What biologists want from their chloride reporters – a conversation between chemists and biologists

As the most abundant anion in the body, chloride plays crucial roles in physiology across diverse cell types. As such, dysfunctional chloride homeostasis leads to a number of serious diseases. Correct chloride flux is maintained by diverse and often tissue-specific families of anion channels, which exhibit low selectivity among other biological anions but are referred to as chloride channels because of the predominance of chloride. In mature neurons, the intracellular chloride concentration ([Cl⁻]_i) is a primary determinant of inhibitory synaptic action potential (Kaila et al., 2014; Medina et al., 2014). Under normal conditions, [Cl⁻]_i is kept low by a K⁺-Cl⁻ cotransporter (KCC2, encoded by the gene SLC12A5), allowing activation of the γ-aminobutyric acid (GABA) receptor (GABA_AR) to drive chloride down the electrochemical gradient into the neuron (Doyon et al., 2016). Improper chloride homeostasis is therefore associated with several severe neurological disorders and epilepsies (Ben-Ari et al., 2012; Huberfeld et al., 2007; Payne et al., 2003). In epithelial cells, the chloride channel activity of the cystic fibrosis transmembrane regulator (CFTR) is associated with transcellular water and salt secretion. When this process is dysfunctional, the fluid layer lining the conducting airways cannot remove inhaled pathogens and debris, leading to cystic fibrosis (CF) (Frizzell and Hanrahan, 2012; Saint-Criq and Gray, 2017). Chloride channels have also been identified that are sensitive to cell volume and extracellular pH, which are involved in signaling following cell swelling and hypoxia, respectively (Qiu et al., 2014; Yang et al., 2019a). Finally, intracellular chloride channels have been implicated in endosomal pH regulation, lysosomal degradation, and endoplasmic reticulum (ER) and mitochondria function (Chakraborty et al., 2017; Jia et al., 2015; Kornak et al., 2001; Mindell, 2012; Novarino et al., 2010; Piwon et al., 2000; Ponnalagu and Singh, 2017; Weinert et al., 2010).

Identifying and characterizing these channels has relied heavily on fluorescent chloride reporters. In many cases, screens using genetically encoded halide-sensitive yellow fluorescent protein (YFP) variants have identified the chloride channel involved in a given physiological process (Ullrich et al., 2019; Voss et al., 2014, Qiu et al., 2014; Yang et al., 2019a). However, diverse physiological roles for chloride necessitate reporters that can provide information in various contexts. For example, pH-sensitivity, sensing regime and level of quantification are all serious considerations when employing fluorescent chloride reporters. Because of this, small-molecule sensors, small molecules conjugated to macromolecules, and genetically encoded reporters have all been designed and fine-tuned to fit specific needs (Biwersi and Verkman, 1991; Kuner and Augustine, 2000; Saha et al., 2015). Furthermore, new classes of chloride-sensitive molecules continue to emerge as biologists uncover novel physiological relevance for chloride transport (Amatori et al., 2012; Collins et al., 2013; Kim et al., 2017).

In this Review, we connect chloride physiology to the chemistry of chloride reporters. First, we detail key chloride channels and transporters, selecting those whose physiological relevance has been detailed by use of fluorescent reporters and those whose uncertainties may be resolved by use of future fluorescent reporters. We then present the current repertoire of fluorescent chloride reporters by outlining their development and refinement. Finally, we hold discussions with chloride biologists to understand their experience using sensors in each biological context. By the end, we advise readers on key design challenges that must be addressed with current sensors or overcome with novel chloride sensors.

Cl⁻ has garnered attention largely due to its physiological role in synaptic inhibition in the central nervous system (CNS). Presynaptic release of GABA and glycine activate their cognate postsynaptic receptors (GABA_AR and GlyR, respectively), opening the associated Cl⁻ channel that then drives Cl⁻ down its electrochemical gradient (Fig. 1A) (Doyon et al., 2016; Kaila, 1994; Staley et al., 1995). In mature neurons of the CNS, [Cl⁻]_i is maintained at a low concentration, such that receptor activation causes Cl⁻ influx and hyperpolarization of the neuronal membrane (Fig. 2A). The Cl⁻ efflux necessary to sustain such low [Cl⁻]_i is mediated mainly by the K⁺-Cl⁻ cotransporter KCC2, which couples K⁺ ion movement along its electrochemical gradient to move Cl⁻ against its gradient (Kaila et al., 2014; Medina et al., 2014; Payne et al., 2003). In turn, the appropriate K⁺ gradient is maintained by the Na⁺/K⁺ ATPase. Additionally, the Na⁺-K⁺-Cl⁻ cotransporter (NKCC1, encoded by the gene SLC12A2) mediates Cl⁻ influx and the Na⁺-driven Cl⁻/HCO₃⁻ exchanger (NCBE, encoded by the gene SLC4A10) mediates Cl⁻ efflux (Blaesse et al., 2009; Doyon et al., 2016). Together, these transporters reflect the dynamic ionic equilibrium necessary to allow proper neuronal signaling (Fig. 1A).

