Author response: Environmental enrichment enhances patterning and remodeling of synaptic nanoarchitecture as revealed by STED nanoscopy

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Synaptic plasticity underlies long-lasting structural and functional changes to brain circuitry and its experience-dependent remodeling can be fundamentally enhanced by environmental enrichment. It is however unknown, whether and how the environmental enrichment alters the morphology and dynamics of individual synapses. Here, we present a virtually crosstalk-free two-color in vivo stimulated emission depletion (STED) microscope to simultaneously superresolve the dynamics of endogenous PSD95 of the post-synaptic density and spine geometry in the mouse cortex. In general, the spine head geometry and PSD95 assemblies were highly dynamic, their changes depended linearly on their original size but correlated only mildly. With environmental enrichment, the size distributions of PSD95 and spine head sizes were sharper than in controls, indicating that synaptic strength is set more uniformly. The topography of the PSD95 nanoorganization was more dynamic after environmental enrichment; changes in size were smaller but more correlated than in mice housed in standard cages. Thus, two-color in vivo time-lapse imaging of synaptic nanoorganization uncovers a unique synaptic nanoplasticity associated with the enhanced learning capabilities under environmental enrichment. Editor's evaluation Synapses mediate information transmission in the brain, and part of the synaptic structure called spines are the receiving end of signal transfer between neurons. Using a custom-built superresolution microscope, the study reveals the nanoscale structural dynamics of individual spine shape and its resident scaffolding protein PSD95 simultaneously, in mouse cortex in vivo. Aspects of the structural dynamics are found to differ depending on whether mice have been reared in a simple housing or in an enriched environment, the latter condition being associated with enhanced activity. https://doi.org/10.7554/eLife.73603.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Activity-driven changes of spine dynamics Over an entire lifespan, cognitive, sensory, and motor learning is associated with changes to a specific assembly of synapses which is often termed the memory engram (Poo et al., 2016). Thus, spines emerge, disappear, or change with cellular processes underlying learning, and even ‘remember’ previous sensory experience (Hofer et al., 2009). There is also evidence that learning induces structural and functional synaptic changes similar to long-term potentiation (LTP) protocols (Poo et al., 2016). In this concept of learning, the maintenance of memory critically depends on the stability of the underlying synaptic connections. Synaptic structures, however, are highly volatile intrinsically as such that synaptic connections undergo continuous spontaneous remodeling without any activity (Mongillo et al., 2017; Ziv and Brenner, 2018); for example, intact synapses are formed despite abolishing pre-synaptic release (Sigler et al., 2017) or network silencing (Hazan and Ziv, 2020). Due to spatial resolution constraints, previous in vivo studies mostly focused on the persistency of spines and synapses in terms of their appearance and elimination; the spine size was estimated from fluorescence intensity. Therefore, directly assessed activity-driven changes in synapse or spine head size in vivo are missing. Spine to synapse structural correlation Decades of neuroscience research has shown a tight correlation between structural plasticity, the morphological transformation of spines and synapses, and modifications in synaptic transmission, termed functional plasticity (Yuste and Bonhoeffer, 2001), which was confirmed recently at ultrastructural resolution (Holler et al., 2021). Functional changes are linked to anchoring amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPAR) via transmembrane AMPAR-regulatory proteins (TARPs) to the post-synaptic density (PSD) scaffolding protein PSD95 (Compans et al., 2016; Herring and Nicoll, 2016). AMPAR-mediated currents were shown to increase simultaneously with the spine head directly after LTP induction. Recent evidence suggests that not all synaptic structures follow this fast dynamic; as such, accumulations of PSD95 increase in vitro with a delay of ~1 hr after inducing LTP, whereas the spine heads increase in less than a minute (Bosch et al., 2014; Meyer et al., 2014). Thus, the relation between spine size and PSD organization (Arellano et al., 2007; Cane et al., 2014; Harris et al., 1992; Meyer et al., 2014) may be temporally dynamic and complex during synaptic change and the remodeling of PSD and spine head plasticity might be more independent than previously thought. Here, we investigate the structural correlation at nanoscale of the spine and post-synapse in vivo using mice reared in enriched environment representing increased activity conditions and normal housing representing baseline