Age-dependent and region-specific alteration of parvalbumin neurons, perineuronal nets and microglia in the mouse prefrontal cortex and hippocampus following obesogenic diet consumption

Emergent evidence demonstrates that excessive consumption of high fat and high sugar (HFHS) diets has negative consequences on hippocampal and prefrontal cortex (PFC) function. Moreover, the delayed maturation of the PFC including the late development of parvalbumin-expressing (PV) interneurons and perineuronal nets (PNNs) may promote vulnerability to HFHS diet-induced nutritional stress. However, the young brain may have some resistance to diet-induced neuroinflammation. Thus, we examined the impact of a HFHS diet commencing either in adolescence or adulthood in male mice. PV interneurons, PNNs and microglia were assessed using immunohistochemistry. We observed greater numbers of PV neurons and PNNs in the hippocampus and the prelimbic and infralimbic PFC in adult mice in comparison to our younger cohort. Mice that consumed HFHS diet as adults had reduced numbers of hippocampal PV neurons and PNNs, which correlated with adiposity. However, we saw no effects of diet on PV and PNNs in the PFC. HFHS diet increased microgliosis in the adult cohort, and morphological changes to microglia were observed in the PFC and hippocampus of the adolescent cohort, with a shift to activated microglia phenotypes. Taken together, these findings demonstrate different regional and age-specific effects of obesogenic diets on PV neurons, PNNs and microglia.

Age effects were observed on the colocalisation of PV+ neurons and PNNs (Fig. 3  We examined whether associations were present between WAT mass and hippocampal regional PV and PNN markers in each age group using linear regressions. We found a significant correlation between WAT mass and PV+ neurons in the CA2/3 of the adolescent, but not adult cohort ( HFHS diet consumption changed PFC microglia density and morphology. Microglia undergo extensive morphological restructuring in response to exogenous and endogenous environmental events that may underlie central nervous system dysfunction 52 . Microglia were immunostained with ionized calcium binding adaptor molecule-1 (IBA-1) and fluorescent intensity and morphology was assessed. Overall, adult mice had greater PFC microglial density compared to adolescents (Fig. 4A Microglia can be visually classified using morphological characteristics that align to functional phenotypes 47 . As such, microglia were classified into one of three morphological states: (a) ramified microglia were defined Mice that consumed HFHS diet across adolescence had proportionally fewer (C,D) ramified 'resting' microglia in all three regions and more hypertrophied 'active' microglia in the PrL and OFC. In contrast, no diet-related changes (C,E) in PFC microglia morphology was seen in mice that consumed HFHS diet as adults. Data were assessed using two-way ANOVAs and *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Representative photomicrographs were taken at 40X magnification (green-IBA-1 + microglia), scale bars show 50 μm for IBA-1 + density and 10 μm for IBA-1 + morphology. www.nature.com/scientificreports/ as having multiple radially projecting processes and an oval cell body, (b) hypertrophied microglia, which were classified by their enlarged darkened soma and shorter, thicker, and less branched processes, and (c) amoeboid microglia, where are defined as having densely stained, enlarged cell bodies with few short, if any, processes (Fig. 4C). Hypertrophied and amoeboid morphologies reflect activated states often seen in obesity 53 , whereas ramification reflects a resting, or surveillance state. Using these classifications, we observed significant effects of HFHS diet consumed across adolescence on the morphology of microglia (diet × morphology-PrL: F (2,28) = 13.94, p < 0.0001; IL: F(2,28) = 15.51, p < 0.0001; OFC: F(2,28) = 23.07, p < 0.0001, Fig. 4D). Post hoc analyses revealed that mice that consumed HFHS diet across adolescence had proportionally fewer ramified 'resting' microglia (PrL p = 0.0002, IL p < 0.0001, OFC p = 0.0001), and more hypertrophied 'active' microglia (PrL p = 0.021, OFC p = 0.008) than controls. In contrast, no diet differences in PFC microglia morphology was seen in in adult mice (diet × morphology PrL F(2, 28) = 0.496, p = 0.615; IL F(2, 28) = 0.31, p = 0.736; OFC F(2, 28) = 1.478, p = 0.245, Fig. 4E).