KCC2 is posited to be the primary determinant of this dynamic equilibrium because neuronal cells with low levels of KCC2 exhibit excitatory GABA and glycine currents due to high [Cl⁻]_i (Klein et al., 2018; Rivera et al., 1999; Szabadics et al., 2006). Furthermore, proper balance between NKCC1 and KCC2 during development and mature neuronal KCC2 function are compromised in numerous neurological disorders, such as autism spectrum disorders, neuropathic pain, Down's syndrome and Huntington's disease (Coull et al., 2003; Dargaei et al., 2018; Deidda et al., 2015; Tyzio et al., 2014). Additionally, numerous disease-causing KCC2 variants have been identified in human epilepsies (Moore et al., 2017). Owing to the centrality of these transporters in neurological disorders, therapeutic strategies targeting KCC2 and NKCC1 at the mRNA and protein levels are fast emerging (Mahadevan and Woodin, 2016; Mahadevan et al., 2017; Tang et al., 2019).

Conversely, excitatory currents are predominantly regulated through exocytosis of synaptic vesicles containing glutamate, but there are additional important non-vesicular mechanisms, such as anion channels, that contribute to glutamate release (Kimelberg et al., 1990; Zhou and Danbolt, 2014). One example is the ubiquitously expressed volume-regulated anion channel (VRAC), which upon cell swelling releases Cl⁻ and organic osmolytes, such as glutamate, to mediate a decrease in cell volume (Osei-Owusu et al., 2018). The leucine-rich repeat-containing protein 8A (LRRC8A; also known as SWELL1, Fig. 1A) and its homologs were identified as the subunits forming the pore of VRAC (Qiu et al., 2014; Voss et al., 2014). SWELL1 is also implicated in lymphocyte development, insulin secretion in pancreatic β-cells, neuron–glia interaction and apoptotic volume decrease (Kang et al., 2018; Planells-Cases et al., 2015; Platt et al., 2017; Yang et al., 2019b).

Secretion by epithelial cells is critical to the function of several organs and is best characterized in the respiratory tract and in exocrine glands. A thin film called the airway surface liquid (ASL) protects epithelial cells that lie along the conducting airways leading to the lungs. Particles and pathogens that are inhaled are trapped in the upper mucus layer and are removed by cilia in the lower periciliary layer of the ASL (Saint-Criq and Gray, 2017). This process of mucociliary clearance requires the transcellular secretion of water to tightly regulate ASL hydration. In humans, transcellular secretion is regulated by Cl⁻ efflux across the apical membrane into the extracellular space predominantly via CFTR, with some involvement of the Ca²⁺-activated Cl⁻ channel TMEM16A (also known as ANO1) (Fig. 1B). Cl⁻ transport is the driving force for paracellular passive Na⁺ movement, and this increased salt concentration generates an osmotic force for water secretion (Fig. 2B). In parallel, CFTR-mediated secretion of HCO₃⁻ counteracts H⁺ secretion by the H⁺/K⁺ ATPase subunit ATP12A to maintain pH homeostasis of the ASL (Shah et al., 2016). In CF, defective CFTR function therefore compromises formation of the periciliary layer, and the diminution of the fluid layer leads to the accumulation of mucus and bacteria (Frizzell and Hanrahan, 2012). Interestingly, mice with dysfunctional CFTR do not develop the bacterial infections typical of CF, potentially due to expression of other anion channels or decreased expression of ATP12A (Grubb and Boucher, 1999; Guilbault et al., 2007; Shah et al., 2016). CFTR also regulates the activity of other anion cotransporters in the SLC26A family (Fig. 1B), which when dysfunctional lead to conditions, such as chondrodysplasias, Cl⁻ diarrhea and deafness through tissue-specific dysregulation of anion secretion (Alper and Sharma, 2013; Duran et al., 2010; Stewart et al., 2009).

Identified as the ‘cystic fibrosis gene’ in 1989, CFTR was posited to encode or regulate a Cl⁻ ion channel because epithelia from CF patients are impermeable to Cl⁻ (Quinton, 1983; Riordan et al., 1989; Rommens et al., 1989). In many CF patients, F508 in CFTR is absent, which prevents its proper folding and maturation in the ER and leads to its degradation (Du et al., 2005). Thus, current therapeutic efforts are focused on small molecules that promote the proper folding of mutant CFTR. As another example, Ca²⁺-activated Cl⁻ channels (CaCCs) significantly contribute to epithelial anion secretion in all cell types. The first to be molecularly identified was TMEM16A (Caputo et al., 2008; Schroeder et al., 2008; Yang et al., 2008). TMEM16A is a therapeutic target that could act as a potential compensatory mechanism when CFTR is dysfunctional. Drugs that increase intracellular Ca²⁺ can indirectly activate TMEM16A to offset low CFTR activity (Cuthbert, 2011). Further, TMEM16A is upregulated in various cancers, and its overexpression is associated with reduced survival and increased metastasis (Ayoub et al., 2010; West et al., 2004).

Acidosis occurs when extracellular pH falls below 7.35 due to increased lactic acid production and expression of carbonic anhydrase (Fig. 2C) (Jamali et al., 2015). Acidosis accompanies many pathological conditions, such as ischemic stroke and inflammation (Capurro et al., 2015; Yingjun and Xun, 2013). Following ischemic stroke, extracellular pH as low as 6.0 causes neuronal death, while during inflammation, it causes pain. Acid-sensing ion channels (ASICs) that elicit inwardly rectifying cation currents have long been implicated in these signal transductions (Wemmie et al., 2013). However, acid-sensitive outwardly rectifying (ASOR) currents have also been associated with acidosis-induced cell death, albeit by a less-understood molecular mechanism (Wang et al., 2007).

