Perineuronal nets are phagocytosed by MMP‐9 expressing microglia and astrocytes in the SOD1G93A ALS mouse model

Perineuronal nets (PNNs) are an extracellular matrix structure that encases excitable neurons. PNNs play a role in neuroprotection against oxidative stress. Oxidative stress within motor neurons can trigger neuronal death, which has been implicated in amyotrophic lateral sclerosis (ALS). We investigated the spatio‐temporal timeline of PNN breakdown and the contributing cellular factors in the SOD1G93A strain, a fast‐onset ALS mouse model.


INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterised by motor deficit symptoms, progressive muscle weakness, and ultimately respiratory failure and death usually occurring within 4 years of initial diagnosis [1,2].Motor neuron death is a pathological hallmark of ALS [3], which is indicated by the loss of large alpha-motor neurons (α-MNs) in the ventral horn of the spinal cord [4,5].Mutations in the gene encoding Cu, Znsuperoxide dismutase-1 (SOD1) are responsible for $20% of familial ALS (fALS), and 5% of sporadic ALS (sALS), which has no monogenic inherited component [6][7][8].The SOD1 protein is a free radical scavenger, and its mutation results in the accumulation of reactive oxygen species that cause oxidative damage to cellular components [9,10].The SOD1 mouse model with a missense mutation substituting glycine for alanine at position 93 (G93A) recapitulates many aspects of human ALS pathology and has therefore been used extensively to study fALS [11,12].While the exact mechanism of ALS aetiology remains unknown, the loss of perineuronal nets (PNNs) has been implicated in disease progression [13,14].
PNNs are extracellular matrix structures that encase excitable neurons and their proximal dendrites.They are well known to play a role in neuroplasticity [15] and in neuroprotection against oxidative stress [16,17].PNNs are composed of a hyaluronan backbone that serves as a scaffold for the attachment of chondroitin sulfate proteoglycans such as aggrecan via tenascin and other link proteins [18][19][20].
These neuroprotective nets are decreased in the ventral horn regions of both SOD1 G93A mice at end-stage disease [14] and in mutant symptomatic SOD1 rats [13].PNN loss specifically around α-MNs has not been investigated, and the timing of PNN breakdown relative to motor deficit symptoms and what may cause the loss of PNN in ALS is unknown.
Of these, MMP-9 has been shown to degrade PNNs by cleaving aggrecan, a proteoglycan that is a common core structural component of many PNNs [23][24][25][26].SOD1 G93A mice at the post-symptomatic stage have increased levels of MMP-9 in the spinal cord [27].The cellular source of the elevated MMP-9 is unknown in these ALS model mice; however, MMP-9 in the ventral horn has shown to be localised to vulnerable α-MNs that are lost by end-stage disease in SOD1 G93A   and WT mice but absent in α-MNs that are spared in ALS.MMP-9 has also been shown to be expressed in microglia of end-stage SOD1 G93A mice; however, their level of expression in relation to PNN breakdown and α-MN loss is unknown [28].Genetic ablation of MMP-9 is protective in the SOD1 G93A mouse as well as other ALS model mice (e.g., TDP-43 mutant mice), with researchers observing delayed muscle denervation, prolonged α-MN survival, and increased overall lifespan following MMP-9 reduction [28,29].α-MNs of SOD1 G93A mice may be overexpressing MMP-9, thus acting as the source of the increased MMP-9 observed in the spinal cord [27,[30][31][32].However, it remains unclear whether PNNs breakdown specifically around α-MNs and if MMP-9 expression is increased in the α-MNs of ALS mice.Because PNNs and MMP-9 have not been studied together in an ALS mouse model, we characterised the spatio-temporal expression of PNNs and MMP-9 in the ventral horn of SOD1 G93A mice throughout disease progression, including the time course of PNN breakdown and its phagocytosis by surrounding activated glia.

Ethics statement
The University of Queensland Animal Ethics Committee approved all experimental procedures (AE000535; AE000114).All experimental procedures complied with the regulations and policies regarding ani-

Key Points
• Perineuronal nets (PNNs) break down around α-motor neurons at disease onset in ALS model mice.
• MMP-9, a known PNN protease in microglia and astrocytes, is upregulated at the time of PNN breakdown.
• Microglia and astrocytes act to phagocytose PNNs during disease progression in ALS model mice.
• Upon PNN breakdown around α-motor neurons, these neurons show increased expression of 3-nitrotyrosine, a marker for protein oxidation, which could render them vulnerable to death.

Immunofluorescence
Immunostaining was performed on 3-6 serial, L5 segment lumbar spinal cord sections at intervals of five μm (i.e., 75 μm apart).In brief, sections were rehydrated (xyleneÂ3, ethanolÂ2, 90% ethanol, 70% ethanol, running distilled water for 2 min each) and antigen retrieval was performed using sodium citrate buffer (10mM sodium citrate, 0.05% Tween 20, pH 6.0) for 10 min using a microwave (900-watt output).Sections were blocked in 0.1% Triton X-100 in PBS (T-PBS) containing 5% normal donkey serum (Sigma-Aldrich, Missouri, USA) for 1 hr.Sections were incubated with selected primary antibodies diluted in blocking buffer at RT for 18 hrs, followed by a wash (3 Â 10 min with 0.01% T-PBS).A cocktail of primary antibodies was used to stain for α-MNs, PNNs, and proteins of interest (Table S1, Data S2).Staining procedures were standardised between sections of WT and SOD1 G93A mice, as variables such as antibody concentration, incubation time, and incubation temperature may impact fluorescence intensity measurements [36,37].Sections with no primary antibodies were stained with secondary antibodies as negative controls for nonspecific background staining.Following staining, images were acquired using the Leica DMi8 inverted fluorescent microscope and Leica DMi8 SP8 laser point scanning confocal microscope (Supplementary Methods in Data S1).

Tissue extraction and enzyme-linked immunosorbent assay (ELISA) for total MMP-9 quantification
Mice were anaesthetized with sodium pentobarbitone (60-80 mg/kg intraperitoneally, Virbac, NSW Australia), and spinal cords were extracted and rapidly frozen in liquid nitrogen.The lumbar section of the spinal cord, as identified by the widening of the spinal cord from the dorsal view [38], was cut out and homogenised in ice-cold 1Â RIPA buffer (#9806, Cell Signalling Technology, Massachusetts, USA) with 1Â Protease Inhibitor Cocktail, EDTA-free (#87785, Thermo Fisher Scientific, Massachusetts, USA) using sonication.Homogenised tissue samples were centrifuged at 12,000g at 4 C for 15 min and supernatants were collected for analysis.Samples were diluted 20 times and an enzyme-linked immunosorbent assay (ELISA) was performed with a commercially available mouse MMP-9 (total) ELISA Kit (Quantikine MMPT90, R&D Systems, Minneapolis, USA) as per manufacturer's guidelines.All samples were assayed in duplicate, and results were corrected for total protein concentration.

PNN quantification
Large lumbar α-MNs were selected based on the following criteria: (i) NeuN-positivity, (ii) soma location within the ventral horn of the spinal cord [39], (iii) soma surface area greater than 450μm 2 [40,41], (iii) a multipolar cell body [41], and (iv) and clearly defined cytoplasm containing a nucleus [42].PNN analysis was conducted semiautomatically using an ImageJ plugin (https://github.com/NickCondon/NeuronQuantification; see Supplementary Methods).A region of interest around α-MNs was segmented based on anti-NeuN staining.The automatic segmentation was not able to select for only α-MNs in the ventral horn; therefore, regions of interest were manually added and removed to analyse only neurons of interest.The fluorescence intensity of PNNs (based on anti-aggrecan staining) in a 2 μm band extending from the perimeter of each α-MN was measured (see Supplementary Methods in Data S1).The average intensity of PNNs around α-MNs was calculated for each mouse.The PNN intensity measurements for SOD1 G93A mice were normalised and compared with their age-matched WT controls.The semi-quantitative analysis of PNNs using fluorescence intensity has been established by previous researchers [43][44][45][46][47].

Fluorescence intensity quantification
Fluorescence intensity analysis was also conducted semi-automatically using an ImageJ plugin (https://github.com/NickCondon/NeuronQuantification).Following the selection of α-MNs, microglia, and astrocytes, the fluorescence intensity of the antibody-labelled target protein inside each cell was analysed.Quantification of glial cells (especially astrocytes) can be challenging due to their morphological complexity.Therefore, only GFAP-labelled astrocytes and IBA-1-labelled microglia which had DAPI-labelled nuclei present in their cell body were counted (Figure S5 in Data S2).The average intensity measurement in the SOD1 G93A mouse was normalised and compared with its age-matched WT control.

Statistical analyses
Data was tested for normality using a Q-Q plot.Normally distributed data allowed for the use of an unpaired, parametric Student's ttest [48].Data was also tested for equal variance using an F-test.

RESULTS
PNNs breakdown around α-MNs in SOD1 G93A mice at disease onset  2E).This is consistent with previous studies suggesting MMP-9 expression in α-MNs is a marker for vulnerable, fast fatigable α-MNs destined to die in ALS [28].By contrast, low MMP-9 expressing α-MNs were not lost at any stage throughout the disease (Figure 2F).Interestingly, when comparing MMP-9 expression intensity in α-MNs between WT and SOD1 G93A mice, there is a trend of decreasing MMP-9 expression in α-MNs of SOD1 G93A mice beginning at the onset stage (0.4647 ± 0.08486, p < 0.05, n = 7) compared with WT mice (1.000 ± 0.1699, n = 8) (Figure 2G).This is also consistent with our MMP-9 ELISA assay which showed a trend of decreasing MMP-9 levels in the lumbar spinal cord of onset and mid-stage SOD1 G93A mice compared with age-matched WT controls (Figure S3 in Data S2).
This led us to hypothesise that pathological MMP-9 within the spinal cord of SOD1 G93A mice does not originate from α-MNs themselves, but rather from surrounding microglia and astrocytes in the ventral horn; increased non-neuronal MMP-9 in the shape of microglia and astrocytes supports this notion (Figure 2D 0 , D 00 ').

Activated glia are known to secrete MMP-9 and break down
PNNs [52][53][54], suggesting local recruitment of MMP-9 secreting, activated glia specifically to α-MNs may contribute to PNN breakdown at disease onset.To investigate this possibility, we examined the timeline of microglia and astrocyte recruitment to the ventral horn of SOD1 G93A mice compared with age-matched WT controls.
In addition to secreting MMP-9, microglia have been shown to break down PNNs through engulfment [54].To investigate whether microglia and astrocytes in the spinal cord also contribute to PNN breakdown via phagocytosis, the engulfment of PNN components by microglia and astrocytes was analysed in SOD1 G93A mice and agematched WT mice.Oxidative stress marker 3-NT is overexpressed by α-MNs in onset and mid-stage SOD1 G93A mice PNNs are thought to be neuroprotective against oxidative stress [16], and oxidative stress is a key factor in motor neuron degeneration in ALS [58][59][60].We, therefore, investigated the levels of oxidative stress in α-MNs during disease progression in SOD1 G93A mice compared with age-match WT mice.The expression of oxidative stress markers 3-nitrotyrosine (3-NT) and 8-oxo-2 0 -deoxyguanosine (8-oxo-dG), which indicate protein [61] and DNA oxidation [62], respectively, were investigated in α-MNs.3-NT fluorescence intensity was shown to be significantly increased in the α-MNs of onset stage SOD1 G93A mice (2.813 ± 0.5520, n = 7) compared with WT (1.000 ± 0.9005, p < 0.05, n = 7; Figure 7A-D).This increase was also observed at mid-stage disease (Figure 7D).This result contrasts with the expression of 8-oxo-dG, which was not increased in α-MNs throughout disease progression in SOD1 G93A mice (Figure S7 in Data S2).
Interestingly, some of the 8-oxo-dG expression which took on the appearance of nuclei was observed in microglia rather than in α-MNs in both WT and SOD1 G93A mice, which may be linked to the activation of microglia [63,64].8-oxo-dG expression which does not colocalize with neurons or microglia may be from the nuclei of astrocytes or other pre-immune cells not stained (Figure S8 in Data S2).

DISCUSSION
To the best of our knowledge, herein we report the first characterisa- was first observed at disease onset in SOD1 G93A mice.The timing of net breakdown correlated with increased microgliosis and an overall increase in non-neuronal MMP-9 present in the ventral horn of SOD1 G93A mice.MMP-9 is released by microglia and degrades aggrecan, a core proteoglycan of PNNs [52,53].In addition, increased phagocytosis of the PNN component aggrecan by microglia was observed at disease onset in our ALS mouse model, which may contribute to the loss of PNNs surrounding α-MNs.The expression of oxidative stress marker 3-NT was also shown to be increased in the affected α-MNs of onset stage SOD1 G93A mice.PNNs are neuroprotective against oxidative stress, and the loss of nets is implicated in increased oxidative stress in affected cells [65].Further to this, PNN breakdown was exacerbated at mid-stage disease in SOD1 G93A mice, with increased MMP-9 expression in microglia and astrocytes, as well as an increase in engulfment of PNN components by both glia in the ventral horn.
PNNs breakdown around α-MNs in SOD1 G93A mice at disease onset Previous studies have looked at PNNs in the spinal cord of SOD1 G93A mice [3] and mutant SOD1 rats [13]; however, neither study tracked the timing of PNN breakdown relative to motor deficit symptoms.
Furthermore, these studies did not measure PNN integrity specifically around α-MNs.Both factors are important as the timing of PNN breakdown around α-MNs relative to motor deficit symptoms provides information on whether the breakdown of these neuroprotective nets contributes to disease progression or is rather a consequence of α-MN loss.The key pathological hallmark of ALS is the loss of α-MNs that synapse onto muscle fibres [66][67][68][69], resulting in neuromuscular denervation and muscle weakness [70,71].It is important to investigate whether PNNs specifically surrounding these vulnerable α-MNs are affected during disease, rather than looking at PNNs throughout the whole lumbar spinal cord which may not directly affect α-MN survival.If PNN degradation is a trigger for motor deficit symptoms in ALS, then changes in net structure should be observed at the onset stage of the disease prior to major α-MN loss.This was observed in our study with PNN breakdown around α-MNs beginning at disease onset in SOD1 G93A mice.It is also important to note that no changes in PNNs around α-MNs were seen in WT mice or at the presymptomatic stage in SOD1 G93A mice, suggesting that PNN loss is specifically related to disease onset (Figure 1F).
Studies in the past have used Wisteria Floribunda agglutinin (WFA) to label PNNs [72][73][74][75]; however, aggrecan is a particularly interesting chondroitin sulfate proteoglycan (CSPG) to investigate as it is a core PNN protein that has the largest number of chondroitin sulfate chains, the greatest negative charge, and therefore, can sequester the most positively charged iron ions [76].It is also the only CSPG to be essential for protection against iron-induced oxidative stress [77].Further, we found aggrecan to be a superior marker of PNNs around α-MNs compared with WFA, which is typically used to stain PNNs (Figure S2 in Data S2); a finding supported by other researchers [78][79][80].This difference in PNN staining is due to WFA preferentially labelling nonsulfated chondroitin sulfate chains of PNNs rather than the core proteoglycan itself [81].
MMP-9 has been shown to play a pathological role in ALS with elevated levels of MMP-9 being observed in the spinal cord and skin of post-symptomatic SOD1 G93A mice [27].Increased MMP-9 levels have been associated with PNN breakdown in several models of disease [21,25,26,30,82] and reducing MMP-9 has been shown to mediate neuroprotective effects in ALS [28,29].It has been suggested that vulnerable α-MNs, which selectively express MMP-9 in the spinal cord of both WT and ALS mice, may be overexpressing MMP-9 in SOD1 G93A mice.In this study, MMP-9 levels in α-MNs of SOD1 G93A mice were decreased at disease onset (Figure 2G).If MMP-9 inside α-MNs is pathological in ALS, then MMP-9 expression should be increased in α-MNs at disease onset when symptoms begin to manifest.This finding conflicts with previous studies, which knocked out and/or knocked down MMP-9 in SOD1 G93A mice, suggesting that increased MMP-9 in α-MNs drives endoplasmic reticulum stress in fast SOD1 G93A α-MNs [28].If neuronal MMP-9 is a driver of stress in SOD1 G93A mice, then it would be expected that there would be more MMP-9 present in the affected α-MNs, not less, as we have observed in the α-MNs of SOD1 G93A mice (Figure 2).Our findings suggest that the MMP-9 that is secreted and degrades PNNs originates from the increased number of recruited microglia and astrocytes in the ventral horn of SOD1 G93A mice.This also may explain why PNNs around WT α-MNs, which also express high levels of MMP-9, are not degraded.It may be that α-MNs are not actively secreting MMP-9, whereas microglia and astrocytes do [52,53,83].
Increased numbers of MMP-9 expressing microglia and astrocytes are observed in the ventral horn of onset and mid-stage SOD1 G93A mice Despite no increase in MMP-9 expression in SOD1 G93A microglia at disease onset, there was a two-fold increase in the number of microglia containing the same levels of MMP-9; therefore, there was a net increase in the amount of MMP-9 that is secreted from glia present in the ventral horn of SOD1 G93A mice (Figure 3E).This was the same time at which PNNs were first degraded in SOD1 G93A mice (Figure 1F).A further increase in microglia numbers and astrocyte numbers, as well as a significant increase in MMP-9 expression in both microglia and astrocytes, was observed at mid-stage disease (Figure 3E-H) when PNN breakdown was exacerbated (Figure 1F).
Consistent with our findings, MMP-9 expression has previously been observed to colocalize with microglia in end-stage SOD1 G93A mice [28].Activated microglia and astrocytes have been shown to PERINEURONAL NETS ARE PHAGOCYTOSED BY GLIA IN AN ALS MOUSE MODEL secrete MMP-9 [52,53,83].These findings combined with our results suggest that MMP-9 secreted by microglia and astrocytes may be responsible for the cleavage and breakdown of PNNs surrounding α-MNs in SOD1 G93A mice.
These results must be reconciled with our quantitative ELISA assay analysis of total MMP-9 protein concentration within the lumbar spinal cord of onset and mid-stage SOD1 G93A mice compared with age-matched WT controls.The ELISA assay showed a trend of decreasing MMP-9 concentration within the lumbar spinal cord of onset, especially in mid-stage SOD1 G93A mice (Figure S3 in Data S2).This may be because high MMP-9 expressing α-MNs were lost during the onset and mid-stage disease in SOD1 G93A mice compared with age-matched WT controls (Figure 2E; confirmed by Kaplan et al, 2014 [28]).The largest reservoirs of MMP-9 in the spinal cord were large α-MNs, with the intensity of MMP-9 staining within these α-MNs being greater than MMP-9 staining that colocalizes with the glia (Figure 2A-D; Figure S4 in Data S2; [28]).
Further, our results serve to reveal the strength of semiquantification of immunostained sections; namely that immunostained comparisons do provide MMP-9 expression data that is directly linked to the identity of a particular cell type, in this case, motor neurons and glia.This is a critical feature of our study given that motor neurons and glia both express MMP-9 and that during disease progression, we are losing α-MNs, and therefore, the MMP-9 they contain [28].At the same time, we see increased numbers of glia that are expressing MMP-9.Hence, our claim of increased MMP-9 is in the context of glia expressing MMP-9, and not due to the loss of MMP-9 from motor neurons that died.
Our findings suggest that the MMP-9 that is secreted and degrades PNNs originates from the increased number of recruited microglia and astrocytes in the ventral horn of SOD1 G93A mice; an idea supported by previous studies that demonstrate glia do secrete MMP-9 [52,53,83].Further, our findings may also explain why PNNs around WT α-MNs, which also express high levels of MMP-9, are not degraded, suggesting that α-MNs are not actively secreting MMP-9.
The trend of decreased total MMP-9 in the lumbar spinal cord of onset and mid-stage SOD1 G93A we observed in our study must also be reconciled with previous studies which showed increased total MMP-9 in the spinal cord of post-symptomatic SOD1 G93A mice compared with WT controls [27].Fang et al (2010) [27] describe extracting the spinal cord at 'just below the skull and at hip level', this procedure would result in the collection of cervical, thoracic, and lumbar spinal cord regions to be included in their MMP-9 ELISA assay.In our MMP-9 ELISA assay, we included only the lumbar region of the spinal cord to collect only spinal cord regions that show significant α-MNs loss in SOD1 G93A mice [11,33,40].Hence, collection differences in the spinal cord regions might explain these conflicting results.These discrepancies highlight the importance of utilising both quantitative (e.g., ELISA assay) and spatial techniques (e.g., immunofluorescence) when investigating changes in complex in vivo environments.
Microglia and astrocytes engulf PNN components in onset and mid-stage SOD1 G93A mice Microglia and astrocytes facilitate the breakdown of PNNs not only through MMP-9 release but also via engulfment.We first observed increased engulfment of PNN components in the onset stage SOD1 G93A mice by microglia (Figure 4F), concomitant with initial PNN breakdown (Figure 1F).This was followed by a further significant increase in aggrecan engulfment by microglia (Figure 4F), and also by astrocytes at mid-stage disease in SOD1 G93A mice (Figure 5F), as PNN degradation was maintained (Figure 1F).The degree of PNN engulfment by microglia and astrocytes coincided with the initial loss and exacerbation of PNN degradation seen at the onset stage and midstage disease, respectively.The activation and inflammatory states of glia were not investigated in this study; however, qualitative observations of microglia and astrocyte morphology in mid-stage SOD1 G93A mice indicate activation.This is shown through glia having a larger, swollen cell body with short, thicker processes [84][85][86][87][88] (Figure 3B-D) as opposed to a small soma and fine, long processes observed in resting glia [85,89] (Figure 3A).The significant increase in the engulfment of PNN components by microglia and astrocytes at mid-stage disease (Figures 4E and 5E) further indicates activation [90,91].
Microglia in the healthy CNS play a role in synapse formation and the learning of new memories, where small quantities of PNNs are phagocytosed [92][93][94].The significant and pathological phagocytosis of PNNs has previously been observed in other neurodegenerative diseases, such as the increased engulfment of aggrecan in the brains of Alzheimer's disease and Huntington's disease model mice [54,95].Here, we report significant engulfment of PNNs by astrocytes at mid-stage disease in SOD1 G93A mice (Figure 4).The role of astrocytes in the phagocytosis of extracellular components is less known; however, the ability of astrocytes to function as phagocytes upon activation has been previously observed in brain ischaemia mouse studies [96] and in vitro [97].
To investigate what may be driving glial recruitment and activation, the expression of 'find-me signals' was investigated in α-MNs.
CX3CL1 is expressed by α-MNs in mid-stage SOD1 G93A mice The expression of the three 'find-me signals' CX3CL1, S1P and LPC in α-MNs was investigated in onset and mid-stage SOD1 G93A and WT control mice.S1P and LPC expression in α-MNs was not different between SOD1 G93A mice and WT mice at these key stages of disease (Figure S5 in Data S2).Interestingly, we found a significant increase in CX3CL1 expression within α-MNs of mid-stage SOD1 G93A mice, with no changes being observed at disease onset (Figure 6D).This was unexpected as there was a significant increase in the number of microglia in the ventral horn of onset stage SOD1 G93A mice (Figure 3E).If CX3CL1 functions as a 'find-me signal' in SOD1 G93A mice, then its expression should be increased at the onset stage when increased microglia recruitment occured.These findings suggest that in this context, CX3CL1 is not a 'find me' signal for microglia.
Other studies suggest that CX3CL1 can act as a ligand for microglia activation and stimulation of their phagocytotic activity [98][99][100], an idea supported by studies which knocked out CX3CL1 receptor in WT mice that showed decreased PNN degradation and engulfment within the dorsal horn of the spinal cord [47].Taken together our findings are consistent with CX3CL1 acting as a ligand for microglia activation and stimulation of their phagocytotic activity, as increased PNN engulfment per microglia was first detected in mid-stage SOD1 G93A mice (Figure 4E), at the same time when CX3CL1 expression was increased in α-MNs (Figure 6J).There was a significant increase in the expression of 3-NT, a marker of protein oxidation [106], within the α-MNs of both onset and midstage SOD1 G93A mice (Figure 7G), coinciding with the time of initial PNN breakdown (Figure 1F).These findings are consistent with previous observations of increased 3-NT localised to the α-MNs of onset to mid-stage disease SOD1 G93A mice [107]; increased free 3-NT in the lumbar spinal cord of symptomatic mutant SOD1 G37R mice [108]; and increased 3-NT in α-MNs of symptomatic SOD1 G93A mice coexpressing the Copper-Chaperone-for-SOD [109].By contrast, we saw no changes in the expression of 8-oxo-dG within the α-MNs of SOD1 G93A mice at the key disease stages (Figure S7 in Data S2).Furthermore, some of the expression of 8-oxo-dG colocalized with the cell bodies of microglia in both WT and SOD1 G93A mice (Figure S8 in Data S2).Previous studies have reported 8-oxo-dG expression in microglia which display an activated phenotype, further supporting the assumption that the microglia are activated [63,64].metal ions [16,17].These include iron which can be transported through the plasma membrane to stimulate oxidative stress [65].Oxidative stress markers are elevated in ALS patients [104,110].The ability of PNNs to sequester iron ions and prevent them from entering the cell is suggested to be due to their overall negative charge [111].
The glycosaminoglycan side chains of chondroitin sulfate proteoglycans are made up of D-N-acetylgalactosamine β (1-4) and repeating disaccharide units of D-glucuronic acid β (1-3) which contain free carboxylic acid groups that are negatively charged and bind positively charged ions, such as iron [112].This neuroprotective feature of PNNs has been observed in GCLM knock-out mice, a mouse model for schizophrenia [113], rat models of traumatic brain injury [67,77] and experiments injecting FeCl3 in mice lacking PNN components [106].
In summary, our study has revealed the breakdown of PNNs specifically around α-MNs of SOD1 G93A ALS model mice at disease onset.
Our working model postulates that microglia are recruited to the ventral horn of SOD1 G93A mice at disease onset resulting in high levels of MMP-9 to be secreted into the ventral horn.MMP-9 cleaves aggrecan and degrades PNNs [104].Microglia and astrocytes also engulf mal experimentation and were conducted in accordance with the state (Queensland Government Animal Research Act 2001, associated Animal Care and Protection Regulations, 2002 and 2008), and national (Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, eighth Edition) guidelines.
To investigate whether PNNs play a role in neuroprotection, the timeline of PNN breakdown in the lumbar spinal cord of SOD1 G93A mice compared with age-matched WT control mice was determined (Figure1A-E).PNN breakdown was first observed at disease onset with a significant decrease in normalised PNN fluorescence intensity in SOD1 G93A mice (0.6170 ± 0.09277, p < 0.05, n = 7) compared with WT mice (1.000 ± 0.1074, n = 8).The additional net breakdown was observed at mid-stage with a significant decrease in normalised PNN fluorescence intensity in SOD1 G93A mice (0.4118 ± 0.01638, p < 0.05, n = 4) compared with WT (1.000 ± 0.1786, n = 4) (Figure1F).PNN breakdown coincided with the initiation of motor deficit symptoms as indicated by a hind-limb grip strength test, and α-MN loss in SOD1 G93A mice[33].Maintained net breakdown at mid-stage also coincided with the exacerbation of motor deficit symptoms as indicated by the rotarod test, and increased α-MN degeneration previously reported at this stage[50,51].MMP-9 expression is not increased inα-MNs of SOD1 G93A mice PNN loss may occur due to cleavage by MMP-9 [23-26].To investigate this possibility, the timeline of MMP-9 expression in the lumbar spinal cord of SOD1 G93A mice compared with age-matched WT control mice was examined (Figure 2A-D).α-MNs were grouped based on their MMP-9 expression into high and low MMP-9 expressing α-MNs.High MMP-9 expressing α-MNs were preferentially lost in SOD1 G93A mice at disease onset (1.667 ± 0.4933 cells per ventral horn, n = 7) compared with WT (4.000 ± 0.7292 cells per ventral horn, p < 0.05, n = 8) (Figure Figure 3E); therefore, there was a net increase in the amount of MMP-9 present in the ventral horn in onset-stage SOD1 G93A mice.No changes in the number of astrocytes were observed at disease onset between SOD1 G93A mice and age-matched WT mice (Figure3F).By mid-stage disease there was a significant increase in both the number of astrocytes (75.71 ± 10.07 cells per ventral horn, n = 4) and microglia (52.50 ± 10.07 cells per ventral horn, n = 4) in SOD1 G93A mice compared with age-matched WT controls (microglia:

9 ± 8 .
Microglia and astrocytes engulf PNN components in onset and mid-stage SOD1 G93A miceTo analyse the engulfment of PNN components by glia, confocal microscopy z-stack images of the ventral horn stained for ACAN and IBA-1 (Figure4A-D) or ACAN and GFAP (Figure5A-D) were used to create 3D renderings of microglia (Figure4A00 -D 00 ) or astrocytes with engulfed aggrecan components (Figure5A00 -D 00 ) (Supplementary Methods in Data S1).There was no difference in the volume of aggrecan engulfed by microglia (Figure4E) or astrocytes (Figure5E) in onset SOD1 G93A mice compared with WT.However, due to increased microglia numbers in the SOD1 G93A mouse, the total amount of aggrecan engulfed by microglia in the ventral horn was significantly increased in onset SOD1 G93A mice (59.86 ± 6.566 μm 3 per ventral horn, n = 5) compared with WT (21.08 ± 4.639 μm 3 per ventral horn, p < 0.01, n = 5; Figure4F).No difference in the total amount of aggrecan engulfed by astrocytes in the ventral horn was observed between onset SOD1 G93A mice and age-matched WT controls (Figure5F).This observed engulfment by microglia and astrocytes increased by mid-stage disease where there was a significant increase in aggrecan engulfed per microglia (60.42 ± 6.420 μm 3 per microglia, p < 0.01, n = 4; Figure4E) and aggrecan engulfed per astrocyte (20.61 ± 0.7360 μm 3 per astrocyte, p < 0.0001, n = 4; Figure5E) in SOD1 G93A mice compared with WT microglia and astrocytes (microglia: 6.975 ± 1.479, n = 4; astrocyte: 6.831 ± 0.5224, n = 4).Similar to the onset stage, there was also an increase in the total amount of aggrecan engulfed by microglia (3172 ± 337.0 μm 3 per ventral horn, p < 0.01, n = 4; Figure4F) and astrocytes (1560 ± 55.72 μm 3 per ventral horn, p < 0.0001, n = 4; Figure 5F) in the ventral horn of mid-stage SOD1 G93A mice compared with WT controls (microglia: 105.2 ± 22.31 μm 3 per ventral horn, n = 4; astrocyte: 113.707 μm 3 per ventral horn, n = 4).This overall increase in aggrecan engulfment throughout the ventral horn by microglia at disease onset along with the overall net increase in MMP-9 throughout the ventral horn coincided with the timing of PNN breakdown around α-MNs.Glia-induced PNN breakdown requires the recruitment of microglia and astrocytes to the ventral horn of the spinal cord where α-MNs reside.To determine what factors might be driving the recruitment of glia towards α-MNs in ALS, we investigated the expression of 'find-me signals' in the α-MNs of SOD1 G93A mice.CX3CL1 is expressed by α-MNs in mid-stage SOD1 G93A mice 'Find-me signals' are chemoattract molecules released by stressed cells to drive microglia migration towards them.To date, three main find-me signals have been reported in mammals: fractalkine (CX3CL1), sphingosine-1-phosphate (S1P) and lysophosphatidylcholine (LPC) [55-57].We analysed the expression of these three find-me signals in α-MNs throughout the ventral horn during disease progression in SOD1 G93A mice compared with WT mice.We observed a significant increase in CX3CL1 in α-MNs of mid-stage SOD1 G93A mice (3.544 ± 0.4431, p < 0.001, n = 6) compared with WT (1.000 ± 0.2213, n = 7; Figure 6A-D).No differences in S1P or LPC expression were observed between the two genotypes (Figure S6 in Data S2).
tion of PNNs specifically surrounding α-MNs throughout disease progression in a well-characterised ALS mouse model.PNN breakdown F I G U R E 6 Increased CX3CL1 expression is observed in α-MNs during mid-stage disease in SOD1 G93A mice.(A-C) Representative images of CX3CL1 (red) in α-MNs (green) in the ventral horn of the lumbar spinal cord of wild type/strain control (WT, age-matched) mice, and onset (OS) and mid-stage (MS) SOD1 G93A mice.Arrows indicate α-MNs of interest.Dotted lines indicate the border of the ventral horn.Scale bars = 100 μm.(D) The CX3CL1 intensity in the α-MNs of SOD1 G93A mice was normalised to their respective age-matched WT (proportion of WT; mean ± SEM).A Student's t-test was performed between WT (orange bars; OS: n = 8, MS: n = 7) and age-matched SOD1 G93A mice (green bars; OS: n = 7, MS: n = 6) to determine significance.20Â objective on the Leica DMi8 SP8 inverted widefield microscope.F I G U R E 7 Increased 3-NT expression is observed during disease onset in SOD1 G93A mice.(A-C) Representative images of 3-NT (green) in α-MNs (red) in the ventral horn of the lumbar spinal cord of wild type/stain control (WT, age-matched) mice, and onset (OS) and mid-stage (MS) SOD1 G93A mice.Arrows indicate α-MNs of interest.Dotted lines indicate the border of the ventral horn.Scale bars = 100 μm.(D) 3-NT intensity in the α-MNs of SOD1 G93A mice was normalised to their respective age-matched WT (proportion of WT; mean ± SEM).A Student's ttest was performed between WT (orange bars; n = 7, MS: n = 7) and age-matched SOD1 G93A mice (green bars; n = 6, MS: n = 4) to determine significance.20Â objective on the Leica DMi8 SP8 inverted widefield microscope.

Oxidative stress marker 3 -
NT is overexpressed by α-MNs in onset and mid-stage SOD1 G93A mice To investigate if oxidative stress is produced in α-MNs of SOD1 G93A mice following PNN degradation, we stained for common markers of oxidative stress in ALS which include 3-NT and 8-oxo-dG [101-105].
The above findings add support to the neuroprotective role of PNNs; namely, their proposed role in protecting neurons against oxidative stress.PNNs perform this function by acting as a chemical barrier by preventing the entry of positively charged potentially toxic F I G U R E 8 Perineuronal nets (PNNs) are broken down by a process of MMP-9 release from and engulfment by microglia and astrocytes.Shown is our working model of PNN breakdown around α-MNs in the SOD1 G93A ALS mouse model compared with WT, based on our data.(1) Activated microglia and astrocytes are recruited to α-MNs defined by their upregulation of MMP-9 by number, and their morphology.(2) The MMP-9 released by these microglia assist in the breakdown of PNNs around α-MNs and digests the PNN fragments.(3) CX3CL1 triggers microglia to phagocytose PNNs from α-MNs.(4) These α-MNs are then made vulnerable to stress via the upregulation of known intracellular oxidative stress markers such as 3-NT.
PNN components through phagocytosis.The degree of net breakdown was correlated with the number of glial cells in the ventral horn of the spinal cord, the expression of MMP-9 inside microglia and astrocytes, and the number of phagocytosed PNNs engulfed by both glial cells.The loss of these neuroprotective nets may expose α-MNs to iron ions and toxic reactive oxygen species in the extracellular matrix, resulting in increased oxidative stress and eventual α-MN degeneration (Figure8).AUTHOR CONTRIBUTIONS Peter G. Noakes, Mark C. Bellingham, David G. Simmons and Sang Won Cheung conceived the study.Ekta Bhavnani performed the initial PNN experiments in SOD1 G93A mice.Sang Won Cheung performed the presented experiments and analyses.All authors made contributions to the drafting of the final manuscript and presentation of the final figures.