Fig 2: The microglial metalloproteinase Mmp14b restricted ECM accumulation and destabilized newborn synapses.a, Schematic of microglial ablation by adding Mtz to fish water in fish that express nitroreductase (ntr) in microglia and macrophages (Tg(mpeg:gal4);Tg(UAS:NTR-mCherry)). b, Representative images of brevican in synaptic region after vehicle (DMSO) or microglial ablation with 5 mM Mtz in fish water for 24 h. Experiment at 14 dpf. c, Quantification of brevican intensity in synaptic region normalized to mean of vehicle control (vehicle, n = 15 fish; Mtz, n = 12 fish; *P = 0.043, Welch’s t-test). d, Heatmap of absolute expression (counts) of mmp, adam and adamts metalloproteinase genes in zebrafish microglia from bulk RNA sequencing at 28 dpf in optic tectum (OT), midbrain (MB) and hindbrain (HB). Marker genes hexb and p2ry12 are shown for comparison. Reanalyzed from ref. 35. e, Schematic of a dimer of the membrane-associated metalloproteinase MMP14 and design of the mmp14b-HA expression construct for epitope tagging of the C-terminal end with HA. Catalytic domain, transmembrane domain and cytoplasmic domain are indicated. f, Mmp14b-HA transgene expression in an mpeg-GFP microglia expressing mpeg:mmp14b-HA. Inset shows HA signals colocalized with microglial processes (arrowheads). Similar results were observed from n = 2 individual experiments. g, Quantification of Mmp14b-HA intensity in microglial processes versus soma. Lines connect data from the same microglia (n = 6 microglia from 3 fish; *P = 0.015, paired t-test per cell). h, Representative images of brevican staining in synaptic regions of mmp14b−/− and mmp14b+/+ control at 14 dpf. i, Quantification of brevican intensity in synaptic region normalized to mean of mmp14b+/+ control (mmp14b+/+, n = 10 fish; mmp14b−/−, n = 11 fish; **P = 0.0014, Welch’s t-test). j, Representative images of microglia-specific Mmp14b rescue experiment. Chat-PSD95FingR;mmp14b+/+ and ;mmp14−/− fish were injected with mpeg:mmp14b-HA construct at one-cell-stage embryos and analyzed at 14 dpf. k, Quantification of synapse density (Chat-PSD95FingR puncta per µm of dendrite length) in mmp14b+/+ and mmp14−/− with or without microglial-specific rescue with mpeg:mmp14b-HA. Dots show means per fish from n = ~1–4 dendritic segments analyzed per fish (mmp14+/+ no rescue, n = 33 dendrites from n = 16 fish; mmp14b+/+ rescue, n = 31 dendrites from n = 14 fish; mmp14b−/− no rescue, n = 27 dendrites from n = 16 fish; mmp14b−/− rescue, n = 40 dendrites from n = 18 fish; two-way ANOVA for fish, **F(1,60) interaction effect P = 0.0022, asterisks showing Tukey’s multiple comparisons). MG, microglia. l, Schematic of time-lapse imaging to measure synapse turnover in Chat-PSD95FingR;mmp14b+/+ and ;mmp14b−/− fish. m, Representative merged images of synapses in Chat-PSD95FingR;mmp14b−/− and ;mmp14b+/+ fish at t = 0 (pink) and t = 24 (green). Experiments at 10–12 dpf. Pink arrowheads, lost synapses; green arrowheads, newly observed synapses. Nonmerged images are shown in Extended Data Fig. 8d. n, Quantification of newly observed, lost and stable synapses between t = 0 and t = 24 in mmp14b+/+ versus mmp14b−/− fish (mmp14b+/+, n = 14 fish; mmp14b−/−, n = 16 fish; ****P < 0.0001 for newly observed, P = 0.058 for lost, P = 0.61 for stable, Welch’s t-test). o, Schematic of time-lapse imaging to determine the fate of individual synapses in mmp14b−/− fish. Synapses at t = 0 were defined as ‘stable’ synapses. Synapses born between t = 0 and t = 6 were defined as ‘new’ synapses and subsequently followed with the 6-h time point set as t = 0 (lower time course). Experiments