Protein interactions, ranging from the formation of stable complexes within the cell to transient complexes involved in cell signaling pathways, determine protein function. Membrane proteins are basic elements of the cell, allowing the transport of molecules across the membrane and communication with the external environment. Among them, the AQPs are transmembrane channels – organized in highly conserved tetrameric structures in the cell membranes – which facilitate the passage of water and/or small solutes (Maurel et al., 2008). The function of AQPs is controlled by physiological signals as well as the interactions between different AQP monomers or binding with other proteins.

The role of hetero-molecular AQP interactions has been described in plants as a mode to control physiological mechanisms (Maurel, 2007). An interesting point of view is that AQP isoforms can act on other AQP isoforms, which could be of remarkable importance in tissues where multiple isoforms are expressed. Heteroligomerization between PIP1 and PIP2 has been shown to modify the characteristics of water channels in plant cells or Xenopus laevis oocytes. Thus, co-expression of PIP1 with PIP2 was necessary not only to increase PIP1 water permeability (Fetter et al., 2004; Otto et al., 2010), but also for its displacement from the endoplasmic reticulum to the PM (Zelazny et al., 2007, 2009).

The study of the structural basis for heteroligomers formation is still a challenge. In maize, a conserved cysteine (Cys) residue in the loop A of PIP1 and PIP2 proteins has been identified as responsible for the formation of a covalent disulfide bond between the monomers of these two subfamilies. Although this Cys did not directly regulate the activity or trafficking of the two PIPs, the residue increased the oligomer stability under unfavorable conditions, hence controlling PIP abundance in plant cells (Bienert et al., 2012). However, in Beta vulgaris, BvPIP2;1 was not able to translocate BvPIP1;1 to the PM after the co-expression of both isoforms, whereas BvPIP2;2 was assembled with BvPIP1;1. Differences in the heteroligomerization of the PIPs were attributed to differences in the functionality of loop A (Jozefkowicz et al., 2013). Also, recently, Yaneff et al. (2015) postulated that the abundance of PIPs transcripts regulates the PIP1-PIP2 associations, providing specific combinations of PIP1-PIP2 isoforms when plants are exposed to different environmental conditions and conferring on the PIP1-PIP2 pair a role as a functional unit in the related physiological processes (Figure 1).

Less studied are the interactions between TIP isoforms and there is a lack of information about the function of their association. The putative interactions between TIPs isoforms from Arabidopsis have been studied using heterologous expression in yeast (Murozuka et al., 2013). The authors described not only a strong interaction between the TIP1;2, TIP2;1, and TIP3;1 isoforms but also the assembly of TIPs with PIP2;1. Although PIPs were localized in different membranes when compared with TIPs (Maurel et al., 2008), some TIPs were present in the PM (Wienkoop and Saalbach, 2003; Gattolin et al., 2011). In the same way, it has been reported that the form of heterotetramers assembly may condition physiological processes related to plant development. In onion cells, the HvTIP1;2 and HvTIP3;1 isoforms interacted, forming a heterotetramer which increased water permeability in the cells of late ripening seeds. A high level of water permeability was induced in comparison with the expression of HvTIP1;2 homotetramers alone. Thus, tight control of the water status of the cells by HvTIP3;1 was achieved in the seed development and desiccation stages (Utsugi et al., 2015).

In cotton, fiber elongation was found to depend on PIP2s and specific PIP2 isoform combinations markedly enhanced the water permeability in Xenopus oocytes (Li et al., 2013). Also, in the halophyte, Thellungiella halophila, ThPIP1 interacted with ThPIP2 and a non-specific lipid-transfer protein 2, suggesting a role of ThPIP1 in the response to the external signals that trigger multiple physiological processes when plants are exposed to salinity (Qiang et al., 2015).

Therefore, although the formation and dissociation of the heterotetramers of MIPs represent a mechanism for dynamic changes of the membrane permeability, further studies are necessary to discern these functions of the heterotetramers in plants.

Hetero-complexes with AQPs, regulating different biological processes, have been described in several organisms. In humans, 70% of the AQPs interactions are comprised by the C-terminus (Sjöhamn and Hedfalk, 2014), which is the most divergent sequence in the AQP family and frequently appears as the main site of regulation. However, in plants, these interactions are emerging as new mechanisms that determine AQPs functions.

