Elimination of the four extracellular matrix molecules tenascin-C, tenascin-R, brevican and neurocan alters the ratio of excitatory and inhibitory synapses

The synaptic transmission in the mammalian brain is not limited to the interplay between the pre- and the postsynapse of neurons, but involves also astrocytes as well as extracellular matrix (ECM) molecules. Glycoproteins, proteoglycans and hyaluronic acid of the ECM pervade the pericellular environment and condense to special superstructures termed perineuronal nets (PNN) that surround a subpopulation of CNS neurons. The present study focuses on the analysis of PNNs in a quadruple knockout mouse deficient for the ECM molecules tenascin-C (TnC), tenascin-R (TnR), neurocan and brevican. Here, we analysed the proportion of excitatory and inhibitory synapses and performed electrophysiological recordings of the spontaneous neuronal network activity of hippocampal neurons in vitro. While we found an increase in the number of excitatory synaptic molecules in the quadruple knockout cultures, the number of inhibitory synaptic molecules was significantly reduced. This observation was complemented with an enhancement of the neuronal network activity level. The in vivo analysis of PNNs in the hippocampus of the quadruple knockout mouse revealed a reduction of PNN size and complexity in the CA2 region. In addition, a microarray analysis of the postnatal day (P) 21 hippocampus was performed unravelling an altered gene expression in the quadruple knockout hippocampus.

. Expression of the synaptic puncta vGlut1 and PSD95 in hippocampal neurons after 14 and 21 DIV. (a,a',a",a"',b,b',b",b"',c,c',c",c"',d,d',d",d"') Representative images of the immunocytochemical stainings of the presynaptic protein vGlut1 (puncta appearing in magenta) and the postsynaptic protein PSD95 (green puncta) after 14 DIV and 21 DIV for the four conditions (N wt/wt /A wt/wt , N wt/wt /A ko/ko , N ko/ko /A wt/wt , N ko/ko /A ko/ ko ) are shown. The white puncta reflect the colocalization of both proteins and indicate structural synapses. The PNNs were visualized using aggrecan (blue), which is visible in (a,a",b,b",c,c",d,d") for PNN-bearing neurons. In (a',a"',b',b"',c',c"',d',d"'), neurons devoid of PNNs are exemplified. Next to each of the micrographs is a higher magnification of one exemplary neurite (white box in each image) showing the individual synaptic puncta in more detail. Scale bar: 50 µm in D"' . (e,f,g,h) Analysis of the relative increase or decrease of the number of synaptic puncta in percent after 14 (e,f) and 21 DIV (g,h). PNN-wearing knockout neurons after 14 DIV (e) combined with wildtype astrocytes (N ko/ko /A wt/wt ) and knockout astrocytes (N ko/ko /A ko/ko ) showed a significant rise of excitatory synapses. The same applies for PNN-free neurons after 14 DIV (f). PNN-wearing alteration of synaptic puncta was visible in PNN-positive wildtype neurons co-cultivated with knockout astrocytes (Supplementary Table S2a,b).
Summarizing these observations, the elimination of the four ECM molecules TnC, TnR, Neurocan and brevican led to a reduction of inhibitory synapses on PNN-positive as well as PNN-negative knockout neurons cultivated for 14 DIV. This effect levelled off after three weeks where wildtype neurons maintained in the presence of knockout astrocytes displayed an enhanced number of vGAT-positive afferents and inhibitory synapses, irrespective of the presence or absence of PNNs (Fig. 2g,h).
Higher neuronal network activity in the quadruple knockout mouse. To test the influence of the altered proportion of excitatory and inhibitory synapses at hippocampal quadruple knockout neurons on the spontaneous activity, MEA analysis was conducted as established in our laboratory 32 . Via the 60 electrodes of the MEA (Fig. 3a,b), the spontaneous activity of the neuronal network developing in vitro was analysed (Fig. 3c,d,e). Each electrode was individually recorded and the traces (Fig. 3g) were represented in small boxes (Fig. 3f,g), reflecting the spontaneous activity as well as the occurrence of bursts in different areas of the network in culture (Fig. 3c-e).
