Resolvin D1 limits 5-lipoxygenase nuclear localization and leukotriene B₄ synthesis by inhibiting a calcium-activated kinase pathway

Significance

Specialized proresolving mediators, such as resolvin D1 (RvD1), are endogenous molecules that both dampen inflammation without compromising host defense and promote tissue resolution. A prime example is RvD1’s ability to decrease the ratio of proinflammatory leukotriene B₄ (LTB₄) to proresolving lipoxin A₄ (LXA₄), but the mechanism is not known. We have discovered a new calcium kinase signaling pathway through which RvD1 lowers the nuclear:cytoplasmic ratio of 5-lipoxygenase (5-LOX), the common enzyme for LTB₄ and LXA₄ biosynthesis This shift in 5-LOX localization dampens LTB₄ production and enhances LXA₄ production. By providing a new mechanistic understanding of how RvD1 tempers inflammation to facilitate resolution, these findings can help devise new therapeutic strategies for diseases driven by nonresolving inflammation.

Abstract

Imbalances between proinflammatory and proresolving mediators can lead to chronic inflammatory diseases. The balance of arachidonic acid-derived mediators in leukocytes is thought to be achieved through intracellular localization of 5-lipoxygenase (5-LOX): nuclear 5-LOX favors the biosynthesis of proinflammatory leukotriene B₄ (LTB₄), whereas, in theory, cytoplasmic 5-LOX could favor the biosynthesis of proresolving lipoxin A₄ (LXA₄). This balance is shifted in favor of LXA₄ by resolvin D1 (RvD1), a specialized proresolving mediator derived from docosahexaenoic acid, but the mechanism is not known. Here we report a new pathway through which RvD1 promotes nuclear exclusion of 5-LOX and thereby suppresses LTB₄ and enhances LXA₄ in macrophages. RvD1, by activating its receptor formyl peptide receptor2/lipoxin A₄ receptor, suppresses cytosolic calcium and decreases activation of the calcium-sensitive kinase calcium-calmodulin-dependent protein kinase II (CaMKII). CaMKII inhibition suppresses activation P38 and mitogen-activated protein kinase-activated protein kinase 2 kinases, which reduces Ser271 phosphorylation of 5-LOX and shifts 5-LOX from the nucleus to the cytoplasm. As such, RvD1’s ability to decrease nuclear 5-LOX and the LTB₄:LXA₄ ratio in vitro and in vivo was mimicked by macrophages lacking CaMKII or expressing S271A-5-LOX. These findings provide mechanistic insight into how a specialized proresolving mediator from the docosahexaenoic acid pathway shifts the balance toward resolution in the arachidonic acid pathway. Knowledge of this mechanism may provide new strategies for promoting inflammation resolution in chronic inflammatory diseases.

Persistent inflammation and its failed resolution underlie the pathophysiology of prevalent human diseases, including cancer, diabetes, and atherosclerosis (1). Hence, uncovering mechanisms to suppress inflammation and enhance resolution is of immense interest (2–5). Resolution is orchestrated in part by specialized proresolving mediators (SPMs), including lipoxins, resolvins, protectins, and maresins (2), and by protein and peptide mediators (6). A common protective function of SPMs is their ability to limit excessive proinflammatory leukotriene formation without being immunosuppressive (2, 7). Specifically, resolvin D1 (RvD1) is protective in several disease models (8) and limits excessive leukotriene B₄ (LTB₄) production without compromising host defense (7, 9). However, the mechanism underlying these actions of RvD1 is not well understood.

Arachidonic acid (AA) is first converted into 5-hydroperoxyeicosatetraenoicacid (5-HPETE) and then into leukotriene A₄ (LTA₄) by 5-lipoxygenase (5-LOX) (10, 11). Subsequent hydrolysis of LTA₄ by LTA₄ hydrolase yields LTB₄ (10, 11). During inflammation, 5-LOX is phosphorylated and translocates to the nuclear membrane, which favors the biosynthesis of LTB₄ (12–17). However, major gaps remain in our understanding of the relevance of this pathway to primary cells and animal models and how they are regulated by SPMs. Further, it is currently not known how this pathway may influence the biosynthesis of lipoxin A₄ (LXA₄), which is a SPM that also requires 5-LOX. These gaps are critical because although LTB₄ is crucial for host defense, exuberant production underlies the basis for several inflammatory diseases and impairs endogenous resolution programs (11, 18). Moreover, complete blockade of LTB₄ biosynthetic enzymes may compromise host defense; thus, understanding new mechanisms that temper LTB₄ production is essential for translational research in this area (19).

