A highly-selective chloride microelectrode based on a mercuracarborand anion carrier

The chloride gradient plays an important role in regulating cell volume, membrane potential, pH, secretion, and the reversal potential of inhibitory glycine and GABAA receptors. Measurement of intracellular chloride activity, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{\boldsymbol{a}}}_{{\boldsymbol{Cl}}}^{{\boldsymbol{i}}}$$\end{document}aCli, using liquid membrane ion-selective microelectrodes (ISM), however, has been limited by the physiochemical properties of Cl− ionophores which have caused poor stability, drift, sluggish response times, and interference from other biologically relevant anions. Most importantly, intracellular \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bf{HC}}{{\bf{O}}}_{{\bf{3}}}^{-}$$\end{document}HCO3− may be up to 4 times more abundant than Cl− (e.g. skeletal muscle) which places severe constraints on the required selectivity of a Cl− – sensing ISM. Previously, a sensitive and highly-selective Cl− sensor was developed in a polymeric membrane electrode using a trinuclear Hg(II) complex containing carborane-based ligands, [9]-mercuracarborand-3, or MC3 for short. Here, we have adapted the use of the MC3 anion carrier in a liquid membrane ion-selective microelectrode and show the MC3-ISM has a linear Nernstian response over a wide range of aCl (0.1 mM to 100 mM), is highly selective for Cl− over other biological anions or inhibitors of Cl− transport, and has a 10% to 90% settling time of 3  sec. Importantly, over the physiological range of aCl (1 mM to 100 mM) the potentiometric response of the MC3-ISM is insensitive to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bf{HC}}{{\bf{O}}}_{{\bf{3}}}^{-}$$\end{document}HCO3− or changes in pH. Finally, we demonstrate the biological application of an MC3-ISM by measuring intracellular aCl, and the response to an external Cl-free challenge, for an isolated skeletal muscle fiber.

usually difficult to de-convolve into accurate molar units, and in some conditions, sensors can alter the extent and kinetics of the ion concentration changes aimed to be measured. Fluorescence life-time imaging circumvents some of these challenges 9 , but requires specialized and expensive detection systems. In principle, an ISM is devoid of these limitations, and is expected to provide a direct determination of intracellular activity. In fact, ISMs are often used to measure the activities of ions in solutions that are used to calibrate photometric sensors. On the other hand, ISMs cannot track fast changes in ion activity (i.e. in the ms range, but see 13,14 ) and in some cases selectivity is a limiting factor. The key component of an ISM is the ion-selective carrier or ionophore, which endows the electrode with the selectivity required to discriminate between ions of similar nature and the sensitivity (usually down to the sub-micromolar range). In contrast to the case for cations, there is a general lack of highly-selective naturally occurring or synthetic anion ionophores, and in particular for Cl − 15,16 .
We tested ISMs made with several commercially available Cl − ionophores, and unlike our experience with cation ionophores (H + , Na + ) we found their responses far from ideal. These Cl − ionophores included: tributyltin (TBT), chloride ionophores I-IV (Selectophore TM , Millipore-Sigma), and the antibiotic 3,4,4′trichlorocarbanilide 17 . We found ISMs fabricated with these compounds had one or more of the following problems: lack of linearity, sub-or supra-Nernstian responses, drift, hysteresis, poor solubility in liquid membranes, sensitivity to blockers of chloride channels or transporters, and poor selectivity for Cl − over − HCO 3 . We also tested a chemically diverse group of recently published compounds described as Cl − carriers for biological membranes: compound-1 18 , compound 4H 19 , cholapod-3, and decalin-13 16,20,21 , and found them not appropriate to build ISMs, mostly demonstrating solubility problems.
