PA28α overexpressing female mice maintain exploratory behavior and capacity to prevent protein aggregation in hippocampus as they age

Abstract With age, protein damage accumulates and increases the risk of age‐related diseases. The proteasome activator PA28αβ is involved in protein damage clearance during early embryogenesis and has demonstrated protective effects against proteinopathy. We have recently discovered that adult female mice overexpressing PA28α (PA28αOE) have enhanced learning and memory, and protein extracts from their hippocampi prevent aggregation more efficiently than wild type. In this study, we investigated the effect of overexpressing PA28α on aging using C57BL/6N×BALB/c F2 hybrid mice. We found that the hippocampal anti‐aggregation effect was maintained in young adult (7 months) to middle‐aged (15 months) and old (22 months) PA28αOE females. While the PA28αOE influence on learning and memory gradually decreased with aging, old PA28αOE females did not display the typical drop in explorative behavior—a behavioral hallmark of aging—but were as explorative as young mice. PA28αOE lowered PA28‐dependent proteasome capacity in both heart and hippocampus, and there was no indication of lower protein damage load in PA28αOE. The life span of PA28αOE was also similar to wild type. In both wild type and PA28αOE, PA28‐dependent proteasome capacity increased with aging in the heart, while 26S and 20S proteasome capacities were unchanged in the timepoints analyzed. Thus, PA28αOE females exhibit improved hippocampal ability to prevent aggregation throughout life and enhanced cognitive capabilities with different behavioral outcomes dependent on age; improved memory at early age and a youth‐like exploration at old age. The cognitive effects of PA28αβ combined with its anti‐aggregation molecular effect highlight the therapeutical potential of PA28αβ in combating proteinopathies.


| INTRODUC TI ON
As we age, damage to proteins and other macromolecules accumulates and causes deterioration of cellular and organ functions.
Means to ensure a healthy proteome through maintained proteostasis is envisioned to prevent age-related disease and to prolong life span (Basisty et al., 2018;Kennedy et al., 2014;Koga et al., 2011;Lopez-Otin et al., 2013). PA28αβ is a protein complex with interesting proteostatic effects in this context. During early embryogenesis in mice, cells rid themselves of aging-related deleterious protein damage (Hernebring et al., 2006), and this process requires PA28αβ (Hernebring et al., 2013). PA28αβ is important for recovery from re-oxygenation injury in the mouse heart (Li, Horak, et al., 2011), and its overexpression increases viability upon oxidative stress of both neonatal rat cardiomyocytes (Li, Powell, et al., 2011) and mouse embryonic fibroblasts (MEFs; Pickering & Davies, 2012).
In a phenotypic screen of mice overexpressing PA28α (PA28αOE), we found that PA28αOE females exhibit improved learning and memory (Adelöf et al., 2018). Multiple parameters demonstrated this effect, including both learning in the shuttle box passive avoidance test and intersessional habituation in an open-field test. These improved cognitive functions correlated to a decreased depressive-like behavior, as PA28αOE female mice also displayed a marked increase in mean active time in the forced swim test (Adelöf et al., 2018).
In this work, we investigate the effects on aging of PA28α overexpression, using the C57BL/6N×BALB/c F2 hybrid background (herein abbreviated F2 hybrid). Hybrid mice are preferable to use in aging studies, because (i) compared to inbreds, they are less prone to develop strain-specific diseases at old age, (ii) they do not display strain-specific behavior as inbreds, and (iii) the heterozygosity of F2 hybrids reflects heterogenetic populations better than inbreds (Adelöf et al., 2019;Miller & Nadon, 2000). The timepoints used to analyze progression of biochemical and behavioral changes are 7 months of age, which represent fully grown mature adults, 15-month-olds that are considered middle-aged, and 22 months of age that constitute the old cohort (Adelöf et al., 2019;Flurkey et al., 2007). PA28α overexpression has been shown to stabilize PA28β at the protein level (Li, Powell, et al., 2011), and this applies to all PA28αOE tissues successfully examined for PA28β protein content: MEFs, striatum/frontal cortex (Adelöf et al., 2018), and eye lens .
PA28αβ is involved in proteasome-dependent degradation, binding to and activating the proteolytic 20S proteasome core. However, the PA28αβ-dependent 20S proteasome activity was not induced in PA28αOE mice. Instead, protein extracts of hippocampus from female PA28αOE mice exhibited enhanced ability to prevent aggregation (Adelöf et al., 2018), indicating a chaperone-like function of PA28αβ in hippocampus, which previously has been found also in reticulocyte lysates (Minami et al., 2000). Hippocampus is central for memory formation and spatial navigation, and age-related hippocampal degeneration is generally considered an important cause of cognitive decline at advanced age. We were therefore particularly interested in whether the reduced protein aggregation in hippocampus is maintained in PA28αOE females as they age; and if so, which effects this would have on markers of aging, life span, and health span.

