Membrane tension‐dependent conformational change of Isoleucine 106 of loop B diminishes water permeability in FaPIP2;1

Abstract

Aquaporins (AQPs) are membrane proteins specialized in facilitating water transport across membranes. Mechanical stress is one of the various stimuli that regulate AQPs. Briefly, there are several studies that report a decrease in permeability upon an increase in membrane tension. However, the molecular details of this mechanosensitive (MS) response are still a matter of debate. Our work attempts to close that gap in knowledge by providing evidence of a conformational change that occurs inside the pore of the strawberry aquaporin FaPIP2;1. Via osmotic shock experiments and molecular dynamics (MD) simulations, we found that a residue of loop B, I106, is key to the blocking of the permeation pathway and such a change is almost exclusively found under membrane tensile stress. In detail, osmotic shock experiments exhibited a nonlinear increment in water fluxes for increasing osmolarities, evidencing a decrease in the FaPIP2;1 permeability. MD simulations under membrane tension showed the same trend, with a significant increase in states with a low water permeability. The latter was correlated with a conformational change in I106 that generates a permeation barrier of around 18 kJ mol⁻¹, effectively closing the pore. This work constitutes the first report of a PIP type aquaporin reacting to tensile stress in the membrane. Our findings could pave the way to test whether this conformational change is also responsible for mechanical gating in the other MS aquaporins, both those already reported and those still waiting to be found.

1. INTRODUCTION

Aquaporins (AQPs) represent a vast family of transmembrane proteins responsible for selectively facilitating the transport of water and other small molecules across lipid bilayers. Their ubiquitous presence spans across all kingdoms of life (Ishibashi et al. 2017).

AQPs belong to the superfamily of Major Intrinsic Proteins (MIPs), identified by the presence of the NPA motif occurring twice within their sequence, possibly originated by a genetic duplication phenomenon during evolution (Zardoya and Villalba 2001). Their widespread distribution is accompanied by a significant level of conservation in both their sequence and structure (Ozu et al. 2022). Once in the membrane, all known AQPs assemble into tetramers. Each monomer comprises six transmembrane helical segments and two loops (B and E) that form short helical segments embedded in the membrane, which contain the fingerprint NPA motifs (Ozu et al. 2018).

Each monomer constitutes a functional water channel with an hourglass‐resembling shape (Jung et al. 1994). The structure reveals two wide atria, one facing the extracellular medium and the other facing the cell interior. Both atria extend towards the center of the membrane in a narrow pore, that is, the water pathway (Murata et al. 2000; Ren et al. ²⁰⁰⁰; Sui et al. ²⁰⁰¹). The diameter of the water pathway allows the passage of only one water molecule at a time, so water crosses this region forming a single file (de Groot and Grubmüller ²⁰⁰¹; Tajkhorshid et al. ²⁰⁰²). The single‐file region is 15–20 Å long, and it is delimited by the selectivity filter (SF)—also known as the aromatic‐Arginine filter (ar/R)—at the extracellular extreme, and the cytoplasmic mouth or cytoplasmic extreme. The mid‐part of the single‐file region is at the membrane center and aligns with the NPA filter. These three regions—the SF, the intracellular mouth or extreme of the single file, and the NPA filter—are key regulation sites in the water pathway (Ozu et al. 2018).

Both ar/R and NPA filters have a well‐defined function. In water‐transporting AQPs, the SF is a size‐selective barrier for the transported molecules (Beitz et al. 2006; Fu et al. ²⁰⁰⁰; Hub and de Groot ²⁰⁰⁸; Sui et al. ²⁰⁰¹; Thomas et al. ²⁰⁰²), while the NPA filter constitutes an electrostatic barrier for proton passage and governs the dipole orientation of water molecules (de Groot and Grubmüller ²⁰⁰⁵). In contrast, the cytoplasmic extreme of the single‐file region lacks a unique function, as some residues located within this site were described to be involved in various regulatory effects in animal and plant AQPs. In human AQP4, H95 behaves as a pH‐dependent gate (Kaptan et al. 2015). However, in PIP2‐type plant aquaporins the pH regulation is different. The mechanism was described in detail in SoPIP2;1 and involves the movement of the intracellular loop D towards the cytoplasmic atria. Hence, loop D blocks the cytoplasmic entrance defining a closed conformation that is stabilized by interactions between key residues from loops D and B (Frick et al. 2013; Törnroth‐Horsefield et al. ²⁰⁰⁶). The response of PIP aquaporins to pH is highly conserved in the PIP2 subfamily, with the mechanism believed to be conserved as well (Scochera et al. 2022). Besides the pH regulation, other regulatory mechanisms have been described in AQPs from different origins, such as phosphorylation, trafficking, intracellular calcium concentration, voltage, and even heterotetramerization (Garate et al. 2011; Hub et al. ²⁰¹⁰; Mom et al. ²⁰²¹; Ozu et al. ²⁰²²). Other studies have even suggested that AQPs would act as osmosensors, with this function potentially being more relevant than their capacity to transport water (Hill et al. 2004). In this regard, mechanosensitive AQPs (MS‐AQPs) could potentially be the solution for osmosensitivity (Hill and Shachar‐Hill ²⁰¹⁵), particularly in responding to situations of abrupt and potentially harmful cell swelling caused by hypo‐osmotic shocks. These ideas gained relevance with the discovery of MS‐AQPs. To date, numerous studies provide evidence that certain and specific aquaporins can show mechanical regulation (Ozu et al. 2023). However, the details of the molecular mechanism are still unknown.

In the study of MS‐AQPs, experimental evidence demonstrates that the water permeability (p _f ) decreases as membrane tension increases. This response seems to be a general feature of mechanosensitivity in aquaporins since it has been observed in members from different life domains (Ozu et al. 2023), that is, AQY1 from yeast (Soveral et al. 2008), VvTIP2;1 from grapevine (Leitão et al. 2014), BvTIP1;2 from the red beet (Goldman et al. 2017), AQP4 from rat (Tong et al. 2012), and AQP1 from human (Ozu et al. 2013). However, a PIP2‐type aquaporin—the BvPIP2;1 from red beet—showed no signs of mechanosensitivity (Goldman et al. 2017). This is intriguing because human AQP1 and BvTIP1;2 are mechanosensitive (Goldman et al. 2017; Ozu et al. ²⁰¹³), but AQP1‐type and PIP2‐type aquaporins share a common ancestor, which is different from the ancestor of AQP8‐type and TIP‐type aquaporins (Soto et al. 2012). The fact that BvPIP2;1 is not a MS‐AQP suggests that PIP2‐type aquaporins lost their mechanosensitive properties during evolution. However, BvPIP2;1 is not a usual PIP2, since it differs from other PIP2 members in two conserved features: (1) it does not form heterotetramers with PIP1 members and (2) it shows a different response to pH changes (Jozefkowicz et al. 2016). Therefore, the possibility that typical PIP2‐type AQPs present mechanosensitive features is still open.

Unlike mechanosensitive ion channels, where the molecular elements involved in gating mechanisms are well characterized, in MS‐AQPs, the structural components governing mechanical gating and the underlying molecular mechanisms remain elusive (Cox et al. 2018; Ozu et al. ²⁰²³). In this regard, we recently proposed that the cytoplasmic extreme of the single‐file region is a key site for mechanical regulation in AQPs (Ozu et al. 2018). We suggested that the clues for a mechanosensitive mechanism are placed in the cytoplasmic extreme of the single file due to four reasons: (1) this region is formed by residues from loop B and transmembrane segments that are donors/acceptors of H‐bonds, such as glycine and histidine; (2) the water osmotic permeability (p _f ) decreases as the number of H‐bonds that water molecules form with spanning residues increases along the water pathway (Horner et al. 2015); (3) a GxxxG sequence—characteristic of transmembrane segments (Russ and Engelman 2000)—is present in loop B just before the first NPA motif, and sequences of this type have been shown to be involved in mechanosensitivity in ion channels (Balleza 2011); and (4) molecular dynamics (MD) studies in AQP4 showed that the p _f diminishes as membrane thickness decreases, due to an elongation of the single file in the cytoplasmic extreme of the water pathway (Tong et al. 2016).