The interaction of AQPs with other proteins or nucleic acids involved in osmotic adjustment processes has been described. Xin et al. (2014) showed that transgenic Arabidopsis plants expressing SlTIP2;2 possessed enhanced tolerance of salinity. In addition, yeast two-hybrid and BiFC analysis in protoplasts determined the physical binding between SlTIP2;2 from tomato (Solanum lycopersicum) and the UDP-galactose transporter (Xin et al., 2014).

The functionality of the AQP TsMIP6, from the halophyte T. salsuginea, was characterized in transgenic rice (Oryza sativa). The expression of this gene was enhanced under salinity (Sun et al., 2015). In addition, TsMIP6 specifically interacted with a member of the glycoside hydrolase family in yeast cells. The authors suggested that the relationship between TsMIP6 and the hydrolase could mediate the osmotic balance in the plant cells and that TsMIP6 could also be involved in the transport of solutes that participate in the maintenance of the osmotic potential.

The pore-gating phosphorylation sites in AQPs were affected in inserted T-DNA mutants for the SIRK1 protein, an active kinase with increased activity in the presence of external sucrose (Wu et al., 2013). Thus, a direct interaction and regulation of the AQPs via phosphorylation by SIRK1 were implicated in this sucrose-specific osmotic response. In maize plants, a physical interaction between the PM syntaxin SYP121, a SNARE protein, and the AQP isoform Zm-PIP2;5 was described, with evidence that the SNAREs coordinated membrane trafficking and the activity of AQPs and plant K⁺ channels (Besserer et al., 2012). This mechanism of coordination may suppose an adjusted regulation of the movement of water and ions in growing cells, guard cells, or cells under drought stress.

Also, in Arabidopsis, autophagic degradation of PIP2;7 in the vacuole was regulated by a protein–protein interaction under drought stress. The Arabidopsis tryptophan-rich sensory protein/translocator (TSPO), a multi-stress regulator, interacted with the AQP isoform at the level of the endoplasmic reticulum and Golgi membranes, resulting in a decreased abundance of PIP2;7 protein in the PM that prevented cell water loss (Hachez et al., 2014b). The correct delivery of this PIP to the PM involved specific interactions with two syntaxin proteins, SYP61 and SYP121 (Hachez et al., 2014a). An additional anchoring role of PIP2;7 has been suggested by Kriechbaumer et al. (2015); the presence of the AQP in the plasmodesmata proteome and the interaction with the plant reticulum family proteins tubulate (RTNLB) may confirm the connection between the desmotubule and PM. In agreement with this idea, the heterologous expression of the OsPIP1;3 AQP of rice in yeast showed a direct interaction with the hairpin protein Hap1 from Xanthomonas oryzae pv. oryzae (Xoo), the pathogen that causes bacterial blight of rice. This may favor the virulence of plant pathogenic bacteria, with the AQPs acting as anchor elements (Ji and Dong, 2015). These interaction studies will form the basis for future research related to the AQP–protein interaction network, with implications in cellular functions and pathways.

In living organisms, membranes are in the liquid-crystalline phase, in which there is considerable freedom of motion for the phospholipids. The membranes are characterized by high rotation of the C-C bonds of the fatty acyl chains and lateral diffusion of the phospholipids within the phase of the membrane. The amplitude of motion, described by an order-parameter, which describes the time-averaged disposition in space of each group of atoms in the acyl chains (Los and Murata, 2004), is inversely proportional to the fluidity (Laroche et al., 2001).