Here, we report that RvD1, by suppressing the activation of the calcium-sensing kinase calcium-calmodulin-dependent protein kinase II (CaMKII), decreases the phosphorylation and nuclear localization of 5-LOX and thereby limits LTB₄ biosynthesis. These results provide a mechanistic understanding of how RvD1 tempers proinflammatory responses to facilitate a rapid transition to resolution.

Results

RvD1 Suppresses AA-Stimulated LTB₄ by Blocking P38/MK2-Mediated 5-LOX Phosphorylation and Nuclear Localization.

We first showed that 1 nM RvD1 enhanced AA-stimulated LXA₄ generation (Fig. 1A) and blocked LTB₄ formation in both bone marrow-derived macrophages and zymosan-elicited peritoneal macrophages (Fig. 1B and SI Appendix, Fig. S1 and S2A). We conducted an RvD1 dose–response experiment and found that the suppression of AA-stimulated LTB₄ generation was close to maximal at 1 nM RvD1 (SI Appendix, Fig. S3). These actions of RvD1 were mediated through its receptor formyl peptide receptor2/lipoxin A₄ receptor (FPR2/ALX) (20), as both an FPR2/ALX blocking antibody (Fig. 1C) and the FPR2/ALX antagonist WRW4 (SI Appendix, Fig. S2B) blocked the ability of RvD1 to reduce AA-stimulated LTB₄ generation. LXA₄, another ligand for FPR2/ALX (21, 22), also blocked LTB₄ generation (Fig. 1D). LTB₄ is generated from LTA₄, a product of 5-LOX, through the action of LTA₄ hydrolase (Fig. 1E) (11). To help pinpoint the step in this pathway at which RvD1 was acting, we circumvented the 5-LOX step by incubating macrophages with LTA₄ instead of AA. As expected, LTA₄ significantly increased LTB₄ level, but RvD1 did not block this increase (Fig. 1F), indicating that RvD1 was acting upstream of the LTA₄ hydrolase. These data prompted us to explore the hypothesis that RvD1 was acting at the level of 5-LOX.

Fig. 1.

RvD1 (D1) suppresses AA-stimulated LTB₄ a in a receptor-dependent manner upstream of LTA₄ hydrolase. (A, B, and D) BMDMs were preincubated with vehicle control, 1 nM RvD1, or 1 nM LXA₄ for 15 min, followed by incubation with 10 μM AA for 40 min. The media were then assayed for LXA₄ by liquid chromatography (LC)-MS/MS (A) and LTB₄ by ELISA (n = 6 for LC-MS/MS and n = 3 for ELISA; mean ± SEM; *P < 0.05 versus AA). In the experiment in A, LXB₄ was not detected by LC-MS/MS. (C) As in B, but the cells were pretreated with IgG control (black bars) or anti-ALX/FPR2 IgG (gray bars) for 1 h at 37 °C (n = 3; mean ± SEM; *P < 0.05 versus all other groups). (E) Scheme of LTB₄ biosynthesis. (F) As in B, except 10 μM LTA₄ was used instead of AA (n = 3; mean ± SEM; **P < 0.05 versus vehicle; n.s., nonsignificant). LTA₄ showed no cross reactivity with the LTB₄ ELISA.