ASOR currents were first described in rat Sertoli cells and HEK-293 cells, and subsequently in diverse mammalian tissues (Auzanneau et al., 2003; Nobles et al., 2004). The biophysical characteristics of the ASOR channel have been established (Lambert and Oberwinkler, 2005; Sato-Numata et al., 2013). However, the molecular identity of the channel responsible was only very recently identified as TMEM206, also known as proton-activated Cl⁻ channel (PAC or PACC1, Fig. 1A), in two independent screens (Ullrich et al., 2019; Yang et al., 2019a). PAC is involved in acid-induced neuronal cell death and PAC-knockout mice are partially protected from ischemic brain injury (Yang et al., 2019a).

As the most abundant anion in the body, Cl⁻ also mediates the lumenal acidification of organelles. Organelle acidification is well studied in lysosomes and in secretory organelles. Lumenal pH in these organelles is crucial to their functions – post-translational modifications along the secretory pathway, ligand trafficking in endosomes and macromolecule degradation in lysosomes are all highly dependent on pH (Mindell, 2012). Proper organelle acidification has been shown to be compromised in cancer, drug resistance and autism spectrum disorders (Rivinoja et al., 2006; Ullman et al., 2018; Weisz, 2003). The protons needed to establish organelle acidity are actively pumped against their gradient by a V-ATPase (Ohkuma et al., 1982). Without the movement of ions of the appropriate charge (so-called counterions), ATPase action would generate a voltage difference that would impede further pumping. Thus, regulated anion influx, cation efflux or a combination of both are necessary for organelle acidification (Fig. 2D).

The first evidence for the role of Cl⁻ in organelle acidification came from purified lysosomes, which failed to accumulate a weakly basic dye when Cl⁻ in the solution was replaced with SO₄²⁻ (Dell'Antone, 1979). The application of a voltage-sensitive dye revealed that Cl⁻ and other anions abolished lysosomal membrane potential more significantly than K⁺ ions (Harikumar and Reeves, 1983). The main candidates posited to be responsible for this Cl⁻ flux were members of the ClC (proteins are also known designated as CLCN) family of Cl⁻ channels and transporters (Fig. 1B). For example, ClC-7-knockout mice suffer from osteopetrosis due to their inability to acidify the ruffled border in osteoclasts, which impairs bone resorption; it is therefore hypothesized that ClC-7 facilitates acidification of lysosomes, which in turn fuse with the plasma membrane to form the acidic ruffled border (Kornak et al., 2001). ClC-7 knockdown reduced lysosomal staining of LysoTracker, implying a hypoacidification of lysosomes (Graves et al., 2008). However, more accurate measurements of lysosomal pH in neurons, fibroblasts and macrophages of ClC-7-knockout mice have shown that acidification is unimpaired (Kasper et al., 2005; Kornak et al., 2001; Lange et al., 2006; Steinberg et al., 2010). It is therefore yet to be established if the Cl⁻ current provided by ClC-7 directly facilitates lysosomal acidification. Additionally, ClC-4 knockdown reduces transferrin receptor recycling due to defective endosome acidification (Mohammad-Panah et al., 2003). Hepatocyte endosomes from ClC-3-knockout mice showed low [Cl⁻] and high pH (Hara-Chikuma et al., 2005a). Similarly, endosomes in kidney proximal tubule cells of ClC-5-knockout mice failed to acidify and showed reduced [Cl⁻] (Hara-Chikuma et al., 2005b).

Organelles on the secretory pathway also undergo progressive acidification, and V-type ATPase activity and counterion flux are implicated. However, there is a severe lack of knowledge related to the underlying transporters here due to the difficulties associated with purifying intact organelles and distinguishing between transiting and resident ion channels (Judah and Thomas, 2006; Paroutis et al., 2004). Large anion conductance has been identified in purified rat liver Golgi (Nordeen et al., 2000; Thompson et al., 2002). The Golgi anion channel GOLAC may facilitate acidification, but is yet to be characterized (Fig. 1B). In addition, a screen for mutant cell lines with delayed protein transport identified the anion channel G protein-coupled receptor 89 (GPR89), also called the Golgi pH regulator (GPHR, Fig. 1B) as responsible for impaired glycosylation and defective luminal Golgi pH (Maeda et al., 2008). Finally, pancreatic β-cells of ClC-3-knockout mice displayed defective insulin exocytosis and secretory granule acidification, implicating ClC-3 in the pH regulation of insulin secretory granules (Deriy et al., 2009a).

CFTR has been implicated in pH regulation based on early evidence that the Golgi was mildly hypoacidified in epithelial cells derived from CF patients (Barasch et al., 1991); this was hypothesized to lower sialyltransferase activity, a known abnormality in CF (Haggie and Verkman, 2009b). Modest hypoacidification of endosomal pH was also observed in fibroblasts lacking CFTR or expressing ΔF508 CFTR (Biwersi and Verkman, 1994). Separately, pH measurements in alveolar macrophages of CFTR-knockout mice revealed defective lysosomal acidification, thereby leading to phagolysosomes being too alkaline to destroy bacteria (Di et al., 2006). The two most common CFTR mutations, ΔF508 and G551D, also hampered bacterial degradation and showed defective acidification of lysosomes and phagosomes (Deriy et al., 2009b). However, other studies have since challenged these claims, as the use of more reliable assays revealed that organelle acidification is independent of CFTR function (Dunn et al., 1994; Haggie and Verkman, 2007, 2009a; Lukacs et al., 1992; Seksek et al., 1996).

