Binding of the protein ICln to α-integrin contributes to the activation of IClswell current

IClswell is the chloride current induced by cell swelling, and plays a fundamental role in several biological processes, including the regulatory volume decrease (RVD). ICln is a highly conserved, ubiquitously expressed and multifunctional protein involved in the activation of IClswell. In platelets, ICln binds to the intracellular domain of the integrin αIIb chain, however, whether the ICln/integrin interaction plays a role in RVD is not known. Here we show that a direct molecular interaction between ICln and the integrin α-chain is not restricted to platelets and involves highly conserved amino acid motifs. Integrin α recruits ICln to the plasma membrane, thereby facilitating the activation of IClswell during hypotonicity. Perturbation of the ICln/integrin interaction prevents the transposition of ICln towards the cell surface and, in parallel, impedes the activation of IClswell. We suggest that the ICln/integrin interaction interface may represent a new molecular target enabling specific IClswell suppression in pathological conditions when this current is deregulated or plays a detrimental role.

www.nature.com/scientificreports www.nature.com/scientificreports/ ICln 23 is a highly conserved, ubiquitously expressed and multifunctional protein that plays a critical role in cell homeostasis and survival 24 . ICln was suggested to be the channel or part of the macromolecular complex underlying the activation of ICl swell 25,26 . Despite being mainly located in the cytosol in resting cells, ICln is transposed towards the plasma membrane upon hypotonic challenge 27,28 or other stimuli 29 . In addition, ICln establishes complex interactions with diverse proteins belonging to different subcellular compartments and functional modules 21,28,[30][31][32] . For instance, ICln regulates the process of splicing and small nuclear ribonucleoproteins (snRNP) biogenesis by associating with several Smith antigen (Sm) proteins of the spliceosomal core 33 and controlling their methylation [34][35][36][37] .
The cytoplasmic tail of human integrin αIIb contains the highly conserved motif (1020)KVGFFKR(1026) (single letter amino acid code; the numbers indicate the amino acid positions within the primary sequence of human αIIb integrin; NCBI Reference Sequence: NP_000410.2), which regulates integrin activation and acts as a recognition site for various intracellular proteins, including ICln 22 . As the KxGFFKR motif (x denotes any amino acid) is common to several α integrin isoforms 38 , it is important to understand whether the ICln/integrin interaction can also occur in nucleated cells and identify the corresponding function. Here we show that the interaction between ICln and the intracellular domain of integrin α-chain is essential for the activation of ICl swell , and is independent from the occupancy of integrin α by extracellular ligands.

Results icln establishes a direct molecular interaction with the intracellular domain of the integrin αIIb chain.
To verify whether the ICln-integrin interaction is restricted to the process of platelet activation, we performed fluorescence resonance energy transfer (FRET) experiments in living NIH-3T3 fibroblasts kept in isotonic solution and transiently transfected with Canis familiaris ICln and the transmembrane and intracellular portion (amino acids 987-1039) of the human integrin αIIb chain (αIIb-short, αIIb s ) (Fig. 1A,B). Deletion of the extracellular portion of αIIb avoided perturbation of a possible ICln/αIIb interaction by an extracellular signal. The cDNA sequences coding for ICln and αIIb s were cloned into vectors to produce fusion proteins with fluorophores suitable for FRET experiments, i.e. CFP-ICln and αIIb s -YFP respectively (see Methods). After expression, αIIb s -YFP clearly targeted to the plasma membrane, with a partial retention in intracellular compartments ( Fig. 1A and Supplementary Fig. S1). The FRET efficiency of CFP-ICln and αIIb s -YFP determined in the plasma membrane region of transfected cells following photobleaching of the FRET acceptor YFP was significantly higher (p < 0.01) with respect to that of the corresponding negative control (expression of CFP alone and αIIb s -YFP) (Fig. 1B, Supplementary Table S1 and Fig. S1). Similar experiments performed on NIH-3T3 fibroblasts fixed in 3% paraformaldehyde and kept in a phosphate buffered saline led to similar results (data not shown). Also in HEK293 Phoenix cells, the FRET efficiency of CFP-ICln and αIIb s -YFP determined in the plasma membrane region was significantly higher than that determined in control cells transfected with CFP and αIIb s -YFP (Supplementary Table S1). The ratio of the FRET efficiency between CFP-ICln and αIIb s -YFP and the stochastic interaction given by the FRET efficiency between CFP and αIIb s -YFP in the two cells types was similar, being 2.9 and 2.7 in NIH-3T3 and HEK293 Phoenix cells, respectively. Accordingly, in sensitized emission N-FRET experiments, ICln and αIIb s produced an N-FRET in isotonic solution that was significantly higher (p < 0.01) compared to the control (Fig. 1C,D, isotonic, and Supplementary File S1). Following exposure of cells to a hypotonic bath solution, the N-FRET of ICln and αIIb s significantly increased over time, while FRET of CFP and αIIb s -YFP did not significantly change (Fig. 1C,D and Supplementary File S1). These results show that in isotonic, as well as hypotonic conditions, ICln is located in close proximity to αIIb s , and therefore strongly suggest a direct molecular interaction between the two molecules. over-expression of αiib s activates icl swell . To test whether the interaction between ICln and the integrin α chain might have a functional implication on ICl swell , the whole-cell patch-clamp technique was used. HEK293 Phoenix cells were transfected with the pIRES2-EGFP vector with or without αIIb s cDNA (see Methods). The bicistronic αIIb s -pIRES2-EGFP vector was used to obtain the simultaneous expression of αIIb s and the transfection marker EGFP as two independent proteins in the same cell, thus allowing for the identification of cells expressing αIIb s by fluorescence microscopy. After the seal was realized and the whole-cell configuration was obtained, ICl swell was elicited by reducing the extracellular osmolarity. Hypotonic shock induced the activation of a large chloride current with the biophysical fingerprints of ICl swell , characterized by outward rectification and slow voltage-and time-dependent inactivation at positive potentials 6 in both control and αIIb s over-expressing cells (Fig. 1E). The current density-to-voltage relation determined after a 10 min exposure to extracellular hypotonic solution (Fig. 1F), as well as the current density-to-time relation (Fig. 1G), showed a strong up-regulation of ICl swell in cells over-expressing αIIb s compared to control cells (p < 0.0001). These results suggest an involvement of the αIIb chain in the activation of ICl swell .
