNF protein expression in a time-dependent manner, and the significant effects appeared at 3h in both hippocampal neurons and astrocytes ( Figure 1C,D, P < 0.01 vs. control). Furthermore, the effect of hypoxia and TNF-α on BDNF protein expression was confirmed by measuring BDNF mRNA levels ( Figure S1). We also showed that hypoxia and TNF-α treatment reduced cell viability in hippocampal neurons ( Figure 1E) and in astrocytes ( Figure 1F), and the significant effects appeared after 3h treatment. Thereafter, these treatment conditions were applied in the following experiments to study the potential mechanisms.

| Propofol reversed hypoxia-and TNF-αmodulated BDNF reduction and cell viability in rat hippocampal neurons
To observe the effects of propofol on hypoxia-and TNF-α-modu- 100 μM) also induced BDNF protein expression, which was inhibited by TNF-α (40 ng/mL, 3h) treatment ( Figure 2B, P < 0.01 vs. control, P < 0.05 vs. TNF-α). In contrast, we found that even 100 μM propofol had no or minor effect on BDNF protein expression in astrocytes in response to hypoxia or TNF-α ( Figure 2C,D). Furthermore, the effect of propofol on hypoxia-and TNF-α-modulated BDNF protein expression was confirmed by measuring BDNF mRNA levels ( Figure S2). Meanwhile, our data indicated that propofol (25, 50, and 100 μM) rescued cell viability in hippocampal neurons ( Figure 2E), but not in astrocytes ( Figure 2F). Also, please note that 0.1% DMSO, the solvent for propofol, had no effect on BDNF protein production in hippocampal neurons or astrocytes ( Figure 2). Therefore, we ruled out the role of DMSO. More importantly, we inferred that 25 μM propofol might be the minimally effective concentration that reversed hypoxia-and TNF-α-inhibited BDNF production in hippocampal neurons, and accordingly we focused on the mechanism responsible for the beneficial effect of 25 μM propofol.

| Hypoxia and TNF-α had no effect on TrkB expression, truncation, or phosphorylation in rat hippocampal neurons and astrocytes
Rat hippocampal neurons and astrocytes were cultured and exposed to hypoxia (5% O 2 ) or TNF-α (40 ng/mL) treatment for different duration (0, 1, 2, 3, 6, and 12h), and the expression, truncation, as well as phosphorylation of TrkB were measured. As shown in Figure 4, we reported that hypoxia had no effect on the expression, truncation, or phosphorylation of TrkB in rat hippocampal neurons ( Figure 4A) and in astrocytes ( Figure 4B). Also, TNF-α had no effect on the expression, truncation, or phosphorylation of TrkB in rat hippocampal neurons ( Figure 4C) and astrocytes ( Figure 4D).