In the present study, we combine both experimental and in silico approaches to study both the mechanosensitivity in a plant aquaporin from the PIP2‐type subfamily—FaPIP2;1 from strawberry—and the molecular features involved in the functional response. Our results demonstrate that (i) FaPIP2;1 is a MS‐AQP and (ii) strongly suggest that I106, from loop B, is a key residue at the cytoplasmic extreme of the single‐file region that flips into the lumen of the water channel with tension increments, leading to a reduction in permeability by blocking the water permeation pathway. Overall, this research represents a crucial first step towards uncovering the molecular determinants of mechanosensitive gating in MS‐AQPs.

2. RESULTS

2.1. FaPIP2;1 behaves like mechanosensitive aquaporins

To study the mechanosensitivity of FaPIP2;1 we employed the same strategy used for hAQP1 (Ozu et al. 2013) and BvTIP1;2 (Goldman et al. 2017). Osmotic shock experiments were performed using Xenopus oocytes injected with 5 ng of FaPIP2;1 cRNA. Each pool of oocytes was separated into three groups. Then, each group was subjected to osmotic gradients of different magnitude (40, 90, and 140 mosmol Kg_w ⁻¹). This experimental series was repeated with oocytes from at least three different frogs. In each experiment, the initial water flux was determined from the slope of the relative volume time course during the first 10 s of the osmotic response. Figure 1a shows a representative experiment. As expected, the higher the gradient, the faster the cell volume changes (Figure 1a). However, the relationship between water flux (J _w ) and the applied gradient (Δosm) is not linear in FaPIP2;1‐injected oocytes (Figure 1b), indicating that J _w is not as high as expected according to the osmosis law (Equation (1)). This is a characteristic behavior of mechanosensitive aquaporins, not observed in gramicidin channels and the PIP2‐type aquaporin BvPIP2;1 (Goldman et al. 2017; Ozu et al. ²⁰¹³). Both control experiments performed with water‐injected oocytes and amphotericin‐incubated oocytes exhibit linear responses on their J _w –Δosm curves (Figure 1b), following the osmosis law. Like in mechanosensitive aquaporins (Goldman et al. 2017; Ozu et al. ^{2013, 2023}), the J _w –Δosm relationship evidences a water permeability (p _f ) decrease in FaPIP2;1‐expressing oocytes but not in controls injected with water or incubated with amphotericin (Figure 1c). To discard possible effects of the vitelline envelope of the oocyte we evaluated its role on these mechanical responses (Figure S1, Supporting Information). The results indicate that although the vitelline envelope can constitute a physical barrier for the incorporation of amphotericin to the oocyte membrane (Figure S1a), its presence has not any relevant effect on the FaPIP2;1 mechanosensitive response (Figure S1b,c). Altogether, the experimental results indicate that FaPIP2;1 responds to osmotic changes like a MS‐AQP.

FIGURE 1.

FaPIP2;1 shows the functional response of mechanosensitive aquaporins. (a) Relative volume (V/V₀) time course in FaPIP2;1‐injected oocytes (5 ng RNA/oocyte) under different osmotic gradients (Δosm = 40; 90; 140 mOsmol Kg_w ⁻¹). The record with water‐injected oocytes (control) tested with Δosm = 140 mOsmol Kg_w ⁻¹ is also shown. Results are presented as mean ±SEM from one typical experiment with 5–6 oocytes in each condition. (b) Relationship between water flux (J _w ) and Δosm registered 10 s after the onset of the hypo‐osmotic conditions shown in (a). Control experiments with amphotericin B‐incubated oocytes are also shown. Results are presented as mean ±SEM from three independent experiments. (c) Osmotic permeability coefficient (p _f ) determined from data shown in (b). Membrane tension values were calculated from mechanical characterizations performed previously in Xenopus oocytes (Goldman et al. 2017; Ozu et al. ²⁰¹³). The tension values reached at 10 s from the onset of the osmotic gradient were 11.0 ± 0.2; 11.3 ± 0.2; and 11.6 ± 0.3 mN m⁻¹ for Δosm = 40; 90; and 140 mOsmol Kg_w ⁻¹, respectively. The symbols (*) indicate significant differences (p < 0.05) with the 40 mM osmotic gradient.

2.2. A tension increase in MD simulations reduces water permeation in FaPIP2;1

Despite the wealth of molecular details available concerning water conduction in AQPs (due to the existence of high‐resolution structures and computer simulations), evidence on the molecular determinants associated with a potential mechanosensitive gate remains elusive. To bridge this gap, we developed a structural model of FaPIP2;1 using the crystallographic structure of SoPIP2;1 as a template, an ortholog from Spinacia oleracea that exhibits the highest sequence similarity (Törnroth‐Horsefield et al. ²⁰⁰⁶). As the volume increase driven by an osmotic shock rises the membrane tension (Ozu et al. 2013, 2023), then we conducted classical molecular dynamics (MD) simulations with sequential increments in membrane tension, beginning form the condition of membrane tension equal zero.

Figure 2a illustrates the relationship between the FaPIP2;1 single pore or unitary water permeability coefficient (p _f ) and tension in MD simulations. The results show a nonlinear decay of p _f with increasing tension, in line with the in vitro experiments with Xenopus oocytes (Figure 1c). We observe that even at 6 mN m⁻¹ well below the tension estimated from the cell swelling experiments, FaPIP2;1 exhibits a significant reduction in its permeability. Figure 2b presents the local p _f values along the permeation pathway. A clear distinction emerges between a non‐stretched and a stretched membrane system between the −10 and 6 Å marks. This is the region delimited by the NPA‐conserved motif and the cytoplasmic end of the permeation pathway. In this area, a significant decrease in p _f is observed when tension is applied (Figure 2b). This reduction reaches its peak at around the −4 Å mark, this is where H105 and I106 from loop B are located, both depicted in Figure 2c in red and yellow, respectively. Therefore, these residues are strong candidates for a tension‐associated gate in FaPIP2;1. Thus, we delved into tension‐dependent conformational changes of these residues and assessed whether they are related to a tension‐dependent permeability reduction in FaPIP2;1.

FIGURE 2.

Membrane‐tension increments induce permeability decrease within the channel's single‐file region, specifically in the vicinity of the loop B NPA motif. (a) Average (over four pores and three replicas) unitary permeability coefficient of FaPIP2;1 obtained from molecular dynamics simulations under different membrane tension values. The errors represent the SEM of the average p _f of each replica. Simulations under control conditions (0 mN m⁻¹) were run for 500 ns. Then, increasing membrane tension values were sequentially applied in four 125 ns‐steps. The values of p _f shown are those of the pore region average, from z = 12 to z = −12 (Å). (b) p _f variation along the water pathway. Results represent the p _f value observed in the steady state reached with each tension tested. SEM represented as color filled regions. (c) Cartoon representation of one FaPIP2;1 monomer in open state. Both key residues from the selectivity filter (SF) Histidine 216 (violet) and Arginine 231 (orange), as well as Histidine 105 (red) and Isoleucine 106 (yellow) at the cytoplasmic extreme of the single‐file region are represented in CPK. The visualization render was performed with VMD (Humphrey et al. 1996).

