region of interest
Non-technical summary Neurons communicate with each other with synapses using chemical messengers. The major synapses in the cerebral cortex utilize glutamate as a messenger and are made on special submicron structures, called dendritic spines. Dendritic spines are diverse in their size and densely packed in the cortex. Therefore, an optical technique for application of glutamate to single spines (two-photon (TP) uncaging) has been intensively used to clarify their functions in vitro. We have here extended 2P uncaging to living adult brain, and found that spine sizes display tight correlations with their functions, such as rapid glutamate sensing and an increase in cytosolic Ca2+ concentrations, even in vivo, as they were reported for in vitro preparations. Our data suggest that the structure and motility of dendritic spines play a key role in the adult brain function.
Abstract Two-photon (2P) uncaging of caged neurotransmitters can efficiently stimulate individual synapses and is widely used to characterize synaptic functions in brain slice preparations. Here we extended 2P uncaging to neocortical pyramidal neurons in adult mice in vivo where caged glutamate was applied from the pial surface. To validate the methodology, we applied a small fluorescent probe using the same method, and confirmed that its concentrations were approximately homogenous up to 200 μm below the cortical surface, and that the extracellular space of the neocortex was as large as 22%. In fact, in vivo whole-cell recording revealed that 2P glutamate uncaging could elicit transient currents (2pEPSCs) very similar to excitatory postsynaptic currents (EPSCs). A spatial resolution of glutamate uncaging was 0.6–0.8 μm up to the depth of 200 μm, and in vivo 2P uncaging was able to stimulate single identified spines. Automated three-dimensional (3-D) mapping of such 2pEPSCs which covered the surfaces of dendritic branches revealed that functional AMPA receptor expression was stable and proportional to spine volume. Moreover, in vivo 2P Ca2+ imaging and uncaging suggested that the amplitudes of glutamate-induced Ca2+ transients were inversely proportional to spine volume. Thus, the key structure–function relationships hold in dendritic spines in adult neocortex in vivo, as in young hippocampal slice preparations. In vivo 2P uncaging will be a powerful tool to investigate properties of synapses in the neocortex.
Optical stimulation can overcome the limitation of classical electrical stimulation: an electrode is often not sufficiently thin and flexible to selectively stimulate neuronal structures of interest. One such methodology, optogenetics, utilizes light-sensitive channels or pumps, the genes of which are transfected into a certain set of cells in the brain tissues, and subsequent stimulation of their proteins causes changes in the membrane potentials (Zhang et al. 2007; Gradinaru et al. 2010). Another methodology, two-photon (2P) uncaging of neurotransmitters, directly activates native neurotransmitter receptors via activation of caged agonists with a high spatiotemporal resolution (Matsuzaki et al. 2001) when these caged compounds are applied directly to preparations. These two optical methodologies have distinct features. For example, the selectivity of stimulation is mainly achieved by the specificity of gene expression in optogenetics, while it is achieved by the precise location of optical activation of chemical compounds in 2P uncaging. Optogenetics is often more suitable for studying network properties (Wang et al. 2007; Zhang et al. 2007; Gradinaru et al. 2010), while 2P uncaging is more suitable for studying subcellular neuronal structures such as synapses (Matsuzaki et al. 2004; Bloodgood & Sabatini, 2005; Noguchi et al. 2005; Beique et al. 2006; Asrican et al. 2007; Harvey & Svoboda, 2007; Honkura et al. 2008; Tanaka et al. 2008; Lee et al. 2009; Kantevari et al. 2010; Govindarajan et al. 2011) and dendrites (Losonczy et al. 2008; Branco et al. 2010).
Although optogenetics has already been applied to many in vivo preparations, 2P uncaging of neurotransmitters has until now never been applied in vivo. This is because methods for the reliable and quantitative application of small chemical compounds to brain tissues have not been established. The lack of in vivo 2P uncaging methodology has hampered the direct characterization of synaptic structures and functions. For instance, although structures of dendritic spines, particularly spine head volumes, are known to show tight correlation with their function in young hippocampal slice preparations (Matsuzaki et al. 2001; Noguchi et al. 2005; Beique et al. 2006; Asrican et al. 2007; Zito et al. 2009), it has not been clarified whether the same rule holds in adult neocortex in vivo. This point is particularly important since long-term changes in spine volumes have been reported in adult forebrains, and related to learning and memory (Hofer et al. 2009; Kasai et al. 2010b; Roberts et al. 2010).
