Optogenetic photochemical control of designer K+ channels in mammalian neurons

Doris L. Fortin, Timothy W. Dunn, Alexis Fedorchak, Duane Allen, Rachel Montpetit, Matthew R. Banghart, Dirk Trauner, John P. Adelman, Richard H. Kramer


Currently available optogenetic tools, including microbial light-activated ion channels and transporters, are transforming systems neuroscience by enabling precise remote control of neuronal firing, but they tell us little about the role of indigenous ion channels in controlling neuronal function. Here, we employ a chemical-genetic strategy to engineer light sensitivity into several mammalian K+ channels that have different gating and modulation properties. These channels provide the means for photoregulating diverse electrophysiological functions. Photosensitivity is conferred on a channel by a tethered ligand photoswitch that contains a cysteine-reactive maleimide (M), a photoisomerizable azobenzene (A), and a quaternary ammonium (Q), a K+ channel pore blocker. Using mutagenesis, we identify the optimal extracellular cysteine attachment site where MAQ conjugation results in pore blockade when the azobenzene moiety is in the trans but not cis configuration. With this strategy, we have conferred photosensitivity on channels containing Kv1.3 subunits (which control axonal action potential repolarization), Kv3.1 subunits (which contribute to rapid-firing properties of brain neurons), Kv7.2 subunits (which underlie “M-current”), and SK2 subunits (which are Ca2+-activated K+ channels that contribute to synaptic responses). These light-regulated channels may be overexpressed in genetically targeted neurons or substituted for native channels with gene knockin technology to enable precise optopharmacological manipulation of channel function.

  • potassium
  • excitability
  • light-activated channel

k+ channels constitute a superfamily of integral membrane proteins that are crucial for many physiological processes. K+ channels regulate cellular excitability by controlling resting potential, action potential duration, the magnitude and duration of afterhyperpolarizations, and the propensity for repetitive firing (Brown 1990). K+ channels regulate neurotransmitter release from presynaptic terminals (Dodson and Forsythe 2004) and integration in postsynaptic dendrites (Johnston et al. 2003; Takagi 2000). In nonexcitable cells, including lymphocytes, smooth muscle, and glia, K+ channels participate in the control of cell volume, proliferation, and cell migration. The importance of K+ channels in normal cellular physiology is underscored by the finding that many human-inherited diseases, including cardiac arrhythmias and epilepsy, are caused by mutations in K+ channels (Abbott 2006; Kullmann 2002; Mulley et al. 2003; Sanguinetti and Spector 1997).

Neurons coexpress many different types of K+ channels, and most are heteromeric, containing two or more different types of subunits. This complexity makes it difficult to attribute unambiguously a particular physiological function to a particular type of K+ channel. In some cases, small molecules, peptide toxins, or antibodies that block a particular type of K+ channel with high specificity have been discovered, facilitating evaluation of channel function. However, high specificity implies high affinity, and such blockers are often irreversible. Channel function can also be deduced by evaluating deficits that result from genetic knockout of a specific K+ channel gene. Although this approach has provided many interesting results, developmental alterations and compensatory up- or downregulation in the expression of other channels and receptors are complicating factors in interpreting the phenotype of knockout mice (Hoffman 2008).

Here, we describe a chemical-genetic strategy that allows optopharmacological control of mammalian K+ channels. The chemical component of the system is a synthetic photoswitch molecule called MAQ. We previously used MAQ to bestow light sensitivity onto a modified Drosophila Shaker K+ channel (Banghart et al. 2004). This channel, which we named “SPARK,” contains an engineered cysteine that serves as the MAQ attachment site, along with mutations that enhance voltage-dependent activation and remove fast inactivation, causing the channels to be constitutively active at typical neuronal resting potentials. Exogenous expression of SPARK in mammalian neurons silences action potential firing. Subsequent attachment of MAQ bestows light sensitivity by an artificial light-sensitive gate that blocks or unblocks the channels under different wavelengths of light, restoring or resilencing firing. Like many other optogenetic tools (Deisseroth et al. 2006; Kramer et al. 2009; Miesenböck 2009), the SPARK conductance was superimposed on the native ionic conductances of neurons, enabling an experimenter to take charge of the cell and regulate activity by overriding its intrinsic electrophysiological properties.

The four varieties of photosensitive K+ channels described in this paper can be exogenously expressed in neurons of interest to enable optical control over different electrophysiological properties. These channels will allow an investigator to manipulate the function of the targeted channel and consequently specific properties of cells with light. With the exception of a cysteine substitution in an extracellular loop (the S5-P loop) to enable photoswitch attachment, the channels described in this paper are identical to wild-type proteins, with no apparent phenotypic differences. This is consistent with structure-function studies showing that mutations in this region do not affect channel gating or permeation. Hence, it will be possible to generate knockin mice in which each of these minimally altered proteins substitutes for its native protein counterpart. Subsequent exposure to MAQ will allow specific and precise optopharmacological knockout of these channels, providing a powerful means for evaluating their functional roles in cells and networks.