2.3. Functionally closed states are more populated under tension

AQPs possess both highly conserved sequences and structures. However, they exhibit a broad range of water conduction capacities. This suggests that, in many cases, open (highly conductive) and closed (low conductive) states are influenced by small movements in the lateral chains of the residues lining the pore (Hashido et al. 2007). Consequently, determining the openness of the protein based on significant structural changes is not a straightforward task. Given this limitation, we functionally defined the open and closed states of the channel based on the permeation events (PE) rates. We defined two representative functional states: one associated with a high rate of permeation events (Open), and the other with a low rate of permeation events (Closed); for more details see Figure S2. With this functional definition, many structural factors could render these states, such as configurations that modify the SF region, alter the single water configuration, or block the pore, among others. Another point is that these closed states also appear under no tension. Spontaneous closing events have been observed in other aquaporins as well (Garate et al. 2011). As shown in Table 1, there is a monotonic increase in the abundance of closed states as tension increases, reaching a peak of 54.1% with the highest tension. Thus, functionally closed states seem to be related to tension increase. As we already established a significant reduction in p _f within the region close to H105 and I106, in the next section we will evaluate whether there are conformational changes in these two residues that can be associated with a tension increment and closed states.

TABLE 1.

Key functional and structural metrics per tetramer under tension.

Tension (mN m−1)	p f (10−14 cm3 s−1) a	Monomers in closed states (%)	I106 in closed state (%)	H105 in closed state (%)
0	2.6 ± 0.8 (2.4 ± 0.5)	38.73 (17.64)	2.04 (7.37)	18.35 (1.73)
6	1.4 ± 0.1 (1.58 ± 0.6)	45.74 (63.99)	7.09 (46.94)	18.04 (0.00)
12	1.2 ± 0.3	48.19	24.89	14.77
24	1.1 ± 0.2	51.82	24.66	23.92
50	1.1 ± 0.3	54.12	31.96	25.01
Total under tension	1.2 ± 0.5	49.97	22.15	20.44

Note: Unitary osmotic permeability (pf), the number of closed states and structural metrics (the number of states with I106 and H105 flipped, denoted as I106 in closed state and H105 in closed state, respectively) were averaged across tetramers (n = 12) from all replicas (N = 3). In all cases, metrics were estimated, and counts were recorded from a 125 ns simulation in each treatment (out of a total of 500,000 conformations with 1000 conformations per ns sampled). pf values were obtained for each tetramer. Values in parentheses were obtained from two replicas: 2 ms under 0 (control) followed by 2 ms at 6 mN m⁻¹.

^a p _f values were averaged over the four monomers (per replica) and then for the three or two replicas (long simulations). The errors represent the SEM of the three replicas.

2.4. I106 obstructs the water pathway at the cytoplasmic extreme of the single‐file region in a tension‐dependent manner

In line with the p _f reduction under tension, there is an increase in the population of closed states. Can we pinpoint key residues in these functionally defined closed states? In Figure 2, H105 and I106 appeared as strong candidates for a tension‐associated gate due to their proximity to the region where there was a noticeable p _f reduction. To characterize the different conformational states of these residues and their relationship with closure and tension, we determined the distributions of the distance of their sidechains to the pore center. Likewise, we monitored the distributions of the distance between the SF residues, H216 and R231 to assess any narrowing of this region related to tension (Figure S3). I106 is next to the first NPA motif (Figure 3a) and in the conformed monomer it is between the cytoplasmic extreme of the single‐file region and the NPA filter (Figure 2d). Figure 3b presents a representative snapshot of the I106 displacement towards the center of the pore that effectively closes the water pathway, which is evidenced by a decrease in the slope of accumulated PE. This displacement is the result of, on average, an 80° rotation of the 𝜓 torsional angle. Is this displacement related to tension? In Figure 3c we present the I106 distance distributions collected for all monomers at different tensions. In all cases, the distributions are bimodal with two normal distributions peaking at 7 and 1.7 Å, respectively. As these two distributions are well separated, members of the former and latter distributions are distinct states. As tension increases, there is a shift towards smaller distances, revealing that the I106 movement towards the pore center is tension‐dependent and blocks the pore. As closed states are not necessarily defined by a unique structural feature, when grouping by states we also observe the shift towards lower distances, further confirming the relationship among tension, the I106 displacement, and closure (Figure 3d). The same analysis performed on H105 does not reflect a clear relationship between tension and displacement towards the pore center (Figure S3a, top). However, this tension‐independent displacement is related to a pore closure (Figure S3a, bottom). When comparing both distributions simultaneously for samples grouped by states and treatment (no tension, and tension) as shown in Figure S4, we observe that the closures triggered by H105 or I106 are independent. As shown in Table 1, of closed conformations in the control group (no tension), around 20% exhibited the rotation of H105 and 2% displayed the rotation of I106. Conversely, flipped H105 populations under tension still represent 20% of closed conformations, while flipped I106 populations monotonically increase in the close states, reaching a value of around 32%. Another potential source of closed states is the narrowing of the SF region. As shown in Figure S4b, the distributions of the H216‐R231 distances are altered with tension changes, with higher tensions partially promoting a narrower SF which is slightly more abundant in closed states (Figure S4b, bottom). However, when comparing to I106, the effect on the SF is not so consistent and pronounced (for all tension values) since the narrowing of the SF is less than 2 Å, whereas I106 moves around 5 Å. To check for finite‐time effects, we ran two additional MD replicas at the lowest tension, that is, 6 mN m⁻¹. These simulations consisted of 2.0 μs free MD and immediately after, a 2.0‐μs simulation subjected to the aforementioned tension. As shown in Figure S5 and Table 1, a p _f reduction is observed and the I106 distribution is shifted towards lower values while there are no changes in the populations of H105 distances. It is interesting to notice that for these long simulations, the percentage of I106 flipped states within the closed states group is around 60%, even higher than the values reported for the largest tension. This suggests that (i) I106 is responsive to relatively low‐tension values and (ii) the closing process is at least in the order of microseconds. Overall, our MD data is consistent and strongly suggests that FaPIP2;1 displays a tension‐dependent conformational change in I106 that promotes a closed state and points towards a valve‐like closing mechanism.

FIGURE 3.

The membrane tension‐dependent water permeability decrease observed in FaPIP2;1 is related to the conformational change of I106. (a) Schematic representation of FaPIP2;1. Both H105 (soft red) and I106 (yellow) are next to the highly conserved NPA motif (magenta) from loop B. (b) Time course of the I106 δ‐carbon position, relative to the center of the permeation pathway (orange registry). The cumulative permeation events is superimposed (black time series). Extracellular views of the channel with I106 at a side of the lumen (open state; on the left) or flipped to the center of the permeation pathway (closed state; on the right) are shown above the graph. (c) Distribution of the distances between I106 δ‐carbon and the center of the pore. Tension increments are represented by different colors, from light green (0 mN m⁻¹; control condition) to light red (50 mN m⁻¹). (d) Distribution of the distances between I106 δ‐carbon and the center of the pore in functionally defined open or closed states.

2.5. The tension‐mediated flipping of I106 alters the water structure in the single‐file region

To understand how the tension‐dependent conformational change of I106 affects the permeation capabilities of FaPIP2;1, we estimated the hydration profiles of open and closed states determined by the tension‐favored I106 displacement. In Figure 4a, we present the water density—relative to the bulk (P/P ₀)—along the permeation pathway. A clear drop in the probability of finding water molecules in the I106 region is observed. This indicates that I106 has significant impact on the ability to allow water passage through the pore when this residue is blocking the pathway. The fact that the water line is interrupted is a major obstacle in permeation, effectively stopping it (Figure 3b).