Phosphorylation of 5-LOX at Ser271 by the P38-activated kinase mitogen-activated protein kinase-activated protein kinase 2 (MAPKAPK2) (MK2) leads to nuclear localization of 5-LOX, which allows 5-LOX to align with other enzymes required for LTB₄ synthesis (17, 23–25). We found that RvD1 blocked AA-stimulated phosphorylation of both kinases, which is a measure of their activation state (Fig. 2 A and B). FPR2/ALX is a G protein-coupled receptor, which led us to explore the possible role of cAMP. We found that 8-bromo-cAMP blocked RvD1’s ability to decrease p-MK2, whereas the Rp-cAMP mimicked RvD1’s actions (SI Appendix, Fig. S4). These results are consistent with a receptor-mediated response involving Gi proteins, which is in line with previous studies exploring the mechanism of action of the FPR2/ALX ligand LXA₄ (26, 27). Most important, the P38 inhibitor SB20380 blocked both 5-LOX nuclear localization (Fig. 2C and SI Appendix, Fig. S5) and AA-stimulated LTB₄ production (Fig. 2D) to a similar extent as RvD1, and the fact that SB20380 and RvD1 were not additive in blocking LTB₄ suggested that RvD1 and P38 were in the same pathway. As more direct proof, we showed that RvD1 decreased AA-stimulated phospho-Ser271-5-LOX (Fig. 3A). We then transfected 5-LOX-deficient (Alox5^−/−) macrophages with DYKDDDDK (FLAG)-tagged plasmids encoding either wild-type or S271A 5-LOX. Similar transfection efficiency for the two vectors was confirmed by anti-FLAG FACS (SI Appendix, Fig. S6). We found that S271A 5-LOX mimicked the ability of RvD1 to suppress nuclear 5-LOX and LTB₄ in AA-treated macrophages, and the effects of RvD1 and the mutant were nonadditive (Fig. 3 B and C). These data, along with the previously known role of nuclear 5-LOX in LTB₄ biosynthesis (17), suggest that RvD1 limits LTB₄ synthesis by blocking AA-stimulated P38-MK2 activation and the subsequent phosphorylation and nuclear localization of 5-LOX.

Fig. 2.

RvD1 decreases nuclear localization of 5-LOX through inhibition of p38-MAPKAPK2 (MK2) signaling. (A and B) BMDMs were preincubated with vehicle control or RvD1 and then incubated with 10 μM AA for 5 min. Cell lysates were immunoblotted for phospho- and total p38 or MK2 and quantified by densitometry (n = 3; mean ± SEM; ***P < 0.001 versus AA). (C) BMDMs treated similar to those in C were permeabilized and stained with Alexa-488 anti-5-LOX antibody (green) and counterstained with the nuclear stain DAPI (blue). Cells were viewed by confocal microscopy at 40× magnification; the macrophage in each image is outlined. (Scale bar = 5 µm.) Images were analyzed by ImageJ software for mean fluorescence intensity of nuclear versus nonnuclear 5-LOX staining in 5–7 cells per field (n = 3 separate experiments; mean ± SEM; *P < 0.05 or **P < 0.01 versus AA; n.s., nonsignificant). (D) BMDMs were incubated with 10 µM SB203580 or vehicle for 1 h before sequential RvD1 and AA treatment. LTB₄ was assessed by ELISA and expressed relative to the value for AA alone (n = 3; mean ± SEM; **P < 0.01 versus AA; n.s., nonsignificant).

Fig. 3.

AA-induced nuclear 5-LOX and LTB₄ is suppressed in macrophages expressing S271A-5-LOX. BMDMs incubated as in Fig. 1A. (A) Flow cytometric analysis for anti-phospho-Ser²⁷¹-5-LOX fluorescence (n = 3; mean ± SEM; *P < 0.05 versus AA). (B and C) BMDMs were transected with FLAG-tagged wild-type 5-LOX or S271A 5-LOX for 72 h before incubations. In B, the cells were then permeabilized and stained with Alexa-488 anti-Flag antibody (green), counterstained with the nuclear stain DAPI (blue), and viewed by confocal microscopy at 40× magnification. Representative cells, with the fluorescence image superimposed on the phase-contrast images and quantification of mean fluorescence intensity of nuclear versus nonnuclear FLAG staining of 5–7 cells per field are shown. (Scale bar = 5 µm.) In C, LTB₄ was assayed by ELISA and expressed relative to the value for AA alone. For B and C, n = 3, mean ± SEM; **P < 0.05 versus AA/wild-type; n.s., nonsignificant.

RvD1 Limits the P38/MK2/Nuclear 5-LOX/LTB₄ Pathway by Suppressing AA-Induced CaMKII Activation.

AA has been shown to rapidly increase intracellular calcium (Ca²⁺) (28), which prompted us to investigate whether a calcium-signaling protein might play a role in this pathway. We were particularly interested in CaMKII because CaMKII has been shown to activate P38 and MK2 in other settings (29, 30). Stimulation of macrophages with AA led to an increase in phospho-CaMKII, the enzyme's active form, and phospho-CaMKII was decreased by RvD1 in a FPR2/ALX-dependent manner (Fig. 4A). The FPR2/ALX ligand LXA₄ also blocked AA-stimulated phospho-CaMKII and phospho-MK2 (SI Appendix, Fig. S7). As a possible mechanism for suppression of CaMKII activation, we found that RvD1 suppressed the rise in cytosolic calcium that occurs with AA treatment of macrophages (SI Appendix, Fig. S8 A and C).