| Propofol induced TrkB phosphorylation in rat hippocampal neurons
We treated rat hippocampal neurons and astrocytes with different concentrations of propofol (1, 5, 10, 25, 50, and 100 μM) for 1h, followed by hypoxia (5% O 2 , 3h) or TNF-α (40 ng/mL, 3h) treatment, F I G U R E 1 Hypoxia and TNF-α reduced the expression of BDNF and cell viability in hippocampal neurons and astrocytes. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of BDNF against GAPDH, and were presented as mean ± standard deviation. Hypoxia treatment for 0h was considered as normoxic condition and served as control. TrkB. We noticed that in rat hippocampal neurons, propofol had no effect on TrkB expression or truncation ( Figure 5A). Interestingly, propofol (50 and 100 μM) induced TrkB phosphorylation, regardless of whether cells were exposed to hypoxia, TNF-α, or not ( Figure 5A, P < 0.05 vs. control). However, propofol had no effect on TrkB F I G U R E 2 Propofol reversed hypoxia-and TNF-α-modulated BDNF reduction and cell viability in hippocampal neurons. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of BDNF against GAPDH, and were presented as mean ± standard deviation. (A) In hippocampal neurons, propofol induced BDNF expression, which was inhibited by hypoxia (5% O 2 , 3h) treatment. (B) In hippocampal neurons, propofol induced BDNF expression, which was inhibited by TNF-α (40 ng/mL, 3h) treatment. (C) In astrocytes, propofol had no effect on hypoxia-modulated BDNF expression. (D) In astrocytes, propofol had no effect on TNF-α-modulated BDNF expression. (E) In hippocampal neurons, propofol rescued cell viability, which was reduced by hypoxia or TNF-α treatment. Data were expressed as the percentage of absorbance of treated cells compared with that of untreated control cells. (F) In astrocytes, propofol had no effect on hypoxia-or TNF-α-mediated cell viability expression, truncation or phosphorylation in astrocytes ( Figure 5B).
Thereafter, we intended to investigate the mechanism responsible for 50 μM propofol-induced TrkB phosphorylation in hippocampal neurons.
Consistently, although hypoxia and TNF-α had no effect on the activation of Cdk5 ( Figure 6B), it was activated by 50 μM propofol but not 0.1% DMSO. In addition, hypoxia, TNF-α, propofol, and DMSO had no effect on the expression of Cdk5 and p39 ( Figure 6C). Then, we applied siRNA technology to confirm the involvement of p35 and Cdk5 in propofol-mediated TrkB phosphorylation. As shown in Figure 6D, we demonstrated that the siRNA targeting p35, p39, and Cdk5 could effectively diminish the expression of p35, p39, and Cdk5, respectively. More importantly, we revealed that blockade of p35 and Cdk5 alleviated propofol-induced TrkB phosphorylation, while blockade of p39 had no such effect ( Figure 6E). To confirm the relationship between p35 and Cdk5, we blocked p35 expression

