Simultaneous Optical Recording in Multiple Cells by Digital Holographic Microscopy of Chloride Current Associated to Activation of the Ligand-Gated Chloride Channel GABAA Receptor

Chloride channels represent a group of targets for major clinical indications. However, molecular screening for chloride channel modulators has proven to be difficult and time-consuming as approaches essentially rely on the use of fluorescent dyes or invasive patch-clamp techniques which do not lend themselves to the screening of large sets of compounds. To address this problem, we have developed a non-invasive optical method, based on digital holographic microcopy (DHM), allowing monitoring of ion channel activity without using any electrode or fluorescent dye. To illustrate this approach, GABAA mediated chloride currents have been monitored with DHM. Practically, we show that DHM can non-invasively provide the quantitative determination of transmembrane chloride fluxes mediated by the activation of chloride channels associated with GABAA receptors. Indeed through an original algorithm, chloride currents elicited by application of appropriate agonists of the GABAA receptor can be derived from the quantitative phase signal recorded with DHM. Finally, chloride currents can be determined and pharmacologically characterized non-invasively simultaneously on a large cellular sampling by DHM.


Introduction
Regulated and selective transport of ions mediated by ion channels underlies various fundamental cellular processes. Ionic channels represent an important group of therapeutic targets which are modulated by a range of currently prescribed drugs. In particular, chloride channels are involved in several physiological functions such as cell volume regulation, transmembrane fluid transport, muscular activity and neuroexcitability (for review see [1]). Their dysfunction is observed in over a dozen human pathological conditions, including several that affect the nervous system such as epilepsy and certain psychiatric diseases (for review, see [2], [3]). The development of drugs targeted to these chloride conductances represents an important field for developing novel pharmaceutical agents (for review, see [4]). In general, electrophysiology (patch clamp) remains the most accurate technique for analyzing and quantifying the effectiveness of a drug on an ionic conductance. Thus, this approach has been widely used for chloride currents mediated by neuronal GABA A receptors associated with chloride conductance during an inhibitory synaptic transmission. Nevertheless, despite an exceptional fidelity and precision, patch clamp are not suitable for multiple compound screening since this approach is technically demanding, with very low throughput capacity and labor-intensive. Recently, the development of automated electrophysiology has substantially improved the throughput. However, the capability of currently available automated system is not yet compatible with primary screening of large random compounds sets.
Another possibility is the use of fluorescent dyes whose fluorescence intensity is related to the intracellular concentration of ions, an approach widely used for visualizing calcium (for review, [5]). However, this technique has not achieved a general applicability because of several methodological drawbacks, notably for chloride, including sensitivity and specificity, limitations inherent to the loading protocol as well as to photobleaching. In addition, the quantitative determination of chloride fluxes with non-electrophysiological methods amenable to screening approaches has been challenging, in particular because the transmembrane ratio of chloride is low (10:1) and the equilibrium potential of chloride is generally close to the resting membrane potential of cells, two factors that result in transmembrane fluxes of limited amplitude, raising therefore issues of sensitivity.
Recently, significant progress has been made in Quantitative Phase Microscopy (QPM) techniques [6], [7], [8], [9] that enable to obtain full-field quantitative phase imaging of transparent living cells, allowing to visualize cell structure and dynamics. In contrast to the non-invasive phase contrast (PhC), initially proposed by F. Zernike and by Nomarski's differential interference contrast (DIC), which provides qualitative information about cell structure, QPM, provides a quantitative measurement of the phase shift induced by a transparent specimen on the transmitted wavefront. The phase shift, or the optical path difference (OPD) containing considerable information about the cell morphology as well as intracellular content related to the refractive index properties, can be regarded as a powerful endogenous contrast agent.