The cell-permeable (1020)KVGFFKR(1026) integrin peptide impairs the translocation of ICln to the plasma membrane and the activation of icl swell . The translocation of ICln to the plasma membrane after hypotonic challenge was verified by sensitized emission N-FRET in living NIH-3T3 fibroblasts exposed to the synthetic palmitoylated peptide (1020)KVGFFKR(1026), corresponding to the amino acid sequence of the ICln recognition domain within the integrin αIIb chain 22 . Control cells were treated with the synthetic palmitoylated peptide KAAAAAR. Palmitoylation allows the peptides applied to the extracellular solution to access the intracellular environment 22,39 , with preferential, although not exclusive, targeting to the plasma membrane. Cells expressed the fusion protein CFP-ICln and the membrane label YFP-mem (i.e. the YFP protein tagged with an N-terminal palmitoylation sequence, see Methods). As previously reported 27,28 , hypotonicity induced a significant increase of FRET between CFP-ICln and YFP-mem, thereby indicating a transposition of ICln towards the cell membrane ( Fig. 2A). Interestingly, pre-incubation of cells with the palmitoylated (1020)KVGFFKR(1026) www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ peptide not only prevented the expected FRET increase following hypotonic stress, but indeed led to a FRET decrease. This indicates that the peptide not only abolished the translocation of ICln normally observed in hypotonicity, but additionally decreased the amount of the membrane-bound ICln compared to isotonic conditions (Fig. 2B). In contrast, preincubation of cells with the palmitoylated KAAAAAR control peptide did not affect the hypotonicity-induced translocation of ICln towards the plasma membrane (Fig. 2C). Importantly, patch-clamp experiments in whole-cell configuration showed that the impaired translocation of ICln toward the cell surface was paralleled by an impaired activation of ICl swell . Specifically, the palmitoylated (1020)KVGFFKR(1026) peptide applied to the bath solution, and not the palmitoylated KAAAAAR peptide, dramatically impaired (Fig. 2D) or prevented (Fig. 2E) the activation of ICl swell in hypotonicity-stimulated HEK293 Phoenix cells. Larkin et al. previously speculated that the amino acids (84)AKFEEE(89) of ICln might be involved in binding the (1020) KVGFFKR(1026) motif 22 . However, the cell-permeable peptide (84)AKFEEE(89) did not significantly affect ICl swell activation compared to the unrelated cell-permeable peptide ELFNDG (Fig. 2F). Therefore, we concluded that the ICln amino acids (84)AKFEEE(89) unlikely represent the main integrin interacting site.

The (61)ISLHA(65) ICln motif is the prime integrin binding site.
In order to identify the binding site of the integrin (1020)KVGFFKR(1026) motif in ICln, we applied 1 H-15 N HSQC spectroscopy. Binding of the (1018)MWKVGFFKRNR(1028) integrin peptide to ICln was probed by using both full-length Canis familiaris ICln (cICln, Fig. 3A) and a truncated form (TcICln, Fig. 3A) lacking the C-terminal part. TcICln folds into a pleckstrin homology domain-like structure consisting of two nearly orthogonal antiparallel β-sheets forming a β-barrel capped by an α-helix near the C terminus ( 40 and Fig. 3). Figure 3B shows an overlay of NMR spectra of unligated (green) and ligated (red) TcICln. It can be seen that most of the residues are unchanged and only a subset of residues (labelled by their position in the primary sequence) display small but significant chemical shift changes. The most pronounced chemical shift changes were observed for S62 (shown in red in Fig. 3A,C), I61 and A65 (shown in orange in Fig. 3A,C). Most interestingly, a continuous stretch of residues displayed significant chemical shift changes (I61-A65). Additionally, the following residues (shown in yellow in Fig. 3A) exhibited chemical shift changes: S55, Y58, L63, Y79, V80, A84, S90 and residues in the C-terminal α-helix (E121, M127, C128, C130, A132 and L133).
