**2. Results**

#### *2.1. Characterization of N2a Cell Growth and Initial Viability*

The N2a CLL 131 cell line used in this study displayed a typical growth curve characterized by a short lag phase (slow cell growth), followed by a log phase (exponential proliferation of cells and consumption of nutrients of the culture medium) until a maximum density of ≈ 100,000 cells/well is reached, and a late stationary/senescence phase (reduced cell proliferation) (Figure 1).

**Figure 1.** Characteristic growth pattern of the neuroblastoma (N2a) cell line used in this study. Densities in 96-well microplates were extrapolated from growth experiments conducted over a 4 days period in 25 cm<sup>2</sup> culture flasks using a 10% fetal bovine serum (FBS) culture medium. Data represent the mean ± standard deviation (SD) of one experiment (N2a cells at 383 passages (P)), with *n* = 10 counts for each point. Coe fficients of variation (CV) ranged from 10.8% to 26.3%.

The linearization of the log phase of the growth curve was defined by the following equation:

$$\mathbf{Y} = 0.0578\mathbf{X} + 7.7948 \text{ (}\mathbf{r}^2 = 0.9974\text{)}\tag{1}$$

in which X is the culture time (hours) and Y is the Ln-transformed cell number. Based on this equation, it was concluded that the N2a growth curve was characterized by a 9.8 h lag phase (Y = Ln (4288)) and that the cell number increased by two-fold after an additional 12 h. Moreover, a cell seeding density of 50,000 ± 10,000 cells/well allowed reaching a maximum cell density of ≈ 100,000 cells/well after 22 h culture time. For more convenience, a culture time of 26 h post-seeding was selected in all further experiments.

Next, the N2a cell initial viability after 26 h growth was compared at ten di fferent cell seeding densities ranging from 10,000 to 100,000 cells/well and in two culture conditions (5% and 10% FBS growth medium) (Figure 2). Results showed that (i) the highest absorbance value is consistently obtained at a cell seeding density around the maximum number of cells supported by microplate wells, and (ii) absorbance values increase in proportion with MTT incubation times (Figure 2). For instance, the maximum absorbance value measured after 26 h culture time and 45 min MTT incubation time was 1.4 while selecting a cell seeding density of 50,000 ± 10,000 cells/well (vertical dotted lines, Figure 2) allowed to reach absorbance values comprised between 1 and 1.25 (horizontal dotted lines, Figure 2). This absorbance range was considered optimal when the detection of a decrease in cell viability is sought. Finally, no di fferences were observed when using either 5% or 10% FBS growth medium, which indicates possible saving opportunities on this expensive reagen<sup>t</sup> at this step of the CBA-N2a.

Seeding density (cells/well)

**Figure 2.** Initial viability of N2a cells observed in 96-well microplates after 26 h growth in 5% FBS (full line) and 10% FBS (dotted line) culture medium, at different cell seeding density. Six distinct MTT incubation times were also tested: 15 min (blue); 25 min (green); 35 min (orange); 45 min (black); 55 min (pink); 65 min (red). Data represent the mean ± SD of one microplate (N2a cells at 536 P), each point tested in six wells. Mean CVs were <3%. Absorbance values were measured at 570 nm via the MTT assay.

Based on these results, all further CBA-N2a experiments were conducted as follows:


#### *2.2. Characterization of N2a Cell Final Viability*

The second step of CBA-N2a is the exposure of N2a cells to VGSC activators or inhibitors, in OV− or OV+ conditions. Following an additional culture time of 19 h overnight, the final viability of N2a cells was assessed as previously described.

#### 2.2.1. N2a Cell Final Viability in OV− Conditions

A final cell viability lower than the initial cell viability (as measured in the RCV control) was observed only with 1% FBS growth medium at cell seeding density > 40,000 cells/well (Figure 3). Conversely, all other growth media allowed to reach a final viability higher than the one displayed by the RCV control, regardless of the cell seeding density (Figure 3), suggesting a complete renewal of the culture medium with 2% FBS growth medium is sufficient to ensure a stable cell viability during 19 h additional culture of the toxin exposure. A better stability of cell viability was consistently achieved at cell seeding densities > 40,000 cells/well. In other words, working at a cell seeding density of 50,000 ± 10,000 cells/well will allow to reach an absorbance value ≥1, close to the one measured in the RCV control (Figure 3).