The number of spikes (Fig. 4a) which correspond to action potentials was significantly increased in knockout neurons co-cultured with knockout astrocytes (N ko/ko /A ko/ko ) after 14 DIV (4306.84 ± 217.87 (p < 0.001)) and 21 DIV (8762. 16  Bursts consist of single spikes and occur simultaneously at a variety of electrodes in a short time interval. When knockout neurons were combined with wildtype astrocytes (N ko/ko /A wt/wt ), the spike number was nearly the same as for the control, indicating a contribution of the astrocytic phenotype into the process of network activity. Similarly, the spike frequency was upregulated in cultures with knockout neurons co-cultivated with knockout astrocytes at both examined time points (14 DIV: 7.15 ± 0.33, p < 0.001; 21 DIV: 14.60 ± 0.70, p < 0.001) compared to the control (14 DIV: 3.91 ± 0.24; 21 DIV: 7.28 ± 0.43). With progressing cultivation time, the neuronal networks knockout neurons (g) co-cultivated with knockout astrocytes (N ko/ko /A ko/ko ) showed a significant enhancement of structural synapses in culture compared to the control (N wt/wt /A wt/wt ) after 21 DIV. No significant differences were revealed after 21 DIV for the number of synaptic puncta of PNN-negative knockout neurons (h) compared to the control condition. Statistics: Five independent experiments (biological replicates N = 5) were performed choosing randomly 20 neurons with (n = 20) and 20 neurons without PNNs (n = 20) per each condition. In sum, 800 neurons were analysed. Data are expressed as mean ± SEM (ANOVA and Scheffé post hoc test for PSD-95 data sets and Kruskal-Wallis test for vGlut1 and Colocalization data sets, p ≤ 0.05). (2019) 9:13939 | https://doi.org/10.1038/s41598-019-50404-9 www.nature.com/scientificreports www.nature.com/scientificreports/ (a,a',a",a"',b,b',b",b"',c,c',c",c"',d,d',d",d"') Representative images show the immunocytochemical stainings of hippocampal neurons after 14 and 21 DIV with the presynaptic marker vGAT (magenta puncta) and the postsynaptic marker gephyrin (green puncta) to detect inhibitory synapses in all four conditions (N wt/wt /A wt/ wt , N wt/wt /A ko/ko , N ko/ko /A wt/wt , N ko/ko /A ko/ko ). Colocalization of these proteins leads to the appearance of white puncta, which we define as structural inhibitory synapses in the neuronal cultures. To distinguish between neurons with (a,a",b,b",c,c",d,d") and without PNNs (a',a"',b',b"',c',c"',d',d"'), aggrecan (blue) was used to visualize the PNNs. The higher magnification next to each image (white box in each image) displays one representative neurite to allow for a more detailed view on the distribution of the synaptic puncta. Scale bar: 50 µm in d"' . (e,f,g,h) The total number of the vGAT, gephyrin and colocalized puncta was quantified and the percentage of the increase/decrease of the number of synaptic puncta was thus determined after 14 (e,f) and 21 DIV (g,h). Irrespective of whether the neurons were covered by PNN (e) or not (f), the knockout neurons gained complexity, more bursts arose and a higher spike frequency in bursts emerged. Knockout neurons grown in the presence of knockout astrocytes (N ko/ko /A ko/ko ) showed a significantly higher spike frequency in bursts of 92.18 ± 4.23 Hz (p < 0.001), which occurred after 14 DIV compared to 66.61 ± 2.44 Hz of the control (N wt/wt /A wt/ wt ). When cultivated for 21 DIV, these differences between the two conditions were still present (N wt/wt /A wt/wt : 86.17 ± 2.97, N ko/ko /A ko/ko : 94.32 ± 2.95 (p = 0.031). Interestingly, knockout neurons co-cultured with wildtype astrocytes (N ko/ko /A wt/wt ) showed an opposite effect concerning their spike frequency in bursts, namely a reduction with a value of 48.54 ± 1.38 Hz (p < 0.001) after 14 DIV. After 21 DIV no significant difference could be observed (75.15 ± 1.91 Hz (p = 0.347)).
