Homer1a attenuates glutamate-induced oxidative injury in HT-22 cells through regulation of store-operated calcium entry

Calcium disequilibrium is extensively involved in oxidative stress-induced neuronal injury. Although Homer1a is known to regulate several neuronal calcium pathways, its effects on, or its exact relationship with, oxidative stress-induced neuronal injury has not yet been fully elucidated. We found that Homer1a protected HT-22 cells from glutamate-induced oxidative stress injury by inhibiting final-phase intracellular calcium overload and mitochondrial oxidative stress. In these cells, stromal interactive molecule 1 (STIM1) puncta, but not the protein level, was significantly increased after glutamate treatment. Store-operated calcium entry (SOCE) inhibitors and cells in which a key component of SOCE (STIM1) was knocked out were used as glutamate-induced oxidative stress injury models. Both models demonstrated significant improvement of HT-22 cell survival after glutamate treatment. Additionally, increased Homer1a protein levels significantly inhibited SOCE and decreased the association of STIM1-Orai1 triggered by glutamate. These results suggest that up-regulation of Homer1a can protect HT-22 cells from glutamate-induced oxidative injury by disrupting the STIM1-Oria1 association, and then by inhibiting the SOCE-mediated final-phrase calcium overload. Thus, regulation of Homer1a, either alone or in conjunction with SOCE inhibition, may serve as key therapeutic interventional targets for neurological diseases in which oxidative stress is involved in the etiology or progression of the disease.


Increased levels of Homer1a attenuated glutamate-induced HT-22 cell injury. To test the role of
Homer1a in glutamate-induced HT-22 injury, we first analyzed the protein levels of Homer1a and Homer1b/c at 1, 6, 12 and 24 h after glutamate treatment. As shown in Fig. 1A,B, the levels of Homer1b/c protein did not change after glutamate treatment. In contrast, the protein levels of Homer1a showed significant elevation at 6, 12 and 24 h after the glutamate treatment (Fig. 1A,C).
We then constructed lentivirus-based shRNA (D/E) and overexpression (O/E) vectors to regulate the protein levels of Homer1a. As shown in Fig. 1D, transfection of Homer1a D/E vectors for 48 h decreased Homer1a protein levels by approximately 75% compared with that of the control + glutamate and vector + glutamate groups 6 h after glutamate treatment. Furthermore, transfection of Homer1a O/E vectors for 48 h significantly up-regulated the Homer1a protein levels by approximately 3 fold as compared with that of the control + glutamate and vector + glutamate groups. Since no significant changes were observed between the Scr-shRNA group and the vector group, the Scr-shRNA group has not been shown in our experiments.
Consistent with previous studies, glutamate treatment caused a dose-dependent decrease in cell viability, with a 38.55% decrease in cell viability at 5 mM, and a 64.56% decrease at 10 mM (Fig. 1E). Cells that were not treated with glutamate were included as non-injured controls. Corresponding results were obtained with an LDH releasing assay (Fig. 1F). The decrease in Homer1a protein levels with the D/E vector significantly mitigated glutamate-induced HT-22 cell injury as evidenced by a further reduction in cell viability (Fig. 1E) and a further increase in LDH release (Fig. 1F) at both concentrations. In contrast, increased levels of Home1a protein from using the O/E vector significantly improved HT-22 cell survival and mitigated LDH release against glutamate treatment (Fig. 1E,F). Taken together, these results indicate that increased levels of Homer1a protein attenuated glutamate-induced HT-22 cells injury.

Increased Homer1a protein levels attenuated glutamate-induced calcium overload and mitochondrial oxidative levels. Previous evidence has indicated that both intracellular ROS accumulation
and Ca 2+ overload were the main causes of glutamate-induced HT-22 cell injury. Therefore, to understand the mechanism of Homer1a-mediated neuroprotection, we measured intracellular calcium dynamics and mitochondrial oxidative levels using live-cell imaging of cells transfected with either a genetically encoded Ca 2+ indicator construct (pCMV-B-GECO1) or a mitochondrial ROS reporter construct (pMitotimer). As shown in Fig. 2A, when compared with that of non-treated control group, glutamate treatment significantly increased intracellular calcium levels in the control + glutamate and vector + glutamate groups; however, this change was not apparent in the first 4 h after glutamate treatment. Homer1a O/E significantly attenuated glutamate-induced intracellular Ca 2+ levels ( Fig. 2A), as compared with that of the control + glutamate or the vector + glutamate groups. In contrast, decreased Homer1a protein levels using the D/E vector significantly increased intracellular Ca 2+ levels at 6, 7, and 8 h after glutamate as compared to the control + glutamate or vector + glutamate groups.