In short, the role of Cl⁻ in organelle acidification is surprisingly still unclear, and much of the evidence is controversial, especially with respect to CFTR. One of the sources of this ambiguity is the complicated subcellular and tissue distribution of the relevant Cl⁻ channels and transporters. Furthermore, the complete lack of appropriate Cl⁻ reporters has proved to be a major roadblock in clarifying the role of Cl⁻ in organelle acidification.

Cl⁻ plays more subcellular roles beyond its involvement in organelle acidification. In the ER, the activity of a few Cl⁻ channels is implicated in Ca²⁺ homeostasis and thereby ER stress (Fig. 2E). This was first shown in the sarcoplasmic reticulum (SR) membrane in smooth muscle cells, and subsequently also in the ER of epithelial cells (Hirota et al., 2006; Neussert et al., 2010; Pollock et al., 1998). There is compelling evidence that implicates bestrophin-1 (BEST1), a calcium Cl⁻ cotransporter mutated in vitelliform macular dystrophy (VMD) or Best disease, as having such a role. Bestrophin-1 was first characterized on the basolateral membrane of retinal pigment epithelia (RPE) and found to regulate cytosolic Cl⁻ and fluid secretion, similar to other CaCCs (Kunzelmann et al., 2007; Sun et al., 2002). However, it also localizes to the ER, which is the largest intracellular Ca²⁺ store (Fig. 1B). Ca²⁺ release from the ER is reduced in primary RPE cells lacking bestrophin-1 and mouse models of Best disease (Barro-Soria et al., 2010; Zhang et al., 2010). Another putative ER Cl⁻ channel, Cl⁻ channel CLIC-like 1 (CLCC1), is also implicated in ER stress, potentially through the disruption of Ca²⁺ homeostasis (Fig. 1B) (Jia et al., 2015).

Endosomal Cl⁻ levels are crucial for proper ligand trafficking and cargo degradation within the lysosome. The two Cl⁻ transporters implicated in these processes are ClC-5 and ClC-7, which are Cl⁻/H⁺ exchangers that facilitate Cl⁻ influx. ClC-5 is primarily expressed in kidney and intestinal epithelia where it resides in early endosomes. Mutations in ClC-5 cause Dent's disease, which is characterized by proteinuria (Devuyst et al., 1999; Günther et al., 1998; Lloyd et al., 1996). ClC-7 is ubiquitously expressed and localizes with its β-subunit Ostm1 to late endosomes and lysosomes (Kornak et al., 2001). Mice lacking ClC-7 or Ostm1 exhibit lysosomal storage disorders in neurons and renal cells, as well as irregular lysosomal morphology (Kasper et al., 2005; Lange et al., 2006). Evidence that ClC-5 and ClC-7 play physiological roles beyond supplying counterions for protons comes from mouse models where a single mutation uncouples ion transport and converts the exchangers into unidirectional transporters, designated as ClC-5^unc or ClC-7^unc. Such a transporter retains an endosomal acidification role, but lacks the role provided by exchange. Indeed, renal endosomes from ClC-5^unc mice acidify normally but exhibit proteinuria and impaired endocytosis (Novarino et al., 2010). Similarly, lysosomal pH in ClC-7^unc and ClC-7^−/− mice is unaffected, but the storage disorder or osteopetrosis phenotype of ClC-7^unc mice is as severe as in ClC-7^−/− mice, indicating that ClC-7 affects lysosomal function without altering acidification (Fig. 2F) (Weinert et al., 2010). Interestingly, when mice express a transport-deficient mutant of ClC-7, designated as ClC-7^td, that is capable of undergoing protein–protein interactions, they show normal pigmentation and less severe neurodegeneration, suggesting a role for ClC-7 interaction partners in those processes (Weinert et al., 2014).

Finally, the nuclear envelope, nucleoplasm and inner mitochondrial membrane also host members of the versatile Cl⁻ intracellular channel (CLIC) family (Fig. 1B). Interestingly, CLICs are mostly found in the soluble state, but in response to cues, such as oxidation and pH changes, insert into cellular membranes and act as anion channels under certain conditions (Domingo-Fernández et al., 2017; Shukla et al., 2009). They participate in diverse processes, such as cell cycle control, actin remodeling, vesicular pH regulation, membrane potential regulation and apoptosis, but the pathways are poorly defined (Argenzio and Moolenaar, 2016). Five out of the six CLIC homologs contain the canonical nuclear localization signal (NLS) KKYR, and have been observed to translocate to the nucleoplasm or outer nuclear membrane (Gururaja Rao et al., 2018). Despite the absence of mitochondrial targeting sequences, some isoforms localize to the inner mitochondrial membrane (Ponnalagu et al., 2016). Mitochondrial CLICs are posited to regulate apoptosis, given that Cl⁻ channel blockers inhibit reactive oxygen species (ROS)-induced and p53-mediated apoptosis (Fernández-Salas et al., 2002; Heimlich and Cidlowski, 2006; Suh et al., 2004). In addition, other anion channels under the umbrella term of the inner mitochondrial anion channel (IMAC) are involved in mitochondrial membrane potential oscillations, yet remain molecularly uncharacterized (Fig. 1B) (O'Rourke, 2007; Ponnalagu and Singh, 2017; Tomaskova and Ondrias, 2010). Thus, our current understanding of mitochondrial and nuclear Cl⁻ homeostasis is limited, due to the low abundance of these channels or their low probability of being open, as well as the inability to perform unbiased subcellular screens using current reporters. All of the aforementioned Cl⁻ channels and exchangers are described in Table S1.

