2005 Nature Publishing Group http://www.nature.com/naturemethodsARTICLESRapid neurotransmitter uncaging in spatiallydefined patternsShy Shoham1,2,4,5, Daniel H O’Connor1,2,5, Dmitry V Sarkisov2,3 & Samuel S-H Wang1,2Light-sensitive ‘caged’ molecules provide a means of rapidly andnoninvasively manipulating biochemical signals with submicronspatial resolution. Here we describe a new optical system forrapid uncaging in arbitrary patterns to emulate complex neuralactivity. This system uses TeO2 acousto-optical deflectors tosteer an ultraviolet beam rapidly and can uncage at over 20,000locations per second. The uncaging beam is projected into thefocal plane of a two-photon microscope, allowing us to combinepatterned uncaging with imaging and electrophysiology. Byphotolyzing caged neurotransmitter in brain slices we cangenerate precise, complex activity patterns for dendriticintegration. The method can also be used to activate manypresynaptic neurons at once. Patterned uncaging opensnew vistas in the study of signal integration and plasticityin neuronal circuits and other biological systems.Neurons integrate synaptic signals from many thousands ofinputs. Understanding the resulting information processing is acentral theme in experimental and computational neuroscience.Multiple inputs can interact to amplify1, attenuate2 or modulateone another. Thus, results obtained by activating one or a fewinputs at a time, as is commonly done in brain slice experiments,do not fully capture the complexity of signal processing bysingle neurons.This potential complexity of signal integration suggests the needto manipulate biochemical and electrical events across dendritesand in multiple neurons with fine spatial and temporal resolution.An attractive option for performing such manipulations is the useof optical approaches in the form of photochemical release of cagedneurotransmitters and second messengers3,4. In optical uncaging, abiologically active molecule is inactivated through covalent attachment of a caging group, is introduced into tissue and is thenconverted to its active form by a flash of light. A large variety ofcaged compounds is available3, including agonists for neurotransmitter receptors such as glutamate, GABA, acetylcholine andbiogenic amines and intracellular messengers such as calcium(in which a chelator of calcium is the molecule caged) andinositol-1,4,5-trisphosphate (IP3). Uncaging approaches open thepossibility of using light to probe semi-intact tissue noninvasively.Caged compounds are useful whenever control of cellularbiochemistry is needed on subsecond time scales. In addition toapplications to neurophysiology, caged compounds have beenuseful in studying other biological problems requiring comparabletime resolution such as secretion5, muscle activation6, fertilization7and nuclear signaling8. An intriguing recent advance is the development of a caged inhibitor of protein synthesis9.Uncaging light pulses are usually delivered to a single fixedlocation. To overcome this limitation we have developed a systemthat rapidly deflects and modulates the uncaging beam usingacousto-optical deflectors (AODs). AODs allow extremely rapidaccess to many locations in tissue and have been used previouslyin some commercial (Noran Inc.) or custom-built10,11 imagingsystems. Our system uncages at over 20,000 locations persecond, considerably faster than older uncaging systems thatsteer beams with modified galvanometer mirrors and mechanicalshutters12,13. Here we present the basic physical characteristicsof this system and demonstrate its application to several problemsin neuroscience.RESULTSGeneral system designOur system design is illustrated in Figure 1a and described in detailin the Supplementary Note online. The ultraviolet light source isa frequency tripled Nd:YVO4 laser (DPSS Corp.; 50–60 ns,l ¼ 355 nm pulses at a 100-kHz repetition rate with averagepower 4400 mW) whose beam is expanded threefold in diameterand directed through two crosswise-oriented tellurium dioxide(TeO2) AODs and a two-lens 1:1 telescope into the optical pathof a two-photon microscope14. The AOD assembly, lenses and aniris are spaced at intervals approximately equal to the lens focallength f, thus forming a 4-f system. Such an arrangement transfers abeam that pivots around the AOD axis to a pivot point at the backaperture of the objective, thus allowing scanning of the beam in thefocal plane15. The laser output is gated using its Q-switchingcontrol so that output pulses are emitted only after the AOD hassettled at a new value. Control pulses set the repetition rate of thelaser to be the same as the AOD switching rate. At the resultingpulse rates the energy per pulse is reproducible (s.d. divided by1Departmentof Molecular Biology, 2Program in Neuroscience and 3Department of Physics, Lewis Thomas Laboratory, Washington Road, Princeton University,Princeton, New Jersey 08544, USA. 4Present address: Faculty of Biomedical Engineering, Technion 32000, Haifa, Israel. 5These authors contributed equally to this work.Correspondence should be addressed to S.S.-H.W. ([email protected]).RECEIVED 25 MAY; ACCEPTED 22 AUGUST; PUBLISHED ONLINE 21 OCTOBER 2005; DOI:10.1038/NMETH793NATURE METHODS VOL.2 NO.11 NOVEMBER 2005 837