Here we report that a Cl-selective anion carrier based on a trinuclear Hg(II) complex containing carborane-based ligands, [9]mercuracarborand-3 (MC3), that was previously used in polymer membrane Cl − sensors 22 is also ideal for liquid membrane Cl-selective ISMs. MC3 is a macrocyclic structure bearing a pre-organized Lewis acid cavity that selectively complexes with Cl − and functions as an anion carrier in the liquid membrane ISM. We show the MC3-ISM is highly selective for Cl − over other biologically relevant anions ( − HCO 3 , lactate, − PO 4 ) and is insensitive to Cl-channel blockers (9-anthracene carboxylic acid), inhibitors of cation-chloride co-transporters (bumetanide, furosemide), or changes in pH and is therefore well-suited for biophysical applications to quantitatively and accurately measure a Cl i .

MC3 synthesis.
To our knowledge, [9]mercuracarborand-3 (MC3, CAS Registry Number 146219-95-6) is not commercially available and so it was synthesized using air-free Schlenk techniques as described by Hawthorne and co-workers 23 . An oven dried Schlenk flask was charged with 144 mg (0.5 mmol) of ortho-carborane (Boron Specialties), filled with an N 2 atmosphere, and 5 mL of anhydrous diethyl ether (Et 2 O) (Fisher, inhibitor free) were added via syringe. The reaction vessel was cooled to 0 °C using an ice bath, 0.85 mL of 2.5 M n-BuLi in hexane (Sigma-Aldrich) was added dropwise to the rapidly stirring reaction mixture. After addition of the n-BuLi solution, the ice bath was removed and the reaction was stirred at room temperature for 2 hours. Next, 315 mg of Hg(OAc) 2 (Mallinckrodt) was added to the reaction mixture under a flow of N 2 and the reaction mixture was stirred at room temperature for 16 hours. Upon completion of the reaction, the reaction was quenched with 5 mL of H 2 O, the organic layer was collected, the aqueous portion was extracted with 5 mL Et 2 O. The organic portions were combined and washed with 5 mL H 2 O, then the organic layer was dried with MgSO 4 and filtered through 2 cm of silica gel using Et 2 O as the eluent. The filtrate was evaporated to dryness using a rotary evaporator to yield a white solid. Due to difficulty with purifying the product by recrystallization from Et 2 O, an alternative purification method was developed as follows. The white solid was washed with hexanes, dissolved in 10 mL of 20% (v/v) hexanes in Et 2 O and left in a fume hood to evaporate the Et 2 O portion which resulted in precipitation of an offwhite powder. The supernatant was decanted and the solids were loaded onto a 3 cm plug of silica gel wetted with 40% acetone (v/v) in hexanes and eluted with 40% acetone (v/v) in hexanes. The resulting filtrate was evaporated to dryness to yield 50 mg (14% yield) of an off-white solid which contains >95% MC3 by 11  Micropipette silanization. Micropipettes were pulled from 1.5 mm borosilicate capillaries with micro-filaments (BF150-86-10, Sutter Instruments). A 4-step pulling protocol was optimized to obtain a short taper (~8 mm) leading to a sub-micron tip opening, using a horizontal puller (P97, Sutter Instruments). When filled with a saline mimicking the ion composition of myoplasm (see solutions), the microelectrodes had tip resistances of 10-12 MΩ. The use of capillaries with micro-filaments greatly reduced the filling-time in the manufacture of ISMs (~5 min compared to >30 min for plain capillaries), without compromising the sealing of the organic phase to the inner surface of the capillary wall. Capillaries were silanized as supplied from the vendor, with no pre-treatment. For silanization, micropipettes were laid horizontally on a support in a Pyrex glass jar (with lid, ~500 ml). One hundred μl of chlorotrimethylsilane (Sigma) was placed in the jar and allowed to evaporate at RT for ~5 min. The silane vapor had free access to a micropipette's interior and exterior surfaces. Then, the micropipettes were baked at 250 °C for at least 4 h (usually overnight). As occurs with all liquid-junction microelectrodes, silanization was essential to insure the mechanical stability and high electrical resistance between the hydrophobic liquid membrane and the inner wall of an ISM. Silanized micropipettes