Bidirectional Neuronal Actuation by Uncaging with Violet and Green Light

Abstract

We have developed a photochemical protecting group that enables wavelength selective uncaging using green versus violet light. Change of the exocyclic oxygen of the laser dye coumarin-102 to sulfur, gave thio-coumarin-102, a new chromophore with an absorption ratio at 503/402 nm of 37. Photolysis of thio-coumarin-102 caged γ-aminobutyric acid was found to be highly wavelength selective on neurons, with normalized electrical responses > 100-fold higher in the green versus violet channel. When partnered with coumarin-102 caged glutamate, we could use whole cell violet and green irradiation to fire and block neuronal action potentials with complete orthogonality. Localized irradiation of different dendritic segments each connected to a neuronal cell body, in concert with 3-dimenional Ca²⁺ imaging, revealed that such inputs could function independently. Chemical signaling in living cells always involves a complex balance of multiple pathways, use of (thio)-coumarin-102 caged compounds will enable arbitrarily timed flashes of green and violet light to interrogate two independent pathways simultaneously.

Multi-color photochemical orthogonality is facile for fluorescence imaging^[1]. A crucial feature of the fluorescent dyes used is that those that absorb longer wavelengths are effectively transparent at shorter wavelengths, hence each optical channel can be quantified separately. For example, rhodamine has an absorption minimum that coincides with the absorption maximum of fluorescein. Further, both dyes have minima that allow blue emitting dyes to be imaged without co-excitation of the red and green dyes.

Chromophores that can be used as photochemical protecting groups with analogous spectral properties have proved much more difficult to develop^[2]. The ideal would be that any wavelength could be applied in an arbitrarily timed order to evoke unique effects. Synthetic organic chemists are familiar with such orthogonality strictures when using ground state (i.e. thermal) reactions. The Nobel Prize winning chemist Bruce Merrifield expressed this eloquently in 1977: “An orthogonal system is defined as a set of completely independent classes of protecting groups. In a system of this kind, each class of groups can be removed in any order and in the presence of all other classes^[3].” Of course, ground state chemistry can take advantage of different reagents, solvents, and temperatures to obtain selectivity. In contrast, photochemistry can use only photons to induce selectivity. Further, the requirements for wavelength-orthogonal deprotection of biological signaling molecules are even more refined, as all reactions must be carried out under the same conditions, at ambient temperatures in physiological buffer.

Here, we describe the development of a new photochemical protecting group (called thio-coumarin-102, or “SC102”, Fig. 1a) having action/inaction absorptivities that are well matched to the green and violet optical channels (Fig. 1b). Further, coumarin-102 (C102) has an absorption maximum in the violet, making it an excellent chromatic complement to SC102 (Fig. 1b). Since photolabile (or caged) neurotransmitters have been applied widely in neuronal physiology^[4,5], we decided to use the two major neurotransmitters, glutamate (Glu) and g-aminobutyric acid (GABA), as test beds for the efficacy of SC102 and C102 in a biological context. We synthesized SC102-GABA (1) and C102-Glu (2) and examined chromatically selective photolysis of these probes individually and together on neurons using violet and green light. We meet the technical challenge of being able to use visible light for orthogonal actuation in the way Merrifield prescribed^[3].

Figure 1.

Structures and absorption spectra of C102, DEAC450 and SC102 and synthesis of SC102-GBA (1) and C102-Glu (2). (a) Structures of the C102, DEAC450 and SC102 chromophores. (b) Extinction coefficients of compounds shown in (a) in aqueous buffer at pH 7.4. (c) Reagents: a. MeMgBr. b. GABA-Boc, EDC. c. Lawesson’s reagent. d. TFA. e. L-Glu-Boc/^tBu, EDC. 1-4 were isolated as TFA salts (not shown for clarity).

The synthesis of 1 and 2 is shown in Fig. 1c. Aldehyde 5^[6] was transformed into secondary alcohol 6 with MeMgBr (51%). Carbodiimide coupling with Boc protected GABA gave 7 (88%). Treatment of this coumarin with Lawesson’s reagent in refluxing toluene allowed regioselective thionation of the exocyclic coumarin oxygen to produce thio-coumarin 8 (62%), which was deprotected quantitatively with TFA to 1. Carbodiimide coupling of 6 with protected L-glutamate gave 9 (71%), which was deprotected with TFA to give 2 (75%, after HPLC). Treatment of 7 with TFA gave C102-GABA (3) in a quantitative yield. Also, we synthesized its des-methyl analog 4 (Fig. 1c, box. See Supporting Information) and SC102-GABA-Fmoc (10), to allow absorption detection of acid release (See Supporting Information).

