MT5-MMP promotes neuroinflammation, neuronal excitability and Aβ production in primary neuron/astrocyte cultures from the 5xFAD mouse model of Alzheimer’s disease

To generate mixed neuronal–glial cell cultures, we used WT, MT5^−/−, Tg and TgMT5^−/− mice in a C57BL6 genetic background as previously described [13, 17]. All the experimental procedures were conducted in agreement with the authorization for animal experimentation attributed by the French Ministry of Research to the laboratory (research project: APAFIS#23040-2019112708474721 v4). Briefly, pregnant females were deeply anesthetized with xylazine (15 mg/kg) and ketamine (150 mg/kg) (Ceva Santé animale, Libourne, France), and E16 embryos were extracted from the uterine horns and cerebral cortices dissected. All the culture media, fetal bovine serum (FBS), reagents and supplements for cell culture were purchased from ThermoFisher Scientific (Villebon-sur-Yvette, France). Cortices were placed into cold HBSS1X medium and dissociated for 10 min at 37 °C in HBSS1X containing DNAse I (10 µg/mL) and 0.1% trypsin. Reaction was stopped by the addition of a DMEM solution containing 10% FBS and further mechanical dissociation was performed through a pipette cone. After centrifugation for 5 min at 300×g, 3.10⁵ cells/well were plated onto 6-well plates pre-coated with poly-l-lysine (10 μg/mL, Sigma-Aldrich, Saint-Quentin Fallavier, France) for 2 h in DMEM medium containing 10% FBS and 1% penicillin/streptomycin (P/S). This medium was further replaced by Neurobasal containing B27, 1% glutamine and 1% P/S for 11 days in vitro (DIV) without anti-mitotic agent. Cells were treated or not with IL-1β (10 ng/mL, PeproTech, Neuilly-sur-Seine, France) and/or DAPT (10 μM, Tocris, Bio-Techne, Lille, France) or proteasome inhibitor MG132 (5 μM, Enzo Life Science, Lyon, France), 24 h before being collected in either RIPA buffer (Sigma-Aldrich) for western blot (WB) analysis or collected for RNA extraction. For immunocytochemistry (ICC) experiments, cells were plated at 1.10⁵ density on 24-well plates on coverslips pre-coated with 500 μg/mL of poly-l-lysine. For electrophysiological experiments, cells were plated as described above for ICC and recorded between 11 and 14 DIV.

Cell viability was evaluated using the 3-(4,5-dimethylthiazol-2yl)-2,5-diphenyl tetrazolium bromide (MTT) assay (Sigma-Aldrich), which measures mitochondrial activity in living cells. A solution at 5 mg/mL was prepared into Neurobasal and then added to cultures at a final concentration of 0.5 mg/mL for 3 h at 37°C, 5% CO₂. Media were fully removed and 200 μL of DMSO added, then 100 μL of DMSO were transferred into a 96-well plate and absorbance (OD) at 550 nm was read in a spectrophotometer. Data were calculated as the percentage of living cells = (transfected cell OD₅₅₀/control cell OD₅₅₀) × 100. The mean values ± SEM were obtained from at least five animals by genotype.

An empty AAV10 or encoding human C99 under control of the synapsin-1 promoter (AAV-empty or AAV-C99 thereafter) were previously described [29]. WT and MT5^−/− neurons were transduced at 6 DIV with 2 μL (at 5.10¹² vg/mL, MOI = 2.5 × 10⁴), treated at 10 DIV with DAPT (10 μM) and recovered at 11 DIV for WB analyses.