A higher amount of spikes in bursts appeared in the (N ko/ko /A ko/ko ) condition compared to the control condition. Whereas in the control condition only 43.79 ± 2.18% spikes after 14 DIV and 50.37 ± 2.02% spikes after 21 DIV appeared within a burst, 60.62 ± 1,85% (p < 0.001) spikes in bursts after 14 DIV and 71.50 ± 1,90% (p < 0.001) spikes in bursts after 21 DIV occurred in the (N ko/ko /A ko/ko ) condition. There was again a significant decrease of this analysed parameter in cultures of knockout neurons cultured with wildtype astrocytes (N ko/ko / A wt/wt ) after 14 DIV with a percentage of 23.96 ± 1.31% (p < 0.001) spikes in burst, which attained control levels after 21 DIV, with a percentage of 50.78 ± 1.93% (p = 1.000). The duration of bursts was also measured using MEA analysis (Fig. 4f, Supplementary Table S3a,b). In the control condition (N wt/wt /A wt/wt ) an average burst lasted for 341.88 ± 19.57 ms after 14 DIV and for 284.94 ± 12.83 ms after 21 DIV. No change in the mean duration of a burst was seen for knockout neurons grown with knockout astrocytes (N ko/ko /A ko/ko ) after 14 DIV, which showed similar results of 294.96 ± 8.84 ms (p = 0.057) after 14 DIV and 341.72 ± 10.66 ms (p < 0.001) after 21 DIV, which also applies for all other conditions compared to the control condition. After 21 DIV, all conditions showed a significant rise in the burst duration in comparison to the control. The only striking difference of the burst duration was observable after 14 DIV for knockout neurons cultured with wildtype astrocytes (N ko/ko /A wt/wt ). Here, the burst duration was significantly protracted to 456.95 ± 26.44 ms (p = 0.032) instead of 341.88 ± 19.57 ms as detected in the control.
In conclusion, the MEA analysis revealed that the deletion of four critical ECM constituents resulted in an increased network activity in the cultures of hippocampal neurons.
Reduction of pnns in vivo in the quadruple knockout mouse. The ECM molecules TnC, TnR, brevican and neurocan are constituents of PNNs, therefore it appeared of interest to analyze the impact of the elimination of these genes on the formation of PNNs in vivo. The marker WFA was used to visualize PNNs in frontal brain slices of the hippocampus of wildtype (Fig. 5a,c,e,g,i) and knockout mice (Fig. 5b,d,f,h,j).
The analysis of the PNNs in vivo was focused on the hippocampus of 15, 20, 25, 30 and 35 days old mice. Within the CA2 region of the hippocampus, an area consisting of numerous PNNs developed starting at P15 and evolved further along with the maturating hippocampus. This PNN territory was present in the wildtype as well as in the knockout hippocampus. A small proportion of the fluorescence in the CA2 region could be due to the diffuse interstitial matrix. However, it has recently been demonstrated that pyramidal neurons in the CA2 region are a major source of perineuronal nets in the hippocampus and positive for WFA and aggrecan 33 . This agrees with recent studies that emphasize that the WFA signal in CA2 is PNN-specific 37,38 . Consistent with this interpretation, a higher magnification image obtained using confocal laser scanning microscopy revealed that the major part of the fluorescence was concentrated in PNNs around neuronal cell bodies, suggesting a specific labeling of these structures (Fig. 5k-n). The PNN areas were determined for the different postnatal stages of the developing hippocampus (P15-P35) by measuring the WFA-positive area (Fig. 6). The quantification revealed a significant downregulation of the PNN area in the quadruple knockout hippocampus at P15, P20, P25, and P30 (Fig. 6e). The PNN area amounted to 42558.46 ± 3223.24 µm 2 (p < 0.001) at P15, 62810.40 ± 5409.42 µm 2 (p = 0.05) at P20, 63750.31 ± 3300.47 µm 2 (p < 0.001) at P25 and 63620.73 ± 5082.00 µm 2 (p < 0.001) at P30 compared with the values of the wildtype of 64155.52 ± 3273.71 µm 2 at P15, 78005.23 ± 5021.09 µm 2 at P20, 90590.67 ± 6055.96 µm 2 at P25 and 104404.09 ± 7588.34 µm 2 at P30 (Supplementary Table S4a,b). In the wildtype hippocampus, the area size increased over time until P30. No significant difference of the PNN size was detectable between the wildtype and knockout hippocampus at P35. At this stage the PNN area of the knockout hippocampus amounted to 73468.03 ± 8223.63 µm 2 and of the area size of the wildtype hippocampus encompassed 90108.04 ± 6145.99 µm 2 (p = 0.120), indicating an alignment of their PNN area size towards adulthood.
