An early enriched experience drives targeted microglial engulfment of miswired neural circuitry during a restricted postnatal period

Brain function is critically dependent on correct circuit assembly. Microglia are well‐known for their important roles in immunological defense and neural plasticity, but whether they can also mediate experience‐induced correction of miswired circuitry is unclear. Ten‐m3 knockout (KO) mice display a pronounced and stereotyped visuotopic mismapping of ipsilateral retinal inputs in their visual thalamus, providing a useful model to probe circuit correction mechanisms. Environmental enrichment (EE) commenced around birth, but not later in life, can drive a partial correction of the most mismapped retinal inputs in Ten‐m3 KO mice. Here, we assess whether enrichment unlocks the capacity for microglia to selectively engulf and remove miswired circuitry, and the timing of this effect. Expression of the microglial‐associated lysosomal protein CD68 showed a clear enrichment‐driven, spatially restricted change which had not commenced at postnatal day (P)18, was evident at P21, more robust at P25, and had ceased by P30. This was observed specifically at the corrective pruning site and was absent at a control site. An engulfment assay at the corrective pruning site in P25 mice showed EE‐driven microglial‐uptake of the mismapped axon terminals. This was temporally and spatially specific, as no enrichment‐driven microglial engulfment was seen in P18 KO mice, nor the control locus. The timecourse of the EE‐driven corrective pruning as determined anatomically, aligned with this pattern of microglia reactivity and engulfment. Collectively, these findings show experience can drive targeted microglial engulfment of miswired neural circuitry during a restricted postnatal window. This may have important therapeutic implications for neurodevelopmental conditions involving aberrant neural connectivity.


| INTRODUCTION
Neurodevelopmental conditions, which can arise due to alterations in the pattern of neural connectivity, are pervasive and largely lack effective treatment.Microglia are becoming increasingly recognized for their multifaceted roles in neural development, disease, and function.The potential for microglia to correct aberrant neural wiring, however, is not well characterized.Recent work in mice suggests that early life experience could unlock a capacity for microglia to drive repair and rescue function of miswired circuits.
Correct assembly of neural circuits is vital for effective nervous system function.In Ten-m3 knockout (KO) mice, connectivity is disrupted.Here, the initial guidance of the ipsilateral (uncrossed) retinal ganglion cell (RGC) axons is altered (Glendining et al., 2017), causing a dramatic and stereotyped disruption to visuotopic mapping (Leamey et al., 2007).In wild type (WT) mice, ipsilateral retinal inputs to the dorsolateral geniculate nucleus (dLGN) are restricted to a small dorsomedial patch that is visuotopically aligned with contralateral retinal inputs.In Ten-m3 KO mice, ipsilateral retinal terminals extend into far ventrolateral dLGN (Leamey et al., 2007), profoundly disrupting the neural representation of visual space (Merlin et al., 2013).The well characterized nature of normal binocular mapping, and its stereotyped disruption in Ten-m3 KOs, presents an informative model to assess the effectiveness and mechanisms of interventions which aim to correct the wiring defects.
In mice, environmental enrichment (EE)-a paradigm which enhances physical, cognitive, and social experience relative to standard housing conditions in laboratory animals-has been shown to have numerous beneficial effects.These include acceleration of neural circuitry development (Cancedda, 2004;Landi et al., 2007;Wang et al., 2013) and enhancement of plasticity in adults (Baroncelli et al., 2010;Ehninger & Kempermann, 2003;Jaepel et al., 2017;Reshef et al., 2014;Ziv et al., 2006).Our recent work has highlighted a novel capacity of EE from birth to drive corrective change to the miswired visual pathway of Ten-m3 KO mice: removal of the most visuotopically mismapped retinal inputs to the dLGN along with improvements to a usually defective, visually mediated behavior (Blok et al., 2020;Eggins et al., 2019).Importantly, sensitivity to the EEinduced correction was limited to the first few postnatal weeks and appeared to involve enhanced levels of microglial reactivity around this time (Rogerson-Wood et al., 2022).
Despite the increasing awareness of the importance of microglia in circuit plasticity (Faust et al., 2021;Prinz et al., 2019;Schafer et al., 2012), efforts to understand how EE impacts microglial function have largely centred around its capacity to influence the microglial-response after immune insult (Augusto-Oliveira & Verkhratsky, 2021).The ability of EE to drive corrective changes in miswired circuitry and rescue function is potentially of great therapeutic importance.Here, we aimed to determine the timing of microglial sensitivity to this experience-based intervention, and to establish whether EE can drive microglia to selectively engulf ectopic axonal terminals in the visual pathway of the Ten-m3 KO model of miswiring.

| MATERIALS AND METHODS
All procedures were performed with approval from the University of Sydney Animal Ethics Committee and conformed with NHMRC (National Health and Medical Research Council) Australian Government guidelines for the care and use of laboratory animals.All mice were housed in climate-controlled rooms ($23.50C, 40%-70% humidity) at a University of Sydney Laboratory Animal Services (LAS) facility on a fixed 12 h/12 h light/dark cycle.Standard mouse chow and water were provided ad libitum.

| Mice
Generation of the Ten-m3 KO line has been previously described (Leamey et al., 2007).The line was originally generated using a purely C57Bl/6 background but was subsequently crossed and maintained on a C57Bl/6-Sv129 background to improve homozygous Ten-m3 KO pup survival rate.Pups on this background can be of varied pigment.Because alternations in the ipsilateral visual pathway of albino mice have been previously shown (Dräger, 1974;Drager & Olsen, 1980), only pigmented mice were used.Homozygous [Ten-m3À/À(KO)] mice were obtained for this study through the crossing of heterozygous females with heterozygous or homozygous males-only heterozygous dams were used, as we have previously seen qualitative evidence of variable maternal care in homozygous KO dams.Breeding was undertaken in standard environment (SE) housing with SE parentage.
A few days before birth, pregnant dams were transferred from their breeding cages into EE housing, or maintained in SE (see Blok et al., 2020;Eggins et al., 2019, for detailed housing descriptions).
Although for ethical reasons EE commenced from a couple of days before birth, EE from birth is used here to describe pups of dams born into the EE cages to clearly distinguish it from EE commenced at later developmental stages which is common in other protocols: we have previously shown that EE commenced at weaning (3 weeks postnatal) has no impact of correction of miswired retinogeniculate projections (Blok et al., 2020;Eggins et al., 2019).Powdered standard mouse chow mixed with the drinking water and "Necta H 2 O" hydration gel (Able Scientific, Australia) were included in the cage ad libitum from postnatal day (P)14 in both SE and EE reared pups to provide support near weaning.Dams were removed from litter cages when pups were approximately 3.5 weeks old.Litters were of mixed genotypesheterozygotes, homozygotes, and wildtypes.To distinguish, genotyping was undertaken (between P10 and P21) using tail and/or ear biopsies as described previously (Leamey et al., 2007).