The QPM that we have developed, called digital holographic microcopy (DHM), has the ability to explore cell dynamics by providing, from a single recorded hologram, quantitative phase images of living cells with a nanometric axial sensitivity [10]. Practically, an original numerical processing of holograms allows not only to calculate the phase shift but also to reconstruct the whole wavefront diffracted by the specimen [11] and consequently to compensate for aberration [12] and experimental noise (time drift, vibration, defocusing, etc.) thus ensuring a high phase stability making possible to explore biological processes across a wide range of time scales, from milliseconds to hours. Thus, DHM appeared to us as a promising method to monitor ion channel activity because of the capacity of DHM to monitor subtle changes of the intracellular refractive index and cell volume resulting from transmembrane water fluxes associated with the activity of specific ion conductances. We have recently demontsrated this application of DHM to the pharmacological study of glutamatergic ionotropic receptors [13].
In this study, we set out to quantitatively measure from the DHM phase signal the dynamics of chloride currents in response to the modulation of chloride conductances in a well-established biological model used for drug-discovery by High Throughput Screening (HTS). To this aim we have used HEK293 cells, a cell line widely used for pharmacological studies, transfected to express the neuronal ligand-gated channel GABA A . The ionic nature of the measured currents was validated by pharmacological analysis as well as by comparing the reversal potential determined by DHM with that measured electrophysiologically with conventional I/V relationships achieved by patch-clamp. Furthermore, the replacement of chloride (Cl 2 ) by thiocyanate (SCN 2 ) in the extracellular medium amplified the GABA-induced optical signal, permitting the pharmacological characterization the GABA A receptor activity simultaneously in large numbers of unpatched cells. Finally, an original mathematical analysis of the phase response determined by DHM, quantitatively predicts the chloride trans-membrane current, thus providing the possibility to quantitatively measure currents without electrode.

Results
For all experiments, the transfected HEK cells (HEK GABA ) cultures had a cell density such that HEK GABA were in contact with neighbouring cells (at least 60% of confluency). Their morphologies were identical to those of the non-transfected HEKs (HEK norm ) namely polygonal (figure 1A). In terms of electrical properties, HEK GABA had a resting potential of 230.860.8 mV (range: 220 to 245 mV; n = 63) and an input resistance of 233613 MV (Range: 150 to 500 MV). There were no significant differences in terms of electrical properties between HEK GABA and HEK norm (V m : 234.661.7 mV; p.0.05; R inp : 214636; p.0.05; n = 10).
GABA triggers a phase shift in the optical signal on HEK GABA At 2100 mV, bath perfusion of GABA (3 mM, 30 s) on HEK GABA led to a transient increase of phase signal (DQ = 5.4661.38u; n = 22), while a similar application of GABA had no effect on HEK norm (DQ = 0.3360.55u; n = 6; p,0.005) (figure 1A). In the presence of a GABA A receptor antagonist, picrotoxin (30 mM), the phase response evoked by GABA application was significantly reduced (282617%, p,0.05, n = 10) (figure 1B), while the specific agonist, muscimol (1 mM; 30 s, n = 6) mimicked the effect of GABA (figure 2A2). These 1 st set of results indicate that the optical signal obtained after application of GABA was associated with the activation of GABA A receptors expressed in HEK GABA . Finally, the amplitude of the phase shift depended both on the concentration of GABA (EC 50 = 3.4 mm) and the length of application ( Determination of the reversal potential of Cl 2 from the amplitude of the phase shift evoked by GABA application Through the patch pipette, we have clamped the membrane at various holding potentials. For a given cell, at 2100 mV, the application of GABA resulted in a transient increase of phase signal (DQ = 4.8562.99u; n = 7) as described above (figure 2A1). In contrast, at +40 mV, GABA triggered a transient strong decrease of phase signal (DQ = 212.4263.39; n = 7) (figure 2A1). In reporting the maximal amplitude of the phase shift as a function of the membrane potential, we obtained a relationship which we called ''Phase/Voltage'' (Q/V), in analogy with the ''current/ voltage'' relationship (I/V) (figure 2A1). In doing so, we were able to detect an outward rectification and also to determine the reversal potential of the ion involved, here Cl 2 (E Cl ), in this case 226 mV (figure 2A1; table 1). This value was close not only to the theoretical value of E Cl (E Cl(Th) = 233 mV) calculated with the Nernst equation using values taken from our experimental conditions (table 1), but also to that obtained by classical electrophysiology with the I/V curve (226 mV; figure 2A1; table 1). At this stage, it is important to not that the large changes in voltage used in this study (from 2100 mV to +40 mV), in absence of GABA, were not able to trigger a detectable optical signal by themselves (not shown), reinforcing the fact that the optical signal is only triggered by application of GABA.