Protein concentration was determined using a Bio-Rad DC™ protein assay kit (Bio-Rad, Marnes-La-Coquette, France). Proteins (30 μg) were loaded and run on 10–15% SDS-PAGE gels, or 4–20% Tris–Glycine pre-casted gels or low molecular weight 16% Tris–Tricine pre-cast gels (ThermoFisher Scientific) and transferred onto nitrocellulose membranes (Dutscher, Brumath, France). After blocking, membranes were probed with the following antibodies directed against MT5-MMP (1/500, our own antibody previously described [13]), or APP N-terminal fragment (22C11, 1/1000, Millipore, Merck Millipore, Molsheim, France), APP C-terminal fragment (APP-CTF, 1/1000, Sigma-Aldrich), human Aβ (6E10, 1/500, Ozyme, Saint-Cyr l’Ecole, France), Aβ/C99 (82E1, 1/100, IBL America, Illkirch-Graffenstaden, France), GFAP (1/1000, Millipore), IL-1β (1/500, PeproTech), N-cadherin (1/1000, BD Biosciences, Le Pont de Claix, France), MAP-2 (1/500, Sigma-Aldrich), β-III tubulin (1/1000, Sigma-Aldrich), LRP-1 (1/1000, Abcam, Cambridge, United Kingdom), RAGE (1/1000, Abcam), LDLR (1/1000, Proteintech Europe, Manchester, United Kingdom), LRP-8 (1/250, Abcam), Histone 3 (1/1000, Abcam), Na⁺/K⁺ ATPase (1/1000, Abcam), β-actin (1/5000, Sigma-Aldrich), GAPDH (1/5000, Sigma-Aldrich), and then incubated with horseradish peroxidase-conjugated secondary IgG antibodies (Jackson Immunoresearch, Interchim, Montluçon, France). Note that depending on the molecular weight of the proteins studied and thus the gels used, GAPDH, β-actin or ponceau S staining of the membrane [30, 31] were used as loading and normalization controls. Immunoblot signals were visualized using the ECL chemiluminescence kit (Dutscher) and quantified using Fiji/Image J software (NIH). Note that immunoblots were represented in separated columns when bands from the same membrane were not adjacent.

Cytoplasmic, membranous, and nuclear fractions were prepared from cell lysates using a ProteoExtract^® Subcellular Proteome Extraction Kit (Calbiochem, Merck Millipore, Molsheim, France) according to the manufacturer’s instructions. The purity of each fraction was analyzed by western blot as described above, using antibodies against GAPDH, Na⁺/K⁺ ATPase and Histone 3 for cytoplasmic, membranous and nuclear fractions, respectively.

We used gelatin zymography on culture supernatants to assess changes in the levels of MMP-2 and MMP-9, also known as gelatinase A and B, respectively. As previously described [32], equal amounts of serum-free supernatants in non-denaturing and non-reducing conditions were subjected to zymography according to the manufacturer’s recommendations (ThermoFisher Scientific). Gels were scanned using GeneTools software.

Total RNA was extracted from 11 DIV cells using the Nucleospin RNA kit (Macherey–Nagel, Hoerdt, France) according to the manufacturer’s recommendations. All the reagents for RT-qPCR experiments were purchased from ThermoFisher Scientific. Single-stranded cDNA was synthesized from 500 ng of RNA using the High-Capacity RNA to cDNA™ kit adapted for quantitative PCR. Twenty-five ng of cDNA were subjected to a qPCR reaction using the Fast Real-Time PCR System. For each experiment, cDNA samples were analyzed in duplicate and relative gene expression was obtained using the comparative 2^−(ΔΔCt) method after normalization to the Gapdh (Mm99999915_g1) housekeeping gene [32, 33]. The expression of the following genes was measured: Mmp24 (Mm00487721_m1), Mmp14 (Mm00485054_m1), Il-1β (Mm01336189_m1), Gfap (Mm01253033_m1), Ccl2 (Mm00441242_m1), Mmp2 (Mm00439498_m1), Mmp9 (Mm00442991_m1), Ide (Mm00473077_m1), Ace (Mm00802048_m1), Ece (Mm01187091_m1), Tnfa (Mm00443258_m1), Bace1 (Mm00478664_m1), Psen1 (Mm00501184_m1), Adam10 (Mm00545742_m1), Lrp1 (Mm0046458_m1), Lrp8 (Mm00474023_m1), Ldlr (Mm00440169_m1), Ager (Mm00545815_m1), APP (Hs00169098_m1), PSEN1 (Hs00997789_m1).

After 11 DIV, our neural cultures were fixed for 15 min with AntigenFix (Diapath, MM France, Brignais, France) and blocked with PBS1X, BSA 3%, 0.1% Triton X-100 (blocking solution) for 1 h. Primary antibodies GFAP (Dako France, Trappes, France), Iba1 (Wako, Sobioda, Mont-Bonnot Saint-Martin, France), β-III tubulin (Sigma-Aldrich), were used at 1/500 dilution. Appropriate AlexaFluor-coupled secondary antibodies were used at 1/800 dilution. Hoechst 33342 (0.5 μg/mL) was used to stain the nuclei. Antibodies and Hoechst were diluted in blocking solution. Omission of the primary antibody was used as control and no immunostaining was observed. Samples were mounted using Prolong Gold Antifading reagent on Superfrost glass slides (Dutscher). Images were taken and processed using a confocal microscope (LSM 700) and Zen software (Zeiss, Jena, Germany).