| Anterograde tracing
All experimental mice received an intraocular injection orientated to avoid damage to the retinal ventral-temporal crescent, the location of ipsilaterally originating RGCs, of the anterograde tracer Cholera Toxin Subunit B (CTB) (Recombinant), Alexa Fluor 488 or 594 conjugate (Invitrogen, Cat# C22841 or C22842, 1% w/v in 0.1 M phosphate buffer (PB), pH = 7.4) into the vitreous chamber of one eye under anesthesia (2%-3% inhalation of isofluorane-oxygen mixture) on either P14, P17, P20, P24, or P29 (age accurate to within 24 h) as previously described (Eggins et al., 2019).Approximately 24 h following tracer injection, mice were euthanased with an overdose of sodium pentobarbitol (>100 mg/kg i.p) and transcardially perfused with 0.9% saline immediately followed by 4% paraformaldehyde (PFA) in 0.1 M PB (Eggins et al., 2019).Mouse brains were excised, postfixed (2-6 h in 4% PFA in 0.1 M PB), cryoprotected (2-7 days in 30% sucrose), given a randomly assigned 4-digit identification number, and sectioned coronally on a freezing-sliding microtome set to 60 μm thickness.If storage was required before sectioning, embedded brains were snap frozen using powdered dry ice and stored in airtight packaging at À80 C until use.To confirm the completeness of CTB uptake by the retina, the extent of terminal labeling within comparable rostral contralateral dLGN was routinely checked.Any cases with poor labeling were excluded from analysis.

| Immunolabeling
Six sections from each mouse, which encompassed the rostral third of the dLGN ipsilateral to the injected eye (the region where the ventrolateral displacement of ipsilateral retinal terminations is most pronounced in Ten-m3 KO mice; Leamey et al., 2007) were chosen (blinded to housing, sex, or age) for subsequent processing.Sections were rinsed (0.1 M PB with 0.3% TritonX-100 [PB-Tx]), incubated for 2 h in blocking solution (2% v/v NGS in PB-Tx) and then placed in primary antibody incubation solution (rabbit anti-Iba1, Wako, Cat# 1919741, 1:500 and rat anti-CD68, Abcam, Cat# ab53444, 1:1000 in blocking solution) overnight (18-26 h at room temperature).

| CD68 labeling analysis
For CD68 labeling time-course, two imaged sections were chosen for analysis (see also "Experimental Design and Statistical Testing").Each low power z-stack (10Â objective; 2.32 μm z-stack interval) first had the primary region of interest (ROI; 80 μm Â 80 μm around the ventrolateral border of CTB-labeled ipsilateral terminals), control region of interest (cROI; 80 μm Â 80 μm around the dorsomedial border of ipsilateral terminal labeling), as well as the outline of the dLGN defined and saved using the ROI Manager tool.CD68 labeling then underwent background subtraction (rolling-ball function, radius 5 [applied 4 times]).Maximum intensity projections (MIPs) of CD68 labeling were then created for each section, and after a second background subtraction (4Â rolling-ball function, radius 5), binary images were acquired using the inbuilt "Default" thresholding algorithm.For each binarized image, the Analyse Particles tool was used to measure "average particle size" and "%area" for the whole dLGN and the ROIs.Fold change ratios of "average particle size" and "%area" were then determined for each ROI in each section, calculated as the value of a given ROI divided by the value for the associated dLGN outlined area.

| Engulfment assay
For the Imaris based engulfment assay, high power confocal z-stacks (63Â objective; 0.5 μm z-stack interval) of the primary (ROI; 127 Â 127 μm 2 around the ventrolateral border of ipsilateral retinogeniculate terminals) and control regions of interest (cROI; 127 Â 127 μm 2 around the dorsomedial border of ipsilateral retinogeniculate terminals) were first preprocessed using Image J (FIJI, NIH, USA).An attenuation correction plugin (Biot et al., 2008) was applied to compensate for the decrease in fluorescence intensity with increased imaging depth (all channels).Background fluorescence was then subtracted (all channels) using the rolling-ball background correction function.A rolling-ball radius of 50 for the Iba1 channel, and 15 for the CTB and CD68 channels, were used.Some z-stacks also had the "De-speckle" and/or "Remove Outliers" function applied (choice guided by the type of background noise present).To facilitate microglia surface rendering (Schafer et al., 2012(Schafer et al., , 2014)), a mean filter (size of 1.5) was additionally applied to Iba-1 channels.Preprocessed images were transferred from FIJI to Imaris (Oxford Instruments, UK) for quantitative analysis using the "Image from FIJI" function of the ImarisXT Imaris-ImageJ/FIJI Bridge.
To identify relevant microglia for the quantitative engulfment assay, a manual identification of microglia (based on overall morphological labeling indicated by Iba-1 labeling) was first undertaken using three-dimensional (3D) visualization and manipulation capacities of Imaris (Surpass Mode).All microglia which had a whole cell body contained entirely within the ROI, and that did not abut the ROI border, were included: microglia which had incomplete Iba1 labeling were excluded from analysis.The later 3D surface rendering of fluorescence volumes (CD68 and engulfed CTB) was achieved using the Automatic Surface Creation tool within Imaris.Background subtraction (local contrast) thresholding was done with the "Smooth" function enabled (1.85 μm diameter of largest sphere which fits into object).Thresholding values were established empirically through blinded (housing and age) visual assessment undertaken prior to rendering.
To isolate CD68 labeling from single microglia, an integral component of a single cell-engulfment assay (Schafer et al., 2014), the inbuilt semiautomatic "Filament Tracing" (autopath mode) function in Imaris using microglia morphology indicated by Iba-1 labeling (Sasaki et al., 2001) was used, followed by the 3D "Convex Hull" function (Convex Hull Xtension package available from Imaris Open) to crop CD68 labeling from individual microglia (undergoing analysis) out of the whole ROI image (Figure 3a).Isolated CD68 fluorescence for each microglia were surface rendered (surface area detail 0.1 μm), irrelevant surfaces (those not part of current microglia undergoing analysis) manually deleted, and the resultant "CD68 Surface" used to mask off external CTB fluorescence (to identify CTB fluorescence within CD68 lysosomes).The CD68-internalized CTB fluorescence (if present) was then surface rendered (surface area detail 0.1 μm), Imaris "Unify Surfaces" function applied (to facilitate automatic recognition of multiple disjointed surfaces as part of one volume), and the total CTB volume within the given microglial grouping of CD68 recorded (automatic measure in Imaris-"Volume").
The following was determined for each mouse (ROI and cROI data analyzed separately): the density and proportion of total microglia containing phagocytosed CTB within their CD68 lysosomes ("CTB + microglia"), and the median aggregate volume of engulfed CTB taken up by individual microglia (median chosen due to skewed data and/or presence of replicate outliers in some samples; see below).