FIGURE 4.

Closed states modify the hydration profile in the FaPIP2;1 single‐file region. The ordering of water molecules in the single file of FaPIP2;1 is altered by membrane tension increments. (a) Relative water number density with respect to the bulk (P ₀) along the permeation pathway. Average number of water molecules along the single‐file region of FaPIP2;1. Data were grouped by the conformational state of I106. SEM represented as color filled regions. (b) PMF calculated from the hydration profile (a), showing a noticeable energetic barrier in the area corresponding to the flipped Ile106. SEM represented as color filled regions.

As data in Figure 4a are essentially a relative probability of finding water within the pore, we can translate these hydration profiles into a potential of mean force (PMF) that provides a thermodynamic interpretation of the I106 pore blockage. As shown in Figure 4b, the closed I106 state generates a considerable free‐energy barrier of around 18 kJ mol⁻¹ when compared to the open state. Therefore, it is evident that spontaneous water permeation would be rather unlikely when I106 blocks the pore, as shown in Figures 3b and S6.

For completeness, we also estimated the hydration profiles for the tension independent H105 closed states. As shown in Figure S7, and consistent with a closed state, there is dehydration around the blocked region. However, given the size and polarity of the residue when compared to I106, both the dry region and the energetic barrier (10 kJ mol⁻¹) are smaller for this state.

2.6. The FaPIP2;1‐I106A mutant loses the mechanosensitivity response in MD simulations

Since the presented results point at I106 as being the mechanosensitive gate, it is reasonable to think that a single‐point mutation at this position could impact the tension‐driven closed state of FaPIP2;1. Therefore we developed a FaPIP2;1‐I106A mutant system to perform MD simulations in the same conditions as those tested for the FaPIP2;1 wild‐type system.

Figure 5a shows the mean p _f values—obtained from six replicas—for the single‐file region of the FaPIP2;1‐I106A mutant system. In contrast to FaPIP2;1 WT (Figure 2a), the average osmotic permeabilities obtained with different tension values are not significantly different in the mutant system (Figure 5a), revealing that the replacement of Ile by Ala at position 106 produces the loss of response to tension.

FIGURE 5.

Mutation I106A does not exhibit the same response to tension as wild type. (a) Average (over four pores and six replicas) unitary permeability coefficient of FaPIP2;1 I106A mutant obtained from molecular dynamics simulations under different membrane tension values. The errors represent the SEM of the average p _f of each replica. Simulation protocols were the same as for the wild type, but with six replicas instead of three. The values of pf shown are those of the pore region average, from z = 12 to z = −12 (Å). (b) Axial p _f , binned along the permeation pathway, for all 24 monomers (four per replica, six replicas). Results represent the p _f value observed in the steady state reached with each tension tested. SEM represented as color filled regions.

In addition, the analysis of p _f values along the single‐file region of the water pathway shows very similar behaviors for all conditions tested, including the control without tension, evidencing no tension dependence (Figure 5b). As expected, the p _f reduction observed between −4 and −8 Å in FaPIP2;1 wild type (Figure 2) is not present in the mutant (Figure 5b). Furthermore, p _f is remarkably unaffected further towards the SF, suggesting that the impact of I106 might not be just local, but rather affects p _f along the whole single‐file region.

Moreover, we ran two replicas of 4 μs for the I106A mutant (Figure S8). Due to the high computational cost of these simulations, only one tension value was applied. Immediately after a 2‐μs run under no tension condition, a 2‐μs run was performed with a 12 mN m⁻¹ applied tension. The results show no noticeable effect of tension on the osmotic permeability of the FaPIP2;1‐I106A mutant (Figure S8). A summary of all the key metrics is presented in Table S1.

3. DISCUSSION

The capacity to sense and respond to mechanical forces is a feature of transport membrane proteins, such as ion and water channels (AQPs). However, while mechanosensitivity is characterized at the molecular level in mechanosensitive ion channels, no evidence about the mechanism in AQPs existed until now (Ozu et al. 2023).

This research's findings constitute a groundbreaking contribution to the field of MS‐AQPs in two aspects. This work reports the first MS‐AQP from the PIP2 subfamily, and at the same time, the first evidence of molecular events driven by membrane tension changes in the single‐file region of the water pathway that leads to water permeability reduction.

3.1. Mechanosensitivity of FaPIP2;1 fills a gap in the phylogenetic relationship among MS‐AQPs

Our experimental results demonstrate that FaPIP2;1 behaves like MS‐AQPs reported in previous works, from both our and other groups (Ozu et al. 2018, 2023). Like human AQP1 (Ozu et al. 2013) and BvTIP1;2 (Goldman et al. 2017), FaPIP2;1 shows a nonlinear decrease in water permeability (characterized by the p _f coefficient) with increments in the osmotic gradient (p _f –Δosm curve). This is evidenced by a deviation from linearity in the osmotic flux versus the osmotic gradient curve (J _w –Δosm), when tested in Xenopus oocytes under controlled conditions that allows discarding other effects (Goldman et al. 2017; Ozu et al. ²⁰¹³).

The amount of RNA injected in experiments shown in Figure 1 guarantee that all the conditions tested have the same initial p _f on average. The injection of aquaporins in oocytes results in p _f values that raise linearly with the injected amount of cRNA until 10–12 ng/oocyte. Although differences in expression levels could exist between oocyte batches, our experience with FaPIP2;1 and other aquaporins shows that the injection of 5–10 ng/oocyte—with a 48–72 h incubation time—guarantee a significant expression level and the same p _f value (not significantly different) in different oocyte batches (Alleva et al. 2010; Goldman et al. ²⁰¹⁷; Ozu et al. ²⁰¹³; Yaneff et al. ²⁰¹⁴).

Control experiments with amphotericin revealed that the vitelline envelope is a physical barrier for the incorporation of this polypeptide to the oocyte plasma membrane, but it does not affect the mechanosensitivity of FaPIP2;1 (see Figure S1). The mechanical effect of VE was previously studied in non‐injected oocytes (Kelly et al. 1997), indicating that it has not significant mechanical effect under 10% volume increments. In MS‐AQPs, mechanical gating occurs at volume changes of 3%–6% volume increments (Goldman et al. 2017; Ozu et al. ²⁰¹³). In addition, the elastic properties of the oocyte during the osmotic response are independent of the osmotic gradient magnitude (Goldman et al. 2017).

As was previously reported, the progress of the osmotic response in Xenopus oocytes with (Goldman et al. 2017; Ozu et al. ²⁰¹³) or without (Kelly et al. 1997; Kelly and Macklem ¹⁹⁹¹) AQPs expression produces a continuous increase in membrane tension. Then, the kinetics of the osmotic response—which depends on p _f and Δosm—explain why oocytes facing different Δosm reach different membrane tensions at the same tested time. Briefly, the swelling response driven by an osmotic gradient produces simultaneous volume and pressure increase (Goldman et al. 2017). Consequently, membrane tension increases as well. Therefore, if all oocytes have the same p _f at the beginning of the osmotic shock (t ₀), then experiments with higher gradients will proportionally produce higher J _w at the same time after t ₀ (e.g., at 10 s). This is observed with gramicidin (Ozu et al. 2013) and amphotericin channels in oocytes (results presented in Figure 1 of this manuscript), and with non‐MS‐AQPs like BvPIP2;1 (Goldman et al. 2017). According to the osmotic law, under controlled conditions, the proportionality constant is p _f (Ozu et al. 2013, 2023). However, if the J _w –Δosm relationship deviates from linearity (i.e., the observed J _w values are lower than expected), it indicates that p _f is becoming lower with the development of the osmotic response. This means that aquaporins are closing with membrane tension increase (Goldman et al. 2017; Ozu et al. ²⁰¹³). This can be explained as follows. By controlling aquaporin expression, it can be assumed that p _f at t ₀ is the same for all gradients tested (Goldman et al. 2017; Ozu et al. ²⁰¹³). Then, higher gradients will produce higher J _w , and consequently, higher tension changes per unit of time. And this will progress with time. At short times, gradient dilution effects and area changes effects can be discarded, but p _f changes are significant (Goldman et al. 2017; Ozu et al. ²⁰¹³). Therefore, when testing J _w at a given time (e.g., 10 s), the obtained values will be lower with higher gradients, functionally evidencing the mechanical gating of aquaporins.

