Age-Dependent Sex Differences in Perineuronal Nets in an APP Mouse Model of Alzheimer’s Disease Are Brain Region-Specific

Our results show that the density of PNNs in wild-type and APP^NL-F/NL-F mice is region-specific (Figure 1). The highest levels of PNNs were shown in the LEC and PRS regions compared to CA1 in both age groups studied, 2–5 months wild-type (F_2,24 = 131.2; p < 0.0001) and APP mice (F_2,24 = 120.8; p < 0.0001), as well as 12–16 months old wild-type (F_2,24 = 83.90; p < 0.0001) and APP^NL-F/NL-F mice (F_2,24 = 79.94; p < 0.0001) (Figure 1).

The density of PNNs was observed to be significantly higher in the older cohorts studied for both sexes of APP^NL-F/NL-F mice. For example, when comparing the abundance of PNNs between the two age windows used from APP^NL-F/NL-F mice, there was an approximately 2-fold difference in the PNN density in CA1, as well as an approximately 2-fold difference in PNN density in the LEC in the older male cohort (F_1,42 = 41.72; p < 0.0001), illustrating an age-dependent increase in the abundance of PNNs. This was consistent for the APP^NL-F/NL-F females, where there was an approximately 2-fold higher density of PNN in CA1 and LEC regions in 12–16-month-old APP females when compared to 2–5-month-old APP female mice (F_1,42 = 41.72; p < 0.0001). As for the wild-type mice, we noticed an insignificant higher level of PNNs related to age.

To further investigate the sex differences associated with PNN density, we focused on the three brain regions CA1 (Figure 2A), LEC (Figure 2D), and PRS (Figure 2G). Two-way ANOVA test performed showed no sex differences in the density of PNNs between males and females for both 2–5-month-old wild-type mice (F_2,24 = 1.138; p = 0.2967) and 2–5-month-old APP^NL-F/NL-F mice (F_1,24 = 1.937; p = 0.1768).

As for the role of sex in influencing the PNN density in the different brain regions of the older APP^NL-F/NL-F mouse model of the AD cohort, our data suggest a sex-specific alteration in PNN density in LEC (F_1,16 = 13.53; p = 0.0020) and PRS (F_1,16 = 10.35; p = 0.0054). Results from performing a two-way ANOVA showed ~1.3-fold higher PNN density in APP^NL-F/NL-F females when compared to their age-matched APP^NL-F/NL-F male counterparts in both LEC and PRS (shown in Figure 2F,I).

Analysis of CA1 from immunoperoxidase labelling did not reveal a statistical sex difference in the abundance of the PNNs; therefore, to validate these results further, z-stack images from immunofluorescence labelling (that provided a better depth of tissue area analysed) were analysed. The data from z-stack confocal images revealed an approximately 2-fold higher PNN density in APP^NL-F/NL-F males compared to age-matched APP^NL-F/NL-F females (F_1,16 = 8.3; p = 0.0106). Figure 3D shows a significant difference in PNN density when comparing older wild-type males and females to the age-matched APP^NL-F/NL-F mice (F_1,16 = 44; p < 0.0001), of which neuroanatomical details can be found in Table 1. Therefore, our results from immunoperoxidase and immunofluorescence analysis show sex-dependent alteration in PNN density during the late stages of AD that are region dependent.

For the older cohort representing the late stage of AD (12–16 months) in CA1 and LEC (Figure 2C and Figure 2F, respectively), our data show a higher PNN density in APP^NL-F/NL-F mice when compared to age-matched wild-type mice. In the CA1 region, results from performing a two-way ANOVA test revealed an approximately 2-fold higher level in PNN density when comparing 12–16-month-old APP^NL-F/NL-F male mice to the age-matched wild-type male mice (Figure 2C), and an approximately 2-fold higher level in PNN density when comparing 12–16-month-old APP^NL-F/NL-F females to age-matched wild-type females in the same region (F_1,16 = 45.55; p < 0.0001) (Figure 2C). This was consistent with the LEC region, where we found an approximately 2-fold higher level in PNN density for the males and a 2- fold higher abundance of PNNs for the females when comparing old APP^NL-F/NL-F mice to age-matched wild-type mice (F_1,16 = 96.24; p < 0.0001) (Figure 2F). No significant differences were found between wild-type and APP^NL-F/NL-F mice in the PRS region in the 12–16-month-old mice (F_1,16 = 2.253, p = 0.1528) (Figure 2I), suggesting that this region is not affected by AD.