Oxidative stress can also cause mitochondrial damage, so we next used fluorescent microscopy to examine the oxidative levels of the mitochondria in these cells. Under physiological conditions, mitochondria in the control group exhibited a low red/green ratio (Fig. 2B,C, column 1). However, 12 h after glutamate insult, the mitochondria in the control + glutamate and vector + glutamate groups appeared high oxidative stress in these cells as evidenced by high red/green ratio (Fig. 3B,C, Column 2 and 3). Decreased levels of Homer1a protein in the Scientific RepoRts | 6:33975 | DOI: 10.1038/srep33975 cells transfected with the D/E vector indicated a significant increase of oxidative levels as compared to the control + glutamate and vector + glutamate groups ( Fig. 3B; evidenced by a higher red/green ratio). Additionally, increased Homer1a protein levels in the cells transfected with the O/E vector significantly alleviated the oxidative stress ( Fig. 3B,C, Column 5) as compared to the control + glutamate and vector + glutamate groups. From these results, we conclude that Homer1a may protect against glutamate-induced HT-22 cell injury by restoring intracellular calcium homeostasis and reducing mitochondrial oxidative stress.

Glutamate-induced HT-22 injury redistributed STIM1 into puncta within the cell. Since HT-22
cells lack iGluRs, SOCE might be the main source of calcium influx in these cells. Thus, we further evaluated the change of two components of SOCE, namely STIM1 and Orai1, in this cell model system. As shown in Fig. 3A,B, the levels of STIM1 protein did not change after glutamate treatment. Additionally, although the expression of Orai1 seemed to slightly increase 6 h after injury, these changes were not statistically significant (Fig. 3A,C).
Previous studies have indicated that the cellular distribution of SOCE components greatly affected SOCE function. Therefore, we visualized the change of STIM1 distribution by transfecting the STIM1-YFP and ER.E2-Crimson constructs. Forty-eight h after transfection, the cells were treated with glutamate, or not, for 6 h. As shown in the Fig. 3D, the STIM1-YFP showed a distributed dispersion pattern, with predominant localization in the ER under normal physiological conditions. However, at 6 h after the glutamate treatment, the STIM1-YFP was differently distributed and appeared as STIM1-YFP puncta, suggesting that glutamate promotes STIM1 oligomerization. Compared with the control group, the STIM1-YFP puncta were significantly increased at 6 h the after glutamate treatment (Fig. 3E). We also detected the distribution of Orai1 within the cell using an Orai1-YFP construct, finding that Orai1 primarily distributed on the cell membrane, and no apparent redistribution of Orai1 after glutamate treatment was observed (data not shown). These results led us to conclude that glutamate treatment induces a redistribution of STIM1 into puncta within the cell.

Inhibition of SOCE mitigated glutamate-induced HT-22 cells injury.
To further elucidate the role of SOCE in glutamate-induced HT-22 cells injury, we altered STIM1 levels using lentivirus-based STIM1-shRNA or STIM1-CRIPSR/Cas9 vectors to either down-regulate (D/E) or knockout (KO) the expression of STIM1, respectively. Consistent with our previous studies, STIM1-shRNA significantly decreased STIM1 protein levels Twentyfour h after seeding the cells into 60-mm plates, the HT-22 cells were exposed to glutamate (10 mM) for either 1, 6, 12 or 24 h. The proteins levels of Homer1b/c and Homer1a were evaluated by western blotting (A). The relative expressions of Homer1b/c (B), and Homer1a (C) were calculated as percent of the optical density of the control group. After infection with either Homer1a shRNA or Homer1a O/E lentiviral vectors, the HT-22 cells were exposed to glutamate (5 mM) for 6 h, after which the proteins were harvested for western blotting assay (D). Forty-eight h after lentiviral infection, followed by 24 h of glutamate treatment, cellular viability was assessed by CCK-8 assay as a percentage of control group (E), and the cell injury was detected by LDH release assay (F). HT-22 cells in the control group were not subjected to lentiviral infection or glutamate injury. Data are presented as mean ± SEM from four experiments; *p < 0.05. # p < 0.05 vs. control + glutamate or vector + glutamate group. by 80.27% as compared to the control group (Fig. 4A). Thirty-six h after transfection, at a low MOI of 5, the levels of STIM1 protein were notably decreased in EGFP-positive HT-22 cells transfected with STIM1-shRNA group as compared with that of EGFP-positive HT-22 cells transfected with scrambled shRNA (Scr-shRNA) or EGFP-negative non-transfected HT-22 cells (Fig. 4B, Row 1). Additionally, three different sgRNAs were used to KO the expression of STIM1. As shown in Fig. 3A, sgRNA #1 had the highest efficacy, with STIM1 protein levels almost undetectable even after overexposure of the western blot. Similar results were obtained using immunofluorescence staining (Fig. 4B, Row 2).