conditions. PSD95 nanoorganization By applying superresolution microscopy we and others have shown that PSD95 is often arranged in clusters, rings, or a more complex nanopattern, and that this nanoorganization undergoes pronounced intrinsic structural changes on the time scale of minutes to hours (Hruska et al., 2018; MacGillavry et al., 2013; Wegner et al., 2018). Recently, it was shown that such a sub-synaptic nanoorganization exists at the pre- and post-synaptic site and that both are aligned in the so-called nanocolumns (Hruska et al., 2018; Tang et al., 2016). Computational studies show that changes of the shape of the PSD (Franks et al., 2003) or modest clustering of AMPAR (Savtchenko and Rusakov, 2014) is a highly efficient way to modify synaptic strength without altering the total amount of receptors. Moreover, AMPAR current amplitude drops significantly already at 50 nm offset between pre-synaptic glutamate release site and AMPAR cluster (Haas et al., 2018). Therefore, activity-driven structural rearrangement at the nanoscale to align post-synaptic receptors to pre-synaptic release sites may be faster than the incorporation of new molecules to induce changes in the synaptic strength. This study Here, we apply nanoscale stimulated emission depletion (STED) microscopy to map temporal changes of the spine head and PSD95 accumulations simultaneously in vivo, and address (1) the plasticity of spine heads and synapses at baseline at time scales similar to LTP processes; (2) the correlation between PSD95 and spine head size changes; (3) the plasticity of the PSD95 nanoorganization; and (4) whether enhanced activity modifies the structure and/or plasticity of these measures. For this purpose, we employ environmental enrichment (EE) where a mouse is provided with multi-sensory stimulation, cognitive activity, social interactions, and physical exercise, which modifies the degree of plasticity and dynamics of cortical sensory circuits. EE has been shown to accelerate the maturation of new neurons, increase pre- and post-synaptic protein levels (Nithianantharajah et al., 2004), and facilitate synaptic plasticity (Greifzu et al., 2014; Nithianantharajah and Hannan, 2006). Structural changes include increased dendritic branching and spine density (Gelfo et al., 2009; Leggio et al., 2005) and induction of spine formation (Yang et al., 2009). However, it is unknown whether the effects of EE leave their mark only at the level of spine formation and elimination such as observed by Yang et al., 2009, or whether EE also affects the dynamics and nanostructure of all individual spines and PSDs. To unambiguously assess these relationships between spine morphology and PSD95, we set up a virtually crosstalk-free two-color STED microscope for in vivo superresolution imaging of EGFP and Citrine. We utilized a transcriptionally regulated antibody-like protein to visualize endogenous PSD95 without introducing overexpression artifacts (Gross et al., 2013). Together with a membrane label, we simultaneously recorded temporal morphological changes of the spine head and PSD95 nanoorganization and compared its dynamics between mice raised in EE and in standard cages (control, Ctr). By investigating the synaptic dynamics in the mouse visual cortex, a brain region known to exhibit enhanced plasticity under EE conditions (Baroncelli et al., 2010; Kalogeraki et al., 2017) and accessible to light microscopy, we demonstrate that spine head and PSD95 size distributions decrease in variability while spine heads increase in average size under EE. In addition, PSD95 underwent much stronger directional changes in size and reorganization of their nanoorganization under EE, while the average change in amplitude was smaller compared to the controls. Dynamical changes of PSD95 and spine head size correlated only mildly. Results Virtually crosstalk-free two-color STED Microscope To unambiguously dissect the dynamics of distinct synaptic components and examine their interactions, we established a virtually crosstalk-free two-color STED microscope for imaging of EGFP and EYFP or Citrine, respectively. The challenge for in vivo two-color STED microscopy is to find an in vivo compatible pair of fluorescent molecules with similar emission spectra so that it can be depleted with the same STED beam, but at the same time can be temporally or spectrally separated (Willig et al., 2021). Previous attempts with STED microscopy of EGFP and EYFP utilized a single excitation wavelength and two-color detection, which suffered from high crosstalk, and therefore required a linear unmixing of the channels (Tønnesen et al., 2011). Channel unmixing, however, requires large signal-to-noise levels. To reduce crosstalk and thus avoid the necessity of channel unmixing, we extended our previously described in vivo STED microscope (Willig et al., 2014) by an additional two-color excitation and detection to selectively excite the green or yellow fluorescent protein (Figure 1A). We utilized 483 nm pulses to excite EGFP and detected its spontaneous fluorescence