Abstract
Two-photon uncaging of glutamate is widely utilized to characterize structural plasticity in brain slice preparations in vitro. In this study, we investigated spine plasticity by using, for the first time, glutamate uncaging in the neocortex of adult mice in vivo. Spine enlargement was successfully induced in a smaller fraction of spines in the neocortex (22%) than in young hippocampal slices (95%), even under a low magnesium condition. Once induced, the time course and mean amplitudes of long-term enlargement were the same (81%) as those in vitro. However, low-frequency (1–2 Hz) glutamate uncaging caused spine shrinkage in a similar fraction (34%) of spines as in vitro, but spread to the neighboring spines less frequently than in vitro. Thus, we found that structural plasticity can occur similarly in the adult neocortex in vivo as in the hippocampus in vitro, although it happens stringently in a smaller subset of spines.
Introduction
Most excitatory synapses in the brain form on dendritic spines. The volume of dendritic spines is tightly correlated with the functional expression of glutamate receptors in the young hippocampus in vitro (Matsuzaki et al., 2001; Smith et al., 2003; Beique et al., 2006; Asrican et al., 2007; Holbro et al., 2009; Zito et al., 2009) and in the adult mouse neocortex in vivo (Noguchi et al., 2011). Spine volume changes have been associated with long-term potentiation and depression of synapses in hippocampal preparations (Zhou et al., 2004; Kopec et al., 2007). Such volume changes eventually cause the generation and elimination of spines (Yasumatsu et al., 2008; Bhatt et al., 2009; Holtmaat et al., 2009; Xu et al., 2009; Kasai et al., 2010; Hayashi-Takagi et al., 2015). Impaired structural plasticity induces pathological states of neuronal circuits (Fiala et al., 2002; Kasai et al., 2010; Forrest et al., 2018).
Two-photon uncaging of caged glutamate compounds (Matsuzaki et al., 2001) is the only available method that reliably stimulates single spines. It is widely used to characterize spine structural plasticity in vitro. Spine enlargement is most robustly induced by uncaging caged glutamate in the absence of external magnesium (Mg2+) so that N-methyl-D-aspartic acid (NMDA) receptors are maximally activated (Matsuzaki et al., 2004; Noguchi et al., 2005; Harvey et al., 2007; Honkura et al., 2008; Lee et al., 2009; Govindarajan et al., 2011; Bosch et al., 2014). Spine shrinkage can be induced by low-frequency uncaging (Hayama et al., 2013; Oh et al., 2013; Noguchi et al., 2016). However, assessing spine plasticity with two-photon uncaging has never been characterized in vivo because of difficulties in uncaging in vivo. The characteristics of structural plasticity is unknown in the adult mouse neocortex in vivo.
We previously established a glutamate uncaging method in vivo in which a caged glutamate compound is applied on the surface of the brain. This method allows the compound to spread into the superficial extracellular space of the neocortex for free diffusion (Noguchi et al., 2011). We now extend our study to focus on the structural plasticity of dendritic spines in vivo.
Results and Discussion
Spine enlargement in vivo
Two-photon uncaging of the caged glutamate compound was applied to single spines of tuft dendrites of layer 5/6 pyramidal neurons in the visual cortex of adult mice in vivo (Noguchi et al., 2011). We used the green fluorescent protein (GFP)-expressing M mouse line or the yellow fluorescent protein (YFP)-expressing H mouse line in which a subset of layer 5/6 pyramidal neurons was selectively labelled. Mice were anesthetized with urethane and xylazine and placed under a microscope objective lens using an imaging chamber that was firmly attached on the skull of the mouse (Figure 1A). To activate NMDA receptors effectively, the recording chamber was superfused with artificial cerebrospinal fluid containing no magnesium (Mg) ions. Caged glutamate was thereafter superfused (Figure 1A and Figure 1–figure supplement 1A). Spine head volume (VH) fluctuations before uncaging were quantified as coefficients of variation (CVs) (Figure 1–figure supplement 1B). The CV of in vivo neocortex spines (15% ± 16% [mean ± standard deviation (SD)]; 227 spines) was not larger than that of hippocampal slices (21%)(Matsuzaki et al., 2004), which ensured the stability of our recording conditions.