| Enhanced aggregation prevention in female PA28αOE hippocampal extracts is maintained with aging
As outlined in the Introduction, we previously found that protein extracts of hippocampus from 7-month-old (mature adult) female PA28αOE mice exhibit an enhanced ability to prevent aggregation compared to wild type (Adelöf et al., 2018). Littermates to these mice were followed in a lifespan study, and cohorts were analyzed in the same way as the 7-month-olds, at the ages of 15 months (middle-aged) and 22 months (old). We found that the increased capacity of aggregation prevention in hippocampal extracts from PA28αOE female mice was maintained with aging (p = 0.0187, Mixed-effects model; Figure 1a). In male mice, we observed no difference between wild type and PA28αOE (ns, Mixed-effects model; Figure 1b).

|
The improved aggregation prevention of female PA28αOE hippocampus is not due to increased levels of Hsp90, Hsc70, Hsp40, or altered serum β-estradiol Aggregation of heat-denatured luciferase is blocked in vitro by Hsp90 binding, and-in addition to Hsp90-the subsequent refolding requires Hsc70, Hsp40, and PA28αβ (Minami et al., 2000 and references therein). PA28αβ is implicated to provide a transitory binding site during transfer of the substrate from Hsp90 to Hsc70 in the early events of refolding (Minami et al., 2000). An upregulation of Hsp90, Hsc70, or Hsp40 in PA28αOE female hippocampi could explain the improved prevention of luciferase aggregation, and thus, we analyzed the levels of these proteins in hippocampal extracts.
While the amount of PA28α protein was increased fivefold in the hippocampus of PA28αOE mice (Figure 1c), there was no difference with genotype or age in protein levels of Hsp90 (Figure 1d), Hsc70 (Figure 1e), or Hsp40 ( Figure 1f). These results identify PA28αβ as a limiting factor in handling heat-inactivated luciferase in female-but not male-hippocampi. The sex discrepancy of PA28αOE proteostatic effect points toward a possible involvement of sex hormones. In addition, estrogens are known to impact hippocampal functions (Liu et al., 2008), and thus, we wanted to analyze whether PA28αOE females had altered estrogen signaling. However, we found that serum β-estradiol levels and S105 phosphorylated estradiol receptor β levels did not differ between PA28αOE and wild type females in any of the cohorts ( Figure S1), indicating that the chaperone-like effect observed in PA28αOE female hippocampi is likely not linked to estradiol signaling.

