Classical and Alternative Activation of Rat Microglia Treated with Ultrapure Porphyromonas gingivalis Lipopolysaccharide In Vitro

The possible relationship between periodontal disease resulting from the infection of gingival tissue by the Gram-negative bacterium Porphyromonas gingivalis (P. gingivalis) and the development of neuroinflammation remains under investigation. Recently, P. gingivalis lipopolysaccharide (LPS) was reported in the human brain, thus suggesting it might activate brain microglia, a cell type participating in neuroinflammation. We tested the hypothesis of whether in vitro exposure to ultrapure P. gingivalis LPS may result in classical and alternative activation phenotypes of rat microglia, with the concomitant release of cytokines and chemokines, as well as superoxide anion (O2−), thromboxane B2 (TXB2), and matrix metalloprotease-9 (MMP-9). After an 18-h exposure of microglia to P. gingivalis LPS, the concentration-dependent responses were the following: 0.1–100 ng/mL P. gingivalis LPS increased O2− generation, with reduced inflammatory mediator generation; 1000–10,000 ng/mL P. gingivalis LPS generated MMP-9, macrophage inflammatory protein 1α (MIP-1α/CCL3), macrophage inflammatory protein-2 (MIP-2/CXCL2) release and significant O2− generation; 100,000 ng/mL P. gingivalis LPS sustained O2− production, maintained MMP-9, tumor necrosis factor-α (TNF-α), and interleukin-6 (IL-6) release, and triggered elevated levels of MIP-1α/CCL3, MIP-2/CXCL2, and cytokine-induced neutrophil chemoattractant 1 (CINC-1/CXCL-1), with a very low release of lactic dehydrogenase (LDH). Although P. gingivalis LPS was less potent than Escherichia coli (E. coli) LPS in stimulating TXB2, MMP-9, IL-6 and interleukin 10 (IL-10) generation, we observed that it appeared more efficacious in enhancing the release of O2−, TNF-α, MIP-1α/CCL3, MIP-2/CXCL2 and CINC-1/CXCL-1. Our results provide support to our research hypothesis because an 18-h in vitro stimulation with ultrapure P. gingivalis LPS resulted in the classical and alternative activation of rat brain microglia and the concomitant release of cytokines and chemokines.

The purpose of our investigation was to experimentally test our hypothesis that the exposure of neonatal rat microglia to ultrapure P. gingivalis LPS in vitro would result in pro-inflammatory/classical and/or anti-inflammatory/alternatively microglia activation, and the release of pro-inflammatory and anti-inflammatory mediators. Our data provide strong experimental evidence for the proposed working hypothesis, because ultrapure P. gingivalis LPS activated both pro-inflammatory/classical and/or anti-inflammatory/alternatively microglia phenotypes in vitro, and, while less potent than Escherichia coli (E. coli) LPS in stimulating TXB 2 , MMP-9, IL-6 and interleukin 10 (IL-10) generation, it appeared more efficacious in enhancing release of O 2 − , TNF-α, MIP-1α/CCL3, MIP-2/CXCL2 and CINC-1/CXCL-1.

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia O 2 − Generation
Neuronal injury via oxidative stress as a result of reactive oxygen species generated by microglia have been thought to be involved in neurodegenerative diseases [15,17,20]. We have shown that E. coli LPS treatment of rat microglia will enhance O 2 − generation in a concentration-dependent manner in vitro [17]. O 2 − generation was determined in microglia tissue culture supernatants, as described in the Materials and Methods section. As shown in Figure 1, panel A, neonatal rat microglia released O 2 − in a concentration-dependent manner when treated with either E. coli or P. gingivalis LPS for 18 h.
Maximal and statistically significant O 2 − release was observed at both 5 × 10 4 to 10 5 ng/mL P. gingivalis LPS. In contrast, with E. coli LPS, the positive control used in these experiments, rat microglia showed maximal O 2 − release at 0.1 and 1 ng/mL, as shown previously [17]. Thus, in our study, P. gingivalis LPS was 10,000-fold less potent than E. coli LPS in stimulating O 2 − generation from neonatal rat microglia in vitro.