Cell culture, plasmids, and transfection.

HEK-293T cells were grown in DMEM containing 10% FBS. Cells were plated at 10–20 × 103 cells/cm2 on poly-l-lysine-coated glass coverslips and transfected using the calcium phosphate method. Recordings were performed 24–48 h after transfection. Hippocampal neurons were prepared from neonatal rats, plated at 50 × 103 cells/cm2 on poly-l-lysine-coated coverslips, and grown in MEM containing 5% FBS, 20 mM glucose, B-27 (Invitrogen), glutamine, and MITO+ Serum Extender (BD Biosciences) and transfected 7 days after plating. Recordings were performed 14–25 days after plating.

Adeno-associated virus injection.

Three- to eight-week-old C57BL/6J mice were injected with adeno-associated virus (AAV) encoding a tetracycline transactivator (tTA)-sensitive bidirectional promoter driving small-conductance Ca2+-activated K+ channel type 2 (SK2) Q339C in one direction and green fluorescent protein (GFP) in the other. A helper virus was coinjected to produce the tTA protein. Both viruses (2-μl total volume) were injected stereotaxically at coordinates corresponding to the hippocampus as per the Paxinos mouse atlas using a Quintessential Stereotaxic Injector (Stoelting, Wood Dale, IL). Animal care and experimental protocols were approved by the University of California, Berkeley (UC Berkeley), Animal Care and Use Committee.

Hippocampal slice preparation.

Hippocampal slices were prepared from mice at least 10 days postinjection. Animals were anesthetized with an intraperitoneal injection of ketamine-xylazine cocktail before being perfused with ice-cold artificial cerebrospinal fluid (aCSF; in mM: 125 NaCl, 2.5 KCl, 25 NaHCO3, 1.25 NaH2PO4, 2.0 CaCl2, 1.0 MgCl2, 12 glucose) equilibrated with 95% O2-5% CO2. Hippocampuses were removed and transferred into a slicing chamber containing sucrose-aCSF (in mM: 75 sucrose, 87 NaCl, 2.5 KCl, 21.4 NaHCO3, 1.25 NaH2PO4, 0.5 CaCl2, 7 MgCl2, 1.3 ascorbic acid, 20 glucose) equilibrated with 95% O2-5% CO2. Transverse hippocampal slices (300 μm) were cut with a Vibratome (VT1000S; Leica Instrument, Leitz, Nussloch, Germany) and transferred into a holding chamber containing aCSF and equilibrated with 95% O2-5% CO2. Slices were incubated at 34°C for 35 min before chemical modification protocol.

Photoswitch treatment.

Cultured cells were incubated at 37°C in the dark for 5 min with 1 mM DTT in cysteine-free DMEM, rinsed, and then incubated for 15 min with 25–100 μM MAQ in cysteine-free DMEM. Hippocampal slices were incubated in 2 mM tris(2-carboxyethyl)phosphine (TCEP) for 10 min in aCSF equilibrated with 95% O2-5% CO2 and then washed 3× with 10-ml aCSF. This was followed by 30-min incubation in 50 μM MAQ before slices were washed 3× with 10-ml aCSF.

Electrophysiological recordings and photocontrol of K+ currents.

For cultured cells, recordings were made in the whole cell patch-clamp configuration in extracellular solution (in mM: 138 NaCl, 1.5 KCl, 1.2 MgCl2, 5 HEPES, 1.5 CaCl2, 10 glucose; pH 7.4) using borosilicate pipettes (4–7 MΩ) with intracellular solution (in mM: 10 NaCl, 135 K-gluconate, 10 HEPES, 2 MgCl2, 2 MgATP, 1 EGTA; pH 7.4). For recording of cells transfected with Kv7.2, the intracellular solution contained (in mM) 112 K-gluconate, 10 KCl, 20 KOH, 20 HEPES, 10 EGTA, 5 Na-ATP, 0.25 Na-cAMP; pH 7.2. For cells transfected with SK2 channels, we recorded in the inside-out patch configuration using pipettes with 1- to 2-MΩ resistance. Pipette solution contained (in mM) 150 NaCl, 10 KCl, 10 HEPES, 1 MgCl2, 3 CaCl2; pH 7.4. Bath solution contained (in mM) 160 KCl, 0.5 MgCl2, 1 EGTA, 10 HEPES and 1.5 μM free CaCl2; pH 7.4. Free calcium concentration was calculated using MAXCHELATOR (http://maxchelator.stanford.edu/webmaxc/webmaxcE.htm).