We examined whether associations were present between WAT mass and prefrontal microglial density and morphology in each age group using linear regressions. We found significant correlations between IBA-1 density and WAT mass in the adult cohorts in the PrL ( We examined whether correlations were present between WAT mass and hippocampal IBA-1 fluorescence or hypertrophied in each age group using linear regressions. We found significant correlations between IBA-1 fluorescence and WAT mass in the adult and adolescent cohorts in the CA1 (Fig. S4A  Furthermore, we found a significant correlation between the hypertrophied 'active' morphology and WAT mass in the CA1 in the adult but not the adolescent cohort (

IBA-1 fluorescence in the PrL PFC was associated with decreased numbers of PV neurons and
PNNs in adult mice. We examined whether associations were present between IBA-1 fluorescence and PFC PV and PNN markers in each age group using linear regressions. We found a significant correlation between IBA-1 fluorescence and PV neuron counts in the PrL in the adults but not adolescent cohort ( In contrast, no diet-related changes (C,E) in microglia morphology were seen in mice that consumed HFHS diet as adults. Data were assessed using two-way ANOVAs and *p < 0.05, **p < 0.01, ***p < 0.001. Representative photomicrographs were taken at 40X magnification (green-IBA-1 + microglia), scale bars show 50 μm for IBA-1 + density and 10 μm for microglia morphology.

Discussion
Here, we show that consumption of a HFHS diet for 5 weeks commencing in either adolescence (P26) or adulthood (P68) had age-and region-specific effects on PV neurons, PNNs and microglia within regions of the murine PFC and hippocampus. Moreover, our immunohistochemical examination demonstrated that populations of PV neurons and PNNs increased with age in the mPFC, but did not change in the hippocampus. Our observation that PV-expressing neurons and their surrounding PNNs have delayed developmental trajectories in the medial PFC, but not the hippocampus, supports previous studies that showed PNN development in the medial PFC extends through adolescence 17,54-56 . We observed increased PV and PNN counts in the PrL and IL PFC in the adult cohort compared to the adolescent cohort of mice. Our results contrast aspects of previous studies that found significant increases in PV and PNN counts in the PrL and IL PFC between juvenile (P24) and adolescent rats (P35), but no significant differences between adolescent and adult rats (P70) 17 . However, the increases in PV and PNNs in the PrL and IL with increased age observed in our experiment indicate that there is a more enduring age-dependent refinement within the mouse medial PFC, which may extend further into adulthood than previously observed 17 . Moreover, we extended our histological examination to encompass the OFC, another frontal cortical region that is involved in behavioural regulation, where we did not see differences in PV and PNN counts between the age groups, which indicates that populations of PV neurons and PNNs in the OFC reach maturity before the PrL and IL PFC.
An explanation for the increase in PV immunoreactivity between our age groups of mice may be a developmental increase in the expression of PV-protein in fast-spiking interneurons, which is reflected by increased PV-immunoreactivity and mRNA during the process of brain maturation 21,22 . Of note, we also found that colocalisation of PNNs with PV neurons in the IL PFC was greater in adult than adolescent animals. This indicates that the IL PFC likely has the latest critical window of maturation in mice, as the envelopment of PV neurons with PNNs is widely acknowledged to be a marker of the closure of critical windows of plasticity in neurodevelopment. Although the exact molecular orchestration of the development of the PFC is still largely unknown, behavioural studies functionally demonstrate that the IL PFC is implicated in mood regulation and fear memory extinction, behaviours that typically develop in late adolescence 57,58 . This functional observation combined with our cytoarchitectural findings indicates that the IL PFC is likely one of the last frontal cortical regions to reach maturity. Studies that have examined human post-mortem tissue samples of the frontal cortices and hippocampus have also shown that PV neurons are present from birth, although at low levels, and increase to peak levels at 2 years of age 59 . PNNs begin to form as early as the second month of life but do not reach a mature appearance until around 8 years of age 59 . Interestingly, longitudinal proteomic and transcriptomic data from human dorsolateral PFC samples show a dynamic reduction in the expression of neurocan, a lectican that forms glycosaminoglyan side chains of PNNs after birth 60 . Aggrecan, another lectican, increases in expression across development and into adulthood, suggesting that it is continuously used in the remodelling of PNNs 59 .