| Heart and hippocampus of PA28αOE mice exhibit decreased PA28-20S activity
Because of the protective effects of PA28α overexpression upon cardiomyopathy and re-oxygenation stress of the heart (Li, Horak, et al., 2011;Li, Powell, et al., 2011), we focused on heart, as well as hippocampus, for the further investigation of biochemical effects of PA28α overexpression during aging.
The proteasome system consists of the 20S proteasome core that interacts with different activators regulating substrate access to 20S proteolytic chamber. The capacity of proteasome complex peptidase activity can be assayed using extraction and assay conditions that optimize the complex composition of interest (Hernebring, 2016). We unexpectedly found that PA28α overexpression decreased PA28-20S proteasome capacity both in heart and hippocampus (p heart = 0.0018, Figure 2a; p hippocampus = 0.0024, Figure 2b; Mixed-effects model), but had no effect on peptidase activities of while 26S and 20S activity did not change with aging in either heart or hippocampus. Furthermore, on comparing heart PA28α protein levels in PA28αOE and wild type, we can conclude that PA28α was increased sevenfold in PA28αOE and that the age-related induction of PA28-20S proteasome capacity in heart was not caused by cumulative PA28α protein levels in this organ ( Figure 2g).

| PA28α overexpression does not reduce damage load in heart or hippocampus
PA28αβ has been repeatedly linked to protein damage control.
Protein damage is expected to increase with aging and is generally considered at the core of age-related organ deterioration (Anisimova et al., 2018;Koga et al., 2011;Levine, 2002). Thus, we assayed protein carbonylation, a common reporter of oxidized proteins, and the deleterious protein-adduct metabolic byproduct advanced glycation end products (AGEs; specifically N ɛ -carboxymethyllysine, CML) in the tissues of aging PA28αOE mice.
We found that protein damage increased with aging in heart (p carbonyls = 0.0003; Figure 3a, p CML < 0.0001; Figure 3c, Mixedeffects model), but not in hippocampus, for either protein carbonylation ( Figure 3b) or CML (Figure 3d). In addition, PA28α F I G U R E 1 Aggregation prevention capacity of hippocampal protein extracts from (a) female mice at 7, 15 and 22 months of age and (b) male mice at 4, 15 and 22 months of age (all from F2 hybrid lifespan analysis except male 4-month-olds which were C57BL/6N). Turbidity reduction is a measure of how much the presence of hippocampal protein extracts from wild type (WT) and PA28αOE can prevent aggregation of heat-sensitive luciferase at 42°C (*p = 0.0187; Mixed-effects model).
Values are mean ± SEM; n = 3-4. Protein levels of (c) PA28α, (d) Hsp90, (e) Hsc70, and (f) Hsp40 in hippocampal protein extracts from 4, 15, and 22 months old PA28αOE and wild type mice (all from F2 hybrid lifespan analysis except 4-montholds which were C57BL/6N). Values are mean ± SD; n = 4 except for Hsp40 n WTF4 = 3. Insets are representative western blots of the different ages (cropped from the same membrane for each analysis) least not in heart or hippocampus, which was unexpected considering its proposed role in degrading protein damage.

| PA28α overexpression in PA28αOE does not prolong life span
While protein damage often correlates to life span (Anisimova et al., 2018), an ability of tissues to resist protein aggregation upon aging could possibly confer proteostatic effects benefitting longevity, even though the levels of protein damage is unaltered. However, there was no effect of PA28α overexpression in the PA28αOE mouse model on median or maximum life span, in either males or females ( Figure 4; Table 1). Neither did analyses of body size, weight, body composition, and core temperature display any difference between wild type and PA28αOE (Table S1).