The ability to measure [Cl⁻] using fluorescent reporters began with the identification of Cl⁻-sensitive dyes and the discovery that YFP is sensitive to halides. This was a major advance over previous approaches that relied on electrophysiology, which although highly precise, are not amenable for high-throughput studies or for probing intracellular Cl⁻ channels and transporters. Hence, the ability to study Cl⁻ homeostasis was greatly advanced by the discovery of Cl⁻-sensitive fluorescent probes. The first of these were synthetic quinoline-based dyes, namely 6-methoxy-N-(3-sulfopropyl) quinolinium (SPQ), N-(ethoxycarbonylmethyl)-6-methoxyquinolinium (MQAE) and 6-methoxy-N-ethylquinolinium (MEQ) (Fig. 3A) (Biwersi and Verkman, 1991; Verkman, 1990). When these dyes are excited, they collide with Cl⁻ ions in their excited states and return to the ground state through a non-radiative path, otherwise known as dynamic collisional quenching. A high [Cl⁻] thus decreases their fluorescence. These dyes are insensitive to pH and HCO₃⁻ and show microsecond response times, yet suffer from some critical limitations. First, the dyes are not ratiometric. Thus, the fluorescence readout depends on the uptake of the probe, its cellular distribution and the optical thickness of the sample (Arosio and Ratto, 2014). Second, they suffer from photobleaching (Geddes et al., 2001). Third, probes like MEQ are cell impermeable, and must be reduced to the cell-permeable diHMEQ, before being re-oxidized to the Cl⁻-sensitive form (Ashton et al., 2015). Finally, their degree of quenching is concentration dependent as they self-quench at high dye concentrations (Kaneko et al., 2004). In spite of these limitations, Cl⁻ indicators have provided useful qualitative information. For example, MEQ was used to confirm that Cl⁻ influx into neurons was caused by GABA receptor activation (Inglefield and Schwartz-Bloom, 1997), and SPQ was used to estimate intracellular Cl⁻ by flow cytometry (Pilas and Durack, 1997). In addition, the low-wavelength excitation problem has been addressed with longer wavelength acridinium-based variants, such as N-methylacridinium-9-carboxamide (MACA) and lucigenin (Fig. 3B) (Biwersi et al., 1994; Kovalchuk and Garaschuk, 2012).

Shortly after, YFP was discovered to be halide-sensitive from the observation that its pK_a was dependent on the concentration of halide or nitrate ions (Wachter and Remington, 1999). Recognizing the potential to develop this into a reporter for intracellular Cl⁻, the protein was engineered for higher halide affinity. The introduction of an additional mutation, H148Q, improved the K_d for Cl⁻ from 777 mM to 154 mM, and X-ray crystallography revealed a specific halide-binding site (Jayaraman et al., 2000; Wachter et al., 2000). YFP-H148Q is advantageous over small molecules because of its higher photostability, longer excitation wavelength, an improved ability for subcellular targeting and cellular retention. It was successfully applied to identify novel CFTR agonists in high-throughput screens (Galietta et al., 2001c). Subsequently, further enhancements of its Cl⁻ affinity into physiologically relevant regimes led to the development of the variant YFP-H148Q-I152L. This variant has a K_d of 88 mM for Cl⁻, making it suitable to measure Cl⁻ in a physiological setting (Galietta et al., 2001b). Similar mutants have been used to identify GABA receptor agonists and CaCC inhibitors in high-throughput screens (De La Fuente et al., 2008; Kruger et al., 2005; Rhoden et al., 2007).

A major limitation faced by all the aforementioned Cl⁻ sensors is the lack of a second emission wavelength that is insensitive to Cl⁻, which could be used to normalize for reporter distribution. The absence of ratiometry makes quantitative measurement highly laborious, if not impossible. To overcome this, one must conjugate a Cl⁻-sensitive moiety to one that is Cl⁻ insensitive. For small-molecule indicators, this was first achieved by chemically linking such two moieties in the synthesis of bis-DMXPQ; here, the Cl⁻-sensitive SPQ is linked to the Cl⁻-insensitive 6-aminoquinolinium (AQ) through a rigid xylyl spacer (Fig. 3A) (Jayaraman et al., 1999). Upon excitation at 365 nm, bis-DMXPQ showed Cl⁻-sensitive emission at 450 nm and a Cl⁻-insensitive emission at 565 nm. Other similar probes connect the Cl⁻-sensitive 6-methoxyquinoline (MQ) to Cl⁻-insensitive groups, such as 5-aminofluorescein (AF), dansyl (DS), and 4-amino-1,8-napthalic anhydride (NA), using various linkers to generate ratiometric reporters (Fig. 3A) (Li et al., 2012, 2014; Ma et al., 2018).