Next to the area size, the corrected total cell fluorescence (CTCF) of the WFA-positive area in the CA2 region of the hippocampus was measured (Fig. 6c,c' ,d,d' ,f ). As previously seen for the area size, the fluorescence intensity was also significantly diminished in the quadruple knockout hippocampus at P15, P20, P25 and P30 (P15: 1963962. 39  showed a significantly reduced number of synaptic puncta. PNN-wearing knockout neurons co-cultivated with knockout astrocytes for 21 DIV (N ko/ko /A ko/ko ) exhibited a significant lower number of gephyrin puncta in comparison to the control (N wt/wt /A wt/wt ) (g), whereas the number of structural synapses was not significantly altered. Knockout neurons combined with knockout astrocytes and without PNNs showed no significant change in the number of their synaptic puncta (h). Statistics: Five independent experiments (N = 5) were performed choosing randomly 20 neurons with (n = 20) and 20 neurons without PNNs (n = 20) per each condition. In sum, 800 neurons were analysed. Data are expressed as mean ± SEM (Kruskal-Wallis test, p ≤ 0.05). www.nature.com/scientificreports www.nature.com/scientificreports/ (5529794.88 ± 430019.63, p = 0.108). In conclusion, the maturation of PNNs is significantly delayed in the quadruple knockout hippocampus but normalized after postnatal day 30 to attain the wildtype control levels in adulthood (Supplementary Table S4a,b).

Alterations in the gene expression in the quadruple knockout hippocampus at postnatal day 21.
Considering the fact that the PNN maturation was significantly retarded in the P21 hippocampus of the quadruple knockout mouse, this time point was chosen for the comparative analysis of gene expression using microarrays.
Overall 34473 genes were examined of which 438 were significantly altered in the quadruple knockout hippocampus. Because we had detected differences in synapse formation, we were particularly interested in genes associated with the ECM or synaptogenesis. A volcano plot and a clustering heatmap highlight the dysregulated genes of interest of these functional annotations, which are either up-or downregulated (Fig. 7a,b)    Using MEA analysis, the spontaneous activity of neuronal networks derived from the hippocampus of either wildtype or quadruple knockout mice was examined. Different parameters were detected and quantified for the four conditions (N wt/wt / A wt/wt , N wt/wt /A ko/ko , N ko/ko /A wt/wt , N ko/ko /A ko/ko ) including the number of spikes (a), the number of bursts (b), the spike frequency (c), the spike frequency in bursts (d), the percentage of spikes in bursts (e) and the mean burst duration (f). Most analysed parameters were enhanced in the neuronal networks of knockout neurons cultured with knockout astrocytes (N ko/ko /A ko/ko ) compared to the control (N wt/wt /A wt/wt ). For example the number of spikes increased significantly to almost twice the control level after 14 DIV as well as 21 DIV and the number of bursts in culture was significantly enhanced too. This also resulted in a higher spike frequency as well as in a higher percentage of spikes in bursts in the knockout neurons grown with knockout astrocytes. The only parameter that remained unchanged between these two conditions was the mean burst duration. Statistics www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ collagens Col1a2 (−1.35-fold, p = 0.0017), Col4a3 (−1.16-fold, p = 0.0014) and Col27a1 (−1.21-fold, p = 0.0098) were significantly downregulated, as well as Spon2 (−1.33-fold, p = 0.0002) (Fig. 7c). Spon2 is known to facilitate neurite outgrowth of hippocampal neurons 39,40 . In contrast, the ECM genes Gpc3 (1.40-fold, p = 0.0089) and Calr (1.16-fold, p = 0.0038) were significantly upregulated in the quadruple knockout hippocampus. Interestingly, genes related to neuronal development and function (Fig. 7d) (Fig. 7e). Chondroitin sulfate proteoglycans such as Cspg4, Cspg5 or Ptprz1 that could potentially compensate the knockout were also analysed, but no significant difference could be observed in their expression levels (Fig. 7b). A tabular overview of the data can be found in the supplement (Supplementary Table S5).