Membrane tension values were calculated for the experiments shown in Figure 1 from mechanical characterizations performed previously in Xenopus oocytes (Goldman et al. 2017; Ozu et al. ²⁰¹³). The tension values reached at 10 s from the onset of the osmotic gradient were 11.0 ± 0.2; 11.3 ± 0.2; and 11.6 ± 0.3 mN m⁻¹ for Δosm = 40; 90; and 140 mOsmol Kg_w ⁻¹, respectively. In comparison, FaPIP2;1 responds to higher tensions than AQP1 (about 3 mN m⁻¹; Ozu et al. ²⁰¹³), but similar to BvTIP1;2 (about 10 mN m⁻¹; Goldman et al. ²⁰¹⁷).

Experiments performed with AQPs in other experimental models also showed a nonlinear decrease in p _f with membrane tension increments (Leitão et al. 2014; Soveral et al. ²⁰⁰⁸), membrane compression (Tong et al. 2012), or membrane thickness decrease (Tong et al. 2016).

Besides AQY1 from yeast (Soveral et al. 2008), members of the TIP subfamily (VvTIP2;1 and BvTIP1;2) in plants (Goldman et al. 2017; Leitão et al. ²⁰¹⁴) or the AQP1‐type (AQP1 and AQP4) in mammals (Ozu et al. 2013; Tong et al. ^{2012, 2016}) are mechanosensitive. This is intriguing because PIP and TIP evolutive divergence seems to have occurred before the divergence between PIPs and AQP1‐type aquaporins from animals (Soto et al. 2012). Moreover, the only PIP2‐type aquaporin tested did not show mechanosensitive response (Goldman et al. 2017). The results presented in this work demonstrate that another PIP2 member is indeed mechanosensitive. The different response of FaPIP2;1 and BvPIP2;1 to membrane tension changes may be because the first one is a typical PIP2 (Alleva et al. 2010; Yaneff et al. ^{2014, 2016}) while the second one is atypical, since it does not form heterotetramers with PIP1 members and shows a different response to pH changes compared with other PIP2 members, including FaPIP2;1 (Jozefkowicz et al. 2013).

The relevance of our results raises in the context of evolution because they demonstrate that PIP2‐type aquaporins can be mechanosensitive, and they provide data to the discussion about the phylogenetic evolution of mechanosensitivity in AQPs.

A high percentage of PIPs is highly conserved in plants. In addition, PIP2‐type aquaporins are the main water gateway in the plant plasma membrane. Particularly, FaPIP2;1 has a high p _f and a complex pH heteromerization regulatory mechanism for adjusting membrane water permeability (Yaneff et al. 2014). Now, MS is added to the combo.

3.2. The molecular features of mechanosensitivity in AQPs have begun to be elucidated

Although mechanosensitivity was reported in AQPs from organisms of different Kingdoms, the molecular features of the mechanism were unknown until now. Through MD simulations, we have unveiled a crucial event in the mechanosensitive mechanism involving the conformational change of I106 in Loop B when membrane tension increases. Our results reveal that I106 flips towards the conduction pathway of FaPIP2;1, obstructing the pore and inducing a rearrangement in the single file of water molecules that we believe leads to the reduction of water permeability.

According to experimental reports, the oocyte rupture occurs above 100 mN m⁻¹ (Kelly et al. 1997; Kelly and Macklem ¹⁹⁹¹; Ozu et al. ²⁰¹³). Thus, at most we have applied stresses that are 50% below the ones that produce oocyte rupture and at most the membrane area increases by 11% with respect to the no‐tension conditions (see Figure S9). For our long μs MD, the applied tension of 6 mN m⁻¹ is even below the estimated stress produced on the oocyte by the osmotic pressure (between 10 and 12 mN m⁻¹), and for the control experiments, this is in the linear regime with respect to fluxes (see Figure 1b). In our simulations we have applied stresses that are in the range of 6–50 mN m⁻¹. This is 50% below the stress generated by the osmotic pressure to 50% below oocyte rupture, respectively.

It is interesting to compare the applied stress forces with forces that generate conformational changes in proteins. For example, using both MD techniques (Lu et al. 1998) and atomic force microscopy (Rief et al. 1997) it has been shown that forces required to stretch and unfold a protein, in this case Titin Immunoglobulin, were in the range of 150–300 pN. Another atomic force microscopy study measured forces around 200 pN for unfolding a crystalized hAQP1 (Möller et al. 2003). For our system sizes, the range of applied forces are approximately between 70 and 500 pN along the X and Y direction. Thus, the applied forces are also in line with forces that produce conformational changes in proteins. Nonetheless, it is important to note that the applied stress affects the whole membrane‐protein system and is not directly applied in a single direction, thus the effective force that the protein feels must be lower when compared to these unfolding studies.

In our MD simulations, following a thinning of the membrane caused by the imposed tensions (Figure S9), we found induced effects in the single‐file region in both connections to the extracellular and cytoplasmic vestibules (Figures 2b and S5c). However, we found that only the latter had an impact on permeability (Figure 2a). Unlike the slight tension‐induced changes in the SF (Figure S3b), tension‐dependent rotation of I106 towards the pore's lumen (Figure 3) led to significant dehydration within the single‐file region (Figures 4 and S6). The effect of dehydration due to hydrophobic blocking by I106 could explain the tension‐induced permeability reduction observed in FaPIP2;1. An underlying energetic barrier due to the hydrophobic nature of the I106 side chain further supports this theory, making this segment of the pore inaccessible to the water molecules trying to move through (Figure 4b).

Qualitatively, the distance distributions towards the permeation pathway of both I106 and H105 are similar: the leftmost peak in Figures 3c,d and S4a is located at approximately the same distance to the channel center (~1.7 Å), and both amino acids have roughly similar volumes. However, it is surprising that the effect on the water column and the free energy profile (Figures 4 and S7) differs for the closures mediated by each of these residues. Why does Isoleucine cause greater dehydration and a higher energy barrier if both residues narrow the channel similarly in terms of distance? The answer probably lies in the apolar nature of Ile106, which effectively plugs the permeation pathway. Conversely, the hydrophilic side chain of H105 likely allows water to get closer to the plug, lowering the energetic barrier established by its rotation towards the permeation pathway.