were stored dry in a closed jar (WPI Jar-E215) to avoid dust adhesion, and were used for up 2 weeks with no noticeable deterioration. cl − sensing liquid membrane and iSM half-cell saline. The ISM liquid membrane consisted of a short column (200-300 μm) of a hydrophobic "chloride cocktail" lodged in the tip of the micropipette. The chloride cocktail was made by dissolving the MC3 and a lipophilic cationic additive (tridodecylmethyl ammonium chloride, TDMAC, Sigma) in the water -immiscible organic solvent 2-nitrophenyl octyl ether (NPOE, Sigma). We used 10% MC3 and 2.5% TDMAC in NPOE (w/volume). No other mixing ratios or solvents were studied. The ISM half-cell consisted of 100 mM NaCl, backfilled into the microelectrode and in contact with an Ag/AgCl reversible electrode. iSM fabrication. We devised a simple and efficient method to fabricate ISMs. Starting with previously silanized micropipettes, we routinely made six ISMs in less than 1 h. To fabricate an MC3-ISM both the chloride cocktail and the reference saline were sequentially backfilled into the micropipette, while viewing under a stereomicroscope. First, an excess amount of chloride cocktail (2-3 times the final volume of the liquid membrane) was delivered as close as possible to the micropipette tip using a 34G microneedle (Quickfil 34G-5, WPI, for which the polyimide coating was stripped from the end). The organic phase spontaneously moves towards the tip of the micropipette by capillary action and fills the shaft in 3-5 min. Initially using excess chloride cocktail speeds up this filling step. Once the tip is filled, the chloride cocktail volume is reduced as much as possible by suction with a new 34G needle. Finally, the saline solution is delivered (using another 34G needle), assuring that a clean interface is formed between the two liquids. Care must be taken to avoid air bubbles in the organic phase or interface, otherwise the ISM should be discarded.
Prior to use, each MC3-ISM was individually tested using a 3 point calibration including solutions of 1, 10 and 100 mM Cl (see solutions), with a Cl of 0.64, 6.46, and 68.5 mM respectively. Tested MC3-ISMs were stored in closed jars (to avoid evaporation and dust) with tips submersed in a filtered (0.2 μm) solution containing 10 mM NaCl; and could be used after 1-2 weeks from fabrication with no detectable deterioration.
Measurement of iSM potential. A two-channel high input impedance (>10 15 Ω) amplifier (FD223a, WPI) was used to measure the electrical potential of MC3-ISMs. One channel was connected to the Ag/AgCl wire of an MC3-ISM. The second channel was connected to a standard ("sharp") microelectrode having 10-15 MΩ tip resistance when filled with electrode solution (see solutions). This microelectrode served as a reference electrode, whose value (V ref ) reflects changes in junction potential that may occur upon solution exchanges.  Solutions containing potential interfering anions were prepared by adding 10 mM of Na-lactate, NaHCO 3 , Na 2 HPO 4 or NaSCN to a mixture of the 150 − Cl and 0 − Cl solutions, and then performing a similar serial dilution as described above to obtain the desired [Cl − ]. Drug-containing solutions were made from highly concentrated stock solutions of 9-anthracenecarboxylic acid (9-ACA, 500 mM in DMSO) or bumetanide (BMT, 50 mM in DMSO). DMSO final concentration was kept below 1:1000, and controls were run to exclude direct effects of solvent.
All chemicals were purchased and used as received from Millipore-Sigma, unless otherwise noted (vide supra). Muscle fiber ISM measurements. To test the MC3-ISM in a cellular context, we used enzymatically dissociated skeletal muscle fibers from the flexor digitorum brevis (FDB) of the mouse, as previously described 26 .
All mouse procedures were approved by the University of California Los Angeles Institutional Animal Care and Use Committee. All methods performed in this study were conducted in accordance with the guidelines and regulations of the IACUC. Fibers were plated in the same chamber as above, and impaled with a MC3-ISM and a reference micro-electrode. Before impalement, both MC3-ISM and reference micro-electrode were zeroed in the presence of 100 − Cl Tyrode.