However, these ratiometric reporters suffer from significant self-quenching and a low signal-to-noise ratio. One solution is to link both moieties to a macromolecule scaffold, which was first achieved by conjugating the Cl⁻-sensitive 10,10′-bis(3-carboxypropyl)-9,9′-biacridinium dinitrate (BAC, Fig. 3B) and the Cl⁻-insensitive tetramethylrhodamine (TMR) to dextran (Sonawane et al., 2002). This had the added advantage of a stable localization in endosomes, and the sensor was used to show the role of Cl⁻ ions in endosome acidification (Sonawane et al., 2002). However, this strategy suffers from batch-to-batch variability because the degree of conjugation cannot be precisely controlled. To improve this, our sensor Clensor uses a double-stranded DNA backbone to achieve quantitative Cl⁻ mapping across the entire physiological regime of Cl⁻ by displaying BAC on one strand and a normalizing Alexa Fluor 647 (AF647) fluorophore on the other in a precise 1:1 stoichiometry (Fig. 3C) (Prakash et al., 2016; Saha et al., 2015). The DNA portion also acts as a negatively charged ligand for scavenger receptors, which, after binding, traffic the sensor along the endolysosomal pathway, after which it can be directed to other organelles by using aptamers (Dan et al., 2019; Modi et al., 2009, 2013, 2014; Narayanaswamy et al., 2019; Thekkan et al., 2018). This leads to a negligible batch-to-batch variation, which enables a quantitative measurement in different genetic backgrounds with heterogeneity coming only from the biological system (Chakraborty et al., 2016; Jani et al., 2019; Krishnan and Simmel, 2011). A Clensor variant that also includes a pH-sensitive moiety, called ChloropHore, enables the simultaneous measurement of lumenal pH and Cl⁻ to resolve populations of lysosomes in cells derived from patients with lysosome storage disorders (Fig. 3D) (Leung et al., 2019). Of course, the small-molecule conjugates still suffer from photobleaching of the photolabile BAC dye, but sequestering the reporters in the small lysosomal compartment increases its effective concentration such that the BAC signal is bright enough to take single readings. To perform dynamic imaging with any small-molecule Cl⁻ reporter, a correction factor for photobleaching would need to be derived and applied, complicating analysis.

In addition to making small-molecule Cl⁻ sensors ratiometric, there has been much effort to modify the halide-sensitive YFP to make it quantitative. The first such sensor was Clomeleon, which is comprised of cyan fluorescent protein (CFP) and a YFP variant, connected via a flexible peptide linker (Fig. 3E) (Kuner and Augustine, 2000). Because the emission of CFP at 485 nm overlaps well with YFP excitation and the two proteins are in close proximity, they can act as a fluorescence resonance energy transfer (FRET) pair. High Cl⁻ quenches YFP fluorescence, leading to low FRET efficiency. However, obtaining quantitative information from Clomeleon is extremely difficult, as its Cl⁻ affinity is highly variable and far from physiological relevance (Arosio and Ratto, 2014). For example, its K_d for Cl⁻ has been reported to be anywhere from 87 to 167 mM, while physiological cytosolic Cl⁻ ranges from 3 to 60 mM (Kuner and Augustine, 2000). The variation is likely due to the steep dependence of Cl⁻ affinity on pH, as the K_d spans two orders of magnitude within a pH range of 6 to 8 owing to the pH sensitivity of FRET efficiency and Cl⁻ binding (Bregestovski and Arosio, 2012). Especially in high Cl⁻, where fluorescence emission is low, this pH dependence leads to significant errors in the readout of Cl⁻ concentration. This has precluded its application in acidic environments or in contexts where Cl⁻ fluxes may be coupled to environmental pH changes.

The most recent Cl⁻ reporters are Clomeleon variants with optimized sensing characteristics, as well as new sensing scaffolds based on Cl⁻-sensitive molecules. The first improvement on Clomeleon significantly enhanced its affinity for Cl⁻ by replacing the YFP variant with a mutant of higher affinity (Markova et al., 2008). This sensor, called Cl-sensor, has a K_d of 30 mM, and thus is perfectly positioned for the physiological regime (Markova et al., 2008). A cell-free screen of random Clomeleon variants to study the influence of the residues in the halogen-binding site led to a double mutant possessing a Cl⁻ K_d of 21.2 mM (Grimley et al., 2013). The improved sensor SuperClomeleon uses this mutation; it also replaces CFP with the brighter donor Cerulean and incorporates a shorter linker to increase the signal-to-noise ratio (Grimley et al., 2013).

However, both Cl-sensor and SuperClomeleon suffer from pH sensitivity due to the T203Y mutation, which is essential for tight Cl⁻ binding to YFP. In fact, at low Cl⁻, variations in YFP fluorescence are as high as 15% between pH 6.8 and 7.2, with changes in the K_d being the main source of error (Rhoden et al., 2007). This leads to significant errors when applying these sensors in neurons, which exhibit sizable pH shifts (Raimondo et al., 2012). One approach to circumvent this limitation has been to simultaneously measure pH and correct for changes in K_d. ClopHensor achieved this by fusing a DsRed monomer to EGFP-T203Y through a flexible linker (Arosio et al., 2010). EGFP-T203Y is a self-ratiometric pH sensor with a pK_a of 6.8, such that the pH can be derived from the ratio of excitation between that at 488 nm and at 458 nm. Furthermore, the T203Y mutation introduces the Cl⁻-binding site in EGFP, and the Cl⁻-induced fluorescence change can be normalized to the DsRed intensity. Thus, Cl⁻ concentrations can be derived from the ratio of the emission intensity of EGFP to that of DsRed (excitation ratio 458 nm:560 nm), and corrected using the K_d value at the observed pH. However, DsRed aggregation has hindered its applicability, and has been replaced with tandem-Tomato as in ClopHensorN (Raimondo et al., 2013). DsRed has also been replaced by LSSmKate2, which has a large enough Stokes shift that it is excited in the same regime as EGFP but emits further red (Paredes et al., 2016; Sulis Sato et al., 2017). This allows quantification of pH and Cl⁻ using only two excitation wavelengths, thereby improving temporal resolution.