To further understand the role of SOCE in our injury model, we treated the cells with SOCE inhibitors. As shown in Fig. 4C, in the absence of extracellular Ca 2+ , treatment of the control cells with TG induced a significant calcium release, followed by an immediate calcium entry after restoration of extracellular Ca 2+ , indicating the existence of SOCE in HT-22 cells. Cells treated with STIM1 D/E, STIM1-KO, 2-APB, SKF96365, and 5J-4 all had a significant decrease in SOCE-mediated calcium entry, as evidenced by the reduction in calcium entry peak value (Fig. 4C,D) and area under curve (AUC, Fig. 4C,E). However, cells treated with STIM1-shRNA were not as successful at attenuating SOCE-mediated calcium entry as the STIM-KO and SOCE inhibitors. No significant change of TG-induced calcium release was observed among all groups (Fig. 4C-E). Finally, we evaluated the neuroprotection of SOCE inhibition against glutamate-induced HT-22 cell injury. After glutamate treatment, decreased cell viability (Fig. 4F) and increased LDH release (Fig. 4G) were observed in cells transfected with vector alone as compared to untreated controls. This effect was significantly attenuated in STIM1-KO cells and cells treated with either 2-APB, or 5J-4; however, cells transfected with STIM1-shRNA and cells treated with SKF-96365 did not demonstrate a significant improvement in either measure. Taken together, these results demonstrate that SOCE might be a key mediator of glutamate-induced HT-22 cell injury.

Up-regulation of Homer1a reduced TG-induced SOCE-mediated calcium entry.
After determining the roles of Homer1a up-regulation and SOCE inhibition in glutamate-induced HT-22 cell injury, we further investigated the effect of Homer1a on SOCE. Transfection with the Homer1a shRNA vector significantly increased TG-induced calcium release, as evidenced by an increased AUC (Fig. 5A,C); however, this was not seen with the peak value (Fig. 5A,B). No significant change of calcium release was observed among the other three groups examined, including the control cells, cells + vector or cells transfected with the Homer1a O/E vector ( Fig. 5A-C). SOCE-mediated calcium entry was also affected by decreasing Homer1a levels. As shown in Fig. 5A,C, cells transfected with the Homer1a shRNA vector resulted in a slight, but significant, intracellular calcium increase after extracellular calcium was restored; as seen with the calcium release, the peak value of calcium entry was not significantly different from the controls (Fig. 5A,B). In contrast, HT-22 cells treated with the Homer1a O/E vector had a significant decrease both in the peak value (Fig. 5A,B) and the AUC (Fig. 5A,C) as compared with the control cells or cells transfected with vector only. These results demonstrated that an increase in Homer1a levels reduced TG-induced SOCE-mediated calcium entry. Up-regulation of Homer1a decreased glutamate-induced STIM1-Orai1 interaction. After confirming the inhibition of Homer1a on SOCE-mediated calcium entry, we wanted to further elucidate how Homer1a might affect SOCE machines. In particular, we examined the interaction of Homer1a with both STIM1 and Orai1, two of the main components of SOCE. Co-IP and subsequent western blotting were conducted using normal HT-22 cells and HT-22 cells stimulated with TG (2.5 μ M) for 5 min. Under physiological conditions, we detected an interaction of these three proteins (STIM1, Homer1a, and Orai1; Fig. 6A-C, Columns 1 and 3). After treatment with TG, there was a significant decrease in the interaction of Homer1a with STIM1 (Fig. 6A,B, Column 2) and a significant increase of the interaction of STIM1 with Orai1 ( Fig. 6A,C, Column 4) as compared with untreated control HT-22 cells. However, significant change of Homer1a-Orai1 interaction was not induced by TG treatment (Fig. 6A, Columns 1 and 2; 5B, Columns 3 and 4). Additionally, immunofluorescence staining showed that partial co-localization of Homer1a and STIM1 also existed under physiological conditions (Fig. 6D). These results suggest that some associations existed among Homer1a, STIM1, and Orai1, either directly or indirectly. Next, we detected the effects of Homer1a D/E or O/E on STIM1-Orai1 association 6 h after glutamate treatment. Cells transfected with the Homer1a shRNA vector slightly, but not significantly, increased STIM1-Orai1 association (Fig. 6E, Column 2; 6F, Column 5). In contrast, cells transfected with the Homer1a O/E vector significantly reduced STIM1-Orai1 associations (Fig. 6E, Column 3; 6F, Column 6). Taken together, these data indicate that Homer1a might play a negative role in regulating STIM1-Orai1 interaction.