Total Aβ38, Aβ40 and Aβ42 levels in culture supernatants were assessed by ELISA using the V-PLEX Plus Aβ Peptide Panel 1 (4G8) Kit (Meso Scale Discovery, Rockville, Maryland, USA) according to the manufacturer's recommendations. The MSD kit uses 4G8 as a capture antibody that recognizes both mouse and human Aβ. Specific Aβ38, Aβ40 and Aβ42 were used for detection. Therefore, the measured Aβ levels are a mixture of endogenous mouse Aβ and human Aβ derived from the processing of human APP. Human Aβ40 levels in culture supernatants after AAV-C99 infections were evaluated using the human Aβ40 ELISA kit (#KHB3481, ThermoFisher Scientific). For detection of IL-1β and MCP-1 in supernatants, we used the murine IL-1β and MCP-1 ELISA Development Kits (PeproTech). IL-6 protein levels were measured using V-PLEX Proinflammatory Panel 1 mouse Kit (K15048D-1, Meso Scale Discovery, Rockville, Maryland, USA) and analyses were done using a QuickPlex SQ 120 instrument (MSD) and DISCOVERY WORKBENCH^® 4.0 software. All these assays were used as recommended by the manufacturers.

All values represent the means ± SEM of the number of independent cultures indicated in the figure legends. For statistical analyses, we used ANOVA followed by a Fisher’s LSD post hoc test and set the statistical significance at p < 0.05. Analyses were performed with the GraphPad Prism software (San Diego, California USA).

Cultures were roughly estimated at 2/3 of neurons (β-III tubulin⁺) and 1/3 of astrocytes (GFAP⁺), while no microglia was detected (Iba1⁺). A representative image of WT neural cells, treated or not with IL-1β, is shown in Fig. 1D. Moreover, levels of the β-III tubulin (Fig. 1E) and MAP-2 (not shown) neuronal markers were stable across genotypes in all experimental conditions, as revealed by WB. Likewise, there was no change in the content of the astrocytic marker GFAP (Fig. 1F). Moreover, the MTT test confirmed no cytotoxic effects associated with genotypes or IL-1β treatment in our conditions (Fig. 1G).

IL-1β is a key cytokine in AD [20] that stimulates its own expression [35, 36] as well as that of other inflammatory mediators, including Il-6, Ccl2 and Tnfa [37–40]. In pace with these data, IL-1β induced its own mRNA in WT (174%) and Tg (109%) cultures (Fig. 2E), but not in cells lacking MT5-MMP (MT5^−/− and TgMT5^−/−) (Fig. 2E). Ccl2 mRNA levels, which were relatively low in basal conditions, respectively reached 5000% and 4000% increases in WT and Tg cells after IL-1β exposure (Fig. 2F). Again, the stimulating effect of IL-1β on Ccl2 was hampered by MT5-MMP deficiency, with mRNA levels significantly reduced by 55% and 49% in MT5^−/− and TgMT5^−/− cells compared with their respective WT and Tg controls. IL-1β treatment upregulated Il-6 levels by 318% in WT, 79% in MT5^−/− and 188% in TgMT5^−/− cells, but had no effect in Tg cells, whose Il-6 levels were down by 45% compared to IL-1β-treated WT (Fig. 2G). On the contrary, unlike other neuroinflammatory mediators, Tnfa expression was not affected by IL-1β in the present experimental conditions (Fig. 2H). Overall, MT5-MMP deficiency modulates the expression of Il-1β, Ccl2 and Il-6 while sparing that of Tnfa, which is otherwise unaffected by genotypes or 10 ng/mL IL-1β treatment.