Seeding density (cells/well)

**Figure 3.** Final viability of N2a cells observed at different seeding densities when cultured 26 h in 5% FBS culture medium followed by 19 h in fresh growth medium supplemented with 1% FBS (red), 2% FBS (green), 3% FBS (orange), 4% FBS (pink) and 5% FBS (black). The initial viability measured in the Reference Cell Viability (RCV) control microplate after 26 h of growth is represented by the blue dotted curve. Data represent the mean ± SD of one microplate (N2a cells at 537 P), each point tested in six wells. Absorbance values were measured at 570 nm via the MTT assay, after a 45 min MTT incubation time.

Based on these results, the following procedure was selected to ensure that, at this step of the CBA-N2a, a final cell viability close to the initial cell viability is obtained in OV− conditions:


#### 2.2.2. N2a cell Final Viability in OV+ Conditions

As expected, absorbance data measured at 0/0 μM (OV− conditions) was close to the one of the RCV control (Figure 4). At O/V concentrations ranging from 10/1 to 80/8 μM, a "protective effect" was observed as evidenced by an increase in cell viability of ≈ 20% above the RCV control (Figure 4). Between 80/8 and ≈ 110/11 μM, although a slight decrease in absorbance data was observed, these values were consistently found above the RCV control (Figure 4), suggesting a slightly toxic effect, although regarded as "non-destructive effect" of O/V treatment on N2a cells (Figure 4). At higher concentrations, however, cell viability decreased in a dose-dependent manner down to the level observed for DMSO control, which was indicative of a "destructive effect" of O/V treatment on N2a cells (Figure 4). Above 300/30 μM, the effect of O/V on cell viability was considered "lethal" with the complete elimination of cell viability (Figure 4). Concerning the variability of O/V treatment obtained across a wide range of N2a cell passages from 384 to 804 P, the CVs were below 12.4% for the "non-destructive" O/V treatment between 80/8 and 100/10 μM selected for the detection of VGSC activators. When seeking to detect VGSC inhibitors, the absorbance values were close to 0 showing non significative CVs above 45% for the "destructive" O/V treatment between 270/27 and 300/30 μM.

**Figure 4.** Dose-response curve of N2a cells when exposed to increasing concentrations of Ouabain and Veratridine (O/V) treatments ranging from 0/0 to 360/36 μM. Data represent the mean ± SD of one microplate in six independent experiments corresponding to cell passage numbers of 384, 542, 800, 801, 803 and 804 P, each point run in triplicate. The mean absorbance ± SD values corresponding to the RCV control (horizontal green line and dotted green lines) and DMSO control (horizontal red line) were determined at 1.105 ± 0.096 and 0.042 ± 0.001, respectively. Absorbance values were measured at 570 nm via the MTT assay, after a 45 min MTT incubation time. The four different effects induced by increasing concentrations of O/V treatment on N2a cells, i.e., protective, non-destructive, destructive and lethal are also represented.

Based on these results, the O/V treatment conditions best adapted to the type of activity we seek to detect on target cells (e.g., activation or inhibition of VGSCs on N2a cells) were defined as follows:


#### *2.3. Characterization of the Unspecific E*ff*ects of Solvent and Dry Extract on N2a Cell Viability*

The issue of potential solvent toxicity and sample extract matrix effects on N2a cell viability was also taken into consideration.

## 2.3.1. Solvent Effects

Using methanol (MeOH), a progressive loss in N2a cell viability is observed in 100/10 μM OV+ conditions at solvent concentrations ≥0.8%, giving a mean absorbance value ≈ 25% lower than the one measured in the RCV control at the highest concentration tested (Figure 5a). In OV− conditions, an opposite effect is observed as evidenced by an increase in cell viability also at concentrations ≥0.8% (Figure 5a). Conversely, regardless of the OV conditions, DMSO consistently induced an abrupt decrease in cell viability at concentrations ≥1% until the complete elimination of viable N2a cells (Figure 5b). These findings indicate that both solvents can be used provided a concentration <0.8% is selected and that the maximum concentration of solvent can start at 0.5%. Practically, the use of non-volatile DMSO was preferred for the preparation of stock solutions (especially for toxin standards) as it is non-hazardous and likely to ensure stable toxin concentration over long periods of storage.