| Anatomical analysis
Only mice which had largely uniform intensity and good contrast of ipsilateral CTB label across the rostral dLGN coronal sections were used for anatomical analysis-those with very faint/absent label, or an obviously distorted "angle of cut" were excluded.Analysis of sections was undertaken in a randomized fashion (housing, sex, and age cohorts intermingled) to avoid an inadvertent "cohort effect".Anatomical analysis was modeled after one described previously (Eggins et al., 2019).Briefly, the dLGN was outlined, defined, and saved using the ROI Manager tool (ImageJ) from low power z-stacks (10Â objective; 2.32 μm z-stack interval) of coronal sections.dLGN area was measured using the "Measure" tool (ImageJ).Ipsilateral RGC CTB labeling then underwent a background subtraction (rolling-ball function, radius 100, applied four times).MIPs of ipsilateral CTB label were created for each section, a second background subtraction (rolling-ball, radius 100, applied four times) along with a de-speckling ("De-Speckle") function applied, and binary images acquired (inbuilt ImageJ thresholding algorithms).Using the ImageJ high-low lookup table option, the brightness of each image was increased such that the brightest CTB pixels were sitting just below being "Over-Exposed."Binary images were then acquired (inbuilt ImageJ thresholding algorithms).
To determine if there were any region-specific changes to ipsilateral RGC CTB label, the long axis of each coronal dLGN section was divided into thirds (dorsal, middle, and ventral).For this, the "Fit Rectangle" function was applied to the given dLGN outline for each binarized section and the external area "Filled" (ImageJ).The image was then rotated (using the Transform function) until the long axis of the rectangle sat flush with the horizontal plane/axis, and the area external to the bounding rectangle cropped.A macro was then batch applied to all resultant cropped and rotated images to measure the following: the total ipsilateral label, and the ipsilateral label in each image sub-region (dorsal, middle, and ventral third).The following was then calculated for each imaged section: the area occupied by the total binarized (ipsilateral) CTB label, and the label in each dLGN subregion (dorsal, middle, and ventral third), relative to the total area of the associated dLGN outline (%).

| Experimental design and statistical testing
For all analyses (CD68, engulfment, anatomy), mice used in each treatment group came from a minimum of three different litters.Sample sizes were chosen based on previous studies.Two adjacent coronal brain sections from each mouse were used: sections encompassing the most rostral ipsilateral RGC-CTB labeled terminals, but before the ipsilateral RGC terminal "zone" completely abuts the optic tract.Pups of both sexes were included in approximately equal numbers, according to availability.All sectioning, imaging, and analyses were performed blind to age, sex, and housing group.For the engulfment assay, all microglia (variable numbers) whose cell body was contained entirely within the given ROI and was not directly abutting the ROI border, were included.

| RESULTS
We have previously provided evidence of EE-induced corrective pruning in the visual thalamus of Ten-m3 KO mice enriched from birth (Eggins et al., 2019), which was underway at around 3-4 postnatal weeks (Rogerson-Wood et al., 2022).This pruning was specific to the locus of the most mismapped ipsilateral retinal inputs, where evidence of EE-driven microglial recruitment and reactivity was also observed.
No evidence of increased levels of microglial reactivity was detected in the visual thalamus of age-matched SE or EE WT mice, nor any evidence of EE-driven pruning of retinal inputs at this stage (Rogerson-Wood et al., 2022).To better characterize potential microglial involvement we first assayed a series of ages from P15 to P30 in Ten-m3 KO mice enriched from birth, and compared them with SE housed KO mice at the same age points using changes in CD68 (cluster of differentiation 68 or "macrosialin") labeling and distribution as markers for microglial reactivity.In the surveillant microglial state, CD68 is reported to be expressed at low levels with a punctate distribution, whereas when microglia are activated, CD68 levels are higher with a more globular distribution (Aono et al., 2017;Chistiakov et al., 2017;Lier et al., 2021;Schafer et al., 2012).The timepoints assessed were based on our previous studies (Eggins et al., 2019;Rogerson-Wood et al., 2022).Given the lack of effect of EE from birth on pruning of retinal afferents or any evidence of EE-induced microglial reactivity at this developmental stage in WT mice (Rogerson-Wood et al., 2022), we here focused our efforts solely on the EE-driven effects in Ten-m3 KOs.