Selectivity coefficient, K Cl X pot
, . The relative selectivity of an MC3-ISM for Cl − over a test interfering anion was determined by the fixed interference method 27 . The concentration of the test interfering anion, [X − ], was maintained at a fixed value of 10 mM, and the MC3-ISM /potential was measured as the [Cl − ] was varied from 10 nM to 100 mM. The MC3-ISM potential was plotted against log(a Cl ) and the intersection of the two linear portions (Nernstian for high [Cl − ] and saturating for low [Cl − ]) determined the value of a Cl that was used to calculate the selectivity coefficient for Cl − over X − as: Values are reported as the mean ± standard error of the mean.

Results
A cationic additive increases the cl − responsiveness an MC3-based ISM. We first tested the response of liquid membrane ISMs constructed using a cocktail containing only MC3 (10%) and NPOE. These electrodes were only weakly sensitive to [Cl − ] changes, as shown for an exemplary response in Fig. 1A. This response is sub-Nernstian, with a slope of 9.2 ± 0.67 mV/decade (n = 3) on a plot of V Cl as a function of log 10 (a Cl ) (Fig. 1C, blue symbols). This result is in agreement with previous data 22 , demonstrating that polymeric membranes containing MC3 are insensitive to Cl − unless the membranes are doped with cationic compounds. www.nature.com/scientificreports www.nature.com/scientificreports/ When the liquid membrane cocktail was supplemented with a lipophilic cationic additive, TDMAC (2.5%, w/w), the potentiometric response became highly sensitive to [Cl − ] changes between 1 and 100 mM (Fig. 1B). For brevity, we refer to microelectrodes with this complete liquid membrane cocktail as MC3-ISMs, but it should be recognized they are doped with TDMAC. The voltage response of MC3-ISMs was stable and showed no signs of hysteresis. Over this range of [Cl − ], the MC3-ISM response was nearly Nernstian, with a slope of 54.6 ± 0.18 mV/ decade (n = 24), as shown in Fig. 1C (black symbols). The reference electrode had very small voltage changes in response to these solution exchanges (2-3 mV, not shown), and these small deviations were subtracted from V ISM to obtain V Cl values shown in Fig. 1.
We did not explore other combinations of MC3 and TDMAC since the 10% and 2.5% (w/w), respectively, mixture tested gave optimal results. The need for a lipophilic cation stems from the fact that MC3 is a neutral carrier, and as such it is expected to produce a limited anion response unless a counterion for the mercuracarborand complex is added 22 . A limited number of trials were also performed with solutions prepared by serially diluting 100 mM NaCl with a Na 2 SO 4 -based solution (Fig. 2B) that is commonly used for "chloride-free" conditions in studies on skeletal muscle fibers (Methods, 0 -Cl solution). This alternative method maintains a high ionic strength for all test values of [Cl − ] and provides a check for evidence of interfering anions in a mammalian physiologic buffer. A plot of steady-state V Cl as a function of log 10 (a Cl ) shows that between 0.1 mM and 100 mM the response was linear (Fig. 2C). This range exceeds the expected values of intracellular a Cl to be encountered in mammalian cells (3 to 20 mM, or even up to 100 mM in pathological contexts such as glioma 28 ). Moreover, the response for dilutions with pure water was near-linear down to an activity of 0.01 mM. The saturation of V Cl at about −200 mV for the NaCl: Na 2 SO 4 mixture reveals an interfering anion effect of − SO 4 2 , although this was detectable only at low a Cl levels (<0.1 mM) that are not encountered in cells.

Linearity and cl
The average slope of the linear response range was 52.8 ± 0.63 mV/decade (n = 14). While this slope is slightly smaller than the theoretical value of 58 mV, it is comparable to values reported for MC3 electrodes fabricated in solid membranes 22 . Those electrodes also showed a sub-Nernstian response, with the slope being smaller for lower molar ratios of TDMAC to MC3.  The response time of a liquid-junction ISM is usually faster as the width of the organic phase decreases at the tip of the electrode 13,14 . The width for the liquid membrane of our MC3-ISMs was ~300 μm, and we did not explore whether the time constant could be reduced by decreasing the width of the organic phase. In any event, a time response in the order of a few seconds is sufficient to study intracellular [Cl − ] changes expected to occur in several tens of second to minutes in biological settings.  15 . Selectivity for Cl − was greatly improved by using anion-selective carriers, e.g. Mn(III) porphyrins 30 , but the physiochemical properties of these first-generation carrier electrodes were unfavorable with instability, drift, and non-Nernstian behavior. The high selectivity for Cl − over − HCO 3 and near-Nernstian behavior of MC3-based solid membrane electrodes 22 showed tremendous promise for developing a Cl-selective liquid membrane ISM suitable for biological applications.