We compared the thermal stability of 3 and 4 in aqueous buffer at physiological pH (7.4) at RT. Consistent with previous reports^[7], des-methyl 4 was much less stable than 3 (Fig. S1a). Further, we found 1 was very stable under these conditions, with no SC102-OH (11) being detected after 4 d (Fig. S1b). Surprisingly, C102-Glu was much less stable, having a thermal half-life of about 3 days at RT at pH 7.4 (Fig. S1c). The glutamate side chain has a pKa = 4.25, whereas GABA is 4.53, we infer this subtle difference is very important for the stability of 1 and 2. Also, we verified that an acid was uncaged by irradiation of 10, photolysis yielded the starting materials, namely 6 and GABA-Fmoc (Fig. S1d).

We determined the quantum yield (ϕ) of photolysis of 2 with violet light. Classic ferrioxalate actinometry gave a value of 0.10 (Fig. S2), whereas photon counting using a calibrated light meter gave a value of 0.08 (Fig. S3). The latter method was used for 1, using 473 nm light, to give a ϕ of 0.13 (Fig. S4, see Supporting Information). We found 1 has an extinction coefficient (e) of 26,000 M⁻¹ cm⁻¹ in aqueous buffer. These values are similar to thio-DEAC^[8] and indicate 1 is photolyzed efficiently with green light (ϕ.e = 3,300). The SC102 chromophore does have significant bathochromic shift of the absorption maximum, when compared to thio-DEAC^[8] and DEAC450 (Fig. 1a), to the green region, with a maximum at 503 nm (Fig. 1b). Further, the absorption minimum is shifted to the violet region, with a minimum at 402 nm that absorbs 37x less than the peak at 503 nm. This complements nicely the maximum of 2, which has an e = 25,000 M⁻¹ cm⁻¹ at 407 nm in aqueous buffer (C102^[9] has e = 22,000 M⁻¹ cm⁻¹ in EtOH). When 1 was irradiated at 530 and 405 nm, we found that the extent of photolysis was proportional to the ratio of absorptivities at these wavelengths. No photolysis of 2 at 530 nm was detected by UPLC (Fig. S5). Finally, the rate of uncaging of 2 was estimated to be 1.2 ns (see Supporting Information).

These photochemical data for 1 and 2 indicated they might form a near-ideal duo for chromatically selective photolysis of glutamate and GABA on living neurons. All neurons contain receptors for these neurotransmitters^[10]. Glutamate opens ligand-gated ion channels (AMPA receptors) that cause cell depolarization, which, if sufficient, takes the membrane potential past the threshold for an action potential spike^[11]. GABA activates ligand-gated ion channels (GABA-A receptors) that allow Cl⁻ to pass selectively through the plasma membrane^[12], this conductance typically counters the effects of glutamate on neurons^[13]. The balance of these effects governs the production of nerve impulses in the brain ^[14].

SC102-GABA was bath-applied to hippocampal brain slices acutely isolated from mice. CA1 pyramidal cells were patch-clamped and held at −60 mV. The internal solution we used placed the driving force for Cl⁻ into the cell, and whole field photolysis (1 mW, 10 ms) with a green LED (530 nm) produced large currents (Fig. 2a). The same energy with a violet LED (405 nm) produced very little current (Fig. 2a). The pathlength (0.4 mm) and concentration of 1 (0.070 mM) in these experiments resulted in significant inner filtering of green light at the sample (>95%) when compared to the violet channel. This is because the latter is only absorbed at < 5% of the former (Fig. 1b). Thus, when this factor is taken into account, we found that the same energy for 530 nm and 405 nm produced 115-fold less response at 1 mW in the violet channel (Fig. 2a). We recorded spontaneous inhibitory post-synaptic currents (sIPSCs), and found they were not perturbed by the presence of 1 (amplitude 13.04 +/− 1.42 pA, see Fig. S6).

Figure 2.

Chromatically selective photolysis of SC102-GABA (1) and C102-Glu (2) on pyramidal neurons. The caged compounds were bath applied to different neurons. Photolysis wavelengths of 405 nm and 530 nm are indicated by violet and green colors, respectively. (a) Current traces from photolysis of 1 (0.070 mM) with green and violet light. Inset: summary of responses from 6 cells ((Mean of differences = −0.99, SD of differences = 0.0078, P < 0.0001, two-tailed t-test, n = 6). (b) Current traces from photolysis of 2 (0.1 mM) with violet and green light. (c) Voltage responses from a neuron caused by photolysis of 2 with violet and green light.