**Figure 5.** Dose-response curves of N2a cells when exposed to increasing concentrations of two solvents, in OV− (open symbols) and OV+ (solid symbols) conditions at 100/10 μM (final concentrations); (**a**) MeOH (-/) and (**b**) DMSO (Δ/) were tested using N2a cells at 551 and 549 P, respectively. Data represent the mean ± SD of one microplate, with each point run in triplicate. Absorbance values were measured at 570 nm via the MTT assay, after a 45 min MTT incubation time. The initial cell viability in the RCV control was 1.139 ± 0.021 and 0.995 ± 0.031 and the final cell viability in the absence of O/V treatment (COV−) control was 1.217 ± 0.025 and 1.047 ± 0.023 for (**<sup>a</sup>**,**b**), respectively. The dotted vertical line corresponds to the maximum solvent concentration 0.5% for solvent interferences.

Consequently, a dilution of at least 1:200 giving the first final concentration (C1) of the toxin standard/sample extract stock solutions in MeOH or DMSO was achieved in all further CBA-N2a experiments, as follows:


#### 2.3.2. Biological Matrix effects

In this study, the dry extract weights (DEW) obtained from 10 g of fish flesh ranged from 2.7 to 4.3 mg (Section 5.2), but they have been shown to sometimes vary up to ten-fold in other studies (data not shown). When tested at similar concentrations in OV− conditions, all extracts induced a final cell viability slightly above the one observed in the RCV control at concentrations ≥1500 pg/μL, whereas cell viability could increase up to 70% as observed for sample Emer05 at 10,000 pg/μ<sup>L</sup> (or 30.08 ± 5.92 μg fish flesh equivalent/μL) (Figure 6). Above this concentration, a decrease in cell viability rapidly occurred, suggesting an unspecific cytotoxicity on N2a cells likely due to matrix effects (Figure 6).

**Figure 6.** Dose-response curves of N2a cells in OV− conditions when exposed to increasing concentrations of LF90/10 dry extracts prepared from *Chlorurus microrhinos* Cmic02 (blue), Cmic19 (red), *Epinephelus merra* Emer13 (black) and Emer05 (green) fish samples. Data represent the mean ± SD of three independent microplates (N2a cells at 800, 801 and 803 P), each point run in triplicate. Absorbance values were measured at 570 nm via the MTT assay, after a 45 min MTT incubation time. The cell viability in the RCV control was determined at 1.143 ± 0.009, while the final cell viability in COV− control was 1.037 ± 0.053. The dotted vertical line corresponds to the maximum concentration of dry extract (MCE = 10,000 pg/μL) for matrix interferences.

Based on these results, the MCE for LF90/10 fractions was set at 10,000 pg/μ<sup>L</sup> for this study in all further experiments. Prior to CBA-N2a toxicity analysis, dry extract solutions of fish samples were thus prepared as follows:


This revisited CBA-N2a resulted in a practical guide that is presented in the Supplementary Materials.

#### *2.4. Application to the Detection of VGSC Activators and Inhibitors*

#### 2.4.1. N2a Cell Initial Viability

To test the relevance of the six parameters revisited in Sections 2.1–2.3, this improved protocol was further applied to the detection of different toxin standards acting on VGSCs.

First, the initial viability of N2a cells prior to toxin exposure was assessed by measuring absorbance values in RCV control using the MTT colorimetric assay in three independent experiments. The metabolization of MTT into formazan by N2a viable cells resulted in a uniform blue color visible in all wells, proof that a high cell layer confluence was attained. Following the addition of DMSO and cell lysis, a dark purple color was released in each well (Figure S1).

The net absorbance data obtained for RCV controls in all three experiments were indicative of a high viability of N2a cells (mean value between 1 and 1.25) and were also highly reproducible, with CVs below 5% (Table 1).

\*


**Table 1.** Initial viability of N2a cells as assessed in RCV controls (cell layer ≈ 100,000 cell/well obtained after 26 h of growth).

 Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 662, 663 and 666 P).