Enlargement of the spines could be induced by two-photon glutamate uncaging repeated 60 times at 1 Hz adjacent to the spine heads (Figures 1B and 1C). Volume changes varied among individual spines; however, the averaged time course showed a transient increment phase, followed by a stable plateau phase (Figure 1D). For spines showing >30% enlargement, the peak enlargement (10–30 min) and sustained phase of enlargement (>60 min) were 109% ± 24% (the mean ± the standard error of the mean: 16 spines/10 dendrites/10 mice) and 50% ± 12%, respectively. These values were similar to those of CA1 pyramidal neurons in vitro (Matsuzaki et al., 2004). Enlargement lasting more than 30 min occurred in 8 of 16 enlarged spines (Figure 1E) and was confined to the stimulated spines (Figures 1D and 1F).
Enlargement was recorded only in a small fraction of spines (22% of 74 spines/20 dendrites/18 mice; Figure 1E), compared with the fraction in the hippocampus in vitro (approximately 95%) (Matsuzaki et al., 2004). In spines without enlargement (ΔVH <30%), the average enlargement was negligible (−0.6% ± 2.5%) (Figure 1D). The stringency in spine enlargement was not because of technical reasons; the enlargement was induced mostly in one spine (0–4 spines; average, 0.8 spine) among several spines (1–7 spines; average, 3.7 spines) that were simultaneously stimulated. This conclusion was quantitatively supported by the fact that the amplitude of the enlargement of stimulated spines was uncorrelated with the distance of the spine from a spine showing significant enlargement (Figure 1G). In these studies, we selected small spines (Figure 1–figure supplement 2A) in which enlargement could be induced in the most pronounced manner. The enlargement was uncorrelated with spine neck length, depth, and mouse age (Figures 1–figure supplement 2B-D).
Spine shrinkage in vivo
We used a solution containing a physiological concentration (1 mM) of Mg2+ to study spine shrinkage (Noguchi et al., 2016). Several spines on a dendrite were stimulated by low-frequency two-photon glutamate uncaging (2.8 spines/dendrite, 1–2 Hz, 5–15 min) (Figure 2A). Stimulated spines showed a large volume reduction (Figures 2A and 2B, spine “S1”). The spine volume gradually reduced in vitro (Hayama et al., 2013; Noguchi et al., 2016) (Figure 2C). We found that 34% of the stimulated spines had shrunk (–ΔVH >30%, 15 of 44 spines/18 dendrites/8 mice) and that the mean amplitude at 20–50 min was 19% ± 4% (n = 44), which was similar to the findings of the young hippocampus in vitro (23% ± 7%, n = 8) (Noguchi et al., 2011). The shrinkage was long-lived (>80 min) in most (73%) spines (Figure 2D). Shrinkage was absent when we added the NMDA receptor antagonist APV in the perfusion solution (Figure 2C; Figure 2–figure supplement 1A).
Spine shrinkage spread to neighboring spines, which also occurred in hippocampal slice culture samples (Hayama et al., 2013; Noguchi et al., 2016). We calculated the average spine volume of the stimulated spines and the neighboring spines at 20–50 min from the onset of stimulation (Figure 2E). We found that the diffusion of spine shrinkage was only significant in spines next to the stimulated spines (<3 μm). Only 12% of spines within 3 μm of a stimulated spine showed shrinkage (–ΔVStimulated >30%; Figure 2–figure supplement 1B). Thus, the spread of spine shrinkage was more stringent in vivo than in vitro in which shrinkage spread to 71% of spines within 3 μm from a stimulated spine and to 38% of spines within 7 μm (Hayama et al., 2013; Noguchi et al., 2016).