at 498–510 nm (Det1, Figure 1B). EYFP (or Citrine) was excited at 520 nm and detected mainly at 532–555 nm (Det2, Figure 1B). The 483 and 520 nm excitations were switched on and off alternately while recording two consecutive line scans to temporally separate the EGFP and EYFP/Citrine emissions. To determine the crosstalk between both channels, we investigated one-color labeled cells expressing EGFP or EYFP. Live-cell imaging showed that we have designed a virtually crosstalk-free microscope with only ~8% crosstalk of EYFP in the EGFP channel (channel 1) and ~5% of the EGFP signal in the EYFP channel (channel 2) which does not require linear unmixing (Figure 1C). STED was performed by a 595 nm laser beam passing a vortex phase plate to create a donut-shaped intensity profile in the focal plane of the objective as previously described (Eggeling et al., 2015). Figure 1 with 1 supplement see all Download asset Open asset Virtually crosstalk-free two-color in vivo stimulated emission depletion (STED) microscopy. (A) Custom-designed in vivo STED microscope with pulsed 483 nm (Exc1) and 520 nm (Exc2) excitation and 595 nm STED laser. APD: avalanche photon detector, BP: bandpass filter, Det: detection, DM: dichroic mirror, OPO: optical parametric oscillator, QWP: quarter wave plate, VPP: vortex phase plate. See Material and methods section for details. (B) Excitation (dashed line) and emission (solid line) spectrum of EGFP and EYFP and wavelength regions for selective excitation and detection. (C) Cultured hippocampal neurons expressing the actin marker Lifeact (LA)-EGFP or LA-EYFP. Excitation of EGFP at 483 nm close to its excitation maximum (B) and detection at 498–510 nm (Det1) reduced the crosstalk to 5% in channel 2 (C). EYFP or Citrine was excited at 520 nm close to its maximum (B) and detected at 532–555 nm (Det2) which resulted in a low crosstalk of 8% in channel 1 (C). (D) In vivo STED microscopy image of apical dendrite in layer 1 of the visual cortex of an anesthetized mouse. Labeling of membrane (myr-EGFP-LDLR(Ct)) and PSD95 (PSD95.FingR-Citrine) visualized the spine morphology and PSD95 nanoorganization at superresolution (D, inset). Images are smoothed and represent maximum intensity projections (MIPs). No unmixing was performed due to the low crosstalk. Two-color in vivo STED microscopy of endogenous PSD95 and spine morphology Having established the virtually crosstalk-free detection of two STED approved fluorescent proteins (Nägerl et al., 2008; Willig et al., 2006), we set out to simultaneously superresolve the nanoorganization of the PSD protein PSD95 and the associated spine morphology in vivo. To this end, we generated recombinant adeno-associated viral (AAV) particles encoding fusion proteins under control of the human Synapsin promoter (hSyn). To visualize PSD95, we expressed the antibody-like protein PSD95.FingR, which has been shown to label endogenous PSD95 (Gross et al., 2013; Willig et al., 2021) attached to the fluorescent protein Citrine. The expression of PSD95.FingR-Citrine is controlled by a transcriptional regulation system that prevents expression after saturation of the binding sites and therefore reduces background (Gross et al., 2013). For spine morphology, we expressed myr-EGFP-LDLR(Ct), a combination of a myristoylation site (myr), EGFP, and the C-terminal (Ct) cytoplasmic domain of low-density lipoprotein receptor (LDLR), a potent marker for the dendritic membrane (Kameda et al., 2008; Willig et al., 2021). In order to reduce the density of labeled neurons and thus overlapping fluorescence signals, the fluorescent labels were incorporated into AAVs in reverse orientation in the vector between double-floxed inverted open reading frames (DIO). Expression of the labels was enabled by Cre recombinase encoding AAVs co-injected at low concentration (Willig et al., 2021). Hence, the density of labeled neurons was tuned by the dilution of the Cre expressing AAV and the brightness of the PSD95 and membrane labeling by the concentration of the respective AAVs; thus the density of labeled cells and their brightness were independent variables. We co-injected the three AAVs into pyramidal cell layer 5 of the visual cortex. Three to six weeks after transduction, the mice were anesthetized and a cranial window was inserted over the visual cortex. To perform motion and aberration-free imaging at nanoscale resolution, the cranial window needs to be of highest quality. As described in detail in Steffens et al., 2020, critical steps involve a craniotomy that is as atraumatic as possible and does not damage the cortical surface when drilling or removing the bone plate and dura mater. Moreover, the gap between the brain surface and the cover glass needs to be negligibly small and the right choice of the dental cement is important to avoid bending of the cover glass. With such an optimized cranial window, STED microscopy visualized dendrites, spines, and attached PSD95 assemblies in layer 1 of the visual cortex at superresolution (Figure 1C) with a resolution of at least ~70 nm