Inflammatory challenge with IL-1β strongly upregulated MCP-1 levels in all genotypes, but they remained significantly lower by 29% in TgMT5^−/− cultures compared with Tg (Fig. 3D). The burst in IL-1β concentration detected by ELISA in cell supernatants after treatment likely reflects the addition of recombinant cytokine, although it cannot be excluded that a small portion results from endogenous synthesis. In any case, IL-1β levels in TgMT5^−/− cells were significantly lower (45%) compared with Tg (Fig. 3E). IL-6 levels were also strongly upregulated by IL-1β, but in this case the major change was the 38% reduction in Tg levels compared to WT and the recovery of IL-6 levels in TgMT5^−/− cells to near WT values (Fig. 3F). After treatment with recombinant IL-1β, a 17 kDa immunoreactive band matching the size of the active form of the cytokine, and absent in untreated cultures, was detected by WB in cell supernatants, (Fig. 3G). The band intensity was reduced in TgMT5^−/− (62%) and MT5^−/− (56%) cells, compared with Tg and WT, respectively (Fig. 3G). The reductions in extracellular IL-1β, prompted us to assess its content in cell lysates, inferring a possible increase in cellular uptake of the cytokine and consequent intracellular accumulation. However, this was not the case, as cell lysates also revealed a 46% reduction of IL-1β levels in TgMT5^−/− cells compared with Tg, while no differences were observed between MT5^−/− and WT cells (Fig. 3G).

These results raised the possibility that IL-1β may be more efficiently eliminated in the microenvironment of MT5-MMP-deficient cells. Two MT5-MMP homologs, MMP-2 and MMP-9 (also known as gelatinases A and B, respectively), have been shown to degrade IL-1β, thus likely contributing to the resolution of inflammation under certain circumstances [41]. We therefore evaluated a possible compensatory upregulation of these soluble MMPs that could explain IL-1β degradation upon MT5-MMP deficiency. However, no changes were observed in the mRNA levels encoding MMP-2 and MMP-9 (Fig. 3H). Moreover, a single band of gelatinolysis with the expected molecular weight of pro-MMP-2 appeared in highly sensitive gelatin zymograms, and this band remained stable across genotypes or after IL-1β treatment (Fig. 3I). The lower molecular weight active form of MMP-2 (~ 64 kDa) and MMP-9 (~ 90 kDa) were virtually undetectable in these conditions (Fig. 3I).

We next asked whether, more generally, proteolytic activities located in intracellular or extracellular environments could explain the putative degradation of IL-1β. To this end, we incubated recombinant IL-1β for 24 h in cell-free conditioned media from cell supernatants or lysates. In this case, IL-1β content remained unchanged between genotypes over a 24-h period (Fig. 3J). Taken together, these data suggest that recombinant IL-1β is taken up by cells and degraded more efficiently intracellularly in MT5-MMP-deficient cells, rather than by extracellular proteinases.

Figure 5D shows representative traces for each genotype in untreated (left) and treated (right) cultures. We recorded miniature global post-synaptic currents (gPSCs) in gap-free mode during 5 min, with the voltage clamped at − 50 mV. gPSCs were then analyzed off line and selected individually using the pClamp routine (Fig. 5D). The mean peak amplitude of gPSCs ranged from − 7 pA to − 25 pA, with a maximum value for Tg neurons and a minimum for the two MT5-MMP-deficient genotypes (Fig. 5E). The peak amplitude significantly increased by 69% in untreated Tg neurons compared with WT cells. Such increase was not observed in TgMT5^−/− neurons, which had 72% lower levels compared with Tg neurons. Likewise, the increase in peak amplitude in Tg neurons was prevented by IL-1β, which decreased levels to 64% of the value in untreated Tg neurons (Fig. 5E). In terms of basal event frequency, Tg neurons had a 536% higher value than WT neurons (Fig. 5F). Again, the increase was prevented in TgMT5^−/− neurons, where the levels decreased by 75% with respect to Tg. IL-1β treatment significantly exacerbated frequency in all genotypes, but the increase was surprisingly particularly important in the MT5^−/− and TgMT5^−/− groups, with > 1000% in both compared with their untreated controls (Fig. 5F). In comparison, frequency augmented by 500% in WT-treated neurons and by only 81% in Tg-treated neurons, relative to their untreated controls (Fig. 5F). Although basal frequency was already much higher in Tg cells compared to WT cells, IL-1β was still able to increase frequency in Tg cells by nearly twofold (Fig. 5F).