F I G U R E 3
The beneficial effect of propofol on BDNF expression was mediated through regulating the phosphorylation of ERK and CREB. (A) In hippocampal neurons, hypoxia (5% O 2 , 3h) and TNF-α (40 ng/mL, 3h) increased the phosphorylation of ERK, which was attenuated by 25 μM propofol, 10 μM PD98059, or 10 μM KO-947. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of phosphorylated ERK against ERK, which was normalized with GAPDH, and were presented as mean ± standard deviation. (B) In hippocampal neurons, hypoxia (5% O 2 , 3h) and TNF-α (40 ng/mL, 3h) increased the phosphorylation of CREB at Ser142 (p-CREB Ser142 ), which was reversed by 25μM propofol, 10μM PD98059, or 10μM KO-947. (C) In hippocampal neurons, hypoxia (5% O 2 , 3h) and TNF-α (40 ng/mL, 3h) reduced the phosphorylation of CREB at Ser133 (p-CREB Ser133 ), which was reversed by 25μM propofol, 10μM PD98059, or 10μM KO-947. (D) In hippocampal neurons, hypoxia and TNF-α reduced BDNF expression, which was reversed by 25μM propofol, 10μM PD98059, or 10μM KO-947, and the beneficial effect of propofol on BDNF production was abolished by 10μM Ceramide C6 by siRNA technology and examined Cdk5 activity. As shown in Figure 6F, we revealed that the propofol-induced Cdk5 activation was alleviated by blockade of p35 and Cdk5, rather than p39 expression ( Figure 6F). TrkA has been identified as the preferred receptor for NGF, TrkB for BDNF, and TrkC for NT-3/4/5. 29 Prior findings highlighted the role of p75NTR in neuronal pathological conditions such as traumatic brain injury, intracranial hemorrhage, ischemic/hypoxia stroke, and neurological diseases, by modulating neuronal apoptosis, axonal elongation and degeneration, myelination, cell proliferation, and synaptic plasticity. 30 In contrast, Trk receptors could activate signaling pathways that are important for cellular activities, including cell proliferation, survival, and axon, as well as dendrite outgrowth. 31 In the current in vitro study, we mainly examined matured form of BDNF, and accordingly, we only focused on TrkB receptor. After BDNF binding, TrkB undergoes dimerization, followed by phosphorylation of intracellular tyrosine kinase residues which act as docking sites for adaptor proteins that recruit additional kinases, leading to activation of intracellular signaling pathways. The activation of BDNF/ TrkB is required for neuron differentiation, survival, synaptic plasticity, and neurotransmitter regulation, while dysregulation of BDNF/ TrkB contributes many pathological processes, including traumatic brain injury, brain ischemic injury, and neurodegenerative diseases. 13 It is known that BDNF/TrkB dysregulation was correlated with several harmful insults, such as oxidative stress and inflammation. 32 In the current study, we focused on two factors (hypoxia F I G U R E 4 Hypoxia and TNF-α had no effect on full-length TrkB (fl-TrkB) expression, truncation (t-TrkB), or phosphorylation (p-TrkB) in rat hippocampal neurons and astrocytes. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of p-TrkB or t-TrkB against fl-TrkB, which was normalized with GAPDH, and were presented as mean ± standard deviation. (A) In rat hippocampal neurons, hypoxia had no effect on the expression, truncation, or phosphorylation of TrkB. (B) In astrocytes, hypoxia had no effect on the expression, truncation, or phosphorylation of TrkB. (C) In rat hippocampal neurons, TNF-α had no effect on the expression, truncation, or phosphorylation of TrkB. (D) In astrocytes, TNF-α had no effect on the expression, truncation, or phosphorylation of TrkB and inflammation), which are major players in the pathogenesis of neurological disorders, and two cell types (hippocampal neurons and astrocytes), which are major sources of BDNF in CNS. We found that both hypoxia and inflammation reduced the expression of BDNF in hippocampal neurons and astrocytes ( Figure 1). However, they had no effect on TrkB expression, truncation, or phosphorylation ( Figure 4). Since we only focused the role of mature BDNF in this study, we did not examine p75NTR, TrkA, or TrkC. In addition, it is known that TrkB has two isoforms: truncated TrkB (TrkB-TC) and full-length TrkB (TrkB-FL). TrkB-TC may act as negative modulators of TrkB-FL. A previous study showed that excitotoxic stimulation of cultured rat hippocampal neurons with glutamate down-regulated TrkB-FL while up-regulated TrkB-TC, which resulted in dysregulation of BDNF/TrkB signaling. 33 Nevertheless, we found neither hypoxia nor TNF-α affected the truncation of TrkB (Figure 4). Interestingly, our findings differ from a previous animal study that reported an increase in BDNF and TrkB expression in chronic cerebral ischemia in the hippocampus of aged rats. 24 We postulated that the discrepancy could be due to two reasons: firstly, we examined acute hypoxia and inflammation rather than chronic ischemia; secondly, our study was carried out in neurons rather than in aged animals. We concluded that in hippocampal neurons and astrocytes, hypoxia and inflammation may cause dysregulation of BDNF/TrkB pathway mainly through modulation of BDNF expression. Our data indicated that hippocampal neurons and astrocytes responded differently to hypoxia and TNF-α. In other words, in different cells, the significant F I G U R E 5 Propofol induced TrkB phosphorylation in hippocampal neurons. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of p-TrkB or t-TrkB against fl-TrkB, which was normalized with GAPDH, and were presented as mean ± standard deviation. (A) In hippocampal neurons, propofol had no effect on TrkB expression or truncation, but induced TrkB phosphorylation. (B) In astrocytes, propofol had no effect on TrkB expression, truncation, or phosphorylation (B)