Meanwhile, new classes of Cl⁻-sensitive molecules are emerging, suggesting potential sensing opportunities beyond the scope of current sensors. For example, the first turn-on Cl⁻ sensors have been realized that use metal complexes. When complexed with cadmium(II) (Cd²⁺), the fluorescence of nitrobenzooxadiazole (NBD) increases with increased Cl⁻. This fluorescence enhancement is due to increased photoinduced charge transfer of the NBD-N aromatic amine into NBD, because the halide–metal interaction weakens the interaction metal ion and the NBD-N aromatic amine (Amatori et al., 2012). Based on the hypothesis that a ligand that forces the metal ion to be closer to the aliphatic amines would improve the switch-on sensor, a bis-NBD complex was realized (Fig. 4A) (Amatori et al., 2014). Another example is the interlocked squaraine rotaxane, comprising a deep-red squaraine dye coordinated by a tetralactam macrocycle (Fig. 4B) (Collins et al., 2013; Gassensmith et al., 2010). Cl⁻ reversibly translocates the macrocycle away from the squaraine, increasing its fluorescence. Additionally, turn-on Cl⁻ sensors have been developed using dyes that exhibit aggregation-induced emission. For example, the fluorescence of the dye 1⁺ increases upon aggregation in the presence of Cl⁻ and is selective over other biological anions in water and in acidic conditions (Fig. 4C) (Watt et al., 2015). While none of these probes have been applied in living cells, they could be applied in the extracellular space to visualize Cl⁻ secretion in CF.

Another recently developed class of fluorescent Cl⁻ sensors uses citrate-based biodegradable photoluminescent polymers (BPLPs) that can be processed into polymeric micelles for imaging (Kim et al., 2017). The Cl⁻-recognition moiety is a conjugated ring formed by citrate and L-cysteine that emits blue fluorescence, which is quenched with increased [Cl⁻] (Fig. 4D). However, the quenching mechanism requires protons and thus it is functional only below pH 2.3. Modifications that eliminate the charged substituents could abolish this pH dependence. Even so, their simple synthesis and sensitivity upon immobilization make BPLPs suited to diagnose CF by detecting abnormal levels of Cl⁻ in biological fluids, such as sweat, which is currently the gold standard for diagnosis of CF. Additionally, Cl⁻-selective electrodes and optodes have been engineered for fiber optical measurement of Cl⁻ (Barker et al., 1998; Brasuel et al., 2003; Pospíšilová et al., 2015).

Finally, a naturally occurring YFP that displays turn-on sensitivity to Cl⁻ seems promising. Unlike genetically encoded sensors based on GFP from Aequorea victoria (avGFP), YFP from the jellyfish Phialidium sp. (phiYFP) is self-ratiometric (Tutol et al., 2019a). Cl⁻ binding alters the pK_a of the chromophore Y66 such that the phenolic form is favored. However, the Cl⁻ dissociation constant is out of the physiological range even at acidic pH, making it impractical for deployment at physiological pH. Similarly, the tetrameric YFP from Brachiostoma lanceolatum (lanYFP) has a Cl⁻-binding pocket (Tutol et al., 2019b). Monomerization of lanYFP maintains the Cl⁻-binding pocket to form mNeonGreen, a turn-on sensor for Cl⁻. However, sensitivity to Cl⁻ depends on K143 being decarboxylated, such that the sensor cannot function above pH 5.5. Despite their current limitations, these scaffolds can be optimized to minimize their pH sensitivity and tune Cl⁻ affinities to be compatible with physiological levels. All of the aforementioned Cl⁻ reporters are described in Table S2.

Because of the reporter-specific intricacies that make them unsuitable in certain biological contexts, we engaged scientists interested in different aspects of Cl⁻ biology in a discussion on particular pain-points related to current Cl⁻ sensors, and identified what would be the desirable characteristics for upcoming Cl⁻ reporters.

One major issue discussed by all of the researchers involved in this dialogue is the level of quantification they need from their reporter. In general, genetically encoded reporters are ideal when simple data acquisition and detectable readouts are more important than quantitative information. For example, Cl-sensor has been useful to compare intracellular Cl⁻ between neurons from wild-type and mutant mice with altered KCC2 function, and to reveal the predicted overall increase in intracellular Cl⁻ (Ludwig et al., 2017). On the other hand, the halide-sensitive YFP variants were critical for the identification of therapeutics that target CFTR, TMEM16A and SLC26A3 (De La Fuente et al., 2008; Galietta et al., 2001a,c; Haggie et al., 2018; Ma et al., 2002; Namkung et al., 2011). Additionally, these variants were perfectly suited to molecularly identify TMEM16A, TMEM206 and LRRC8A as Cl⁻ channels (Caputo et al., 2008; Qiu et al., 2014; Yang et al., 2019a). However, because of the complicated pH sensitivity of these reporters, it is practically impossible to use them to detect small changes in Cl⁻ or to obtain quantitative information. This is especially an issue because Cl⁻ channels, most notably CFTR and GABA_AR, are permeable to HCO₃⁻, which affects pH fluctuations. Despite these complications, halide-sensitive YFP variants have also provided a valuable qualitative understanding of KCC2 activity and dysregulation (Boffi et al., 2018; Ludwig et al., 2017; Wimmer et al., 2015). However, to obtain quantitative information, ratiometric reporters that are insensitive to pH are needed. For example, Clensor and MEQ-TMR-dextran have been used to show that high lumenal Cl⁻, caused by the activity of ClC-7, is necessary for lysosomal degradation (Chakraborty et al., 2017; Weinert et al., 2010). In addition, pH-correctable genetically encoded reporters such as ClopHensor can be applied, but their complex analysis makes using the small-molecule conjugates more feasible.