|
The EE driven localized-CD68 upregulation in the Ten-m3 KO visual thalamus occurs in a brief window after postnatal day (P)18, but before P30 To establish the time-course for the potential EE-driven localized microglial-reactivity in Ten-m3 KO mice, we assayed CD68-labeling in the dLGN at five developmental timepoints-P15, P18, P21, P25, and P30-in both SE and EE (from birth) cohorts of Ten-m3 KO mice: P15 CD68, a transmembrane lysosomal glycoprotein, is a marker for microglial reactivity (Lier et al., 2021;Schafer et al., 2012).The anterograde tracer CTB was used to unambiguously label ipsilateral retinal axons (see Methods).
Qualitative analysis of combined CD68 labeling and anterograde tracing of ipsilateral retinal terminals (Figure 1a-p; Figure S1) revealed SE Ten-m3 KO mice as having a largely uniform labeling across the dLGN at all ages examined (Figure 1a, e, i, m; Figure S1a).Ten-m3 KO mice enriched from birth until P15 (Figure S1c) or P18 (Figure 1c) also showed no notable localized changes to CD68 labeling across the nucleus.When EE-P21 mice were assessed, however, a marked increase in CD68 reactivity was observed at the far ventrolateral extremity of ipsilateral RGC terminals in three of the eight mice analyzed (Figure 1g).A more robust effect was seen in EE-P25 where six out of the eight mice analyzed showed a clear increase in CD68 labeling in the corresponding location (Figure 1k).Moreover, higher powered images of this region (Figure 1b,d,f,h,j,l,n,p) revealed that CD68 particles appeared reliably larger and more "globular" in EE-P21 (Figure 1h) and EE-P25 samples (Figure 1l), strikingly different from the small and punctate CD68 clumps seen at E-P15 (Figure S1d), EE-P18 (Figure 1d) and EE-P30 (Figure 1p), as well as all ages in SE visual thalamus (Figure 1b,f,j,n, Figure S1b).
Since our qualitative analysis indicated that CD68 labeling was uniformly low under SE or EE conditions between P15 and P18, we used P18 as the starting point for our quantitative analysis for comparison with later stages.We performed our analyses on primary (ROI) and control (cROI) regions of interest defined by the ventrolateral and dorsomedial border of the ipsilateral terminals, respectively.
To further determine whether the change in CD68 average particle size fold change ratio for the ROI was unique, a similar comparison between SE and EE KOs was conducted on the cROI (the dorsomedial border of ipsilateral retinogeniculate terminals; Figure 2c).nor between EE-P21s, and EE-P30s ( p = 1.000; Figure 2f).
An effect of age (univariate ANOVA with age, housing, and sex as fixed factors: F [3, 48] = 3.971, p = .013)for CD68 "percentage area" fold change ratios was observed at the cROI (Figure 2c): no other main effects of housing, sex, or interactions were detected.Pairwise comparisons revealed no significant differences between any groups (see right two columns of Table 2 for descriptive statistics).
Our analyses showed no effect of sex (see Table S1), nor any interactions between this parameter and any of the other factors assessed (housing Â age; age Â sex; housing Â age Â sex; Table S1).
Since no evidence of an effect was observed on EE-induced microglial reactivity (see data from males and females graphed separately: Together these results show that EE (from $birth) drives increased CD68, both in terms of particle size and % area -indicators of microglial reactivity-at a site previously established to exhibit corrective pruning of miswired ipsilateral terminals within the visual thalamus of Ten-m3 KO mice (Eggins et al., 2019;Rogerson-Wood et al., 2022).Further, this change occurred within a brief window of time commencing after P18 and ceasing by P30, suggesting that there is a "critical period" for this EE-induced microglial reactivity.This effect was not impacted by sex and was seen specifically at the area of pruning in EE mice at this time.
3.2 | Microglial-engulfment of mismapped ipsilateral retinal terminals in the Ten-m3 KO visual thalamus is heightened when localized CD68upregulation is at its highest Microglial-engulfment of neuronal presynaptic elements has been previously mechanistically linked with the developmental pruning of wild type RGC terminals in the visual thalamus at other developmental stages (P5-9 and $P40 Schafer et al., 2012Schafer et al., , 2016)).Heightened CD68 expression has also been shown to be associated with increased microglial engulfment-activity (Lier et al., 2021;Schafer et al., 2012).
Definitive evidence of microglial engulfment of terminals requires a more thorough assay, however.Thus, in order to test whether the EEinduced localized CD68 upregulation, found to peak here in P25 Ten-m3 KO mice, was indeed indicative of heightened-levels of microglial engulfment of miswired ipsilateral retinal terminals, an engulfment assay adapted from previous work (Schafer et al., 2014; see also Methods) to directly quantify CTB label (ipsilateral RGC axon terminal segments) within CD68+ microglial lysosomes (Figure 3a-c) was undertaken on microglia (labeled with Iba 1, Figure 3a-c The density and proportion of total (ROI) microglia containing phagocytosed CTB within their CD68 lysosomes ("CTB+ microglia") was compared (Figure 3f-i).We also asked whether the volume of engulfed CTB taken up by individual microglia differed between groups (Figure 3j,k).Consistent with previous findings (Figures 1 and   2; Rogerson-Wood et al., 2022), microglia from the corrective pruning locus (ROI) in EE-P25 Ten-m3 KO mice showed qualitative phenotypes typical of being more "activated": more globular CD68 labeling, less ramified morphology (larger soma, shorter fatter branches), and clumped (uneven) cell distribution (Figure 3).
Finally, a housing Â age interaction (univariate ANOVA with age and housing as fixed factors: F [1, 12] = 4.795, p = .049)was detected for the volume of CTB engulfed by individual microglia (Figure 3j) at the ROI (Figure 3d).Subsequent pairwise comparisons showed that EE-P25s had a significantly elevated volume of engulfed CTB (μm 3 ) compared with SE-25s ( p = .009),but there was no difference between EE-P18s and SE-P18s ( p = .983;see Table 3 [right] for descriptive statistics).EE-P25s also had a significantly elevated volume of engulfed CTB (μm 3 ) compared with EE-P18s (p = .022),but there was no difference between SE-P25s and SE-P18s ( p = .651; Figure 3j; see also Table 3, right columns).
Together these results indicate that EE (from birth) drives increased microglial-engulfment that, like the localized CD68-upregulation, appears "targeted" toward the most-inappropriately mapped ipsilateral RGC terminals in the Ten-m3 KO visual thalamus.Further, this increase in microglial-engulfment displayed an age-specificity not present at P18 but notable at P25 that aligned with the time course of localized CD68-upregulation, suggesting a similar critical window inbetween P18 and P30.To further determine the locus of EE-driven reduction in ipsilateral terminals, CTB label contained within the dorsal, middle, or ventral regions of the dLGN was determined by dividing the long (dorsomedial-ventrolateral), axis of the dLGN into thirds (Figure 4a-d, dividing lines within dLGN outline; see Methods, also (Eggins et al., 2019;Rogerson-Wood et al., 2022)).As with total ipsilateral terminal area, ipsilateral terminal area within each third (dorsal, middle, and ventral) was analyzed as a percentage of total dLGN area.
The dorsal region alone exhibited no effect of housing (univariate ANOVA with age and housing as fixed factors: F [1, 20] = 0.004, p = .953)but did show a small, significant effect of age (F [1, 20] = 5.559, p = .029).No age Â housing interaction was detected (F [1, 20] = 0.019, p = .892).Pairwise comparisons within housing groups, however, revealed no differences between EE-P18s and EE-P30s, nor between SE-P18s and SE-P30s ( p = .132;see Table 5 for descriptive statistics; Figure 4g).Thus, while there is a small agerelated change in the proportion of terminals in the dorsal region of the dLGN, no effect of housing was observed.
T A B L E 3 Descriptive statistics of microglial-engulfment measures for both the ROI (left 2 columns) and cROI (right 2 columns) in Ten-m3 KO mice.
Age CTB+ microglia/mm 3 %CTB+ microglia Engulfed volume (μm 3 ) Note: Left: CTB+ microglia expressed as the number of cells per mm 3 within the ROI for SE and EE groups at ages P18 and P25, Middle: % CTB+ microglia compared with total microglia in ROI, and Right: Volume of CTB engulfed by individual microglia within the ROI expressed in μm 3 for SE and EE groups at ages P18 and P25 engulfed volume of CTB (right); see also Figure 3 and text for more details.Mean ± standard deviation are presented with sample size in parentheses.
These findings point to the notion that the EE-driven removal of ipsilateral retinal inputs in the Ten-m3 KO visual thalamus occurs from the ventrolateral border, in alignment with previous results (Eggins et al., 2019;Rogerson-Wood et al., 2022).Crucially, we show that sig-   We demonstrate here that there is a window during early postnatal life where experience can induce a highly localized increase in the reactivity of microglia to selectively engulf miswired neural projections to drive normative changes.Previous studies have demonstrated that altered experience, including EE, can influence microglialengulfment (Augusto-Oliveira & Verkhratsky, 2021;Li et al., 2022).
Much of this work, however, has focused on how experience modulates microglial reactivity as part of the inflammatory response to pathological insult.For example, adult voluntary-wheel running has been found to drive a normalization in the surplus hippocampal-CA3 synapses caused by in utero maternal immune inflammation via increased microglial-engulfment (Andoh et al., 2019).In a similar vein, voluntary exercise during adolescence (P23-34) has been shown to prevent the usual reductions in microglial engulfment in the hippocampal dentate gyrus that come with extended sleep deprivation (Tuan et al., 2021).The current study, to the best of our knowledge, is the first to show that EE can drive targeted microglial engulfment of ectopically wired components of dysfunctional neuronal circuits that arise due to genetic disruption of circuit assembly.
Our findings correlate with, and substantially extend, previous work which has shown that there is a critical period for EE-induced correction of miswired ipsilateral projections in the Ten-m3 KO retinogeniculate pathway.Exposure to EE for a period of 6 weeks, commencing from birth but not from weaning (P21) or adulthood, drove removal of the most visuotopically mismapped ipsilateral retinal inputs in the dLGN of Ten-m3 KO mice (Eggins et al., 2019), as well as improvements in a defective, visually mediated behavior (Blok et al., 2020).Importantly, this process was recently shown to be in progress by around P25 (Rogerson-Wood et al., 2022).The demonstration that EE from birth induces increased microglial reactivity at the site of correction, and that this commences between P18 and P21, fits well with these earlier observations.Most notably, these findings provide a potential explanation as to why EE commenced at P21 (or later) was unable to drive the pruning of miswired terminals in the dLGN (Eggins et al., 2019), suggesting that exposure to EE must be initiated some days prior to the onset of experience-induced localized changes in microglial reactivity to enable this corrective process.
These insights set the stage to determine more precisely how much earlier than the observed localized microglial reactivity EE needs to commence to trigger microglia to perform corrective pruning.
Current literature suggests that microglia are active regulators of pruning, rather than merely scavengers of pruned terminals (reviewed in Whitelaw et al., 2022).Although our data do not directly address this question, the presence of CTB labeled terminal endings engulfed by microglia seen here are consistent with them playing an active role in pruning.It remains possible, however, that the microglial cells are merely scavenging processes that have already been pruned via some other mechanism just prior to the time of assessment.The establishment of the timeline of EE induced, microglia-mediated circuit repair for this circuit will provide an important basis for future studies to assess the role of microglia more definitively via their inactivation using agents such as PLX3397 (which depletes the microglial population; Elmore et al., 2014) and minocycline (which dampens their phagocytic activity; Luo et al., 2023;Schafer et al., 2012;Sellgren et al., 2019).
T A B L E 5 Descriptive statistics for regional analysis comparing percentage of dLGN regions occupied by ipsilateral terminals within in the dorsal (left column pair), middle (middle column pair), and ventral (right column pair) thirds of the dLGN.Regardless, a close correlation of significant changes in CD68 percentage area, size of CD68+ particles, and the timing of microglialengulfment were observed, suggesting that these changes in CD68 labeling serve as an effective marker of active microglial-engulfment.This is consistent with previous studies, assessing similar metrics in different neural systems and developmental timepoints (Aono et al., 2017;Schafer et al., 2012).
4.2 | EE induced wiring correction may be distinct from WT retinogeniculate developmental processes The timing of the EE-mediated microglial pruning described here raises questions as to whether this experience-induced reactivity merely takes advantage of normal developmental processes occurring at the time or whether enrichment is driving distinct mechanisms.The window of corrective pruning identified in this study overlaps with an already defined "critical-period" ($P20-30) of visual-sensitivity for the WT retinogeniculate synapse (Hooks & Chen, 2006, 2008), suggesting a potential relationship.The changes reported in these previous studies, however, do not appear to correlate well with the EE-driven pruning observed here in Ten-m3 KO mice.Notably, large scale changes to the axonal scaffold (Hong et al., 2014) and associated increases in microglialengulfment of axonal fragments (Schafer et al., 2016) were only reported to occur after the closure of this critical-period (after P30) and were only assayed in contralateral monocular regions of WT visual thalamus.During P18-P30, only a clustering of presynaptic boutons (on a maintained, broad axon arbor) has been noted (Hong et al., 2014), quite different from the significant loss of aberrant ipsilateral axonal inputs and elevated levels of localized microglial engulfment observed here in enriched Ten-m3 KO mice.
Recent studies have looked at changes in retinogeniculate connectivity in more "binocular" regions of the dLGN during this same developmental window (Li et al., 2023).Interestingly, unlike in the current study, ipsilateral retinal inputs were shown to be largely conserved in WT mice, with contralateral inputs identified as undergoing most of the functional changes (Li et al., 2023) Although the expression pattern of Ten-m3 indicates it is present in neurons, we cannot currently exclude the possibility that it is also expressed in microglia.It is therefore possible that, since the mice used in this study are constitutively lacking Ten-m3 (Leamey et al., 2007), microglial function may be compromised in this model.
Accordingly, while it seems unlikely, the ipsilateral miswiring itself may be a direct result of microglial dysfunction, and EE could compensate for this by facilitating localized microglial reactivation and engulfment of mismapped RGC terminals seen here (Eggins et al., 2019;Rogerson-Wood et al., 2022).While a change in microglial reactivity cannot be ruled out in Ten-m3 KOs, a loss of function seems very unlikely given that segregation of ipsilateral and contralateral RGC terminal domains, a process shown to be dependent on microglial-mediated pruning (Schafer et al., 2012), is not affected in Ten-m3 KOs, and follows the same developmental time course as in WTs (Glendining et al., 2017).Further studies are required to definitively establish whether Ten-m3 is expressed in microglia, and/or has any direct role in their function.

