Skip to main content
eLife logoLink to eLife
. 2018 Jun 22;7:e34700. doi: 10.7554/eLife.34700

Chronic 2P-STED imaging reveals high turnover of dendritic spines in the hippocampus in vivo

Thomas Pfeiffer 1,2,, Stefanie Poll 3,, Stephane Bancelin 1,2,, Julie Angibaud 1,2, VVG Krishna Inavalli 1,2, Kevin Keppler 4, Manuel Mittag 3, Martin Fuhrmann 3,‡,, U Valentin Nägerl 1,2,‡,
Editor: Karel Svoboda5
PMCID: PMC6014725  PMID: 29932052

Abstract

Rewiring neural circuits by the formation and elimination of synapses is thought to be a key cellular mechanism of learning and memory in the mammalian brain. Dendritic spines are the postsynaptic structural component of excitatory synapses, and their experience-dependent plasticity has been extensively studied in mouse superficial cortex using two-photon microscopy in vivo. By contrast, very little is known about spine plasticity in the hippocampus, which is the archetypical memory center of the brain, mostly because it is difficult to visualize dendritic spines in this deeply embedded structure with sufficient spatial resolution. We developed chronic 2P-STED microscopy in mouse hippocampus, using a ‘hippocampal window’ based on resection of cortical tissue and a long working distance objective for optical access. We observed a two-fold higher spine density than previous studies and measured a spine turnover of ~40% within 4 days, which depended on spine size. We thus provide direct evidence for a high level of structural rewiring of synaptic circuits and new insights into the structure-dynamics relationship of hippocampal spines. Having established chronic super-resolution microscopy in the hippocampus in vivo, our study enables longitudinal and correlative analyses of nanoscale neuroanatomical structures with genetic, molecular and behavioral experiments.

Research organism: Mouse

Introduction

Dendritic spines form the postsynaptic structural component of most excitatory synapses in the mammalian brain. They constitute computational units of information processing that underlie essentially all higher brain functions (Nimchinsky et al., 2002; Sala and Segal, 2014; Yuste and Bonhoeffer, 2004) and play a crucial role in brain disorders such as autism spectrum disorder and Alzheimer's disease (Dorostkar et al., 2015; Südhof, 2008). Spine structure is closely linked to synapse function, as the size of spine heads scales with synaptic strength (Matsuzaki et al., 2001; Noguchi et al., 2011) and the shape and number of spines can be modified by the induction of synaptic plasticity (Matsuzaki et al., 2001; Engert and Bonhoeffer, 1999; Nägerl et al., 2004; Tønnesen et al., 2014; Zhou et al., 2004) and by sensory experience (Holtmaat et al., 2006; Keck et al., 2011).

Rewiring of neural circuits by spine plasticity is considered a key neurobiological mechanism of memory formation (reviewed in [Nimchinsky et al., 2002; Sala and Segal, 2014; Yuste and Bonhoeffer, 2004; Kasai et al., 2010]). Notably, a recent study showed that optically induced shrinkage of potentiated spines disrupts newly acquired motor skills (Hayashi-Takagi et al., 2015), indicating a causal link between spine plasticity and memory. While experience-dependent plasticity of dendritic spines has been a consistent finding across mouse cortex in vivo (Holtmaat and Svoboda, 2009), very little is known about it in the hippocampus. However, this is a major knowledge gap, because this neural structure constitutes the archetypical memory center of the brain, and hippocampal brain slices and primary cell cultures are the dominant model systems for the study of synaptic plasticity mechanisms.

Imaging dendritic spines in the hippocampus in vivo is challenging because of its remote location more than 1 mm below the surface of the mouse brain. A pioneering study accomplished this with two-photon (2P) microscopy, but only over a period of a few hours (Mizrahi et al., 2004). Recently, approaches based on a ‘hippocampal window’ (Gu et al., 2014) or micro-endoscopy (Attardo et al., 2015) enabled ‘chronic’ 2P imaging over several weeks. However, 2P microscopy inevitably lacks the spatial resolution to visualize important details of spine morphology, such as spine necks (Bethge et al., 2013), and even struggles to resolve individual dendritic spines on CA1 pyramidal neurons, leaving a high fraction of them undetected (Attardo et al., 2015).

While regular light microscopy typically detects a spine density of around 1 spine/µm (Gu et al., 2014; Attardo et al., 2015; Brigman et al., 2010), electron microscopy (EM) reports ~3 spines/µm on CA1 hippocampal pyramidal neurons in rats (Harris et al., 1992) and in stratum radiatum of mice (Bloss et al., 2018). By comparison, spine density is about ten times lower in many cortical areas, for example ~0.24 spines/µm for pyramidal neurons in layer 5 of mouse barrel cortex (Holtmaat et al., 2006).

Given the limited spatial resolution of 2P microscopy, we turned to super-resolution stimulated emission depletion (STED) microscopy (Klar et al., 2000; Hell, 2007) to improve the visualization of dendritic spines in the intact hippocampus of living mice. We used a home-built STED microscope based on 2P excitation (2P-STED) (Bethge et al., 2013; Ter Veer et al., 2017) and equipped it with a long working distance objective to reach the deeply located hippocampus. We adopted a ‘hippocampal window’ technique (Gu et al., 2014; Schmid et al., 2016; Dombeck et al., 2010), where a portion of the overlying somatosensory cortex is surgically removed and replaced by a metal cylinder sealed with a cover slip, providing stable optical access to the CA1 area of the hippocampus.

We demonstrate that our new approach offers substantially improved spatial resolution and image quality compared to regular 2P microscopy in mouse hippocampus in vivo. Using transgenic mice with fluorescently labeled pyramidal neurons, we measured spine density on basal dendrites of pyramidal neurons in stratum oriens of the CA1 area and compared results obtained with 2P and 2P-STED microscopy in vivo as well as with STED microscopy in fixed hippocampal sections. Furthermore, we carried out repetitive 2P-STED in vivo imaging over a 4-day period to measure spine turnover.

Our analysis showed a two times higher spine density than reported by conventional 2P microscopy, and around 40% of all spines turned over within 4 days, suggesting a high level of circuit remodeling in the hippocampus in vivo. Furthermore, detailed morphological analysis revealed that primarily small spines were affected by spine turnover.

Results

2P-STED microscopy with a long working distance objective

We set up in vivo STED microscopy of dendritic spines in mouse hippocampus to track their morphological dynamics over the course of several days. We used a custom-built 2P-STED microscope (Figure 1A) (Bethge et al., 2013; Ter Veer et al., 2017) in combination with a modified ‘cranial window’ technique to gain high-quality optical access to stratum oriens in the CA1 area of the hippocampus (Gu et al., 2014; Dombeck et al., 2010). We surgically removed the overlying somatosensory cortex and inserted a metal tube sealed with a cover slip as a physical place holder (Figure 1B) (Gu et al., 2014; Dombeck et al., 2010). To bridge the distance between the surface of the skull and the alveus located right above the hippocampus, we used an objective with a long working distance yet relatively high numerical aperture (Nikon N60X-NIR: WD 2.8 mm, NA 1.0).

Figure 1. 2P- STED microscopy of dendritic spines in the hippocampus in vivo.

(A) Schematic of the custom-built upright 2P-STED microscope. A Ti:Sapphire laser emits light pulses at 834 nm with 80 MHz repetition rate. The laser pumps an optical parametric oscillator (OPO) to obtain pulsed STED light at 598 nm. A glass rod and a polarization-maintaining fiber (PMF) stretch the STED pulses. The STED doughnut is engineered by a helical 2π phase mask in combination with λ/2 and λ/4 wave plates. The second Ti:Sapphire laser tuned to 900 nm with 80 MHz repetition rate served for two-photon (2P) excitation and is synchronized to the STED light pulses. The 2P and STED beam are combined using a dichroic mirror (DM1). Both beams are swept over the specimen using a galvo-based x-y scanner and a z-focusing device. Scan (SL) and tube lens (TL) image the scan mirrors into the back focal plane of the objective and ensure that the expanded laser beams overfill the back aperture of the objective. Fluorescence is de-scanned and guided to the avalanche photodiode (APD) via a dichroic mirror (DM2) and filters (EF). Electro-optical modulators (EOM) serve for quick adjustments of the beam intensity. (B) Schematic visualizing the combined use of the hippocampal window preparation and a long working distance objective (top). A 2P overview image showing the cell body and basal dendrites of a GFP-labeled pyramidal neuron in hippocampal CA1 in vivo (bottom). Maximum intensity projection (MIP) of ten z-sections with 2 µm z-steps. (C) Quantification of the lateral resolution with 40 nm fluorescent nanospheres. Upper panel: Representative comparison of 40 nm beads imaged by 2P and 2P-STED microscopy. Lower panel: Paired comparison of mean full-width at half maximum (FWHM) of line profiles obtained from 40 nm nanospheres (p<0.0001, paired t-test; n = 17 beads). (D) Representative images obtained from a GFP-labeled dendrite acquired by 2P and 2P-STED microscopy. Insets show line profiles fitted with Lorentzian functions and FWHM values obtained from the indicated dotted lines of the spine neck (MIP of three z-sections). (E) Paired measurements of spine neck widths imaged in 2P and 2P-STED mode (p<0.0001, paired t-test; n = 29 spine necks, 11 dendrites, 6 mice).

Figure 1—source data 1. Data for panel C.
DOI: 10.7554/eLife.34700.006
Figure 1—source data 2. Data for panel B.
DOI: 10.7554/eLife.34700.007
Figure 1—source data 3. Data for Figure 1—figure supplement 1.
DOI: 10.7554/eLife.34700.008

Figure 1.

Figure 1—figure supplement 1. Axial PSF measurement: 2P versus 2P-STED.

Figure 1—figure supplement 1.

(A) 2P and 2P-STED axial PSF. Scale bar: 500 nm. (B) FWHM of the axial PSF measured on 170 nm fluorescent beads in 2P (1200 ± 14 nm) and 2P-STED (1218 ± 24 nm) modes (n = 17 beads; paired t-test; p=0.52).
Figure 1—figure supplement 2. Measurements of spine neck diameters and direct comparison of 2P versus 2P-STED.