#### 2.4.2. Detection of VGSC Activators and Inhibitors

In OV− conditions (upper half of the microplates), the same dark purple color was observed for COV− control and all concentrations tested regardless of the toxin standard (Figure S1). Conversely, in OV+ conditions (bottom half of the microplate), a progressive fading of this dark purple color was observed with increasing concentrations of P-CTX3C and PbTx3, until complete discoloration, signing the total death of N2a cells (Figure S1). Otherwise, a progressive darkening of the pale color observed at low concentrations occurred with increasing concentrations of STX and dc-STX, signing a progressive restoration of N2a cell viability (Figure S1).

For VGSC activators, net absorbance values showed that COV− and COV+ control absorbance data were close to that of the RVC control. The same observation applies to VGSC inhibitors except in OV+ conditions where absorbance data of COV+ controls were in the range of 0.1.

Inter-assay comparison in three independent experiments showed that net absorbance data are reproducible with CVs of 6.1% and 5.4% for COV− controls, and 1.6% and 4.5% for COV+ controls of P-CTX3C and PbTx3, respectively. Likewise, net absorbance data of COV− controls showed CVs of 6% and 3.9% of STX and dc-STX, respectively, but 41.4% and 37.5% for COV+ controls, these latter values being close to 0 absorbance.

In the presence of the two VGSC activators, net absorbance data of all nine concentrations obtained in OV− conditions remained close to the ones measured in COV<sup>−</sup>, COV+ and RCV controls regardless of the toxin concentrations tested, whereas in OV+ conditions, a sigmoidal dose-response curve was obtained for both P-CTX3C and PbTx3 (Figure 7a,b). In the presence of the two VGSC inhibitors, net absorbance data of all nine concentrations obtained in OV− conditions remained close to the ones measured in COV− and RCV controls regardless of the toxin concentrations tested, whereas in OV+ conditions, a sigmoidal dose-response curve was obtained for both STX and dc-STX (Figure 7c,d).

Table 2 details several characteristic parameters of CBA-N2a curves. The EC50 values roughly corresponded to twice that of EC80 for VGSC activators and to 2.5-fold higher than that of EC20 for VGSC inhibitors. Overall, high reproducibility was found for top and bottom absorbances as well as negative Hillslopes with CVs < 5% for P-CTX3C and PbTx3, and CVs < 10% for STX and dc-STX (Table 2). Regarding VGSC activators, CVs were ≈ 19% and between 17% and 21% for EC80 and EC50 values, respectively. For VGSC inhibitors, CVs of EC20 and EC50 varied between 10.3% and 17.5% for STX, and between 19.2% and 26.3% for dc-STX. Comparing the potency within VGSC acting toxin family, P-CTX3C was approximately 3300-fold more potent than PbTx3 and STX was approximately 4-fold more potent than dc-STX (Table 2).

All these findings showed that this newly improved CBA-N2a allowed for specific and sensitive detection of VGSC activators and inhibitors.

Based on these results, quality check (QC) controls were defined as follows:


The QC controls are used to verify the specific action of VGSC acting toxins (Figure S2) and their concentration is selected at the EC50 of a given standard toxin.

**Figure 7.** Dose-response curves displayed by N2a cells when exposed to nine increasing concentrations of toxin standards after 19 h exposure time, in OV− (open symbols) and OV+ (solid symbols) conditions at 100/10 μM for voltage gated sodium channels (VGSC) activators and 270/27 μM for VGSC inhibitors (plain symbols): (**a**) P-CTX3C (-/•); (**b**) PbTx3 (-/); (**c**) (STX) (♦/); (**d**) dc-STX (Δ/). Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 662, 663 and 666 P), with each concentration run in triplicate. Net absorbance values were 1.101 ± 0.067 and 1.077 ± 0.058 in COV− controls (*n* = 3), and 1.209 ± 0.014 and 1.226 ± 0.055 in COV+ controls (*n* = 3) for (**<sup>a</sup>**,**b**), respectively, for VGSC activators. Net absorbance values were 1.170 ± 0.046 and 1.113 ± 0.067 in COV− controls (*n* = 3), and 0.136 ± 0.051 and 0.145 ± 0.060 in COV+ controls (*n* = 3) for (**<sup>c</sup>**,**d**), respectively.