From combining information of permeability values along the channel (Figure 2) and pore radius reduction by I106 (Figures 3 and 4) the idea that only the narrowing of the cytoplasmic region is linked to channel closure naturally arises. This is further supported by the fact that the distance distribution between the two key residues of the SF (H216 and R231) does not significantly show distinct populations between open and closed states, nor do they strongly correlate with tension (Figure S3b). This evidence is in line with our proposal that the cytoplasmic extreme of the single‐file region was involved in mechanosensitivity (see section 1). However, our results do not match the proposed hypothesis. First, the role of I106 was unexpected and residues from the GxxxG sequence seem uninvolved. In our simulations, the pore blocking at the cytoplasmic extreme has two different origins: one associated with the flip of H105 under control conditions (no tension applied), and the other associated with the tension‐induced flip of I106. Both situations exhibit a clear relationship with closure (Table 1). Moreover, the flipping of one residue impairs the flipping of the other (Figure S4). Thus, the proposition is that both H105 and I106 are flipping in an alternating manner. This alternation is constant, so the channel does not lay in a closed state for a long time. The situation changes when tension increases, shifting the balance to the flipped position of I106, which determines a tension‐induced closed state. Therefore, we propose that the function of FaPIP2;1 as a water‐transporting channel is consistent with a three‐state model, with two closed states and one open state at the cytoplasmic side (Figure 6). One of the two closed states is determined by the spontaneous flipped position of H105. The other closed state is due to the tension‐induced flipping of I106, with the mutual exclusion of both states (Figure S4).

FIGURE 6.

Schematic representation of a possible three state model for mechanosensitivity in FaPIP2;1. Our proposal of a model in which there are two possible closed states C_H and C_I, but only the last of which is favored by changes in tension. The energy required for water to permeate through the pore is higher than for the two other conformational states, possibly due to the hydrophobic nature of Isoleucine. His105 and Ile106 are colored in red and yellow, respectively.

Simulations with the FaPIP2;1‐I106A support this proposition. Our proposed model includes two closed states that do not occur simultaneously, one of which is tension dependent. Therefore, the removal of the alleged tension gate should also remove mechanosensitivity. Indeed, our results show that with an Alanine in position 106, tension does not impact water permeation through the FaPIP2;1 pore (Figure 5). Moreover, H105 maintains its non‐MS behavior, showing no noticeable variations in the percentage of correlation with closed states, as tension increases (Table S1). This is compelling evidence in favor of the importance of loop B as a mechanosensitive gate in FaPIPI2;1, and it points, in particular, to the I106 residue as key in the response to membrane tension stresses.

Following the simulations, we designed the FaPIP2;1‐I106A mutant and performed osmotic experiments. Unfortunately, preliminary experiments showed no conclusive results. The p _f values obtained with the FaPIP2;1‐I106A mutant were low in comparison to the wild‐type, just a bit higher than non‐injected oocytes. However, the FaPIP2;1‐I106A mutant showed a linear J _w –Δosm relationship—which is expected for a tension independent response (Goldman et al. 2017; Ozu et al. ²⁰¹³; Ozu et al. ²⁰²³)—with a significantly higher slope from non‐injected oocytes, which means a significant p _f . Since low p _f values could be associated with low expression levels, we also performed confocal fluorescence experiments to test the localization of the EYFP‐FaPIP2;1‐I106A mutant in the oocytes. Unfortunately, preliminary experiments could not confirm the presence of the mutant in the plasma membrane. Although functional experiments showed promising functional results, they are not conclusive yet. In addition, localization experiments would evidence troubles for expression of the mutant in the plasma membrane. Reported evidence indicates that some PIP mutants do not express at the plasma membrane of Xenopus oocytes (van Wilder et al. ²⁰⁰⁸; Yaneff et al. ^{2014, 2015}). It is not known why a single‐point mutation can affect the localization in Xenopus oocytes. More experiments are being designed to deal with this problem and confirm the loss of mechanosensitivity in the FaPIP2;1‐I106A mutant.

A previous study performed on human AQP1 showed that the mechanosensitive response of AQPs is reversible (Ozu et al. 2013). Now, imagine that the membrane can be deformed following swelling and shrinking events, cyclically, and that these changes are under frequency control. Here, according to the model described in the previous paragraphs and depicted in Figure 6, it can be supposed that the time I106 stays in the flipped position would depend on the frequency of membrane deformation. So, if the mechanical features described here for FaPIP2;1 were common to all MS‐AQPs, then emerges the question of whether the transport capacity of AQPs is altered in cells under the effects of frequency‐associated deformations, such as contractile cells.

By other side, plant cells are exposed to high and abrupt osmotic changes. The cell wall protects them from bursting, and, in some cases, cell turgor impacts on the rigidity of the stem (such as in herbaceous plants) and serves as mechanical support. Therefore, mechanosensitivity in AQPs could be a key player in the response to abrupt turgor changes, producing fine‐tuning of membrane water permeability.

This research not only advances our understanding of MS‐AQPs gating but also paves the way for future studies in this field. For the first time, using MD simulations we have established a direct link between functional characterization of MS‐AQPs and some of the molecular components within their structure that likely govern mechanosensitivity.

4. MATERIALS AND METHODS

4.1. Oocyte isolation

Adult female Xenopus laevis frogs (Nasco, Fort Atkinson, WI) were maintained in a room with controlled temperature (18°C) and a 12‐h light–dark cycle. Each frog was kept in an individual tank with filtered water and was fed twice a week. Frogs were anesthetized for surgery by immersion in 0.3% tricaine (MS222) and the oocytes were removed and prepared as previously described (Goldman et al. 2017; Ozu et al. ²⁰¹³). All procedures were performed according to the rules defined by the local Council for the Correct Use and Care of Laboratory Animals, which complies with the EU Directive 2010/63/EU.

4.2. cRNA cloning and isolation

The RNA synthesis was performed as previously reported (Yaneff et al. 2014). The FaPIP2;1 gene (GQ390799) is cloned into expression vectors derived from pT7Ts (Alleva et al. 2010; Mut et al. ²⁰⁰⁸), commonly used for expression in X. laevis oocytes. The clone is flanked by the 5′ and 3′ UTR regions of the X. laevis β‐globin gene, which gives stability to the cRNA injected into the oocytes and ensures correct protein synthesis (Preston et al. 1992). To translate the RNA, the plasmid was linearized using the Xbal restriction enzyme and polymerized using the T7 mMESSAGE mMACHINE Ultra kit (Ambion, Austin, TX). The synthesis products were resuspended in RNase‐free DEPC‐water at a final concentration of 0.1 μg μl⁻¹, supplemented with recombinant RNAsin (Promega). Two 1 μl aliquots were taken to verify the purity of the product and determine its concentration using agarose gel electrophoresis and fluorescence quantification using the Qbit fluorometer and its RNA Quant‐iT RNA Assay Kit (Invitrogen, UK). The rest of the material was aliquoted and stored at −20°C until it was used.

4.3. cRNA microinjection

The cRNA microinjection was performed as previously published (Goldman et al. 2017; Ozu et al. ²⁰¹³). Briefly, oocytes were injected with a commercial microinjector (Drummond Nanoject), with 50 nl of 0.1 μg μl⁻¹ cRNA solution. For control experiments, oocytes were injected with the same volume of H₂O‐DPEC. All injected oocytes were kept for 48 h at 18°C in ND20 1X solution (in mM: 20 NaCl, 2 KCl, 1 MgCl₂, 1.8 CaCl₂, 5 HEPES; pH = 7.4), supplemented with 150 mM mannitol (to equal the osmolality of the oocyte cell, i.e., 200 mOsm Kg_w ⁻¹) and 1 μg ml⁻¹ gentamicin (GIBCO/Life Technologies, Grand Island, NY).

4.4. Incubation with amphotericin B

Incubation of Xenopus oocytes with Amphotericin B was performed according to a previously published protocol (Capurro et al. 1994). Briefly, oocytes were incubated for 20 min with 10 mg ml⁻¹ Amphotericin B (previously dissolved in DMSO) in ND20 1X supplemented with 150 mM mannitol and 1 μg ml⁻¹ gentamicin. In some experimental series the vitelline envelope (VE) was removed before or after the incubation with Amphotericin B.