The envelopment of PV neurons with PNNs not only marks the closure of developmental windows of heightened neuroplasticity, but also provides a protective microenvironment around neurons 37 . Neurons without PNNs have increased sensitivity to environmental and endogenous toxins, which can lead to functional dysregulation 33,34,36 . The biomolecular mechanisms underpinning the endogenous modulation of PNN integrity are yet to be fully elucidated. Possible mediators include extracellular proteolytic enzymes called matrix metalloproteinases (MMPs) that are secreted by neurons and glia. Activated MMPs cleave structural components of PNNs, and so permit the dynamic remodelling of synapses and alterations to neuronal plasticity 48,61 . However, the upregulation of MMP activity is seen in states of neuroinflammation and neurodegenerative diseases 49 , where MMPs secreted by microglia could aberrantly cleave PNNs and expose neurons to toxins.
Interestingly, we did not see effects of HFHS diet on PV neuron immunoreactivity, PNNs or their colocalisations in regions of the PFC. Previous studies have shown that intermittent daily access to a HFHS diet in rats, which causes binge-like behaviours, reduced PV neuron immunoreactivity and PNNs in the PFC 24,26 , and high fat diets (> 60% fat) have been related to reduced PNNs in the PrL and orbitofrontal cortex in male rats 24 . This finding suggests that diet formulations and administration protocols may have differential effects on PFC PV and PNNs, and also supports studies that have observed distinct cross-species expression patterns of PNNs between mice and rats 62 .
In contrast to the PFC, we saw deleterious effects of HFHS diet on populations of PV neurons and PNNs in the hippocampus, particularly in adult mice. Diet-associated reductions of PV neuron immunoreactivity and PNNs were observed within the CA1 of the adult cohort, suggesting that this region may be more vulnerable to nutritional stress. This aligns with previous observations of pronounced obesity-induced cognitive deficits in spatial memory 14,15,63,64 , as PV-mediated inhibitory neurotransmission within the CA1 is required for the stabilisation of place cells 65 . The most prominent diet-evoked changes to PV neurons and PNNs were observed in the adult cohort, supported further by negative associations between adiposity and numbers of PNNs within the CA1 and CA2/3 in the adult cohort. Thus, PV neurons and PNNs in the younger brain may have some resilience to HFHS diet-induced dysregulation within the hippocampus. This may be because of increased remodelling of PNNs during early life which would allow for the protective microenvironment to be repaired more efficiently.
Our observation of changes to PNNs and PV neuron populations in the hippocampus align with studies showing that hippocampal function is vulnerable to the negative functional consequences of HFHS diet and obesity 8,12,[14][15][16] . Obesity is associated with an increase in circulating inflammatory cytokines, which are generated by enlarged and increased populations of adipocytes in white adipose tissue. Increased inflammation leads to disruption of tight junction proteins such as claudin-5 and occludin within the blood brain barrier (BBB) 66 . The increased permeability of the BBB allows the influx of leukocytes, and engages microglia to produce proinflammatory cytokines 67 . These mechanisms are proposed to underlie both the onset of microgliosis and have further downstream damage to neuronal function. www.nature.com/scientificreports/ Mice that consumed HFHS as adults showed increased microgliosis in the PrL PFC and all hippocampal subregions measured by IBA-1 fluorescence compared to the adolescent cohort, which was also associated with increased adiposity. Given that microglia respond to cytokines, this observation suggests that there may be some resistance to HFHS-induced neuroinflammation in younger animals, which has been previously observed in rats 44 . However, when we examined morphological changes in microglia-from resting (ramified) to active (hypertrophied) states-we found a significant reduction in ramified microglia and related increases in hypertrophied microglia within the PFC of mice exposed to the HFHS-diet during adolescence, but not as adults, which was positively correlated with adiposity. The switch between microglial ramification to reactive phenotypes in the PFC is associated with alterations in neuronal activity 74 , and may be linked to the more pronounced functional changes in PFC-mediated cognitive tasks observed in mice fed obesogenic diets across adolescence 6 . We therefore suggest that the obesity-evoked activation of microglia could provide a neuroprotective mechanism in the adolescent brain; however, this may have enduring functional consequences on neuroplasticity. Ramified microglia are involved in the pruning of excess synapses during neurodevelopment 75,76 and control of the neurogenic niche within the dentate gyrus 77 . As such, the activation of microglia by neuroinflammation may divert their role from the refinement of neuronal interactions and plasticity, leading to a derailment of neurodevelopmental trajectories.