To confirm that changes in phase signal were associated with the flow of Cl 2 , we modified the concentration of this anion in the patch pipette ([Cl 2 ] intrapip .) from 44 mM to 139 mM. Accordingly, the values of the reversal potential for Cl 2 obtained with the Q/ V and I/V relationships were shifted to less negative values (respectively 27 mV and 22 mV; n = 6) and were similar to the value of E Cl(Th) calculated under these new conditions (24 mV) (figure 2B1; table 2). These results clearly show that the electrochemical properties of a given ionic conductance (here the conductance for Cl 2 ) can be determined by DHM with the same precision as that achieved with classical electrophysiological approaches.
Ionic co-transporters NKCC and KCC are not significantly involved in the genesis of GABA A -induced optical signal It has been established that an alteration in chloride homeostasis linked to a transient but strong activation of GABA A receptors results in an activation of ionic co-transporters KCC and NKCC [14], two membrane proteins co-transporting ions (especially Cl 2 ) and water molecules (for review, see [15], [16]), in order to restore a normal chloride homeostasis on both sides of the plasma membrane. To determine the possible involvement of these ionic co-transporters in the genesis of the GABA A -induced optical signal, furosemide, a broad spectrum blocker of KCC and NKCC has been used. In presence of this blocker (100 mM), the optical signal induced by application of GABA (3 mM, 30 s) was not significantly modified regardless the membrane potential (2100 mV and +40 mV) (figure 1E), suggesting that these two proteins do not contribute to the optical response mediated by the opening GABA-activated chloride conductance. Simultaneous measurement of multiple optical recordings of Cl 2 flux in unpatched HEK GABA cells The next step consisted in detecting optical signals from unpatched cells in order to explore the possibility of performing simultaneous multiple recordings. As shown in figure 3A, the optical responses triggered by GABA (3 mM, 30 s) from unpatched HEK GABA were heterogeneous resulting either in an increase (25% of recorded cells; n = 47) or a decrease (19%; n = 38) or no detectable changes (56%; n = 105). This result is consistent with the fact that the resting potential of HEK cells (mean  To overcome this problem, we altered the Cl 2 reversal potential by replacing in equimolar amounts most of the extracellular Cl 2 with SCN 2 (modified ACSF, see Material and Methods and File SI). In such modified ACSF, application of GABA (3 mM, 30 s), results in a strong positive phase shift simultaneously in almost all HEK GABA cells (up to 85-90% of recorded cells; figure 3B1) corresponding mainly to a Cl 2 efflux (see File S1 and figure S1). The value of the peak amplitude was significantly higher in modified ACSF (4.44+/20.59u; n = 174; p,0.005) than in normal ACSF (0.74+/20.29u; n = 188).
An interesting point is that the EC50 for GABA applied on unpatched HEK GABA perfused with modified ACSF is around 1.4 mM (figure 3C), a value slightly inferior to that determined for patched HEK GABA cells clamped at 2100 mV in normal ACSF (3.4 mM; figure 1) but in excellent agreement with the literature [17]. A control application (vehicle solution) had no effect on phase shift (20.11+/20.39u; n = 80; figure 3D) demonstrating that the optical effect observed was linked to GABA. Moreover, the pharmacological characteristics of optical responses linked to    figure 3D). Finally, in the presence of diazepam (10 mM), a benzodiazepine known to potentiate the action of GABA on GABA A receptors, the optical response associated with GABA application was significantly increased (6.75+/20.82u; n = 112; p,0.005; figure 3D) confirming that the effect of GABA is associated with the activation of GABA A receptors. Taken together these data clearly demonstrate that DHM is a non invasive method (no dye and no electrode) able to characterize pharmacologically the activity of a ligand-gated channel (here the GABA A receptor) simultaneously on a large sample of individual cells. These characteristics could be of great interest for HTS for GABA A receptor/chloride channel modulators.