The MC3-ISM is insensitive to changes in pH or
We measured the sensitivity of MC3-ISMs to interference from changes in pH or [ − HCO 3 ], as a test of suitability for these ISMs to be used for intracellular determination of a Cl . First, we tested for MC3-ISM sensitivity to ΔpH from 6.5 to 8.5, which covers the entire physiologic range of expected cytoplasmic values. In these HEPES-buffered − HCO 3 -free solutions, a three-point Cl − calibration response (1, 10, 100 mM) for an MC3-ISM was not affected by a pH shift from 7.4 to 6.5 or from 7.4 to 8.5 (Fig. 4A, left and right, respectively). Next, we measured the response of MC3-ISMs in constant 10 mM Cl − , a typical value for myoplasmic Cl − in skeletal muscle fibers 31,32 , upon which was imposed step changes in HCO 3 − from 1 to 50 mM (Fig. 4B), a concentration range far beyond those expected to occur in vivo either in physiological or pathological conditions (i.e. ~13 mM 33 ). The potentiometric response of MC3-ISMs was completely insensitive to the presence of − HCO 3 at concentrations as high as 50 mM. This is a remarkable feature of the MC3 Cl-selective carrier, and makes MC3-based ISMs the electrode of choice for biological purposes.
Chloride selectivity of the MC3-ISM. The selectivity of the MC3-ISM was assessed for Cl − over various physiologically relevant anions including lactate, phosphate, thiocyanate and bicarbonate. The fixed interference method, as described by IUPAC was used 27 , wherein the concentration of the test interfering anion was held constant at 10 mM (as a Na + salt), and V Cl was measured in solutions with varying [Cl − ] between 100 mM and 100 nM. We found that the MC3-ISM response is not significantly affected by the presence of lactate, phosphate, or bicarbonate for [Cl − ] from 100 mM to values as low as 100 μM (Fig. 5). On the other hand, the MC3-ISM was almost insensitive to Cl − in the presence thiocyanate (Fig. 5, magenta diamonds). From the data in Fig. 5, we calculated the selectivity coefficient, K Cl X pot , , for each test anion (Table 1) which gives a selectivity series: thiocyanate > Cl ≫ bicarbonate > phosphate ≈ lactate. Importantly, the MC3-ISM was 125-fold more sensitive to Cl − than to − HCO 3 , which explains the lack of an effect for − HCO 3 in Fig. 4 and confirms the MC3-ISM is suitable for measuring intracellular a Cl without interference from fluctuating levels of − HCO 3 .    www.nature.com/scientificreports www.nature.com/scientificreports/ interference was based on detecting a distortion of the Nernstian behavior of an MC3-ISM over an operating range of 1 mM to 100 mM [Cl − ], while the Cl − agents were added at concentrations typically used for studies of muscle. We found that the MC3-ISM is almost insensitive to 9-ACA (Fig. 6A,D), with a very small effect (reduced V Cl response) in 200 μM 9-ACA when [Cl − ] was 5 mM or lower. Studies in skeletal muscle that are designed to eliminate the ClC-1 conductance typically use 100 μM 9-ACA 36 . With this concentration of blocker, the V Cl response remained Nernstian down to a [Cl − ] of 2.5 mM (the lowest expected physiologically), demonstrating that MC3-ISMs can be used to measure intracellular a Cl in studies with 9-ACA. The MC3-ISM response was insensitive to BMT (Fig. 6B,E). At 20 μM, a concentration in 10-fold excess of that used for complete inhibition of NKCC1, there was no detectable distortion of the MC3-ISM Nernstian behavior over a [Cl − ] range from 1 mM to 100 mM. BMT concentrations up to 100 μM have been used for some investigations 37 , although in our experience this approaches the solubility limit in physiological saline. Even for a 1:500 dilution of a 50 mM stock BMT solution in DMSO (nominally 100 μM BMT), the MC3-ISM was distortion-free down a [Cl − ] of 2.5 mM. The MC3-ISM response was mildly attenuated in the presence of 100 μM FUR in