We therefore developed a methodology to apply 2P uncaging of neurotransmitters to the neocortex of adult mice in vivo. We first examined the diffusion of a small fluorescent probe from the pial surface into the brain parenchyma, and estimated the actual concentrations of caged agonists applied to the pyramidal neurons in the brain up to 200 μm from the pial surface. We then established the conditions where 2P glutamate uncaging could stimulate synapses at the level of single spines, and found that the strong structure–function relationship of spines exists for spine volume and functional expression of AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptors. The relationship was stably maintained over many minutes, suggesting that the structure of spines plays an important role in the synaptic transmission. Moreover, we found that Ca2+ signalling induced by 2P uncaging showed inverse correlation with spine volume, suggesting that the neocortex shares similar learning rules as the hippocampus.
Surgery of mice
All animal procedures were approved by the Animal Experiment Committee of the University of Tokyo. Procedures were in accordance with the University of Tokyo Animal Care and Use Guidelines and are also in compliance with the policies and regulations of The Journal of Physiology as elaborated by Drummond (2009). We anaesthetized male C57BL6/J mice (4–6 weeks old) with intraperitoneal injections of urethane (Sigma-Aldrich, St Louis, MO, USA) and xylazine (Rompun, Bayer Health Care, Leverkusen, Germany) at 1.2 g and 7.5 mg (kg body weight)−1, respectively, which were supplemented with a subcutaneous of an analgesic, ketoprofen (20 mg (kg body weight)−1; Merial, Duluth, GA, USA). A steel plate with a recording chamber was attached to the skull using cyanoacrylate glue (Aron-alpha, Toagosei, Tokyo, Japan) (Fig. 1A), such that the recording chamber was attached to the skull just above the visual cortex (3.0 mm posterior, 2.5 mm lateral to bregma) (Paxinos & Franklin, 2001), and the plate was then tightly fixed to the platform with nuts. We then removed the skull using a pair of forceps and a dental drill, which was fixed to a stereotaxic instrument. The dura mater was carefully removed using a pair of fine forceps to minimize any pressure applied to the brain surface. We then placed a semicircular glass coverslip (cut from a circular one; ∼85 μm thickness, 2.7 mm diameter; Matsunami Glass, Kishiwada, Japan) to cover approximately half of the exposed brain surface. The coverslip was fixed using dental acrylic (ADFA, Shofu, Kyoto, Japan) (Fig. 1A) or a stainless wire. Mice were supplied with humidified oxygen gas and warmed to 37°C using a heating pad (FST-HPS, Fine Science Tools Inc., North Vancouver, Canada) (Supplemental Fig. S1).
The recording chamber was constructed on the stainless steel plate (Fig. 1). The plate had a hole with a diameter (5 mm) larger than that of the skull opening (2.1 mm). The chamber (c) had a solution reservoir made of silicon rubber (Fig. 1A), and was heated to 32–37°C by circulating warm water through tubing (t, Fig. 1A) made of stainless steel or silicone. We superfused the chamber with artificial cerebrospinal fluid (ACSF) for 0.5–1 h (Supplemental Fig. S1), removed the ACSF, and then added 200 μl of a solution containing a fluorescent probe or caged compound. In some experiments, to maintain the osmotic pressure (∼310 mosmol kg−1) of the bathing solution for a long experiment (>1 h), pure water was slowly added using a syringe pump (1 μl min−1; KDS-100, KD Scientific Inc., Holliston, MA, USA), and the osmolarity of the solution was checked at the end of the experiment.
In vivo two-photon imaging and uncaging were performed using an upright microscope (BX61WI; Olympus, Tokyo, Japan) equipped with a FV1000 laser scanning microscope system (FV1000, Olympus) and a water-immersion objective lens (LUMPlanFI/IR 60× with a numerical aperture of 0.9, Olympus) (Supplemental Fig. S1). The system included two mode-locked femtosecond-pulse Ti–sapphire lasers (MaiTai, Spectra Physics, Mountain View, CA, USA). The laser, which was used for uncaging, was set at 720 nm (Matsuzaki et al. 2001). The other laser, which was used for imaging, was set at 830 and 950 nm for Alexa 594 and Alexa 488, respectively. Each light path was connected to the microscope via an independent scan head and acousto-optic modulator. For 3-D reconstruction of dendrite images, 21–40 XY images separated by 0.5 μm were stacked by summation of fluorescence values at each pixel. Uncaging laser power was typically set at 6 exp(d/120) mW, where d (μm) is the depth of the dendrite from the pial surface.