| Enhanced memory and anti-depressive-like behavior in PA28αOE females are not maintained as they age
In the 7-month-old female PA28αOE mice, we observed improved memory compared to wild type, both in the shuttle box passive avoidance test and in habituation, as well as a reduced depressivelike behavior using the forced swim test ( Figure 5; Adelöf et al., 2018). At 22 months of age, however, PA28αOE and wild type females performed similarly in tests of memory (Figure 5a), habituation ( Figure 5b,c), and depressive-like behavior ( Figure 5d). We also found that the reduced depressive-like behavior of PA28αOE at the 7-month timepoint was not caused by altered fat mass (Figure 5e), which has been shown to directly influence swimming behavior in the forced swim test (Adelöf et al., 2019). In males, there were no differences between PA28αOE and wild type in memory, habituation, or mobility at 7 months of age (Adelöf et al., F I G U R E 2 PA28-20S proteasome capacity in (a) heart (****p 7-22 < 0.0001, **p W T-OE = 0.0018; Mixed-effects model) and (b) hippocampus (**p W T-OE = 0.0024; Mixed-effects model), 26S proteasome capacity in (c) heart and (d) hippocampus, and 20S proteasome capacity in (e) heart and (f) hippocampus; all from F2 hybrid lifespan analysis. Values are mean ± SEM; heart: n = 11-12, hippocampus: n = 5-6. PA28α overexpression in (g) heart from PA28αOE and wild type (WT) mice from F2 hybrid lifespan analysis. Images below graph are representative western blots of protein extracts from the different ages (cropped from the same membrane) 2018), and similar results were seen at the 15 and 22 months' timepoints ( Figure S2).

| Old female PA28αOE mice display a youthlike explorative behavior
During mouse aging, exploratory activity is known to markedly decrease (Adelöf et al., 2019;Fahlström et al., 2011). The inclination to explore can be measured by analyzing the mobility of an animal during the first few minutes when introduced to an unfamiliar environment. In the 7-month-old female mice, we found no behavioral differences between PA28αOE and wild types during naïve open-field testing ( Figure 5; Adelöf et al., 2018). In the 22 months' cohorts, wild types displayed as expected an "aged" behavior, with horizontal activity reduced by 54% compared to 7-month-olds during the first 5 min in the open field test (p = 0.0006, Student's t test; Figure 5f). In contrast, female PA28αOE 22-month-olds were as active as 7-month-olds during the first 5 min and thus maintained their exploratory behavior with aging to a markedly higher degree than wild type (p = 0.036, Student's t test; Figure 5f).

F I G U R E 3
Protein damage levels in heart and hippocampus from young adult, middle-aged and old wild type and PA28αOE mice. Carbonylated proteins in (a) heart (****p 7-22 < 0.0001) and (b) hippocampus from PA28αOE and wild type (WT) mice. N ɛ -carboxymethyllysine (CML) in (c) heart (**p 7-22 = 0.003, Mixed-effect model) and (d) hippocampus. All tissues were from the F2 hybrid lifespan analysis except 4-month hippocampi that were from C57BL/6N. Values are mean ± SEM; heart: n = 11-12, hippocampus: n = 7-8. Images to the right of graphs are representative western blots of protein extracts from the youngest and oldest cohorts (size markers are approximations based on pre-stained protein standards) Lifespan analysis of PA28αOE and wild type C57BL/6N × BALB/c F2 hybrid mice displayed as interval of minimum and maximum survival curves. Lifespan of (a) female and (b) male PA28αOE mice (interval between dashed lines) and wild type mice (WT, interval between solid lines). Timepoints for behavioral analyses and organ harvesting of naïve cohorts of mature adult (7 months), middle-aged (15 months), and old (22 months Values are mean ± SEM; n WT7F = 9-10, n OE7F = 6, n WT15F = 12, n OE15F = 10, n WT22F = 7, n OE22F = 8-13 Standing on hind legs-rearing-is considered both exploration and vertical mobility (Tanaka et al., 2012). Age effects on rearing activity during the first 5 min in the open-field test displayed trends similar to that of exploratory horizontal activity; 22-month-old PA28αOE females reared as often as 7-month-old wild type and PA28αOE females, which was about twice the frequency of wild types at 22 months of age ( Figure 5g).
Exploratory behavior also decreased with aging for males, especially in wild type ( Figure S3;