For cultured cells, illumination was provided using a xenon lamp (175 W) with band-pass filters (379 ± 17 and 500 ± 8 nm). At the back of the objective, light output was 6 mW/cm2 for 380-nm light and 2 mW/cm2 for 500-nm light. When measured through a ×20 objective and normalized to the focal area at the specimen plane, light output was 1.6 and 0.4 mW/mm2 for the 380- and 500-nm light, respectively.

CA1 pyramidal cells in hippocampal slices were visualized with modified DOT imaging contrast optics (Axioskop FS2; Carl Zeiss, Thornwood, NY) and a charge-coupled device (CCD) camera, and recordings were made in the whole cell patch-clamp configuration. Patch electrodes (2∼3 MΩ) were filled with a solution containing (in mM) 135 K-gluconate, 8 NaCl, 1 MgCl2, 10 HEPES, 4 MgATP, 0.3 Na2GTP, and 10 phosphocreatine; pH 7.26. Electrophysiological records were filtered at 5 kHz and sampled at 20 kHz. The input resistance was determined from an ∼30-pA (500-ms) hyperpolarizing current-injection pulse following each event. All recordings were performed at room temperature, and we evaluated only those cells with a stable series resistance (80% compensated) and a resting membrane potential between −75 and −55 mV. For recording small conductance Ca2+-activated K+ (IsK) tail currents, cells were depolarized from −55 to +20 mV for 100 ms followed by a return to −55 mV (Hammond et al. 2006).

For current-clamp recordings, we only used cells with stable input resistance and resting potential (−70 to −50 mV), and we rejected recordings where the series resistance changed by >20%. Glass micropipettes filled with saline were positioned in the stratum radiatum to stimulate presynaptic axons via a stimulus isolation unit (A-M Systems, Sequim, WA). For excitatory postsynaptic potentials (EPSP) measurements, a bias current was applied to maintain the membrane potential at −60 mV. SR-95531 (2 μM) and CGP-55845 (1 μM) were added to reduce GABAA and GABAB contributions, respectively.

For neurons in brain slices, fluorescence illumination was generated by an X-Cite 120 light source, and exposures were ∼2 s in duration. Slices were alternately exposed to 500-nm light to induce block of SK2 channels and then exposed to 380-nm light to unblock the channels, at least three consecutive times.


Design of light-regulated K+ channels.

The photoswitch used for regulating K+ channels (Fig. 1A) contains a maleimide (M) that tethers the molecule to a genetically engineered cysteine, a photosensitive azobenzene linker (A), and a pore-blocking quaternary ammonium group (Q). In darkness, MAQ is in the low energy trans configuration, but 380-nm light photoisomerizes the molecule to the cis configuration. The cis form of MAQ spontaneously relaxes to the trans form over several minutes, but in 450- to 520-nm light relaxation is accelerated. Given sufficient light intensity, relaxation can occur within milliseconds after light exposure. Hence, after conjugating MAQ to the channel, the blocker can be toggled in and out of the pore by photoswitching the azobenzene with 500- and 380-nm light, respectively, allowing regulation of ion conduction (Fig. 1B). Photocontrol of channel activity can be achieved with great temporal and spatial precision, enabling acute and reversible blockade of K+ channel function.

Fig. 1.

Optopharmacological strategy to control the activity of K+ channels. A: the synthetic photoswitch MAQ consists of a cysteine-reactive maleimide, a photoisomerizable azobenzene linker, and a pore-blocking quaternary ammonium (QA; blue). MAQ undergoes trans-to-cis isomerization on illumination with 380-nm light. Exposure to 500-nm light, or prolonged time in darkness, returns the molecule to the trans configuration. B: MAQ covalently attaches to a genetically engineered cysteine located at the appropriate distance from the pore of a K+ channel. C: sequence alignment of different K+ channels. We engineered cysteines at position equivalent to E422C in Kv1.3 (P374C), Kv3.1 (E380C), Kv7.2 (KCNQ; E257C), and small-conductance Ca2+-activated K+ channel type 2 (SK2; Q339C). S5 and S6 denote the 5th and 6th transmembrane domains. P denotes the K+ selectivity filter.

Light-sensitive Kv1.3.

Members of the mammalian Kv1 family of voltage-gated K+ channels are highly homologous to Drosophila Shaker K+ channels, so we began by engineering a photosensitive version of Kv1.3. In neurons, Kv1 family channels are targeted to axons (Gu et al. 2003) where they contribute to membrane repolarization during the falling phase of the action potential (Fadool et al. 2004). Kv1.3-containing channels are particularly important for maintaining tonic firing during sustained depolarization (Kupper et al. 2002) and regulate the differentiation of neuronal progenitors into neurons, making the channels an intriguing drug target for neurodegenerative disease (Peng and Huss 2010; Wang et al. 2010). Kv1.3 also plays important physiological and pathological roles in many nonneuronal cell types, including T cell lymphocytes, where it has become an important target for potential therapeutic modulation of the immune system (Beeton et al. 2006), and platelets, where it is the only voltage-gated K+ channel expressed (McCloskey et al. 2010).