for Citrine and ~84 nm for EGFP (Figure 1—figure supplement 1A, B). The inset in Figure 1D reveals a perforated nanoorganization of endogenous PSD95 that would not be detectable with conventional in vivo light microscopy due to the insufficient resolution. Due to the low crosstalk, a computational post-processing such as unmixing was not required. We observed mainly dendrites that expressed both, EGFP and Citrine, and only rarely found dendrites expressing either Citrine or EGFP alone. EE housed mice exhibit less variability in the size of spine head and PSD95 assemblies while their heads are larger than in control Having established the two-color in vivo STED microscopy, we addressed the question of whether spines or PSDs exhibit systematic differences in size between EE and Ctr mice. EE mice were housed in a large, two-floor cage equipped with three running wheels for physical exercise, a labyrinth for cognitive stimulation, a tube and a ladder to change between the two levels, and were kept in groups of up to 12 female mice to allow manyfold social interactions (Figure 2—figure supplement 1A–C). Ctr mice were raised in pairs in standard cages without any equipment (Figure 2—figure supplement 1D, E). The mice were transduced with AAVs as described above and 3–6 weeks after transduction, they were anesthetized and a cranial window was implanted above the visual cortex. Imaging commenced about 2.5 hr after onset of the anesthesia. We acquired two-color z-stacks mainly of dendrites, which were in parallel to the focal plane. To analyze differences in morphology of the dendritic spines, we encircled the spine head and PSD95 at their largest extent and computed the respective area (Figure 2A, see Materials and methods section for details). We only considered spines containing a PSD95 label and therefore most likely form functional synapses. Occasionally we observed long, thin spines without a head, sometimes called filopodia in the literature, which were not analyzed. The spine head area correlated with the PSD95 area for EE and Ctr (Figure 2B), corroborating EM and two-photon microscopy studies (Arellano et al., 2007; Cane et al., 2014; Meyer et al., 2014). The histograms of spine head (Figure 2C) and PSD95 area (Figure 2D) were positively skewed and the distributions were significantly different between EE and Ctr mice. We determined a median spine head area for EE housed mice of 0.527 (interquartile range [IR]: 0.359–0.844) µm2 and 0.462 (IR: 0.290–0.787) µm2 for Ctr mice. The area of PSD95 on the same set of spine heads was 0.158 (IR: 0.107–0.269) µm2 in median for EE mice and 0.161 (IR: 0.092–0.286) µm2 for Ctr mice which is in accordance with previous EM studies reporting 0.15 µm2 for the PSD area in the mouse visual cortex (Harris and Weinberg, 2012). This size of the PSD95 area of layer 5 pyramidal neurons is slightly larger than our previously reported diameter of 354 nm which corresponds to ~0.10 µm2 for a circular distribution, obtained in a ubiquitously expressing PSD95-EGFP knock-in mouse (Wegner et al., 2018); therefore, the larger size of the PSD95 area could reflect the larger size of the spine heads of layer 5 pyramidal neurons (Konur et al., 2003). To further dissect these differences, we next plotted the histograms of the logarithm of spine head (Figure 2E) and PSD95 areas (Figure 2F). All four histograms (Figure 2E and F) are symmetric and well described by a Gaussian function. This is in line with prior work showing that the spine and PSD95 fluorescence is log-normally distributed in general, which is predicted by multiplicative processes of ongoing spine plasticity (Hazan and Ziv, 2020; Loewenstein et al., 2011). The Gaussian function fitting the spine head area is shifted to the right and is narrower for EE housed mice (Figure 2E). This is manifested by a significantly larger spine head area (Figure 2G) and smaller variance of the size distribution (Figure 2I) of EE housed mice, which might imply a preferential loss or adaptation of small spines (Figure 2E). The distributions of the logarithm of the PSD95 area (Figure 2F) are almost centered and thus the average area is not significantly different between EE and Ctr mice (Figure 2H). However, the variance (Figure 2J) is smaller for EE housed mice which is also visible in the Gaussian distribution that is narrower for EE housed mice (Figure 2F), that is, having less extreme values. Since previous studies often used the brightness of PSD95 assemblies or spines, as a of we also the PSD95 brightness and found a of between the brightness and nanoscale size (Figure 2—figure supplement indicating that the brightness is a of synaptic We also the spine density which was not significantly different between EE and Ctr housed mice (Figure 2—figure supplement show that EE a in spine head size and variability of the post-synaptic both of which have been shown to with synaptic strength (Holler et al., 2021). Figure 2 with 2 see all Download asset Open asset distributions of spine head and PSD95 area are sharper and show larger heads