Discussion
The present studies have established a role of Homer1a in oxidative stress injury, negatively affecting STIM1-Orai1 associations, inhibiting SOCE-mediated final-phrase calcium influx, and even alleviating glutamate-induced HT-22 cell death. After treating HT-22 cells with glutamate, the protein levels of Homer1a increased, while the STIM1 localization into puncta within the cell occurred without any apparent change in either STIM1 or  Orai1 protein levels. Both increasing Homer1a levels through transfection of the O/E vector and inhibiting SOCE with either SOCE inhibitors or CRISPR/Cas9-mediated STIM1 KO significantly protected HT-22 cells from glutamate-induced oxidative stress injury. Furthermore, TG-induced SOCE-mediated calcium influx and glutamate-provoked STIM1-Orai1 association was attenuated by Homer1a O/E. Therefore, given the role of calcium homeostasis disturbance in oxidative stress injury, therapeutic strategies that can increase Homer1a protein levels and/or inhibit SOCE might serve as potential interventions for neurological diseases in which oxidative stress is involved in either the disease etiology or progression.
Homer1a, as an immediate early gene (IEG), is dynamically regulated by various stimuli, although the mechanisms involved has not been fully elucidated 28 . In line with our previous studies 12,27,29 , our current study found Homer1a protein levels were significantly increased, peaking 6 h after glutamate treatment. Furthermore, the increased levels of Homer1a attenuated glutamate-induced HT-22 cell death, which agrees with other studies in chronic inflammatory pain, traumatic brain injury, and some neuropsychiatric disorders 12,16,30 . Together, these studies suggest that the role of Homer1a in neuroprotection is to prevent detrimental or lethal injury to cells. Furthermore, our study found that the neuroprotection of Homer1a was attributed to its ability to inhibit final-phrase calcium overload and reducing mitochondrial oxidative levels, which is in line with our previous study in H 2 O 2 -induced PC12 cell injury 27 . Similar results have also been reported in which the inhibition of calcium influx, by using CoCl 2 or the use of a Ca 2+ free buffer alleviated glutamate-induced HT-22 cells death [9][10][11] . However, the exact mechanism by which Homer1a reduces mitochondrial oxidative levels were not investigated within the scope of this study, and would be warranted in a future study. SOCE, a crucial regulating mechanism of cellular calcium homeostasis, is dynamically regulated 21 , and previous studies have demonstrated that SOCE is involved in oxidative stress injury 22 . For example, Henke, N. et al. found that Orai1 mediates the detrimental Ca 2+ influx in programmed cell death induced by glutathione depletion or glutamate treatment 8 . Some other investigators found inhibition of STIM1 limits lipopolysaccharide-induced vascular inflammation 25 , alleviated 6-hydroxydopamine-induced oxidative stress 31 , and relieved hydrogen peroxide-induced oxidative injury 26 . In this study, we found changes in the localization of STIM1 after glutamate treatment, with an increase in the presence of STIM1 puncta despite the fact that there was not a detectable change in STIM1 protein levels. This suggests that STIM1 oligomerized after the glutamate treatment. Similar changes in STIM1 protein localization were obtained in COS7 cells after buthionine sulfoximine or H 2 O 2 treatment 32 . In line with this data, a notable increase of [Ca 2+ ] cyt was observed at this timepoint in long-term calcium imaging; however, there was not a significant change in either Orai1 protein levels or its cellular distribution (data not shown). Taken together, these studies suggested that glutamate provoked the redistribution of STIM1, and may cause STIM1-mediated calcium influx. Furthermore, inhibition of SOCE by STIM1 KO, or SOCE inhibitors (2-APB and 5J-4) significantly inhibited TG-induce calcium entry and promoted cell survival against glutamate. However, although treatments of STIM1 D/E and SKF96365 