| DISCUSS ION
We herein demonstrate that the aggregation prevention effect in hippocampus of PA28α overexpressing young adult female mice is maintained as they age. These animals displayed a youth-like behavior at advanced age; specifically, 22-month-olds explored unfamiliar environments to the same extent as 7-month-olds. This is indicative of a healthy aging phenotype. As young adults, PA28α overexpressing females had improved learning/memory and reduced depressive-like behavior. These cognitive effects of PA28α overexpression gradually decreased with age, and there were no differences in these parameters comparing 22-month-old wild type and PA28αOE mice. We neither observed any effects on life span or protein damage aging markers. Our results indicate that the elevated capacity to prevent aggregation in hippocampus in PA28αOE females exerts proteostatic effects that cognitively manifests in different ways as the mouse ages: as improved memory in young, and enhanced exploratory behavior in old mice.
The known roles of PA28αβ are schematically summarized in Figure 6; on one hand, PA28αβ exhibits proteasome activation and on the other hand chaperone-like/anti-aggregation functions. As laid out below, the molecular shift between these two roles is likely intricately regulated to maintain cellular functions and protein homeostasis though challenges of metabolic stress, inflammation, and aging. PA28αβ as a proteasome activator (function A in Figure 6) is important for generating (at least some) immunopeptides for presentation at the major histocompatibility complex I (MHC-I; Cascio, 2014;Vigneron & Van den Eynde, 2014), and for protein damage control during early embryogenesis, possibly constituting a proteome rejuvenation check-point (Hernebring et al., 2006(Hernebring et al., , 2013. PA28αβ as a chaperone, however, (function B in Figure 6), can protect against desmin-related cardiomyopathy (DRC), a proteinopathy induced by the missense αB-crystallin mutation CryAB R120G (Li, Horak, et al., 2011). The reasons for why we categorize PA28αβ as exhibiting a chaperone function against DRC is outlined in previous work (Adelöf et al., 2018). Finally, as shown herein, PA28αβ's chaperone-like functions promote memory and learning at young age and exploration at old age.
That we found no effect of PA28α overexpression on life span or age-related protein damage was initially unexpected since PA28αβ has been repeatedly associated to the degradation of proteins carrying such deleterious modifications (Hernebring et al., 2013;Pickering & Davies, 2012). Notably, the PA28αβ variant with chaperone-like F I G U R E 6 Schematic presentation of PA28αβ's known roles as proteasome activator and as a chaperone. PA28αβ has been linked to diverse physiological functions, including peptide generation in the immune system, protein damage control during early embryogenesis, heart homeostasis (e.g., protection against desmin-related cardiomyopathy, DRC), and-as presented in this workcognitive functions. The molecular basis of PA28αβ's role is likely dual and can be separated into its 20S proteasome stimulating (Function A) and chaperonelike (Function B) activity, possibly orchestrated by PTMs. These two molecular roles seem to act in different physiological functions, as outlined in the figure function is likely exclusively overrepresented in PA28αOE, and overexpression of the 20S proteasome activating PA28αβ may still affect proteostasis in a manner that could lead to prolonged life span.
Such mechanism could theoretically apply to humans as well, since deleterious protein damage (CML) has been shown to dramatically decrease also as human embryonic stem cells (hESCs) differentiate (Barandalla et al., 2017). However, no one has yet to our knowledge investigated PA28-20S proteasome activity during differentiation of hESCs.
Apart from changes in protein expression, activity of proteasome complexes is known to be regulated by alterations in complex composition and by posttranslational modifications (PTMs) of both 20S core and its activators (reviewed in Kors et al., 2019). As indicated in Figure 6, the function of PA28αβ may be regulated at the protein and/or protein complex level by differences in PTMs of PA28αβ's two functional variants, and this PTM is possibly phosphorylation.
An early study reported that phosphorylation of PA28 is required for proteasome activation, as phosphatase treatment abolished PA28-dependent proteasome activity in rabbit reticulocyte extracts (Li et al., 1996). This discovery has been overlooked since in vitro proteasome capacity studies are commonly performed using re- Levels of protein carbonyls vary greatly between brain regions in mice. Dubey and colleagues found that protein carbonyls increased with aging of C57BL/6 mice by 30% in hippocampus and 85% in striatum, while there was no difference at all in carbonyl levels with age in hindbrain (Dubey et al., 1996). A 50% increase in protein carbonyls in hippocampus with age in C57BL/6 mice (Choi et al., 2004) and a 200% increase in CML levels in C57BL/6JNia mice (Thangthaeng et al., 2008) have also been reported.
Though reports of unchanged protein damage levels in hippocampus during aging are scarce, our results anyhow indicate an innate difference between how hippocampus and heart age. This notion is supported by a recent study comparing gene expression of age-related biological pathways in different rat tissues, in which hippocampus showed a distinctive transcriptomic profile. In addition to displaying the smallest fold change of upregulated genes, hippocampus demonstrated unaltered gene expression in pathways associated to mitochondrial function that were the most downregulated by aging in other tissues (Shavlakadze et al., 2019). This may reflect that hippocampus is able to uphold mitochondrial integrity and sustain proteostasis upon aging much more efficiently than other tissues.
Such situation would result in a postponed accumulation of protein damage upon aging. Protein carbonyls have been shown to accumulate with age in tissues in a biphasic manner, with an exponential increase in roughly the last third of life span (Levine, 2002). The 22month timepoint represents animals with around 1/6 left of their life span, and it is possible that the shift in protein damage accumulation rate occurs even later than this in hippocampus.
Heart and hippocampus also differed in proteasome activity, considering that PA28-20S activity increased with age in heart (by 67%) but not in hippocampus. Remarkably, the increase in PA28-20S activity with age in heart occurs without an induction of PA28α levels, strengthening the hypothesis of an unknown molecular regulation behind PA28αβ's dual function as outlined in Figure 6. Proteasome activity is generally considered to decrease with age, though results are conflicting (as reviewed; Koga et al., 2011). However, it is important to note that proteasome activity in cell extracts may be distinct from that in tissue, since aging may induce ATP deficiency and/or protein aggregate interactions that can hamper the proteolytic machinery in vivo and these could be reversed by protein extraction.  (Davis et al., 2004). In rodents, differences in exploratory behaviors have been linked to changes in gene expression in several brain regions (such as the prefrontal cortex, amygdala, and periaqueductal gray; Nelovkov et al., 2006) and specific neuromodulatory responses are known to be induced only by novel stimuli (Schomaker & Meeter, 2015). Thus, exploratory activity differs in the neuronal activation as compared to learning and memory, resulting in a distinction of the two behavioral effects of PA28αOE mice from a neuronal response perspective.
The enhanced hippocampal chaperone-like capacity observed in female but not male PA28αOE mice could explain the cognitive effects in PA28αOE females. Although we do not know the biological cause to this sex discrepancy, it is most likely not directly linked to sex hormonal factors, since we found no differences in the levels of serum β-estradiol or S105 phosphorylated estradiol receptor β between PA28αOE and wild type females in any of the cohorts. There may be indirect effects, and there are sex-specific differences in the brain (McCarthy, 2008); female rodents have for example higher density of dendritic spines in hippocampus than males, which changes for the opposite under stressful conditions when males instead obtain greater density of dendritic spines (Shors et al., 2001). However, the increased exploratory behavior of PA28αOE males compared to wild type males at 15 months of age demonstrates that PA28αOE has an effect also in males, although not as pronounced as in females. This effect could in theory also be linked to a chaperone-like function, albeit present in other brain regions than hippocampus or in a context dependent manner not detected at termination of these animals.