for mice housed in enriched (EE) than in standard cages. (A) emission depletion (STED) of dendritic spines and associated PSD95 assemblies Images are maximum intensity enhanced for area Spine heads and PSD95 assemblies were encircled to the (B) correlation of spine head and PSD95 area in Ctr and EE housed mice. are and correlation is and from distributions of spine head area (C) and PSD95 area (D) are positively skewed and significantly different between EE and Ctr housed mice of single large are F) data as shown in but values. represent Gaussian to the respective and variance of data shown in F) with of mice and Figure data 1 Spine head and PSD95 assembly Download correlation of PSD95 and spine head area changes over minutes to hours We showed above a correlation between the size of PSD95 assemblies and the spine head (Figure which previous of close structural correlation between PSD and spine size (Arellano et al., However, on which time scale the dynamic changes between these two are linked in vivo has We thus whether and how temporal changes of these were Therefore, we performed time-lapse STED microscopy of spine morphology and PSD95 for EE and Ctr housed mice for up to hr as described above and temporal changes of these (Figure Imaging commenced hr after onset of the and the time-lapse was recorded at without a We recorded STED at different of was recorded for three time at a time of either (Figure and or Over these time the spines were mostly a spine was or a new of these dynamic spines PSD95 and they were therefore most likely highly dynamic filopodia and For time the size of the spine head and PSD95 assembly was as described The average spine head area and PSD95 area not show large changes over the time of (Figure Thus, we not changes as a of or and observed We computed size changes over time so that both, and changes were symmetric with a of Figure of the changes of the spine head over changes of the PSD95 area for time (Figure and changes of all time (Figure the data are over all four changes to Figure supplement 1A–C). linear a but correlation between 0.15 and Figure such that an increase of PSD95 area is more by an increase in spine head area and (Figure data however, also or correlated that is, PSD95 areas on a spine or (Figure PSD95 assemblies larger changes in size than spine which is from the distributions in Figure or the median changes (Figure supplement 1D, spine heads by and by while PSD95 assemblies by and by To further the of correlated and we performed a 1 is by the which the data the maximum of variance and the of is the variance this The vector was for all three time in the (Figure and therefore for positively correlated The is by to and therefore in the (Figure for correlated The variance of and thus correlated changes is larger at all time than that of that is, changes (Figure Figure supplement Thus, we found the positively correlated changes also a large of changes a temporal of spine head and PSD95 For all time the variance of and was for Ctr housed mice, not (Figure To Ctr EE housed mice, we the variance over all time (Figure as the changes not significantly Thus, the variance of Ctr housed mice was significantly for but not for This that correlated changes much less in the EE housed mice indicating a between changes in spine head size and PSD95 Figure with 1 supplement see all Download asset Open asset changes of spine head and PSD95 area are positively correlated with a larger variability for control than environmental enrichment (EE) housed mice. of spine heads and PSD95 of time-lapse in vivo two-color stimulated emission depletion (STED) microscopy for EE (A) and Ctr (B) housed mice at time of and Images are smoothed and shown as maximum intensity (C) of spine head and PSD95 areas of Ctr and EE housed mice over changes in spine head and PSD95 area after and time and changes of all time are from and correlation is represent the components 1 and 2 of the for Ctr and EE respectively. and of changes plotted in variance of and variance over all time EE of temporal changes between spine head size and PSD95 correlation and of spine head size and area of PSD95 assemblies correlation a PSD95 assembly on a spine and of mice and for EE and for of spines with PSD95 EE, are for example, and Figure data 1 Spine head and PSD95 assembly sizes for time Download In the synaptic structure is highly dynamic and spine heads and PSD95 assemblies change in size by already after The variance of these changes is smaller and positively correlated changes are more pronounced for EE housed mice. that EE housed mice undergo smaller but more morphological changes under anesthesia. of PSD95 area is different between EE and Ctr housed mice Since our were log-normally indicating a multiplicative dynamic, we whether the temporal changes were also regulated by multiplicative processes (Hazan and Ziv, 2020; Loewenstein et al., 2011). Thus, we plotted the synaptic size after different time as a function of the original size (Figure linear a and for spine head and PSD95 for both EE and Ctr