| EE may drive unique microglial signaling mechanisms
A striking feature of our results was the high level of specificity of the EE-induced effects on microglia, both temporally and spatially.This is even more remarkable when one considers that EE is an intervention that impacts the whole animal over multiple weeks.Despite this, microglial reactivity in the dLGN was highly localized to the most profoundly miswired terminals over the course of a few days.This suggests that there is a developmental critical period during which EE enables microglia to identify and specifically prune miswired inputs.
The signaling molecules that underlie this targeted removal of neuronal inputs have yet to be identified.A major vision-sensitive signaling interaction (Fn-14-TWEAK) shown to be instructive for retinogeniculate synapse removal specifically during the period of visionsensitivity ($P20-30; see above), was found not to influence microglial-engulfment (Cheadle et al., 2020), likely discounting its direct involvement here.Studies in other systems provide precedence for novel epigenetic drivers of plasticity that can be uniquely switched-on by early exposure to EE (Arai et al., 2009;Baroncelli et al., 2016;Li et al., 2006).Previous molecules identified as playing a direct "instructive" role for retinogeniculate synapse elimination via microglial-engulfment include phosphatidylserine (Scott-Hewitt et al., 2020) and the complement proteins C1q (Bialas & Stevens, 2013;Cong et al., 2022) and C3 (Schafer et al., 2012).JAK2 has also been shown to be involved in retinogeniculate synapse elimination during monocular segregation in the dLGN (Yasuda et al., 2021) and could be involved.
Given the seeming mechanistic involvement of microglia, it is conceivable that EE from birth may not only promote the corrective pruning via increasing the instructive messaging microglia receive, but prime the microglia themselves, increasing their capacity to engulf and remove aberrant connections.This could involve the upregulation of microglial receptors, such as MerTK (Bolton et al., 2022), the GPR56 splicing isoform (Li et al., 2020), or CR3 (Schafer et al., 2012) that facilitate microglial engulfment of synapses.EE could also act by decreasing the levels of global negative-regulators of microglial engulfment, such as CD22 (Pluvinage et al., 2019).Future work should investigate whether any of these signaling pathways are mediating/driving the EE-driven corrective pruning characterized here, or whether distinct signaling systems are at play.et al., 2023;Thompson et al., 2016).Importantly, these changes could only be induced during the WT retinogeniculate "critical period" of visual sensitivity (P20-30).Distinct alterations in ipsilateral eye input mapping in V1, as well as a global suppression of V1 activity levels, have been seen in Ten-m3 KO mice (Merlin et al., 2013).Further, feedback input from V1 has been shown to be highly dynamic (Kirchgessner et al., 2020), and directly influenced by arousal-state (Reinhold et al., 2023).EE may therefore help drive distinct patterns of feedback from V1, to instruct activity-dependent processes that promote pruning of the miswired projections.Further work will be required to assess this possibility.