Figure 1—figure supplement 2.

(A) 2P-STED image of a dendritic stretch with intensity profiles (orange), Lorentzian fit (white) and FWHM measurements (white horizontal line). Two-pixel line profiles were taken on single sections at the locations indicated by the white dotted lines (for illustrative purposes). FWHM values indicate that 2P-STED microscopy allows for sub-diffraction imaging in the hippocampus in vivo . MIP of three z-sections of a GFP-labeled basal dendrite imaged ~10–15 µm below the coverslip. (B, C, D) Dendrites imaged consecutively in 2P and 2P-STED mode showing the improvement in lateral resolution. MIP of three z-sections.
Figure 1—video 1. Hippocampal window approach afforded high level of sample stability.
Download video file (3.3MB, mp4)
DOI: 10.7554/eLife.34700.005
The hippocampal window preparation offers excellent tissue stability for imaging in the hippocampus of the anesthetized mouse. The 2P image z-stack (100 × 100 × 60 µm3; 200 nm pixel size; 2 µm z-steps) illustrates the virtual absence of motion artifacts, which would impede the acquisition of super-resolved STED images.

To verify that this objective is compatible with STED imaging, we measured the point-spread function (PSF) of the microscope using fluorescent nanospheres (diameter: 40 nm). The full-width at half maximum (FWHM) was 54 ± 2 nm in the x-y plane in 2P-STED mode, compared with 325 ± 5 nm in regular 2P mode (p<0.0001, paired t-test; n = 17 beads; Figure 1C), demonstrating that our super-resolution approach delivered a six-fold nominal gain in lateral spatial resolution. In contrast, the axial resolution remained unchanged in 2P-STED (Figure 1—figure supplement 1).

2P-STED microscopy of dendritic spines in the hippocampus in vivo

We then used the 2P-STED microscope to image dendritic spines on basal dendrites of CA1 pyramidal neurons (stratum oriens) in anesthetized mice. We used transgenic mice (Thy1-HYFP/+ and Thy1-MGFP/+, 4–12 months old) that expressed either GFP or YFP in a subset of CA1 pyramidal neurons (Feng et al., 2000). Focusing on dendrites close to the coverslip (5–20 µm), spines were much more clearly delineated in 2P-STED than in 2P mode (Figure 1D, Figure 1—figure supplement 2). In particular, spine necks were better resolved with 2P-STED as evidenced by paired measurements of spine neck widths (2P-STED: 147 ± 8 nm; 2P: 369 ± 6 nm; p<0.0001, paired t-test; n = 35 spine necks, 17 dendrites, 6 mice; Figure 1E).

Image distortions caused by animal breathing and heartbeat normally pose a serious challenge for imaging in vivo, in particular STED microscopy, which is especially sensitive to brain motion due to its high spatial resolution. However, we only encountered mild levels of image blur (Figure 1—video 1), making it not necessary to apply motion correction (see Discussion).

Dendritic spine density of CA1 pyramidal neurons in vivo

We determined spine density of basal dendrites of CA1 pyramidal neurons, comparing it in images acquired both in 2P-STED and 2P mode (Figure 2A). There was a significant difference between the two modes (2P-STED: 2.13 ± 0.10 µm−1; 2P: 1.61 ± 0.07 µm−1; p<0.0001, paired t-test; n = 82 dendrites, 6 mice; Figure 2B and Figure 2—figure supplement 1), where 2P-STED detected on average 32% more spines than 2P in a direct comparison (Figure 2B). This increase was due to the improved detection or discrimination of (1) spines that either had short necks or that extended into the z-direction, which were otherwise obscured by the fluorescence signal of the dendritic shaft, and (2) closely clustered spines, which appeared merged in the 2P images (Figure 2A).

Figure 2. Density of spines on basal dendrites of CA1 pyramidal neurons in vivo.

(A) CA1 basal dendrite imaged with 2P (top) or 2P-STED (bottom) microscopy. Filled arrowheads highlight spines discerned in 2P-STED, but not 2P mode. Open arrowheads indicate spines that could only be visualized in 2P-STED mode as they were otherwise masked by the blurry fluorescence of the dendrite in the 2P image (MIP of two z-sections). (B) Measured spine densities in consecutive 2P and 2P-STED acquisitions of the same dendrites (p<0.0001, paired t-test; n = 82 basal dendrites, 6 mice). (C) Image of a basal dendrite obtained from fixed brain tissue acquired on a confocal STED. (D) Dendritic spine density on basal dendrites in fixed hippocampal tissue (n = 37 basal dendrites, 6 mice). (E) Geometric model to extrapolate the spine density in 3D. Spines cannot be detected when they are inside the ‘blind zone’, depending on the dimensions of the morphology and microscope PSF. (F) Extrapolation of spine density in 3D following the model in (E) for in vivo (black line) and fixed tissue measurements (red line).

Figure 2—source data 1. Data for panel B.
DOI: 10.7554/eLife.34700.012
Figure 2—source data 2. Data for panel D.
DOI: 10.7554/eLife.34700.013

Figure 2.

Figure 2—figure supplement 1. Methodology of dendritic spine density analysis in vivo.

Figure 2—figure supplement 1.

(1) MIP of a basal dendrite acquired using 2P-STED in vivo. (2 - 9) Identification of dendritic spines in single z-planes. Magenta arrowheads indicate spines and magenta outlines highlight the spine shape. (10) Maximum intensity projection (MIP) of identified spines. (11) Measurement of dendritic length. (12) Overlay of (10) and (11) and analysis of spine density.
Figure 2—video 1. z-stack of a fixed dendrite imaged with one-photon STED microscopy.
Download video file (236.5KB, mp4)
DOI: 10.7554/eLife.34700.011
STED z-stack of a fixed basal dendrite of CA1 pyramidal neuron acquired on commercial one-photon STED microscope (20 nm pixel size, 200 nm z-steps).

While our 2P-based mean value for spine density (1.61 µm−1) is around 45% higher than the values reported in the light microscopy literature (~1.1 µm−1) (Gu et al., 2014; Attardo et al., 2015), the mean value based on 2P-STED is about twice as large as this (2.13 µm−1). We also counted spines in fixed brain slices obtained from the same animals (Figure 2C and Figure 2—video 1). Using a commercial STED microscope with one-photon excitation, we measured a density of 2.68 ± 0.08 µm−1 (n = 37 dendrites, 6 mice; Figure 2D), which comes closer to the values reported by EM.

To understand where the remaining discrepancy might come from, we made a simple geometrical model to account for the limited axial resolution of our STED approach, which had left some spines invisible (Figure 2E). According to the model, which took into account the actual dimensions of dendritic morphology and the PSF of the microscope, about a quarter of the spines could not be detected in vivo because of this problem. Correcting the measured values by this fraction, we calculated a spine density of 2.91 ± 0.14 µm−1 for the in vivo data and 3.15 ± 0.09 µm−1 for the fixed tissue data (Figure 2F), which effectively closed the gap to the ‘ground truth’ EM values.

Dendritic spines undergo high morphological turnover in vivo

To determine the kinetics of spine emergence and elimination, which is a matter of controversy (Gu et al., 2014; Attardo et al., 2015), we performed repeated 2P-STED imaging of individual dendritic stretches over 4 days (day 0, 2 and 4; Figure 3A). To retrieve a particular dendrite in consecutive imaging sessions, we used tissue landmarks such as blood vessels, fluorescent cell bodies and dendrites.

Figure 3. Turnover of spines on basal dendrites of CA1 pyramidal neurons in vivo.

(A) Repetitive imaging of basal dendrites in CA1 area using 2P-STED microscopy. The upper panel shows low-magnification overviews containing the dendrite of interest highlighted with a white box. The lower panel shows the corresponding 2P-STED images of the dendrite over time. The images represent single z-planes. Dendritic spines with blue arrowheads were stable between imaging sessions. Red arrowheads mark lost spines, and green arrowheads mark new ones. The axonal bouton (AB) is marked by a white arrowhead. The numbering of spines is continuous. (B) Quantification of spine density over 4 days (n = 14 dendrites, 3 mice). (C) Quantification of the 4-day survival fraction of dendritic spines. (D) Fraction of lost spines and (E) fraction of new spines. Thin grey lines represent the measurements of single dendrites.

Figure 3—source data 1. Source data of the parameters underlying Figure 3 extracted from the turnover data set.
DOI: 10.7554/eLife.34700.017
Figure 3—source data 2.
DOI: 10.7554/eLife.34700.018
Figure 3—source data 3. Data for Figure 3—figure supplement 1.
DOI: 10.7554/eLife.34700.019

Figure 3.

Figure 3—figure supplement 1. 2P vs 2P-STED measurement of spine turnover in vivo.

Figure 3—figure supplement 1.

(A) Quantification of the 4-day survival fraction in 2P and 2P-STED (n = 14 dendrites, 3 mice). (B) Fraction of lost spines and (C) fraction of new spines in 2P and 2P-STED (n = 14 dendrites, 3 mice). Thin grey lines represent the measurement of single dendrites. In all panels, orange and black lines represent 2P and 2P-STED, respectively.
Figure 3—figure supplement 2. 2P-STED time-lapse imaging of hippocampal dendrites in vivo.

Figure 3—figure supplement 2.

2P-STED could be used for time-lapse imaging of hippocampal dendrites as bleaching of the volume-filling fluorophore was negligible and no apparent phototoxicity was induced. MIP of three z-sections of a YFP-labeled dendrite imaged every 10 min.

First, we quantified the average spine density for each time point, which remained stable over the entire observation period (day 0: 2.31 ± 0.10 µm−1; day 2: 2.28 ± 0.08 µm−1; day 4: 2.30 ± 0.10 µm−1; n = 14 dendrites, 3 mice; Figure 3A,B and Figure 3—source data 1).