It has been suggested that sex differences play a role in the depletion of PNNs due to the activation of microglia in AD; therefore, we investigated whether PNN density was associated with regions of activated microglia and whether these components were in close proximity to dense Aβ plaques in the CA1 (Figure 3) of 12–16-month-old male and female APP^NL-F/NL-F age-matched to wild-type mice. Microglia regulate PNN components, including CSPGs, which are bound and labelled by WFA, while CD68⁺ was used as a marker of activated microglia. As mentioned, in the CA1, we observed an approximately 2-fold higher density of WFA+ PNN in APP^NL-F/NL-F males compared to APP^NL-F/NL-F females (F_1,16 = 8.3; p = 0.0106). We also found a significantly higher density of PNNs when comparing wild-type to age-matched APP^NL-F/NL-F mice for both males and females, which was quantified as an approximately 4-fold higher density for males and an approximately 3-fold higher density for females (F_1,16 = 44; p < 0.0001) (Figure 3D). This correlated with significantly higher density of Aβ plaques for the APP^NL-F/NL-F mice compared to their age-matched wild-type mice, with a 14-fold increase in Aβ density for males and a 9-fold increase in density for the females F_1,16 = 52.79, p < 0.0005) (Figure 3B). As for the CD68⁺ microglia quantification (Figure 3C), we found an approximately 2-fold higher CD68⁺ microglia density for the males and an approximately 3-fold higher density for the females when comparing 12–16-month-old wild-type mice to age-matched APP^NL-F/NL-F mice (F_1,16 = 85.05; p < 0.0001); however, no sex-related differences were observed when quantifying the Aβ plaques (F_1,16 = 0.07330; p = 0.790), and the CD68⁺ microglia (F_1,16 = 0.1660; p = 0.6891) in the CA1 region (Figure 3B,C). All comparison values are presented in Table 2.

To further investigate any association of sex differences in microglia in AD, we performed a WGCNA, which showed one microglial module (yellow) with significant correlations with the traits tissue, age, genotype, and sex (Figure 4A), indicating that the expression of the genes in this module is affected by changes in the named traits. The grey module represents unassigned genes and is disregarded. No genes listed in the perineuronal net term list GO:0072534 were present in this module. We performed gene enrichment analysis at FDR-corrected p < 0.05, which showed terms relating to infection and immune response as well as phagocytosis (Figure 4B).

The density of PNNs varied according to the different brain regions. This is in accordance with previous findings that suggested differences in the distribution patterns of PNNs in the brain [27], as well as with more recent findings that showed that PNNs are not uniformly expressed in all brain regions [28]. These differences are believed to be associated with the diverse functionalities that are exhibited by PNNs in the brain, which encompass their neuroprotective role and their role in regulating synaptic plasticity, signalling, and memory, which are all implicated in distinct neurological diseases [4]. Our results demonstrated that the regions with the highest abundance of PNNs were the LEC and PRS in both the wild-type and APP^NL-F/NL-F cohorts from both age windows. These findings correlate with the data from [27], where the PRS showed the highest frequency of labelled PNNs. Our results also showed no significant difference in the PRS regions of the 2–5-month-old and 12–16-month-old wild-type mice when compared to their respective age-matched APP^NL-F/NL-F mice. However, we did observe a difference in the abundance pattern in older females in our AD model. The lack of change in the males corroborates the hypothesis that the PRS is devoid of plaques with dystrophic neurites in early-onset, late-onset sporadic, and familial AD [13,14,15,16], and in males of mouse models of AD [29], suggesting preservation of this region during the pathogenesis of AD. However, there is a lack of comparison made with data from tissue from females, and our data highlights this underrepresentation. Thus, our results reveal brain-region-specific differences in PNN density; however, to obtain a clearer understanding of these structures, further investigations will be required to determine the underlying cause of these differences in PNN density in these brain regions and how that can affect AD pathogenesis.

The present study suggests that the density of PNNs is age dependent, with the levels of PNN density increasing with age in both genotypes studied in several brain regions, which is consistent with previous studies performed in post-mortem brain tissue [30]. Interestingly, the LEC and CA1 regions are the first regions to be affected during early AD [31] and were the regions that showed significantly altered density of PNNs in later stages of AD, which corroborates the results of previous studies [32,33].

Overall, we observe a greater density of PNNs in our AD mouse model at the late stages of AD compared to the age-matched control mice. This change could be suggestive of the theory of a “compensatory protective mechanism” to prevent neurodegeneration and damage caused by the progression of AD [34] as a result of the accumulation of amyloid beta plaques [11]. Others have also suggested that PNNs have a neuroprotective role [1,5,35], which is further supported by studies that found elevated levels of hyaluronic acid, CPSGs, as well as tenascin proteins in AD models [35]. These are components of the ECM structure of PNNs [5,8,24]. Therefore, an increase in their levels could essentially be linked to an increase in the formation of PNNs.

