Extracellular matrix remodeling through endocytosis and resurfacing of Tenascin-R

Abstract

The brain extracellular matrix (ECM) consists of extremely long-lived proteins that assemble around neurons and synapses, to stabilize them. The ECM is thought to change only rarely, in relation to neuronal plasticity, through ECM proteolysis and renewed protein synthesis. We report here an alternative ECM remodeling mechanism, based on the recycling of ECM molecules. Using multiple ECM labeling and imaging assays, from super-resolution optical imaging to nanoscale secondary ion mass spectrometry, both in culture and in brain slices, we find that a key ECM protein, Tenascin-R, is frequently endocytosed, and later resurfaces, preferentially near synapses. The TNR molecules complete this cycle within ~3 days, in an activity-dependent fashion. Interfering with the recycling process perturbs severely neuronal function, strongly reducing synaptic vesicle exo- and endocytosis. We conclude that the neuronal ECM can be remodeled frequently through mechanisms that involve endocytosis and recycling of ECM proteins.

Fig. 1. TNR molecules endocytose and subsequently resurface in neurons.

a Neurons were treated with sulfo-NHS-S-S-biotin to label cell-surface proteins. Following a 6-h incubation, allowing for internalization, remaining cell-surface proteins were stripped of their labels with glutathione. After a further 18 h of incubation, allowing for recycling, the neurons were again treated with glutathione. Lysates representing surface, endocytosed or non-recycled pools, were collected on streptavidin-coupled beads, immunostained for TNR, immobilized on glass slides and imaged with confocal microscopy. b Example beads collecting TNR pools, or controls incubated without primary antibodies. Scale bar = 2 µm. c A quantification of TNR fluorescence intensity normalized to the ‘surface’ mean in the corresponding experiment, indicates that a large fraction of TNR molecules endocytose within 6 h, and many subsequently resurface within 24 h. As positive/negative controls, the synaptic vesicle protein Syt1, well-known to recycle, and the intracellular protein calmodulin were tested. The plots are scaled by the ratio between the ‘surface’ mean for these proteins and that of TNR. N = 4 (TNR) and 3 (Syt1/calmodulin) independent experiments with >100 (TNR) and >50 (Syt/calmodulin) beads. d, e As additional controls, the lysosomal marker LAMP1, known to endocytose but scarcely recycle, and myelin basic protein (MBP), which should not endocytose, were also tested. Scale bar = 1 μm. e Quantification of LAMP1/MBP fluorescence intensity, normalized to the ‘surface’ mean of the corresponding experiment. N = 3 independent experiments with > 50 beads. Statistical significance was evaluated using repeated-measures one-way ANOVA, (c TNR: F_1.153,3.458 = 28.29, **p = 0.009; Syt1: F_1.007,2.014 = 62.98, *p = 0.015; calmodulin: F_1,2 = 0.016, p = 0.912; e LAMP1: F_1.293,2.585 = 19.6, *p = 0.028; MBP: F_1.52,3.041 = 28337, ***p < 0.001), followed by the Holm-Sidak multiple comparisons test comparing ‘surface’/‘endocytosed’ and ‘endocytosed’/‘non-recycled’ (c TNR: *p = 0.032, *p = 0.021; Syt1: *p = 0.044, *p = 0.044; calmodulin: p = 0.933, p = 0.993; e LAMP1: *p = 0.045, p = 0.162; MBP: ***p < 0.001, p = 0.068). Data represent the mean ± SEM, dots indicate individual experiments. Source data are provided in Source Data file.