Using a similar protocol, we tested 2 on CA1 neurons. When we bath-applied 2 at 0.1 mM, photolysis with the violet LED evoked large currents, whereas currents from the green LED were not detectable (Fig. 2b). Note that this concentration of C102-Glu must produce a significant inner filtering of light at 405 nm, which “biases” the photolysis conditions away from the violet channel. Nevertheless, even doubling the green energy to a level that evoked > 1 nA from GABA uncaging (Fig. 2a) evoked no current from irradiation of 2 (Fig. 2b). Voltage responses from photolysis of 2 with violet light were graded (Fig. 2c). The highest violet energy evoked multiple action potential (AP) spikes, whereas the same green energy produced no response (Fig. 2c). The data in Fig. 2c illustrate the most crucial feature of the AP, namely that it is “all or none”^[11-13], as only when sufficient glutamatergic current reaches the cell body of the neuron can this trigger the opening of the voltage-gated sodium channels on the axon initial segment.

Having established 1 and 2 can be photolyzed with excellent chromatic selectivity, we went on to test these probes in two-color uncaging experiments. As a prequel to this, we established we could use 1 to block action potentials evoked by current injection from a patch pipette (Fig. 3a), with a small uncaging delay changing this effect (Fig. 3b). These effects are similar to our previous studies^[2b]. Next, we co-applied a solution of 1 and 2 to a patch-clamped neuron and irradiated with violet or green light alone, or with a combination of both green and violet light (Fig. 3c,d). Violet light evoked AP spikes consistently, however, when green light was applied immediately before violet irradiation, AP spikes could be blocked reliably. Interestingly, when a 20-30 ms gap was introduced between green and violet flashes, no block was achieved (Fig. 3c). However, we found that it was possible to overcome somewhat the effect of a delay by increasing the GABA dosage (Fig. 3d).

Figure 3.

Two-color photolysis of SC102-GABA (1) and C102-Glu (2) on pyramidal neurons. (a) Photolysis of 1 (0.07 mM) with green light (G, 530 nm) blocked action potential (AP) spikes generated by current injections from the patch pipette (n = 3 cells, 5 experiments, P = 0.00015, one-way ANOVA). (b) Example of a 10-ms time delay of green irradiation can change the number of AP spikes. (c) Photolysis with violet (V, 405 nm) and green light of a mixture of 1 (0.07 mM) and 2 (0.08 mM) co-applied to neurons could fire and block AP spikes (n = 3 cells, 5 experiments, P = 0.033, one-way ANOVA). A 20 ms space between green and violet flashes resulted in a failure of the AP block. But increasing the GABA dosage for a fixed delay (d) can enhance AP block.

These data are consistent with the concept that the overall balance of excitatory glutamate and inhibitory GABA inputs is crucial to triggering the AP spike^[13]. The arrival of the effects of dendritic excitation are summed temporally in the cell body, and we and others have shown temporal compression of excitation enhances glutamate input non-linearly^[2b,5]. Analogously, if GABA input arrives too early, its contribution to membrane potential dissipates due to transmitter clearance, diffusion, and re-uptake, and thus excitation dominates. In other words, the all-or-none nature of the AP makes it, in effect, a coincidence detector of the timing and amount of neurotransmitter inputs^[13], as seen in Fig. 3.

Our experiments using whole-field irradiation of patch-clamped neurons probe the global interplay of receptor activation across entire neurons as detected at the cell body. We turned to local 1-photon laser uncaging, in concert with 2-photon, laser scanning microscopy^[15], to examine how focused receptor activation is integrated by neurons in different compartments.