4.5. Osmotic experiments

To the establishment of osmotic gradients, three solutions were used:

(1) ND20 1X + 110 mM mannitol (160 mOsm Kg_w ⁻¹).

(2) ND20 1X + 60 mM mannitol (110 mOsm Kg_w ⁻¹).

(3) ND20 1X + 10 mM mannitol (60 mOsm Kg_w ⁻¹).

The osmolality was measured with a vapor pressure osmometer (Wescor 5520).

To perform the osmotic experiments, each oocyte was first recorded during 20 s under isosmotic (ND20 1X + 150 mM mannitol) conditions. Then, the solution was changed by one of the hypoosmotic solutions already described. So, gradients of 40, 90, and 140 mOsm Kg_w ⁻¹ were established with solutions 1, 2, and 3, respectively. The oocyte under hypo‐osmotic conditions was recorded for 2–3 min. The experiments were recorded with a webcam (VXS6000 Microsoft, CA) mounted on a trinocular microscope (Olympus SZ40; Olympus Co., Japan). The experiment was recorded in avi video format, using the AMCaP v9.20 software (http://noeld.com/propgrams.asp?cat.vuideo#AMCap).

4.6. Determination of the osmotic permeability coefficient (p _f )

The osmotic permeability coefficient (p _f ) was determined from the osmotic law,

$$J_w=P_f\bullet A\bullet V_w\bullet ∆\mathrm{osm},$$ (1)

where J _w , A, V _w , and Δosm are the osmotic flux, the membrane surface area, the molar volume of water and the osmotic gradient, respectively. To determine J _w , the volume of the oocyte was calculated from video records assuming a spherical shape. Then, J _w was calculated as the slope of linear fitting of the volume‐time curve to the first 10 s of the osmotic response. The surface area was calculated from volume. V _w was assumed to be constant and equal to 18 ml mol⁻¹ in the conditions of the experiments. Finally, Δosm was established as already described. To obtain volume values from video recordings, the images were analyzed with the ImageJ software (developed by Wayne Rasband, Research Services Branch, National Institute of Mental Health, Bethesda, USA; version 1.37, http://rsb.info.nih.gov/ij/). To obtain J _w the GraphPad Prism software (V.5) (https://www.graphpad.com) was used.

4.7. FaPIP2;1 modeling

To date, there is no experimentally resolved protein structure of FaPIP2;1. Therefore, as a template, we employed the crystals of Spinacia oleracea (SoPIP2;1) from the Protein Data Bank, which have a sequence identity of 76% and a coverage of 98%, both metrics defining a zone of confidence for comparative modeling (Krieger et al. 2003). The resulting comparative model, which was generated by the satisfaction of spatial restraints method, excluded the first 23 and the last 5 amino acids, which were not resolved in the crystal model of SoPIP2;1, resulting in shortened terminal ends. The closed SoPIP2;1 conformation (PDB 1Z98; Törnroth‐Horsefield et al. ²⁰⁰⁶) structure, which has a 2.1 Å resolution, was used for all but loop D, which blocks the pore. Loop D was instead modeled using PDB 2B5F (Törnroth‐Horsefield et al. ²⁰⁰⁶) as a template, which has a 3.9 Å resolution. Modeling was performed using the Modeller v9.22 software (Sali and Blundell 1993). We added a disulfide bond between CYS 75 of monomers A and B; and C and D. We also phosphorylated S121 and S280 of each chain, in accordance with the crystal of the constitutively open state of SoPIP2;1 (Törnroth‐Horsefield et al. ²⁰⁰⁶). The X. laevis oocytes' interior is at a pH of 7.6–7.7. At this pH, basic residues are positively charged, and acidic residues are negatively charged. Given the standard pKA value for histidine (6.04) and the aforementioned pH, 99% of all Histidines should be in their neutral state. The simulation system was formed by embedding this FaPIP2;1 homotetramer into a 1‐palmitoyl‐2‐oleoyl‐sn‐glycero‐3‐phosphocholine (POPC) bilayer and solvating it with water to form a rectangular system; its dimensions were 124 × 120 × 120 [ų] with a total number of 177 K atoms. The whole system was built using the VMDv1.9.3 software (Humphrey et al. 1996). For our long simulation, we took the FaPIP2;1 tetramer and built the system anew, this time employing CHARMM‐GUI (Jo et al. 2008; Lee et al. ²⁰¹⁶). For this system we used Hydrogen Mass Repartition to allow for a higher integration timestep (Balusek et al. 2019; Hopkins et al. ²⁰¹⁵). We also used CHARMM‐GUI to generate the I106A system, replacing all four I106—one per pore—with Alanines.

4.8. Molecular dynamics simulations

We studied the response of FaPIP2;1 to mechanical stresses by applying surface tension to the membrane plane. Tension values used for stretching the membrane were 6, 12, 24, and 50 mN m⁻¹ applied in successive increments, each for a duration of 125 ns, following the methods published by another group (Xie et al. 2014). First, we conducted tests on a pure POPC system to estimate the final dimensions and the membrane integrity of a stretched system at 50 mN m⁻¹ (Figure S9).

In aquaporins, on average, the permeation events occur in the range of nanoseconds, around 1 event per 1 to 2 nanoseconds and our test simulations showed that the membrane dimension upon tension converge after 10 ns (see Figure S9). Thus, our simulation times are two orders of magnitude higher with respect to the typical permeation rates of aquaporins and membrane changes. Moreover, there are four pores which are independent, thus increasing the sampling by a factor of four.

For the protein + membrane system, we conducted an energy minimization and then progressively raised the temperature from 0 to 298 K with the protein's alpha carbons fixed, and then simulated 600 ns without tension, the first 100 of which were not considered for analysis. Following this stress‐free simulation, we ran 125 ns at each surface tension, each directly following the previous one. Molecular dynamics (MD) simulations were carried out using the NAMDv3.0 simulation engine (Brooks et al. 2009; Phillips et al. ²⁰²⁰) and the CHARMM36 parameter set (Huang and MacKerell Jr ²⁰¹³) with periodic boundary conditions and explicit solvent (TIP3P water model; Jorgensen et al. ¹⁹⁸³). Pressure (normal to the membrane, i.e., z‐axis) and temperature were kept at around 1 atmosphere and 298 K, respectively, using Langevin dynamics and the Nosé‐Hoover method (Martyna et al. 1994). The tension variations in the membrane plane were isotropic. A 12 Å cut‐off was applied for Lennard‐Jones and real space electrostatics, with a smoothing function applied between 10 and 12 Å. Long‐range electrostatics were estimated using the Particle Mesh Ewald method (Darden et al. 1993). All bonds involving hydrogens were kept to their reference values employing the SHAKE algorithm (Ryckaert et al. 1977). The integration step used was 2 fs for bonded and real space interactions and 4 fs for reciprocal space electrostatics. To ensure statistical significance, we ran three independent replicas following the same incremental procedure. For our long simulations, we performed two simulations with a tension value of 6 mN m⁻¹, and the timestep used for integration was 4 fs. Total simulation times for this longer simulation were 2.2 μs without tension, the first 100 ns of which were not considered, and 2 μs with tension. We also developed a model of the FaPIP2;1‐I106A mutant and performed MD simulations with and without tension increments, as was done with the wild type. Worth noting is that we observed a wider variance in the p _f estimations at no tension. Thus, for this mutant we performed six replicas instead of three for the incremental tension protocol. This wider variance is expected, as we have removed a rather bulky hydrophobic residue that is close to the membrane and replaced it with a smaller one. In this way, H105 has much more freedom to oscillate between open and closed states. Likewise, as we did not observe any effects of the tension protocol, for the long 2 μs, we employed a higher value of 12 mN m⁻¹, which is still within physiological values. Overall, the cumulative simulated time was of around 25 μs.