**Table 2.** Dose-response curve parameters of cell-based assay (CBA)-N2a when detecting VGSC activators or inhibitors.


\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 662, 663 and 666 P). \*\* OV+ conditions: 100/10 and 270/27 μM final concentrations for VGSC activators and inhibitors, respectively. ND = not determined.

#### 2.4.3. Composite Toxicity Estimation of VGSC Activators in Fish Samples

Four fish samples were analyzed: two steephead parrot-fish (*Chlorurus microrhinos*, Scaridae, Cmic02 and Cmic19) and two honeycomb groupers (*Epinephelus merra*, Serranidae, Emer05 and Emer13). To evaluate the composite CTX-like toxicity, a set of standards must be run in parallel with samples

in order to calibrate an experiment properly. For this purpose, P-CTX3C standard was tested in parallel with fish matrix, in the absence vs. under non-destructive O/V treatment at 85.7/8.57 μM (final concentrations), as the EC50 of standards are further needed to infer toxin content in biological samples.

A high repeatability and reproducibility of viability data were obtained for RCV, COV<sup>−</sup>, COV+ controls as well as QCOV− (Table 3), with CVs < 7%, whereas a lower reproducibility was observed for QCOV+ absorbance data with a CV of 17.3% inherent to the cytotoxicity of PbTx3 at EC50 (Section 2.4.2).

**Table 3.** Assessment of five viability controls useful to validate the detection of VGSC activators in fish matrix using the revisited CBA-N2a.


\* Data represent the mean ± SD of three microplates in one experiment (N2a cells at 810 P). \*\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P).

Moreover, statistical analyses by means of the Wilcoxon test showed no significant difference between RCV control and COV<sup>+</sup>, and between QCOV− and COV− for both intra- and inter-assay (Table S1). Conversely, significant differences were observed between QCOV+ and all other controls. In addition, the comparison between intra- and inter-assay values showed no significant differences for COV+ values, nor between RCV control and COV+ values showing that final viability was similar to initial viability under non destructive O/V treatment (Table S1).

The dose-response curves thus obtained for P-CTX3C (data not shown) allowed to establish another set of EC80 and EC50 (Table 4) which were compared to those of Table 2 (Section 2.4.2). Statistical comparison by means of the Wilcoxon test showed no significant differences between EC80 and EC50 values obtained under two distinct O/V treatments, i.e., 100/10 and 85.7/8.57 μM with *p*-values > 0.4 (Table S2).

**Table 4.** Variabilities of EC80 and EC50 for two VGSC activators (P-CTX3C and PbTx3) using the revisited CBA-N2a as assessed under non-destructive O/V treatment conditions.


\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 662, 663 and 666 P). \*\* Data represent the mean ± SD of three microplates in one experiment (N2a cells at 810 P). \*\*\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P).

Further, the limit of detection (LOD) and the limit of quantification (LOQ) of the CTX-like toxicity in fish were determined using the MCE of 10,000 pg dry extract/μL and EC80 and EC50 values obtained for P-CTX3C exclusively run in parallel with fish samples under 85.7/8.57 μM OV treatment. The LOD and LOQ showed high repeatability with CVs < 9.5% (intra-assay), whereas a lower reproducibility was noticed with CVs > 23.9% (Table 5). The Wilcoxon test also confirmed no significant differences between intra- and inter-assay data with *p*-values of 0.7 and 0.4 for LOD and LOQ, respectively. Hence, a mean LOD value was established at 0.089 ± 0.017 ng/mg of dry extract for P-CTX3C, with LOQ values about twice that of LOD.


**Table 5.** Estimation of limit of detection (LOD) and limit of quantification (LOQ) of the CTX-like toxicity in fish flesh using the revisited CBA-N2a under non-destructive O/V treatment at 85.7/8.75 μM.

\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P). \*\* Data represent the mean ± SD of three microplates in one experiment (N2a cells at 810 P).