Similar results were obtained for the binding of the (1020)KVGFFKR(1026) motif to full-length cICln, encompassing also the conformationally flexible C-terminus (Ct) 41 . In this experiment, the peptide (1018) MWKVGFFKRNRPPLEEDDEEGE(1039), which corresponds to the entire Ct of integrin αIIb and therefore includes the (1020)KVGFFKR(1026) motif, was used. Comparison of spectra of full-length unligated and ligated cICln shows that only a small number of peaks are shifted, while the majority of peaks remained more or less unchanged, again indicating a specific interaction. Overall, the magnitude of chemical shift changes is small, presumably due to weak binding and small structural adaptions upon ligand binding. It should also be kept in mind that 1 H-15 N HSQC spectra report on the protein backbone and not on side-chains, which are expected to be closer to a potential ligand. Residues which were most significantly altered upon binding were: A65, M81, N83 and A108. The most pronounced chemical shift change was observed for A65 (shown in red in Fig. 3A). It should be noted that the other residues displaying chemical shift changes (M81, N83 and A108, shown in orange in Fig. 3A) are located in the vicinity of A65 and part of a hydrogen bond network bridging the β-strands of the β-sandwich. Small chemical shift changes were observed also for residues located in the C-terminal α-helix (E121, M123, M127, C128 and C130, shown in yellow in Fig. 3A). Slight chemical shift changes found for residues in the flexible (and structurally loosely defined) loop region connecting β-strands β6 and β7 (E92 and E96) and in the flexible C-terminal tail (Y168 and T182) might be due to transient contacts of these regions with the well-folded core (PH) domain of ICln and are therefore not shown in Fig. 3A.
We thus conclude that the β4-β5 loop and the β5 β-strand comprising the ICln amino acids (61)ISLHA(65) represent the main interaction site for the (1020)KVGFFKR(1026) motif, while chemical shift changes observed for the other residues are in part due to small local structural alterations in adjacent β-strands.  (65) motif not only abrogated the FRET increase between CFP-ICln and YFP-mem normally observed after hypotonic stress, but led to a FRET decrease compared to the isotonic condition (Fig. 4A,B), again indicating an interference with ICln transposition and localization to the plasma membrane. Accordingly, patch-clamp experiments in whole-cell configuration showed that the Caption in B also refers to D. (E) Original recordings obtained by whole-cell patch-clamp in hypertonic (top) or hypotonic (bottom) conditions in control (left panels) or αIIb s -transfected (right panels) HEK293 Phoenix cells stimulated with voltage increments as depicted in the lower inset. Cells over-expressed both αIIb s and the transfection marker EGFP as separate proteins, or EGFP only. (F) The current density-to-voltage relationships measured 10 min following hypotonic shock, and G, the current density-to-time relationships were fitted with second order polynomials, following application of the extra-sum of squares F test (αIIb s , hypotonic, versus control, hypotonic: p < 0.0001).
www.nature.com/scientificreports www.nature.com/scientificreports/ palmitoylated (57)EYPTISLHALSRDR(70) peptide applied to the bath (extracellular) solution (Fig. 4C,E), and not a palmitoylated peptide with a scrambled amino acid sequence (RIYSASDLEPRLTH, pal-control, Fig. 4D,E), strongly inhibited ICl swell in HEK293 Phoenix cells. The effect of the peptide (57)EYPTISLHALSRDR(70) on ICl swell was also investigated by adding it to the pipette filling (intracellular) solution used in patch-clamp experiments. This manoeuvre allows the peptide to gain access to the intracellular space after establishing the whole-cell configuration. The non-palmitoylated (57)EYPTISLHALSRDR(70) peptide did not interfere with the activation of the current (Fig. 5A). In contrast, adding the palmitoylated peptide to the intracellular space significantly reduced ICl swell (Fig. 5B).
Integrity of the (61)ISLHA(65) motif is essential for the association of ICln to the αiib-chain and the activation of icl swell . To further confirm the identity of the ICln amino acid(s) involved in the interaction with the integrin α chain, amino acid substitutions (I61G, S62G, A65N) were made within the integrin-recognition domain of Canis familiaris ICln (cICln, Fig. 6A) and FRET was determined by sensitized emission in NIH-3T3 fibroblasts transiently transfected with wild type (wt) or mutated (mut) CFP-ICln and αIIb s -YFP. The FRET between CFP-ICln mut and αIIb s -YFP measured in the plasma membrane region in isotonic and hypotonic solutions was significantly lower with respect to that measured between CFP-ICln wt and αIIb s -YFP (Fig. 6B). These results confirm that the ICln amino acids I61, S62 and A65 are critically involved in the interaction with αIIb s . To investigate the functional consequences of the impaired ICln/integrin interaction, ICl swell was measured by patch-clamp experiments in whole-cell configuration in HEK293 Phoenix cells In (E) cells were pre-incubated for 10 min in a hypertonic solution containing 100 µM of the palmitoylated peptides, and then stimulated with hypotonic solution. In (D,F) the current density (I) was normalized for the current value measured immediately before the addition of the peptides to the bath (I 0 ). Data were fitted by linear regression (D,F) or with second order polynomials (E), following application of the extra-sum of squares F test (in D and E, KVGFFKR versus KAAAAAR: p < 0.0001; in F, AKFEEE versus ELFNDG: not significant).
www.nature.com/scientificreports www.nature.com/scientificreports/ transfected with ICln wt, ICln mut or a control vector. Similar to previous results 23 , overexpression of ICln significantly up-regulated ICl swell . In contrast, in cells expressing ICln mut, ICl swell did not differ from the endogenous current measured in control cells (Fig. 6D,E). ICln mut resulted functionally impaired compared to the wild type www.nature.com/scientificreports www.nature.com/scientificreports/ also following expression in NIH-3T3 fibroblasts ( Supplementary Fig. S2). These results, together with those shown in Figs 2, 4 and 5, strongly indicate that interfering with the ICln/integrin interaction impairs the activation of ICl swell .