We found that the prestimulation spine volume showed a weak but insignificant correlation with spine shrinkage (Figure 2–figure supplement 1C). Spine retraction occurred during spine shrinkage (Figure 2F) (Hayama et al., 2013); however, spine shrinkage was insignificantly correlated with retraction (ΔSpine length; Figure 2–figure supplement 1D). We did not observe any interspine distance dependency in the induction of spine shrinkage (Figure 2–figure supplement 2A). Spine shrinkage was also insignificantly correlated with the initial spine neck length, dendritic depth, and age of mice (Figure 2–figure supplement 2B–D).
Stringent structural plasticity of dendritic spines in the neocortex in vivo
We found that two-photon uncaging could induce prominent plasticity of spine structures in the adult neocortex in vivo that was similar to that of the hippocampus in vitro. The major difference was the low success rate of spine enlargement in vivo (22% vs. 95% in vitro), which was not caused by technical factors (Figure 1G). The success rate in inducing shrinkage was similar to that of the hippocampus, albeit its spread in the neocortex was limited. It remains to be clarified why enlargement is restricted in the neocortex, and whether it may occur after repeated reactivation in vivo. In summary, spine structural plasticity occurs in a stringent manner in the neocortex in vivo, which may provide a cellular basis for slow learning in the cortex (Lisman et al., 2001).
Materials and Methods
Surgery for the in vivo mouse experiment
All animal procedures were approved by the Animal Experiment Committee of the University of Tokyo (Tokyo, Japan). Procedures were conducted in accordance with the University of Tokyo Animal Care and Use Guidelines. The surgical procedure was previously described (Noguchi et al., 2011). In brief, we anesthetized adult mice expressing GFP or YFP in a subset of neurons: Thy1 GFP in the M line [GFP-M] or YFP in the H line [YFP-H]. Eighteen mice, aged 148 ± 129 days (expressed as the mean ± the SD), were used for the enlargement condition (YFP-H, 14 mice; GFP-M, 4 mouse). Eight mice, aged 70 ± 19 days, were used for the shrinkage condition (YFP-H, 7 mice; GFP-M, 1 mice). They were anesthetized with intraperitoneal injections of urethane and xylazine at 1.2 g/kg body weight and 7.5 mg/kg body weight, respectively, which were supplemented with the subcutaneous administration of the analgesic ketoprofen (20 mg/kg body weight). A steel plate with a recording chamber was attached to the skull by using cyanoacrylate glue so that the recording chamber was attached to the skull just above the visual cortex (3.0 mm posterior, 2.5 mm lateral to the bregma) (Paxinos & Franklin, 2001). The plate was then tightly fixed to the metal platform. We then removed the skull using a pair of forceps and a dental drill, which was fixed to a stereotaxic instrument (Narishige, Tokyo, Japan). The dura mater was carefully removed using fine forceps and a microhook to minimize any pressure applied to the brain surface. We then placed a semicircular glass coverslip to cover approximately one-half of the exposed brain surface (Figure 1A). The coverslip was fixed using dental acrylic (Fuji-Lute BC; GC Corp., Tokyo, Japan) or a stainless wire. The mice were supplied with humidified oxygen gas and warmed to 37°C with a heating pad (FST-HPS; Fine Science Tools Inc., North Vancouver, Canada).
Two-photon in vivo imaging and uncaging
In vivo two-photon imaging and uncaging were conducted using an upright microscope (BX61WI; Olympus, Tokyo, Japan) equipped with a FV1000 laser scanning microscope system (Olympus) and a water-immersion objective lens (LUMPlanFI/IR 60× with a numerical aperture of 0.9; Olympus). The system included two mode-locked femtosecond-pulse titanium-sapphire lasers (MaiTai; Spectra Physics, Mountain View, CA, USA). The laser was set at 720 nm and used for uncaging (Matsuzaki et al., 2001). The other laser was set at 980 nm and used for imaging. Each light path was connected to the microscope via an independent scan head and acousto-optic modulator. For the 3D reconstruction of the dendrite images, 21–40 XY images separated by 0.5 μm were stacked by summing the fluorescence values at each pixel. 4-Methoxy-7-nitroindolinyl (MNI)-glutamate or 4-carboxymethoxy-5,7-dinitroindolinyl (CDNI)-glutamate was custom-synthesized by Nard institute Ltd. (Amagasaki, Japan) or purchased from Tocris Bioscience (Bristol, UK) was perfused in the recording chamber in artificial cerebral spinal fluid (ACSF).