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia LDH Generation
To determine whether an 18-h treatment with either P. gingivalis or E. coli LPS in vitro was toxic to rat neonatal microglia, we determined LDH presence in microglia culture supernatants as described in the Materials and Methods section. As shown in Figure 1, panel B, supernatants of microglia treated with P. gingivalis LPS (0.1-10 5 ng/mL) did not yield significant LDH release when compared with untreated microglia (0). In contrast, although not statistically significant, E. coli LPS (1-100 ng/mL) treated microglia supernatants revealed a concentration-dependent increase in LDH release. Thus, P. gingivalis LPS did not appear to affect the microglia cell membrane integrity in the concentration range (0.1-10 5 ng/mL) used in the in vitro experiments, but was indeed able to elicit microglia activation during the 18-h treatment, and the concomitant generation of O 2 − (Figure 1), MMP-9, cytokines and chemokines (Figures 2-4).

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia TXB 2 Generation
TXB 2 is an eicosanoid shown to be produced by LPS-activated microglia and hypothesized to be implicated in neuroinflammation [17,21,22]. TXB 2 release was determined in microglia tissue culture supernatants by ELISA (see the Materials and Methods section). As shown in Figure 1, panel C, microglia treated with P. gingivalis LPS for 18 h showed no significant release of TXB 2 at any of the concentrations tested in this study. In contrast, microglia treatment with E. coli LPS resulted in TXB 2 production, which was statistically significant at 100 ng/mL, as previously reported [17].

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia MMP-9 Generation
MMP-9 is a class of zinc-metalloproteinase which is involved in extracellular matrix degradation and has been reported to be generated by rat microglia stimulated with either E. coli or V. vulnificus LPS in vitro [17,[22][23][24]. MMP-9 production was determined in microglia tissue culture supernatants by ELISA (see the Materials and Methods section). As shown in Figure 2, panel A, P. gingivalis LPS-treated microglia released statistically significant levels of MMP-9 at 10 5 ng/mL LPS, the concentration at which maximal O 2 − release was also observed ( Figure 1, panel A). In contrast, after E. coli LPS treatment, microglia released statistically significant MMP-9 at 1 ng/mL LPS. Thus, P. gingivalis LPS was 100,000-fold less potent than E. coli LPS in activating statistically significant MMP-9 production.
O2 in a concentration-dependent manner when treated with either E. coli or P. gingivalis LPS for 18 h. Maximal and statistically significant O2 -release was observed at both 5 × 10 4 to 10 5 ng/mL P. gingivalis LPS. In contrast, with E. coli LPS, the positive control used in these experiments, rat microglia showed maximal O2 -release at 0.1 and 1 ng/mL, as shown previously [17]. Thus, in our study, P. gingivalis LPS was 10,000-fold less potent than E. coli LPS in stimulating O2 -generation from neonatal rat microglia in vitro. and TXB 2 (C) release was determined as described in the Materials and Methods section. The data are expressed as the means ± SEM of triplicate determinations from several independent experiments (n). * p ≤ 0.05, *** p ≤ 0.001, **** p ≤ 0.0001 LPS versus untreated control (0 ng/mL LPS).

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia TNF-α and IL-6 Generation
The pro-inflammatory cytokine TNF-α has been shown to be involved in neurodegenerative diseases [15]. We and others have reported that LPS-stimulated microglia release TNF-α in vitro [17,22,25,26]. TNF-α release was assessed in microglia tissue culture supernatants with a Milliplex MagPix Multiplex Array (see the Materials and Methods section). As shown in Figure 2, Toxins 2020, 12, 333 6 of 20 panel B, after P. gingivalis LPS stimulation in vitro microglia released statistically significant TNF-α at 10 5 ng/mL. In contrast, upon E. coli LPS treatment, microglia showed statistically significant TNF-α release at 10 ng/mL LPS. Thus, P. gingivalis LPS was observed to be 10,000-fold less potent than E. coli LPS in stimulating statistically significant TNF-α production.
The pro-inflammatory cytokine IL-6 has been observed in cellular survival, stress responses, and neuroinflammation [27]. We and others have reported that LPS-stimulated microglia produce IL-6 in vitro [22,25,26]. IL-6 release was assessed in microglia tissue culture supernatants with a Milliplex MagPix Multiplex Array (see the Materials and Methods section). As presented in Figure 2, panel C, maximal IL-6 release was observed upon exposure to E. coli LPS (10 ng/mL) and P. gingivalis LPS (10 5 ng/mL), a similar concentration that triggered maximum TNF-α generation ( Figure 2, panel B). However, the magnitude of IL-6 release was several-fold greater than TNF-α. Thus, P. gingivalis LPS in our study was 10,000-fold less potent than E. coli LPS in stimulating rat microglia in vitro to generate the cytokines TNF-α and IL-6. The pro-inflammatory chemokine MIP-1α/CCL3 appears to be involved in granulocyte recruitment to damaged brain regions [28]. We and others have reported that E. coli LPS stimulated microglia release MIP-1α/CCL3 in vitro [22,23,29,30]. MIP-1α/CCL3 release was determined in microglia tissue culture supernatants with a Milliplex MagPix Multiplex Array (see the Materials and Methods section). As shown in Figure 3, panel A, P. gingivalis LPS-stimulated microglia released statistically significant MIP-1α/CCL3 between 10 4 and 10 5 ng/mL compared to untreated controls. In contrast, E. coli LPS-stimulated microglia showed a statistically significant release of MIP-1α/CCL3 between 1 and 100 ng/mL. Thus, P. gingivalis LPS when compared with E. coli LPS appeared 10,000-fold less potent in stimulating statistically significant MIP-1α/CCL3 production.