| Animal care and termination
All mice were housed with a 12:12 h light-dark cycle (dawn: 5.30-6.00, dusk: 17.30-18.00) at room temperature (21°C) with controlled humidity (45%-55%) under daily surveillance. Water and regular chow diet were given ad lib (R3; energy percentage: 12% fat, 62% carbohydrates, and 26% protein and a total energy content of 3 kcal/g. Lactamin). Female mice were initially cohoused four by four, but the number decreased as mice died or were euthanized over time in the lifespan experiment. Males were housed in groups of three but changed to single housing at 6 months of age due to aggression and fighting. To maintain representative male cohorts and to not select for dominant males, eleven cages with signs of fighting were removed from the study. Cages were cleaned every week (cohoused animals) or every 2 weeks (single housed animals), and nesting material (paper, cardboard houses, and wooden sticks) was transferred to new cages upon cleaning.
Euthanization took place under anesthesia (5% isoflurane) by decapitation. In the lifespan study, animals were euthanized upon signs of ill-health (e.g., weak posture, inactivity, shabby fur, failure to eat or drink, enlarged organs and tumors; in line with Adelöf et al., 2019).
At termination of mice in the behavioral study cohorts, organs and tissues were collected and directly transferred to dry ice, and kept frozen in −80°C until biochemical analyses. Animal work was carried out in accordance with EU Directive 2010/63/EU for animal experiments, and the study was performed following the ethical certificate approved by the Animal Ethics Committee in Gothenburg, Sweden (Permit Number: 164-2015).