Next, we measured the levels of murine Aβ38, Aβ40 and Aβ42. Aβ38 was not detected in our cultures (not shown) and we found no increase of either species in Tg compared with WT cultures (Fig. 6C). However, MT5-MMP deficiency significantly reduced Aβ40 levels in MT5^−/− (48%) and TgMT5^−/− (41%) cells, compared with WT and Tg cells, respectively (Fig. 6C). Although IL-1β did not affect intragenotype Aβ40 levels, it caused a significant reduction in Tg (33%) and MT5^−/− cells (49%) compared with WT (Fig. 6C). Basal Aβ42 levels were approximately tenfold lower than those of Aβ40. The lack of MT5-MMP in this case reduced by 44% the levels of Aβ42 only in TgMT5^−/− cells compared with Tg (Fig. 6D). IL-1β had not effect on Aβ levels (Fig. 6C and D). We conclude that MT5-MMP deficiency downregulates Aβ40 and Aβ42 levels in developing neural cells in culture and that this trend is not modified by IL-1β after 24 h of incubation.

Human Aβ40 was detected only in Tg and TgMT5^−/− cells, albeit at relatively low concentrations (~ 35 pg/mL), as revealed by a human-specific ELISA kit (Additional file 2A), indicating that the Thy1 neuronal promoter was functional to drive human transgene expression and efficient metabolization of human APP. This is consistent with previous data showing activation of the Thy1 promoter at 4–5 DIV [46] and with the detection of hAPP and hPSEN1 mRNAs in our cultures (Additional file 2B and C). Genotype or IL-1β treatment did not affect the content of hAβ40, hAPP or hPSEN1 gene expression in any way in our experimental conditions (Additional file 2A–C).

Cellular receptors such as low-density lipoprotein receptor-related protein 1 (LRP1) [47–50], LRP8 [51], LDLR [52] or RAGE [53] may also affect APP metabolism by either modulating the activities of β- and γ-secretase and/or directly Aβ levels through endocytosis. In this context, Lrp1 and Lrp8 mRNA expression was stable in all experimental groups (Fig. 7B). In contrast, Ldlr mRNA content was significantly upregulated by 231% in TgMT5^−/− compared with Tg, and IL-1β did not alter this trend. The RAGE receptor encoded by the Ager gene, showed no differences upon IL-1β stimulation. Nevertheless, under basal conditions, MT5^−/− and Tg cells expressed significantly more Ager than WT cells (67% and 92%, respectively) (Fig. 7B). Next, we assessed the protein content of these receptors by WB. In our experimental conditions, LRP8 was undetectable and no differences were observed between genotypes and treatment for LRP-1 and LDLR (Fig. 7C). Tg cultures expressed significantly higher levels of RAGE compared to WT (200%) and TgMT5^−/− (194%) (Fig. 7C). IL-1β treatment did not modulate RAGE content in all genotypes.

Overall, there was no clear evidence of transcriptional regulations that could explain the downregulation of Aβ content upon MT5-MMP deficiency. The results also indicated that incubation of 10 ng/mL of IL-1β for 24 h did not impact the expression of genes encoding potential modulators of Aβ balance, and confirmed overall no influence of the cytokine in global APP/Aβ metabolism in our experimental settings.

An important finding of this work is that MT5-MMP content is higher in Tg cells, suggesting a modulating effect of AD transgenes on this proteinase. Such regulation highlights that the impact of MT5-MMP in AD, previously described in adult mice [13, 17], may actually begin at a stage well before the onset of obvious pathological and clinical signs. Of note, MT5-MMP deletion did not alter the expression of MT1-MMP, MMP-9 or MMP-2, all close MT5-MMP homologues, also involved in the control of APP metabolism and amyloidogenesis [33, 34, 54, 55], implying no compensatory regulation by these proteinases upon MT5-MMP deficiency. The content of β-III tubulin and GFAP, as well as the values of the MTT test, were similar across all experimental groups. This lack of cytotoxicity contrasts with a previous report showing cell demise in 5xFAD primary cortical neurons at 7 DIV [56]. In this study, no astrocytes were reported and cell density was fivefold lower than ours, which could explain a microenvironment that facilitates neuronal vulnerability. In addition, the apparent lack of toxicity mediated by IL-1β compared to previous studies [57, 58] may be related to different experimental settings used, including concentrations and time of exposure to the cytokine.