Effect of P. gingivalis LPS on Neonatal Rat Brain Microglia IL-10 Generation
The anti-inflammatory and immunosuppressive cytokine IL-10 [36] has been shown to be generated by E. coli LPS-treated mouse, rat, and human microglia [22,[36][37][38]. IL-10 release was determined in microglia tissue culture supernatants with a Milliplex MagPix Multiplex Array (see the Materials and Methods section). As shown in Figure 4, P. gingivalis LPS-treated microglia only showed an increase in IL-10 generation at 10 5 ng/mL compared to untreated controls. In contrast, E. coli LPS stimulated microglia released statistically significant IL-10 at 10 ng/mL compared to untreated controls.
Toxins 2020, 12, x FOR PEER REVIEW 8 of 20 The anti-inflammatory and immunosuppressive cytokine IL-10 [36] has been shown to be generated by E. coli LPS-treated mouse, rat, and human microglia [22,[36][37][38]. IL-10 release was determined in microglia tissue culture supernatants with a Milliplex MagPix Multiplex Array (see the Materials and Methods section). As shown in Figure 4, P. gingivalis LPS-treated microglia only showed an increase in IL-10 generation at 10 5 ng/mL compared to untreated controls. In contrast, E. coli LPS stimulated microglia released statistically significant IL-10 at 10 ng/mL compared to untreated controls.
As shown in Figure 5, panels A and B, confocal fluorescence imaging confirmed the presence of remaining MMP-9 within 1 ng/mL E. coli LPS-treated microglia, but none in the cytosol of 100,000 ng/mL P. gingivalis LPS-treated microglia, suggesting all MMP-9 had been released into the tissue culture supernate after the microglia had been treated for 18 h in vitro with P. gingivalis LPS.   Figure 3) into the tissue culture media was observed to be in the µg range. In order to determine whether MMP-9, IL-6, MIP-1α/CCL3, MIP-2/CXCL-2 and CINC-1/CXCL-1 were present in 1 ng/mL E. coli LPS-treated and 100,000 ng/mL P. gingivalis LPS-treated microglia after the 18 h in vitro exposure, we investigated the presence of these mediators in microglia by confocal fluorescence imaging (see the Materials and Methods section).
As shown in Figure 5, panels A and B, confocal fluorescence imaging confirmed the presence of remaining MMP-9 within 1 ng/mL E. coli LPS-treated microglia, but none in the cytosol of 100,000 ng/mL P. gingivalis LPS-treated microglia, suggesting all MMP-9 had been released into the tissue culture supernate after the microglia had been treated for 18 h in vitro with P. gingivalis LPS. However, as shown below in Figure 6 (panels A and B), Figure 7 (panels A and B), Figure 8 (panels A and B) and Figure 9 (panels A and B), confocal fluorescence imaging confirmed the presence of IL-6, MIP-1α/CCL3 and MIP-2/CXCL-2 within the cytosol of 1 ng/mL E. coli LPS-treated microglia as well as in 100,000 ng/mL P. gingivalis LPS-treated microglia. CINC-1/CXCL-1 was not observed in P. gingivalis LPS-treated microglia cytosols. These observations suggested that, in contrast to MMP-9 generation, release of IL-6, MIP-1α/CCL3, MIP-2/CXCL-2 and CINC-1/CXCL-1 by microglia into the tissue culture supernates was ongoing after an 18-h treatment of microglia with either E. coli LPS or P. gingivalis LPS. Thus, we conclude that the generation of these mediators appeared to be a dynamic process that for some cytokines may require more than 18 h to be completed. However, as shown below in Figure 6 (panels A and B), Figure 7 (panels A and B), Figure 8 (panels A and B) and Figure 9 (panels A and B), confocal fluorescence imaging confirmed the presence of IL-6, MIP-1α/CCL3 and MIP-2/CXCL-2 within the cytosol of 1 ng/mL E. coli LPS-treated microglia as well as in 100,000 ng/mL P. gingivalis LPS-treated microglia. CINC-1/CXCL-1 was not observed in P. gingivalis LPS-treated microglia cytosols. These observations suggested that, in contrast to MMP-9 generation, release of IL-6, MIP-1α/CCL3, MIP-2/CXCL-2 and CINC-1/CXCL-1 by microglia into the tissue culture supernates was ongoing after an 18-h treatment of microglia with either E. coli LPS or P. gingivalis LPS. Thus, we conclude that the generation of these mediators appeared to be a dynamic process that for some cytokines may require more than 18 h to be completed.