ARTICLESInfraredlaserbScan lensDichroicDetectordUV laser150100500Tube lensNDIris filtersxy2200468Control signal (V)exy15010050001050 100 150 200Intended location (µm)f100ffffDichroicBrain sliceIntensity (a.u.) 2005 Nature Publishing Group http://www.nature.com/naturemethodsλ/4Beam waveexpander plate AODsc200Actual location (µm)ScanmirrorsLocation (µm)a8060402000.7 µm3 4 5 6 7 8Position control voltage (V)Figure 1 A system for patterned uncaging and two-photon imaging. (a) An ultraviolet uncaging laser beam is projected into the focal plane of a two-photonmicroscope. The AODs, lenses and iris approximately form a 4-f imaging system. The inset shows the lateral resolution of a single uncaging spot. (b) Dependenceof two-dimensional uncaging location on the control voltage, as measured by uncaging dried fluorescein dextran. The voltages control the frequency of theradio-frequency wave used to excite the two (x and y) acousto-optic crystals. (c) Comparison of the expected and actual uncaging after linearization of theresponse locations in b by a fourth-order polynomial. (d) Uncaging pattern at 45-ms access time per point and with adjustments to laser intensity tocompensate for position-dependent variations in transmission efficiency. Bar, 50 mm. (e) Uncaging pattern for a randomly-accessed grid pattern at 27-msaccess time per point. Bar, 50 mm. (f) Light energy reaching the specimen at different diffraction angles (magenta) is made more uniform (black) by reducingthe first-order beam intensity at angles of higher transmission efficiency.mean ¼ 2–3%), and is better than the laser manufacturer’sspecification for the fastest pulse rates.For steering the ultraviolet laser beam, a radio-frequency shearpressure wave propagating through the two TeO2 crystals (model2DS-150-50-.364; Brimrose Corp.) at the corresponding speed ofsound (617 m/s) creates periodic variations in refractive index,forming effective gratings that steer the beam. Using the iris, thefirst-order deflection beam, which contains the most power, ispassed while other beams (mixed zero and/or higher-order) areblocked. We chose a shear-mode TeO2 device over a longitudinaldevice in which the speed of sound is higher (4,260 m/s). Higherwave speeds potentially shorten switching times but also impose aneed for elongated deflectors and relatively complex designs for twodimensional systems (A. Bullen, V. Iyer, S.S. Patel & P. Saggau,Soc. Neurosci. Abstr. 25, 601.13, 1999; and ref. 16). The transmissionof TeO2 is not as consistent for near-ultraviolet (330–380 nm) lightas for visible and infrared light, so crystals were selected by themanufacturer for high transmission. In our system with the AODpower switched off, the two TeO2 crystals selected transmit B75% ofthe entering light. Because light throughput depends on the diffraction efficiency of the AOD, which for our devices depends on thecircular polarization of the incoming beam, polarization is adjustedusing a quarter-wave plate. After these adjustments the light energyarriving at the specimen is up to 300 nJ per pulse at B500 mWaverage laser power. This is reduced to the desired amount of energywith a neutral-density filter wheel and by modulating the radiofrequency power. An aperture is placed in the image plane to blockhigher diffraction orders as well as the parked beam.The ultraviolet uncaging beam emerging from the AOD isprojected through two lenses and then merged with the infraredimaging beam using a dichroic mirror. Both beams are thendirected through a 40 objective (Leica model UVI, 0.8 NA)838 VOL.2 NO.11 NOVEMBER 2005 NATURE METHODSinto a single shared focal plane. To ensure that the focal planes ofthe ultraviolet and infrared beams were near to one another, weused an objective with good transmittance at both wavelengths(Leica objective model UVI; transmission at 355 nm: B60%, at830 nm: B75%) and wide-band correction for intrinsic variationsin the index of refraction. Photolysis can be achieved in a variety ofaxial positions relative to the image plane by moving