We previously found increased levels of IL-1β in the brain of 3-day-old 5xFAD pups [59] prior Aβ accumulation, and later at 2 months, along with the onset of Aβ accumulation [34]. Interestingly, MT5-MMP deletion resulted in decreased IL-1β levels in the brains of 5xFAD mice at prodromal stages of the pathology, indicating functional interactions between IL-1β and MT5-MMP [13]. Consistent with this idea, we show here that basal IL-1β levels are higher and those of Aß are stable in Tg cells compared to WT, arguing for regulation of inflammation in young cells by a non-Aβ related mechanism. Moreover, MT5-MMP deficiency reduces the neuroinflammatory response to IL-1β as well as the basal levels of IL-1β and MCP-1. The effect of genotype/IL-1β was cytokine-selective, as shown by the lack of effect on Tnfa, and a more complex behavior of IL-6, whose reduction in Tg cells after IL-1β was recovered in TgMT5^−/−. Downregulation of MCP-1 in MT5-MMP-deficient cells could dampen the system's ability to recruit microglia/macrophages to the site of inflammation, thereby helping to limit the progression of an exacerbated inflammatory response. Overall, these data echo a previous study showing that systemic injection of IL-1β did not trigger an inflammatory response in the PNS of adult MT5-MMP-deficient mice [10]. In that case, MT5-MMP-deficient prevented proper N-cadherin processing eventually disrupting the crosstalk between sensory neurons and mast cells [10]. Unchanged levels of N-cadherin or its proteolytic fragments in our MT5-MMP-deficient cells argue against this possibility. Alternatively, our data imply instead that cells lacking MT5-MMP could degrade IL-1β in a more efficient manner. This idea is indirectly supported by recent data showing that non-catalytic interactions of MT5-MMP promote C99 degradation by the proteasome and, to a lesser extent, by lysosomes [32]. Although IL-1β clearance is not well understood, it has been suggested that low levels of IL-1β stimulate autophagolysosomal function and attenuate inflammation in cell cultures, while higher cytokine concentrations (> 200 pg/mL) have the opposite effect [60, 61]. Whether MT5-MMP may act as an interactor/chaperon for IL-1β and/or the IL-1β/IL-1R1 complex, as it is the case for APP [13, 17, 32], needs further investigation.

To investigate the impact of MT5-MMP deficiency on basal synaptic activity, we measured spontaneous network synaptic events as a landmark for each genotype. In agreement with previous reports (see for review [62]), our Tg (5xFAD) neurons showed increased synaptic activity, as illustrated by higher amplitude and frequency, which is considered as a sign of hyperexcitability. At 11–14 DIV, with WT and Tg cells showing equal levels of Aβ (see Fig. 6), it is unlikely that the peptide influences the hyperexcitability observed in Tg neurons. The latter showed increased levels of MT5-MMP and IL-1β compared with WT, and more interestingly TgMT5^−/− neurons do not show hyperexcitability and show control values of IL-1β levels. Together, this raises the possibility of a coordinated action of MT5-MMP and IL-1β in promoting neuronal hyperexcitability. Although, IL-1β has diverse and sometimes divergent effects on neuronal activity depending on cell-type, cytokine concentration and duration of the stimulus [63], several studies highlight various mechanisms of IL-1β-mediated excitability: NMDA receptor stimulation of Ca²⁺ influx [64], inhibition of GABA-evoked currents [65] or prevention of the inhibitory effect of cannabinoid CB1 receptor on glutamate release [66]. In this context, it is possible that the onset of neuronal hyperexcitability observed in 5xFAD mice [67, 68] takes place during development, in which case, lower levels of IL-1β in MT5-MMP-deficient neurons could help to attenuate this process in the long run.