| The protective property of propofol against hypoxia-and TNF-α-mediated BDNF/TrkB dysregulation
Propofol is an intravenous anesthetic widely used in clinical anesthesia and sedation. In addition, it has a variety of biological effects on the protection of organs, including brains, 34 hearts, 35 and kidneys. 36 Currently, the neuroprotective property of propofol in the CNS and the underlying mechanism are of great interests. A large amount of in vitro studies revealed that propofol may improve BBB function, 21 protect neuron apoptosis 18 and autophagy, 17 and maintain microglia function. 37 In addition, animal studies demonstrated that propofol may improve brain function in rats with ischemia-reperfusion injury 38 and may ameliorate neuroinflammatory injury in rats. 39,40 Recently, the role of BDNF/TrkB signaling in the neuroprotective property of propofol gains interests. An animal study indicated that propofol may protect chronic ischemic cerebral injury in aged rats via modulating BDNF/TrkB pathway. 24 In that animal study, it was reported that low-dose propofol (10 mg/kg, intraperitoneally) promoted the expression of BDNF and TrkB, but high-dose propofol (50 mg/kg, intraperitoneally) inhibited their expression. Consistently, our in vitro study demonstrated that 25-50 μM propofol induced BDNF expression in hippocampal neurons which were exposed to hypoxia or TNF-α ( Figure 2). Meanwhile, we found propofol had no effect on TrkB expression, while increased its phosphorylation regardless of whether the hippocampal neurons were exposed to hypoxia/TNF-α or not ( Figure 5). We postulated that the difference in the amount of propofol administration and the difference in experiment model may account for the discrepancy. In contrast, our data implied that astrocytes may not be a target for propofol with regard to BDNF/TrkB dysregulation (Figures 2 and 5). We note that the beneficial concentration of propofol was 25-50 μM, which is within the plasma range of propofol during general anesthesia and is clinically relevant. Accordingly, we concluded that propofol may regulate hypoxia-and TNF-α-mediated BDNF/TrkB dysregulation, through BDNF expression and TrkB phosphorylation in hippocampal neurons, but not in astrocytes. Here, in the present study, our data suggested that ERK/CREB signaling is involved in hypoxia-and TNF-α-mediated BDNF/TrkB dysregulation ( Figure 3). More specifically, we believed that ERK/ CREB signaling plays a key role in the beneficial effect of propofol on BDNF production. The pivotal role of ERK/CREB in BDNF production has previously been shown in the brain of mice 44 and rats. 45 It is well known that CREB could be phosphorylated by protein kinases such as protein kinase A (PKA), protein kinase C (PKC), PI3K, CAMKII, and ERK at different site such as Ser133 and Ser142, and it is recognized that most kinases induce p-CREB Ser133 , which increases CREB transcriptional activity, while some kinases induce p-CREB Ser142 , which decreases its activity. Although p-CREB Ser133 F I G U R E 6 Propofol-induced TrkB phosphorylation was carried out via modulating p35 expression and Cdk5 activation in hippocampal neurons. (A) Propofol induced the expression of p35. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of p35 against GAPDH, and were presented as mean ± standard deviation. (B) Propofol induced the activation of Cdk5. Data were expressed as relative fluorescence compared with that of untreated control cells, and were presented as mean ± standard deviation. One hundred percent activity was set for control cells. (C) Propofol had no effect on the expression of Cdk5 and p39. The panel was a representative experiment, and GAPDH served as loading control. (D) Transfection efficiency of siRNAs against p35, p39, and Cdk5 was evaluated by Western blot. Untransfected neurons served as normal control, and control siRNA-transfected neurons served as transfection control. The panel was a representative experiment. (E) Transfection of siRNA against p35 or Cdk5 alleviated propofol-induced TrkB phosphorylation, while transfection of siRNA against p39 had no such effect. The upper panel was a representative experiment and the lower panel was the summary of densitometric data from five separate experiments. GAPDH served as loading control. Data were expressed as normalized ratio of protein band density of p-TrkB against fl-TrkB, which was normalized with GAPDH, and were presented as mean ± standard deviation. (F) Blockade of p35 and Cdk5 alleviated Cdk5 activation. Data were expressed as relative fluorescence compared with that of untreated control cells, and were presented as mean ± standard deviation. One hundred percent activity was set for control cells has already been shown to be correlated with BDNF production in rat model 46 and in rat cortical neurons 47 as well as in mouse hippocampal neurons, 48 the role of p-CREB Ser142 has rarely been investigated. Given our data showing that ERK inhibitor attenuated hypoxia-and TNF-α-mediated phosphorylation of CREB Ser142 and CREB Ser133 , and that the presence of ERK activator markedly abolished the effects of propofol on CREB phosphorylation (Figure 3), it is likely that ERK functions upstream of CREB. While our data did not prove a definitive role of CREB in BDNF production, our finding of correlation between the phosphorylation of CREB and propofol-induced BDNF production is a novel aspect of the current study, particularly the finding of an increase in p-CREB Ser133 and a decrease in p-CREB Ser142 .