The GABA A receptor-mediated current (I GABA ) can be derived from the phase signal Our experimental recordings showed that, while the two types of signals (electrical and optical) provide the same information on GABA A receptor properties, their kinetics are strikingly different (figures 2A1 & 4 A1). Thus, at 2100 mV, the rise time (t rise ) or decay time (t decay ) of the phase shift (t rise : 85.069.3 s; t decay : 243620.2 s, n = 8) was significantly longer than for the I GABA (t rise : 23.962.3 s; p,0.005; t decay : 47.864.7 s; p,0.005). Similarly, at +40 mV, the kinetic constants were also significantly longer for the phase shift (t rise : 92.5615.3 s; t decay : 220619.0 s) than for I GABA (t rise : 29.361.6 s; p,0.05; t decay : 57.261.3 s; p,0.005; n = 6). These measures of kinetics suggest nevertheless a strong relationship between the current and phase signal generated by GABA, where the current is a parameter representing a number of charges per unit of time and the phase signal would be a mirror of the accumulation of these same charges during the total opening time of the conductance.
As shown in File S2, this relationship between the current (here I GABA ) and the DHM phase signal (Q t ) is explicitly given by equation 2: where e Ã GABA represents the effective volume variations per number of net charges transported through the membrane, which takes into account of any volume variations, including the nonelectrogenic ones.
In order to quantitatively derive I GABA from the rapid phase response, several parameters from equation 3 must be known. Practically, V 0 (initial volume of HEK cells) as well as the parameters r (a parameter related to the cell deformation) and s (a parameter taking into account transport of substances, including proteins, across cell membranes) have been measured by the decoupling procedure [10], [18]. Consequently, a single parameter remains to be determined, namely e Ã GABA , to calculate the current. Practically, as far as the rapid phase is considered, the decoupling procedure has permitted to demonstrate that s value is not statistically different from 1 when HEK cells are considered, indicating that the intracellular refractive index presents a variation mainly resulting from a dilution or a concentration of the intracellular content by ions and water fluxes (see File S2). As far as HEK cells are considered the typical values of the parameter r are concerned, they are within the range 0.5-0.8, reflecting the fact that the cell deformation associated with the volume change is not isotropic (r = 0.33) but preferentially along the z-axis. Practically, figure 4 shows examples of currents derived from the phase signal at three different holding potentials (2100, 240 and +40 mV) as calculated with equation 2. For these calculations, the parameter e Ã GABA in equation 2, was computed by performing a least squares fit, in order to minimize the sum of the square of the deviations between the measured current from the phase derived current (I pred. ) ( figure 4). It must be stressed that the adjustment of this single parameter e Ã GABA allows obtaining an I pred. in good agreement with the measured current. Typical values for the parameters e Ã GABA lie within the range of 90-110 mm 3 /nC for HEK cells. With these values of e Ã GABA , I pred. is also in good agreement with the recorded current I GABA when a modified ACSF is considered( Figure 4B1).
Finally, in order to fully take dvantage of the possibility afforded by DHM to optically measure transmembrane currents, we have determined the phase-derived currents corresponding to the mean current over a population of unpatched HEK GABA within the same culture. In modified ACSF condition, equation 2, with parameter e Ã GABA set at values in the range indicated above, provides an I pred. from unpatched HEK GABA cells in good agreement with the recorded I GABA from patched cells ( figure 4B2). More importantly, we were able to optically measure the potentiation of I GABA by diazepam from unpatched cells. The validity of this optical measure is confirmed by the recording of I GABA potentiation in patched cells ( figure 4B2). This stresses the capability of DHM to quantitatively determine characteristic currents of a large cell sample in a non-invasive way (without the use of electrodes or fluorescent dyes), thus making this approach amenable to HTS procedures.