Discussion
Although it is known for a few years that PNNs contribute to the process of synaptogenesis, it is not yet clear to which extent the different molecules are involved therein 9,10,41 .
Because of the manifold and complex character of the PNN structure, it is not surprising that neurons with PNN, as well as quadruple knockout neurons with a defective PNN composition display considerable changes in their synapse formation. Several studies of the recent years have shown that one of the main features of the ECM and PNNs is to restrict the formation of neurites 42 and their synapses with concurrent protection of neurons towards harmful substances or events 43 . These studies relied on the application of the bacterial derived enzyme chondroitinase ABC (ChABC), which degrades the chondroitin-4-sulfate, chondroitin-6-sulfate, dermatan sulfate and the hyaluronan glycosaminoglycan chains 10,44 . When hippocampal neurons were treated with ChABC in vitro, synapse formation was boosted 6 . Different from the ChABC-treated cultures, the quadruple knockout neurons miss the CSPGs neurocan and brevican, but still express aggrecan, the PNN marker used in this study, and possibly other CSPGs such as phosphacan 45 . These may be expressed similar to the wildtype control because the gene array analysis did not reveal substantial modifications of CSPG genes.
TnR is a pivotal constituent of PNNs and its elimination leads to a change of PNN distribution, composition and function 46 . A recent study concerning hippocampal neurons from TnR-deficient mice reported a modified PNN organization around dendrites that could be rescued by adding TnR and aggrecan into the culture medium 21 . With regard to physiological parameters, TnR-knockout mice displayed an enhancement of excitatory synaptic transmission in the hippocampal CA1 region 47 . This is in agreement with our observation that an increased number of excitatory synapses resulted in enhanced synaptic transmission in the neuronal network of the quadruple knockout mouse, which was recorded via MEA analysis in vitro.
It is conceivable that the missing expression of TnC and TnR is compensated by other glycoproteins. For example, the tenascin family comprises the four genes TnC, TnR, TnW and TnX, but neither TnW nor TnX are elevated in the quadruple knockout mouse 48,49 . In contrast, the two glycoproteins fibulin-1 and -2 of the ECM are clearly upregulated on the protein level in the quadruple knockout brains. This is accompanied by an altered hyaluronan deposition in the mutant ECM 28 . Interestingly, fibulin-1 mutations are associated with cognitive deficits in the human and fibulin-2 mediates TGF-beta1 effects on adult neural stem cells 50,51 .
A former study already reported the synaptogenesis of hippocampal quadruple knockout neurons in an in vitro co-culture system. The loss of TnC, TnR, brevican and neurocan resulted in a significant increase of synaptic density within 14 DIV that transited to a decremented synapse number after 21 DIV. Neurons coated with PNNs were characterized by an attenuated synapse number after both 14 and 21 DIV 29 .