Neurons cultured from mice were loaded with a Ca²⁺ dye (Asante-Red) using the AM ester technique, and 2 was photolyzed on individual dendrites using a 405-nm laser (Fig. 4a). Our microscope has two pairs of galvanometers, allowing independent control of uncaging and imaging lasers^[2b,15]. Uncaging produced robust intracellular Ca²⁺ transients caused by local depolarization activating voltage-gated Ca²⁺ channels only on each targeted segment. The use of an “all optical” approach (uncaging with imaging^[16,17]) provides 3-dimensional information (i.e. time-lapse, x/y scanning) not possible with electrophysiology (Figs. 2 and 3). Next, we co-applied 1 and 2 and focused the lasers onto the cell body. The Ca²⁺ signals evoked by violet light were reduced when green light from a 532-nm laser preceded the violet with no time gap (Fig. 4b). However, if a 50 ms interval was added before the violet light, no reduction was detected (cf. Fig. 3c). Removal of the interval restored suppression, and of green light, the full signal (Fig. 4b). We examined this idea further by comparing the effects of focused irradiation of two separate dendrites emerging from the same cell (Fig. 4c). Consistent with theoretical studies^[13], we found that inhibition could be tuned locally to modulate excitation in one dendritic branch independently of any impact on the other branch (Fig. 4c). Crucially, these effects were binary, with local inhibition producing no effects on the other branch (i.e. G₁ perturbs V₁ but not V₂, and vice versa).

Figure 4.

All optical, 3-color interrogation of bidirectional dendritic signaling. Neurons were loaded with Asante Red using the AM ester method. Then, 2 was bath-applied (0.1 mM) without (a) or with (b,c) 1 (0.02 mM) to neurons. Frame-scan, 2P imaging at 1000 nm (0.8-1.4 Hz) allowed imaging of intracellular Ca²⁺ signals evoked by photolysis with 405-nm (violet, V) and 532-nm lasers (green, G). Colored dots indicate laser targets. (a) Sequential, repetitive irradiation of dendritic segments produced branch-localized signals in a reproducible manner. (b) Violet irradiation of a cell body produced excitation that could be reduced by prior green irradiation. Such inhibition failed if a 50-ms spacing was introduced between green and violet pulses. (c) Branch-specific violet irradiation produced local Ca²⁺ signals that could be suppressed by green uncaging on the same segment, but not the opposing segment (data are representative of experiments on 9 cells, when suppression of the Ca²⁺ signal was always greater than 50%).

Using two colors of light for photochemical deprotection of different molecules is a topic of long standing interest^[2,18,19], with the ideal being that either wavelength can be applied in an arbitrarily timed order to evoke unique effects^[2,20]. However, none of the previous studies focus on creating a chromophore with appropriately shifted absorption maximum and minimum in a region fully compatible with modern microscope objectives. Here, we develop a new photochemical protecting group with such properties (Fig. 1) that enables arbitrarily ordered two-color photolysis with violet and green light in complex biological systems for the first time. The caged compounds could be co-applied at similar concentrations, giving a solution with similar absorptivities for each probe. Even under such conditions it is noteworthy that wavelength selective responses could be evoked cleanly.

The use of the julolidine nucleus was crucial for our work. Widely used as part of normal coumarins, there are, to our knowledge, no reports of thio-coumarin-102 in the photochemical literature (but see^[21] for elaborately substituted variants). The absorption minimum of this new chromophore (Fig. 1b) is useful in creating a caging chromophore that is almost unresponsive to violet light (Figs. 2a, S5). While coumarin-102 has been used for photochemical release of fully protected amino acids in organic solvents^[22], no reports of its use in physiological studies have appeared. Since most confocal microscopes have 405 nm and 532 nm lasers, coumarin-102 and thio-coumarin-102 caged compounds could be used with such systems. Further, the very simplicity of the LED photolysis system we used^[20] (cost less than $2,000) puts orthogonal uncaging within easy reach of any investigator with a fluorescent microscope.

Coumarins have been used for uncaging a range of biomolecules containing acid functionalities. For example, cAMP^[23], ATP^[24], lipids^[25], DNA fragments^[26], and structurally related natural products have all been uncaged with coumarin-based photosensitive protecting groups. Such work suggests that SC102 could be applied directly to these biomolecules, if wavelength orthogonal physiological experiments were required. Such SC102-caged compounds could be partnered with one of a wide range of short wavelength caged compounds that have been developed already^{[20, 27]}.

Synthetic organic chemistry was transformed by the development of protecting groups that could be removed selectively using thermal reactions^[28]. Here we show that single-atom tuning^[21] of coumarin-102 to thio-coumarin-102 (Fig. 1) creates a new chromophore that offers orthogonal photochemical removal on a par with the best ground state (i.e. thermal) reactions. Wavelength selective photolysis of the (thio-)coumarin-102 caged neurotransmitters (Fig. 1) is highly efficacious in complex biological tissue (Figs. 2-4), suggesting our new protecting group could have wide application in many future scenarios.