Imaging of Alexa 488 in the brain
The bathing solution was changed to ACSF containing 100 μm Alexa 488 (Molecular Probes, Eugene, OR, USA) and 200 μm Trolox to measure the axial gradient of fluorescence. Time-lapse XZ images of the neocortex were obtained for 60 min. The efficiency of fluorescence collection E(z) at each depth within the brain tissue was calibrated by injecting 5 μl of 5 mm Alexa 488 in ACSF into the contralateral lateral ventricle (LV; +0.5 mm posterior, +1.2 mm lateral to bregma) (Paxinos & Franklin, 2001). We did not remove the dura for calibration experiments to avoid losing the dye. The fluorescence intensity was maximal for 1–3 h after application when XZ images of the neocortex were obtained for calibration using the recording chamber. The mice were then anaesthetized with 2–4% isoflurane and subjected to transcardiac perfusion with 20 ml of PBS followed by 30 ml of 4% paraformaldehyde (PFA). The neocortex was post-fixed in 4% PFA at 4°C for >12 h, washed several times with PBS, sliced at 100 μm thickness using a tissue slicer (VT1000, Leica, Wetzlar, Germany), and then imaged using the 2P microscope. In the pipette calibration method, a tip-closed glass micropipette filled with 100 μm Alexa 488 was imaged at several depths of the brain parenchyma (Oheim et al. 2001).
A region of interest (ROI) for obtaining the fluorescence value at each depth was composed of 10 × 2 pixels (4.1 μm × 4.0 μm), and the background values were obtained at a depth >500 μm, which was subtracted from the signal value. The relative concentration of the dye at a certain depth of the brain parenchyma C(z) was estimated by C(z) = (FS(z)/FS(0))/(FLV(z)/FLV(0)) where FS(z) and FS(0) represent fluorescence intensities at an arbitrary depth and at the surface, respectively, under the surface-loading condition (Fig. 1), and FLV(z) and FLV(0) represent fluorescence intensities at an arbitrary depth and at the brain surface, respectively, under the LV-loading condition (Supplemental Fig. S3). We estimated the relative spine head volume by measuring the total fluorescence intensity of stacked images of spines relative to that of the largest spine in the mapping region.
The recording chamber was first continuously superfused with fresh ACSF (125 mm NaCl, 2.5 mm KCl, 1 mm MgCl2, 2 mm CaCl2, 1.25 mm NaH2PO4, 26 mm NaHCO3 and 20 mm d-glucose), which was bubbled with 95% O2 and 5% CO2, at 0.25 ml min−1. The bathing solution was then changed to ACSF containing 1 μm TTX (Nacalai, Kyoto, Japan), 200 μm Trolox (Sigma-Aldrich) and 20 mm MNI-glutamate (Matsuzaki et al. 2001) (synthesized by us or purchased from Tocris Bioscience, Bristol, UK). Whole-cell patch pipettes were made from borosilicate glass capillary (GD-1.5, Narishige, Tokyo, Japan or GC150F-10, Harvard Apparatus, Holliston, MA, USA) using a puller (MF-83, Narishige). The pipettes were filled with a solution containing 138 mm potassium gluconate, 4 mm MgCl2, 4 mm ATP (sodium salt), 0.4 mm GTP (sodium salt), 10 mm disodium phosphocreatine, 10 mm EGTA-KOH, 10 mm Hepes-KOH (pH 7.2) and 0.04 mm Alexa 594 (Molecular Probes). For calcium imaging, 0.3 mm Oregon Green BAPTA-1 (OGB-1) and 0.03 mm Alexa 594 were included, EGTA was omitted and potassium was replaced with cesium.