Cl − solutions of 2.5 mM and lower (Fig. 6C,F). intracellular cl − activity, a Cl i , measured with an MC3-ISM. We verified the performance of the MCS-ISM as a sensor of intracellular Cl − , by recording a Cl i in a skeletal muscle fiber isolated from the flexor digitorum brevis of the mouse. The fiber was impaled with a standard reference electrode to measure the resting potential (V m Fig. 7, upper panel) and also with a Cl-selective MC3-ISM (V ISM ) which reports the combined effect of V m and V Cl . The offset for V ISM = 0 mV was set in a calibration solution with 100 mM [Cl − ]. The difference between these two signals, V Cl = V m − V ISM (Fig. 7, lower panel) is proportional to log a ( ) In control Tyrode solution (140 mM Cl − ), the resting membrane potential was stable (V m = −80 mV), as was V Cl = −65 mV, corresponding to a Cl i = 6 mM. Upon switching the perfusion to SO 4 -Tyrode (nominally zero Cl − ) the intracellular a Cl i exponentially fell to about 1.5 mM because of Cl − efflux through the high resting conductance of ClC-1 channels. These changes are fully reversible, as shown by the recovery of a Cl i when the perfusion with Cl − Tyrode was resumed at 500 sec in Fig. 7.
The transient response of V m is also consistent with decrease of a Cl i , followed by recovery 4 . At the onset of Cl-free Tyrode solution (Fig. 7, 80 sec), V m depolarized because the equilibrium potential for Cl − suddenly shifted from about −77 mV to a positive value. Intracellular Cl − content then declined over the next 30 s, which shifted E Cl to more negative potentials and caused V m to relax back toward −80 mV. When extracellular Cl − is suddenly www.nature.com/scientificreports www.nature.com/scientificreports/ restored (Fig. 7, 500 sec), V m had a hyperpolarizing undershoot because intracellular a Cl i was now lower than baseline. As a Cl i increased back to the normal basal level, V m relaxes to the initial value of −80 mV. Similar responses were observed in 8 additional muscle fibers.

Discussion
The development of a highly-selective and sensitive chloride sensor in polymeric membrane electrodes, based on the preorganized macrocyclic Lewis acid MC-3 as the anion carrier 22 , was a major technological advance. Prior electrode designs using anion exchangers (e.g. TBT or TDMAC) lacked selectivity, which in these electrodes was dominated by anion lipophilicity (Hofmeister series). Improved Cl − selectivity was achieved with metalloporphyrin carriers (e.g. Mn(III), chloride ionophore I Sigma-Aldrich), but interference from  39 (e.g. chloride ionophore II, Sigma-Aldrich), but these compounds were suboptimal for liquid membrane ion-selective microelectrodes because of sluggish response times (t 90% ~ 10 s), electrical instability with drift and non-Nernstian responses, or poor solubility in NPOE. In contrast, MC-3 is chemically stable, and MC-3 based polymeric membrane electrodes have more consistent Nernstian responses, high selectivity for Cl − , and faster settling times of a few seconds. The MC3 ionophore represents a unique platform for chloride sensing. These properties stem from the molecular architecture of this structure which includes a precise triatomic arrangement of Hg(II) sites separated by rigid ortho-carboranyl boron-rich cluster ligands. Such a trinuclear metal receptor arrangement allows MC3 to bind anions in a multivalent fashion. Furthermore, electron-withdrawing nature of the C-bound ortho-carborane clusters, similar in magnitude to the perfluoroaryl substitutents 40,41 , enhances the Lewis acidity of the Hg(II) centers critical to the anion binding. Finally, the sterically encumbering nature of the boron cluster ligands enhances the kinetic stability of the corresponding organometallic complex and creates a well-defined pocket around the Hg-based centers modulating the anion binding selectivity.