IL-1β elicited cell responses such as preventing an increase in amplitude in Tg cells, which could be interpreted as a homeostatic cellular response to the inflammatory burst. A possible post-synaptic mechanism of IL-1β underlying such effect could be the cytokine-mediated decrease in the content and phosphorylation of the AMPA-GluR1 subunit at the post-synaptic membrane [69]. IL-1β may also act as pre-synaptic neuromodulator [70, 71], which is consistent with the increased frequency we observed in all genotypes. However, the magnitude of the effect of MT5-MMP deficiency is surprising, given the weak inflammatory response of MT5^−/− cells to IL-1β (see Fig. 3), raising the possibility that MT5-MMP could differentially affect various IL-1β signaling pathways mediated [63] or not [72] by membrane receptors. Furthermore, as the increase in capacitance correlates with the increase in membrane surface area and pre-synaptic vesicle fusion [69], a parallelism could be established with the observation of increased gPSCs and capacitance in IL-1β-treated MT5^−/− neurons. However, no general conclusion can be drawn, as this correlation was not validated in the other experimental groups. Alternatively, the lack of MT5-MMP could prevent the formation of sAPP95/sAPPη, recently suggested to bind the GABA_BR1a and inhibit pre-synaptic neurotransmitter release [73]. Even if we found no sAPP around 85–95 kDa, we cannot exclude that a small functional pool of this WB-undetectable fragment reaches the synapse. Although further research is needed to better understand the novel effects reported here, MT5-MMP deficiency appears to prevent AD genotype-related synaptic dysfunction under basal conditions, while exacerbating IL-1β-induced neuronal excitability.

In contrast to previous observations in the brains of adult 5xFAD mice [13, 17, 34], C99 did not accumulate in developing neural cells. Yet, murine Aβ implicitly proved the formation at some point of its immediate precursor, C99. We assume that the latter is formed at a low pace and/or that it is promptly degraded by the proteasome or autophagolysosome [32], and even by α-secretase to yield C83 [32, 74]. Likely, all or some of these mechanisms are active in developing neural cells, thus preventing the neurotoxic effects of C99 accumulation [29, 45, 75]. In contrast to C99, DAPT did rescue C83 levels, which were not altered by MT5-MMP deficiency. This is in agreement with data showing stable levels of C83 in the frontal cortex of 5xFAD mice lacking MT5-MMP [17]. Unlike CTF modulation, MT5-MMP deficiency resulted in a decrease of Aβ levels, which could not be correlated with increases in Aβ-degrading enzymes. In fact, the observed reduction in neprilysin (Mme) is somewhat counterintuitive, unless it actually underlies a negative feedback response to a potential increase in neprilysin activity. Analysis of genes involved in Aβ transport/clearance (e.g., Lrp1, Lrp8, Ldlr, Ager) also revealed no clear evidence of a transcriptional mechanism that could explain the modulation of Aβ content.

The above data denote the capacity of MT5-MMP to control Aβ levels, echoing our pioneer work in vivo [13], and demonstrate that functional interactions of MT5-MMP with APP/Aβ may already happen in developing neural cells. However, the inability to detect C99 in these cells led us to question whether MT5-MMP might also control C99 as it does in adult mice [13, 17]. We addressed this question by providing evidence that the levels of transduced human C99 were downregulated in MT5-MMP-deficient primary neurons and, therefore, also those of Aβ. It is noteworthy that C83 derived from overexpressed C99 was also downregulated in MT5-MMP-deficient neurons, in contrast to the lack of effect on constitutive C83 likely resulting from APP processing. Thus, the control of C83 by MT5-MMP seems to depend mostly on whether it is generated from APP or C99. Taken together, these data indicate that young neurons have the potential to prevent the accumulation of endogenous C99 and thus prevent the derived detrimental consequences. This work also highlights that MT5-MMP deficiency facilitates C99 degradation in these neurons, which supports our recent results in HEK cells showing that deletion of C-terminal domains of MT5-MMP does indeed lead to C99 degradation [32].

Additional file 1. Immunoblots representing subcellular distribution of C83 detected with the APP-CTF antibody in primary cortical cultures at 11 DIV. Fractions are represented with their loading controls: for fraction 1, cytosolic—GAPDH; for fraction 2, membranous—Na + /K + ATPase, and for fraction 3, nuclear—Histone 3. Cells were treated or not with DAPT (10 μM). AAV-C99 (right) indicates a positive control. WT cells were infected for 5 days with AAV-C99 and recovered at 11 DIV. Note that only C83 levels were detectable with DAPT treatment.Additional file 2. A Measurement of human A levels (pg/mL) in Tg and TgMT5-/- cultures using the ThermoFisher Scientific ELISA kit. B and C. mRNA levels of hAPP and hPSEN1 analyzed by RT-qPCR in Tg and TgMT5-/- cultures and normalized with Gapdh as housekeeping gene. Black bars represent control (untreated) conditions and grey bars IL-1β treated conditions (10 ng/mL for 24 h). Values are the mean ± SEM of 3–5 independent cultures by genotype.