Another set of considerations is how well the fluorescence characteristics suit the context. For example, reporters that are bright and not prone to photobleaching are needed for dynamic imaging. In these scenarios, genetically encoded sensors are better. However, even among small-molecule reporters, those conjugated through macromolecules are more capable of dynamic imaging. Often, conjugates such as MEQ-TMR-dextran show acceptable photostability and quantum yields to obtain quantitative dynamic measurements. For example, it has been used to show defective Cl⁻ accumulation in lysosomes lacking ClC-7 and in ClC-7^unc and ClC-7^td lysosomes (Weinert et al., 2010, 2014). In addition, the spectral regime occupied by the reporter is important. Most Cl⁻ reporters use the GFP channel, therefore precluding the use of the GFP channel for simultaneous sensing or protein labeling. Since ratiometric Cl⁻ sensors use one or two additional excitation wavelengths, this limitation is exacerbated. Finally, the sensing regime is a legitimate concern. Most genetically encoded reporters have a low affinity for Cl⁻, making them unusable in physiological contexts, but suitable for screens using a higher affinity anion, such as iodide, as a proxy. For physiological applications, it is instead better to use a small-molecule sensor, such as Clensor, which can accurately report on Cl⁻ across the entire physiological regime.

One final factor is the capacity of the reporter to be delivered and retained inside the desired compartment. In additional to measuring Cl⁻ fluxes across the cell membrane, researchers would also like to obtain this information for the membranes of intracellular organelles and proteoliposomes. The growing number of Cl⁻ channels and transporters found in various organelles requires new Cl⁻ reporters specifically targeting different intracellular membranes. On the other hand, proteoliposomes offer an in vitro environment to study a Cl⁻ transporter in isolation, and have been used to show the Cl⁻ channel activity of many proteins, such as TMEM16A and the CLC family (Park and MacKinnon, 2018; Ran et al., 1992; Terashima et al., 2013). In the case of whole-cell Cl⁻ measurements, the delivery and abundance of the reporters is a major concern. Although quinoline-based small-molecule reporters show sufficient cell permeability, acridinium-based dyes exhibit less permeability (Fig. 3A,B). For organelles, the reporter must have the capacity to be targeted, using a signal peptide or chemical tag, and function in the unique lumenal environment. Small-molecule conjugates, such as Clensor and BAC-TMR-dextran, are good choices in these scenarios, as they show robust localization along the endocytic pathway and are functional in acidic conditions. Finally, in the case of in vitro measurements within proteoliposomes, the most important issue is retention within the liposome. Because the efflux of small molecules from liposomes confounds dynamic measurements, especially in cases where the channel of interest has scramblase activity, such as TMEM16F, alternative non-fluorescence-based methods are often applied.

No single reporter will satisfy these needs in every biological context, which is why the field is always open to fine-tuning current reporters and developing entirely new scaffolds. Hopefully, future reporters fill the current gaps in our current sensing capabilities; we need brighter small molecules that are not prone to photobleaching, genetically encoded reporters that have a simpler way to correct for pH dependence, and sensors that are functional within organelles distinct from the endocytic pathway.

As discussed above, fluorescent chloride reporters are not a ‘one size fits all’ solution. Genetically encoded reporters have distinct advantages and disadvantages as compared with small-molecule probes, and novel classes offer new sensing capabilities. Probing each physiological role of chloride transporters requires a reporter with a unique set of characteristics tailored to the particular application (Fig. 5). For example, probing an intracellular channel presents distinct sensing considerations from a plasma membrane channel, and high-throughput screens necessitate different reporters than detailed quantitative characterizations. Rather than compiling a list of characteristics constituting an ideal sensor, we have highlighted features that make each sensor suited to a given scenario. As discussed here, the capacity to glean molecular or quantitative information on chloride transport in diverse biological contexts clearly depends on the appropriate selection of chloride reporters. By taking advantage of the unique advantages of available reporters and improving reporter designs, we can better address chloride dysregulation in diverse disease conditions. Brighter small-molecule sensors that are not prone to photobleaching will transform our ability to visualize Cl⁻ flux and thereby target it in diseases such as epilepsy and stroke. Genetically encoded sensors with simpler corrections for pH dependence will allow us to screen for channels in acidic compartments and obtain quantitative information in user-friendly stable cell lines. Finally, reporters localized to currently unexplored organelles will reveal molecular information on channels in the mitochondria, ER, and Golgi that have disease relevance but unexplained functions.

We thank Verenice Noyola for assistance in drawing schemes for figures. M.Z. thanks the NIH Chemistry-Biology Interface (CBI) Predoctoral Training Program for support. K.C. thanks the Schmidt Science Fellows Program, in partnership with the Rhodes Trust, for support.