Fig. 2. Dynamic TNR molecules emerge at synapses, in an activity-dependent fashion.

a To monitor dynamic TNR molecules, surface epitopes were blocked by incubating with non-fluorescent TNR antibodies (gray). After some time, fluorophore-conjugated antibodies were applied (red) to reveal newly-emerged epitopes (dark blue). b Newly-emerged epitopes 2/4/12 h post-blocking, (epifluorescence). Scale bar = 10 µm. c Fluorescence intensity, normalized to non-blocked neurons. N = 3 independent experiments, ≥10 neurons per datapoint. Repeated-measures one-way ANOVA: F_1.089,2.179 = 790.8, ***p < 0.001, followed by Fisher’s LSD: ‘2 h’/‘4 h’:*p = 0.041; ‘4 h’/‘12 h’:*p = 0.032; ’12 h’/‘no blocking’: **p = 0.002. d All TNR epitopes (left) or newly-emerged epitopes 12 h post-blocking (right) were revealed (magenta, STED imaging). Membranes of a subset of neurons were labeled using sparse DiO labeling (green, confocal imaging). Presynapses were identified by VGlut1 (blue, STED imaging). Scale bars: 1 µm (full images), 500 nm (insets). Bottom: hundreds of synapses were averaged by centering synapse images on the VGlut1 puncta and orienting the dendritic DiO signals vertically. “All” TNR epitopes cover the entire bouton-dendrite area (left), while newly-emerged epitopes are enriched in the bouton region (right). e Quantification of TNR exchange at synapses (as in c, measuring exclusively TNR at VGlut1-labeled synapses). N = 3 independent experiments with >100 synapses. f Comparison of newly-emerged TNR epitopes 12 h post-blocking in cultures treated with bicuculline (40 µM), or CNQX (10 µM) and AP5 (50 µM). Intensity is normalized to the corresponding control (DMSO). N = 3 experiments, ≥10 neurons per datapoint. One-way ANOVA on rank: F_2,6 = 42, ***p < 0.001, followed by Dunn’s multiple comparisons test: ‘ctrl’/‘CNQX + AP5’: **p = 0.003; ‘ctrl’/‘bic’: *p = 0.042. Data represent the mean ± SEM, dots indicate individual experiments. g, Analysis of 2-color-STED images (as shown in d). Synaptic enrichment is substantially higher for newly-emerged epitopes. N = 3 independent experiments with >100 synapses. Repeated-measures one-way ANOVA on log-transformed data: F_1.977,3.954 = 24.13, **p = 0.006, followed by Fisher’s LSD: ‘new’/‘all’ epitopes: *p = 0.024 (dendrites); *p = 0.036 (axons). Data represent the mean (line) ± SEM (shaded region). Source data are provided in Source Data file.

Fig. 3. The emergence of TNR epitopes is dependent on synaptic weight.

a The TNR epitopes in the ECM were blocked as in the previous experiments, and 12 h later the cultures were incubated with fluorophore-conjugated TNR antibodies (magenta) and with fluorophore-conjugated antibodies for the intra-vesicular domain of Syt1 (green), which reveal the synaptic vesicle pool that undergoes exo- and endocytosis (the active pool). The size of this pool is a measure of the activity of the respective boutons. The panels show example synapses with different active vesicle pools, imaged in STED (TNR) and confocal (Syt1). Scale bar = 300 nm. The graph shows the mean fluorescence intensities normalized to the median intensity of the respective experiment. The Syt1 intensities are binned to include an equal number of synapse images. An analysis of the correlation of the TNR signal at Syt1-labeled synapses indicates that the TNR signals correlate strongly with the size of the active vesicle pool in the respective boutons. N = 3 independent experiments, with >1100 synapses per datapoint, Spearman’s ⍴ = 0.927, ***p = 6.489 × 10⁻⁷ (two-sided). b Newly-emerged TNR epitopes (magenta) were labeled after 12 h as in panel a, and the neuronal plasma membrane was visualized with DiO (green). The panels show example spines with different sizes, imaged in STED (TNR) and confocal (DiO). Scale bar = 300 nm. The graph shows the mean fluorescence intensity of TNR and the mean synapse area, normalized to the median values in the respective experiment. The synapse area values are binned to include an equal number of synapse images. An analysis of the correlation of the TNR signal at DiO-labeled spines indicates that the TNR signals correlate strongly with the size of the dendritic spine for newly-emerged TNR epitopes. N = 3 independent experiments, with >280 synapses per datapoint, Spearman’s ⍴ = 0.862, ***p = 3.601 × 10⁻⁵ (two-sided). All data represent the mean ± SEM, with dots indicating individual experiments. Source data are provided in Source Data file.