| Maternal care can be affected in dams housed in EE conditions
The EE protocol used in this study involved a series of augmentations to the home-cage of mice which were initiated a few days prior to birth.It is therefore likely that part of the impact of EE on the pups would have been indirect, via the dams, particularly for the first couple of weeks when the pups were predominantly in the nest, and the interaction with the mother was a major part of their experience.
Direct effects of EE likely started to become more prominent once the pups began to explore the environment independently at around eye opening (Simonetti et al., 2009).The increased novelty and complexity of the environment is likely to promote more movement and social interaction which in turn enhances the experience (Kempermann, 2019).As EE can alter maternal care (Sale, 2018;Sparling et al., 2020), it is likely that the impact observed here on microglial-mediated corrective pruning could, at least in part, be due to changes in these interactions.Further studies will be required to differentiate the contribution of maternal engagement from direct effects of EE during the first few postnatal weeks in developing mice.

| EE-inducible microglial reactivity provides as a potential therapeutic tool
Recent work has provided evidence that early environmental intervention can impact the severity of genetically driven changes affecting neurodevelopment (Guthrie et al., 2023;Klin et al., 2020;Whitehouse et al., 2021).While optimal timing for such interventions have yet to be determined (e.g.Towle et al., 2020), the work presented here strongly suggests that, at least for the miswired visual pathway in Ten-m3 KO mice (Leamey et al., 2007), experience-driven microglia-mediated correction can only be activated if animals are exposed to EE during a specific, early window of neurodevelopment.
Over-exuberant/altered patterns of connectivity have been reported in various human neurodevelopmental conditions including autistic spectrum disorders (Openshaw et al., 2023;Takeguchi et al., 2022;Xie et al., 2023) and associated mouse models (Bertero et al., 2018;Carroll et al., 2021;Jain et al., 2020;Kennedy et al., 2020;Smith et al., 2019).Notably, an increases in CD47, a molecule which inhibits microglial-mediated phagocytosis, is upregulated in cells derived from people with autism (Li et al., 2021).The capacity to increase experience-driven microglial reactivity to drive normative changes, as demonstrated here, could potentially be of enormous benefit therapeutically in individuals affected by such conditions.Future work determining the potential for early EE to correct aberrant wiring in other established neurodevelopmental models will be required to assess this possibility, potentially in combination with pharmacological manipulations.