Taking advantage of the extensive similarity between the amino acid sequence of Shaker and Kv1.3 (Fig. 1C), we introduced a cysteine in the extracellular loop (P374C) at the position equivalent to that in SPARK to bestow photosensitivity onto Kv1.3. Because the affinity of Kv1 family channels for the pore-blocker TEA is low, we introduced an additional point mutation (H401Y) that increases the affinity for TEA to the micromolar range (Kavanaugh et al. 1991). The resulting channel was expressed in HEK-293T cells, and photosensitization was tested by recording whole cell currents after 15 min of MAQ treatment. The mutated Kv1.3 channels indeed did become sensitive to light, with 500-nm light blocking ∼50% of the voltage-activated current and 380-nm light relieving channel blockade (Fig. 2, A and C). Currents through photosensitized Kv1.3 could be repeatedly blocked by 500-nm light and unblocked by 380-nm light (Fig. 2A). For cells expressing wild-type Kv1.3 channels, application of MAQ resulted in no detectable photosensitization. Pretreatment of cells expressing the photosensitive Kv1.3 with another cysteine-reactive reagent, MTSET, prevented subsequent photosensitization by MAQ (Fig. 2, B and C), confirming that MAQ photosensitization is mediated by cysteine conjugation.

Fig. 2.

Engineering a photosensitive Kv1.3. A: introduction of P374C and H401Y and subsequent treatment with MAQ photosensitizes Kv1.3. Whole cell recording from a MAQ-treated HEK-293T cell expressing Kv1.3 P374C and H401Y. The cell was held at −60 mV and stepped to +20 mV every 5 s. The resulting current was measured and plotted over time and wavelength changes. Illumination with 500-nm light blocks the channel, reducing ion flow, whereas illumination with 380-nm light unblocks the channel. B: when the genetically engineered cysteine is protected by treatment with cysteine-reactive reagent MTSET before treatment with MAQ, there is no photosensitization and no regulation of current with light. Kv1.3 channels exhibit cumulative inactivation with very slow recovery (Kupper et al. 2002) that may account for the small decline in currents observed in A and B. C: fraction of current regulated with light in cells expressing Kv1.3 P374C and H401Y treated with MAQ (48.2 ± 16.2%) and pretreated with MTSET to protect the cysteine and prevent attachment of MAQ (0.05 ± 5.5%). The fraction of current photoregulated is defined as the difference between the current under 380- and 500-nm light divided by the amount of current in 380 nm. Data represent average ± SD (n = 6–13).

Light-sensitive Kv7.2.

Heteromultimeric channels composed of primarily Kv7.2 and Kv7.3 subunits underlie the M-current, a slow voltage-gated current in neurons and cardiac muscle that is activated at membrane potentials below action potential threshold (Brown and Passmore 2009). M-current is named as such because it is inhibited by muscarinic receptors, providing a potent mechanism for neurotransmitter modulation of excitability. Mutations in Kv7.2 cause channelopathies in humans, including inherited juvenile epilepsy and peripheral nerve hyperexcitability (Maljevic et al. 2008).

To generate light-regulated Kv7.2-containing channels, we again used MAQ in conjunction with a cysteine-scanning approach. Kv7.2 E257C channels showed the greatest degree of photosensitization, with 500-nm light blocking ∼34% of the current elicited during at step from −70 to −40 mV (Fig. 3A) and a similar reduction in the tail current on repolarization to −70 mV (Fig. 3B). Introduction of a cysteine at position K255 and G256 completely eliminated photosensitization, whereas cysteine substitution at N258 resulted in no current on the cell surface. Interestingly, cysteine substitution at D259 appears to result in “reverse” photoswitching, with 380-nm light blocking the channel, although the effects of light were not as pronounced as with E257C (Fig. 3C). These data underscore the need for precise positioning of the MAQ attachment site to achieve maximal photosensitization of the engineered channel.

Fig. 3.

Photosensitization of Kv7.2 channels. A: engineering of a cysteine in Kv7.2 (E257C) and subsequent treatment with MAQ enables photoregulation of the channel. Whole cell recording from a cell expressing Kv7.2 E257C treated with MAQ. Current was elicited by stepping to −40 mV from a holding voltage of −70 mV. Illumination with 500-nm light blocks the channel, whereas the 380-nm light unblocks the channel. B: tail current through Kv7.2 E257C is also photoregulated after treatment with MAQ. Voltage protocol to elicit tail current is the same as in A. C: fraction of current regulated in cells expressing Kv7.2 K255C (−1.4 ± 6.2%), G256C (−0.7 ± 1.1%), E257C (33.8 ± 8.3%), and G259C (−14.2 ± 18.3%). There was no detectable Kv7.2 current in cells expressing the mutant N258C (N/A). Currents were elicited as in A. The position of the engineered cysteine is crucial for photosensitization. Data represent averages ± SD (n = 3–9).