Discussion
Both neuroinflammation initiation and resolution, as a result of CNS infections [18] have been reported to be associated with microglia phenotypes described as either pro-inflammatory/classical or anti-inflammatory/alternatively activated [16]. One extensively studied in vivo and in vitro activator of microglia is LPS [8], which appears to activate microglia via the lipid A portion of the LPS macromolecule, resulting in the time-and concentration-dependent release of pro-inflammatory mediators, such as matrix metalloproteinases, metabolites of arachidonic acid, cytokines, chemokines and free radicals, such as O2 - [17,19,22]. Interestingly, the recent observation that P. gingivalis LPS has been detected in the human brains, putatively supports the proposed hypothesis that it might activate brain microglia [6,7].
The first aim of this research was to test the hypothesis that ultrapure P. gingivalis LPS might induce an in vitro pro-inflammatory/classical activation microglia phenotype. Our experimental observations appear to provide support for our hypothesis: First, and similar to E. coli LPS, which

Discussion
Both neuroinflammation initiation and resolution, as a result of CNS infections [18] have been reported to be associated with microglia phenotypes described as either pro-inflammatory/classical or anti-inflammatory/alternatively activated [16]. One extensively studied in vivo and in vitro activator of microglia is LPS [8], which appears to activate microglia via the lipid A portion of the LPS macromolecule, resulting in the time-and concentration-dependent release of pro-inflammatory mediators, such as matrix metalloproteinases, metabolites of arachidonic acid, cytokines, chemokines and free radicals, such as O 2 − [17,19,22]. Interestingly, the recent observation that P. gingivalis LPS has been detected in the human brains, putatively supports the proposed hypothesis that it might activate brain microglia [6,7]. The first aim of this research was to test the hypothesis that ultrapure P. gingivalis LPS might induce an in vitro pro-inflammatory/classical activation microglia phenotype. Our experimental observations appear to provide support for our hypothesis: First, and similar to E. coli LPS, which was used previously as a positive control [19], we observed that ultrapure P. gingivalis LPS-treated rat brain microglia produced O 2 − in a concentration-dependent and statistically significant manner.
Second, ultrapure P. gingivalis LPS-treated microglia produced the pro-inflammatory MMP-9, but no significant TXB 2 production, thus contrasting with E. coli LPS-triggered microglia TXB 2 release as we have previously reported [19,22]. Third, ultrapure P. gingivalis LPS-treated microglia released the following pro-inflammatory cytokines and chemokines in a concentration-dependent manner, with MIP-1α/CCL3 generation being the highest, followed by MIP-2/CXCL2, CINC-1/CXCL1, IL-6 and TNF-α. Fourth, although ultrapure P. gingivalis LPS appeared less potent than E. coli LPS in activating a rat pro-inflammatory/classical microglia phenotype and was less efficacious in stimulating the generation of TXB 2 , MMP-9, and IL-6, in contrast, the release of O 2 − , TNF-α and the chemokines MIP-1α/CCL3, MIP-2/CXCL2 and CINC-1/CXCL1 was increased when compared to their production by E. coli LPS-treated microglia. Thus, our experimental data supports our working hypothesis, namely that ultrapure P. gingivalis LPS (0.1-10,000 ng/mL) activates a rat pro-inflammatory/classical brain microglia phenotype in vitro, and, in contrast with E. coli LPS, P. gingivalis LPS does not seem to be toxic to the microglia cell in vitro, as shown by the minimal LDH release we observed in our experiments. A second objective of this investigation was to determine whether treatment of microglia with ultrapure P. gingivalis LPS might result in an anti-inflammatory/alternatively activation phenotype, and anti-inflammatory cytokine IL-10 release, a cytokine that appears to be involved in tissue repair