Unexpectedly, in our approach WT astrocytes were not capable to rescues the PNN formation by knockout neurons. Astrocytes in vitro secrete brevican, neurocan and TnC. However, TnC is not a major component of perineuronal nets 52 while TnR is mainly expressed by oligodendrocytes and not by astrocytes 10 (and is only present at low concentration in the co-culture of WT astrocytes with KO neurons, if at all. Lecticans do contain binding sites for hyaluronic acid and TnR 53 . Several reports suggest that CSPGs like aggrecan or brevican are incorporated into perineuronal nets through their interaction with TnR and hyaluronic acid 14,53,54 . However, TnR is absent from the knockout neurons in co-culture. This may explain why in our model the incorporation of astrocytic CSPGs into PNNs was impaired. This agrees with a study that emphasized the functional importance of TnR for the formation of PNNs 21 . For the formation of structurally and functionally intact synapses, specific molecules are necessary. In former studies, it was demonstrated that the astrocyte-secreted factors thrombospondin and hevin lead to the development of structural, but silent synapses 55,56 . In line with this, Gpc4 and Gpc6 were recently also discovered as CA2 region of knockout mice. At P35, the PNN areas size was not significantly different between knockout and wildtype. (f) Quantitative analysis of the PNN fluorescence intensity (CTCF) in the CA2 region of the hippocampus of mice at P15, P20, P25, P30 and P35. The fluorescence intensity was significantly reduced within the hippocampus of knockout compared to wildtype mice. The only exception was at P35, when no significant alteration of the PNN fluorescence intensity was detectable. Statistics: Three independent experiments (biological replicates N = 3) were performed and the PNN area size of four hippocampi (n = 4) was examined. Data are expressed as mean ± SEM (F-test and unpaired student's t-test, p ≤ 0.05). (2019) 9:13939 | https://doi.org/10.1038/s41598-019-50404-9 www.nature.com/scientificreports www.nature.com/scientificreports/ astrocytic factors, which stimulate the excitatory synapse formation by regulating glutamate receptor clustering 57 . The analysis of the gene expression performed in this study revealed an upregulated expression of Gpc3 in the hippocampus of P21 quadruple knockout mice, which is probably an explanation for the higher number of excitatory synapse number and the elevated activity in the quadruple knockout cultures.
A recently published study has demonstrated an altered synaptic plasticity in the dorsal dentate gyrus of the quadruple knockout hippocampus that translated into an enhanced short-term depression in conjunction with These are depicted in form of a volcano plot, which includes upregulated genes (red dots), downregulated genes (blue dots) and unaffected genes (grey). (b) The hierarchical clustering of genes in the P21 hippocampus uncovered a differential gene expression pattern in the quadruple knockout compared to the wildtype mouse using three dependent hippocampus samples of siblings (n = 3). The cluster heat map comprises significantly up-and downregulated genes of interest concerning the annotations "ECM" as well as "neurons/synapses" including Gpc3, Gabrq, Gad2, Wnt7, Syt9, Sod3, Dnm3, Cript, Sema4c, Calr, Cplx3, Col4a3, Col27a1, Adamts13, Grin2d, Col1a2, Spon2 and Rapsn. The colour shift of light blue to light red indicates the expression change of these genes, that is genes that appear in blue were down-whereas genes illustrated in red were upregulated in the hippocampus of the quadruple knockout mouse. www.nature.com/scientificreports www.nature.com/scientificreports/ a modified frequency-dependence in vivo 30 . This changed synaptic plasticity in vivo is also detected in in vitro neuronal networks performed in this study. Here, the spontaneous network activity was significantly increased in quadruple knockout cultures. These data are in agreement with the report of Dityatev and colleagues who found a higher excitability of hippocampal neurons in culture after ChABC treatment 14 . Furthermore, hippocampal neuronal cultures treated with hyaluronidase developed an epileptiform activity due to the occurrence of superbursts 58 . Thus, it is evident that the loss of pivotal ECM molecules leads to a higher neuronal activity independent of whether the ECM molecules are digested by appropriate enzymes or genetically eliminated.
The initiation of PNN formation in the murine hippocampus detectable by WFA occurred between postnatal stage P10 and P15. In the adult hippocampus, the PNN formation resembled the expression in the wildtype hippocampus, indicating a progressive compensation of the four missing ECM molecules TnC, TnR, brevican and neurocan, possibly by the upregulation of other ECM molecules. The results found for the in vivo PNN formation fit perfectly with the in vitro findings 29 , both demonstrating a diminished PNN formation in the postnatal quadruple knockout mouse. A recent study demonstrated that especially the CA2 excitatory pyramidal neurons are surrounded by PNNs. When mice were kept in an enriched environment, the PNN expression was enhanced, indicating a correlation between PNN development and environmental stimuli within the juvenile brain 33 . Interestingly, PNNs have been proposed as indispensable mediators of the modulation of synaptic plasticity by the transcription factor Otx2 59-61 . An impaired PNN formation and structure accompanied with altered synaptic plasticity is associated with a couple of psychiatric and neurological diseases such as schizophrenia and epilepsy, as reviewed recently 62 .