The present study demonstrates that sex is a factor affecting PNN density in different brain regions including the CA1, LEC, and PRS. This sexual dimorphism in PNN density may be connected to changes and differences in hormone levels between males and females and a possible connection between PNNs and oestrogens, which is understudied. This corroborates a previous study that showed that PNN-associated neurons in the mPOA (medial preoptic area) express oestrogen receptor α, which is suggested to influence the signalling mechanism affecting the assembly of the PNNs during the reproductive cycle in rodents [36].

Evidence also suggests a high density of oestrogen receptors in the hippocampus [37], and interestingly, they are colocalised extensively with parvalbumin (PV) cells in various brain regions including the hippocampus [38]. This suggests a potential connection of oestrogen with PNNs that encapsulate PV cells, and a mechanism whereby oestrogen can regulate neuronal excitability via inhibitory interneurons, which aligns with previous studies that hypothesise that oestrogen receptors influence hippocampal synaptic plasticity and neural signalling, which may affect cognition and memory [39].

Furthermore, oestrogens are thought to have several roles in AD [23], including neuroprotection and influencing the regulation of amyloid beta pathology by reducing the levels of Aβ and preventing its accumulation [40], which are functions comparable to those of PNNs [5]. Therefore, it may be plausible that males require a higher abundance of PNNs in brain regions such as the CA1 to obtain neuroprotection against factors such as oxidative stress, which they are more susceptible to (Griffiths et al., 2019) [20], as they do not have the potential extra protection that can be provided by the oestrogen hormones. The lack of experimental connection between PNNs and oestrogens is a limitation of the current study; it needs to be investigated further and could potentially prove a novel approach to understanding the sex-related differences in AD prevalence.

Microglia have a complex association with PNNs. They are involved in the regulation of PNN density in the normal brain through the remodelling of the ECM structural components, including CSPGs, tenascins, and hyaluronan and proteoglycan link proteins [41,42]. They also contribute to the upregulation of these components during neuronal injury [43,44]. However, microglia are not the primary facilitators of this role, as during their depletion, PNN density increases [44,45,46,47]. Thus, the microglial function in the turnover of PNNs appears to be more focused on degradation. Activated response microglia (ARM) have been implicated in the extensive loss of PNNs in AD as PNN components have been observed within microglia both in the mouse model and in humans [18].

Many recent studies have investigated the sex-related differences affecting the abundance and activation states of microglia, from development to maturation, under various neuroinflammatory conditions [48,49,50]. Microglial cells isolated from female mouse models of AD shift from a homeostatic state towards an activated response (ARM) state faster than age-matched male mice [51]. This change involves more glycolytic activity and impaired phagocytosis in female mice, which causes reduced amyloid clearance in comparison with males. The sex-related changes in microglial morphology and increased Aβ plaque density were consistent in post-mortem cortical tissue from individuals with AD and age-matched controls [52]. Our results in mice are consistent with these findings and show higher levels of dense Aβ plaques in aged female mice. We observed no sex-related differences in CD68+ microglia; however, morphological changes were not investigated. Due to sex differences affecting microglial function in disease, we performed WGCNA on a publicly available dataset to assess associations with microglial genes and sex as a trait.

Our WGCNA also revealed a module of genes with a weak correlation with sex, meaning that their expression is associated with differences between sexes. Annotation of this set of genes using the KEGG pathway showed an enrichment for Fc-gamma receptor-mediated phagocytosis. Fc-gamma receptors (FcγRs) are receptors on the surface of cells that elicit immune responses including phagocytosis. The expression of FcγRs on microglial cells is reported to be increased during ageing and AD [53,54,55], but how these receptors are associated with microglial clearance of Aβ plaques and PNNs warrants further investigation.

All the procedures in this study were carried out in accordance with the British Home Office regulations under the Animal Scientific Procedure Act 1986, under the project licence (awarded in March 2018) PPL: P1ADA633A held by the principal investigator, Prof. Afia Ali. All procedures were approved by both internal and external UCL ethics committees, and in accordance with the ARRIVE guidelines for reporting experiments involving animals [56]. A total of 50 animals (disease model and wild-type) were used in this study. The animals had ad libitum access to food and water and were reared in cages with a maximum of five inhabitants, with a day:night cycle of 12 h:12 h.