Discussion
Although integrins are in general recognized as osmosensors in RVD 7,42 , the molecular mechanisms by which they influence the swelling-activated chloride current ICl swell are incompletely understood. This lack of information was due in part to the molecular identity of VRACs remaining elusive for more than 30 years 43,44 . In 2014, two independent laboratories identified the leucine-rich repeat containing protein 8 A (LRRC8A) as an essential component of the channel that mediates ICl swell 45,46 . Recently, the structure of the human and Mus musculus homo-hexameric LRRC8A channel was described by cryo-electron microscopy [47][48][49][50] . Overall, these studies provide the basis for understanding the precise mechanism of activation of VRACs and identifying their molecular partners. LRRC8A and other members of the LRRC8 family form heteromers, of which the composition controls the VRAC single-channel conductance, rectification, anion selectivity and inactivation kinetics 46,51 . When reconstituted into lipid bilayers in a two-compartments system, purified LRRC8 complexes formed anion channels activated either by osmolality gradients between two compartments both containing hyperosmolar solutions, or low ionic strength in the absence of osmotic gradients between the two compartments 51 . These findings suggest that LRRC8 complexes may include a sensor for osmolarity and low ionic strength. However, other authors recently showed that low ionic strength is neither sufficient to activate VRAC on intracellular compartments, nor indispensable to keep plasma membrane-localized VRACs active, and rather suggested that the diacylglycerol/protein kinase D pathway specifically and potently activates VRAC 52 . Therefore, in a cellular context, other factors, www.nature.com/scientificreports www.nature.com/scientificreports/ including binding partners, post-translational modifications 53 , second messengers or even other channels 54 , may contribute to VRAC activation besides a mere reduction in the intracellular ionic strength.
The composition of the extracellular matrix modulates the characteristics of ICl swell through integrin receptors, and knock-down of β1 integrin impaired ICl swell activation in skeletal muscle-derived C2C12 cells 55 . The integrin β chains seem to play a fundamental role in this context. The research group of Browe and Baumgarten showed that mechanical stretch of β1 integrin leads to the activation of a chloride current with the biophysical and pharmacological fingerprints of ICl swell 56-60 , thus suggesting that the role of integrins in RVD would be an outside-in transduction of a mechanical stimulus with ICl swell as the final effector. Accordingly, knockdown of integrin β1 also reduced RVD of adherent Ehrlich ascites cells 42 , and knockdown of integrin β3 almost completely blocked ICl swell activation in basilar artery smooth muscle cells 61 . As integrin β3 co-precipitated with Src and the ClC-3 protein, these authors suggested that the integrin β3/Src/ClC-3 signalling pathway might be involved in the activation of ICl swell 61 .
In order to elucidate the molecular events underlying ICl swell activation, we set out to explore the role of the interaction between integrins and ICln. ICln interacts with the (1020)KVGFFKR(1026) motif of the platelet-specific integrin αIIb, thereby influencing platelet activation 22,[62][63][64] . We found that ICln can interact with the α-chain also in cells unrelated to platelets, in isotonic as well as hypotonic conditions (Fig. 1A-D). Moreover, overexpression of the transmembrane and intracellular portion of the αIIb chain (αIIb s ) upregulated ICl swell (Fig. 1E-G). Omission of the extracellular portion of the integrin αIIb chain was adopted to avoid modulation of ICl swell by a possible alteration of cell adhesion to the extracellular substrate, thereby allowing for an unbiased exploration of the intracellular signalling events towards ICl swell activation. Importantly, this approach showed www.nature.com/scientificreports www.nature.com/scientificreports/ that the ICln/integrin interaction occurs regardless of the activation state or occupancy of the integrin α-chain by extracellular ligands. These findings are in agreement with those of Larkin et al., which showed that ICln bound equally to purified αIIb and to integrin from resting or activated platelets, and that binding was not affected by direct integrin activation with Mn ++ or by inhibitors of integrin occupancy 63 .