in the CNS [39]. We observed the following results: First, in contrast with E. coli LPS, ultrapure P. gingivalis LPS was less potent in triggering anti-inflammatory IL-10 generation after a 18 h in vitro, as depicted by the right shift observed in the dose-response curve that is presented in Figure 4. Second, while E. coli LPS induced statistically significant anti-inflammatory cytokine IL-10 generation, in contrast, the ultrapure P. gingivalis LPS treatment of microglia resulted in IL-10 generation, which though enhanced, was not statistically significant. Thus, additional studies will be required to demonstrate whether ultrapure P. gingivalis LPS may also trigger an anti-inflammatory/alternative activation phenotype in rat brain microglia in vitro. Furthermore, whether systemic P. gingivalis LPS may activate an anti-inflammatory/alternatively phenotype in rat brain microglia in vivo remains to be studied in future investigations.
It is of importance to discuss potentially new lines of research that have emerged from our studies of the in vitro effects of ultrapure P. gingivalis LPS on rat neonatal microglia. First, it would be significant to compare the biological activity of ultrapure P. gingivalis, which has been prepared by removing lipoprotein, with that of standard P. gingivalis LPS, and determine whether both pro-inflammatory/classical and/or anti-inflammatory/alternative rat microglia phenotypes are observed in vitro [11]. Second, testing whether LPS isolated from other P. gingivalis strains will also activate pro-inflammatory/classical and/or anti-inflammatory/alternative microglia phenotypes in vitro becomes a significant question worthy of further investigation [40]. Third, our in vitro study with ultrapure P. gingivalis LPS involved studying the pro-inflammatory/classical and/or anti-inflammatory/alternatively activation of neonatal rat brain microglia and the concomitant mediator response. Thus, determining whether adult rat microglia might also be activated by ultrapure P. gingivalis LPS would be of interest, as adult microglia have been shown to generate considerable PGE 2 [41], and so the pro-inflammatory/classical and/or anti-inflammatory/alternative adult microglia activation phenotypes and concomitant pro-inflammatory and anti-inflammatory mediators released may differ from those we have observed using neonatal microglia in this investigation. Fourth, because P. gingivalis LPS has recently been observed in the human brain [7], and primary human microglia have been shown to release pro-inflammatory O 2 − , TXB 2 and TNF-α [42] and anti-inflammatory IL-10 [43], the effect of ultrapure P. gingivalis LPS on the activation of human microglia in vitro should be investigated. Fifth, in vivo studies to determine whether systemic P. gingivalis LPS may result in rat brain microglia pro-inflammatory/classical and/or anti-inflammatory/alternative activation phenotypes should be considered as systemic LPS has been reported to cause brain inflammation [44,45]. Sixth, the possible effect of microglia's circadian clock on the in vitro and/or in vivo activation of pro-inflammatory/classical and/or anti-inflammatory/alternative microglia activation phenotypes by P. gingivalis LPS should be considered in the design of both in vitro and in vivo experiments [46]. The proposed lines for further in vitro and in vivo research with P. gingivalis LPS will, in our view, help determine microglia pro-inflammatory/classical and/or anti-inflammatory/alternative activation phenotypes, and will hopefully contribute to novel therapeutic and diagnostic strategies for periodontal disease as well as the neuropathologies that have been hypothesized to implicate the Gram-negative bacterium P. gingivalis.

Conclusions
In conclusion, the current study, which demonstrates that in vitro ultrapure P. gingivalis LPS triggers both the classical and alternative activation of rat brain microglia, furthers our current understanding of P. gingivalis LPS's potential toxicity to the brain immune system.