Next, we examined spine turnover by counting all spines that appeared and disappeared from one imaging session to the next. Furthermore, we calculated the ‘survival fraction’, which is the percentage of spines that were present on day 0 and visible during the subsequent imaging sessions. The survival fraction was 78.1 ± 3.6% for day 2, and 60.8 ± 3.4% for day 4 (Figure 3C).

We then calculated the fraction of spines lost and gained between the imaging sessions. The fraction of lost spines was 21.2 ± 3.6% from day 0 to 2, and 24.7 ± 3.0% from day 2 to 4 (Figure 3D), while the fraction of new spines from day 0 to 2 was 20.4 ± 2.6%, and 21.9 ± 3.0% from day 2 to 4 (Figure 3E). These results indicate that the rates of spine loss and gain were balanced and did not change over the sessions, which is consistent with the constant spine density we observed.

By comparison, spine turnover was less visible in 2P mode, amounting to a survival fraction of 74.9 ± 2.2% after 4 days (Figure 3—figure supplement 1).

Spine turnover affects primarily small spines

Given the detailed morphological information in the 2P-STED images, we examined if there was a relationship between spine turnover and morphology. In cortex, spines with a large head have been shown to be more stable than filopodial and thin spines (Holtmaat and Svoboda, 2009; Knott et al., 2006), but it is unknown whether such a link also exists for the hippocampus.

After deconvolution of the images and 3D reconstruction (see Materials and methods), we analyzed the morphology of all spines that were visible on any of the imaging sessions (i.e. on days 0, 2 and/or 4; Figure 4), dividing the spines into two groups according to their ‘observed persistence’. Spines that were visible on all three imaging sessions (Persistence >2 days) had larger heads than spines that were visible on only one or two sessions (Persistence ≤2 days), suggesting that small spines are more short-lived than large ones (>2 days: 0.03 μm3 and 0.01–0.06 μm3 versus ≤ 2 days: 0.02 μm3 and 0.01–0.04 μm3, median and interquartile range; Figure 4A,B; p<0.0001).

Figure 4. Structure-dynamics relationship of hippocampal spines.

(A) 3D reconstruction of a dendrite imaged on days 0, 2 and 4. Spines persisting for more than 2 days (#0–8, blue), and 2 days or less (#9–20, salmon) are illustrated. (B) Spine head volumes measured on reconstructed dendrites (p<0.0001, Mann-Whitney test; n = 14 dendrites, 3 mice; box plot shows median and 10, 25, 75 and 90th percentiles). (C, D) 3D morphology plots visualizing the populations of spines observed persistent for more than 2 days and 2 days or less (C), and their affiliation to identified clusters 1, 2 and 3 (D) (see also Figure 4—figure supplement 1A). Plotted are, the ratio of mean head to neck diameters (ØHead/ØNeck), spine length and maximum head diameter (Ømax Head). (E) Quantification of spine proportions within identified clusters, distinguishing spines of different persistence (>2 days versus ≤2 days). (F) Table summarizing the morphological parameters utilized for cluster analysis: ØHead/ØNeck, Ømax Head and length of spines, for spines that persist for >2 days (blue) and ≤2 days (salmon). Data are represented as median and interquartile range (25th–75th percentile). Significant differences are marked by asterisks (***p<0.001, ****p<0.0001, unpaired t-test or Mann-Whitney test; for all comparisons see Figure 4—figure supplement 1B; n = 14 dendrites, 3 mice).

Figure 4—source data 1. Underlying data for Figure 4.
elife-34700-fig4-data1.xlsx (109.7KB, xlsx)
DOI: 10.7554/eLife.34700.022

Figure 4.

Figure 4—figure supplement 1. Cluster analysis of spine morphology.

Figure 4—figure supplement 1.

(A) Dendrogram displaying the Euclidean distances of individual spines and their respective assignment to one of the three identified clusters. (B) Statistical comparison of morphological parameters between spines of different observed persistence (>2 days versus ≤2 days; n = 14 dendrites, 3 mice).

Refining this analysis, we plotted key morphological parameters (dendritic spine length, maximum head diameter, ratio of mean head to neck diameters) in three dimensions. The 3D plot revealed a clear difference in the distributions of the morphological parameters depending on the observed persistence of the spine (Figure 4C). A cluster analysis (‘Agglomerative Hierarchical Clustering’ analysis based on Euclidian distance in the parameter space, see Materials and methods) revealed at least three distinct populations of spines (Figure 4—figure supplement 1), which resemble the different categories commonly used in the literature to classify dendritic spines (‘small’ ≅ cluster 1, ‘thin’ ≅ cluster 2 and ‘mushroom-like’ ≅ cluster 3; Figure 4D). The latter category is generally composed of long spines with high head to neck diameter ratios. Notably, whereas around 50% of more persistent spines exhibited a mushroom-like morphology, less persistent spines rarely (7%) belonged to this category (Figure 4E). Even within cluster 3, less persistent spines had on average smaller heads and lower head to neck diameter ratios (Figure 4F).

Discussion

We established chronic super-resolution imaging of dendritic spines in the intact hippocampus of living mice. It is based on a home-built upright 2P-STED microscope (Bethge et al., 2013; Ter Veer et al., 2017) equipped with a long working distance water immersion objective and a ‘hippocampal window’ to reach this deeply embedded brain structure (Gu et al., 2014; Dombeck et al., 2010).

STED microscopy has been used before for imaging dendritic spines in vivo, but only in superficial cortical layers and for one-off and acute imaging sessions (Berning et al., 2012; Willig et al., 2014). Considering the importance of the hippocampus for learning and memory, it is of great interest to develop and improve approaches for faithful detection and monitoring of dendritic spines there.

Dendritic spines commonly serve as a morphological proxy for excitatory synapses, allowing inferences on synaptic strength and functional connectivity, for instance when their shape, size or number change over the course of an electrophysiological or behavioral experiment (Yuste and Bonhoeffer, 2004; Holtmaat and Svoboda, 2009). However, since spines have nanoscale dimensions and are densely packed in light scattering tissue, an accurate and quantitative anatomical readout is oftentimes difficult to obtain (Bethge et al., 2013).

Spine density has been shown to vary between 2 and 4 µm−1 on proximal apical dendrites in stratum radiatum in adult rats (Harris et al., 1992; Harris and Stevens, 1989) and mice (Bloss et al., 2018) according to serial section transmission EM, and between 2 and 3 µm−1 on basal oblique dendrites in stratum oriens in adult mice according to array tomography (Bloss et al., 2016). By comparison, the spine densities we detected in stratum oriens lie within the range of these measurements, while the two recent in vivo 2P microscopy studies reported values less than half the amount (Gu et al., 2014; Attardo et al., 2015).

As our STED approach only improved the lateral resolution, spines that protruded into the axial direction were still difficult to detect, explaining why our density values are still below those reported by EM. Indeed, correcting our data based on a simple geometric model removed the residual difference. In the near future, our method stands to benefit from ongoing technical advances, for instance in 3D beam shaping, which will improve axial resolution (Gould et al., 2012; Patton et al., 2016), and should enable nearly complete spine detection in vivo.

We could only image spines in the stratum oriens within 20 µm from the cover slip; beyond this distance, image quality degraded rapidly. This depth limitation was likely due to optical aberrations caused by mismatches in refractive index between the immersion medium, the glass cover slip and the sample (n ≈ 1.37 for brain tissue [Lue et al., 2007]). Unlike other objectives we have used for STED microscopy in brain tissue (Bethge et al., 2013; Urban et al., 2011), the new objective did not have a correction collar to reduce spherical aberrations, which otherwise probably would have permitted greater imaging depth. However, the use of adaptive optics is bound to allow for deeper imaging (Ji, 2017) and thus to reach dendrites in stratum radiatum and other areas of the hippocampus.

We suspect that the near total absence of motion artifacts in the images was primarily due to two effects: firstly, the implanted metal tube probably stabilized the brain mechanically, and secondly, blood pulsations are likely much weaker in the hippocampus where the vasculature is mostly formed by capillaries and has fewer large arteries than in superficial cortex (Marinković et al., 1992).

Faithful and complete detection of spines across space and time is absolutely critical for an accurate assessment of spine turnover. Failure to distinguish closely spaced spines will lead to an underestimate of spine turnover, because events of individual spines appearing or disappearing within a merged cluster will inevitably be missed, giving a false impression of spine stability (Attardo et al., 2015). Conversely, temporal fluctuations, for instance when spines rotate in or out of the optical axis, will lead to an overestimate of spine turnover. Moreover, biased detection of large spines might distort the measurements of spine turnover as large spines are reportedly less structurally plastic than small spines (Matsuzaki et al., 2004), as we have also shown here.

The spine detection problem was highlighted by two recent studies that pioneered chronic imaging in the hippocampus in vivo (Gu et al., 2014; Attardo et al., 2015) and reported very different lifetimes for spines on apical and basal dendrites of CA1 pyramidal neurons. The study by Gu et al. (Gu et al., 2014) reported that ~96% of spines survived for at least 16 days in stratum radiatum, whereas the study by Attardo et al. (Attardo et al., 2015) predicted that the entire population of spines in stratum oriens turned over within a month.

Both studies used 2P microscopy with relatively low NA optics and presented images of comparable quality, reporting similar average spine densities (~1.1 µm−1) that were stable over time. While Gu et al. assessed spine turnover directly from the 2P images, Attardo et al. went a step further and used mathematical modeling to account for the effect of merged spines on apparent spine turnover. Importantly though, without applying the mathematical correction, their survival fraction was around 80% after 21 days, indicating that direct analysis of 2P images is highly problematic when it comes to establishing spine turnover rates.

Our super-resolution approach documented a spine survival fraction of 60% after 4 days, providing direct and unambiguous evidence that spines can turn over extremely rapidly in CA1 stratum oriens, supporting one of the main modeling-based conclusions of the Attardo et al. study. However, since we imaged just three time points over a relatively short period, it is hard to extrapolate our results in time and determine the extent to which the observed kinetics hold for the entire spine population. In fact, our observation that larger spines were more stable is a sign for the existence of multiple, kinetically distinct spine populations, unlike the Attardo et al. study, which argued for a single population.