Significant dysregulation of genes of interest of the ECM or genes associated with neuron biology were detected by a comparative transcriptome analysis of the P21 quadruple knockout compared to the wildtype hippocampus using microarrays.
Like the previously discussed Gpc3 also other molecules contribute to synapse formation and synaptic transmission. For example, in the quadruple knockout mouse Sema4c is upregulated, which is known as interaction partner of different PSD95 isoforms in neocortical cultures as well as in the adult brain of mice 63 . Cplx3 that was significantly downregulated in the quadruple knockout hippocampus regulates the synaptic neurotransmitter release of both excitatory and inhibitory synapses 64 . The molecule Grin2d, also known as GluN2D, was reduced in the quadruple knockout hippocampus. By using Grin2d-deficient mice, its contribution in synaptic transmission of hippocampal interneurons of young mice as well as of the pyramidal CA1 neurons of newborn pups has been proven 65 . Furthermore, Grin2d has been found in hippocampal interneurons and enhances their activity due to its involvement in synaptic transmission 66 .
Thus, available data concur with our observation that a modified number of excitatory and inhibitory synapses prevail in quadruple knockout neurons in vitro. The excitation versus inhibition balance is considered very important in the context of neuropsychiatric disease 67 . A shift of the balance can result from neurodevelopmental abnormalities targeting the inhibitory interneurons 68 or from structural changes at the synapse 69 . Our observations suggest that the ECM intervenes in the balance of excitation and inhibition in neural networks. A recent review elegantly summarizes potential mechanisms of action of PNNs 70 . Brevican for example, which is missing in our mouse model controls the location of AMPA and potassium channels and might therefore affect synapse dynamics 71 . Neurocan is known to influence the binding of neural cell adhesion molecule (NCAM) and ephrin type-A receptor 3 (EPHA3) at perisomatic synapses 72 . TnR, another protein missing in the quadruple knockout mouse is known to be associated with HNK-1 and necessary for the maintenance of the excitatory and inhibitory balance in CA1 region 47 . Beyond these PNN-associated compounds also proteins released by astrocytes influence synaptogenesis, for example thrombospondins and the gabapentin receptor 55,73 , Hevin and SPARC 56 , or Pentraxin 3 that promotes excitatory synapse formation by fostering the clustering of AMPA glutamate receptors 74 . In the light of these observations, the interactions of ECM molecules with neuronal receptors may represent a promising theme for analyzing this issue in future studies. cell culture. Co-cultivation of astrocytes and neurons. Hippocampal neurons were obtained from embryonic mice and co-cultured with astrocytes in an indirect contact as described previously 31 (See supplemental data S6).
Electrophysiological recordings. Recordings of the spontaneous activity were performed for 10 min with a sampling frequency of 20 kHz. The data of the individual electrodes were collected using the software MC_Rack (version 3.9.0 by Multi Channel Systems MCS GmbH). The field potentials of raw data were corrected with a high pass filter which had a frequency of 200 Hz. The spontaneous activity was detected with a spike detector only if the value exceeded a threshold value 4.5-fold higher than the standard deviation. The following parameters were (2019) 9:13939 | https://doi.org/10.1038/s41598-019-50404-9 www.nature.com/scientificreports www.nature.com/scientificreports/ set for the measurements: Maximal interval to start burst, 10 ms; Maximal interval to end burst, 100 ms; minimal interval between bursts, 210 ms; minimal duration of burst, 50 ms; minimal number of spikes in bursts, 5. immunocytochemistry. Staining  Staining of inhibitory synapses. The staining procedure of inhibitory synapses was similar to the previously described protocol by Dobie and colleagues 36  With the help of a peristaltic pump (Cole-Parmer GmbH), the blood was exchanged with a 0.9% (w/v) NaCl solution (Carl Roth GmbH & Co. KG) with 0.2% (v/v) of 25 000 I.E./5 ml heparin-sodium (Ratiopharm GmbH). Thereafter, 4% w/v PFA was applied for 10-15 min for fixation of the brain. Decapitation as well as the quick removal of the brain from the skull followed. The brain was transferred into 4% w/v PFA and incubated overnight at 4 °C. On the next day, the brain was placed into 30% w/v sucrose (Thermo Fisher Scientific Inc.). After approximately two days, the brain was embedded with tissue tec freezing medium (Leica Microsystems GmbH; Cat. No.: 14020108926) on dry ice and stored at −20 °C. Afterwards, the frozen brains were cut into 14 µm thick frontal sections (interaural: 1.86 mm, bregma: −1.94 mm) using a cryostat CM 3050 S (Leica Microsystems GmbH).