Fig. 4. Dynamic TNR molecules are endocytosed in neurons over hours, and recycle with a periodicity of ~3 days.

a Newly-emerged TNR epitopes were labeled 4 h post-blocking, and monitored by live epifluorescence imaging. Arrowheads indicate neuronal somas. Scale bar = 10 µm. Quantification of the intensity in somas (normalized to t₀ timepoint) indicates significant internalization. N = 5 independent experiments, 1-4 neurons per datapoint. Friedman test (χ²₆ = 25.46, ***p < 0.001), followed by two-sided Dunn’s multiple comparisons test (‘6 h’: *p = 0.033; ‘8 h’: **p = 0.005; ‘10 h’: **p = 0.005; ‘12 h’: **p = 0.002). Data represent mean ± SEM, dots indicate individual experiments. b Internalized TNR in axons vs. dendrites. Epitopes were allowed to internalize for 12 h, followed by surface stripping with proteinase K. The remaining signal was imaged with confocal microscopy in neurites visualized using DiO (green), with axons identified by immunostaining AnkyrinG (blue). Scale bar = 4 µm (1 µm zoom). Quantification of the signal density reveals no differences between dendrites and axons. N = 4 experiments, ≥10 neurons per datapoint. Two-sided paired t-test (t = 0.741, p = 0.513). Data represent mean ± SEM, dots indicate individual experiments. c Newly-emerged TNR epitopes were labeled 6 h post-blocking. The fraction present on the surface of neurites was measured at different intervals by imaging neurons in epifluorescence before and after stripping with proteinase K. Quantification of the fluorescence ratio before/after stripping (normalized to the ‘0d’ timepoint) reveals peaks of TNR resurfacing at 3 and 6 days post-labeling (~3 day periodicity). Amounts stripped at ‘3d’ and ‘6d’ are significantly higher than at ‘1d’ and ‘2d’, or ‘5d’ and ‘7d’. N = 4 independent experiments, 5 before/after images per datapoint. Kruskal-Wallis followed by Fisher’s LSD (Days 2, 3, 4: H₂ = 8.29, *p = 0.016, ‘3d’/‘2d’; *p = 0.046; ‘3d’/‘4d’: **p = 0.005; Days 4, 5, 6: H₂ = 6.74, *p = 0.036, ‘6d’/‘5d’: *p = 0.022, ‘6d’/‘7d’: *p = 0.028). Scale bar = 20 µm. Data represent the mean (lines) ± SEM (shaded regions); dots indicate individual experiments. Source data are provided in Source Data file.

Fig. 5. The 3 day-long recycling observed by labeling with Fab fragments or His-tagged TNR.

a Assay to label molecules completing a full endocytosis/resurfacing cycle. (1) Surface TNR epitopes are blocked with TNR Fab fragments and non-fluorescent secondary nanobodies. (2) 4 h later, newly-emerged epitopes are tagged with new Fab fragments, without secondary nanobodies. (3) Following a 12-h incubation, allowing for internalization, newly-emerged epitopes remaining at the surface are blocked with non-fluorescent nanobodies. (4) Immediately afterwards, or 1–3 days later, the newly-emerged and then internalized epitopes that resurfaced are revealed with fluorophore-conjugated secondary nanobodies. b Neurons were imaged in epifluorescence microscopy. Substantial fluorescence is visile at both the 1- and 3-day time points. N = 3 independent experiments, ≥15 neurons per datapoint. Kruskal-Wallis (H₂ = 7.2, *p = 0.0273), followed by two-sided Dunn’s multiple comparisons test: ‘0 d’/‘3 d’: *p = 0.0199. Scale bar = 10 µm. Data represent mean ± SEM, dots indicate individual experiments. c–e Recycling of recombinant His-tagged TNR (rTNR). c rTNR distributes similarly to endogenous TNR, after pulsing neurons with rTNR for 1 h and staining with WFA to label PNNs (epifluorescence). Scale bar = 20 µm. N = 3 independent experiments. d rTNR recycling assay: (1) Neurons were pulsed with rTNR for 1 h, and then incubated for 0–3 days, allowing for internalization and recycling (2). Neurons were fixed immediately (3), or first incubated with proteinase K to remove surface-bound rTNR (3’). Neurons were permeabilized and immunostained with anti-His tag antibodies to reveal all rTNR (4), or internalized rTNR (4’). e At time = 0, rTNR staining was strongly reduced by stripping. At 1d, similar staining was observed in stripped/non-stripped cultures. At 3d, staining was again reduced after stripping. Scale bar = 10 µm. N = 3 independent experiments, 5 before/after images per datapoint. Repeated-measures one-way ANOVA (F_1.044,2.088 = 28,6, *p = 0.03), followed by Fisher’s LSD (‘0 d’/‘1 d’: **p = 0.002; ‘1 d’/‘3 d’: *p = 0.027; ‘0 d’/‘3 d’: p = 0.775). Data represent mean (lines) ± SEM (shaded regions), dots indicate individual experiments. Source data are provided in Source Data file.