Membrane fluidity has been associated with many other processes in membranes, like protein activity. The effect of fluidity change on the function of membrane proteins such as ATPase was reported by Cooke and Burden (1990). They used xenobiotics to help elucidate the influence of fluidity on protein activity but encountered difficulties in their search for a direct relationship. Yeagle (1985) observed that high pressure decreased fluidity and stimulated ATPase activity, whereas addition of cholesterol also decreased fluidity but inhibited ATPase activity. Kitagawa et al. (1995) claimed to have demonstrated a correlation between fluidity and membrane-bound enzyme activity. Subsequently, fluidity changes have been reported to affect the activity of membrane proteins, like P4 ATPases, influencing the ability of the plant to adapt to cold stress (Gomès et al., 2000). In the same way, some investigations have shown that water transport through AQPs can be affected by membrane physical properties such as fluidity (Carvajal et al., 1996; Frick et al., 2013). These reports showed that high fluidity is related to high AQP functionality. Frick et al. (2013) concluded that increased membrane fluidity could affect the conformational state of SoPIP2;1, pushing the equilibrium toward an open conformation. Furthermore, Ho et al. (1994) reported that water in the membrane bilayer was also responsible for modifying fluidity, since increases in membrane water content increased fluidity. Thus, diffusion of water through the bilayer affects its fluidity and the rate of that diffusion depends on the fluidity, as determined by changes in lipid composition (sterols and/or fatty acid saturation). Also, recent findings indicate that membrane hydration increases the space in the acyl chains (Disalvo, 2015). Therefore, due to an increased presence of water molecules at the lipid/protein interface, a consequence of increasing unsaturation, an increase in AQP permeability has been observed (Lee, 2004). In this context, it should always be remembered that the PM is a heterogeneous entity comprising different domains, each with its own fluidity and viscosity (Marrink et al., 2009). Therefore, although it is now possible to obtain a reasonable quantification of the overall motion parameters in terms of membrane structure and mobility, and their influence on AQP functionality, they should be interpreted very carefully.

Viscosity, as a property of a fluid, is the internal friction a consequence of the cohesive forces between molecules (Lee et al., 1989). The term microviscosity was defined by Shinitzky et al. (1971) as the harmonic mean of the effective viscosities opposing the in- and out-of-plane rotations of the fluorescent probe. An effective definition of these terms that would allow translation of one to the other is, in fact, elusive, especially in those situations where microviscosity values are determined using steady-state fluorescence anisotropy values (Engel and Prendergast, 1981). Thus, in lipid bilayers, it is possible to define a relationship between fluidity and viscosity, both being dynamic parameters. Although studies of microviscosity in relation to AQPs functionality are very scarce, microviscosity has been reported to be involved in the alteration of the conformation of transmembrane channels, changing the activity of the proteins (Foulkes, 1998).

This property, that is dependent on the physical state of the membrane lipid bilayer and on the water in the membrane, has been related to AQPs functionality – its maximum influence being in the response of plants to stress conditions (Carvajal et al., 1998).

The fact that the microviscosity of lipid membranes is higher than that of water bestows on AQPs the critical parameter related to the nearly frictionless transport of water molecules through them. This phase has been identified recently as the possible phase transition temperature, after selecting specific activation energies and activation volumes for confined (cylindrical) geometry (Kwang-Hua, 2015). With this approach and model-based calculations, the study produced results that can be applied to the diverse fields of research related to transmembrane water transport via AQPs, thereby shedding further light on the microviscosity/fluidity-AQPs relationship

In a study of the microviscosity of the hydrocarbon zone in the isolated tonoplast and PM, using a fluorescent diphenylhexatriene probe, it was demonstrated that both membranes do not differ in this parameter in the phase state of their lipid bilayer (Trofimova et al., 2001). The study pointed out that the observed difference in water permeability does not depend on the state of the lipid phase and probably reflects the dissimilar functional activity of PM and tonoplast AQPs. However, the study did not take into account that the tonoplast and PM possess different amount of AQPs – that result in differing osmotic water permeabilities (Maurel et al., 2008), independently of the physical properties of the membrane.

It has always been assumed that the lipid composition of the membrane is the major factor determining its physical characteristics. The function of sterols in plants was thought to be related primarily to their ability to affect membrane structure and water permeability (Graziani and Livne, 1972). However, they can also affect the packing of the membrane bilayer, interacting with fatty acid side-chains of phospholipids and integral membrane proteins, increasing fluidity (Cooper, 2000) and bilayer water permeability (Da Silveira et al., 2003). In relation to this, beside the fact that a high sterol content increases fluidity and thereby AQPs functionality, other studies indicate that the highest local concentrations of PIP-AQPs correspond to tightly packed, sterol-enriched domains (Belugin et al., 2011). These domains are correlated with higher water permeability that probably corresponds to both AQPs and the lipid bilayer.