Fig. 4.

RvD1 suppresses the p38-MK2-LTB₄ pathway by inhibiting CaMKII. (A) BMDMs were treated as in Fig. 1A, except IgG control or blocking anti-ALX/FPR2 IgG were included; V, vehicle control. Cell extracts were immunoblotted for p-CaMKII and β-actin loading control and quantified by densitometry (n = 3, mean ± SEM; *P < 0.05 versus AA). (B) BMDMs from control (Camk2g^fl/fl) mice or Camk2g^fl/flLysMCre^+/− mice were incubated as in A (without blocking antibody), immunoblotted for p-MK2 and β-actin, and quantified by densitometry. (C) 5-LOX nuclear versus nonnuclear localization was assayed as in Fig. 2C. (Scale bar = 10 µm.) (D) LTB₄ was assessed by ELISA. For B–D, n = 3, mean ± SEM; **P < 0.05 versus AA-control; n.s., nonsignificant.

To show causation related to CaMKII, we studied macrophages from Camk2g^flflLysMCre^+/− mice, which lack the macrophage isoform, CaMKIIγ (SI Appendix, Fig. S9). The increase in phospho-MK2 by AA was significantly decreased in CaMKII-deficient macrophages compared with control Camk2g^flfl macrophages, indicating that CaMKII was upstream of MK2. Moreover, RvD1’s ability to limit phosphorylation of MK2 was abolished in CaMKII-deficient macrophages, suggesting that RvD1 acts in the same pathway as CaMKII (Fig. 4B). Most important, AA-stimulated 5-LOX nuclear localization and LTB₄ production were decreased in macrophages lacking CaMKII, and the suppressive actions of RvD1 and CaMKII deficiency were not additive (Fig. 4 C and D). These data support the premise that RvD1 signals through CaMKII to limit LTB₄ production.

To further prove causation, we transduced macrophages with an adenovirus expressing constitutively active T287D CaMKII (CA-CaMKII), which mimics autophosphorylation and is thus autonomously activated in the absence of bound calcium (31). Adenoviral transfection itself did not alter the pathway, as RvD1 blocked AA-stimulated MK2 phosphorylation in adeno-LacZ control cells to a similar degree as in nontransduced wild-type macrophages (SI Appendix, Fig. S10A and earlier). In contrast to the situation with wild-type macrophages, RvD1 was unable to block AA-stimulated MK2 activation, 5-LOX nuclear localization, and LTB₄ production in CA-CaMKII-transduced macrophages (SI Appendix, Fig. S10 A–C), suggesting that RvD1 acts by blocking CaMKII activation.

RvD1 also blocked AA-stimulated phospho-MK2 and LTB₄ generation in human monocyte-derived macrophages, and this effect was dependent on both FPR2/ALX and an RvD1 receptor unique to human cells, GPR32 (SI Appendix, Fig. S11 A–C). Moreover, human macrophages transduced with dominant negative K43A CaMKII (31) exhibited diminished AA-stimulated LTB₄ generation that was not further decreased by RvD1 (SI Appendix, Fig. S11D). These data suggest similar pathways of RvD1-mediated LTB₄ suppression in mouse and human macrophages.

The RvD1-CaMKII-p38-MK2 Pathway Is Functional in Vivo.

To examine whether the RvD1-LTB₄ signaling pathway is operable in vivo, we used a model of acute zymosan-induced peritoneal inflammation in which controlled release of LTB₄ by resident macrophages in the early stages of inflammation is a critical determinant for swift resolution (18). As predicted, RvD1 treatment decreased zymosan-induced LTB₄ generation (SI Appendix, Fig. S12C) and polymorphonuclear neutrophil (PMN) infiltration (SI Appendix, Fig. S12D) (9). Most important, these events were accompanied by a decrease in phospho-p38 (SI Appendix, Fig. S12A) and phospho-MK2 (SI Appendix, Fig. S12B) in exudate macrophages.