Fig. 6. An overview of organelles involved in the trafficking of newly-emerged TNR epitopes.

a Newly-emerged TNR epitopes were labeled 12 h post-blocking, concurrently with the application of LysoTracker™ Green, to label acidic organelles. After a 6-h incubation, allowing for internalization, surface TNR was stripped with proteinase K. Neurons were imaged live (epifluorescence). Scale bar = 4 μm. >70% of internalized TNR is present in acidic organelles. N = 3 independent experiments, ≥4 neurons per datapoint. Data represent mean ± SEM, dot indicate individual experiments. b–g To identify the compartments containing internalized TNR, newly-emerged TNR epitopes were labeled 12 h post-blocking and allowed to internalize for 6 h, after which remaining surface-bound TNR was stripped with proteinase K. The neurons were fixed and immunostained with organelle markers. Shown are 2-color-STED images of TNR (magenta) and organelle markers (green): caveolin1, Rab11a (recycling endosomes), LAMP1 (lysosomes), TGN38 (trans-Golgi network) and calreticulin (ER). The right side of each panel shows zoomed views of the dashed boxes. Arrowheads indicate colocalizing signals. Scale bar = 2 μm (full images), 500 nm (zoomed images). g Quantification of % TNR spots colocalizing with organelle markers, compared to a negative control (using non-specific primary antibodies). TNR colocalizes significantly with ER, TGN, LAMP1, Rab11a and caveolin. N = 3 independent experiments, ≥10 neurons per datapoint. One-way ANOVA (F_8,18 = 4.284, **p = 0.005), followed by Fisher’s LSD to compare all markers with ‘neg ctrl’ (Caveolin1: **p = 0.002; Rab5: p = 0.099; Rab7: p = 0.126; Rab11a: *p = 0.017; Rab11b: p = 0.169; LAMP1: **p = 0.005; TGN38: ***p < 0.001; Calreticulin: ***p < 0.001). Data represent mean ± SEM, dots indicate individual experiments. h A fraction of newly-emerged TNR localizes to dendritic Golgi outposts following endocytosis. Newly-emerged TNR epitopes (magenta) were labeled 4 h post-blocking, and allowed to internalize over 12 h. The neurons were fixed and immunostained with TGN38 to identify dendritic Golgi outposts^,. Representative images, taken with confocal microscopy, are shown. Arrowheads indicate colocalizing signals. Scale bar = 2 µm. N = 4 independent experiments. Source data are provided in Source Data file.

Fig. 7. TNR recycling is mediated by integrins.