In vivo enlargement of dendritic spines
For the in vivo spine enlargement experiments, the cortical surface was first superfused with magnesium-free ACSF (ACSF w/o Mg2+) containing 125 mM NaCl, 2.5 mM KCl, 3 mM CaCl2, 0 mM MgCl2, 1.25 mM NaH2PO4 26 mM NaHCO3, 20 mM glucose, and 10 μM tetrodotoxin (Nacalai, Kyoto, Japan). This solution was bubbled with 95% oxygen and 5% carbon dioxide for approximately 30 ± 15 min (expressed as the mean ± the SD; 20 dendrites). The bathing solution was then changed to ACSF w/o Mg2+ containing 20 mM MNI-glutamate or 10 mM CDNI-glutamate and 200 μM Trolox (Sigma-Aldrich, St. Louis, MO, USA), thereby enabling its diffusion into the cortical extracellular space approximately 15 min before the uncaging experiments. Two-photon uncaging was aimed at the tip of the spines, and repeated 60 times at 1 Hz. The power of the uncaging laser was typically set at 10 mW with an activation time of 0.6 ms. We expected that transient currents similar to miniature excitatory-postsynaptic currents were roughly elicited at this laser power; however, in this experiment we did not fine-tune the power along the cortical depth (Noguchi et al., 2011). For each experiment, 2–8 spines (average, 4.6 spines) were stimulated along a dendrite. We studied 52 spines/15 dendrites/14 mice with MNI-glutamate, and 22 spines/5 dendrites/4 mice with CDNI-glutamate. The success rate of enlargement was 25% and 13%, respectively. The solution was pooled in a small reservoir (2 mL) (Figure 1A). We constantly added pure water (after determining its flow rate empirically) to the reservoir to maintain the osmotic pressure of the solution at approximately 320 mOsm/kg. The solution was warmed at 37°C on the chamber by using circulating hot water (Figure 1A). All physiological experiments were conducted at 37°C.
In vivo shrinkage of dendritic spines
For the spine shrinkage experiments, the cortical surface was superfused with ACSF containing 2 mM CaCl2 and 1 mM MgCl2. The bathing solution was then changed to ACSF that additionally contained 200 μM Trolox and a caged compound (i.e., 20 mM MNI-glutamate or 10 mM CDNI-glutamate). We studied 39 spines/16 dendrites/7 mice with MNI-glutamate, and 5 spines/2 dendrites/1 mice with CDNI-glutamate. The success rate of shrinkage was 36% and 25%, respectively. Repetitive stimulation was conducted at 1–2 Hz for 5–15 min with laser powers similar to those used for enlargement (~10 mW). As a control, the stimulation was also conducted in the presence of 50 mM D-2-amino-5-phosphonovaleric acid (APV), which is an NMDA receptor antagonist with MNI-glutamate.
Analysis of the spine volume
Spine head volumes were estimated from the total fluorescence intensity by summing the fluorescence values of stacked images of the 3-D data, as previously reported using Image-J software (NIH, Bethesda, Maryland, USA)(Noguchi et al., 2011). When the image showed axon fibers overlapping with the target dendrite at different image depths, the spine head volume in the dendrite was calculated by partially summing the fluorescence values of sequential five Z-images by taking the moving average of the image stack along the Z-direction to avoid axonal fibers. A dendritic spine is near the diffraction limit of a two-photon microscope; therefore, the partially summed values (2 μm range in the Z-direction) should contain nearly the entire spine volume. Thus, the maximum value of the Z-moving average images allows good approximation of the total Z-summing of the stacked images.