Staining of cryosections. First, the cryosections were incubated in blocking solution for 1 h at RT in a wet chamber, followed by the incubation with primary antibodies. The primary antibodies were diluted in PBT1 (PBS, 0.1% w/v BSA, 0.1% v/v Triton X-100) containing 5% v/v goat serum (Dianova) as follows: neurofilament 200 (NF200), polyclonal, rabbit Affymetrix GeneChip ® analysis. Using Affymetrix GeneChip ® Mouse Gene 2.0 ST Arrays, a whole-transcript expression analysis of >30,000 RefSeq transcripts of P21 hippocampus of wildtype and quadruple knockout mice siblings was performed in triplicates. RNA quantity was assessed using the NanoDrop 1000 (Thermo Scientific Nano Drop Technologies), and RNA quality control was carried out with the RNA 6000 Nano Assay (Agilent 2100 Bioanalyser) to ensure that the samples had an RNA integrity number (RIN) of at least 9. All further processing of total RNA was performed according to the Ambion whole-transcript Expression kit and the Affymetrix GeneChip whole-transcript terminal labeling and control kit manuals. The cDNA fragment size was checked using the 2100 Bioanalyser (fragment size between 50 and 200 bp). www.nature.com/scientificreports www.nature.com/scientificreports/ Fluidics Station 450 (program: FS450 0002) and scanned on a GeneChip ® Scanner 3000 7 G (both Affymetrix). Raw image data were analysed with GeneChip ® Command Console ® Software (AGCC). Gene expression intensities were normalized (Robust Multichip Average 76 ) and summarized using Affymetrix ® Expression Console ™ Software. Transcripts that were differently expressed more than 1.5-fold with an Anova p-value less than 0.05 between the analysed samples were addressed as regulated.
The heat-Map was established using AltAnalyse 77 version 2.1.0 with following parameters: cosine hopach for column clustering and cosine weighted for row clustering. The rows were normalized relative to the row median. The volcano plot was constructed using Graph Pad Prism 6. experimental design and statistical analysis. In general, the data are expressed as mean ± SEM. The statistical significance is given by the p value: p ≤ 0.05 = *, p ≤ 0.01 = ** and p ≤ 0.001 = ***. The program IBM SPSS Statistic (Version 20) was used for statistical evaluation. The exact number of experimental repetitions can be found in the figure legends.
Microscopy. The confocal laser-scanning microscope LSM 510 meta operated with the ZEN2009.Ink software (Carl Zeiss Microscopy GmbH) was used for image acquisition of the immunocytochemical stainings. Different layers were recorded with an interval of 0.25 µm using 630-fold magnification. For generating two-dimensional images, pictures were overlaid. For quantification of synaptic puncta, the program ImageJ was used with the Plugin "Puncta Analyser" from Barry Wark; licensed under http://www.gnu.org/copyleft/gpl.html) and the following settings were adjusted: rolling ball radius = 50 pixel, size (pixel 2 ) = 2-infinity and circularity = 0.00-1.00.

PNNs in vivo.
The immunological staining of the cryosections was recorded using the microscope Axio Zoom. V16 (Carl Zeiss Microscopy GmbH). A prominent PNN-positive area was found in the CA2 region of the hippocampus 33 . To quantify the WFA-intensity and -area, the PNNs in the micrographs were manually bordered, the corrected total cell fluorescence (CTCF) was determined as previously described 78 and the area was measured using ImageJ.