| ERK/CREB and p35/Cdk5 were involved in the beneficial effect of propofol against hypoxia-and TNF-α-mediated BDNF/TrkB dysregulation
In addition, our data implicated a role of p35/Cdk5 in propofol-modulated BDNF/TrkB pathway, especially on TrkB, based on our data showing that blockade of p35/Cdk5 abolished the effect of propofol on TrkB phosphorylation ( Figure 5). Cdk5 is a small serine/threonine kinase abundant in postmitotic neurons, and the activation of Cdk5 requires the binding of one of its two specific activators, p35 or p39, in the developing cerebral cortex and hippocampus. 49,50 It is known that p35 and p39 share approximately 60% sequence homology and exhibit differential developmental expression in the brain. The expression of p35 protein is high throughout the embryonic stage, whereas p39 expression increases during postnatal differentiation. Although in vitro experiments suggested that p35 and p39 share similar substrate specificity, they are spatially segregated within neurons and have different biochemical properties. 51 Previous study indicated that p35/Cdk5mediated phosphorylation of target protein is required for hypoxia-induced xanthine oxidoreductase hyperactivation in the lung, 52 and p35/Cdk5 has been shown to be responsible for phosphorylation of TrkB, neurofilament proteins, and tau protein in brain. 50,53 Consistently, we found p35, rather than p39, is critical for TrkB phosphorylation in the hippocampal neurons that were exposed to hypoxia, TNF-α, and propofol ( Figure 6E). Meanwhile, based on the findings that blockade of p35 expression attenuated Cdk5 activity ( Figure 6F), we concluded that p35 functions upstream of Cdk5.
In the present in vitro study, the beneficial concentration of propofol was 25 and 50 μM, which is about 4.5 and 9 μg/mL. In clinical practice, during general anesthesia with propofol, plasma concentration was usually kept at 2-6 μg/mL. It seems 50 μM is beyond physiological concentration range, and may not be achieved during general anesthesia. However, given the fact that in vitro system is different from in vivo system and that the concentration of an agent used in in vitro study maybe 10 times higher than that used in clinical practice, we believe that the protective concentration found in this study is clinically relevant. On the other hand, although this is an in vitro study carried out in hippocampal neurons, it is of great clinical relevance and significance. Our findings may provide rationale for choosing propofol as an intraoperative or sedative agent in those patients who are under the risk of cerebral ischemia or inflammation, and this need to be confirmed by following animal study.

| Limitation
Our study has several limitations. Firstly, we focused on ERK/CREB and p35/Cdk5 in hypoxia-and TNF-α-as well as propofol-mediated regulation of BDNF/TrkB pathway, without investigating detailed analysis of other components in the signaling pathway. Secondly, it is known that p-CREB may be dephosphorylated by phosphatase PP1 and PP2A to keep the balance of its phosphorylation status.
However, in the study, we did not examine the effect of hypoxia, TNF-α, or propofol on the expression and activity of these enzymes.

| CON CLUS ION
In this in vitro study, we reported that in rat hippocampal neurons, propofol via modulating ERK/CREB signaling pathway may reverse hypoxia-and TNF-α-mediated reduction in BDNF. In addition, we demonstrated that in hippocampal neurons, propofol via activating p35/Cdk5 pathway induces TrkB phosphorylation. Taken together, our findings suggested a novel protective effect and mechanism of propofol against hypoxia-and TNF-α-induced malfunction of hippocampal neurons.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.

AUTH O R CO NTR I B UTI O N S
Weiping Tao performed research and wrote the paper; Xuesong Zhang performed research and wrote the paper; Juan Ding performed research and analyzed data; Shijian Yu performed research; Peiqing Ge analyzed data; Jingfeng Han analyzed data; Xing Luo analyzed data and wrote the paper; Wei Cui designed research and revised the paper; Jiawei Chen designed research and revised as well as approved the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.