We now show that the advantages of MC-3 based polymeric membrane electrodes can be realized in a liquid membrane microelectrode form factor. Following from the design of the MC-3 electrodes in PVC membranes 22 , we used NPOE as the liquid organic phase in our ISMs and no other plasticizers were tested. Just as with the PVC membrane electrodes, the potentiometric response to Cl − with MC-3 alone was sub-Nernstian (Fig. 1), and doping the liquid membrane with a lipophilic cationic additive (2.5% TDMAC) markedly increased the ISM sensitivity to near-Nernstian behavior (54.3 mV/decade). Increased amounts of cationic additive in PVC membrane electrodes extended the linear range for high a Cl to about 100 mM, but with a cost of slightly reduced selectivity for Cl − over other anions 22 . Since our ISMs already had linear responses up to the highest concentration of the calibration solution (nominally 100 mM mixture, calculated a Cl = 68.5 mM), we did not test higher amounts of additive. Moreover, in a physiological context the expected intracellular a Cl is in the range of 3 to 20 mM.
The potentiometric responses of MC-3 ISMs were remarkably consistent and stable over long recording periods. Except for the occasional failure in the fabrication of an ISM (e.g. air bubble at the liquid membrane/aqueous electrolyte interface), the variability was exceptionally small in the slope of the linear response range (0.1 mM to 100 mM). A sample of 24 MC-3 ISMs with an average slope of 54.6 mV/decade had a standard deviation of only 0.9 mV/decade and a range of 52.9 to 56.0 mV/decade. By comparison, a sample of 62 ISMs constructed with the "improved" Corning 477913 chloride ionophore was reported to have a mean slope of 52.8 mV/decade with a standard deviation of 5.1 mV/decade and the range of values extended from 35 to 64 mV/decade 42 . The potentiometric response of MC-3 ISMs was stable with no significant drift over ten minutes or more, as shown  (2)(3)(4)6). Moreover, the intracellular response recorded from a skeletal muscle fiber was also stable, as shown by the return to the baseline value after recovery from a 10 min exposure to Cl-free conditions (Fig. 7).
The slow electrical response time of Cl-selective ISMs has been a limiting factor for use of some ionophores. The problem arises from the high resistivity of immiscible liquid membranes with neutral ion carriers in combination with the capacitance of the glass pipette. While reducing the ISM tip resistance with cationic additives, choice of the ionophore, and optimization of the liquid membrane thickness may be used to improve performance, the 10% to 90% response time was reported to be as long as 41 sec for Mn(III) porphyrin Cl-selective ISMs 38 . For macroscopic MC-3 electrodes in PVC membranes, a more favorable rise time of about 10 sec was observed 22 . In our MC-3 ISMs the 10% to 90% response time was about 3 sec (Fig. 3), which is sufficient to monitor changes of intracellular a Cl in muscle fibers that typically occur over tens of seconds to minutes.
The most important advantage of MC-3 ISMs is the high selectivity for Cl over other anions, especially − HCO 3 . Historically, the Cl − ISMs used in several landmark papers showing intracellular a Cl in muscle and heart is higher than expected from passive electrodiffusion were based on proprietary Cl − ionophores from Corning (477315 and later 477913) that had substantial interference from − HCO 3 29,31,32 . This limitation was compounded by the composition of the intracellular milieu, where a Cl is typically 4 mM, and  (Fig. 4) is also important because intracellular pH often changes or is intentionally manipulated during a study. The accompanying fluctuations of intracellular a HCO3 would render the potentiometric response uninterpretable for a poorly selective Cl − ISM. Sensitivity to pH is also a limitation for genetically encoded Cl sensors based on YFP variants, which has led to the development of ratiometric dual-sensor fluorescent proteins (ClopHensor) to monitor both [Cl] and pH simultaneously 43 .