the first lensafter the AOD along its optical axis to shift the focus of theultraviolet beam. For most experiments the lens is used to focusuncaging into the same plane as the infrared beam17.We visualized uncaging by exciting fluorescent beads or byuncaging dried samples of caged fluorescein dextran. The uncagingfocal volume has an approximately Gaussian profile both laterallyin the focal plane (0.7-mm full width at half maximum; FWHM),and axially (10-mm FWHM), corresponding to an NA of approximately 0.3. This is less spatial resolution than the focal volume oftwo-photon excitation (for l ¼ 830 nm and NA 0.7, lateral FWHM¼ 0.4 mm and axial FWHM ¼ 2.6 mm; ref. 18). Further expansionof the uncaging beam entering the AOD introduced more spatialasymmetry in the focal volume and was not done. In brain slices(see Supplementary Data 1 and Supplementary Fig. 1 online),because of scattering, the estimated total excitation energy in thefocal volume decreased with a length constant of 32 mm. The lateraland axial FWHM of the uncaging spot grew slightly in the first30 microns of the slice, reaching values of 3 mm (lateral) and17 mm (axial) at a depth of 25 microns.Beam steering requires measurement of the relationship betweencontrol voltage and uncaging position. We did this by measuringuncaging position for a grid pattern of x and y control voltages(Fig. 1b) and finding mapping functions for interpolation andlinearization (Fig. 1c). It is possible in this way to reliably accessfields of up to B170 mm 170 mm.
The fundamental limit of time resolution is the theoreticalacousto-optical switching time, t ¼ daperture / vacoustic, wheredaperture is the diameter of the aperture and vacoustic is the acousticwave frequency. For a 5-mm transmitted beam diameter, t is 8 ms.An additional limit is the switching time of the radiofrequencydriver, whose nominal switching time was 50 ms (Brimrose modelVFE-150-50-V-B1-F2-2Ch, channels: deflection in x and y andintensity in x and y). The long switching time required us to use aswitching time between uncaging locations of 40 ms or more(Fig. 1d). Faster switching rates resulted in uncaging between theintended locations (Fig. 1e).Each crystal is controlled by two voltages via the radiofrequencyacousto-optical driver. One voltage controls the wave frequencybetween 125 and 175 MHz and directs the first-order diffractedbeam over a 1.7-degree range. The second voltage controls the waveamplitude and thus the first-order beam intensity. The diffractionefficiency in AOD crystals varies as a function of diffraction angle,with a multiple-hump structure that results from a design thatmeets two phase-matching criteria (Fig. 1f). To obtain a uniformuncaging intensity we used a two-step procedure. First, wegenerated a map of the intensity of light using a photodiodeplaced in front of the objective. Next, because the amount oflight per pulse is generally more than sufficient to achieve efficient25Control201510 APV50 APV DNQX–4–202Displacement (µm)4buncaging, we reduced the second control voltage at regions of highintensity. This truncated the peaks of the intensity distribution andgave a relatively flat field (Fig. 1f).Using caged glutamate to emulate dendritic integrationWe applied caged glutamate (MNI-glutamate)19–21 to single andmultiple sites on hippocampal pyramidal neurons in rat brain slices(Fig. 2). We measured the zone of receptor activation by movingthe uncaging location away from a CA1 pyramidal neuron dendrite(Fig. 2a). Responses dropped off to 50% of peak at distancesbetween 1.7 and 2.0 mm from the center (mean s.e.m., 1.9 0.1 mm for three branchlets). In the NMDA-type glutamatereceptor blocker APV (DL-2-amino-5-phosphonovaleric acid;100–200 mM), the responses were tightly localized to dendrites,with half-maximal distances of 0.75 to 1.4 mm (1.0 0.1 mm,7 branchlets; smaller than the distance with no antagonist, one-tailedt test, P o 0.01), comparable to the resolution in dendritic shafts oftwo-photon uncaging22. As uncaging was done over dendrites andnot spines, these responses were likely to include extrasynapticreceptors. Further addition of the non-NMDA receptor antagonist6,7-dinitroquinoxaline-2,3-dione (DNQX) totally abolished theresponses. At single locations, response amplitudes at 10-s intervalswere stable over periods of at least 200 s, indicating that responsescan be measured repeatedly without cumulative phototoxicity.Cell 1Cell 2To test if patterned uncaging can beused to study signal integration alongdendrites, we photolyzed glutamate at tendendritic locations in rapid succession.100 mVIndividual responses were small in both10 mscurrent c