The whole-cell patch clamp was established by the blind method (Margrie et al. 2002) or the shadow patching method (Kitamura et al. 2008). Neurons were voltage clamped at −70 ∼−89 mV (Axopatch 200B, Molecular Devices, Sunnyvale, CA, USA), and the currents were low-pass filtered at 2 kHz and sampled at 25 kHz.
Three-dimensional mapping of glutamate-induced currents
Mapping of AMPARs by uncaging was performed as described previously (Matsuzaki et al. 2001). The XY scanning area of the ROI was divided into several pixels (1 pixel = 0.55 μm), and the currents induced by glutamate uncaging were sampled at an interval of 50 ms. The 2-D scanning was performed at four to nine different Z-axis planes each separated by 1 μm. The Z position was automatically positioned by a piezo objective lens actuator (PI GmbH & Co. KG, Karlsruhe, Germany) with a commercially available program (FV1000MPE, Olympus). Peak current amplitudes, ∼2–9 ms from the onset of uncaging, were assigned to each pixel and displayed by pseudocolour coding. The data were further processed by linear interpolation and stacking along the Z axis by the maximum intensity method (Figs 2F and 3B). Time-lapse mapping of one dendrite was conducted every ∼10 min.
Calcium imaging was conducted as previously (Noguchi et al. 2005). Briefly, whole-cell recording pipettes were filled with a solution containing OGB-1 and Alexa 594, and the neurons were held at −20 mV. Line scanning of the imaging laser was conducted at 167–266 Hz and the fluorescence ratio (G/R) was measured by fluorescence emission acquired at 400–550 nm (‘G’ channel) and 590–650 nm (‘R’ channel) for OGB-1 and Alexa Fluor 594, respectively. Uncaging duration was 0.8 ms and laser power was set at 10 ed/120 mW, where d is the distance from the pial surface. Given that G/R before stimulation (G/R)0 was similar to (G/R)min, we used the value ΔG/R = G/R − (G/R)0 to evaluate Δ[Ca2+]i. We started imaging about 20 min after whole-cell perfusion.
Ranges of values were expressed as mean ± SEM unless otherwise stated. Statistical analyses of the differences in values were performed using one-way ANOVA as shown in Fig. 1F and Fig. S3E. Analysis of the significance between spine-head volume and the peak-current amplitudes or ΔG/R was conducted using a t test of correlation coefficient.
Surface application of chemical compounds to the neocortex
We applied caged compounds to the neocortex surface, which was exposed by removing a part of the skull and dura mater (Fig. 1A) (see Methods). We used a homemade recording chamber for the mouse heads, which simultaneously allowed: (1) the delivery of chemical compounds to the brain surface; (2) in vivo whole-cell recording from the cortical region exposed to the external medium; (3) 2P imaging of fine neuronal structures under the glass coverslip; and (4) 2P uncaging of caged compounds at the identified neuronal structures (Supplemental Fig. S1). The recording chamber was constructed on a metal plate, which was fixed on the skull of the mouse. The metal plate had a hole with a diameter (∼5 mm) greater than that of the skull opening (∼2.1 mm) (Fig. 1A). The skull opening was half covered with a glass coverslip to suppress the pulsation of the brain for 2P imaging. The remaining half of the brain surface was accessible to an external medium and patch pipette. The total volume of the external medium was ∼200 μl during imaging (Fig. 1B).
For reliable application of in vivo 2P uncaging, it is critical to know how small chemical compounds are distributed in the brain tissue when applied to the brain surface. Such distribution depends on the structures of the chamber and the brain. We therefore developed a direct calibration method using a fluorescent dye, Alexa 488 (100 μm), which has a similar molecular weight (MW) as a caged compound that we used; MWs for Alexa 488 and 4-methoxy-7-nitroindolinyl (MNI)-glutamate (Matsuzaki et al. 2001) are 643 and 323, respectively. Fluorescence images could be obtained when we applied Alexa 488 to the chamber (Fig. 1C). The dashed white rectangle in Fig. 1C denotes the area where 2P imaging was prevented by diffraction by the edge of the glass (g). We excluded the columnar region within 60 μm from the edge of the glass in the following experiments.