Discussion
The results reported in this article show that DHM is a simple, reliable and non invasive optical technique for the determination of the pharmacological properties of a chloride conductance simultaneously in a large cell population. Indeed, the optical signal detected with DHM -the quantitative phase response-is an intrinsic one, linked to a physiological process. The illumination system used for DHM is rather conventional (laser diode) and of low power (,200 mW/cm 2 ) [8], [10]. Moreover, with an appropriate analysis of the phase response it is possible to quantitatively obtain the chloride current which generates it, in a strictly non-invasive manner without using recording any electrodes.
The potential offered by DHM to detect and analyse pharmacologically a chloride conductance in a cell population were confirmed by data obtained for GABA A receptors expressed by HEK GABA . Practically, the results show that the DHM optical phase responses are specifically linked to the activation of GABA A receptors expressed by HEK GABA (no optical response on HEK norm ), with pharmacological characteristics in agreement with those already described in the literature [19]. The ionic species underlying the GABA current could be determined by constructing a phase/current plot indicating a reversal potential close to the theoretically-determined equilibrium potential of chloride. It is known that this ligand-gated chloride channel is also permeable to other anions including HCO 3 2 [20]. However, one should note that the permeability of HCO 3 2 through GABA A receptors is 5 to 10 times lower than that of Cl 2 . Moreover, it is important to note that the optical (and electrical) signals recorded from patched HEK GABA cells have been obtained without HCO 3 2 in both bath medium and the intrapipette solution. Consequently, even if we cannot totally rule out the participation of another anion in the GABA-induced optical signal, our results suggest that, in our recording condition, that the optical (and electrical) signals were essentially associated with chloride fluxes. This is particularly interesting, since no reliable optical imaging techniques are available to specifically study Cl 2 dynamics. To the best of our knowledge this is the first example of an optical technique providing without dye, precise information on specific Cl 2 fluxes in a quantitative manner.
Another important point concerns the ''route'' taken by the water molecules accompanying the chloride fluxes. Of course, one route is a passive diffusion through the plasma membrane, a process known to have a slow kinetics. Another possible ''route'' is a direct water flux through the chloride pore of GABA A receptor. However, it has been shown that chloride conductances are generally poorly permeable to water molecules [20]. Furthermore, the fact that a constant e Ã GABA value in equation 2 allows to derive an adequate current response -the early phase of current increase when GABA is applied as well as the delayed phase of the current recovery when GABA is no longer present -is compatible with a low water permeability of chloride conductances. This transport of water during GABA stimulation could be potentially modulated by membrane proteins involved in the transporting of Cl-(and water molecules) such as KCC and NKCC (For review, see [16]). However, our results have clearly shown that there is no contribution of these two ionic co-transporters to the genesis of optical signal.
These results are in accordance with the literature indicating that theses two co-transporters are weakly expressed in native HEK cells [21], [22]. Other proteins transporting Cl 2 , including the HCO 3 2 /Cl 2 exchanger [23], voltage-dependent chloride conductances (CIC family) and/or VRAC (Volume Regulatory Anion Channel) [24] could also contribute to the fluxes of Cl 2 and water triggered by GABA application.. Considering that, in patch clamp recordings, pH was clamped in both compartments (extraand intracellular medium) it is very likely that the proteins involved in the regulation of pH like HCO 3 2 /Cl 2 exchanger are only weakly activated. Concerning a possible involvement of voltagedependent chloride conductances, large changes in voltage performed in our study (from 2100 to +40 mV) without application of GABA did not trigger any detectable phase signal (data not shown), suggesting only marginal, if any, involvement of these chloride conductances in the GABA-induced optical responses. In addition, these type of Chloride channels are weakly expressed in HEK cells [25]. Finally, while one cannot formally exclude a contribution of the VRAC (Volume Regulatory Anion Channel) in the optical signal genesis triggered by GABA, their involvement is very unlikely. Indeed, if VRAC activation were to occur during cell swelling resulting from a massive influx of Cl 2 (and water molecules), then one would expect a VRAC-dependent signal at different membrane resting potentials during an application of GABA, both when membrane is clamped at +40 mV and at 2100 mV. However this can be excluded since a cell shrinkage occurs at 2100 mV as a result of Cl 2 efflux. More importantly, the GABA-induced optical responses, performed on unpatched cell in the NaSCN condition and used to derive currents, correspond to an increase of phase signal i.e. a cell shrinkage, for which a relevant participation of VRAC is very unlikely.