| Study design of lifespan analysis and behavioral phenotypic profiling
The lifespan study was carried out with a total of 200 male and female wild type and PA28αOE mice (n WTfemales = 51, n OEfemales = 60, n WTmales = 48, n OEmales = 41). Cohorts of mice from this lifespan study were subjected to behavioral profiling and organ harvesting at 7, 15, and 22 months of age, when mice were considered mature adults, middle-age and old, respectively (Adelöf et al., 2019;Flurkey et al., 2007). At the time for phenotypic analyses, the mice were exactly 6.6-7.8 ± 0.2 (age at test period start -age at test period end ±SD of age variance in the cohort; 7 months), 14.5-15.6 ± 0.1 (15 months), 21.8-22.5 ± 0.2 (22 months) and no attempt to control for seasonal effects was made (7 months: September-October; 15 months: May-June; 22 months: late December-January). These cohorts of mice were included in the lifespan analysis and marked as censored at the timepoint of first behavioral assessment. To estimate natural life span, we calculated intervals between two survival curves. The minimum survival curve was obtained by counting euthanized animals as if they had died from natural causes and the maximum survival curve by counting these mice as if they were as healthy as their littermates (as described in Adelöf et al., 2019).
Mice were acclimatized for 1-week prior initiation of studies, and behavioral analyses known to be affected by handling (such as activity box) were performed early in the testing period (except for females at 15 months in the activity box analysis which therefore was omitted from data analysis) and tests which affect the mice to a greater extent were performed later in order to give the animals longer recovery time and to not impact subsequent analyses. To ensure complete separation of the sexes and no sexdependent impact of the analyses, the order of the behavioral test-battery differed between females and males. Behavioral tests were single-blind and carried out at daytime (between 10.00 and 14.00) after the mice had been given at least 1 h of acclimatization in the experimental room. Data from the animals from the 7-month timepoint have previously been published (Adelöf et al., 2018) as well as all wild type mice (Adelöf et al., 2019), and an analysis of cataract incidence of all animals in the lifespan study .

| Activity box
General activity, exploratory behavior, and habituation can be meas-

| Forced swim test
The forced swim test analyses signs of depression, generally re-

| Shuttle box passive avoidance test
Learning and memory was assessed though passive avoidance analysis in the shuttle box system (Accuscan Instruments Inc.

| Blood serum preparation and βestradiol detection
Blood was collected from the left atrium of the heart under isoflurane anesthesia, prior to necropsy, incubated at room temperature for 30-45 min, centrifuged to isolate serum, which was stored at −80°C until analysis. β-estradiol levels were detected by the Mouse/ Rat Estradiol ELISA-kit (#ES180S-100; Calbiotech) according to manufacturer's instructions. The β-estradiol serum levels are presented as relative values for each separate timepoint, as we found that both kit performance and β-estradiol in serum at −80°C were not stable over time (months).