We did not observe any overt signs of phototoxicity, such as dendritic ‘blebbing’ during short time-lapse sequences and across the imaging sessions (Figure 3A, Figure 3—figure supplement 2). This absence together with the fact that spine density remained constant indicates that the observed kinetics were real and not an artifact induced by our hippocampal window approach or 2P-STED imaging. The use of the NMDA-receptor antagonist ketamine could be problematic given the important role of NMDA-receptors in hippocampal synaptic plasticity. However, as the drug was applied only for a couple of hours per imaging session and we did not induce any plasticity, it is unlikely that our results were compromised by it. The use of more innocuous drugs or performing the experiments with non-anesthetized and head-fixed animals will be preferable in the future.

As we did not image in the stratum radiatum, we do not know to what extent the divergence in spine turnover between stratum radiatum reported by Gu et al. and stratum oriens (Attardo et al. and our data) reflects methodological or genuine anatomical or physiological differences. In fact, stratum radiatum primarily receives input from CA3 (Somogyi, 2010; Cappaert et al., 2014), whereas stratum oriens also has afferents from entorhinal cortex and amygdala. In addition, there may be intrinsic differences in the postsynaptic neuron, which account for dendrite-specific spine turnover.

Our finding that hippocampal dendrites are subject to ongoing intense anatomical remodeling supports the view of the hippocampus as a highly dynamic structure designed to encode and process new memories, but not as a long-term repository of information (Frankland and Bontempi, 2005). The situation may be quite different in cortical areas, where typically a large fraction of spines is stable for many weeks in adult mice (Grutzendler et al., 2002; Majewska et al., 2006; Trachtenberg et al., 2002), unless they are subjected to behavioral training (Xu et al., 2009; Yang et al., 2009) or sensory manipulations (Hofer et al., 2009; Keck et al., 2008). However, whether and how spine turnover actually reflects memory-relevant functional adaptations in synaptic strength and connectivity remains to be determined.

In summary, the present work adds up to a substantial advance for the study of hippocampal synapses in living mice, extending the scope of super-resolution microscopy to a deeply embedded brain structure that is critical for memory function. By correlating spine-level structural changes with genetic, molecular and behavioral interventions and assays, our chronic super-resolution imaging approach creates manifold opportunities to study the neurobiological mechanisms and functional significance of spine plasticity in the mammalian brain.

Materials and methods

Animals

We used adult female and male transgenic mice (Thy1-HYFP/+ and Thy1-MGFP/+, 4–12 months old) where a subset of pyramidal neurons is fluorescently labeled (Feng et al., 2000). The mice were group housed by gender at a day/night cycle of 12/12 hr. All procedures were in accordance with the Directive 2010/63/EU of the European Parliament and approved by the Ethics Committee of Bordeaux and by the government of North Rhine Westphalia.

Hippocampal window surgery

Chronic hippocampal windows were implanted as described previously (Gu et al., 2014; Schmid et al., 2016), providing optical access to the stratum oriens of the CA1 region of the hippocampus (Figure 1B). In brief, mice were anesthetized with an i.p. injection of ketamin/xylazine (0.13/0.01 mg/g bodyweight) and received subcutaneous injections of analgesics (buprenorphine, 0.05 mg/kg) and anti-inflammatory agents (dexamethasone, 0.2 mg/kg) to prevent the brain from swelling during the surgical procedure. Using a dental drill, a craniotomy of 3 mm in diameter was performed above the right hemisphere (stereotactic coordinates: anteroposterior, −2.2 mm; mediolateral, +1.8 mm relative to bregma). The dura and somatosensory cortex above the hippocampus were carefully aspirated, while leaving the external capsule of the hippocampus intact. Subsequently, a custom-made metal tube sealed with a coverslip on the bottom side (both 3 mm in diameter, height 1.5 mm and 0.13 mm) was inserted into the craniotomy and fixed to the skull with dental acrylic. Following the surgery, mice received analgesics for 3 days (buprenorphine, 0.05 mg/kg, s.c.) and were allowed to recover from the surgery for 4 weeks before experiments started.

Two-photon STED microscopy

We used a custom-made upright 2P-STED microscope (Figure 1A) based on two-photon excitation and stimulated emission depletion (STED) using pulsed lasers (Bethge et al., 2013; Ter Veer et al., 2017). Briefly, a femtosecond mode-locked Ti:Sapphire laser (Chameleon, Coherent, Santa Clara, CA) operating at 80 MHz and emitting light at 834 nm was used in combination with an optical parametric oscillator (OPO BASIC Ring fs, APE, Berlin, Germany) to produce STED light pulses at 598 nm with ~150 fs pulse duration. The pulses were stretched to ~100 ps by passing them through a glass rod and a 20 m long polarization-maintaining fiber. A STED light reflection served to synchronize a second Ti:Sapphire laser (Tsunami, Spectra Physics, Darmstadt, Germany) tuned to 910 nm and running at a repetition rate of 80 MHz, which was used for two-photon excitation of the fluorophores. Synchronization and optimal pulse delay were achieved with phase-locked loop electronics (3930, Lok-to-Clock, Spectra Physics). The STED doughnut was created by passing the STED beam through a vortex phase mask (RPC Photonics, Rochester, NY), which imposed a helical 2π-phase delay on the wave front. Wave plates (λ/2 and λ/4) were used to make the STED light circularly polarized before it entered into the objective. The 2P and STED beam were combined using a long-pass dichroic mirror. The two laser beams were moved over the sample in all three dimensions using a galvanometric x-y scanner (Janus IV, TILL Photonics) combined with a z-focusing piezo actuator (Pifoc, PI, Karlsruhe, Germany). To bridge the physical distance between the surface of the brain and the deeply embedded hippocampus, we employed a long-working distance water-immersion objective (Nikon CFI Apo 60X W NIR, 1.0 NA, 2.8 mm WD). The epi-fluorescence was de-scanned and imaged onto an avalanche photodiode (SPCM-AQR-13-FC, PerkinElmer, Villebone-sur-Yvette, France). Signal detection and peripheral hardware were controlled by the Imspector scanning software (Abberior Instruments, Göttingen, Germany) via a data acquisition card (PCIe-6259, National Instruments). Optical resolution was assessed by imaging 40 nm fluorescent nanospheres (yellow-green fluospheres, Invitrogen) immobilized on glass slides (Figure 1C). Regions of interest were consecutively imaged in 2P and 2P-STED mode (10 × 10 µm2; 10 nm pixel size; 50 µs pixel dwell time). The laser power at the sample was typically around 5–20 mW for 2P and 5–15 mW for STED.

In vivo imaging

Mice were anesthetized with an i.p. injection of ketamin/xylazine (0.13/0.01 mg/g bodyweight) and the eyes were protected with ointment (Bepanthen). The mouse was fixed to a custom-made stereotactic frame and kept at body temperature using a heating pad. For the quantification of spine density, 2P-STED and 2P images of basal dendrites of CA1 pyramidal neurons were acquired applying identical acquisition parameters (10 × 10 × 4–8 µm3; 20–40 nm pixel size; 0.5–1 µm z-step; 20–70 µs pixel dwell time). For the chronic repetitive imaging, the position of the field of view (FOV) was registered in the first imaging session with the help of vascular landmarks and cell bodies of CA1 pyramidal neurons. This allowed for subsequent retrieval of the FOV for each mouse. An overview 2P image z-stack (100 × 100 × 20 µm3; 200 nm pixel size; 1 µm z-steps) was acquired at each time point starting from the coverslip. Subsequently, identified distal stretches of basal dendrites in CA1 stratum oriens (5–20 µm in depth) were imaged in 2P and 2P-STED mode. A single imaging session lasted for ~1 hr and mice woke up in their home cage afterwards. All mice survived the imaging sessions and recovered normally from the anesthesia. Mice were excluded from the analysis if the hippocampal window was faulty or if their fluorescence levels were too low.

Image analysis

The 2P and 2P-STED images were spatially filtered (1 pixel median filter) and their brightness and contrast were individually adjusted. Optical resolution of the 2P-STED microscope was assessed by taking images of 40 nm fluorescent nanospheres. We measured two-pixel line profiles across the nanospheres and perpendicular to the spine neck, and fitted them with a Lorentzian function, whose full-width at half maximum (FWHM) served as a measure of the spatial resolution or the neck width, respectively. Paired spine neck measurements were only done for spines where the necks could be discerned in both the 2P and 2P-STED images. Spine density was determined as described before (Gu et al., 2014; Fuhrmann et al., 2007; Holtmaat et al., 2005). All dendritic spines, which extended out laterally from the dendritic shaft by more than 200 nm, were counted by manually scrolling through the z-stacks. Spine density was calculated as the number of spines divided by the dendritic length in micrometer (Figure 2—figure supplement 1). For the 4-day repetitive imaging data set, the spine density of each dendrite was determined blindly with respect to the time point of acquisition. Lost and new spines over the 4-day interval were identified by scrolling through the z-stacks in a chronological order. Spines were scored lost, if they were under the 200 nm threshold. Spines were considered new, if their position on the dendrite relative to neighboring spines shifted by ≥500 nm (Gu et al., 2014), (Holtmaat et al., 2005). The survival fraction of spines (Fs) was calculated as the number of remaining spines at day t (Nr(t)) divided by the number of spines at day 1 (N(1)), expressed in percent:

Fs=Nr(t)N(1)×100 (1)

The fraction of lost spines (Flost) was assessed for each time point by dividing the number of spines that disappeared on day t (NL(t)) by the total number of spines (N(t)):

Flost=NL(t)N(t)×100 (2)

The fraction of new spines (Fnew) was assessed for each time point by dividing the number of spines that appeared on day t (Nn(t)) by the total number of spines (N(t)):

Fnew=Nn(t)N(t)×100 (3)

Reconstruction of dendritic segments

To enhance contrast, raw 2P-STED image stacks were subjected to deconvolution (Huygens HuCore version 17.4, Scientific Volume Imaging b.v., Hilversum, The Netherlands) utilizing a theoretical PSF based on microscope parameters and classic maximum likelihood estimation (cmle) with a quality stop criterion of 0.01, automatic background estimation and a signal-to-noise ratio (SNR) of 15.