The most important aspect of this imaging technique is the possibility of simultaneous multi-cellular recording which may be very valuable for a possible HTS application, since, in contrast to classical electrophysiological approaches, multiple responses obtained simultaneously from a population of cells can be averaged following the application of pharmacological agents, without disadvantages linked to labelling techniques (loading of dye or contrast agent and/or bleaching). Practically, obtaining such a simultaneous and non-invasive multi-cellular recording, results was achieved through a drastic change of the transmembrane Cl 2 gradient (substitution of the quasi totality of extracellular Cl 2 by SCN 2 in equimolar manner), which allows to obtain values of resting potential (between 220 to 240 mV) that are well separated from the theoretical Cl 2 reversal potential (in this new condition, = E Cl +30 mV). Within this framework, other approaches changing the transmembrane gradient of Cl 2 could also be considered such as for example the co-transfection of the GABA A receptor and KCC2, since KCC2 can change the gradient of Cl 2 concentration without affecting the membrane potential [26]. However the conditions used in the experiments reported here are considerably simpler to achieve.
Finally, a simple mathematical expression, with a single unknown parameter ''e'', specific for the conductance under study, (here ''e Ã GABA '' for GABA A receptor), relating the phase shift to the measured current has been derived and successfully applied to provide a quantitative determination of the ionic current from the DHM optical signal. Practically, the fact that the cell volume is stable before the GABA application i.e DV r (t~0)~0 and since a constant value of e Ã GABA allows to properly calculate the current I GABA from the phase signal, it follows that DV r (t) does not significantly contribute to the rapid phase response. Thus the parameter e Ã GABA %e GABA represents the volume variation associated with the net charge movement transported across the pore of GABA A receptor and can thus inform about the membrane permeability to water.
In conclusion, this study describes a novel application of DHM to analyse at the single-cell level non-invasively and without the use of dyes, the optical signature of a specific ionotropic receptor activity as well as its modulation by specific pharmacological agents. In this case the activity of the GABA A receptor selectively permeable to Cl 2 and its modulation by diazepam, a drug widely used for the management of anxiety disorders in particular are reported. In addition appropriate mathematical treatment of the optical signal affords the possibility to quantitatively determine the dynamics of the current triggered by the GABA A receptor activity, making this technique amenable to use for pharmacological screenings of modulators developed for the management of human pathologies involving dysfunctions of chloride channels.

Cell preparations
HEK 293 cells stably expressing configurations of rat GABA A receptors (HEK GABA ) were generously given by Hoffmann-LaRoche (Basel, Switzerland). Briefly, cDNAs encoding rat GABA A a1, b2 and c2s subunits [27], [28]were subcloned into the expression vectors pIRESpuro2, pIRESneo2 and pIREShy-gro2 vectors (Clontech, Mountain View, CA), respectively. The pIRES/GABAA a1, b2, c2s, constructs were sequenced to confirm their nucleotide sequence and then cotransfected into HEK 293 cells at a ratio of 1:1:2 (plasmid mass ratio) using the lipofectamine 2000 kit according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). Transfected cells were grown in minimal essential medium (Invitrogen) supplemented with 10% fetal calf serum (Invitrogen), 20 mM HEPES (Invitrogen) and 100 U/ml penicillin/100 mg/ml streptomycin (Invitrogen) for 48 hours and then, the cells were transferred to the selection medium containing 0.3 mg/ml puromycin (Clontech, Mountain View, CA, USA), 300 mg/ml hygromycin B (Roche Diagnostics, Mannheim, Germany) and 200 mg/ml G418 (Invitrogen) for the generation of stable cell lines. Cell colonies were isolated and expression of the GABA A a1b2c2s receptor was determined by [ 3 H]flumazenil binding.