affect TG-induced calcium entry, no significant neuroprotection was observed. This result of STIM1 D/E was partially in agreement with Henke, N. et al.'s study 8 ; however, our study found that STIM1 D/E was significantly worse at inhibiting TG-induced SOCE, as compared with that of STIM1 KO. One possible reason for these differences seen between the STIM1 D/E and the STIM1 KO might due to residual STIM1 present in cells transfected with the STIM1 D/E vector only, which were sufficient to mediate glutamate-induced calcium influx. In previous studies, SKF96365 was used as a non-special inhibitor of SOCE. Additionally, some off-targets also have been found, like inducing cell apoptosis 33 . Our study in HT22 cells also found that SKF96365 significantly alleviated H 2 O 2 -induced necrosis, but notably increased H 2 O 2 -induced apoptosis (data not shown). This apoptosis might compromise the neuroprotection of SKF96365-mediated SOCE inhibition. However, more efforts should be made in order to clarify the mechanisms involved of SKF96365's off-target functions.
Homer1b/c has previously been shown to alter both the SOCE machinery and Ca 2+ release-activated Ca 2+ channels 18,20,34 . However, the role of Homer1a has not been clearly defined. Homer1a has been considered as a negative competitor of Homer1b/c due to its lack of a CC domain in C-terminal, which prevents it from self-multimerizing. The present study supported this, as the up-regulation of Homer1a resulted in decreased TG-induced calcium influx and negatively regulated SOCE. Additionally, the associations among STIM1, Homer1a, and Orai1 found in this study are similar to the previously identified associations between STIM1 and Homer1b/c or among STIM1, Homer1b/c, and Orai1 19,20 .
Previous studies have established that Homer1 (including Homer1a and Homer1b/c) interacts with other proteins via its EVH1-like domain, which specifically binds to the proline-rich sequences of PPXXF (where X is any amino acid) 17 , a similar sequences of PXXF in TRPC (LP(P/X)PFN) and phosphoinositide 3 kinase enhancer L (PKPF) [35][36][37] , or a LPSSP sequence 35 . Jardin, I. et al. found that TG and thrombin enhanced the Ca 2+ -dependent interactions among STIM1, Homer1b/c, and Orai1 in human platelets 20 . Likewise, Dionisio, N. et al.'s study has indicated that TG induces interactions of Homer1 with STIM1 and the Ca v 1.2 α 1 subunit 19 . In this study, we found that in response to TG-induced calcium store depletion, the association between STIM1 and Orai1 increased, while the association between STIM1 and Homer1a decreased, while no apparent change of Homer1a-Orai1 association was observed. The lack of change in the Homer1a-Orai1 association might be due to the low expression level of Homer1a at resting conditions. However, upon examination of the amino acid sequences of STIM1 and Orai1, while a similar sequence (NPAHFI) in the C-terminal of STIM1 could be identified, no such sequence could be identified in Orai1. Furthermore, up-regulation of Homer1a significantly reduced glutamate-inducing STIM1-Orai1 association. Therefore, we speculated that Homer1a might dissociate the SOCE complex, and down-regulate SOCE by negatively competing with Homer1b/c.
In conclusion, this study provided experimental evidence that Homer1a is involved in glutamate-induced HT-22 cell death. Overexpression of Homer1a demonstrated a neuroprotective role for this protein in glutamate-induced oxidative stress injury by dissociating the STIM1-Orai1 association and inhibiting glutamate-induced final-phase calcium overload. These results partially uncover the role of Homer1a on SOCE and in oxidative stress injury, Scientific RepoRts | 6:33975 | DOI: 10.1038/srep33975 and might contribute to the development of new therapeutic strategies. Future studies should determine the exact mechanisms involved in Homer1a-mediated reduction in mitochondrial oxidative level, as well as determine the exact interacting targets and other potential binding partner molecules using either mass spectrometry or point mutations.