Based on the "dry extract weight/fresh weight" (DEW/FW) ratio characterizing each fish sample (Section 5.2), the LOD and LOQ of P-CTX3C in fish flesh were determined and expressed in ng of P-CTX3C eq/g of fish flesh (Table 5) for further comparison with the EFSA and US Food and Drug Administration (FDA) advisory level. Although LOD and LOQ of the CTX-like toxicity varied from one fish to another (Table 5) due to differences in DEW/FW ratios, the Wilcoxon test showed no significant differences between intra- and inter-assay data regardless of the fish sample and P-CTX3C used as reference (Table S3). Hence, the mean LOD values were established at 0.031 ± 0.008 (CV = 25.4%) and LOQ = 0.064 ± 0.016 (CV = 24.3%) ng P-CTX-3C eq/g of fish flesh.

Application of the CBA-N2a to the detection of VGSC activators in four fish samples gave two distinct patterns as observed for VGSC activators (Figure S1). When exposed to increasing concentrations of fish extracts (Figure 8a,b), no cytotoxic effects were observed in OV− and OV+ conditions for Cmic02 and Emer13. Conversely, sigmoidal dose-response curves with a negative slope were obtained for Cmic19 and Emer05 fish samples in OV+ conditions (Figure 8a,b), whose ciguatoxicity was previously characterized by fluorescent receptor binding assay (fRBA) and/or LC-MS/MS analyses (Section 5.1.3).

**Figure 8.** Dose-response curves displayed by N2a cells when exposed to increasing concentrations of fish dry extracts (LF90/10) of two herbivorous fishes (**a**) and two carnivorous fishes (**b**) after 19 h exposure time in OV− (open symbols) and OV+ conditions 85.7/8.57 μM (solid symbols): Cmic02 (Δ/), Cmic19 (♦/), Emer13 (-/•) and Emer05 (-/). Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P), with each concentration run in triplicate. The dotted vertical line corresponds to the maximum concentration of dry extract (MCE = 10,000 pg/μL) for matrix interferences.

The sigmoidal dose-response curves thus obtained for Cmic19 and Emer05 were fitted according to the four parameter logistic regression (4PL) model showing high repeatability and reproducibility for top absorbance and Hillslope, with CVs < 10% (Table 6). Absorbance values close to 0 were also consistently obtained at the highest concentration of fish extract tested (Table 6).

The mean EC50 and EC80 values determined in pg of dry extract/μL for Cmic19 and Emer05 also showed a high repeatability with CVs between 5% and 11.2%, respectively, whereas a lower reproducibility of these values was noticed with CVs between 12% and 21.6%, respectively (Table 6). The Wilcoxon test also confirmed no significant differences between intra- and inter-assay data with *p*-values of 0.7 and 0.2 for Emer05 and Cmic19, respectively. Hence, mean EC80 values were established at 63.6 ± 12.4 and 108.5 ± 13.7 pg/μ<sup>L</sup> for Cmic19 and Emer05, respectively, with mean EC50 values 1.6 and 1.9 higher than that of EC80 for Emer05 and Cmic19, respectively. Based on EC50 values, Cmic19 dry extract was 1.5-fold more potent than Emer05 dry extract (Table 6).

**Table 6.** Dose-response curves parameters of CBA-N2a when detecting VGSC activators in fish samples.


\* Data represent the mean ± SD of three microplates in one experiment (N2a cells at 810 P). \*\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P). \*\*\* OV+ conditions: 85.7/8.57 μM final concentrations.

In addition, the toxin contents in fish flesh were also estimated using the EC50 values determined for toxin standard and fish extract (Table 7).

Overall, CTX-like composite toxicity data showed high repeatability and reproducibility with CVs < 13% (Table 7). The Wilcoxon test showed no significant differences between intra- and inter-assays for toxin contents with *p*-values of 0.3 and 0.6 for Emer05 and Cmic19, respectively. Hence, the mean toxin content in Cmic19 and Emer05 was estimated at 6.66 ± 0.68 and 3.31 ± 0.35 ng P-CTX-3C eq/g fish flesh, respectively, indicating Cmic19 was twice as toxic as Emer05.

**Table 7.** CTX-like composite toxicity estimation in two ciguatoxic fish samples using the revisited CBA-N2a.


\* Data represent the mean ± SD of three microplates in one experiment (N2a cells at 810 P). \*\* Data represent the mean ± SD of one microplate in three independent experiments (N2a cells at 795, 797 and 798 P).