The primary sequence of the site of the integrin αIIb-chain responsible for binding to ICln was previously determined to be represented by the highly conserved (1020)KVGFFKR(1026) amino acid motif 22 , and was confirmed in the present study (Fig. 2). However, the identity of the ICln site involved in binding to the www.nature.com/scientificreports www.nature.com/scientificreports/ integrin α-chain remained to be established, and was determined here by NMR spectroscopy to be the (61) ISLHA(65) motif (Fig. 3). FRET experiments in NIH-3T3 cells show that cell-permeable synthetic peptides mimicking the ICln binding site to integrin or the integrin binding site to ICln not only impeded the expected hypotonicity-induced FRET increase between ICln and the plasma membrane probe YFP-mem, but also reduced FRET below the basal levels measured in isotonic conditions (Figs 2B and 4A,B). These data indicate that the cell-permeable synthetic peptides prevented the hypotonicity-induced translocation of ICln towards the cell membrane and also reduced the membrane-resident pool of ICln. To explain our findings, we propose a model ( Supplementary Fig. S3) where the cytosolic, integrin-bound and membrane bound forms of ICln exist in a dynamic equilibrium. In isotonic conditions ( Supplementary Fig. S3A), the cytosolic form of ICln largely prevails and a significant fraction of ICln is bound to integrin α, as denoted by FRET experiments (Fig. 1B,D). Hypotonicity (Supplementary Fig. S3B) induces the transposition of ICln towards the cell membrane ( Fig. 2A), thereby altering the dynamic equilibrium in favor of the membrane-bound ICln, without a substantial increase of the integrin-bound ICln pool (Fig. 1D). The palmitoylated (1020)KVGFFKR(1026) integrin peptide may compete with the integrin α chain endogenously expressed in these cells for binding to ICln ( Supplementary  Fig. S3C). Such a competition would have interfered with and weakened or prevented the ICln/α chain interaction. The palmitoylated (1020)KVGFFKR(1026) peptide may have realized a non-functional interaction with ICln (not sufficiently stable, or inappropriately oriented), which failed in leading to an association of ICln to the plasma membrane. This may have produced an alteration of the dynamic equilibrium between the membrane-bound and unbound ICln in favour of the unbound form, thus leading to the observed decrease in N-FRET between ICln and the membrane probe following exposure to the hypotonic medium (Fig. 2B). These findings indicate that binding to integrin is an essential prerequisite for stabilization of ICln at the plasma membrane in hypotonicity. It is also conceivable that the palmitoylated (1020)KVGFFKR(1026) peptide may have recruited the membrane-resident ICln in a complex that was not available for interaction with the membrane probe YFP-mem. A similar reduction of FRET between ICln and the membrane probe in hypotonicity could be observed with the palmytoilated ICln peptide (Fig. 4A,B). Importantly, the lack of translocation of ICln provoked by the integrin as well as the ICln peptide was accompanied by an impaired activation of ICl swell (Figs 2D,E and 4C). To have an inhibitory effect, the ICln peptide needed to be targeted to the cell membrane (Fig. 5A,B), further suggesting that its molecular target is located in close proximity of the cell periphery. We may envision that only a palmitoylated ICln peptide can reach the correct localization, sufficient local concentration or take the proper orientation to bind α integrin. Therefore, an intracellular, non-palmitoylated water-soluble ICln peptide would be inactive, while a palmitoylated membrane-targeting peptide would bind to and displace ICln from α integrin ( Supplementary Fig. S3D), thus preventing the association of ICln with the plasma membrane and hampering the activation of ICl swell .
Mutations within the (61)ISLHA(65) ICln motif (Fig. 6A) abrogated the ICln interaction with the integrin αIIb chain (Fig. 6B) and rendered ICln inactive, i.e. unable to increase the ICl swell intensity above the endogenous current (Fig. 6C-E). Overall, these experiments indicate that the interaction between ICln and the integrin α-chain promotes the activation of ICl swell . It is important to highlight that the direct physical interaction between ICln and integrin αIIb per se does not require α integrin interaction with extracellular ligands, as it was characterized in a cell system overexpressing a truncated form of the α-chain void of the extracellular domain. We therefore suggest that the α-chain provides a scaffolding system for ICln, facilitating the translocation and/or stabilizing the ICln protein in close proximity of the cell membrane, an essential prerequisite for ICl swell activation. However, we do not exclude modulation of the ICln/integrin interaction by an outside-in signalling pathway in the physiological setting. We propose a model of integrins orchestrating the regulation of cellular volume by acting as osmosensors as well as osmoeffectors within an integrin β -ICln -integrin α -ICl swell pathway ( Supplementary Fig. S4). The osmotic signal, sensed by the β-chain, would trigger an intracellular signalling cascade ending on ICln, thus leading to its recruitment at the sub-membrane region and binding to the α-chain. This would promote the association of ICln with the plasma membrane, its accumulation in this compartment and activation of ICl swell as the final event. It is also conceivable that ICln may dissociate from integrin prior association with the plasma membrane; this hypothesis is consistent with the mild increase of FRET between ICln and integrin α following hypotonic stress (Fig. 1D), while FRET with the membrane probe is more pronounced (Fig. 2A). The mechanism by which ICln would activate ICl swell once recruited at the plasma membrane deserves further investigation. It is conceivable that ICln may take part in the macromolecular complex of the ICl swell channel.
The concept of an outside-in signalling mediated by integrin and triggered by hypotonicity is well supported by several findings. In rat liver, the active conformation of integrin β1 rapidly appeared after hypo-osmotic perfusion, indicating integrin activation in response to hepatocyte swelling. Hypo-osmotic hepatocyte swelling triggered the activation of Src-type kinases, mitogen-activated protein (MAP)-kinases and downstream metabolic events, which were fully abolished by integrin-inhibitory proteins exhibiting an Alanine-Glycine-Aspartic Acid (RGD) motif 7 . In rabbit ventricular myocytes, the activation of ICl swell induced by mechanical stretch of β1 integrin was associated to an activation of focal adhesion kinases (FAK), Src, the angiotensin II receptor, and the EGF receptor [57][58][59][60] . In basilar artery smooth muscle cells, hypotonic solution increased Src phosphorylation in a time-dependent manner, a phenomenon that was significantly enhanced by integrin β3 overexpression and reduced by integrin β3 knockdown, thus suggesting that integrin β3 mediated the Src phosphorylation induced by hypotonic solution 61 .