LPS Decontamination
To inactivate LPS, all glassware and metal spatulas were baked for 4 h at 180 • C. [17]. Sterile and LPS-free 225 cm 2 vented cell culture flasks were from BD Biosciences, San Jose, CA, USA; 24-well flat-bottom culture clusters and disposable serological pipettes were from Costar ® , Corning Inc., Corning, NY, USA. Sterile and pyrogen-free Eppendorf Biopur pipette tips were from Brinkmann Instruments, Inc., Westbury, NY, USA.

Isolation of Rat Neonatal Microglia
Adherence to the National Institutes of Health guidelines on the use of experimental animals and protocol approved by Midwestern University's Research and Animal Care Committee was followed in all experiments. Rat brain neonatal microglia was harvested and characterized as described [22]. The ethic approval board name: IACUC (Institutional Animal Care and Use Committee), Midwestern University; ethic approval code: 941, approval date: 24 January 2017.

Activation of Microglia with LPS (Experimental Protocol)
To determine the effect of ultrapure P. gingivalis LPS, 1.8-2.0 × 10 5 neonatal microglia in DMEM + 10% FBS + 1% penicillin (P) + streptomycin (S) were plated into each well of a 24-well flat-bottom culture cluster. Thereafter, some wells remained untreated shown as 0 on the x axis of Figures 1-4 while other wells were treated with 0.1-100,000 ng/mL P. gingivalis, or E. coli LPS (0.1-100 ng/mL) for 18 h in a humidified 5% CO 2 incubator at 35.9 • C, as neonatal rat microglia are maximally activated by E. coli LPS at this time point [19,22]. Upon termination of the in vitro treatment, media (1 mL) from each tissue culture well was split into two aliquots. One aliquot (0.1 mL) was used to measure LDH levels and the remaining aliquot (0.9 mL) was frozen (−80 • C) for determination of MMP-9, TXB 2 , chemokines and cytokines as described below.

Lactate Dehydrogenase (LDH) Assay
Cellular toxicity following neonatal rat microglia pre-incubation with either P. gingivalis or E. coli LPS was determined spectrophotometrically as described [17,47]. LDH release was expressed as a percent of total LDH released from Triton X-100 (0.1%)-treated microglia.

Assay for Microglia TXB 2 Generation
After an 18-h preincubation of neonatal rat microglia with either P. gingivalis LPS or E. coli LPS, the production of TXB 2 was determined in the supernatants by ELISA (Cat. # 519031, Cayman Chemical Company, Ann Arbor, MI, USA) according to the manufacturer's protocol. Prior to the assay, samples were diluted between 1:10 and 1:100. On the assay characteristics, the Cayman Chemical Company protocol reports that the intra-assay precision of the TXB 2 ELISA was generated from 8 reportable results across two different concentrations of analytes in a single assay (intra-assay %Coefficient of Variation (CV) 8.2-15.3), while the inter-assay precision was generated from two different concentrations of analytes across 11 different assays (inter-assay %CV 9.9-12.9). The results were expressed as pg/mL with the minimum detectable concentration being 5 pg/mL TXB 2 .

Assay for Microglia MMP-9 Generation
After an 18-h preincubation of neonatal rat microglia with either P. gingivalis LPS or E. coli LPS, MMP-9 release into supernatants was assessed by ELISA (Cat. # DY8174, R&D systems, Minneapolis, MN) according to the manufacturer's protocol. Prior to the assay samples were diluted between 1:10 and 1:100. On the assay characteristics, the R&D systems protocol reports that the intra-assay precision of the MMP-9 ELISA was generated from 3 samples of known concentration tested 20 times in a single assay (intra-assay %CV 4.5-6.9), while inter-assay precision was generated from 3 samples of known concentration in 20 separate assays by at least 3 technicians (inter-assay %CV 3.9-6.9). The results were expressed as pg/mL and the minimum detectable concentration was 78.1 pg/mL.

Statistical Analysis of the Data
The data were expressed as means ± SEM of triplicate determinations of several independent experiments, as noted in the Figure legends. Data were analyzed with Prism software package version 7 (GraphPad, San Diego, CA, USA) and tested for normality with the Shapiro-Wilk normality test. Appropriate multiway analysis of variance was then performed on all sets of data. Where significant interactions were encountered, simple effects were tested with a one-way analysis of variance followed by a Dunnett's post-hoc test. Differences were considered statistically significant at p < 0.05 [19].