Electrophysiology recording
Whole-cell recordings were made, and signals were amplified by using Multiclamp 700B amplifiers (Axon Instruments, Union City, CA) and digitized by means of an ITC-1600 interface (Instrutech, Great Neck, NY) to a PC computer running Igor Pro (Wavemetrics, Portland, OR). All currents (sampling interval, 5 kHz) were low-pass filtered (2 kHz

Imaging
Digital Holographic Microscopy (DHM) is a full-field interferometric imaging technique that allows to derive from the quantitative phase signal, minute changes in cell refractive index and volume associated with a variety of biological processes without dye or contrast agent (for review, see [29]), in particular when transmembrane water movements are involved [10]. Indeed, changes in ionic permeability are accompanied by a rapid movement of water through the plasma membrane, in particular when involving Cl 2 [13]. Briefly, as shown by Marquet et al. [8], for each pixel of the DHM images, the phase shift induced by the observed cell is given by the following equation: where l is the wavelength of the lightsource, h the cell thickness, n c the mean intracellular refractive index (along the path length corresponding to the thickness h), and n m the refractive index of the perfusion solution. Consequently, the phase signal depends on two distinct cell parameters: h which provides information concerning cell morphology and volume, and n c whose value is related to the amount of non-aqueous material present in the cell and is essentially determined by the protein content [30]. DHM can therefore quantitatively detect small variations of the phase, which mainly depend on the refractive index of the cell n c and cell morphology and volume derived from h. However, the value of the phase is mainly dictated by the intracellular refractive index rather than by cell morphology [10]. In turn, this refractive index is dependent on the protein content of the cell [30], [10]. Accordingly, entry of water will dilute the intracellular protein content resulting in a decrease in the phase while an exit of water will concentrate the protein content leading to an increase in the phase.
The basic design of our imaging system has been described in [10], [31], [13]. Briefly, holograms are acquired with a DHMT 1000 (Lyncée Tech SA, PSE-EPFL). A laser diode produces the coherent light (l = 683 nm) which is divided by a beam splitter into a reference wave and an object wave. The object wave diffracted by the specimen is collected by a microscope objective and interferes with a reference beam to produce the hologram recorded by the CCD camera. Frequency of hologram acquisition is 0.2 Hz. Reconstruction of the original image from the hologram is numerically achieved by a computer. The reconstruction algorithm provides simultaneous amplitude and quantitative phase images of the cells (Koala software). It is important to note that an extensive quality control of the DHM technique has been published in Rappaz et al. [18].
Pratically, before to start a phase recording we waited for a minimum of 10 min in order to obtain a stable baseline. GABA was added after a minimum of 1 min of stable baseline recording for both the optical and the electrical signals.

Offline analysis
The electrophysiological and optical recordings were analysed by using MATLAB 7.6 (Mathworks Software, Natick, MA) and all curves have been fitted by using ORIGIN 7.5 (Microcal Software, Northampton, MA). GABA concentration-response (or time application-response) profiles were fitted to the following logistic equation: Q/Q max = 1/[1+(EC50/[GABA]) n ], where Q and Q max represented the normalized GABA induced phase shift at a given concentration (or time application) and the maximum phase shift induced by a saturating [GABA], EC50 was the half-maximal effective GABA concentration (or time application), and n was the slope factor. For both, optical and electrical response of GABA application, rise time (t rise ) and decay time (t decay ) correspond to 0-100% peak amplitude.
All data are presented as means 6 SEM. Student's t-test (paired or unpaired) to determine statistical significance (p,0.05).