Light-sensitive Kv3.1.

Kv3.1 subunits are widely expressed, particularly in neurons that fire action potentials at high frequencies, including those in the auditory brainstem and cerebellum (Gan and Kaczmarek 1998). Kv3.1-containing channels deactivate very rapidly after a depolarizing pulse, an important feature thought to enable action potentials to occur in rapid succession. Optical manipulation of Kv3.1 channels would aid in exploring the function of these channels in neuron and other cell types where they are expressed. K+ current through homomeric Kv3.1 channels containing a cysteine at amino acid 380 (E380C) and treated with MAQ were photosensitive at all voltages, but the fraction of current photoregulated was largest at membrane potential more negative than 0 mV (Fig. 4A). Voltage-gated currents elicited by stepping to −10 mV were reduced by ∼70% after switching from 380- to 500-nm light (Fig. 4B). As expected, wild-type channels lacking a genetically engineered cysteine were insensitive to light even after treatment with MAQ (Fig. 4B). Currents through photosensitive Kv3.1 are repetitively blocked by 500-nm light with little decrement in the fraction of current regulated and no apparent photobleaching of the photoswitch (Fig. 4, C and D).

Fig. 4.

Photosensitization of Kv3 family channels. A: steady-state current voltage curves from a voltage-clamped HEK-293 cell expressing Kv3.1 E380C. Currents were elicited by stepping to the indicated voltage from a holding potential of −70 mV and measured in 380-nm (open squares; violet) or 500-nm (closed squares; green) light after treatment with MAQ. B: fraction of current photoswitched after MAQ treatment of cells expressing Kv3.1 E380C (67.0 ± 6.5%) or wild-type Kv3.1 (9.2 ± 3.5%). Currents were elicited by stepping from a resting potential of −70 to −10 mV. Data represent average ± SD (n = 3–9). C: representative currents elicited by stepping from a holding potential of −70 to −10 mV in alternating 500- (green), 380- (violet), and 500-nm (green) light. Currents through photosensitive Kv3.1 are repetitively blocked by 500-nm light with little decrement in the fraction of current regulated and no apparent photobleaching of the photoswitch. D: currents were elicited and measured every 2 s as in C and plotted over time and wavelength changes. Illumination with 500-nm light (green) blocks the channel, whereas illumination with 380-nm light (violet) unblocks the channel. E: channels expressed in HeLa cells composed of up to 2 (heteromers) or 4 cysteine (homomers)-containing subunits are photosensitized to a similar extent by MAQ treatment (homomeric Kv3.4, 5.540 ± 3.6%; homomeric Kv3.4 D420C, 57.2 ± 15.9%; heteromeric Kv3.4 D420C + Kv3.1, 49.8 ± 6.0%; heteromeric Kv3.1 E380C, 75.6 ± 24.2%; heteromeric Kv3.1 E380C + Kv3.4, 59.1 ± 12.6%. Data represent average ± SD, n = 4–5 for each combination).

In fast-spiking neurons, Kv3.1 subunits can heteromultimerize with Kv3.4 subunits, resulting in a channel that repolarizes spikes more rapidly (Baranauskas et al. 2003). Since MAQ should interact with its target channel like any standard quaternary ammonium compound, that is by interacting with a single binding site (25–27), we predicted that introducing a cysteine at position D420 in Kv3.4, the equivalent position to E380 in Kv3.1, would enable MAQ to photosensitize not only homomeric Kv3.4 channels, but also heteromeric Kv3.4/Kv3.1 channels (Fig. 4D). Indeed, the photosensitivity was the same whether the channels comprised 4 identical cysteine-containing subunits (i.e., homomeric mutant channels) or a mixture of cysteine-containing and wild-type subunits (i.e., heteromeric channels with only 1 subunit type mutated). Hence, photosensitization is all-or-none, rather than varying incrementally with the number of tethered MAQ molecules, consistent with the channel possessing a single TEA binding site that mediates photosensitive pore blockade (Blaustein et al. 2000; Heginbotham and MacKinnon 1992; Hille 1967).