After deconvolution, image stacks were 3D reconstructed in IMARIS v 8.4.1 (Bitplane AG, Zurich). Dendrites and associated spines were reconstructed semi-automatically using the Filament Tracer module. Morphological parameters of spines (mean head width, mean neck width, maximum head width, spine head volume and spine length) were measured and exported for further analysis. To quantify persistence, spines were manually traced over time by assigning individual spines of consecutive imaging sessions to each other. Morphological parameters of spines that were present in all three imaging sessions (persistence >2 days) were measured on the last imaging time point. Morphological parameters of spines that were observed for 2 days or less (persistence ≤2 days) were measured at the time point of first appearance (new spines), or at the last time point of presence (lost spines).

Cluster analysis

For the cluster analysis, the spine parameters, ratio of the mean head to neck diameters (ØHead/ØNeck), the maximum head width (Ømax Head) and spine length were taken into account. Agglomerative Hierarchical Clustering Analysis (AHC) was performed using Python’s scikit-learn toolbox. Before cluster analysis, the parameters were centered and scaled to unit variance using the standard scaler from Python’s scikit-learn toolbox. AHC dissimilarity level was calculated based on Euclidean distance. Agglomeration was performed using Ward’s method. By fitting the hierarchical clustering to the population of all analyzed spines, we identified three clusters.

Tissue preparation and immunochemistry

All mice were injected with a lethal dose of pentobarbital (200 mg/kg, i.p., Centravet) and perfused transcardially with saline solution followed by 4% (w/v) paraformaldehyde in 0.1 M Phosphate buffer. Brains were removed, post-fixed in 4% paraformaldehyde for 6–8 hr, and then sectioned at 40 μm in the coronal plane on a vibratome (VT1200, Leica). Free floating sections were blocked and permeabilized with a blocking buffer containing 5% normal goat serum (NGS) and 0.5% TritonX-100 diluted in PBS for 1 hr at room temperature. They were then incubated with the rabbit anti-GFP (polyclonal serum 1/1000, Invitrogen) diluted in PBS containing 0.1% tween and 1% NGS for 24 hr at 4°C; and with the photostable Atto647N-conjugated goat anti-rabbit secondary antibody for 45 min at room temperature (2 µg/ml, Sigma-Aldrich). The sections were mounted directly on coverslips (high-performance coverglass D = 0.17 ±0.005 mm refractive index = 1.526, Zeiss) using Mowiol (Mowiol 4–88 Calbiochem #475904, refractive index = 1.460) for imaging.

Fixed tissue imaging

STED images of fixed brain slices were acquired on a commercial STED microscope (Leica DMI6000 TCS SP8 X, Leica Microsystems, Mannheim, Germany), using a 93X glycerol objective with a numerical aperture of 1.3 that was equipped with a motorized correction collar. The microscope was supplied with a white light laser 2 (WLL2) with freely tunable excitation from 470 to 670 nm. The STED module used a pulsed laser for depletion at 775 nm. Image stacks of basal dendrites of CA1 pyramidal neurons were acquired with a pixel size of 20 nm, a z-step size of 200 nm and at a scan speed of 200 Hz using three line averages and four frame accumulations. Spine density analysis was performed as described for the in vivo image analysis (see Image Analysis).

Quantification and statistics

Quantifications and statistical analysis were performed using Microsoft Excel and GraphPad Prism 7 (GraphPad Software, Inc.). All data were tested for normal distribution using the Shapiro-Wilk normality test. For normally distributed data t-tests, either paired (data of Figures 13) or unpaired (Figure 4F), were performed to test for statistical significance. Not normally distributed data were tested with Mann-Whitney test (Figure 4BFigure 4FFigure 4—figure supplement 1B). All values are represented as mean ± sem, unless stated otherwise.

Acknowledgements

This work was supported by the DZNE, grants from the Deutsche Forschungsgemeinschaft (SFB1089; C01 and B06), CoEN (CoEN3018) and the EU (ERA-NET MicroSynDep) to MF, the Fondation pour la Recherche Médicale (DEQ20160334901) and CoEN (ANR-16-COEN-0003–02) to UVN. TP was supported by a Boehringer Ingelheim Fonds PhD fellowship and a PhD extension grant from the Fondation pour la Recherche Médicale (634FDT20160435677). We thank Mirelle ter Veer for work on the STED microscope, Steffen Burgold for his support with IMARIS, all lab members for comments on the manuscript and the DZNE Light Microscope Facility (LMF), Image and Data Analysis Facility (IDAF) and Animal Research Facility (ARF) for technical support. Fixed tissues imaging was performed at the Bordeaux Imaging Center (BIC), a service unit of the CNRS-INSERM and Bordeaux University, member of the national infrastructure France BioImaging. We thank Patrice Mascalchi (BIC) for technical support.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Martin Fuhrmann, Email: martin.fuhrmann@dzne.de.

U Valentin Nägerl, Email: valentin.nagerl@u-bordeaux.fr.

Karel Svoboda, Janelia Research Campus, Howard Hughes Medical Institute, United States.

Funding Information

This paper was supported by the following grants:

  • Fondation pour la Recherche Médicale DEQ20160334901 to U Valentin Nägerl.

  • COEN ANR-16-COEN-0003-02 to U Valentin Nägerl.

  • COEN CoEN3018 to Martin Fuhrmann.

  • Deutsche Forschungsgemeinschaft SFB1089 to Martin Fuhrmann.

  • ERA-Net MicroSynDep to Martin Fuhrmann.

  • Boehringer Ingelheim Fonds Graduate Student Fellowship to Thomas Pfeiffer.

  • Fondation pour la Recherche Médicale Graduate Student Fellowship 634FDT20160435677 to Thomas Pfeiffer.

Additional information

Competing interests

No competing interests declared.

Author contributions

Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing.

Formal analysis, Investigation, Writing—review and editing.

Investigation, Methodology, Writing—review and editing.

Investigation, Methodology, Writing—review and editing.

Investigation, Methodology, Writing—review and editing.

Resources.

Software, Methodology, Writing—review and editing.

Conceptualization, Supervision, Writing—original draft, Project administration, Writing—review and editing.

Conceptualization, Supervision, Funding acquisition, Writing—original draft, Project administration, Writing—review and editing.

Ethics

Animal experimentation: All procedures of this study were in accordance with the Directive 2010/63/EU of the European Parliament and approved by the Ethics Committee of Bordeaux and by the government of North Rhine Westphalia.

Additional files

Transparent reporting form
DOI: 10.7554/eLife.34700.023

Data availability

Source data files have been provided for Figures 1, 2, 3 & 4 and Figure 1-figure supplement 1 and Figure 3-figure supplement 1.