a, b Assessment of colocalization between recycling TNR molecules and β1-integrin. a Left: newly-emerged TNR epitopes were labeled 12 h post-blocking concurrently with a labeling of surface-bound β1-integrins, by applying fluorophore-conjugated antibodies directed against the extracellular domain of the receptors. The neurons were fixed and imaged with 2-color-STED. Right: newly-emerged TNR epitopes were labeled 4 h post-blocking, concurrently with β1-integrin. Neurons were incubated a further 12 h to allow for internalization, and remaining surface-bound molecules were stripped with proteinase K. The neurons were fixed and imaged with confocal microscopy. Images on the right of each panel show zoomed views of the dashed boxed. Scale bars = 2 µm (full images), 500 nm (zoomed images). b Quantification of % colocalizing TNR signal (for a) shows newly-emerged TNR epitopes colocalize with both cell surface-bound and internalized β1-integrins. The values are significantly higher than negatives controls, relying on non-specific primary antibodies. Controls were imaged in STED/confocal for comparison to images in the left/right panels, respectively. N = 3 (‘surface β1-integrin’ experiments and negative controls), and 4 (‘internalized β1-integrin’) independent experiments, ≥10 neurons per datapoint. Two-sided Student’s t-test (‘surface integrin’ vs. ‘neg ctrl’: t = 11.61, ***p = 0.0003; ‘internalized integrin’ vs. ‘neg ctrl’: t = 3.177, *p = 0.025. Data represent mean ± SEM, dots indicate individual experiments. c To assess whether β1-integrin receptors are required for TNR endocytosis, newly-emerged TNR epitopes were labeled 12 h post-blocking, after which the neurons were immediately incubated with function-blocking anti-β1-integrin antibodies for 6 h. Neurons were then incubated with proteinase K, to remove remaining surface-bound TNR, and imaged with epifluorescence microscopy. A reduction in fluorescence signal is evident in integrin-blocked cultures. Scale bar = 5 μm. d Quantification of the fluorescence intensity confirms that the amount of internalized TNR is significantly reduced following the blocking of β1-integrin receptors. N = 3 independent experiments, ≥15 neurons per datapoint. Two-sided Student’s t-test (t = 3.343, *p = 0.029). Data represent mean ± SEM, dots indicate individual experiments. Source data are provided in Source Data file.

Fig. 8. TNR recycling possibly relates to TNR re-glycosylation.

Newly O-glycosylated proteins were labeled by feeding neurons with azide-modified galactosamine (GalNAz) and/or glucosamine (GlcNAz), which were then revealed by click chemistry. Alternatively, newly-synthesized proteins were labeled by feeding neurons with azidohomoalanine (AHA), which was also tagged using click chemistry. Newly-emerged TNR epitopes were labeled 6 h post-blocking and visualized at the surface. The neurons were imaged with 2-color-STED. Scale bar = 1 µm. Quantification of the colocalization of the signals confirmed that internalized TNR epitopes colocalize significantly with GalNAz or GalNAz+GlcNAz, at levels substantially higher than the minimal AHA colocalization, which is not significantly different from negative controls (relying on non-specific primary antibodies). N = 4 independent experiments, ≥10 neurons per datapoint. Kruskal-Wallis test (H₃ = 9.022, *p = 0.029), followed by two-sided Fisher’s LSD (*p = 0.014 and *p = 0.021 for ‘GalNAz’ and ‘GalNAz+GlcNAz’ respectively). Data represent mean ± SEM, dots indicate individual experiments. Source data are provided in Source Data file.

Fig. 9. Perturbing the recycling TNR pool modulates synaptic function.