Mapping of the residues displaying frequency changes upon binding of the (1020)KVGFFKR(1026) motif on the 3D structure of ICln reveals a distinct binding site on one edge of the β-sandwich (Fig. 3C). Interestingly, this binding mode is analogous to ubiquitin (Ub) binding of the pleckstrin-homology GLUE domain of the human ESCRT-II EAP45 (also called VPS36) 65 . Figure 7 shows a structural superposition of the EAP45 GLUE-Ub complex and ICln using the pleckstrin-homology domain as templates. The structural similarity is clearly visible and www.nature.com/scientificreports www.nature.com/scientificreports/ convincingly supports a similar protein-binding mode. Interestingly, EAP45 GLUE binding to Ub was found to be coupled to phosphoinositol binding, thus suggesting a structural dynamic link between recognition of ubiquinated cargoes and endosomal phospholipids binding in the context of multivesicular bodies sorting. The similar protein-binding mode might reflect a functional resemblance as ICln displays a dynamic and regulated incorporation in membranes depending on changing osmotic conditions. We could detect neither endogenous ICln nor recombinant αIIb s in immunoprecipitates of recombinant αIIb s or endogenous ICln from HEK293 Phoenix cells. This finding is opposed from what was observed in platelets, in which the co-immunoprecipitation was indeed detected 22 , and is indicative of a dynamic interaction, or that only a small fraction of ICln is available for interaction with integrin at a given time. This may represent a characteristic of non-platelet cells and supports the hypothesis that the physiological function of this interaction differs from that found in platelets.
Integrins and the integrin interactome are emerging as novel targets in the control of inflammatory diseases and cancer 66 . Also, as ICl swell is involved in numerous pathophysiological processes 67 , its pharmacological inhibition is regarded as a potential strategy in the treatment of several life-threatening conditions. For example, in the heart, blockers of Cl − currents such as ICl swell may have antiarrhythmic properties 68,69 . Furthermore, pharmacological inhibition of ICl swell antagonizes apoptosis by preventing apoptotic volume decrease and downstream apoptotic events 70 . Hence, ICl swell inhibitors could find application in the treatment of certain forms of heart failure 56 . Although not unanimously 71 , ICl swell blockers have been shown to impair the motility of cells 72 , delay their proliferation rate and induce cell cycle arrest 70 , thus suggesting that ICl swell suppression might be beneficial in the treatment of some forms cancer. In brain, ICl swell contributes to the release of the excitatory neurotransmitters glutamate and aspartate, and is thought to promote the excitotoxic death of neuronal cells in stroke and traumatic brain injury 3 . In addition, involvement of ICl swell in the pathogenesis of exercise-induced asthma was suggested 73 . ICl swell is inhibited by a wide variety of classical non-specific Cl − channel blockers such as 4,4′-Diisothiocyano-2,2′-stilbenedisulfonic acid (DIDS), 5-nitro-2-(3-phenylpropyl-amino) benzoic acid (NPPB) and tamoxifen. Interestingly, tamoxifen is a selective estrogen receptor modulator that is widely used in the treatment and prevention of breast cancer (https://www.cancer.gov/about-cancer/treatment/drugs/tamoxifencitrate). ICl swell blockade may in part contribute to its anti-proliferative actions. 4-[(2-Butyl-6,7-dichloro-2-cyclopentyl-2,3-dihydro-1-oxo-1H-inden-5-yl)oxy]butanoic acid (DCPIB), which is considered the most selective ICl swell antagonist, attenuated microglia activation and neuronal injury in a model of ischemia/reperfusion 74 . Accordingly, astrocyte-specific LRRC8A knockout mice are protected from brain damage after ischemic stroke 75 . Overall, these lines of evidence point to a potential beneficial effect of ICl swell suppression in several pathological conditions.
In summary, the protein ICln directly binds the highly conserved (1020)KVGFFKR(1026) amino acids of the intracellular domain of the integrin αIIb via the (61)ISLHA(65) amino acid motif lying in the β4 -β5 loop and β5 sheet of its pleckstrin-homology-like N-terminal portion. The stability of this interaction is required for the proper activation of the current ICl swell during RVD. The findings presented here show that perturbation of the binding of ICln to the integrin α-chain leads to ICl swell attenuation, and suggest that the ICln/α integrin interaction interface could be exploited as a new pharmacological target to obtain specific ICl swell suppression. www.nature.com/scientificreports www.nature.com/scientificreports/
For FRET experiments, cells were seeded on 6-well plates and grown overnight to 80% confluence. Cells were then transfected with 8-12 μg of polyethylenimine (Sigma-Aldrich) and 2-3 μg of DNA (NIH-3T3 fibroblasts) or with 3 μg of plasmid DNA by the calcium phosphate co-precipitation method (HEK293 Phoenix cells). After 8 hours, the medium was replaced. Twenty-four hours post-transfection, cells were seeded on glass coverslips. FRET experiments were done 48 hours after transfection.
For patch-clamp experiments, cells were seeded into 30 mm diameter Petri dishes, grown overnight to 50% confluence and transfected with 3 μg of plasmid DNA by the calcium phosphate co-precipitation method (HEK293 Phoenix cells) or with 8-12 μg of polyethylenimine and 2-3 μg of DNA (NIH-3T3 fibroblasts). Eight hours after transfection, cells were transferred on glass coverslips (10 mm diameter); electrophysiology measurements were performed 24-32 hours after transfection. The transfection efficacy was about 50%. For patch-clamp experiments on untransfected cells, HEK293 Phoenix cells were seeded directly on glass coverslips and measurements were performed 16-24 hours after seeding.