Also, it has been reported that an increase in the proportion of AQPs in the plant DRM (detergent-resistant PM fraction), which is rich in sterols and is considerably different from the total PM fraction, increased the osmotic water permeability of the PM at low and freezing temperatures and, hence, increased cell survival (Minami et al., 2009). Also, discoveries regarding the regulation of AQP intracellular trafficking and sub-cellular localization in response to environmental stresses, like water shortage and salt stress, revealed that sterol-rich domains are crucial in the cell surface dynamics and endocytosis of PM AQPs (for reviews see Hachez et al., 2013 and Luu and Maurel, 2013). Therefore, the interactions between AQPs and sterol-enriched domains seem to be closely related to AQP membrane functionality (Belugin et al., 2011). All these studies have contributed to the deciphering of AQP sub-cellular trafficking, which has been reported to be related to plant growth and development in Arabidopsis as a response to environmental changes (Boursiac et al., 2005). In fact, the content of sterols seems to be related to the higher or lower resistance to salinity that involves AQPs functionality (López-Pérez et al., 2009; Chalbi et al., 2015). High salinity usually produces an increase in total PM sterols (Silva et al., 2007; López-Pérez et al., 2009). However, the results obtained for halophytes revealed that in Cakile maritima the total PM sterols decreased, reducing membrane rigidity (Chalbi et al., 2015). This was related to low membrane stability, but not to water permeability. However, rather than total sterols, the stigmasterol/sitosterol ratio should be used as the parameter that indicates a swift increase in water permeability. Sitosterol has been shown to be very efficient regarding the regulation of water permeability in plants grown under salt stress (Silva et al., 2007; Basyuni et al., 2012), and it is related closely to AQP functionality (López-Pérez et al., 2009).

In the same way, the results in animals provide evidence that the sterol content in the membrane could shift the balance toward the transport of other molecules, rather than water, by AQPs. Therefore, it has been concluded that sterols, which impart mechanical stability to the membrane, reduce gas permeability through the lipid bilayer but can markedly raise the gas permeability through AQPs. In this sense, Itel et al. (2012) showed that cholesterol can decrease membrane CO₂ permeability by two orders of magnitude in phospholipid vesicles and intact cells. However, the permeability of CO₂, and possibly other gasses, through AQP1 appeared to be increased by three orders of magnitude. Also, studies with kidney cells revealed that cholesterol depletion reduced the diffusion coefficient of AQP3 but not that of AQP5 (Koffman et al., 2015), suggesting that only a subset of AQPs may be associated with lipid rafts and regulated by cholesterol.

The phospholipid composition, in terms of the head-group and acyl chains, also seems to be related to membrane physical properties which influence the activity of proteins. In studies of the PM composition of wheat plants grown under drought stress, it was observed that the phosphatidylcholine/phospha- tidylethanolamine ratio and the level of unsaturation of the fatty acyl chains increased. It was suggested that this produced a more fluid lipid matrix in order to preserve the physiological function of the lipid bilayer (Vigh et al., 1986). Similar findings, plus an increase in free sterol abundance, were observed in water-stressed sunflower PM (Navarri-Izzo et al., 1993). However, after the discovery of AQPs, it was found that the adaptation to stress was dependent on the AQPs rather than on lipid composition, although the phospholipid composition could modulate AQP functionality through effects on the physical properties of membranes, such as fluidity (Carvajal et al., 1996).

Molecular dynamic simulations and crystallographic refinement to determine the localization of DMPC (dimyristoylphosphatidylcholine) lipids around lens-specific animal water channel aquaporin-0 (AQP0) showed that the positions of the constrained lipids in the 2D crystals are defined by the acyl chains rather than the head groups (Aponte-Santamaría et al., 2012). Furthermore, the positions of these lipids are influenced greatly by the local mobility of the protein, whereas specific hydrogen bonds play a secondary role. Therefore, AQPs follow the general mechanism in which membrane proteins diffuse laterally, associated with the two layers of lipids, with the positions of the lipids in the first solvation shell being modulated also by irregularities in the protein interface (Lindahl and Sansom, 2008).