a Assay to perturb TNR recycling: newly-emerged TNR epitopes were labeled 12 h post-blocking with biotinylated antibodies, and bound to large aggregates of antibodies. As control, all other epitopes (non-recycling) were labeled. b STED images of aggregates. Scale bar = 1 µm. c Histogram of aggregate size (FWHM). N = 4 independent experiments, 995 aggregates. d Neurons were incubated with aggregates for 30 min. Synaptic activity was assessed by uptake of Syt1 antibodies (as in Fig. 3). Without stimulation, Syt1 antibodies detect the surface vesicle population (40–50% of actively-recycling vesicles). Stimulation results in signal increase (exo-/endocytosis of new vesicles) in controls, but not in aggregate-treated cultures (epifluorescence). Scale bar = 4 µm. e Quantification of Syt1 fluorescence intensity confirms this observation and indicates that tagging all other epitopes has no effects. N = 4 (‘new epitopes’)/3 (‘all other’) independent experiments, ≥15 neurons per datapoint. Repeated-measures ANOVA on rank (‘new epitopes’: F_1,6 = 12.54, *p = 0.012; ‘all other epitopes’: F_1,4 = 1.5, p = 0.288) for the interaction Stim/ctrl x + /− Aggregates), followed by Sidak’s multiple comparisons test (‘new epitopes’: *p = 0.02, p = 0.419; ‘all other epitopes’: **p = 0.002, ***p < 0.001 for ‘stim’ vs. ‘ctrl’ for untreated and treated neurons, respectively). f, g Effect of recycling perturbation on synapse structure. Dissociated cultures (f) and organotypic hippocampal slices (g) were treated with aggregates for 12 h. Plasma membranes were visualized with DiO (f) or by infection with AAV9-Syn-eGFP (g), in averaged spines or individual examples (insets). Scale bar = 300 nm (f), 500 nm (g). N = 3 independent experiments, >80 (f), >60 (g) synapses per condition. One-way ANOVA (f: F_{2, 6} = 5.269, *p = 0.05) or repeated-measures one-way ANOVA (g: F_{1.041, 2.083} = 20.76, *p = 0.042), followed by Fisher’s LSD (f:** p = 0.005, p = 0.418; g: *p = 0.025, p = 0.16), to compare ‘all other epitopes’/‘new epitopes’ and ‘all other epitopes’/‘Tyrode’, respectively. Data represent mean ± SEM, dots indicate individual experiments (d–g). Source data are provided in Source Data file.

Fig. 10. TNR dynamics are observed in brain slices from adult mice, and are altered in an epilepsy model.

a, b Intracellular TNR in disease models. Hippocampal slices from kainic acid (KA)-induced epilepsy model mice and 5xFAD familial Alzheimer’s disease model mice were immunostained for TNR and the ER marker calreticulin, to enable identification of intracellular (somatic) TNR. All other TNR was presumed extracellular. a Imaged regions (confocal) from mice pre-treated with vehicle or KA. The proportion of intracellular TNR is increased in KA-treated mice. N = 3 mice per treatment, 60 (vehicle) and 67 (KA) regions analyzed. Kruskal-Wallis (H₃ = 27.93, ***p < 0.001), followed by two-sided Dunn’s multiple comparisons test (‘vehicle; extracellular TNR’/‘vehicle; intracellular TNR’: ‘p = 0.932; ‘KA; extracellular TNR’/‘KA; intracellular TNR’: ***p < 0.001; ‘vehicle; intracellular TNR’/‘KA; intracellular TNR’: **p = 0.004). b Similar analysis for 5xFAD mice. No significant differences are observed. N = 3 mice per treatment, 68 (WT) and 29 (5xFAD) regions analyzed. Kruskal-Wallis (H₃ = 3.233, p = 0.357). Scale bar = 5 µm. c–e Isotopic imaging in adult mice suggests intracellular TNR is not newly synthesized. TNR turnover in vivo was measured with correlative fluorescence and isotopic imaging (COIN^,,,,) in brain slices of mice pulsed with isotopically stable ¹³C₆-lysine for 14 or 21 days (previously characterized in). c Top: section stained for TNR and calreticulin (epifluorescence). Bottom: nanoSIMS images of ¹²C¹⁴N⁻ (left) and ¹³C¹⁴N^{− (}middle) secondary ions. The ¹³C¹⁴N⁻/¹²C¹⁴N⁻ ratio image (right) indicates the enrichment of ¹³C. Scale bar = 4 µm. d Zoom of square regions in c. Scale bar = 500 nm. e Quantification of ¹³C¹⁴N⁻/¹²C¹⁴N ratio as fold over the natural abundance level. TNR-enriched areas exhibit the lowest ¹³C enrichment in these cells (lowest newly synthesized protein levels). N = 6 sections from 3 mice per condition. Kruskal-Wallis (H₂ = 167.2, ***p < 0.001), followed by two-sided Dunn’s multiple comparisons test (***p < 0.001 for all comparisons). For all panels: boxes show median (mid-line) and quartiles, whiskers show minimum/maximum values. Outliers were omitted according to inter-quartile range (IQR) proximity (exceeding 1.5*IQR). Source data provided in Source Data file.