| Conclusions
In conclusion, we here provide evidence that microglia are cellular mediators of EE-induced corrective pruning in the retinogeniculate pathway of Ten-m3 KO mice.We show that EE from birth drives microglia to engulf miswired projections in a highly temporally and spatially restricted manner.This suggests the presence of a unique developmental critical period (around 3-4 weeks postnatal in the pathway examined), during which microglia can respond to an environmental intervention that can drive normative changes in neural circuitry.Our study adds to the growing awareness of the multifaceted roles of microglia and importance of early experience in shaping the impact-severity of genetic-risk factors for neurodevelopmental conditions (Klin et al., 2020), paving the way for future studies undertaking manipulations of microglial function to promote corrective pruning.
All microscopy was undertaken blind to age, sex, and housing conditions.At least three of the six coronal brain sections from each mouse which underwent CTB-tracer and immunofluorescence labeling, were chosen for imaging (the most rostral dLGN sections which still contained ipsilateral RGC terminal CTB-labeling).Acquisition of low power confocal z-stacks of the whole dLGN coronal cross-section (2.32 μm z-stack interval) for CD68 labeling and anatomical timecourse analyses was undertaken using a Zeiss LSM 800 confocal microscope with a 10Â NA0.45 Plan-Apochromat dry objective (laser excitation line 488 nm for Alexa Fluor 488, 561 nm for Alexa Fluor 594, and 640 nm for Alexa Fluor 647).Acquisition of high power confocal z-stacks of the ventrolateral border (primary region of interest or "ROI") and the dorsomedial border (control ROI or "cROI") of CTB-labeled ipsilateral retinogeniculate terminals (0.5 μm z-stack interval) for terminal engulfment time-course analysis was undertaken on a Zeiss LSM 800 confocal microscope using a 63Â NA1.40 Plan-APOchromat oil immersion objective (laser excitation lines for Alexa Fluor fluophores as listed above).Using the same imaging setup and parameters, confocal z-stacks of the ventrolateral border of CTB-labeled ipsilateral retinogeniculate terminals were also acquired to qualitatively confirm CD68 labeling time-course.
Microsoft Excel (Microsoft Corporation, WA, USA) and SPSS (IBM Corporation, NY, USA) were used for data processing and graphical representations (significance level α set to 0.05).All statistical analyses used standard univariate omnibus testing (see Results).Each mouse was graphically represented by a single fill-in and colored data point.For all measures except CTB+ microglial density and % CTB+ microglia (of total), this point represented an aggregate value of either section or cell (biological) replicates.A mean aggregate (of section replicates) was used for all time course CD68 labeling and anatomical measures.A median aggregate (of cell replicates) was calculated for the engulfment assay measure of "Engulfed CTB volume per microglia" due to the presence of outliers and skewed distributions in some replicate data sets.Where applicable, replicate values in addition to aggregate values, are depicted on graphs-replicates are indicated by faint gray unfilled circles, grouped in columns according to sample identity.

F
I G U R E 1 Environmental enrichment drives an age-dependent, localized change in CD68 expression at the ventrolateral border of mismapped ipsilateral retinal inputs in the Ten-m3 KO dLGN.(a-p) Representative confocal images of CD68 (lysosomal protein in microglia, yellow) immunolabeled coronal dLGN sections from P18, P21, P25, and P30 Ten-m3 KO mice, either standard-housed (SE) or environmentally enriched (EE) from birth.Ipsilateral retinal terminals were labeled using Alexa Fluor-conjugated CTB (magenta, see Methods).Equivalent coronal dLGN sections along the rostral-caudal axis are presented (rostral extremity shown-see Methods).Images are maximum intensity projections of confocal z-stacks.(a, c, e, i, m, o) Low-power images of a representative section from P18 (a), P21 (e), P25 (i), and P30 (m) SE-KO mice, as well as P18 (c) and P30 (o) EE-KO mice all display a largely uniform distribution of CD68 labeling across the dLGN (see text).(g, k) Low-power images of a representative section from EE-P21 (g) and EE-P25 (k) mice display a distinct upregulation in the pattern of CD68 labeling in a small region of the dLGN: the ventrolateral (VL) extremity of the ipsilateral retinal terminals (boxed area; ROI).This focal change was seen in some (3 of the 8) of the EE-P21s and most (6 out the 8) of the EE-P25s examined.(b, d, f, h, j, l, n, p) Higher power images of corresponding boxed areas depicted in low power samples (ROIs; boxed 126 μm Â 126 μm) show a change in CD68 labeling, with larger more globular clumps in EE-P21 (H) and EE-P25 (L) sections compared with those from all other age and housing cohorts.CTB, cholera toxin subunit β; dLGN, dorsal lateral geniculate nucleus; DM, dorsomedial; P, postnatal day; VL, ventrolateral.Scale bar: 100 μm for (a, c, e, g, i, k, m, o) and 20 μm for (b, d, f, h, j, l, n, p).(ai, ci, ei, ii, mi, oi) low power images of CTB (magenta) and CD68 (yellow) labelling; (aii, cii, eii, iii, mii, oii) low power images of CD68 alone.Unlike the primary ROI, here only a small effect for age was detected (univariate ANOVA with age, housing, and sex as fixed factors; main effect of age: F [3, 48] = 2.886, p = .045):no other main effects of housing, sex, or interactions were observed.Pairwise comparisons revealed no significant differences between any groups (see right two columns of Table 1 for descriptive statistics).The percentage of the ROI positive for CD68 compared with the rest of the dLGN ("percentage area", expressed as a fold change ratio [ROI/dLGN]) was also used as a means of assessing localized changes in microglial activity (see Methods).This analysis revealed an age Â housing interaction (univariate ANOVA with age, housing, and sex as fixed factors: F [3, 48] = 3.384, p = .026)for the ROI (Figure 2b).Subsequent pairwise comparisons indicated that fold change values for CD68 "percentage area" were significantly greater in EE-P21s than SE-P21s ( p = .024)and in EE-P25s than SE-P25s (p = .006),but not detectably different between EE-P18s and SE-P18s ( p = .406),nor between EE-P30s, and SE-P30s ( p = .676;see Table 2 for relevant descriptive statistics; Figure 2f).Within each housing cohort a significant effect of age was detected for EE (univariate ANOVA with age as fixed factor: F [3, 48] = 5.291, p = .003),but not SE groups (F [3, 48] = 1.373, p = .262).Subsequent pairwise comparisons (Bonferroni corrected) in the EE cohort indicated that F I G U R E 1 (Continued) fold change values were significantly greater in EE-P25s compared with EE-P18s ( p = .004)and EE-P30s ( p = .024).No significant differences were found between EE-P25s and EE-P21s ( p = 0.380), between EE-P18s and EE-P21s (p = .527)or EE-P30s (p = 1.000),

Figure
FigureS2), sex has not been included as a factor in subsequent analyses.
No main effect of housing (univariate ANOVA with age and housing as fixed factors: F[1, 12] = 0.453, p = .514)or age (F [1, 12]T A B L E 1 Descriptive statistics for average CD68 positive particle size fold change ratio measurements within ROI (left 2 columns) and cROI (right 2 columns; see Figure2for more details).