Another selectivity-based limitation of earlier Cl − ISMs was interference from drugs used to block Cl − channels (e.g. 9-ACA) or to inhibit Cl − exchangers and cotransporters (e.g. SITS, bumetanide, furosemide). These compounds are often used in studies of Cl transport, where drug-dependent changes in extracellular and intracellular Cl − activity are measured with ISMs. These inhibitors have high lipid solubility and are anions at physiologic pH, and so not surprisingly, substantial anion interference occurs for ISMs using poorly selective ionophores such as Corning 477413. For example, furosemide at 100 μM caused nonlinear potentiometric responses for a Cl ≤ 30 mM, with a corresponding selectivity coefficient of K Cl Fu Pot , = 157 35 . We tested three drugs that inhibit Cl − transport: furosemide, bumetanide, and 9-AC. In furosemide, we detected mild deviations from linear Nernstian behavior for MC3-ISM electrodes when a Cl ≤ 1.6 which is 20-fold lower than for Corning 477413 (Fig. 6C,F). In contrast, the potentiometric response of the MC3-ISM over the entire range of test solutions ([Cl] from 1 to 100 mM) was not affected by 20 μM bumetanide (Fig. 6B), which is ten times the IC 50 to inhibit the Na-K-2Cl 1 cotransporter. For the ClC-1 channel blocker, we detectable interference in 200 μM 9-ACA when [Cl] <10 mM (Fig. 6A). Studies in skeletal muscle typically use 100 μM 9-ACA, and at this lower concentration we detected a change in the MC3-ISM response only for [Cl] <2.5 mM. Overall, the MC3-ISM is highly selective for Cl with a linear Nernstian response over the biologically meaningful operating range of 1 to 100 mM, with no interference from endogenous anions (10 mM − HCO 3 , − PO 4 , or lactate) or bumetanide and very modest effects in high-dose 9-ACA well above the IC 50 .
Finally, we demonstrate that the MC3-ISM reliably measures a Cl in a biological context, the intracellular activity in an isolated skeletal muscle fiber (Fig. 7). The response to an external Cl-free challenge is a stringent test of the MC3-ISM because the resulting decrease of intracellular a Cl would reveal effects from interfering intracellular anions or other sources of distortion as an attenuation of the change in the potentiometric signal. The response in Fig. 7 clearly shows a large response for V Cl = −(V MC3-ISM − V ref ), corresponding to a decrease of intracellular a Cl to less than 2 mM. The record in Fig. 7 also shows the high quality of the signal-to-noise for measuring intracellular a Cl , as well as the stability of the response with no significant drift and a return to the same baseline when external Cl is restored 10 minutes later. Measurement of intracellular a Cl is also of great interest in studies of the nervous system because the Nernst potential for Cl − determines the postsynaptic response to activation of glycine and GABA A receptors. Moreover, intracellular Cl − varies over time, ranging from 3 to 20 mM or higher. Neuronal intracellular Cl − is high in the immature brain, in part due to activity of NKCC1 that promotes Cl − influx, and so receptor activation elicits post-synaptic depolarization (i.e. excitation). With development, NKCC1 decreases and is replaced by KCC that promotes Cl − efflux 3,44,45 . The reduced intracellular Cl − shifts the Nernst potential for Cl − below the resting potential and receptor activation now produces hyperpolarization (inhibition). Derangements of the Cl − gradient in the neurons have been implicated in epilepsy 46 and in anomalous regulation of cell volume that drives proliferation and migration of gliomas 47 . The MC3-ISM offers an improved approach for measuring somatic a Cl for the determination of basal levels and changes that may occur over seconds to minutes -for example in response to inhibition of specific transporters. Of course alternative methods, such as dye fluorescence lifetime 9 or genetically encoded Cl-sensors 10 , are required to measure local a Cl changes within diffusion-restricted spaces in out in the dendrites or for fast transients less than 3 sec.