References

  1. Attardo A, Fitzgerald JE, Schnitzer MJ. Impermanence of dendritic spines in live adult CA1 hippocampus. Nature. 2015;523:592–596. doi: 10.1038/nature14467. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Berning S, Willig KI, Steffens H, Dibaj P, Hell SW. Nanoscopy in a living mouse brain. Science. 2012;335:551. doi: 10.1126/science.1215369. [DOI] [PubMed] [Google Scholar]
  3. Bethge P, Chéreau R, Avignone E, Marsicano G, Nägerl UV. Two-photon excitation STED microscopy in two colors in acute brain slices. Biophysical Journal. 2013;104:778–785. doi: 10.1016/j.bpj.2012.12.054. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bloss EB, Cembrowski MS, Karsh B, Colonell J, Fetter RD, Spruston N. Structured Dendritic Inhibition Supports Branch-Selective Integration in CA1 Pyramidal Cells. Neuron. 2016;89:1016–1030. doi: 10.1016/j.neuron.2016.01.029. [DOI] [PubMed] [Google Scholar]
  5. Bloss EB, Cembrowski MS, Karsh B, Colonell J, Fetter RD, Spruston N. Single excitatory axons form clustered synapses onto CA1 pyramidal cell dendrites. Nature Neuroscience. 2018;21:353–363. doi: 10.1038/s41593-018-0084-6. [DOI] [PubMed] [Google Scholar]
  6. Brigman JL, Wright T, Talani G, Prasad-Mulcare S, Jinde S, Seabold GK, Mathur P, Davis MI, Bock R, Gustin RM, Colbran RJ, Alvarez VA, Nakazawa K, Delpire E, Lovinger DM, Holmes A. Loss of GluN2B-containing NMDA receptors in CA1 hippocampus and cortex impairs long-term depression, reduces dendritic spine density, and disrupts learning. Journal of Neuroscience. 2010;30:4590–4600. doi: 10.1523/JNEUROSCI.0640-10.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Cappaert NLM. Van Strien NM. Witter MP. In: The Rat Nervous System. fourth ed. Paxinos G, editor. Burlington: Elsevier Science; 2014. p. 1053. [Google Scholar]
  8. Dombeck DA, Harvey CD, Tian L, Looger LL, Tank DW. Functional imaging of hippocampal place cells at cellular resolution during virtual navigation. Nature Neuroscience. 2010;13:1433–1440. doi: 10.1038/nn.2648. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Dorostkar MM, Zou C, Blazquez-Llorca L, Herms J. Analyzing dendritic spine pathology in Alzheimer's disease: problems and opportunities. Acta Neuropathologica. 2015;130:1–19. doi: 10.1007/s00401-015-1449-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Engert F, Bonhoeffer T. Dendritic spine changes associated with hippocampal long-term synaptic plasticity. Nature. 1999;399:66–70. doi: 10.1038/19978. [DOI] [PubMed] [Google Scholar]
  11. Feng G, Mellor RH, Bernstein M, Keller-Peck C, Nguyen QT, Wallace M, Nerbonne JM, Lichtman JW, Sanes JR. Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP. Neuron. 2000;28:41–51. doi: 10.1016/S0896-6273(00)00084-2. [DOI] [PubMed] [Google Scholar]
  12. Frankland PW, Bontempi B. The organization of recent and remote memories. Nature Reviews Neuroscience. 2005;6:119–130. doi: 10.1038/nrn1607. [DOI] [PubMed] [Google Scholar]
  13. Fuhrmann M, Mitteregger G, Kretzschmar H, Herms J. Dendritic pathology in prion disease starts at the synaptic spine. Journal of Neuroscience. 2007;27:6224–6233. doi: 10.1523/JNEUROSCI.5062-06.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Gould TJ, Burke D, Bewersdorf J, Booth MJ. Adaptive optics enables 3D STED microscopy in aberrating specimens. Optics Express. 2012;20:20998–21009. doi: 10.1364/OE.20.020998. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Grutzendler J, Kasthuri N, Gan WB. Long-term dendritic spine stability in the adult cortex. Nature. 2002;420:812–816. doi: 10.1038/nature01276. [DOI] [PubMed] [Google Scholar]
  16. Gu L, Kleiber S, Schmid L, Nebeling F, Chamoun M, Steffen J, Wagner J, Fuhrmann M. Long-term in vivo imaging of dendritic spines in the hippocampus reveals structural plasticity. Journal of Neuroscience. 2014;34:13948–13953. doi: 10.1523/JNEUROSCI.1464-14.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Harris KM, Jensen FE, Tsao B. Three-dimensional structure of dendritic spines and synapses in rat hippocampus (CA1) at postnatal day 15 and adult ages: implications for the maturation of synaptic physiology and long-term potentiation. Journal of Neuroscience. 1992;12:2685–2705. doi: 10.1523/JNEUROSCI.12-07-02685.1992. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Harris KM, Stevens JK. Dendritic spines of CA 1 pyramidal cells in the rat hippocampus: serial electron microscopy with reference to their biophysical characteristics. Journal of Neuroscience. 1989;9:2982–2997. doi: 10.1523/JNEUROSCI.09-08-02982.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Hayashi-Takagi A, Yagishita S, Nakamura M, Shirai F, Wu YI, Loshbaugh AL, Kuhlman B, Hahn KM, Kasai H. Labelling and optical erasure of synaptic memory traces in the motor cortex. Nature. 2015;525:333–338. doi: 10.1038/nature15257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hell SW. Far-field optical nanoscopy. Science. 2007;316:1153–1158. doi: 10.1126/science.1137395. [DOI] [PubMed] [Google Scholar]
  21. Hofer SB, Mrsic-Flogel TD, Bonhoeffer T, Hübener M. Experience leaves a lasting structural trace in cortical circuits. Nature. 2009;457:313–317. doi: 10.1038/nature07487. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Holtmaat A, Svoboda K. Experience-dependent structural synaptic plasticity in the mammalian brain. Nature Reviews Neuroscience. 2009;10:647–658. doi: 10.1038/nrn2699. [DOI] [PubMed] [Google Scholar]
  23. Holtmaat A, Wilbrecht L, Knott GW, Welker E, Svoboda K. Experience-dependent and cell-type-specific spine growth in the neocortex. Nature. 2006;441:979–983. doi: 10.1038/nature04783. [DOI] [PubMed] [Google Scholar]
  24. Holtmaat AJ, Trachtenberg JT, Wilbrecht L, Shepherd GM, Zhang X, Knott GW, Svoboda K. Transient and persistent dendritic spines in the neocortex in vivo. Neuron. 2005;45:279–291. doi: 10.1016/j.neuron.2005.01.003. [DOI] [PubMed] [Google Scholar]
  25. Ji N. Adaptive optical fluorescence microscopy. Nature Methods. 2017;14:374–380. doi: 10.1038/nmeth.4218. [DOI] [PubMed] [Google Scholar]
  26. Kasai H, Fukuda M, Watanabe S, Hayashi-Takagi A, Noguchi J. Structural dynamics of dendritic spines in memory and cognition. Trends in Neurosciences. 2010;33:121–129. doi: 10.1016/j.tins.2010.01.001. [DOI] [PubMed] [Google Scholar]
  27. Keck T, Mrsic-Flogel TD, Vaz Afonso M, Eysel UT, Bonhoeffer T, Hübener M. Massive restructuring of neuronal circuits during functional reorganization of adult visual cortex. Nature Neuroscience. 2008;11:1162–1167. doi: 10.1038/nn.2181. [DOI] [PubMed] [Google Scholar]
  28. Keck T, Scheuss V, Jacobsen RI, Wierenga CJ, Eysel UT, Bonhoeffer T, Hübener M. Loss of sensory input causes rapid structural changes of inhibitory neurons in adult mouse visual cortex. Neuron. 2011;71:869–882. doi: 10.1016/j.neuron.2011.06.034. [DOI] [PubMed] [Google Scholar]
  29. Klar TA, Jakobs S, Dyba M, Egner A, Hell SW. Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission. PNAS. 2000;97:8206–8210. doi: 10.1073/pnas.97.15.8206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Knott GW, Holtmaat A, Wilbrecht L, Welker E, Svoboda K. Spine growth precedes synapse formation in the adult neocortex in vivo. Nature Neuroscience. 2006;9:1117–1124. doi: 10.1038/nn1747. [DOI] [PubMed] [Google Scholar]
  31. Lue N, Bewersdorf J, Lessard MD, Badizadegan K, Dasari RR, Feld MS, Popescu G. Tissue refractometry using Hilbert phase microscopy. Optics Letters. 2007;32:3522–3524. doi: 10.1364/OL.32.003522. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Majewska AK, Newton JR, Sur M. Remodeling of synaptic structure in sensory cortical areas in vivo. Journal of Neuroscience. 2006;26:3021–3029. doi: 10.1523/JNEUROSCI.4454-05.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Marinković S, Milisavljević M, Puskas L. Microvascular anatomy of the hippocampal formation. Surgical Neurology. 1992;37:339–349. doi: 10.1016/0090-3019(92)90001-4. [DOI] [PubMed] [Google Scholar]
  34. Matsuzaki M, Ellis-Davies GC, Nemoto T, Miyashita Y, Iino M, Kasai H. Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CA1 pyramidal neurons. Nature Neuroscience. 2001;4:1086–1092. doi: 10.1038/nn736. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Matsuzaki M, Honkura N, Ellis-Davies GC, Kasai H. Structural basis of long-term potentiation in single dendritic spines. Nature. 2004;429:761–766. doi: 10.1038/nature02617. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Mizrahi A, Crowley JC, Shtoyerman E, Katz LC. High-resolution in vivo imaging of hippocampal dendrites and spines. Journal of Neuroscience. 2004;24:3147–3151. doi: 10.1523/JNEUROSCI.5218-03.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Nägerl UV, Eberhorn N, Cambridge SB, Bonhoeffer T. Bidirectional activity-dependent morphological plasticity in hippocampal neurons. Neuron. 2004;44:759–767. doi: 10.1016/j.neuron.2004.11.016. [DOI] [PubMed] [Google Scholar]
  38. Nimchinsky EA, Sabatini BL, Svoboda K. Structure and function of dendritic spines. Annual Review of Physiology. 2002;64:313–353. doi: 10.1146/annurev.physiol.64.081501.160008. [DOI] [PubMed] [Google Scholar]
  39. Noguchi J, Nagaoka A, Watanabe S, Ellis-Davies GC, Kitamura K, Kano M, Matsuzaki M, Kasai H. In vivo two-photon uncaging of glutamate revealing the structure-function relationships of dendritic spines in the neocortex of adult mice. The Journal of Physiology. 2011;589:2447–2457. doi: 10.1113/jphysiol.2011.207100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Patton BR, Burke D, Owald D, Gould TJ, Bewersdorf J, Booth MJ. Three-dimensional STED microscopy of aberrating tissue using dual adaptive optics. Optics Express. 2016;24:8862–8876. doi: 10.1364/OE.24.008862. [DOI] [PubMed] [Google Scholar]
  41. Sala C, Segal M. Dendritic spines: the locus of structural and functional plasticity. Physiological Reviews. 2014;94:141–188. doi: 10.1152/physrev.00012.2013. [DOI] [PubMed] [Google Scholar]
  42. Schmid LC, Mittag M, Poll S, Steffen J, Wagner J, Geis HR, Schwarz I, Schmidt B, Schwarz MK, Remy S, Fuhrmann M. Dysfunction of somatostatin-positive interneurons associated with memory deficits in an alzheimer's disease model. Neuron. 2016;92:114–125. doi: 10.1016/j.neuron.2016.08.034. [DOI] [PubMed] [Google Scholar]
  43. Somogyi P. Hippocampus – Intrinsic Organisation. New York: Oxford University Press; 2010. [Google Scholar]
  44. Südhof TC. Neuroligins and neurexins link synaptic function to cognitive disease. Nature. 2008;455:903–911. doi: 10.1038/nature07456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Tønnesen J, Katona G, Rózsa B, Nägerl UV. Spine neck plasticity regulates compartmentalization of synapses. Nature Neuroscience. 2014;17:678–685. doi: 10.1038/nn.3682. [DOI] [PubMed] [Google Scholar]
  46. Ter Veer MJT, Pfeiffer T, Nägerl UV. Two-Photon STED microscopy for nanoscale imaging of neural morphology in vivo. Methods in Molecular Biology. 2017;1663:45–64. doi: 10.1007/978-1-4939-7265-4_5. [DOI] [PubMed] [Google Scholar]
  47. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E, Svoboda K. Long-term in vivo imaging of experience-dependent synaptic plasticity in adult cortex. Nature. 2002;420:788–794. doi: 10.1038/nature01273. [DOI] [PubMed] [Google Scholar]
  48. Urban NT, Willig KI, Hell SW, Nägerl UV. STED nanoscopy of actin dynamics in synapses deep inside living brain slices. Biophysical Journal. 2011;101:1277–1284. doi: 10.1016/j.bpj.2011.07.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Willig KI, Steffens H, Gregor C, Herholt A, Rossner MJ, Hell SW. Nanoscopy of filamentous actin in cortical dendrites of a living mouse. Biophysical Journal. 2014;106:L01–L03. doi: 10.1016/j.bpj.2013.11.1119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Xu T, Yu X, Perlik AJ, Tobin WF, Zweig JA, Tennant K, Jones T, Zuo Y. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature. 2009;462:915–919. doi: 10.1038/nature08389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Yang G, Pan F, Gan WB. Stably maintained dendritic spines are associated with lifelong memories. Nature. 2009;462:920–924. doi: 10.1038/nature08577. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Yuste R, Bonhoeffer T. Genesis of dendritic spines: insights from ultrastructural and imaging studies. Nature Reviews Neuroscience. 2004;5:24–34. doi: 10.1038/nrn1300. [DOI] [PubMed] [Google Scholar]
  53. Zhou Q, Homma KJ, Poo MM. Shrinkage of dendritic spines associated with long-term depression of hippocampal synapses. Neuron. 2004;44:749–757. doi: 10.1016/j.neuron.2004.11.011. [DOI] [PubMed] [Google Scholar]

