TREK-1 channel activation by PUFAs is independ of acyl chain length
In order to evaluate the TREK-1 activation by PUFAs, we used a cell line overexpressing TREK-1 channel (HEK hTREK-1 cells) (Moha ou Maati et al., 2011). Using the patch-clamp technique in whole-cell configuration, we recorded currents from a ramp protocol from − 100 mV to + 30 ms, as depicted in Fig. 1A. The initial TREK-1 current density of HEK hTREK-1 cells was 15.2 \(\pm\)0.9 pA/pF at 0 mV and varied between 2.3 pA/pF and 54.4 pA/pF (n=130) while the current density in HEK 293T cells was 4.9 \(\pm\)0.9 at 0 mV, varying between 2.3 pA/pF and 10.5 pA/pF (n = 6)(Fig. 1B). The initial current of HEK hTREK-1 cells showed a characteristic outward rectification (Fig. 1C, left) which is not observed in HEK 293T cells (Fig. 1C, right). As espected, the initial TREK-1 current (I0, black line) was inhibited by 10 µM of norfluoxetine (NrFlx). I0 was significantly decreased from 8.8 ± 2.2 pA/pF to 3.7 ± 0.9 pA/pF (p-value=0.02; n = 5) when norfluoxetine was applied and the outward rectification was lost (Fig. 1D). Thus, I0 was carried out mostly by TREK-1 channel.
As previously described, TREK-1 channel is activated by PUFAs, such as arachidonic acid (AA, C20:4 n-6)7, LA8 or DHA20. During PUFA superfusion, the current progressively increased until a steady-state (Fig. 1E). In HEK 293T cells, there was no PUFA-activated conductance, even after 3 min of perfusion. In HEK hTREK-1 cells, we performed dose-response experiments with the shorter and the longer PUFAs, respectively C18:2 n-6 and C22:6 n-3 (Fig. 1F). The effect of PUFAs reached a maximum at 6 µM (Fig. 1F) thus we performed all the whole-cell experiments at 10 µM. The initial average membrane potential (Em0) of HEK hTREK-1 cells was − 69.2 ± 0.7 mV (n = 130) compared to the average Em0 of HEK 293T cells of -34.2 ± 2.4 mV (n = 6 ; Fig. 2A).
To investigate the importance of the acyl chain length and double bounds in TREK-1 activation, we compared the response of 9 PUFAs with various chain lengths, from 18 to 22 carbons with different numbers of double bounds, from 2 to 6 unsaturations and one mono-unsaturated fatty acids (C18:1 n-9) (Table 1). When TREK-1 is activated by PUFAs, Em hyperpolarized to -81.7 ± 0.3 mV (Fig. 2A, n = 130), close to the theorical equilibrium potential of K+ ions (EK = -86.5 mV). TREK-1 was not activated by the C18:3 n-3 PUFA and consequently Em did not change (Fig. 2A).We plotted the relationship between TREK-1 current density at 0 mV in the presence of unsaturated fatty acids (IPUFA) normalized to the initial current density (I0) (this normalized parameter corresponds to the current fold-increase (IPUFA/I0)) and the number of carbons on the acyl chain. As shown in Fig. 2B, there is no statistical correlation between unsaturated fatty acid activation IPUFA/I0 and the acyl chain length (Spearman correlation test : p-value = 0.10), as for the current density of TREK-1 channel after PUFA perfusion (Spearman correlation test : p-value = 0.33) (Fig. 2D). We then plotted the relationship between IPUFA/I0 at 0 mV or current density in response to PUFA perfusion and the number of double bounds in the acyl chains and observed a positive correlation between the number of double bounds and the IPUFA/I0 (Fig. 2C, Spearman correlation test : p-value = 0.04, r = 0.67). However, this potential relationship was not confirmed by the simple linear regression test ( p-value = 0,12). Moreover, no correlation was observed between the number of double bounds in the acyl chains and IPUFA (Fig. 2E; Spearman correlation tests: p-value = 0.25). These statistical analysis reveal that there is at least no relationship between the effect of unsaturated fatty acids on TREK-1 and the acyl chain length and that a potential relationship existing with the number of double bonds requires further investigations. Accordingly, the most potent activators were both one of the shortest one, C18:2 n-6 (IPUFA/I0 = 24.8 ± 3.3; current density = 353.7 ± 22.4 pA/pF), and one of the longest one, C22:6 n-3 (IPUFA/I0 = 29.8 ± 4.4; current density = 276.8 ± 24.5 pA/pF). Conversely, C18:2 n-6 is one of the most potent activator but C18:3 n-3 failed to activate TREK-1 channel (IPUFA/I0 = 1.4 ± 0.1; Current density = 20.7 ± 4.2 pA/pF). Like C18:3 n-3, the saturated stearic acid (C18:0) had no effect on TREK-1 while the mono-unsaturated C18:1 n-9 produced a 7.4 ± 1.9-fold increase of ITREK−1 (n = 12, Table 1). However, statistical analysis failed to discriminate the PUFA’s effects (IPUFA/I0 parameters compared with a nonparametric kruskall-wallis test) probably due to the important variability of the effects.