The fluorescence intensity was gradually increased after its application to the chamber with a time constant of ∼20 min (Supplemental Fig. S2). We performed the imaging and uncaging experiment at >40 min after application of the solution, i.e. when the fluorescence was nearly equilibrated. Fluorescence intensity F(z) at a particular depth of z (Fig. 1D) in the cortex is expressed as αC(z)E(z), where α is the extracellular space (ECS) volume fraction, C(z) the concentration gradient of the dye, and E(z) epifluorescence collection efficiency, which depends on several factors, including the excitation and collection of fluorescence in the tissue by the objective lens, light scattering within the tissue, diameter of skull opening, narrowing of the effective optical opening of the skull by blood vessels, numerical aperture and working distance of the objective lens (Oheim et al. 2001; Beaurepaire & Mertz, 2002). Analysis of the fluorescence decays (Figs 1D and E and S3) gives the concentration gradient C(z) (Fig. 1F), as follows.
We first determined the value of α directly by plotting fluorescence intensities from the external medium into the brain surface (Fig. 1D; along the red line shown in Fig. 1C). The value of α was estimated as the ratio of fluorescence intensities between the external medium and the outermost part of the brain where light scattering within the tissue was minimal. This approach is particularly relevant in 2P imaging, which is least affected by the inner filter effects of fluorescent probes along the optical path (Fig. 1D), unlike one-photon (1P) imaging (Lakowicz, 2006). The value of α thus estimated was 22.4 ± 3.7% (mean ± SD, n= 5). The large value of α obtained in this study was consistent with that obtained in a previous study in which α was obtained by diffusion (Syková & Nicholson, 2008) or self-quenching (Magzoub et al. 2009) measurements in deeper regions of the cortex, supporting that α is similar between the superficial and deeper layers in the cortex. This large α value allows sufficient number of caged compounds to be uncaged for the activation of neuronal receptors.
We next estimated E(z) of the neocortex by injecting Alexa 488 into the lateral ventricle (LV) of the mice whose dura was kept intact. We waited for >1 h for the dye to equilibrate in the cerebral cortex (Supplemental Fig. S3A), and then obtained the images (Supplemental Fig. S3B). An axial fluorescence gradient was still observed (Supplemental Fig. S3C), though it was less steep than that shown in Fig. 1E. To confirm the homogenous distribution of the dye along cortical layers, we fixed the cortex by transcardiac perfusion of PFA since the Alexa 488 dye is fixable. We then cut the coronal section of the neocortex (Supplemental Fig. S3A) and imaged the fluorescence profile along the layers by excluding obvious cell bodies and blood vessels (Supplemental Figs S3D and S4). We found no significant gradient (Supplemental Fig. S3E; n= 3; P= 0.29, one-way ANOVA). Therefore, the fluorescence decay curve shown in Supplemental Fig. S3C purely reflects E(z).
We then obtained C(z) by dividing the fluorescence profile of the surface loading (Fig. 1E) representing αC(z)E(z) by that for the LV loading (Supplemental Fig. S2C) representing αE(z), for the brain region covered with (open circles) and without (filled circles) the coverslip, respectively (Fig. 1F) (see Methods). In the brain region without the coverslip, the fluorescence intensity was 100% of the external medium at the surface, and gradually decayed to 40% at a layer 200 μm below the brain surface. Under the coverslip, however, the fluorescence intensity was only 40% of the external medium, even at the pial surface (29–46%, n= 5), and the dye concentration did not critically depend on the depth of the layer (P= 0.07, one-way ANOVA) and was 43 ± 4% (29–51%, n= 5) up to 200 μm. The coverslip did not significantly reduce the fluorescence (Supplemental Fig. S3B). These data suggest that the dye could flow into the cortex underneath the coverslip mainly from the open side of the cortex.
We also confirmed the epifluorescence collection efficiency E(z) using a previous methodology by imaging a pipette filled with Alexa 488 at several depths of the tissue (Supplemental Fig. S5) (Oheim et al. 2001). The attenuation length constant was 166 ± 7 μm (147–182 μm, n= 5) (Helmchen & Denk, 2005), consistent with that obtained in our LV dye injection experiments (Supplemental Fig. S3C). The pipette calibration method, however, tends to damage brain structures, such as blood vessels. It was also inaccurate because of reflection and refraction of light by the pipette, and because of insufficient volumes of the pipette to fill the focal volume of the objective lens. In effect, the data were highly variable (Supplemental Fig. S5). In addition, the pipette method cannot estimate the precise gradients caused by structures of a particular recording chamber with a glass coverslip, unlike the LV injection method (Supplemental Fig. S3C).