Decision letter

Editor: Karel Svoboda1

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Chronic STED imaging reveals high turnover of dendritic spines in the hippocampus in vivo" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Karel Svoboda as the Reviewing Editor, and the evaluation has been overseen by Eve Marder as the Senior Editor. The following individual involved in review of your submission has also agreed to reveal their identity: Anthony Holtmaat (Reviewer #2).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Imaging and tracking of dendritic spines in vivo poses two distinct challenges: 1. the resolution, contrast and sensitivity need to be sufficiently high to discern spines from the parent dendrite and from one another. 2. dealing with motion artifacts, which can equally degrade the resolution. The first challenge is particularly important in hippocampus, since this is a deep structure with high spine densities. The latter is usually more of an issue in neocortex because it contains a high density of large, pulsating, blood vessels. 2-photon laser scanning microscopy allows imaging dendritic microstructures over time, however its resolution is limited and may not always be sufficient to resolve/detect all spines in the hippocampus. Modeling approaches have been applied to infer dynamics and turnover based on 2P in vivo images (Attardo et al., 2015).

The authors have applied 2P-STED, a super-resolution method, to image spine dynamics in the mouse hippocampus in vivo over 4 days. They dealt with access and motion artifacts by surgically replacing the cortex with a metal cylinder. To showcase the possibilities of the microscope and prep the authors provide paired comparisons between regular 2-photon and 2-photon STED images and measurements of spines. Then they went on to show that, due to the superior resolution, spine densities are closer to the values reported in EM (used as ground truth). The spine turnover rates are higher than what was reported in a previous 2-photon imaging study with a similar preparation (Gu et al. 2014), and similar to the modeling-based measurements derived from microendoscopic images (Attardo et al. 2015).

This paper describes a proof of principle using 2P-STED in vivo and provides a first analysis of structural plasticity data. We have no major concerns as to the validity of the results or their interpretation, but suggest expanding the analysis in order to provide more details about the structure-dynamics relationships.

Major concerns:

1) The spine density reported with STED is still lower than that measured with EM. The authors should produce best-of-class STED images from fixed samples of the same mice and compare the spine densities, this would also control for possible effects of the surgery, loss of resolution in vivo etc.

2) The paper is short, but still a bit flabby. In parts it reads a bit like an infomercial for STED. The resolution advantage over 2p is obvious and should be mentioned once. The effects on measured spine density are also obvious and don't need to be belabored multiple times etc.

3) The data are under-analyzed. One of the main advantages of the current study as compared to the Attardo et al. study, is the information about spine structure (size, length, head size, neck size etc). In cortex, there are clear structure-dynamics relationships, but for hippocampus in vivo this is not known. The authors should expand their analysis into this direction and show whether such relationships between spine dynamics and spine length/size, head size etc exist.

4) The authors argue that the current data set experimentally confirms the modeling data from the Attardo et al. study. Indeed, the first 4 days of imaging suggest that the entire population of spines may be impermanent. However, due to the limited sampling rate, it is not inconceivable that the authors could not detect a more or less stable spine population. To get more insight into this, it would be interesting to analyze the survival rate of new spines, and perhaps model this to estimate the stability and turnover of the entire population. In cortex new spines tend have a half-life in the order of 2 days or so, but in the current data set they could probably live longer. This, together with the finding that about 30% of the entire population disappears over 4 days, may indeed suggest that the whole population of spines is subject to turnover (in contrast to cortex, where the turnover rates to a large extent concerns newly formed spines). The structure-dynamics relationship could also be incorporated into this model.

5) The authors need to discuss or analyze more extensively what were the critical factors in causing the quite different turnover rates between the current study and the Gu et al. paper (one of the co-authors is on both papers!). It is difficult to fathom that the different values (96% stability over 16 days in Gu et al., and 60% stability over 4 days in the current paper) were entirely due to the improved resolution or the hippocampal region. A direct comparison of turnover rates scored in 2P and 2P-STED could have possibly hinted at the underlying factors, or could have identified the type of spines that can be seen turning over in 2P-STED and not in 2P. Furthermore, they should also discuss if long-term 2P-STED could provide better data for cortex or not, given that cortex is usually more prone to movement artefacts?

6) Since this paper focuses heavily on showing the potential of in vivo 2P-STED, we suggest that the authors provide a more substantive comparison between 2P-STED and 2P images in 3D, that is by showing the image stacks that were typically used for scoring, in addition to the projections. This could be done as a figure supplement or as a new figure. This is important since spines are usually scored in 3D (as the authors do too), and resolving them in the axial dimension is most problematic – and probably still difficult in 2P-STED (as the authors comment). This might relate to the remaining difference between the reported spine densities in 2P-STED and EM.

7a) Most of the spine necks measured in 2P-STED are smaller than the PSF in 2P (as measured on beads). For example, there are spines that measure 500-700 nm in 2P, but less than 200 in 2P-STED. Wouldn't one expect the size of the smaller necks in 2P to be limited to the PSF of the microscope? Can the authors explain this phenomenon?

b) Can the authors also explain why the PSF in 2P is highly variable (the FWHM varies from ± 300-450nm)? We assume that the beads were imaged under optimal and standardized conditions. So, what causes the large variability? The PSF in the axial dimension should be presented as well, as this is the dominant factor in the difficulty to discern spines from the parent dendrite in 2P and probably also in 2P-STED.

c) Along similar lines since the PSF, it would be better to measure the PSFs on beads that are embedded in the preparation. This could change the variance and the bounds as seen in Figure 1C, as well as the relationship of the PSF to the measurements in E.

8) The authors should specify the length of dendrite that was analyzed, and provide the number of spines that were counted. For example, subsection “Dendritic spine density of CA1 pyramidal neurons in vivo”, first paragraph and subsection “Dendritic spines undergo high morphological turnover in vivo”, second paragraph.

9) The authors used a parametric test to report differences throughout. Did the authors check for normality of the data? They need to specify this.

10) All dendritic images presented in the manuscript appear saturated, which makes it difficult to judge the accuracy of spine counting and the health of dendrite.

11) The criteria used in counting spines are a bit vague. For example, in Figure 3A, day 1, spine 10 does not seem to be connected to the dendritic shaft; spine 9 seems to be a small bump from the dendritic shaft, whereas similarly-sized bumps (e.g., between spine 18 and 19) are not counted. The methods of counting need to be documented in great and quantitative detail.

eLife. 2018 Jun 22;7:e34700. doi: 10.7554/eLife.34700.029

Author response


1) The spine density reported with STED is still lower than that measured with EM. The authors should produce best-of-class STED images from fixed samples of the same mice and compare the spine densities. This would also control for possible effects of the surgery, loss of resolution in vivo etc.

Taking up your advice, we acquired STED images in fixed hippocampal brain slices obtained from the same transgenic mice we had used for the in vivo experiments. The spine density in these fixed samples approaches the values reported by EM. To understand the remaining difference, we used a geometrical model based on the PSF of our microscope and the dimensions of the measured dendrites. The model makes the point that the limited z-resolution of the STED approach is the main reason, why we still underreport spine density relative to EM. (Revised Figure 2; Figure 2—video 1.)

2) The paper is short, but still a bit flabby. In parts it reads a bit like an infomercial for STED. The resolution advantage over 2p is obvious and should be mentioned once. The effects on measured spine density are also obvious and don't need to be belabored multiple times etc.

We heeded the reviewer’s advice and eliminated the redundancies about the STED benefit.

3) The data are under-analyzed. One of the main advantages of the current study as compared to the Attardo et al. study, is the information about spine structure (size, length, head size, neck size etc). In cortex, there are clear structure-dynamics relationships, but for hippocampus in vivo this is not known. The authors should expand their analysis into this direction and show whether such relationships between spine dynamics and spine length/size, head size etc exist.

We thank the reviewers for raising this important point, which prompted us to extract more morphological information from our time-lapse data set to analyse spine structure-dynamics relationships.

We carried out two types of analyses, which indicate that a structure-dynamics relationship indeed exists for hippocampal spines in vivo, resembling the case in the cortex, where smaller spines generally appear to be less stable.

Firstly, we show that spines that were visible on only one or two days have on average smaller heads than spines that were visible on all three imaging sessions.

Secondly, we performed a mathematical cluster analysis of the morphological parameters, which revealed three distinct spine populations (which incidentally resemble the classical spine types: cluster 1 ≅ small; cluster 2 ≅ thin; cluster 3 ≅ mushroom-like). Whereas 50% of ‘persistent’ spines (i.e. that were visible on all three imaging sessions) appeared in cluster 3, spines that were visible only on one or two sessions rarely (7%) exhibited a mushroom-like morphology. (New Figure 4; new Figure 4—figure supplement 1.)

4) The authors argue that the current data set experimentally confirms the modeling data from the Attardo et al. study. Indeed, the first 4 days of imaging suggest that the entire population of spines may be impermanent. However, due to the limited sampling rate, it is not inconceivable that the authors could not detect a more or less stable spine population. To get more insight into this, it would be interesting to analyze the survival rate of new spines, and perhaps model this to estimate the stability and turnover of the entire population. In cortex new spines tend have a half-life in the order of 2 days or so, but in the current data set they could probably live longer. This, together with the finding that about 30% of the entire population disappears over 4 days, may indeed suggest that the whole population of spines is subject to turnover (in contrast to cortex, where the turnover rates to a large extent concerns newly formed spines). The structure-dynamics relationship could also be incorporated into this model.