The variability in PUFA responses is related to the variable initial current density
Despite an important number of cells, we observed a large variability of the TREK-1 current activation by PUFAs. The severity of the inclusion criteria (see Material and Methods section) suggests that the variability observed in PUFAs responses is inherent to TREK-1 channel.
Figure 3A illustrates the large variability of PUFA effects based on the fold-increase analysis (IPUFA/I0). Te variability of the PUFA-induced current (IPUFA) appeared less wide (Fig. 3B). Indeed, the calculation of the coefficient of variation (CV = \(\frac{SD}{Mean}\)) indicated that the dispersion of the IPUFA/I0 calculation is higher than the dispersion of the IPUFA at steady-state of the activation (Fig. 3C). Thus, CV modification is in accordance with the hypothesis that the variability of I0 might be responsible of the variability of the IPUFA/I0 parameter. We demonstrated that the variability of the IPUFA/I0 parameter did not resulted from the variability of current density at steady-state after the application of PUFAs (IPUFA) (Fig. 4) but from the variability of the initial current I0 (Fig. 5A-B).
To better characterize the relationship between the effects of PUFAs and I0, we plotted “ fold-increase of TREK-1 current”=f(“initial current”). As shown in Fig. 5A, there is a non-linear relationship between IPUFA/I0 and I0. This negative relationship can be linearized by log-transforming the data (Log10) (Fig. 5B, Table 2). Thus, the effects of all PUFAs but C22:5 n-3 depends on I0, independently of the absolute amplitude of IPUFA. This inverse relationship indicates that the variability of I0 is not mainly due to a variable level of TREK-1 expression. To determine if the observed variability is linked to the cellular model that we used, or not, we also performed some experiments on two other models: another stable model of TREK-1 overexpression21 and transiently transfected HEK 293T cells with TREK-1 (pIRES2 KCNK2 WT) (Fig. 5C). The IPUFA/I0 variability observed can be explained by the variety of constitutively active TREK-1 channels in resting condition.
Table 2
Parameters of the linear regression Log10 (IPUFA/I0) = f(Log10(I0)). TREK-1 current was recorded in whole-celle configuration of patch-clamp.The p-value indicates the significativity of the relationship, the R2 indicates the goodness of the fit and the Y-intercept reflects the fold-activation of TREK-1 current for an unitary current as Log(1) = 0.
| n | p-value | R² | Equation | Y-intercept ± SD |
C18:2 n-6 | 20 | 0.0001 | 0.82 | Y = -0.9378*X + 2.343 | 2.3 ± 0.1 |
C20:2 n-6 | 7 | 0.027 | 0.66 | Y = -0.5264*X + 1.621 | 1.6 ± 0.2 |
C20:4 n-3 | 12 | 0.0001 | 0.80 | Y = -0.6305*X + 1.672 | 1.7 ± 0.1 |
C20:4 n-6 | 16 | 0.002 | 0.50 | Y = -0.8315*X + 2.083 | 2.1 ± 0.3 |
C20:5 n-3 | 17 | 0.0008 | 0.53 | Y = -0.7172*X + 1.958 | 1.9 ± 0.2 |
C22:5 n-6 | 11 | 0.0001 | 0.84 | Y = -1.073*X + 2.462 | 2.5 ± 0.2 |
C22:5 n-3 | 12 | 0.371 | 0.08 | Y = -0.3221*X + 1.378 | 1.4 ± 0.4 |
C22:6 n-3 | 20 | 0.0001 | 0.68 | Y = -0.7378*X + 2.133 | 2.1 ± 0.1 |
By linear regression analysis, we determined the Y-intercept which reflects the fold-activation of an unitary current TREK-1 current (Log(1) = 0). The Y-intercept of DPA n-3 cannot be calculated since there was no correlation between IPUFA/I0 and I0. For the 7 others PUFAs, the Y-intercept values allow their separation into 3 groups with the following activation sequence : C22:6 n-3, C22:5 n-6, C18:2 n-6 > C20:5 n-3, C20:4 n-6 > C20:4 n-3, C20:2 n-6 (Table 2). Then, plotting the relationship between Y-intercept and the number of carbons revealed no correlation between the fold-activation of TREK-1 and the acyl chain length of PUFAs (Fig. 5D) or with the number of double bounds (Fig. 5E). In conclusion, TREK-1 activation by PUFAs is not dependent of the acyl chain length and the number of double bond (Fig. 2B-E and Fig. 5D-E).