Western blot. Western blots were performed as previously described 29  to standardize calcium signal values. The values were then plotted against time and shown as F/F 0 (F, fluorescence value at any time; F 0 , fluorescence value at the baseline).

Detection of mitochondrial reactive oxygen level ([ROS] Mito ).
In order to detect [ROS] mito , the HT-22 cells were transfected with the plasmid pMitotimer (Addgene, Cambridge, MA, USA), an engineered reporter gene used to assess mitochondrial content, structure, stress, and damage 39 , using Lipofectamine LTX-plus (Invitrogen, Carlsbad, CA, USA). Images were acquired at the following excitation/emission wavelengths (nm): pMitotimer green (488/518) and red (543/572), and analyzed using Image J (National Institutes of Health, Bethesda, MD, USA). The level of [ROS] mito was reported as the ratio of red/green as previously described 39 .
Detection of STIM1 and Orai1 distribution. To visualize the distribution of SOCE components within the cells, HT-22 cells were transfected with STIM1-YFP, ER.E2-Crimson or Orai1-YFP (Addgene, Cambridge, MA, USA) using Lipofectamine LTX-plus (Invitrogen, Carlsbad, CA, USA) for 24 h, and then subjected to treatments as described above [40][41][42] . The localization of STIM1 and ER was captured using a confocal microscope and analyzed using Image J software (National Institutes of Health, Bethesda, MD, USA). The level of STIM1 puncta was calculated as the average/per cell.
Immunofluorescence staining. After respective treatments, the HT-22 cells were fixed with 4% paraformaldehyde for 10 min, washed with phosphate-buffered saline (PBS), and permeabilized with 0.2% Triton X-100. Next, the HT-22 cells were incubated with primary antibodies overnight at 4 °C as follows: Homer 1a (1:25, Santa Cruz Biotechnology, Santa Cruz, CA, USA) and STIM1 (1:200, CST, Danvers, MA, USA) after blocking with horse serum. Cells were then incubated with Alexa Fluor ® 488 or 594-conjugated anti-goat or anti-rabbit secondary antibodies, respectively (1:400, Life Technologies, Carlsbad, CA, USA), for 2 h at room temperature. After washing the cells with PBS three times, the cells were stained with Hoechst stain to visualize the nuclei (Sigma-Aldrich, St. Louis, MO, USA). Cells were imaged using a confocal microscope (FV10i, Olympus, Tokyo, Japan) in XY-or XYZ-plane fashion. Identical light sensitivity and laser power were used among all groups.

Co-immunoprecipitation (Co-IP).
To evaluate protein interactions, Co-IP experiments were performed using the Pierce ™ Crosslink Magnetic IP/Co-IP Kit (Pierce, Thermo Fisher Scientific Inc., Waltham, MA, USA) according to the manufacturer's instructions. HT-22 cells were cultured in 100-mm dishes, and then harvested in ice-cold lysis/wash buffer supplemented with a proteinase inhibitor cocktail (Roche Applied Bioscience, Indianapolis, IN, USA). The lysate was centrifuge at 13,000 g for 15 min at 4 °C. Protein concentrations in the extracts were determined using a BCA protein assay kit (Pierce, Thermo Fisher Scientific Inc., Waltham, MA, USA). Magnetic beads were crosslinked with non-specific rabbit or goat IgG (2 mg), poly-clonal rabbit anti-STIM1 (1:50; CST, Danvers, MA, USA), or goat anti-Homer1a (1:20; Santa Cruz Biotechnology, Santa Cruz, CA, USA), and then the beads were washed two times with coupling buffer. The protein extracts were combined with the beads and incubated overnight at 4 °C. Following magnetic isolation, the precipitates were washed three times with wash buffer, eluted with the elution buffer, neutralized with neutralization buffer, and prepared for western blotting.
Statistical analysis. Data are expressed as a mean ± SEM, and analyzed by one-way analysis of variance (ANOVA) followed by Bonferroni's multiple comparisons among more than two groups or by unpaired t-test comparison between two groups. All statistical analyses were carried out with GraphPad Prism 6 (GraphPad, San Diego, CA, USA). p < 0.05 was considered as the threshold for significant difference.