In vivo 2P glutamate uncaging
Distribution of functional AMPA-type glutamate receptor at the dendrite of excitatory neurons is tightly correlated with spine volume in slice preparations from young rat (Matsuzaki et al. 2001; Noguchi et al. 2005; Beique et al. 2006; Asrican et al. 2007; Zito et al. 2009), while such a relationship has never been examined in vivo. We therefore applied our in vivo uncaging method to test whether this rule also holds true in vivo (Fig. 2A). We applied a higher concentration (20 mm) of a caged glutamate compound, MNI-glutamate, to the brain surface than that used in slice preparations, since concentrations beneath the coverslip were about 40% of the external medium. Tetrodotoxin (TTX) was also included to prevent the spontaneous activity of neurons. More than 20 min after application of the caged compound, whole-cell patch clamp recording was performed from a neocortex layer 2/3 (L2/3) pyramidal neuron using either the blind method (Margrie et al. 2002) or the shadow patch method (Kitamura et al. 2008). The whole-cell pipette containing Alexa 594 stained the neurons within 20–40 min (Fig. 2B).
We successfully induced transient inward currents by 2P uncaging of MNI-glutamate at the identified spines with laser powers of 12 mW for a duration of 0.6 ms (Fig. 2C and D). The amplitudes of 2pEPSCs were within the range of miniature EPSC amplitude (10.5 ± 0.5 pA, 4.2–31.6 pA, n= 79). In this experiment, the laser powers were chosen so that the powers at the plane of focus (86 μm) were similar to that at the slice preparation when light scattering within the tissue was considered (6 mW = 12 e−86/120 mW) (Matsuzaki et al. 2001), assuming the length constant of light scattering of 120 μm (Helmchen & Denk, 2005). The average amplitudes of 2pEPSCs which were evoked in dendritic branches at various depths of d and with the laser powers of 6 ed/120 mW (Supplemental Fig. S6A) were slightly larger in deeper layers (Supplemental Fig. S6B), possibly reflecting the dendritic cable property.
To obtain 3-D mapping of the amplitudes for 2pEPSCs in a dendritic branch, a region of interest (ROI) was set in an arbitrary position on a dendrite (the red rectangle in Fig. 2C). ROI was divided into appropriate pixels (such as 16 × 10 pixels) with a width of 0.55 μm. Caged glutamate uncaging was performed in every pixel, and current values were represented with a pseudocolour coding (Fig. 2E). The same procedure was applied at several depths to obtain the maximum intensity projection maps (Supplemental Fig. S7A). The amplitudes of currents for each spine were obtained twice in the experiments shown in Supplemental Fig. S7A, where the 3-D maps are quite similar between the two (Supplemental Fig. S7A and B; correlation coefficient: r= 0.94, n= 12 spines), indicating the reliability of the mapping experiments. In Fig. 2F, the maximum intensity projection maps were displayed for the dendritic surface, indicating that sensitivity to glutamate was clustered at spines.
To confirm the stability of our 3-D mapping data, we used the automated mapping system with which one complete 3-D map could be obtained within 1.5–2 min (Methods), and was repeated three times with an interval of ∼10 min in the same dendrites of L2/3 pyramidal neurons (Fig. 3A and B). The distributions in the amplitudes of 2pEPSCs were similar in the three mapping experiments (Fig. 3B). The maximal amplitudes of currents for each spine in the dendrites were stable for 20 min (Fig. 3C). The normalized amplitudes of the maximal amplitudes of currents from many spines in five dendrites were stable (Fig. 3C), although a slight reduction in the average currents was detected. This reduction could be attributed to the increases in the access resistance of the whole-cell recording. Thus, the functional expression of AMPA receptors was stable over time.