Cloning procedures, plasmid constructs and site-directed mutagenesis. Starting from human or
Canis familiaris (from MDCK cell line) cDNA, the ORF of full length ICln was amplified by PCR using standard protocols. The cDNA sequence corresponding to amino acids 987-1039 of the human integrin αIIb chain (αIIb short -αIIb s ) was amplified by PCR from the pcDNA3.1 vector (kindly provided by Prof. N. Moran) containing the full-length cDNA of αIIb. The PCR products encoding ICln and αIIb s were cloned in frame into the mammalian expression vectors pECFP-C1 and pEYFP-N1 (Clontech, Mountain View, CA, USA) respectively, to produce the fusion proteins CFP-ICln (CFP was fused to the Nt of ICln) and αIIb s -YFP (YFP was fused to the Ct of ICln) suitable for FRET experiments. To label the plasma membrane, the pEYFP-Mem vector (Clontech), encoding for YFP targeted to the plasma membrane by a palmitoylation sequence (YFP-mem) was used. CFP and YFP were used as the FRET donor and acceptor, respectively.
For electrophysiology experiments, ICln or αIIb s ORFs were cloned into the XhoI and BamHI restriction sites of the bicistronic mammalian expression vector pIRES2-EGFP (Clontech). The use of vectors bearing an internal ribosome entry site (IRES) allows for the simultaneous expression of two individual proteins (αIIb s or ICln and the enhanced green fluorescent protein (EGFP)) from a single bicistronic mRNA without the production of fusion proteins 77 . Therefore, EGFP expression occurs only if preceded by αIIb s or ICln expression. Single transfected cells could be optically individuated by detecting the fluorescence emitted by EGFP (excitation maximum = 488 nm; emission maximum = 507 nm). Control experiments were conducted in cells transfected with the pIRES2-EGFP vector with no insert and therefore overexpressing only the EGFP protein.
For protein purification, the ORFs of Canis familiaris ICln and its truncation mutant (ICln159X, referred to as TcICln) 40 were cloned in frame with a hexahistidine tag into the bacterial expression vector pET3-His 78 , allowing for the expression of N-terminal His 6 -tagged proteins in Escherichia coli strains.
Mutagenesis of the Canis familiaris ICln sequence was carried out using the Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA, USA) with the following primers: 5′GGAATACCCCACCGGTGGCTT GCATAACGTGTCCAGGG3′, forward, and 5′CCCTGGACACGTTATGCAAGCCACCGGTGGGGTATTCC3′, reverse. All plasmid inserts were sequenced prior to use in experiments (Microsynth AG, Switzerland).

FRET experiments.
Acceptor photobleaching experiments were performed in living or fixed (for 10 min in 3% paraformaldehyde at room temperature) NIH-3T3 or HEK293 Phoenix cells as previously described 79 . Forty-eight hours post-transfection, cells were transferred to a FCS-2 chamber (Bioptechs, Inc. Butler, PA, USA) and imaged with a TCS SP2 AOBS confocal microscope (Leica Microsystems, Wetzlar, Germany), using an argon laser and a 63X (1.4 NA) oil immersion objective. During acquisition, cells were bathed in isotonic solution (in mM: NaCl 90, KCl 5, CaCl 2 2, MgCl 2 2, Glucose 5, HEPES 10, Mannitol 80, pH 7.4) or in phosphate buffered solution. CFP was excited at 458 nm and the fluorescence acquired between 465 and 505 nm; YFP was excited at 514 nm and the fluorescence acquired between 525 and 600 nm. YFP photobleaching in whole cells was performed with 4 sequential illuminations (4 frames, 1024 × 1024 resolution, 200 Hz) at 514 nm (setting the AOTF at 100% and the zoom at 8×). FRET efficiency was calculated from the ratio of the CFP fluorescence evaluated before (CFP pre ) and after (CFP post ) YFP photobleaching, using the formula: www.nature.com/scientificreports www.nature.com/scientificreports/ acceptor images were then acquired after 10, 15 and 20 minutes. The sensitized emission (F sen ) was calculated as already described 79,80 and according to the formula:  80 . The N-FRET indices were calculated in each cell from the mean intensities derived from 3 plasma membrane ROIs according to the formula: . FRET signals were determined in areas of interest of the cell in the membrane region before, 5 and 15 minutes after hypotonic stimulation. n indicates the number of cells analyzed; for each cell, ten ROIs were taken into account [28][29][30] .
To obtain the N-FRET image shown in Fig. 1C, the following image processing steps were carried out. ROIs were designed corresponding with cells expressing only CFP or YFP. From these ROIs, correction factors were measured and calculated. With these factors, sensitized emission images were obtained applying the formula (2) in a pixel-by-pixel manner using the math or the Image calculator tools of ImageJ software. The sensitized emission images were ratioed to the F D image to obtain the N-FRET pictures. To suppress excessive noise in dim parts of the images, the N-FRET images were smoothed with a spatial filter and a mask was created, by setting a threshold equal to the background from this image. Subsequently, this mask was applied to the original, unfiltered N-FRET image 80 . The N-FRET images obtained were then optionally smoothed.