Variable activation rates suggest different PUFA binding affinities for TREK-1 channel
To explore a new mode of action of PUFAs on TREK-1 channel, we compared the kinetics of activation of TREK-1 perfusing PUFAs or ML402, a binding activator of TREK-1. Figure 6A to C represent the mean ± SEM of the normalized current densities (\(\frac{I-{I}_{0}}{{I}_{PUFA}-{I}_{0}}\)) over the time in response to PUFA and ML402. A steady-state of activation was reached when 3 successive traces were stable. The half-activation of TREK-1 by each PUFA and ML402 are presented in the Table 3. We were able to distinguish at least two types of activation rate, a fast one with an half-activation less than 3 min (C18:2 n-6, C22:6 n-3 and ML402) and a slower one above 3 min (C20:2 n-6, C20:4 n-3, C20:4 n-6, C20:5 n-3, C22:5 n-6 and C22:5 n-3). The Fig. 6D shows no statistical correlation between the acyl chain lengh, and thus their ability to incorporate into the membrane, and the half-activation (Spearman correlation test : p-value > 0.99). As reported in Table 3, the averaged half-activation of TREK-1 channel was lower for C18:2 n-6 and C22:6 n-3, both having comparable kinetics to those observed for ML402. Although there were no significant differences between these three compounds and C22:5 n-6 and C20:5 n-3, the kinetics of the latters appeared slower (Fig. 6A, B and D, Table 3). In contrast, C20:4 n-3, C20:4 n-6 and C22:5 n-3 had slower kinetics with an averaged half-activation close to 4 min (Table 3). Since C18:2 n-6 and C22:6 n-3 have the same fast activation and C22:5 n-3 has the slowest one, we assumed that the activation rate of the TREK-1 by PUFAs does not depend on the acyl chain length (Fig. 6D). However, among the PUFAs, there is a negative correlation between the half-activation (min) and the fold-increase of TREK-1 current (IPUFA/I0) (Spearman correlation test : p-value = 0.002 and r = -0.93; Fig. 6E). As the stronger activators are the faster activators of TREK-1, we propose that some PUFAs, such as C18:2 n-6 and C22:6 n-3, have a higher binding affinity for TREK-1 which would allow them to activate it faster and stronger.
Table 3
Parameters of the activation kinetic of TREK-1 by PUFAs and ML402 in whole-cell configuration of patch-clamp. Values indicate the half-activation time (min) and the steady-state (min) of the activation of TREK-1. The statistical comparison of the half-activation time for each PUFAs was performed with a Kruskal-Wallis test followed with the post-hoc Dunn’s test. Two rows having the same letter are not significantly different. Data are expressed as mean \(\pm\) SEM.
| | Half-activation (min) | Statistical differences | Steady-state (min) |
| n | Mean ± SEM | Mean ± SEM |
C18:2 n-6 | 20 | 1.86 ± 0.16 | a | 3.65 ± 0.28 |
C20:2 n-6 | 7 | 4.61 ± 0.30 | b,c | 7.40 ± 0.25 |
C20:4 n-3 | 12 | 4.56 ± 0.24 | b,c | 7.13 ± 0.23 |
C20:4 n-6 | 16 | 3.97 ± 0.31 | b,c | 6.67 ± 0.34 |
C20:5 n-3 | 17 | 3.18 ± 0.30 | b,c | 5.65 ± 0.33 |
C22:5 n-6 | 11 | 3.12 ± 0.30 | a,c | 5.31 ± 0.41 |
C22:5 n-3 | 12 | 4.18 ± 0.35 | b,c | 5.95 ± 0.37 |
C22:6 n-3 | 20 | 2.26 ± 0.19 | a | 4.34 ± 0.27 |
ML402 | 19 | 2.11 ± 0.16 | a | 3.86 ± 0.33 |
Activation of TREK-1 channel by PUFAs is fully reversible.