As suggested, we examined the kinetic data to look for evidence of a more long-lived spine population. We analysed the behaviour of new spines (i.e. spines that appeared on day 2) and observed that 66% of them were still visible during the next imaging session (day 4), indicating that newly formed spines may have a longer half-life in hippocampus than in cortex.

To take a closer look at the possibility of a stable spine population, we extrapolated the fraction of surviving spines by using a simple kinetic model based on a geometrical progression. The number of surviving spines on day n+1 can be expressed as:

un+1=un- F∗un

where un is the number of surviving spines on day n and F is the fraction of lost spines from day n to n+1 among the surviving ones. Note that this definition of F differs from the lost fraction of spines, since it excludes the new spines on day n that are lost on day n+1. Therefore, the number of surviving spines is

un=u0(1-F)n

Fitting the model to the in vivo data set, we estimated F = 22%, which corresponds to a spine half-life of 5.5 days. Moreover, the fit indicates that the survival fraction indeed converges to zero, which is inconsistent with the existence of a stable spine population. However, the ability of this modelling approach to identify kinetically distinct populations is at present not very good, given the limited sampling duration.

By comparison, fitting the model to the 2P data set, yields a spine half-life of 9 days (see Author response image 1). This confirms that data with lower spatial resolution indeed lead to an over-estimation of spine stability, as discussed in the Attardo study.

Author response image 1.

Author response image 1.

Because of the limited sampling and the inherent unreliability of longer-term extrapolations, we decided to omit this data from the manuscript. However, we discuss that these important kinetic issues need to be tested experimentally.

5) The authors need to discuss or analyze more extensively what were the critical factors in causing the quite different turnover rates between the current study and the Gu et al. paper (one of the co-authors is on both papers!). It is difficult to fathom that the different values (96% stability over 16 days in Gu et al., and 60% stability over 4 days in the current paper) were entirely due to the improved resolution or the hippocampal region. A direct comparison of turnover rates scored in 2P and 2P-STED could have possibly hinted at the underlying factors, or could have identified the type of spines that can be seen turning over in 2P-STED and not in 2P. Furthermore, they should also discuss if long-term 2P-STED could provide better data for cortex or not, given that cortex is usually more prone to movement artefacts?

In the revised manuscript we elaborate more about why the Gu et al. study reported a much higher spine stability. The reason is likely a combination of microscope resolution and hippocampal region. However, without a direct experimental comparison (our STED approach limited us to stratum oriens), we cannot provide a quantification for how much Gu et al. over-estimated stability versus how much spines are really more stable in the stratum radiatum.

Having said that, we did observe a considerable level of spine turnover even with just 2P, which one might think supports the case of a biological difference between basal and apical CA1 dendrites. However, our 2P approach, which benefited from a higher objective NA, already detected about 45% more spines than Gu et al., so it is not a fair comparison, and thus the same argument about over-estimating spine stability applies. We have tried to address these questions openly in the Discussion.

Regarding the point about the utility of 2P-STED for cortex, it is likely that improved spatial resolution will still provide better data, even if imaging in superficial brain areas is more prone to movement artefacts. The benefit of STED will be less obvious in as much as sample motion and not optics is the dominant factor behind image blur. (New Figure 3—figure supplement 1.)

6) Since this paper focuses heavily on showing the potential of in vivo 2P-STED, we suggest that the authors provide a more substantive comparison between 2P-STED and 2P images in 3D, that is by showing the image stacks that were typically used for scoring, in addition to the projections. This could be done as a figure supplement or as a new figure. This is important since spines are usually scored in 3D (as the authors do too), and resolving them in the axial dimension is most problematic – and probably still difficult in 2P-STED (as the authors comment). This might relate to the remaining difference between the reported spine densities in 2P-STED and EM.

Indeed, resolving dendritic spines along the optical axis is still difficult, as our new approach did not provide an improved z-resolution. To comply with the reviewer’s suggestion, we provide a new supplementary figure that illustrates how we scored spines. In addition, as mentioned above, we have used a geometrical model to estimate the effect of limited z-resolution on spine counting. Taking this fraction of invisible spines into account, the estimated spine density closely matches the reported spine densities reported by Bloss et al. 2018 and Harris et al. 1992 using EM. (New Figure 1—figure supplement 1; new Figure 2—figure supplement 1.)

7a) Most of the spine necks measured in 2P-STED are smaller than the PSF in 2P (as measured on beads). For example, there are spines that measure 500-700 nm in 2P, but less than 200 in 2P-STED. Wouldn't one expect the size of the smaller necks in 2P to be limited to the PSF of the microscope? Can the authors explain this phenomenon?

We thank the reviewers for pointing out this inconsistency. Revisiting the 2P images, we realized that some of our measurements were contaminated by instances where spines appeared merged in the images with neighbouring spines or other poorly resolved structures due to the lower spatial resolution, yielding artefactually wider neck measurements. By carefully looking at the corresponding 2P-STED images, we could eliminate these cases. In addition, we have now excluded measurements where the R2 of the fit for the line profiles was below 0.8. (Revised Figure 1E.)

b) Can the authors also explain why the PSF in 2P is highly variable (the FWHM varies from ± 300-450nm)? We assume that the beads were imaged under optimal and standardized conditions. So, what causes the large variability?

We also thank the reviewers for spotting this error, due to the inclusion of clustered beads in the analysis. Removing these (based on checking the corresponding 2P-STED image) and applying the threshold described above (fits with R2 < 0.8 were excluded), the variability was greatly reduced. We obtained a value of 325 nm ± 5 nm for 2P, which is very close to the theoretical value of 320 nm for a 1.0 NA objective and 900 nm excitation. (Revised Figure 1C.)

The PSF in the axial dimension should be presented as well, as this is the dominant factor in the difficulty to discern spines from the parent dendrite in 2P and probably also in 2P-STED.

We now present the PSF in the axial dimension as requested. (New Figure 1—figure supplement 1.)

c) Along similar lines since the PSF, it would be better to measure the PSFs on beads that are embedded in the preparation. This could change the variance and the bounds as seen in Figure 1C, as well as the relationship of the PSF to the measurements in E.

We agree with the reviewers that measuring PSFs in situ would be the preferred strategy, but unfortunately for practical reasons we cannot fulfil this request at this stage.

8) The authors should specify the length of dendrite that was analyzed, and provide the number of spines that were counted. For example, subsection “Dendritic spine density of CA1 pyramidal neurons in vivo”, first paragraph and subsection “Dendritic spines undergo high morphological turnover in vivo”, second paragraph.

We now provide a table containing the source data of the spine turnover analysis. (New Figure 3—source data 1.)

9) The authors used a parametric test to report differences throughout. Did the authors check for normality of the data? They need to specify this.

Indeed, we checked normality of the data distributions using the Shapiro-Wilk test. This is stated in the text now.

10) All dendritic images presented in the manuscript appear saturated, which makes it difficult to judge the accuracy of spine counting and the health of dendrite.

We usually adjusted image contrast and brightness so that the spines, especially their necks would be readily visible at the expense of causing some regions in the dendrite to appear saturated. To see what different brightness levels do to the images, we present the images used in the actual figures in a less saturated way, which look very similar (Author response image 2).

Author response image 2.

Author response image 2.

11) The criteria used in counting spines are a bit vague. For example, in Figure 3A, day 1, spine 10 does not seem to be connected to the dendritic shaft; spine 9 seems to be a small bump from the dendritic shaft, whereas similarly-sized bumps (e.g., between spine 18 and 19) are not counted. The methods of counting need to be documented in great and quantitative detail.

We agree with this criticism and have improved the description and transparency of our analysis of spine density, by providing image z-stacks for the reader to scroll through and adding a clarifying paragraph in the Materials and methods section.

We would like to point out that most of the turnover data was analysed independently and blindly (i.e. blind to the order of the imaging session) by the three first authors, which makes us confident about our main finding of high spine turnover.

The revised manuscript contains a table (Figure 3—source data 1) with an overview of the number of spines and dendritic lengths extracted from the turnover data.

In response to the reviewers’ comments, Author response image 3 shows a figure outlining why spines 9 and 10 was counted as indicated and connected to the dendritic shaft. In contrast, the “bump” highlighted by the white arrow head fell below the threshold.

Author response image 3.

Author response image 3.

(New Figure 2—figure supplement 1; new Figure 3—source data 1; new paragraph in Materials and methods.)

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    Figure 1—source data 1. Data for panel C.
    DOI: 10.7554/eLife.34700.006
    Figure 1—source data 2. Data for panel B.
    DOI: 10.7554/eLife.34700.007
    Figure 1—source data 3. Data for Figure 1—figure supplement 1.
    DOI: 10.7554/eLife.34700.008
    Figure 2—source data 1. Data for panel B.
    DOI: 10.7554/eLife.34700.012
    Figure 2—source data 2. Data for panel D.
    DOI: 10.7554/eLife.34700.013
    Figure 3—source data 1. Source data of the parameters underlying Figure 3 extracted from the turnover data set.
    DOI: 10.7554/eLife.34700.017
    Figure 3—source data 2.
    DOI: 10.7554/eLife.34700.018
    Figure 3—source data 3. Data for Figure 3—figure supplement 1.
    DOI: 10.7554/eLife.34700.019
    Figure 4—source data 1. Underlying data for Figure 4.
    elife-34700-fig4-data1.xlsx (109.7KB, xlsx)
    DOI: 10.7554/eLife.34700.022
    Transparent reporting form
    DOI: 10.7554/eLife.34700.023

    Data Availability Statement

    Source data files have been provided for Figures 1, 2, 3 & 4 and Figure 1-figure supplement 1 and Figure 3-figure supplement 1.


    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES