Sex-specific microglial responses to glucocerebrosidase inhibition: relevance to GBA1-linked Parkinson disease

Microglia are heterogenous cells characterized by distinct populations each contributing to specific biological processes in the nervous system, including neuroprotection. To elucidate the impact of sex-specific microglia heterogenicity to the susceptibility of neuronal stress, we analysed the dynamic changes in shape and motility occurring in primary mouse microglia following pro-inflammatory or neurotoxic insults, thus finding sex-specific responses of microglial subpopulations. Male microglia exhibited a pro-inflammatory phenotype, whereas female microglia showed enhanced neuroprotective capabilities associated with the activation of Nrf2 detoxification pathway in neurons. The sex difference in neuroprotective functions is lost by inhibition of glucocerebrosidase, the product of the GBA1 gene, mutations of which are the major risk factor for Parkinson’s disease (PD). This finding is consistent with the increased risk of PD observed in female carriers of GBA1 mutation, when compared with wild type population, suggesting a role for microglial functionality in the etiopathogenesis of PD-GBA1.


Pag 4
risk factor for motor neuron disorders 25 and PD 22 . In this context, another clinically relevant (genetic) risk factor for PD is the presence of specific mutations in the GBA1 gene, which have been detected in up to 5-25% of patients 26,27 . This gene encodes for a lysosomal hydrolase, namely the glucocerebrosidase (GCase): biallelic mutations in GBA1 causes Gaucher Disease (GD) 28 , while heterozygotic carriers do not develop GD but retain the increased risk to develop PD 29 .
Although most studies previously focused on the functional effects of GBA1 mutations in neurons, our recent investigations revealed that GCase inhibition in microglia is sufficient to impair the physiological ability of microglial to protect neurons against oxidative stress and neurotoxic stimuli 30 : this acquired microglia phenotype may contribute to the increased risk of neurodegeneration observed in GBA1 carriers.
To investigate the microglial phenotype due to GCase inhibition, in the current study, we developed and applied a non-invasive imaging approach on primary cultures generated by murine models of both sexes. This original methodology allowed us to record in real time the changes of cell morphology induced by specific pharmacological stimuli, with the aim of associating the dynamic variation in cell shape and motility to the biochemical effects induced by the treatments. With this analysis we found that the effects of GCase inhibition in microglia are sex-dependent, thus showing a greater loss of neuroprotective ability of female's as compared to male's microglia.

Image-based microglia analysis allows detecting functional clusters
To investigate the changes of microglia morphology occurring as a consequence of specific stimuli, we generated an unbiased imaging approach allowing for the dynamic quantification of shape and movement variations of single cells over a fixed period of time. To mimic the physiological microglial environment, we seeded primary adult microglial cells obtained from CX3CR1 +/GFP mice, constitutively expressing GFP 31 , on a layer of neuron-enriched primary culture of cortical cells from syngeneic wild type mice ( Supplementary Fig. 1) known to structurally support microglia microglia over 2 h; the recorded movies were processed with the ImageJ software 32 to obtain morphological and kinetic descriptors for each cell in the acquired field of view ( Fig. 1A and Supplementary Movie 1). Briefly, the background was subtracted from the acquired images, which were in turn binarized using a defined threshold that enabled cluster regions of pixels based on similarities threshold to distinguish cell shapes and generate an object for each cell; then the binarized images were processed to remove noise by smooth and despeckle functions of the software to produce sharp objects. A threshold of 130 µm 2 for the surface size was selected to sort the shapes of cells (microglia) from those originating from cellular debris. The selected shapes were processed to measure two static morphology descriptors: the cell area in square micrometers and the solidity ( Fig. 1B) 33 , the latter defined as the ratio of the area divided by the area of the smallest convex set polygon that contains the cell 32 , thus resulting in a higher solidity for ameboid rather than for ramified shapes, in a range from 0 to 1 values (Fig. 1B). For each cell, measurements were taken in every frame of the time lapse acquisition; median values of these measurements described the predominant morphology during recording and were used to generate the graphs ( Fig. 1C and   1E). Coefficients of variation (CV%) for area and solidity were calculated to obtain numeric descriptors of the dynamic changes occurring during the 2-h measurements ( Fig. 1D and 1F). The CV% of the cell area due to size variation was taken as a surrogate marker of cell contractility, while the CV% of the solidity was considered as a measure of the morphological modifications in terms of complexity.
Since microglia are cells able to sense the environment and migrate in response to specific stimuli 34 the distance traveled by microglia during the recording time was also considered as a parameter inherently linked to their activity: distance was calculated by tracing the shift in the center of mass of each cell occurring frame by frame, in terms of coordinates (x, y). To define the total covered distance, all shifts were summed and converted into µm values (Fig. 1H). Finally, we measured the number of rotations performed by each cell, another descriptor representing microglial dynamics: in Pag 6 order to calculate this parameter, the ellipse in which the cell can be inscribed was identified and used to calculate the angular displacements frame-by-frame, which were in turn added up to obtain the total rotation of the cells during the recording, expressed as angular degree values (Fig. 1G).
To validate the method we analyzed the descriptor changes associated with a well-characterized microglia polarization, namely the one caused by the potent endotoxin lipopolysaccharide (LPS) 35 a strong inducer of a pro-inflammatory microglial phenotype 36 . Male-derived microglial cells were cultivated on the layer of primary neuron-enriched cultures for 24 h and treated with 10 µg/ml LPS; microglia morphology and motility were analyzed and compared with vehicle-treated cells, by processing videos captured from 6 up to 8 h after the treatment, a time point that is associated with high gain of pro-inflammatory features 37 . The experiment revealed that the selected descriptors were effective in detecting and describing specific features of microglia induced by LPS ( Fig. 1 and Supplementary Movie 2) 38,39 . In detail, the area of the analyzed cells did not change during the acquisition ( Fig. 1C and 1D), but the treatment induced an increase in their solidity of about 13% meaning that when microglia were treated with LPS, their shape became more ameboid (Fig. 1F), while cells maintained their complexity across time, since the variation of solidity (CV% solidity) was higher in vehicle-treated and lower in LPS-treated cells (Fig. 1G). As expected, the cell kinetics was also affected by LPS, showing an increase of about 43% in the number of rotations (Fig. 1G), and of about 27% in the covered distance ( Fig. 1H) when compared with vehicle-treated cells.
The results showed that the single-parameter analysis was efficiently identifying phenotypic changes induced by a strong stimulus -as potent as LPS is -occurring in the overall microglial population, but did not provide any detail on the presence of microglia subpopulations with different behavior (Fig. 1 A-H, Supplementary Movie 2). This is particularly important, since microglia shows a peculiar heterogeneity in physiological condition suggesting the existence of various subpopulations reacting differently upon stimulation 40,41 ; we attempted to discriminate these different microglia subpopulations by combining our cellular descriptors in a cluster analysis.
We performed a biparametric analysis to test whether we could distinguish the existence of distinct Pag 7 morpho-functional categories: the medians of morpho-dynamic descriptors were used as a cutoff to assign each cell to a descriptive category representative of a value above or under the media. By using the combination of two parameters, cells were clustered into four different subpopulations. As an example, by analyzing solidity and area ( Fig. 2A) it was possible to generate four clusters representing microglia subpopulations: Cluster 1) "simple & big", Cluster 2) "simple & small", Cluster 3) "complex & big" and Cluster 4) "complex & small". Cluster 1 is composed of cells that have both area and solidity above the median, in contrast, the cells with area and solidity under the media fall in Cluster 4; cells with the bigger area and low solidity fall in Cluster 3, and the cell small and simple in Cluster 2. The categories generated for each parameter are reported in Table 1 and Supplementary  Interestingly, the method was able also to detect that some categories of cells did not respond to LPS and their subpopulation remained unaffected after the treatment, e.g. "steady & static", "variable & motile", "complex & motile", "simple & static" (Fig. 2B). These results demonstrated that the dynamic morpho-functional analysis allowed to discriminate microglial subpopulations differentially responding to specific stimulations. Pag 8 Once validated, the morphometric approach was used to test whether sex-differences could be detected in the dynamic behavior of microglia. To this end, primary brain microglia cells from male or female mice were isolated from adult CX3CR1 +/GFP and seeded on neuron-enriched primary cortical cells from syngeneic wild type mice (mixed population of male and female mice). 24 h after seeding microglia dynamics were recorded for 2 h to identify possible sex-related differences in unstimulated conditions. Indeed, different subpopulations were present in female and male microglia: when compared to male, female microglia showed cluster subpopulations of "big & static", "variable & static", "inactive & complex", "complex & small", "complex & static", and "rotant & static" (Fig. 3). These data suggested that female microglia in physiological conditions are enriched in subpopulations characterized by complex and static cells, possibly interacting with the surrounding micro-environment, with a less pro-inflammatory profile: this is in accordance with previous reports indicating that, in female mice microglia are more dedicated to the maintenance of brain homeostasis, while male microglia are more inclined to perform defensive tasks 19 .

Chemical inhibition of β-glucocerebrosidase (GCase) exerts a differential effect in male and female microglia.
We previously demonstrated that the pharmacological inhibition of microglial GCase with conduritol-B-epoxide (CBE) interferes with the neuroprotective function of microglia 30 . To characterize the microglia morphology in response to GCase inhibition, we carried out the morphofunctional analysis after treating cocultures with 200 µM CBE: this concentration was selected in order to ensure a almost total (-98% activity) inhibition of GCase activity sufficient to selectively interfere with the microglia neuroprotective functions 30 , while with negligible effects on the activity of additional glycosidase targets 43, 44 . The dynamic changes of microglial morphology were recorded at early time points (48 to 50 h after treatment), a time window in which neuronal and microglia mortality due to CBE were virtually absent 30 . The morphometric analysis revealed that GCase inhibition of male microglia changed the phenotype of specific sub-populations, increasing Since it has been reported that microglia can acquire a pro-inflammatory phenotype after long-term GCase inhibition 45,46 , we compared the phenotype triggered by LPS ( Fig. 2B)  Based on these results, we decided to test whether this more pronounced effect of CBE on female microglia also reflected alterations in their neuroprotective functions; indeed, we previously demonstrated that microglia are able to increase neuronal NRF2 transcriptional activity that protect Pag 10 neurons from neurotoxin effects, a mechanism which is impaired by GCase inhibition 30 . We purified microglia from groups of female and male wild type mice treated with vehicle or 100 mg/kg CBE for 3 days to inhibit microglial GCase 30 , the purified microglia was seeded over a neuronal cell layer derived from ARE-luc2 mice (Fig. 6A), transgenic animals in which a luciferase reporter is expressed under the control of the Nrf2 transcription factor 47,48 . This system allowed us to measure the ability of microglia to increase the neuronal Nrf2 activity simply by measuring the luciferase activity in the coculture. Interestingly, female microglia extracted from vehicle-treated mice revealed a more prominent effect in inducing Nrf2 response when compared to male microglia. The effect of CBE treatment, that is expected to reduce neuronal to microglia Nrf2 response 30 , was sufficient to blunt the differences observed between male and female microglia obtained from vehicle-treated mice (Fig. 6B), thus suppressing the neuroprotective action exerted by microglia independently from the sex of origin.
Based on these data, which suggest that a reduction in GCase activity decreases the protective microglial response in female mice, we hypothesized that the normal male predominance seen in PD patient populations would be abolished in PD subjects with GBA1 variants.
We analyzed the AMP-PD database that includes a total of 3497 individuals (Fig. 7). Of these, 1971 (56.4%) were males and 1526 (43.6%) were females. For idiopathic PD cases alone, there were 1236 males (63.4%) and 715 females (36.6%). In the GBA-PD group there were 163 males (57.4%) and 121 females (42.6%). Statistical analysis showed that the male predominance in idiopathic PD is lost in GBA1-associated PD, although this just fails significance at p=0.0525.

Discussion
Microglial cells are characterized by the presence of different subpopulations, which differ in abundance and morphology, and are characterized by distinct genetic programs, protein expression patterns, and ability to respond to environmental stimuli 14,15 . The distribution of these subpopulations follows a spatial-temporal definite pattern: indeed, specific phenotypes can be Pag 11 detected at different evolutionary stages, but they can also coexist simultaneously in brain parenchyma of adult animals 40, 49 . The morphology of microglia is indicative of their functional status 10,50 , thus analysis of microglial shape can anticipate information about the biochemical pathways triggered in these cells by pathophysiological processes 51 . Standard morphological analysis based on immunocytochemistry images provides snapshots of cell shape and offers a static view of the cell population 33 but it does not detect dynamic changes, such as the variation in cell protrusions or changes in migration, features that are certainly key components of microglial biology and allow better deciphering their behavior 52 . In our study, we added the temporal dimension to standard shape analysis by applying time-lapse fluorescence microscopy to our in vitro model of the multicellular condition of the brain, encompassing a co-culture of GFPexpressing primary microglia and primary cortical cells enriched in neurons. In this context, we applied an unbiased imaging-based analysis for each microglia cell of the investigated population, and for each frame of the recorded movies, we measured standard morphological cues that included cell dimension and complexity, together with novel dynamic descriptors able to describe timedependent changes in microglia motility, contractility, rotation, and complexity. The method was effective for detecting changes occurring in pro-inflammatory microglia, which have been previously described 38,53 and include changes in the cellular shape towards the amoeboid morphology, and a general increase in the motility ( Fig. 1 and 2).
The first set of experiments was designed to validate the method, and the results were in line with prior knowledge, but at the same time revealed information on the response of primary microglia to pro-inflammatory stimuli by disclosing resilient subpopulations of cells that did not undergo substantial changes after stimulation. These subpopulations show phenotypes that -following the classification generated by our protocol -are defined by the descriptors "complex & motile", "simple & static", "stationary & motile" and "rotant & static", and display a non-responsive phenotype against LPS stimuli (Fig. 2B). Thus, our morpho-functional analysis provided a direct demonstration that adult microglia exist in different subtypes, each characterized by peculiar Pag 12 shapes, and possibly by different gene expression profiles and function 40,51 , and that these subtypes can differently respond to stimulations.
Once validated, we have applied the morpho-functional analysis to evaluate sex-related differences in the composition of microglia subpopulations. In previous studies, biochemical, morphological, and functional data recognized sex-related differences in microglia revealing that male microglia have a constitutive mild pro-inflammatory phenotype, while female microglia are reminiscent of surveilling microglia, that for definition are stationary and ramified cells sensing the environment 19,54,55 ; these differences were shown to be genetically determined, independent of hormonal status and of the microenvironment, indeed are maintained also when microglia are maintained in culture, when cross-transplanted in a brain of opposite sexes and when microglia were extracted from brains of ovariectomized females, where the hormonal environment was similar to male mice 19 . It is likely that these sex-related differences in microglia are contributing to the differential sex-specific susceptibility and severity of some neurological diseases 12,56,57 .
In our morpho-functional analysis, male microglia, when compared to female cells were enriched in sub-populations defined by the descriptors "small & motile", "simple & motile", "contractile & simple", "steady & motile", "small & simple", "simple & big" (Fig. 3), which are increased when cultures are treated with LPS (Fig. 2B), thus supporting the notion that male microglia show a higher tendency to acquire a pro-inflammatory phenotype than female microglia. In contrast, in female microglia we found a marked presence of subpopulations defined by the descriptors "big & static", "variable & static", "inactive & complex", "complex & small", "complex & static" (Fig. 3), categories that are decreased after LPS treatment (Fig. 2B) supporting the hypothesis that female cells show a less pro-inflamed phenotype, with a profile reminiscent of surveilling microglia 10,19 . Do these differences influence the development and progression of a neurodegenerative process?
Previous data from our lab showed that microglia may contribute to brain neuroprotection by inducing the Nrf2 pathway in neurons through direct contact between microglial cells and neurons. This Nrf2-activation is reduced when microglial GCase is pharmacologically inhibited, an effect Pag 13 that renders dopaminergic neurons more sensitive to neurotoxic stimulations 30 . On the basis of these data, we hypothesized that the reduced microglial neuroprotective functions might contribute to the observed increased risk of PD in carriers of GBA1 mutations and prompted us to analyze the microglial morpho-functionality after GCase inhibition. Interestingly, the morpho-functional analysis on static descriptors demonstrated that inhibition of GCase enriched the microglial population with cells characterized by an ameboid-like ("simple & big") morphology like those observed with LPS (Fig. 4) and typical of the pro-inflammatory activation. Similar peculiar microglial morphologies with bigger soma and less protrusion have been detected in murine and vertebrate GCase deficient models induced by genetic modification 58,59 and in brain areas (such as substantia nigra) of neuropathic GD patients 60,61 , and were often associated with a proinflammatory phenotype. However, with our morpho-metric analysis, the use of dynamic descriptors clearly distinguished the effects of LPS and CBE treatments on microglia, showing that male CBE-treated cells were static and less contractile, a phenotype markedly different from the pro-inflammatory phenotype ( Fig. 2B and 4B), and characteristic of inactive microglia. This is in line with our previous expression data showing that no pro-inflammatory genes were induced by the CBE treatment in immortalized microglia 30 . The more stationary phenotype and the decreased number of protrusions suggest that CBE-treated microglia display a reduced contact surface with the neuron membranes, a condition likely contributing to the decreased Nrf2 expression in neurons 30 . The reduction of Nrf2 levels might increase the risk of neurodegeneration especially in neurons of the substantia nigra that are frequently exposed to oxidative stress due to dopamine metabolism 62 . In the case of GBA1 mutations, the microglial GCase inactivation is constitutive and over time could promote pathways leading to promotes neurodegeneration.
Interestingly, female microglia seem to be more affected by GCase impairment, indeed the morphofunctional phenotype is more divergent from the vehicle when compared to male microglia (Fig. 5): the CBE effect on female microglia increases the subpopulations characterized by a less active behavior (less ramified shape and static) to a greater extent compared to male microglial cells.

Pag 14
Moreover, female microglia also displayed a divergent response as compared to male microglia, indeed CBE treatment increased the subpopulations defined by the descriptors as "small & motile" and "complex & contractile"; to our knowledge, this is the first description that the inhibition of GCase is able to have a different effect on the morphology and motility of male and female microglia.
Intriguingly, the effect of GCase inhibition is more penetrant in female microglia, reducing the superior ability of female microglia to induce Nrf2 in neurons to the same extent found in male microglia (Fig. 6). This dramatic change in female microglia function is likely diminishing the greater neuroprotective ability of female microglia, rendering them comparable to male microglia.
These sex-related morpho-functional differences may have functional consequences: it is tempting to speculate that the increased neuroprotective ability of female microglia could contribute to the 1.5-2 fold reduced risk of developing idiopathic PD observed in female individuals, a sex-bias which appears to be reduced in GBA1-PD patients (Fig. 7) 63-67 . Indeed, the majority of studies report higher female prevalence in GBA-PD, or do not observe sex-related differences [64][65][66][67] suggesting that the protective effect associated with female sex is indeed blunted by GBA mutations 65 , although a firm explanation of this difference with idiopathic PD have not been reported. Our data suggest that the differential effects of these mutations on the microglial phenotype might contribute to the differences observed between idiopathic and GBA-PD in terms of loss of male predominance.
In conclusion we report a novel methodological approach toward the identification of dysfunctions of microglia in models of neurological diseases. The morpho-functional method was revealed to be sufficiently sensitive to recognize phenotypic differences in unstimulated microglia derived from the brain of male or female animals; moreover, the technique demonstrated the existence of discrete subpopulations of microglia, each characterized by specific morphological descriptors, indicative of different specific phenotypes and a differential response to specific stimuli. This novel perspective provides insight into the microglial heterogeneous behaviors that might underlie pathological Pag 15 stimuli in different CNS regions or in the function of sex and age 68 . Indeed, the identified morphofunctional parameters allowed us to describe the morphological changes induced not only by a wellknown pro-inflammatory agent (LPS), but also by the CBE model of reduced GCase activity and GBA1-PD. Our data, for the first time demonstrates that GCase inhibition triggers a specific microglial morpho-functional phenotype associated with a reduced ability of microglia to perform neuroprotective functions, with more dramatic consequences for microglia isolated from female animals: this finding might contribute to the understanding of the sex-related differences clinically observed in idiopathic PD.

Cell cultures
Primary neurons were derived from the cerebral cortex of p0-p1 mice following standard operational procedure using the neural tissue dissociation kit-postnatal neurons (Cat. 130-094-802, Miltenyi Biotec) as previously described 30 . In brief, the brain cortices from six mice were pooled as a single experimental group and subjected to enzymatic and mechanical dissociation, then 150,000 primary neuronal cells were seeded for each well of a poly-L-ornithine coated 24-well plate, replacing half of the medium volume every 2 or 3 days. At day ten, 37,500 primary microglia cells isolated from the whole brain of adult mice (age 3-6 months) were seeded on neuron layer; briefly, the brains from two mice were pooled and subjected to enzymatic and mechanical dissociation and Then the objects with an area greater or equal to 130 μm 2 , were subjected to the "analyze particle" function to calculate the "Area", "Center of mass", "Shape descriptors" and "Feret's diameter" for each object in each frame. The area was converted from pixel to the surface in μm 2 ; the coordinates of the center of mass of each object were used to calculate the distance covered by the cell during the time-lapses. Among the "Shape descriptors", we operated with the solidity, a value that corresponds to the area/convex area of the object; between the "Feret's diameter" values we used the "Feret Angle" to calculate the number of rotations of each object during the recording. An math operation was used to perform the clustering analysis: in brief, for each parameter obtained from the analysis the values of the vehicle and treated cells were used to identify the median parameter for the experiment, this median was used as a threshold to cluster the cells in two groups (over or under Pag 17 the median); the combination of two parameters was used to generate four different clusters (Table   1 and Supplementary Table 1).

Animals and treatments
The animals were fed ad libitum and housed in individually ventilated plastic cages within a temperature range of 22-25 °C under a relative humidity of 50% ± 10% and an automatic cycle of 12 hours light/dark. C57BL/6 and CX3CR1 +/ GFP mice were supplied by Charles River (Charles River Laboratories MGI Cat 2159769, RRID:MGI:2159769 and MGI:J:84544), ARE-luc2 mice were generated in our laboratory 47 . For pharmacological treatments, mice (15-30 weeks old) were administered 100 mg/kg/day CBE or vehicle (PBS) via i.p. injection for 3 days before the purification of microglia.

Luciferase enzymatic assay
Luciferase assays were performed as illustrated previously 48 in brief, microglia-neuron cultures were lysed with luciferase cell culture lysis reagent (Cat. E1531, Promega), and the protein concentration was determined with a Bradford assay 69 . The luciferase activity assay was carried out in luciferase assay buffer by measuring luminescence emission with a luminometer (Veritas, Turner) for 10 s to obtain the relative luminescence units (RLU).

Clinical data
Clinical data were obtained from the Accelerating Medicines Partnership Parkinson's Disease Participants with the following tags were included: "PD", "Genetic Registry PD", "Genetic Cohort PD", "Genetic Registry Unaffected", "Genetic Cohort Unaffected", "Healthy Control". Participants Pag 18 marked as "Prodromal" and "SWEDD" (scans without evidence for dopaminergic deficit) were excluded from the analysis.

Statistical analysis
For the cellular experiment statistical analyses was performed employing Prism 7 (Version 7.00, GraphPad Software Inc., RRID:SCR_002798, http://www.graphpad.com), multiple t-test versus vehicle were used to determine if there were significant differences in means and a p-value lower than 0.05 was considered to indicate statistical significance. For the clinical data statistical analysis was carried out using R (version 4.2.1, RRID:SCR_001905, http://www.r-project.org). Pearson's Chi-squared test was used to compare sex differences between carriers and non-carriers of GBA variants.

Data availability
The data that support the findings of this study are available at DOI 10.5281/zenodo.7360295.