To see if the activation of TREK-1 channel by PUFA is due to an insertion and thus a modification of membrane tension, we looked at the reversibility. We focused on C18:2 n-6, C20:5 n-3 and C22:6 n-3, the most potent activators (Table 1). ML402 activation reversed immediately and 50% of effective reversibility occurred in less than 1 min (Fig. 7A and 7B). C20:5 n-3, had a kinetic of washing (washout time 50%: 0.9 ± 0.1 min) comparable to ML402 (Fig. 7A and 7B). Even though the washout of C18:2 n-6 and C22:6 n-3 was slower than ML402 (washout time 50%: 2.4 ± 0.2 min, 3.7 ± 0.3 min and 0.4 ± 0.04 min, respectively), PUFAs effects were also fully reversed under washing. No correlation between the acyl chain length (Fig. 7C, Spearman correlation test: p-value > 0.99) or the number of double bounds (Fig. 7D, Spearman correlation test: p-value > 0.99) and the time to have 50% of effective reversibility of the TREK-1 activation was found. C18:2 n-6 and C22:6 n-3, that activated TREK-1 at least twice more than ML402 (I/I0: 24.8 ± 3.3, 29.8 ± 4.4 and 9.6 ± 0.9, respectively), had a total reversibility in few minutes. At this point, we cannot exclude a membrane insertion of PUFAs, but we assume that the main effects of PUFAs on TREK-1 activation could be a direct and reversible interaction of PUFAs with the channel as it is known for KCNQ122 and for the Shaker H4 Kv channel23.
Alteration of membrane curvature or fluidity did not explain PUFAs activation of TREK-1 channel.
In order to evaluate the membrane curvature and tension effects on the PUFAs-induced TREK-1 activation, we performed experiments in the inside-out configuration of the patch-clamp technique (+ 30 mV, symmetrical condition: 145 mM KCl). In this configuration, we were able to superfuse molecules at the inner face of the membrane and PUFAs must induce an aopposite curvature of the membrane than in the whole-cell configuration. As shown in Fig. 8A, ML402 superfusion at the inner face of the membrane reversibly increased TREK-1 current (Table 4). Kinetics of activation and reversibility were comparable to those obtained in the whole-cell configuration (Fig. 6 and Fig. 7) suggesting that the ML402-binding site is accessible from the outer and the inner membrane leaflet. Interestingly, a comparable reversible activation of TREK-1 channel was obtained for C18:2 n-6 and C22:6 n-3, 5 µM, suggesting that the membrane curvature is not involved in the activation of TREK-1 channel by PUFAs (Fig. 8B and 8C, Table 4). In the other hand, the inside-out experiments showed that PUFAs activation occured independently of scaffold proteins like PLD2 known to modulate TREK-1 activity in response to disruption of lipid rafts24,25.
Table 4
Descriptive statistics for TREK-1 activation and reversibility kinetics in inside-out configuration of patch-clamp. Values indicate the steady-state time (min) and the time to have 50% of effective reversibility of TREK-1 activation (s).