However, the link between microgliosis and damage to or dysfunction of PV neurons is complex. We did not observe relationships between IBA-1 fluorescence and PV neurons/PNNs in the hippocampus, but did see a negative association between PV neurons/PNNs and IBA-1 fluorescence in the PrL PFC of adult mice. When microglia become activated, they can release more cytokines and can exacerbate neuroinflammation 69 . Due to their high energy demands, PV neurons are particularly susceptible to dysfunction under inflammatory or oxidative conditions [70][71][72][73] . Microglia activation is one indicator of a neuroinflammatory environment 45 , however further studies are needed to examine whether other chemical markers of inflammation or oxidative stress were increased.
Whether the hippocampus is vulnerable to nutritional stress in young mice is less clear-with previous studies showing exacerbation of memory deficits and neuropathological changes seen in diet administration from an early age 13,78 , but other studies in adult rats and mice showing profound and rapid HFHS-dependent memory deficits 14,15 . Developmentally, the maturation of the hippocampus is complete before the PFC, and in keeping with this developmental pattern we did not observe differences between PV neurons or PNN populations within dorsal hippocampal subregions of the two age groups. However, we did see significant increases in PV/PNN colocalisation between the two age groups within the CA1 and CA2/3, indicating that the refinement of PV-mediated plasticity in these region may continue into early adulthood 79 . Furthermore, the dentate gyrus is a key region of adult hippocampal neurogenesis, the disruption of which can lead to pronounced memory impairment 80 . It has previously been shown that high fat or high sugar diets decreased histological markers of neuroproliferation in the dentate gyrus 63,81,82 , indicating that even relatively small shifts in cellular populations in the dentate gyrus can have a profound impact on hippocampal function. Microglia activation is known to have a negative impact on hippocampal neurogenesis 77 which may provide another mechanism by which hippocampal function is disrupted by HFHS diet consumption 82 .
In summary, these data demonstrate that HFHS-diet leads to disturbances to PV neurons, PNNs and microglia in an age-and regionally distinct manner, with effects on PV neuron and PNNs numbers being more pronounced within the adult hippocampus, and some resilience to microgliosis in the hippocampus of adolescent mice. Moreover, we observed protracted changes in PV cell density, PNNs and microglia in the PrL and IL PFC outside of the expected peri-adolescent period, suggesting that these cellular populations continue to develop in these regions. Further studies should investigate the sex-dependent effects of HFHS diets across development, whereby notable sex differences in the PFC has been observed in microglia morphology following immune challenges 83 , PNN numbers 84 , and in adult rats fed high fat diets 27 . While changes in PNNs and PV neurons have been reported in animal models and humans with neuropsychiatric conditions 54 , the functional significance of these changes remain to be delineated with respect to dietary challenges-including obesity. Further studies utilising proteomics as well as whole-transcriptome sequencing will provide more in-depth molecular characterisation of how brain regions are affected by different nutritional states and help pin-point developmental windows of vulnerability. Moreover, measurement of MMP activity in the brain will provide insight into how in obesity could be altering microglia activity and the integrity of PNNs. Critically, a better understanding of the functional relationship between microglia, neurons and PNNs during sensitive windows of neurodevelopment such as adolescence, during ageing and in neuropathological states will provide innovative therapeutic strategies to prevent cognitive decline. Diet administration. Mice were habituated to the housing conditions for 5 days before being allocated to their diet manipulation groups. Diets administration began when mice were either 26 days old (which encompasses adolescence 85 ) or at 68 days old (adult) as shown in Fig. 1A. Mice in the control diet condition had ad libitum access to rodent chow (Teklad Envigo 7913, 18% protein rodent diet, 3.1 kcal/g, 5% fat, 5% fibre). Brains were extracted and stored in 4% PFA for 24 h at 4 °C. After this, they were transferred to 20% sucrose in PBS for 24 h at 4 °C, before being frozen and sliced to 30 μm coronal sections using a cryostat (Leica CM1950S). Four serially adjacent sections were obtained from each brain and stored in cryoprotectant (50% PBS, 25% ethylene glycol, 25% glycerol) at -20 °C. Dorsal hippocampal slices (bregma -1.6 to -2.4 mm) and frontal cortical slices containing prelimbic, infralimbic and orbitofrontal PFC (bregma + 2.3 to + 1.8 mm) were identified using a mouse brain atlas 86 and selected based on their DAPI nuclear staining pattern. IBA-1 staining. IBA-1 immunoreactivity was used an indication of microglia activation 87 . Following blocking and permeabilisation, free floating sections were incubated for 48 h at 4 °C with constant agitation in rabbit anti-IBA-1 antibody (1:2000, Abcam, ab178846) in 2% NGS and 0.1% BSA in PBS-T, followed by 3 PBS washes and a 2 h room temperature incubation goat anti-rabbit AlexaFluor 488 secondary antibody (1:500, Invitrogen, A-11008). After three washes, sections were then mounted to slides and coverslipped as described above.