Tension was applied by modifying the stress tensor for the system. In a regular simulation under control conditions—that is, in the absence of lateral tension straining the membrane—the pressure in the X, Y, and Z axes is the same. However, when applying lateral tension, the X and Y components in the diagonal of the tensor are modified. The left matrix is the non‐stretched system's tenson (assuming that pressure is only controlled in the Z axis), while the right one is the tensor corresponding to the system under tension

$$\left[\begin{array}{ccc}0& 0& 0\\ 0& 0& 0\\ 0& 0& {\sigma }_{\mathit{zz}}\end{array}\right]\to \left[\begin{array}{ccc}{\sigma }_{\mathit{xx}}& 0& 0\\ 0& {\sigma }_{\mathit{yy}}& 0\\ 0& 0& {\sigma }_{\mathit{zz}}\end{array}\right],$$

where ${\sigma }_{\mathit{zz}}$ is the pressure given by the Langevin piston used to control the system pressure. Given a surface tension $\gamma$ and the system's Z dimension $L_z$, the other two components of the stress tensor are

$${\sigma }_{\mathit{xx}}={\sigma }_{\mathit{yy}}={\sigma }_{\mathit{zz}}-\frac{\gamma }{L_z},$$

which becomes the surface tension applied in the X‐Y plane.

4.9. Osmotic permeability

We calculated the osmotic permeability coefficient (p _f ) using the method presented by Ikeguchi's group in 2007 (Hashido et al. 2007), which is based on the work by Zhu et al. (2004). Briefly, for a pore aligned in the z axis, we define a non‐dimensional variable n(t) in its differential form dn,

$$\mathit{dn}={\sum }_{i\in S\left(t\right)}\frac{d{z}_i}{L\left(t\right)},$$ (2)

where S(t) is the number of water molecules inside the pore and L(t) is the length of the pore. dz _i is defined as

$$d{z}_i=\left[z_i\left(t\right)-z_i\left(t-\mathit{\delta t}\right)\right]\mathit{\delta t}=1\mathit{ps},$$ (3)

with z _i being the z coordinate of the ith water within S(t). The integral of n(t) is

$$n\left(t\right)={\int }_0^t\frac{\mathit{dt}}{L\left(t\right)}{\sum }_{i\in S\left(t\right)}d{z}_i.$$ (4)

From Equation (4) we can compute the mean‐squared displacement of n(t) (MSD) and following Einstein relations, the diffusion coefficient for n(t), D _n is

$${⟨n^2\left(t\right)⟩}_t=\frac{1}{M}{\sum }_m{n}^2\left(t\right)\parallel n\left(t=t^{\prime }\right)=0,\mathit{dt}=1\mathit{ps},$$ (5)

$$D_n=\frac{⟨n^2\left(t\right)⟩}{2t},$$ (6)

which is computed in m windows of 1 ns and for multiple time origins. Finally, p _f is obtained by

$$p_f=v_w{D}_n,$$ (7)

where v _w  = V/N _A is the average volume of a water molecule, V (18 cm³ mol⁻¹) and N _A are the standard molar volume of water and Avogadro's number, respectively.

To explore the contributions to the total permeability, we applied an extension of Zhu's p _f method developed by Hashido et al. (2007). The method consists of binning the pore‐defining cylinder and calculating a p _f for each bin. A time series for n(t) and its MSD is shown in Figure S10.

4.10. Distance measurement of ILE106/HIS105 to pore center

We computed the distance between a reference point within the pore and the side chain of ILE106 or HIS 105. We defined the reference of the pore as the geometrical center of the alpha carbon of Valines 95, 110, and 185. For the selectivity filter, we just measured the distance between one of Arginine's side chain nitrogen atoms and the bond between Histidine's gamma carbon and nitrogen. A schematic for these measurements is presented in Figure S11.

4.11. Functional determination of closed and open states

We computed the time series of the accumulated permeation events for each monomer and smoothed it by fitting a spline function to it (Figure S2). Subsequently, we numerically derived this function and considered “open” the states associated with a high derivative and “closed” those with a low or null derivative. In this way, the channel was considered open in the vicinity of isolated permeation events or between a series of permeation events. Due to this calculation method, the amplitude of this vicinity varies depending on how many events occur consecutively in the time series, affecting the derivative. Therefore, the threshold value for the derivative from which the channel was considered open was adjusted so that, for a single event in a time series, conformations 1 ns before and 1 ns after that event exceeded the threshold based on the typical frequency of PEs in AQPs, which is around 1 event/2 ns (Nelson and Cox 2007).

4.12. PMF profiles

The potential of mean force (PMF) was computed from the water density profiles along the pore axis by

$$\mathrm{PM}{\mathrm{F}}_i=-k_{B}T\ln \left({\rho }_i\right),$$ (8)

where k _B is the Boltzmann constant and T is the absolute temperature, in Kelvin. ρi(P/P ₀) is the relative water density. All PMF values were adjusted to set the minimum to zero.

AUTHOR CONTRIBUTIONS

Agustín Caviglia: Methodology; software; writing – original draft; investigation; visualization; formal analysis. Nicolás Espinoza‐Muñoz: Investigation; writing – original draft; visualization; formal analysis; software; methodology. Juan José Alvear‐Arias: Methodology; visualization; writing – original draft; formal analysis; investigation. Luciano Galizia: Investigation; methodology; formal analysis; data curation; writing – review and editing; visualization. Florencia Guastaferri: Investigation; writing – original draft; methodology; formal analysis; writing – review and editing; visualization. Rosario Zimmermann: Writing – review and editing; visualization. Lorena Sigaut: Methodology; visualization; formal analysis. Gabriela Amodeo: Investigation; writing – review and editing; formal analysis; resources; funding acquisition. Carlos González: Writing – review and editing. Marcelo Ozu: Conceptualization; investigation; funding acquisition; writing – original draft; writing – review and editing; formal analysis; project administration; resources; supervision; data curation; methodology; validation; visualization. José Antonio Garate: Conceptualization; investigation; funding acquisition; writing – original draft; writing – review and editing; methodology; software; formal analysis; project administration; supervision; resources; validation; visualization; data curation.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Table S1. Key functional and structural metrics per tetramer under tension, for I106 mutant simulations.

Figure S1. The vitelline membrane does not impact the mechanosensitive behavior of the oocyte membrane.

Figure S2. Method employed for definition of functionally open and closed states.

Figure S3. Distance distributions for both H105 to the center of the pore, and between residues from the selectivity filter.

Figure S4. Distribution of the distance from I106 to the channel center as a function of the homologous distance for H105.

Figure S5. ILE106 and HIS105 distributions for our 2 μs simulations.

Figure S6. Membrane tension increments produce a disruption in the chain the water molecules at the position of Ile106 is disrupted.

Figure S7. The ordering of water molecules in the single file of FaPIP2;1 is altered in H105 mediated closed state.

Figure S8. Long simulations of FaPIP2;1 I106A mutant shows no apparent effect of tension on osmotic permeability.

Figure S9. Characteristic dimensions pertaining to our simulations.

Figure S10. Collective variable n during the simulation, where each row is a replica.

Figure S11. Representation of the distance measurement for all three characteristic distances.

Caviglia A, Espinoza‐Muñoz N, Alvear‐Arias JJ, Galizia L, Guastaferri F, Zimmermann R, et al. Membrane tension‐dependent conformational change of Isoleucine 106 of loop B diminishes water permeability in FaPIP2;1. Protein Science. 2024;33(12):e5204. 10.1002/pro.5204