We assessed the spatial resolution of glutamate uncaging by the smallest current spots in the 3-D mapping data (Matsuzaki et al. 2001). The full-width-at-half-maximum (FWHM) spatial resolutions of current mapping obtained from a dendrite were 0.80 ± 0.05 μm laterally (n= 5) (Fig. 4A) and 1.9 ± 0.3 μm axially (Fig. 4B), and they were independent of the depth of the brain up to 200 μm (Supplemental Fig. S8). The FWHM values were slightly larger than those obtained in the slice preparations (0.6 μm and 1.4 μm) (Figs 4 and S8, green lines) (Matsuzaki et al. 2001). We think this apparent degradation was mainly due to the pulsations of the brain by blood flow and respiration approximately with a distance of 0.1–0.4 μm. The pulsation can blur the current mapping data which were slowly obtained with a semi-random protocol (see Methods). In contrast, the in vivo fluorescent images that were acquired with a fast raster scanning showed the FWHM resolution of 0.45 μm laterally and 1.9 μm axially (Supplemental Fig. S8C and D), which are close to the spatial resolution of our microscope (Supplemental Fig. S8C and D, dashed lines). We also confirmed that the hemi-open glass window of our recording chamber did not degrade the spatial resolution (Fig. 1C): The fluorescent profiles of a small bead with a diameter of 0.04 μm fixed on a coverslip were unaffected by the recording chamber (0.35 μm laterally and 1.6 μm axially) (Supplemental Fig. S8C and D) (Matsuzaki et al. 2001). These data suggest that the actual focal volume of the uncaging was similar to that in the slice preparation (Fig. 4A and B), even though the spatial resolution of the mapping data was slightly degraded. Thus, we can selectively stimulate individual spines using in vivo 2P uncaging.
We therefore obtained the maximum amplitudes of the currents for a spine, and examined how they were correlated with fluorescence intensities of spine heads, i.e. the relative spine head volumes within the dendrite (Methods). We found the correlation between the relative spine head volume and the maximum amplitudes of currents in the spine was very strong (Fig. 4C; r= 0.90, n= 12). Similar current mapping results from 16 dendrites (Fig. 4D) indicate that the functional AMPA receptor distribution is approximately proportional to the spine head volume (r= 0.48–0.99, 0.88 ± 0.03, n= 16).
Finally, we investigated properties of the Ca2+ signalling evoked by in vivo 2P glutamate uncaging. Increases in [Ca2+]i were monitored by fluorescence of calcium-sensitive Oregon Green BAPTA-1 (OGB-1, ‘G’ channel) and calcium-insensitive Alexa 594 (‘R’ channel) simultaneously acquired (Fig. 5A and B) (Noguchi et al. 2005). Neurons were depolarized to −20 mV to facilitate the opening of NMDA receptors. We estimated the increases in [Ca2+]i by ΔG/R elicited by glutamate uncaging (Fig. 5A and B). The ratiometric measurement cancelled out the fluorescence fluctuation from the brain pulsation caused by blood flow and respiration (Fig. 5Ba). We found that the peak values of ΔG/R negatively correlated with spine head volumes (Fig. 5D, two dendrites, r=−0.62, P= 0.018, t test of correlation coefficient). Thus, the spines in adult neocortex in vivo showed similar structure–function relationships as in young hippocampal slice preparations (Noguchi et al. 2005; Sobczyk et al. 2005).
We have established a methodology for in vivo 2P uncaging of neurotransmitters in the neocortex of adult mice. Our LV injection method allowed us to estimate the concentration gradient of the fluorescent probe underneath the coverslip under the same conditions as actual uncaging experiments, and showed that the concentration of the probe was about 40% of the external solution, and nearly homogenous up to 200 μm from the pial surface. We also estimated the ECS volume fraction, α, of the neocortex as 22%, consistent with the one estimated with diffusion (Syková & Nicholson, 2008) and self-quenching measurements (Magzoub et al. 2009). ECS was probably constant along the depth of the cortex, because the fluorescence intensity was constant along all layers of the cortex in the LV injected and fixed brain preparations (Supplemental Figs S3E and S4). Electron microscope (EM) studies often estimated ECS as ∼10% (Rusakov & Kullmann, 1998; Syková & Nicholson, 2008). The smaller ECS obtained by EM might be due to the section's oblique intercellular spaces which are not measured. Our study has thus clarified that the cortical surface is an appropriate preparation for 2P uncaging of neurotransmitters.