To confirm that light-sensitive Kv3.1 can alter activity of neurons, we transfected cultured hippocampal neurons with Kv3.1 E380C. Current-clamp recordings obtained after MAQ treatment from transfected neurons coexpressing GFP showed that light exposure had a significant effect on repetitive firing elicited by depolarizing current injection. For example, the neuron in Fig. 5A fired for almost twice as long when the channels were unblocked in 380-nm light than when they were blocked in 500-nm light, and six out of seven transfected neurons showed similar results (Fig. 5B). Hence, hippocampal neurons expressing Kv3.1 E380C channels fire more action potentials when the channels are in the unblocked state. Indeed, action potentials repolarized more quickly with a larger afterhyperpolarization when the channels were unblocked (Fig. 5C), presumably enabling faster recovery of voltage-gated Na+ channels from inactivation, which promotes repetitive firing.

Fig. 5.

Photosensitization of Kv3.1 channels in neurons enable control of repetitive firing. A: photoregulation enables control of repetitive firing in a neuron expressing Kv3.1 E380C after MAQ treatment. The number of action potentials fired in response to sustained current injection (iinj) is higher under 380-nm (violet) than 500-nm (green) light. Light exposure did not affect resting membrane potential. B: photoregulation of Kv3.1 E380C enables control of action potential firing in transfected neurons. Six of seven cells expressing Kv3.1 E380C and treated with MAQ fire more action potentials in 380-nm light in response to a depolarizing current injection. C: photoregulation of Kv3.1 E380C changes the shape of the action potential in transfected neurons treated with MAQ. Single action potentials were generated by short (10 ms) current injection. The afterhyperpolarization is more pronounced under 380-nm (violet) than in 500-nm (green) light.

Photosensitization of SK2.

SK channels are small conductance Ca2+-activated K+ channels. In contrast to large-conductance potassium (BK) channels for which gating is sensitive to both Ca2+ and voltage, SK channels are sensitive only to Ca2+, which activates the channels by binding to auxiliary calmodulin subunits (Maylie et al. 2004). There are four members of the SK channel family, and three of them, SK1, SK2, and SK3, are expressed in central nervous system neurons (Stocker et al. 2000). In CA1 hippocampal neurons, SK2 channels modulate EPSPs and contribute to long-term potentiation (LTP) of synaptic transmission (Lin et al. 2008; Ngo-Anh et al. 2005).

Based on amino acid sequence alignment with the other photosensitive K+ channels engineered for these studies, we chose position Q339 as the cysteine attachment site for SK2 channels (Fig. 1C). Inside-out membrane patches were obtained from cells expressing SK2 Q339C and treated with MAQ. Channels were activated by applying a solution containing 1.5 μM free Ca2+. The current elicited first in darkness was increased ∼2.5-fold in 380-nm light, consistent with relief of channel block as the MAQ was converted from the trans to the cis configuration. The Ca2+-elicited current was repeatedly attenuated and restored by illuminating with 500 and 380 nm, respectively (Fig. 6A).

Fig. 6.

Photosensitization of SK2 channels. A: inside-out patch recording from a HEK-293T cell expressing SK2 Q339C and treated with MAQ. Ca2+-activated currents were induced with a solution containing 1.5 μM free Ca2+ buffered with EGTA. Ca2+-activated current is larger in 380- than 500-nm light. B: whole cell recording showing light-sensitive currents during a depolarizing voltage ramp (−80 to +80 mV). The pipette contained 1 μM Ca2+ to ensure maintained activation of SK2 channels. Currents were maximal in 380-nm light and reduced in 500-nm light. Apamin (0.5 μM) further inhibited the current, indicating that blockade by trans MAQ was >60% complete.

Light-regulated SK2 current could also be observed under whole cell patch-clamp (Fig. 6B). For these experiments, we used a patch pipette that contained 1 μM free Ca2+ and applied a voltage ramp from −80 to +80 mV. The Ca2+-elicited current was decreased over the entire voltage range by switching from 380- to 500-nm light. We then completely blocked the channels with the specific SK channel inhibitor apamin (Fig. 6B). This allowed us to estimate that >60% of the SK2-mediated current is suppressed in 500-nm light, consistent with our using inside-out patch results (Fig. 6B).