| | | Steady-state (min) | 50% of reversibility (s) |
| n | Concentration (µM) | Mean ± SEM | Mean ± SEM |
C18:2 n-6 | 4 | 5 | 4.6 ± 0.4 | 18.3 ± 6.3 |
C22:6 n-3 | 6 | 5 | 2.9 ± 0.3 | 18.3 ± 4.2 |
ML402 | 6 | 10 | 2.6 ± 0.6 | 15.5 ± 7.6 |
Then, we assessed the membrane fluidity changes during PUFAs application with a pyrenedecanoic acid probe (PDA), analog to lipids. By measuring the ratio of PDA monomer to excimer fluorescence (405nm/470nm ratio), a quantitative assesment of the membrane fluidity can be obtained at different time points by following the ratio modification over time (F/F0-1). We focus our experiments on C18:2 n-6 and C22:6 n-3 the stronger activators of TREK-1 at 10 µM and C18:3 n-3 that failed to activate TREK-1. These 3 PUFAs at 10 µM did not modify the membrane fluidity even after 50 minutes of application, while at 100 µM they induced a decrease of F/F0-1 from t0 compared to the control condition (basic extracellular medium). These results indicate that membrane fluidity is not modified by PUFAs at 10 µM (Fig. 9A-C), at least within 50 minutes of application. In addition, given that TREK-1 activation starts at 1 minute of perfusion of C18:2 n-6 and C22:6 n-3 10 µM (Fig. 6) it is unlikely that PUFA effects on TREK-1 activation are due to an increase in membrane fluidity. Altogether, these data suggest that at least both C18:2 n-6 and C22:6 n-3 PUFAs activate TREK-1 channel by direct interaction with TREK-1 protein and not by a modification of the membrane fluidity.
DHA interacts directly with TREK-1 channel protein in TREK-1 enriched microsomes
In order to assess a potential direct PUFA-TREK-1 interaction, we purified microsomes from hTREK-1/ HEK and native HEK 293T cells and labeled lysine residues of the total proteins to perform affinity measurements using Spectral Shift (SpS). Data of the affinity curve are expressed as mean \(\pm\) SEM of 5 to 9 experiments from different batches of microsomes of “hTREK-1/ HEK” and “native HEK 293T” cells. TREK-1 protein quantity is constant within the 16 samples from microsomes from hTREK-1/ HEK. At first glance, we observed similar affinity from C22:6 n-3 in these two type of microsomes : Kd,TREK−1 ~ 50 µM and Kd,HEK ~ 100 µM highlighting a similar mode of association of C22:6 n-3 within the microsomes. However, the SpS signal displayed subtle differences according to whether or not microsomes were enriched in TREK-1 protein. Knowing SpS signal rises from fluorescence recorded by two individual channel, and having a strong reproducibility from each condition, we can derive the following assumption:
$${\text{I}}_{{\lambda }}\left(\text{T}\text{R}\text{E}\text{K}1,\text{H}\text{E}\text{K}\right)= {\text{I}}_{{\lambda }}\left(\text{T}\text{R}\text{E}\text{K}1\right)+ {\text{I}}_{{\lambda }}\left(\text{H}\text{E}\text{K}\right)$$
(1)
Where for each individual wavelength (λ670 nm and λ650 nm), the fluorescence recorded for the TREK-1-enriched microscomes ( Iλ(TREK1, HEK)) correspond to both the fluorescence from the labelled empty microsomes (Iλ(HEK)) and from the labelled TREK-1 (Iλ(TREK1)).
Being able to isolate the specific fluorescence associated to TREK-1 \(\left({\text{I}}_{{\lambda }}\left(\text{T}\text{R}\text{E}\text{K}1\right)\right)\) within the TREK-1-enriched microsomes (\({\text{I}}_{{\lambda }}\left(\text{T}\text{R}\text{E}\text{K}1,\text{H}\text{E}\text{K}\right))\), we can use individual channel for further calculation using Eq. 2, isolating the SpS signal associated only to TREK-1 by normalizing out the background signal coming for the free microsomes.
$$\text{R}\left(\text{T}\text{R}\text{E}\text{K}1\right)=\frac{{{\text{I}}_{670}(\text{T}\text{R}\text{E}\text{K}1,\text{H}\text{E}\text{K})-{\text{I}}_{670}\left(\text{H}\text{E}\text{K}\right)}_{}}{{\text{I}}_{650}\left(\text{T}\text{R}\text{E}\text{K}1,\text{H}\text{E}\text{K}\right)-{\text{I}}_{650}\left(\text{H}\text{E}\text{K}\right)}$$
(2)
When performing so, the specific TREK-1 dose-response gives a Kd = 44 µM for C22:6 n-3 (Fig. 10). Despite displaying similar affinities, the TREK-1-enriched microsomes shows a statistically better affinity than the microsomes themselves. However, it is worth noticing only a 2-fold increase of affinity, which may highlight a similar binding mode for both interactions. Altogether, this suggests an interaction mediated by the lipid bilayer, such as a membrane insertion followed by interaction with TREK-1 channel.