3. 3 |
Figure 1oi).No visually obvious impact of EE (from birth) on ipsilateral retinal inputs to the dLGN was detected in P18 Ten-m3 KO mice, however (n[EE-P18s] = 6; Figure4b; also Figure1ci).Omnibus testing of main effects for dLGN area showed no age Â housing interaction (univariate ANOVA with age and housing as fixed factors: F [1, 20] = 0.303, p = .588),nor any effects of housing (F [1, 20] = 1.845, p = .190),but did show an effect of age (F [1, 20] = 9.275, p = .006).Subsequent pairwise comparisons within each housing cohort indicated that dLGN area was significantly larger in EE-P30s than EE-P18s (p = .019).The mean was greater in SE-P30s compared with SE-18s ( p = .093)showing a similar pattern as expected, although this did not reach significance (see Table4 for nificant corrective pruning has not yet occurred by P18, a timepoint before any EE-driven localized CD68 increase, or any EE-driven F I G U R E 4 Significant pruning of mismapped ipsilateral retinal inputs in the Ten-m3 KO visual thalamus following EE from birth is detected at P30 but not at P18. (a-d) Coronal sections through the rostral dLGN of SE and EE P18 (a, b) and P30 Ten-m3 KO mice (c, d) following anterograde transport of Alexa Fluor conjugated CTB (yellow) from the ipsilateral retina.The dLGN has been outlined and divided in thirds along its long (dorsomedial-ventrolateral) axis (i.e., the perpendicular lines).In SE and EE P18 (a, b) and SE-P30 mice (c), ipsilateral terminals are distributed in a narrow band which approaches the ventrolateral border of the dLGN.In EE-P30 mice (d), however, the ipsilateral projections do not extend as far ventrolaterally as they do in P18 and SE-P30 mice (arrow).Scale bar: 100 μm.(e-i) Quantitative data from P18 and P30 Ten-m3 KO mice either standard-housed (SE) or environmentally enriched (EE) from birth.(e) Analysis of the total area of the dLGN showed an effect of age: dLGN area was significantly greater in EE-P30s than EE-P18s (p = .019).(f) A significant reduction in the percentage of the dLGN area occupied by ipsilateral terminals in EE-P30s compared with SE-P30s (p < .001)and EE-P18s ( p < .001) was detected.(g, h) No differences in the %dLGN area of ipsilateral terminals targeting dorsal dLGN (G) or the middle dLGN (h) was observed.(i) In the ventral dLGN, however, EE-P30s were found to have significantly less %dLGN ipsilateral label compared with SE-P30s ( p < .001) as well as EE-P18s (p = .004).(e-i) Results shown are of pairwise comparisons.Intensely colored circles represent data values from (aggregate) independent samples (see Methods).Faint gray data points, grouped in columns according to sample identity, represent data values from single section measurements.Six animals per housing condition for both the P30 and for P18 age cohorts were used.Mean bars are depicted.Error bars show standard error of the mean (SEM).CTB, cholera toxin subunit β; DM, dorsomedial; P, postnatal day; VL, ventrolateral.*p < .05,**p < .01,***p ≤ .001.T A B L E 4 Descriptive statistics for measures of dLGN area (left 2 columns) and total ipsilateral retinal terminal label (right two columns) in Ten-m3 KO mice.Age dLGN area (μm 2 ) Total ipsi.Label (%dLGN area) 4 | DISCUSSIONThis study has shown that the enhanced life experience provided by early exposure to EE can drive targeted microglial engulfment of aberrant projections during a defined postnatal window.In the retinogeniculate pathway of the Ten-m3 KO mouse, this period was marked by EE-induced localized changes in the CD68 labeling profile specifically in the region of the most profoundly miswired retinal inputs, which commenced between P18 and P21, peaked at around P25, and had ceased by P30.EE-driven microglial engulfment of the miswired retinal inputs was observed at the peak of this window, but not just before it commenced, nor in a control region.These results correlated well with the time-course of EE-induced corrective pruning of labeled retinal inputs, as determined anatomically.The temporal specificity of the EE-induced microglial reactivity and engulfment provides a likely cellular mechanism for the observed critical period to drive normative changes in an aberrantly wired neural circuit.4.1 | Early exposure to EE triggers microglia to perform corrective pruning (via engulfment) and drive normative change

4. 4 |
Maturation of corticothalamic feedback may contribute Enhanced experience promotes neural activity.It is therefore possible that EE facilitates activity-dependent mechanisms to detect and promote targeted microglial removal of terminals which do not align with their contralateral counterparts.An important component of this capacity could relate to other developmental changes occurring in the dLGN.One candidate is the maturation of feedback from layer VI of primary visual cortex (V1) which has recently been identified as critical for normal WT retinogeniculate development.Inhibiting this cortical feedback in WT mice can trigger aberrant recruitment and/or maintenance of contralateral retinal inputs onto dLGN relay neurons (Li Wood: Conceptualization, Methodology, Formal analysis, Investigation, Writing and editing, Data visualisation and figure creation.Claire S. Goldsbury: Resources, Methodology, Writing and editing, Supervision.Atomu Sawatari: Figure creation, Writing and editing.Catherine A. Leamey: Conceptualization, Writing -Reviewing and Editing, Supervision, Project administration.

Table 4
of the total dLGN area, an age Â housing interaction (univariate ANOVA with age and housing as fixed factors: F [1, 20] = 11.568,p= .003)wasdetected.Subsequent pairwise comparisons showed that ipsilateral terminals took up significantly less area in EE-P30s compared with SE-30s ( p < .001),butthere was no difference between EE-P18s and SE-18s ( p = .527;seeTable4 for descriptive

Table 5
within age groups indicated significantly less ipsilateral label in EE-P30s than SE-P30s ( p < .001),butnodifference between EE-P18s and SE-P18s ( p = 0.188; see Table5for descriptive statistics; Presented are group means ± standard deviation of the measurements.See also Figure4and text for more details.n = 6 for all groups.Mean ± standard deviation are presented.