To test whether light-regulated SK channels can be generated in situ, we used a virus to express SK2 Q339C subunits in hippocampal CA1 neurons. AAV carrying the genes encoding SK2 Q339C and GFP was introduced via transcranial stereotaxic injection into 3- to 5-wk-old mice. No behavioral changes were noted in mice post-AAV injection. Hippocampal slices were obtained >10 days postinjection and treated with MAQ. Cells expressing SK2 Q339C were identified based on coexpression of GFP. Resting membrane properties of fluorescent SK2 Q339C-expressing cells were similar to those of nonexpressing cells (membrane potential = −63 ± 2 mV for control; −62 ± 1 mV for fluorescent; input resistance (Rin) = 201 ± 20 MΩ for control; 208 ± 25 MΩ for fluorescent; access resistance (Ra) = 21 ± 2 MΩ for control; 22 ± 2 MΩ for fluorescent; n = 12–14 for each). Voltage-clamp recordings of SK2 Q339C-expressing neurons showed photosensitive SK channel activity. In 380-nm light, when MAQ is in its cis configuration and SK2 Q339C channels are relieved from block, we observed a slowly relaxing outward tail current after a depolarizing pulse to +20 mV, characteristic of IsK (Bond et al. 2004). This tail current was reduced after illumination with 500-nm light (Fig. 7, A and C). The amount of photoregulation varied considerably in different cells; exposure to 500-nm light reduced the amplitude of IsK measured at 100 ms by ∼30% on average (n = 12), presumably because the photoswitch-ready channels were expressed against a background of native, light-insensitive channels. The IsK recorded from nonfluorescent (and therefore not expressing SK2 Q339C) CA1 neurons from the same slice was not altered by exposure to 380- or 500-nm light (Fig. 7, B and C). Hence, MAQ selectively photosensitizes cysteine-substituted SK2 channels, and light can alter currents carried by these channels in an acute brain slice preparation.

Fig. 7.

Photoregulation of SK2 channels in hippocampal slices. A: whole cell recordings from hippocampal CA1 neurons expressing SK2 Q339C in a MAQ-treated slice. A depolarizing pulse (top) elicited tail currents (bottom) that were larger in 380-nm (green) than in 500-nm light (violet). B: light has no effect on currents measured from a nonexpressing cell in a MAQ-treated slice. C: summary data for photocontrol of the SK-mediated tail current. On average, 500-nm light reduced the tail current measured 100 ms after the pulse by 30 ± 19% (n = 12; average ± SD) for cells expressing SK2 Q339C and by 3 ± 9% (n = 14; average ± SD) for nonexpressing cells. *Significant difference (P < 0.01; paired t-test). D: average excitatory postsynaptic potentials (EPSP) waveform from a SK2 Q339C-expressing neuron in 500-nm light (green), which blocks the SK2 channels, and 380-nm light (violet), which unblocks the channels. E: repeated changes in peak EPSP amplitude with 500- and 380-nm light.

We next asked whether photoregulating SK2 channels could affect synaptic communication in a hippocampal slice. SK2-containing channels are found in dendritic spines of CA1 neurons (Ngo-Anh et al. 2005). Electrical stimulation of presynaptic Shaffer collateral axons causes glutamate release and postsynaptic activation of Ca2+-permeant NMDA receptors in the spines. The resulting Ca2+ influx activates SK channels, providing a repolarizing conductance that reduces the glutamate receptor-mediated depolarization (Ngo-Anh et al. 2005). Consistent with this, pharmacological blockade of synaptic SK channels with apamin causes an increase in EPSP amplitude.

EPSP amplitude was regulated by light in some SK2 Q339C-expressing neurons, but the magnitude of the effect was variable. This is not surprising given that these cells have a mixture of channels possessing mutant and wild-type SK2 subunits. Figure 7D shows a particularly clear example in which 500-nm light resulted in a doubling of EPSP amplitude, consistent with the expected transition from the MAQ-unblocked to MAQ-blocked state. EPSP amplitude could be repeatedly increased and decreased with 500- and 380-nm light, respectively (Fig. 7E). The control of EPSP size with light demonstrates that the SK2 Q339C-containing channels are incorporated in the postsynaptic membrane and function in a similar manner as wild-type SK2 channels.


Engineering “designer” light-activated K+ channels.

A large number of toxins and small molecules have been found that selectively block different K+ channels. However, there are still channel types with no selective blockers, and it remains difficult to find blockers that unambiguously discriminate between different members of a given K+ channel family. By genetically engineering a cysteine attachment site into a single K+ channel subunit, we are generating photoswitch-ready channels with built-in pharmacological specificity. Subsequent treatment of engineered channels with MAQ allows specific, precise, and acute photopharmacological regulation of the channels.

Photoisomerization of MAQ from trans to cis shortens the molecule from ∼20 to ∼13 Å. This shortening will change the position of the quaternary ammonium moiety with respect to its binding site in the pore. To maximize photocontrol of current through the targeted K+ channel, MAQ must be properly positioned on the channel protein such that it can reach the pore in the trans but not the cis configuration. When the SPARK channel was engineered, the intramolecular distances between the Shaker channel pore and several extracellular residues had already been estimated (Blaustein et al. 2000), making the selection of an optimal cysteine attachment site (E422C) a straightforward task (Banghart et al. 2004). Although the K+ channels used here share considerable sequence homology with Shaker, identifying the optimal attachment position required some cysteine scanning. Not surprisingly, photosensitization is acutely dependent on the position of the engineered cysteine. For example, a single amino acid shift in Kv7.2 completely eliminates photosensitization. The strict positional requirement has several possible explanations. Some cysteine positions may tether MAQ too far to enable the quaternary ammonium to reach the pore. Others may tether MAQ so close to the pore that photoisomerization to the cis configuration fails to reduce the effective concentration of quaternary ammonium sufficiently to relieve block. Finally, at some positions, the genetically engineered cysteine may simply be inaccessible to covalent modification. More structural information about the channels is needed to distinguish between these possibilities.