We then compared these responses in Camk2g^flflLysMCre^+/− versus Camk2g^flfl mice and found that macrophage-CaMKII deficiency led to decreases in both zymosan-stimulated LTB₄ and PMN infiltration in a manner that was not additive with RvD1 (Fig. 5 A–C). Note that 5-LOX protein levels were not different between the two groups of mice (SI Appendix, Fig. S13). To investigate whether the suppression of LTB₄ and PMNs by RvD1 was through phosphorylation of 5-LOX at Ser271 in vivo, we injected Alox5^−/− mice with plasmids encoding wild-type 5-LOX or S271A 5-LOX. Flow cytometry verified successful transfection of peritoneal macrophages (SI Appendix, Fig. S14). We found that S271A 5-LOX-transfected mice exhibited significantly reduced LTB₄ generation and PMN infiltration and that this decrease was not additive with RvD1 (Fig. 5 D and E). These combined data demonstrate that RvD1 limits excessive LTB₄ production in vivo by suppressing CaMKII activation and 5-LOX phosphorylation.

Fig. 5.

Evidence for the RvD1–CaMKII–p-5-LOX pathway in vivo. (A) Representative chromatograms of LTB₄ and 6-trans-LTB₄ (335 > 195) in inflammatory exudates of control (Ctrl), Camk2g^fl/fl (Ctrl), or Camk2g^fl/flLysMCre^+/− mice 2 h after i.p. zymosan. (B and C) Quantification of LTB₄ and PMNs in inflammatory exudates of the indicated groups of mice treated with zymosan in the presence or absence of 10 ng i.v. RvD1 (for LTB₄). (D and E) Alox5^−/− mice were transfected with plasmids encoding wild-type or S271A 5-LOX 42 h before sequential RvD1 and zymosan treatments and assayed for LTB₄ and PMNs. For B, n = 3–5/group, mean ± SEM; *P < 0.05 versus both control groups; for C–E, n = 5 mice/group, mean ± SEM; *P < 0.05 or **P < 0.01 versus Zym/wild-type; n.s., nonsignificant; WT, wild-type.

Discussion

An important process in inflammation resolution is the dampening of proinflammatory molecules in a manner that does not compromise host defense (2). Thus, it is essential to understand at a molecular-cellular level how this critical process is achieved. We provide here a new pathway (SI Appendix, Fig. S15) that applies to a previously recognized example of SPM-mediated inflammation suppression; namely, the ability of RvD1 to decrease LTB₄ levels (9). Moreover, we show here that RvD1 increases the level of LXA₄, which is consistent with the hypothesis that intracellular 5-LOX localization affects the balance of LTB₄ and LXA₄. In this context, the most likely scenario is that LXA₄ is released and then acts in a paracrine and autocrine manner via FPR2/ALX.

Our results indicate that the target of RvD1 in AA-treated macrophages is a new CaMKII pathway that promotes p38-MK2 activation and LTB₄ production. RvD1 suppressed CaMKII activation, most likely by blocking calcium entry into the cytosol, as RvD1 blocked the increment in cytosolic calcium effected by AA, ATP, and fMet-Leu-Phe (SI Appendix, Fig. S6). In this regard, RvD1 was recently reported to block histamine-stimulated intracellular calcium in goblet cells in a receptor-dependent manner (32), and the SPM resolvin E1 was shown to block intracellular calcium initiated by the antimicrobial peptide LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES (LL-37) in PMNs (33). Interestingly, another study showed that LL-37 stimulated p38 and LTB₄ formation in PMNs (34), although the signal transduction pathway leading to p38 activation was not elucidated. These findings, together with the data herein, suggest that the ability of SPMs to rapidly and transiently limit intracellular calcium and CaMKII-p38-MK2 signaling may be a fundamental mechanism for preventing excessive or prolonged inflammation (33, 35).

CaMKII has been shown to mediate other cellular processes in diseases in which chronic inflammation is important, raising the possibility of additional beneficial effects of RvD1. For example, CaMKII triggers endoplasmic reticulum-stress-induced apoptosis in macrophages (36), which could be important in diseases in which leukocyte apoptosis and secondary necrosis underlie the pathology, such as advanced atherosclerosis and certain autoimmune diseases (37, 38). Moreover, in the setting of obesity, CaMKII promotes excessive hepatic glucose production and impairs hepatic insulin signaling by activating a p38-MK2-mediated pathway (29, 30). Although the action of RvD1 in hepatocytes remains to be explored, RvD1 has been shown to promote insulin sensitivity in diabetic mice, in part by enhancing insulin signaling (39).