Immunohistochemistry. PV and PNN staining.
Imaging. Immunostained sections were imaged using an EVOS FL digital cell imaging system (Thermofisher Scientific). Tiled images of three unilateral dorsal hippocampal sections and PFC sections per mouse per stain were obtained using a 40X dry objective and 405 nm, 470 nm or 585 nm LED light cubes (depending on the secondary antibody used for the staining protocol described above). The frontal cortex (bregma + 2.3 to + 1.8 mm) was divided into three subfields: Prelimbic (PrL) PFC, infralimbic (IL) PFC and orbitofrontal cortex (OFC) and the dorsal hippocampus (bregma -1.6 to -2.4 mm) was divided into the three subfields: CA1, CA2/3; and DG ( Fig. 1F,G).
From the images, PNN counts, PV cell counts, and colocalised PV cells with PNNs were counted using Imagebased Tool for Counting Nuclei plugin (Centre for Bio-image Informatics, UC Santa Barbara, CA, USA) for NIH ImageJ software (https ://image j.nih.gov/ij/). In the frontal cortices, counts were made from 500 μm × 500 μm areas taken from each subregion. In the dorsal hippocampus, subregional cell counts from the stratum pyramidal, stratum oriens and stratum radiatum of the CA1 and CA2/3, and the molecular, granular and polymorphic layers of the dentate gyrus, including the hilus were made based on anatomical characteristics referenced from The Mouse Brain in Stereotaxic Coordinates Atlas 86 .
For IBA-1 staining, 3 areas (240 μm × 190 μm) from each frontal cortex and hippocampal subregion were captured using 10 sequential z-stacks (1.2 μm) at maximum projection at 40X objective. ImageJ software was used to calculate fluorescence intensity of the IBA-1 stain, expressed as arbitrary units (AU) as has been previously published in diet studies 82 . This was conducted as an objective measure of the mean area of the IBA-1 stain in specific brain regions. As microglia can change shape dynamically within different environments, cell counts do not necessarily provide information regarding activation. Microglia morphology was determined based on visual inspection of the distal arborisation and cell body shape of IBA-1 positive cells based on categories-ramified, hypertrophied or amoeboid microglia, phenotypes that have been previously characterised in multiple studies 47,71,72,88,89 . We used a workflow to visually identify IBA-1 positive microglia phenotypes according to defined morphological characteristics 90 . Ramified microglia were identified as cells with small circular cell bodies and multiple (> 3) thin, highly branched processes. Hypertrophied microglia (also called reactive or primed microglia) had larger cell bodies that were oval or elongated in shape with > 2 branched, wide and enlarged processes. Amoeboid microglia had large cell bodies that had at most two short, rod-like processes without any branching or were completely devoid of any branches 47,88,89 . We coded the slides and a blinded independent observer then quantified them.

Statistical analyses.
All image analysis and cell counts were conducted by an experimenter who was blind to age and diet group allocations. Data are presented as mean ± SEM. Statistical analyses were carried out in SPSS Software (IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY, USA) and GraphPad PRISM Software (Prism Software, version 8.0, Irvine, CA, USA) using repeated measures ANOVAs and two-way ANOVAs with Bonferroni or Fisher's LSD post hoc tests, and data were screened for homoscedasticity by Levene's test. Correlations were examined post hoc using one-tailed linear regressions and significance was set at p < 0.05.