performed at 10–12 dpf. p, Representative image of a single excitatory synapse imaged at t = 0, 6, 12 and 24 shows a ‘new synapse’ born between t = 0 and t = 6 in mmp14b−/− fish. Raw fluorescence images on top and fluorescence overlaid with 3D reconstruction of synapse on bottom. Arrowheads, newborn synapse. Dashed circle, site of newborn synapse. q, Kaplan–Meier plot of survival of individual synapses over time in mmp14b+/+ control (from Fig. 1m) and mmp14b−/− fish (for mmp14b−/−, data from n = 15 fish, n = 331 stable synapses and n = 26 new synapses). Values were plotted as mean ± s.e.m. ****P < 0.0001; **P < 0.01; *P < 0.05. Scale bars, 20 µm (b,h), 5 µm (f), 5 µm (j,m), 2 µm (p). Illustrations in a and e created using BioRender.com.Source data

Fig 3: Additional characterization of mmp14b deficient fish.a) Representative images and quantification of the number of microglia having 4C4 microglial marker in mpeg-GFP fish (mmp14b+/+, n = 8 fish; mmp14b−/−, n = 7 fish; p = 0.468, Welch’s t-test). Dashed lines indicate hindbrain regions. Scale: 100 µm. b) Representative images and quantification of the total length of microglial processes (mmp14b+/+, n = 23 microglia from n = 7 fish; mmp14b−/−, n = 23 microglia from n = 7 fish; p = 0.026*, Kolmogorov-Smirnov test, performed for microglia). Scale: 5 µm. c) Representative images and quantification of synapse density per µm of dendrite length (Chat-PSD95FingR puncta) in mmp14b−/− and mmp14b+/+ control at 14 dpf. Dots represent means per fish from n = 1-3 dendritic segments analyzed per fish (mmp14b+/+, n = 21 dendrites from n = 11 fish; mmp14b−/−, n = 17 dendrites from n = 11 fish; p = 0.0044**, Welch’s t-test for fish). Scale: 5 µm d) Representative images and quantification of SV2 presynapse marker in mmp14b−/− fish at 14 dpf (mmp14b+/+, n = 7 fish; mmp14b−/−, n = 9 fish, p = 0.035*, Welch’s t-test). SV2 intensity normalized to mean of mmp14b+/+ control. Dashed lines indicate hindbrain regions. Scale: 100 µm. e) Schematics of the astrocytic mmp14b-HA expression construct. f) Representative images and quantification of synapse density per µm of dendrite length (Chat-PSD95FingR puncta) with or without astrocytic Mmp14b rescue. Chat-PSD95FingR;mmp14b+/+ and;mmp14−/− fish were injected with slc1a3b:mmp14b-HA construct at one-cell stage embryos and analyzed at 14 dpf. Dots show means per fish from n = 1 ~ 4 dendritic segments analyzed per fish (mmp14b+/+ no rescue, n = 47 dendrites from n = 22 fish; mmp14b+/+ rescue, n = 47 dendrites from n = 23 fish; mmp14−/− no rescue, n = 48 dendrites from n = 24 fish; mmp14−/− rescue, n = 52 dendrites from n = 24 fish; Two-way ANOVA for fish, F(1,89) interaction effect p = 0.0083**, asterisks show Tukey’s multiple comparisons). Scale: 5 µm. g) Non-merged images of time lapse imaging experiment from Fig. 3m. Circles indicate stable synapses. Boxes indicate the region shown in Fig. 3m. Scale: 5 µm. Values were plotted as mean ± SEM.***: p < 0.001; **: p < 0.01; *: p < 0.05; ns: not significant. Source data

Fig 4: Computational modeling supported the existence of two synapse states and a role of ECM in stabilizing newborn synapses.a, Model of synapse dynamics in our system, based on four parameters: birth rate of new synapses (bn), decay rates of new synapses (dn), probability of transition from new to stable synapse (cn→s) and decay rate of stable (ds) synapses. b, Fitting of synapse lifetime data in Fig. 1m from Chat-PSD95FingR fish, based on the two-state model described in a. Densities of synapses for each category of lifetime are shown from raw data (gray) and as predicted by the two-state model (purple). The numbers in parentheses on the x axis indicate the first and last time (in hours) that synapses were detected over 72 h. Note that while experimental analyses in Figs. 1 and 3 tracked only synapses that were present at 0 h (‘stable’) or born between 0 and 6 h (‘new’), the computational model examined all synapses observed over 72 h of imaging to generate this predicted distribution. c, Synapse dynamics calculated from the two-state model in mmp14+/+ and mmp14b−/− fish, showing calculations of the four parameters described in a. Units for synapse birth rate (bn) reflect synapses born per µm per hour, shortened to 1 µm−1 h−1. Birth rate is independent of the number of synapses in the system. Units for rates of synapse decay or stabilization (dn, cn→s and ds) reflect the rates as a proportion of existing synapses, that is, changed synapses per µm per hour divided by total synapses per µm, shortened to 1 h−1. Error bars show ±s.d. ****P < 0.0001, Mann–Whitney U-test. See Supplementary Data 1 for exact P values. d, Graphical data summary. e, Graphical abstract of ECM impact on synapse numbers and dynamics. Illustrations in a, d and e created using BioRender.com.

Fig 5: Validation of MMP14 knock-down in human iPSC derived Microglia.a) MMP14 mRNA levels from MMP14 KD microglia analyzed by RT-qPCR showing two independent shRNAs targeting MMP14, with scrambled shRNA as control. Expression normalized to GAPDH housekeeping gene (n = 3 independent experiments; One-way ANOVA with Tukey’s multiple comparison). b) MMP14 mRNA levels from MMP14 KD and MMP14-GFP rescue microglia analyzed by RT-qPCR. Expression normalized to GAPDH (n = 3 independent experiments; one-way ANOVA, asterisks in the figure show results of Tukey’s multiple comparison). c) Western blot of MMP14 showing MMP14 KD and rescue with MMP14-GFP in microglia. Note higher molecular weight of MMP14-GFP vs. endogenous MMP14. d) Quantification of the microglia cell number in the triculture system with MMP14 KD and MMP14-GFP rescue (n = 3 independent experiments, p = 0.209, one-way ANOVA). e) Representative images and quantification of process length in the microglia mono-culture system with MMP14 KD and rescue. Dots show mean microglial process length per field of view from 24 fields of view over 3 independent experiments. Box-and-whiskers plot: box shows 25-75th percentile, whiskers show min to max. Statistics performed on means per field of view. (Kruskal-Wallis test, asterisks in the figure show results of Dunn’s multiple comparisons). Scale: 100 µm. f) Quantification of the microglia cell number in the microglia mono-culture system with MMP14 KD and MMP14-GFP rescue (n = 3 independent experiments, one-way ANOVA, asterisks in the figure show results of Dunn’s multiple comparison). g) Representative western blot of Brevican in cell supernatant vs. MAP2 loading control in cell lysate from the tri-culture system with MMP14 KD. h) Quantification of brevican in the cell supernatant in tri-culture system with MMP14 KD. Each dot represents mean brevican intensity normalized to no microglia control from n = 4 independent experiments (n = 4 independent experiments, p < 0.0001****, One-way ANOVA, asterisks in figure show results of Tukey’s multiple comparison). i) Representative images and quantification of synapse puncta in the tri-culture system with or without MMP14 KD in microglia. The number of vglut2+ puncta per image field were quantified (n = 6 technical replicates for each group, p = 0.043*, Welch’s t-test). Scale: 10 µm. Values were plotted as mean ± SEM.****: p < 0.0001; *: p < 0.05; ns: not significant. Source data