In humans, the ratio of proresolving LXA₄ to proinflammatory LTB₄ is balanced when inflammation is properly controlled (7, 18), whereas this balance is skewed toward LTB₄ in certain chronic inflammatory diseases (40–42). Although the mechanisms of excess LTB₄ production in these diseases remain to be elucidated, it is possible that defective RvD1 levels or activity play a role through the mechanisms elucidated in this report. For example, atherosclerosis, a disease characterized by failed inflammation resolution (43), is associated with excessive production of LTB₄ (11), and 5-LOX has been shown to be located in the nuclear region of macrophages in human atherosclerotic lesions (44).

In summary, our results provide a mechanistic understanding of how RvD1 carries out one of its most important proresolving functions; namely, suppressing LTB₄ production while boosting LXA₄ synthesis. The new pathway elucidated in this report suggests that therapeutic administration of RvD1 and possibly other SPMs may be particularly beneficial for inflammatory diseases in which excessive CaMKII-p38-MK2 activation or LTB₄ underlies the pathology.

Materials and Methods

LTB₄ Detection in Vitro.

Bone marrow-derived macrophages (BMDMs) were harvested from female C57BL/6J mice (6–8 wk of age) and cultured in DMEM, 10% (vol/vol) FBS, 20% (vol/vol) L cell media containing macrophage colony stimulating factor, glutamine, and penicillin/streptomycin for 7 d. For individual experiments, 3 × 10⁶ macrophages in 300 μL PBS containing calcium and magnesium were incubated with 1 nM RvD1 for 20 min at 37 °C (45), followed by AA (10 μM) or LTA₄ (10 μM) stimulation. After 40 min, the cells were placed on ice and the media were subjected to ELISA analysis. For FPR2/ALX receptor experiments, IgG or anti-FPR2 (10 μg/incubation, 37 °C, 1 h) was added before RvD1 stimulation.

Confocal Microscopy for Intracellular Localization of 5-LOX.

BMDMs were plated on 8-well coverslips (LabTek) and incubated under various conditions, as described in the figure legends. After addition of 5% cold formalin, BMDMs were incubated for 60 min at 4 °C with permeabilization buffer (cat. no. 554715, BD Biosciences) containing anti-5-LOX antibody. Excess antibody was then removed, and the cells were incubated with Alexa 488 anti-rabbit-IgG for an additional 30 min at 4 °C. The cells were counterstained with Hoechst to identify nuclei, viewed on a Nikon A1 confocal microscope, and analyzed using ImageJ software.

Zymosan A-Stimulated Peritonitis.

Six- to 8-wk-old female mice were administered 10 ng RvD1 per mouse by i.v. injection. After 15 min, 200 μg zymosan A per mouse was injected i.p. to induce peritonitis for 2 h, as in ref. 9. All procedures were conducted in accordance with the Columbia University Standing Committee on Animals guidelines for animal care (protocol no. AC-AAAF7107).

In Vivo Transfection.

Plasmids (10 μg) were incubated with 16 mL Jet-Pei-Man in vivo transfection reagent (PolyPlus Transfection; cat. no. 203–10G) for at least 30 min at room temperature. These transfection complexes (1 mL) were injected i.p. into 6–8-w-old female Alox5^−/− mice (Jackson Laboratories). After 42 h, peritonitis experiments were conducted.

Identification and Quantification of Eicosanoids by Liquid Chromatography-MS/MS.

Lipid mediators of interest were profiled using an HPLC system (Shimadzu Prominence) equipped with a reverse-phase (C18) column (4.6 × 50 mm; 5.0 μm particle size) coupled to a triple quadrupole mass spectrometer (AB Sciex API2000), which was operated in negative ionization mode. Multiple reaction monitoring was used to identify and quantify LTB₄ (335 > 195), 6-trans-LTB₄ (335 > 195), and LXA₄ (351 > 115) (46). (For detailed methods see SI Appendix, Methods.)

Statistical Analysis.

Results are presented as means ± SEM. Differences between two groups were compared by paired Student t test or one-way ANOVA after normality testing. P < 0.05 was considered significant.