The knock-in APP^NL-F/NL-F AD mouse model was used for this study [57]. This mouse model was chosen because it faithfully reproduces the effect of AD Aβ pathology without overexpression artefacts in a time-dependent manner. The APP^NL-F/NL-F model has two familial AD (FAD) mutations: KM670/671NL (Swedish) and I716F (Iberian). The former increases β-site cleavage of APP to produce elevated amounts of both Aβ40 and Aβ42, whereas the latter promotes γ-site cleavage at C-terminal position 42, thereby increasing the Aβ42/Aβ40 ratio in favour of the more hydrophobic Aβ42, as seen in clinical AD. Thus, the APP^NL-F/NL-F mouse model shows Aβ accumulation and related pathology in an age-dependent manner, with initial accumulation shown at 6 months [57]. The knock-in line was crossed with C57BL/6 mice, and the resulting heterozygous pairs were used for breeding but excluded from experiments. Only male and female APP^NL-F/NL-F and age-matched wild-type (C57BL/6) mice, aged between 12 to 16 months, were included.

Animals were genotyped via standard polymerase chain reaction using the following four primers: 5′-ATCTCGGAAGTGAAGATG-3′, 5′-TGTAGATGAGAACTTAAC-3′, 5′-ATCTCGGAAGTGAATCTA-3′, and 5′-CGTATAATGTATGCTATACGAAG-3′, as previously described [57].

To provide further assessment and in-depth quantification of PNNs in the CA1 region of the hippocampal regions, confocal microscopy analysing Z-stack immunofluorescence images was performed. From each brain section, Z-stacks at ×20 objective were taken using the Zeiss LSM880 confocal microscope (Zeiss, Baden-Württemberg, Germany) in unison with the Zeiss Zen Black imaging software from the CA1. The regions of interest (ROI), including CA1 (including stratum oriens, stratum pyramidale, and stratum lacunosum), were located using the manual joystick through the ×20 objective lens by systematically searching the slice and consistently evaluating the location with reference to an appropriate mouse atlas. Z stack images were taken at a resolution of 1024 × 1024 pixels with 12–14 Z steps through the depth of the slice and with application of appropriate filters to complement secondary antibody fluorescence: DAPI (405 nm), Alexa 488 (488 nm), Alexa 555 (555/565 nm).

Image analysis was performed with ImageJ software (1.5.3 version) using an automated macro specific to each quantification. The Z-stack images were processed at maximum intensity projection and split into the constituent colour channels. Following this, the IsoData auto thresholding method was used to demarcate signal from background and produce the ROI (Huang and Wang, 1994). The Analyse Particles function was then used (PNN size = size = 100.00–infinity, dense Aβ plaques = size = 50.00–infinity, and CD68+ microglia = size = 0.00–infinity). PNN and CD68+ structures were calculated using % area, which divides the area of all the particles that meet the minimum size limit by the total area of the image field [60,61]. To compare dense plaque accumulation, average area was used as the measure for Aβ structures, which gave the average area of the dense plaques within the field.

Weighted gene co-expression network analysis [62] was performed on cortical and hippocampal microglia isolated from 12- and 21-month-old age-matched male and female APP^NL-G-F and C57Bl/6 wild-type mice accessed from a publicly available dataset GEO: GSE127892↗ [51]. Sala Frigerio et al. [51] analysed over 10,000 microglial cells at four different timepoints; however, in this study we focused on the two later stages. The raw data were processed and normalised using the variance-stabilising transformation method from the DESeq2 package [63]. The WGCNA blockwiseModules function was performed on the normalised counts at a soft-threshold power of β = 3 with a minimum module size of 50 to identify networks correlated with traits of interest: tissue, age, genotype, and sex. ShinyGO 0.77 was then used to perform gene enrichment analysis on the modules of interest using the KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway database [64,65,66].

The statistical analysis was performed using GraphPad Prism version 9.0 for Windows and Microsoft Excel. Based on the differences observed between control and diseased data sets obtained in our preliminary studies, an n ≥ 5 was ideal for this study in order to reveal a statistical difference of >80% power, assuming a 5% significance level and a two-sided test for neuroanatomy experiments.

Various statistical tests were performed depending on the parameters used, and each figure legend details the specific statistical test used. Before performing any statistical test, the normality of the raw data was verified using the Shapiro–Wilk test and a ROUT test to identify potential outliers. In almost all of the cases, no outliers were identified. Where an outlier existed, it was removed from the data pool via the software. When comparing between three or more data sets from two genotypes, a one-way analysis of variance (ANOVA) was performed. When a second factor was taken into account, for example age, a two-way ANOVA was used. After any ANOVA, a post hoc test for multiple comparisons was applied. All p-values below 0.05 were considered significant, and asterisks were added to the presentation of the data as follows: * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001.

To assess the significance of the differences between groups, F-values were also taken into consideration. There are displayed as (F values_(dfn,dfd); p-value), where dfn stands for degrees of freedom numerator, and dfd stands for degrees of freedom denominator. As for the interpretation, larger F values were considered significant, whereas smaller F values were considered non-significant.

All figures displaying error bars represent ± the standard error of the mean. The “n” is given as the number of observations and is equal to the number of animals used, unless otherwise stated.