Patch-clamp recordings. Patch-clamp experiments were performed as previously described 29,81 . Shortly, single cells expressing EGFP were selected by fluorescence microscopy and voltage-clamped using the whole-cell patch-clamp technique. The resistance of the glass pipettes was 3-8 MΩ when filled with the pipette solution (in mM: CsCl 125, MgCl 2 5, EGTA 11, raffinose 50, ATP 2, HEPES 10, pH 7.2 (adjusted with CsOH)). The hypertonic bath solution was composed of (in mM): NaCl 125, CaCl 2 2.5, MgCl 2 2.5, HEPES 10, mannitol 100, pH 7.4 (adjusted with NaOH). Fast exchange of the hypertonic bath solution with a hypotonic bath solution (in mM: NaCl 125, CaCl 2 2.5, MgCl 2 2.5, HEPES 10, pH 7.4) was obtained using a perfusion system with a flow rate of 5 ml/min and a bath volume of ~300 μl. For data acquisition, an EPC-10 amplifier (HEKA Elektronik, Lambrecht, Germany) controlled by a Macintosh computer running Patch Master (HEKA Elektronik) software was used. Access resistance as well as fast and slow capacitance were monitored and compensated throughout the recordings. All current measurements were filtered at 5 kHz and digitized at 50 kHz. To monitor the activation of the swelling activated chloride current (ICl swell ), cells were held at 0 mV and step pulses of 400 ms duration were applied from 0 mV to 40 mV every 20 s. To establish the current to voltage (IV) relationship, step pulses of 500 ms duration were applied every 5 or 10 min from −120 mV to 100 mV in 20 mV increments from a holding potential of 0 mV. For data analysis, Fit Master (HEKA Elektronik) and EXCEL (Microsoft, USA) software were used. The current values were normalized to the membrane capacitance to obtain the current density. Each data set was obtained from cells from at least 3 independent subcultures with control experiments on the same subculture.
Protein expression and purification. cICln and TcICln were overexpressed and labelled with 15 N as N-terminal His 6 -tagged fusion proteins in E. coli BL21(DE3) strain (Agilent Technologies, Santa Clara, CA, USA). For protein expression and uniform labeling with 15 N, transformed E. coli BL21 (DE3) (Agilent Technologies) were first grown in LB-medium (Sigma-Aldrich) and then diluted 1:100 into 1 liter of minimal medium (48 mM Na 2 HPO4, 22 mM KH 2 PO4, 8.56 mM NaCl and 18.7 mM 15 NHCl (Euriso-top, Germany), supplemented with 20 ml of a 20% (w/v) glucose solution (Sigma-Aldrich), 2 ml of 1 M MgSO 4 , 0.3 ml of 1 M CaCl 2 and 100 μg/ml ampicillin) and further cultured at 37 °C in an orbital shaker. Protein expression was induced at an OD 600 of ~0.8 by addition of isopropyl-1-thio-β-D-galactopyranoside to a final concentration of 0.5 mM. Bacterial cells were harvested 4 h after induction. The bacterial pellet obtained from a 1 liter 15 N labelled M9 minimal medium culture was resuspended in 30 ml of lysis buffer (50 mM K 2 HPO4, pH 8.0) and lysed using a French-press. To purify over-expressed proteins, the bacterial lysate was cleared by centrifugation and loaded onto an IMAC column (HisTrap, GE Healthcare, Chicago, IL, USA) pre-equilibrated with lysis buffer. Following extensive washing with lysis buffer, bound protein was eluted with a buffer containing 50 mM K 2 HPO4, pH 8.0 by using a NaCl gradient. Fractions containing cICln or TcICln were pooled, concentrated in Amicon Ultra 10 kDa ultrafiltration units (Millipore, Burlington, MA, USA) and loaded onto a Hiload 16/60 Superdex 75 column (GE Healthcare) equilibrated in 50 mM K 2 HPO 4 , 100 mM NaCl, pH 7.2. Fractions containing cICln or TcICln were collected, pooled and concentrated. For the following NMR experiments the buffer was changed to 50 mM K 2 HPO 4 , pH 6.6. www.nature.com/scientificreports www.nature.com/scientificreports/ NMR spectroscopy. NMR experiments were performed at 25 °C using Inova 800 MHz spectrometers (Varian Inc., Palo Alto, CA, USA). 1 H N -15 N HSQC spectra were recorded with carrier frequencies for 1 H N and 15 N at 4.73 ppm and 120.0 ppm, respectively. The sample contained 0.8 mM recombinant canine ICln, 50 mM K 2 HPO 4 , pH 6.6 in 90% H 2 O and 10% D 2 O. All spectra were processed using NMRPipe/NMRDraw 82 and analyzed with Sparky 83 .
Salts, chemicals and drugs. All salts and chemicals used were of "pro analysis" grade. Palmitoylated and non-palmitoylated synthetic peptides were synthesized and purified by HPLC by the Organic Synthesis Core (Royal College of Surgeons, Dublin 2, Ireland) or purchased from GenScript (Piscataway, NJ, USA).
Statistical analysis and sequence alignment. All data are expressed as arithmetic means ± S.E.M. For statistical analyses, GraphPad Prism software (version 4.03 for Windows, GraphPad Software, San Diego, CA, USA) was used. Significant differences between means were tested by paired or unpaired, two-tailed, Student's t-test as appropriate. The current density-to-time and current density-to-voltage relationships were fitted with linear regression or second order polynomials (Y = A + BX + CX 2 ). For detecting significant differences between those data, the extra-sum of squares F test was applied. Statistically significant differences were assumed at p < 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001). n indicates the number of cells tested, except where otherwise specified. Sequence alignment was done with Gene Runner software, version 6.5.48.

Data Availability
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.