In fact, we were able to induce 2pEPSCs by 2P uncaging which was aimed at identified spines. The spatial resolution of the 2P uncaging for the 2pEPSC was 0.6–0.8 μm up to the depth of 200 μm (Fig. 4A and B). Automated 3-D mapping of glutamate sensitivities demonstrated that expression of functional AMPA-type glutamate receptors was stable over many minutes, and is proportional to the spine volume in adult pyramidal neurons in vivo, as in young hippocampal preparations. This is consistent with previous EM data in chemically fixed cortex, where the number of immuno-gold labelling of AMPA receptors showed a correlation with the length of active zone (Kharazia & Weinberg, 1999). The structure–function relationship is essential to learning and memory (Kasai et al. 2010a), because the relationship was stable over 20 min (Fig. 3B), and because changes in spine volumes in adult neocortex over days have recently been reported in a manner correlated with experiences and learning (Hofer et al. 2009; Roberts et al. 2010).
We could also record Ca2+ transients evoked by 2P uncaging, and found that increases in [Ca2+]i were greater for thinner spines, as found in young hippocampal preparations. Since [Ca2+]i increases are fundamental triggers of synaptic plasticity, the learning rules of spines must have a similarity in the neocortex and hippocampus (Kasai et al. 2010a,b). We found that [Ca2+]i increases were well confined within spine heads (Fig. 5C). The confinement of [Ca2+]i increases in each spine head must be the important basis for independent induction of synaptic plasticity (Kasai et al. 2010b).
Our data suggest that the spatial resolution of 2P uncaging in vivo is similar to the one in slice preparations, and that the resolution can be further improved by adaptive optics (Ji et al. 2010), vector beams (Kozawa & Sato, 2007) or STED (Stimulation Emission Depletion) approaches (Nagerl et al. 2008). We think that 2P uncaging of glutamate is also possible in the cortical layers deeper than 200 μm from the pial surface, although reliable estimation of concentrations of probes becomes more difficult, because the fluorescent signal becomes weaker in deeper layers of cortex. The intra-ventricular application of caged compounds may give a solution to the depth problem when caged glutamate compounds with less antagonistic effects on GABA receptors become available (Fino et al. 2009; Matsuzaki et al. 2010). In the future, spine structural plasticity may be induced by 2P uncaging, and be traced for days for the quantification of long-term synaptic structural plasticity. Thus, in vivo 2P uncaging will be a powerful technique to directly investigate the functions of neocortical synapses and their abnormalities in adult animals.
J.N. and H.K. designed the project; J.N., A.N. and S.W. performed the in vivo mouse brain experiment and analysed the data; G.C.R.E-D. synthesized MNI-Glu; K.K. and M.K. assisted in vivo whole-cell clamp experiments; H.K., M.M. and J.N. designed the set up for the 2P microscope; J.N and H.K. wrote the manuscript and all authors approved the final version.
We thank T. Sasaki, C. Miura, K. Tamura, H. Iwanami, and M. Yoshida for technical assistance, as well as T. Hayama, T. Ichiki, T. Akagi, H. Takehara, S. Okabe, S. Kondo, J. Nabekura, H. Wake, Y. Takatsuru, R. Shigemoto, Y. Fukazawa, T. Isa, K. Seki, M. Yoshida, A. Nambu, S. Chiken, M. Miyata, T. Kimura, Y. Komatsu, T. Kuwaki, A. Nakamura and H. Hirase for their technical advice. This work was supported by Grants-in-Aid for Specially Promoted Area (No. 2000009 to H.K.), Scientific Research (S) (No. 21220006 to M.K.), Scientific Research (C) (No. 21500367 to J.N.), Scientific Research on Priority Areas (Elucidation of neural network function in the brain, No. 20021008 to M.M.), Young Scientist (A) (No. 19680020 to M.M., No. 22680031 to K.K.), Challenging Exploratory Research (No. 22650083 to K.K.), the Strategic Research Program for Brain Sciences (Development of Biomarker Candidates for Social Behavior to M.K.), and the Global COE Program (Integrative Life Science based on the study of Biosignalling Mechanisms to H.K. and M.K.) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), and by National Institute of Health Grants GM65473 (to G.C.R.E.-D. and H.K.) and GM53395 (G.C.R.E.-D.). This work was also supported by a Mitsubishi foundation grant to M. M. and a Research Grant from Human Frontier Science Program to H.K.