Dendritic spines have spontaneous fluctuations in fluorescence because of spontaneous morphological changes, motility, or measurement errors. To determine spine volume fluctuations, we calculated the CV of the in vivo images before glutamate uncaging (14.7% ± 16.1% for 227 spines in the enlargement condition and 12.5% ± 7.9% for 196 spines in the shrinkage condition). We set the limit values of the fluctuation of the baseline as 2 CV (i.e., 30% for the enlargement data; 25% for the shrinkage data) and discarded the data when the fluctuation exceeded the limit value. Stimulated spines and neighboring spines with a prestimulation fluctuation over this limit (i.e., unstable spines) were discarded, as were the neighbors of the unstable stimulated spines. For the spine volume analysis, the average spine volume during 10– 30 min (i.e., enlargement) or 20–50 min (i.e., shrinkage) after the onset of the stimulation was calculated and are indicated as the difference from the baseline volume.
Analysis of the spine neck length and the spine length
The 3-D spine neck length was measured manually using Image-J software (Noguchi et al., 2011). An XYZ stack image was resliced at z = 0.1 μm, and an XZ image of the spine neck was created. The intensity profile was measured along the neck in the XZ image. The half-maximum position of the spine and the parent dendrite was the edge of the spine and the dendrite, respectively. The spine neck length was calculated by subtracting the radiuses of the dendrite and the spine from the distance between peaks of the spine and the dendrite (Figure 1B and Figure 2–figure supplement 2B). For the analysis of the spine length before and after stimulation, the length between the tip of the spine and the edge of the dendrite was measured on the Z-stack images (Figure 2– figure supplement 1D).
Statistical analysis
All data are presented as the mean ± standard error of the mean (SEM) (n indicates the number of spines), unless otherwise stated. Statistical tests of the spines were conducted using Excel-Statistics software (Social Survey Research Information Co. Ltd., Tokyo, Japan), as indicated. Differences from the baseline values were analyzed using the Wilcoxon signed-rank test (Figure 1F; Figure 2E; and Figure 2–figure supplement 1A). Differences between groups were analyzed using the Mann–Whitney rank sum test (Figure 2–figure supplement 1B). The Pearson’s product-moment correlation coefficient was calculated for the scatter plots (Figure 1G; Figure 1–figure supplement 2A–D; and Figure 2–figure supplement 2A–2D). The significance of a correlation coefficient was analyzed by using the t-test.
Competing Interests
The authors declare no competing interests.
Author Contributions
J.N. and H.K. are co-corresponding authors and designed the study; J.N. conducted most imaging experiments; A.N., H.U., T.H., S.Y,. and N.T. assisted in some imaging experiments and in the data analysis; J.N. and H.K. wrote the manuscript. All authors contributed to the editing of the paper.
Acknowledgements
We thank C. Maeda, M. Ogasawara, H. Ohno, K. Tamura, M. Nakajima, C. Matsubara and T. Sasaki for their technical assistance. We also thank N. Ichinohe for the helpful discussion and support. This work was supported by Grants-in-Aids for Science Research (S) (grant no. 26221011 to H.K.), Scientific Research (C) (grant numbers 18K06497 and 26430005 to J.N. and grant no. 2640290 to N.T.), Scientific Research on Innovative Areas (grant number 26111706 to J.N. and grant no. 16H06396 to S.Y.) and the Cooperative Research Program of “Network Joint Research Center for Materials and Devices” (grant no. 20171030 to J.N.), and World Premier International Research Center Initiative (awarded to HK) from the Ministry of Education, Culture, Sports, Science and Technology ([MEXT]; Tokyo, Japan), and Core Research of Evolutional Science and Technology ([CREST]; Tokyo, Japan; grant no. JPMJCR1652 to H.K.) from Japan Science and Technology Agency ([JST]; Tokyo, Japan), and the Strategic International Research Cooperative Program (SICP), Brain/MIND, and Strategic Research Program for Brain Sciences projects (grant no. 17dm0107120h0002) from the Agency for Medical Research and Development (awarded to H.K.).