The degree of MAQ photoswitching is also dependent on the affinity for blockade by external quaternary ammonium, which varies among different K+ channels. Once MAQ becomes tethered, the QA moiety is constrained in space to the volume of a hemisphere with a radius of ∼20 Å, which results in an effective concentration in the tens of millimolar range (Banghart et al. 2004; Kramer and Karpen 1998). However, the affinity of some K+ channels for external QA is so low (Kd > 100 mM) that the channels remain unblocked, even when the tethered MAQ is in the trans state. The location of the external QA binding site is conserved among many K+ channels (position 449 in Shaker), and particular amino acids that determine high or low QA affinity have been characterized (Heginbotham and MacKinnon 1992; Kavanaugh et al. 1991). By selecting the appropriate amino acid for this site, it may be possible to “tune” the extent of block by trans MAQ as well as the extent of unblock after photoisomerization to the cis configuration. In this manner, we have succeeded in altering the extent of photoswitching in the Shaker channel and its mammalian homolog Kv1.3. However, our attempts to increase the QA affinity of Kv4.2 by mutating this site (V489Y) prevented surface expression of the channel. Thus, although it is likely that more light-regulated K+ channel subunits eventually will be added to the list of those we have already engineered, it is difficult to predict which channels will be the most successful.

Applications of designer light-activated K+ channels.

The various K+ channels used in this study differ in their cell-type distribution, gating properties, kinetics, and modulatory control. The different physiological functions of K+ channels suggest that overexpression of light-sensitive versions could be used not only to suppress action potentials with light, but also to fine-tune different aspects of cellular electrophysiology. For example, light-regulated Kv7.2 channels might be useful for controlling resting potential, light-regulated Kv3.1 for controlling accommodation, and light-regulated SK2 for controlling the size of EPSPs.

K+ channel optopharmacology enables selective spatial manipulation of channel function within parts of a single neuron. For example, local illumination can improve our understanding of the role of particular K+ channels in dendritic integration in pyramidal neurons. Photoregulation can also be useful for probing ion channel function in groups of neurons, such as tonotopically distributed neurons in the auditory brainstem, where K+ channel gradients play an important role in frequency tuning (Parameshwaran et al. 2001).

Genetic knockouts have provided important information about ion channel function, but elimination of one type of ion channel can sometimes result in lethality or developmental abnormalities. For example, genetic knockout of Kv1.3 in mice results in alterations in the size and shape of olfactory bulb glomeruli, leading to enhancement of the sensitivity to odorants in vivo (Fadool et al. 2004). Genetic knockout of certain K+ channels can induce compensatory changes in the expression levels of other ion channels. For example, knockout of both Kv3.1 and Kv3.3 disrupts normal thalamocortical oscillations (Espinosa et al. 2008), but eliminating either subunit individually has little effect, possibly resulting from compensatory upregulation of the other subunit or changes in the expression level of entirely different channels. Our approach provides the means for unambiguously revealing the role of a given subunit. In principle, replacement of the native Kv3.1 with the cysteine-containing mutant in knockin mice should provide a “clean” method for acute and reversible optopharmacological control of the channel, without inducing compensatory changes in the expression levels of other proteins.

The optopharmacological strategy described here requires an exogenous synthetic photoswitch molecule. The photoswitch is modular in nature, consisting of a reactive moiety, a photoisomerizable linker and a ligand that binds to channel or receptor. This modularity allows flexibility in the design of each functional group, yielding a combinatorial toolkit for the optopharmacological regulation of protein function. The increasing availability of structural and pharmacological data suggests that this optopharmacology strategy may be extended to other types of channels and receptors, opening a new window into understanding the function of these signaling proteins in neurons and other cells.


This work was supported by the National Eye Institute (Grant EY-018957 to R. H. Kramer), the National Institute of Mental Health (Grant MH-088484 to R. H. Kramer), and the UC Berkeley Nanomedicine Development Center for Optical Control of Biological Function (National Eye Institute Grant PN2-EY-018241).


No conflicts of interest, financial or otherwise, are declared by the author(s).


We thank S. Szobota, J. Patti, and H. Janovjak for useful discussion and A. Desai for purification of AAV. E. Isacoff (UC Berkeley), B. Rudy (New York University), A. Tzingounis (University of Connecticut), and T. Jentsch (Leibniz-Institute, Berlin) generously provided plasmids.


View Abstract