| Analyses of protein expression and protein damage markers by SDS-PAGE and Western blot
Heart (left ventricle) and left hippocampus were manually grinded in Eppendorf tubes and lysed in a modified RIPA buffer (50 mM Na 2 HPO 4 pH7.8; 150 mM NaCl; 1% Nonidet P-40; 0.5% deoxycholate; 0.1% SDS; 1 mM DTPA; 1 mM 4-(2-Aminoethyl)benzenesulfonyl fluoride hydrochloride, AEBSF). Debris was removed by centrifugation at 5000 g (heart) and 10,000 g (hippocampus) for 10 min at 4°C, and protein concentration of the cell extracts was determined using the Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific). There was no difference in hippocampi protein carbonyl content when comparing centrifugation at either 5000 g or 10,000 g as demonstrated in Figure S4, confirmed with ImageJ quantification.
Samples were prepared in loading buffer (100 mM Tris-HCl, 50% Ponceau membrane staining. There were no differences between the sexes in levels of protein carbonyls and CML ( Figure S5), and therefore, the data from male and female tissue were pooled in the statistical analyses.

| Luciferase aggregation prevention
Hippocampus were analyzed for luciferase aggregation prevention capacity as previously described (Adelöf et al., 2018). Specifically, right hippocampi were manually grinded in Eppendorf tubes and lysed in extraction buffer (25 mM Tris/HCl pH 7.8, 100 mM NaCl, 5 mM MgCl 2 , 1 mM ATP, 1 mM AEBSF, and 5% glycerol), and cell debris was removed by centrifugation at 5000 g for 10 min followed by addition of 1 mM DTT after an aliquot was taken for protein concentration determination with the BCA Protein Assay kit (Pierce, Thermo Fisher Scientific). Luciferase (200 nM; SRE0045, Sigma-Aldrich) was denatured at 42°C in 50 mM Tris (pH 7.6) and 2 mM EDTA, in the presence of 5 µg hippocampal protein extracts (4.5 µg for females 7 months) and aggregation of luciferase was measured as light scattering at 340 nm. Sample wells without protein extract were used as positive control (reference indicating maximum aggregation). Protein samples without luciferase were used as negative control of aggregation and extraction buffer only was considered background, neither of these controls changed over time. The chosen timepoint for analysis was when the positive control (without protein extract) had reached 60%-80% of maximum aggregation.
The luciferase aggregation prevention capacity of the hippocampus extracts was calculated as percent non-aggregated luciferase.

| Statistical analysis
All groups were analyzed for normal distribution by Shapiro-Wilk and QQ plots and homogeneity of variance by Levene's test or homoscedasticity plots. Two-tailed independent t test (Student's t test) was used for groups which passed test of normality and equal variance, two-tailed independent unequal variance t test (Welch's test) for normal distributed groups that did not have equal variance, and nonparametric Mann-Whitney test for unevenly distributed groups.
Differences between wild type and PA28αOE at the three different timepoints, age effects, and time × group interactions were calculated with 2-way ANOVA, followed by Sidak multiple comparisons test, for groups with equal n and Mixed-model effect for groups with missing values. Statistics were calculated using GraphPad Prism 8 and IBM SPSS Statistics 27.

ACK N OWLED G EM ENTS
We thank Anne Petersen for valuable support and discussions; Liselotte Andersson for excellent animal husbandry; Johan Forsström and Michelle Porritt for technical assistance; Mikael Bjursell for contributions on phenotyping design; Richard Miller for valuable input on designing experimental mouse lifespan studies; Elin Blomberg for artistic partaking in Figure 6 and the graphical abstract; Thomas Nyström group for allowing us unlimited access to their Odyssey imaging system. This work was supported by the Swedish Foundation for Strategic Research (SSF; ID14-0087) and AstraZeneca AB. None of these organizations had any role in the design, performance, analysis, or writing of this work.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding authors upon reasonable request.