|Home | About | Journals | Submit | Contact Us | Français|
Increased neuronal glucocorticoid exposure may underlie inter-individual variation in cognitive function with ageing in rodents and humans. 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1) catalyses the regeneration of active glucocorticoids within cells (in brain and other tissues), thus amplifying steroid action. We examined whether 11β-HSD1 plays a role in the pathogenesis of cognitive deficits associated with ageing in male C57Bl/6J mice. We show that 11β-HSD1 levels increase with age in CA3 hippocampus and parietal cortex, correlating with impaired cognitive performance in the water-maze. In contrast, neither circulating corticosterone levels nor tissue corticosteroid receptor expression correlate with cognition. 11β-HSD1 elevation appears causal since male transgenic mice with forebrain-specific 11β-HSD1 overexpression (~50% in hippocampus) exhibit premature age-associated cognitive decline aged 18 months in the absence of altered circulating glucocorticoid levels or other behavioural (affective) deficits. Thus excess 11β-HSD1 in forebrain is a cause of as well as a therapeutic target in memory impairments with ageing.
Chronically elevated glucocorticoid (GC) levels are detrimental to the brain, especially to the hippocampus. In the adult hippocampus GC excess potentiates excitatory neurotransmission, disrupts electrophysiological functions such as long-term potentiation (LTP) thought to underlie memory, interferes with learning and recall, promotes dendritic atrophy and may potentiate neurotoxicity (McEwen et al., 1999). Inter-individual differences in plasma GC levels may underpin variation in cognitive function with ageing (Meaney et al., 1995), higher steroid levels associating with and even predicting subsequent cognitive deficits and hippocampal atrophy in rodents (Yau et al., 2002) and humans (Lupien et al., 1998). Elevated GC levels appear causal of cognitive decline with ageing since manipulations which maintain low GC levels from mid-life (adrenalectomy and low-dose GC replacement (Landfield et al., 1981), antidepressant drug therapy (Yau et al., 2002)) prevent the emergence of cognitive deficits with subsequent ageing, at least in rodents.
GC action on target cells is not dependent merely upon hormone levels in the circulation and the density of intracellular glucocorticoid (GR) and mineralocorticoid (MR) receptors in target tissues, but also on pre-receptor metabolism by 11β-hydroxysteroid dehydrogenases (11β-HSDs) (Holmes and Seckl, 2006). The adult rodent and human brain highly expresses only the type 1 isozyme (11β-HSD1) which catalyses the regeneration of active glucocorticoids (cortisol, corticosterone) from inert 11-keto forms (cortisone, 11-dehydrocorticosterone) in neurons (Rajan et al., 1996), thus amplifying cellular GC action. Importantly, aged mice deficient in 11β-HSD1 (11β-HSD1−/− mice) have reduced intrahippocampal levels of corticosterone despite normal circulating concentrations (Yau et al., 2001). 11β-HSD1−/− mice are protected from the normal decline in memory and hippocampal LTP seen with ageing (Yau et al., 2001; Yau et al., 2007) suggesting that 11β-HSD1 is an important control of intraneuronal GC action in vivo. Hence 11β-HSD1 inhibition is a target for therapy of age-associated cognitive disorders (Wamil and Seckl, 2007). Indeed, in 2 small randomised, double-blind, placebo-controlled trials, an 11β-HSD inhibitor improved cognitive function in elderly men and patients with type 2 diabetes (Sandeep et al., 2004). However, these genetic and therapeutic manipulations affect the whole body and 11β-HSD1 is highly expressed in peripheral organs, notably in liver and adipose tissue (Stewart et al., 1999). The primary role of 11β-HSD1 in the brain per se is not clear-cut. For example, changes in peripheral (hepatic) 11β-HSD1 alone alter brain function, at least at the level of the hypothalamic-pituitary-adrenal (HPA) axis (Paterson et al., 2007).
Whilst elevated 11β-HSD1 selectively in adipose tissue occurs in human and rodent obesity (Livingstone et al., 2000; Rask et al., 2002) and appears a plausible cause of metabolic syndrome (Masuzaki et al., 2001; Masuzaki et al., 2003), any primary role of 11β-HSD1 in the CNS in causation of cognitive variation with ageing is unexplored. Here we examined the critical issues of whether endogenous 11β-HSD1 levels vary with cognitive function in the hippocampus of aged mice, and whether specific elevation of 11β-HSD1 in the forebrain (including the hippocampus) has cognitive consequences with ageing.
Male C57BL/6J mice were purchased at 8-10 weeks old (Harlan UK) and maintained in cages of 3-4 on standard chow (Special Diet Services, Essex, UK; product 801190) and tap water ad libitum (lights on 0700-1900) in our animal facilities until ready for experiment at 6m (young) and 24-27m (aged). All procedures were performed in strict accordance with the United Kingdom Animals (Scientific Procedures) Act (1986).
Basal (0800h) tail venesection blood samples were taken for corticosterone assay. At least one week later, animals were tested in the reference memory watermaze task, first learning to escape to a visible platform (submerged but marked with visible tower block protruding 10cm on top) over 4 consecutive days of non-spatial training (3 trials/day curtains around pool to hide visuo-spatial cues), then a hidden platform over 5 consecutive days of spatial training (4 trials/day, no curtains), essentially as described (Yau et al., 2007). One hour after the last spatial training trial, a 60s probe trial was performed with the platform removed. Swim paths and measures of performance were analyzed by Watermaze software (Actimetrics, Evanston, IL). Aged mice showing signs of motor or visual impairments were excluded (2 aged mice were excluded of 16 tested). Mice were killed by decapitation, the brains removed and snap frozen on soft dry ice for cryostat sectioning and in situ hybridization using [35S]UTP-labelled cRNA antisense probes as described (Mattsson et al., 2003) for 11β-HSD1, MR and GR mRNAs in the anterior hippocampus. Slides were dipped in photographic emulsion (NTB-2, Kodak, UK) and exposed at 4°C for 14-21 days.
The rat 11β-HSD1 cDNA-based minigene, previously employed to overexpress 11β-HSD1 in adipose tissue and liver (Masuzaki et al., 2001; Paterson et al., 2004), was fused in-frame at the C-terminus to the influenza virus-derived HA epitope tag by PCR-mediated site-directed mutagenesis. This was inserted downstream of the CamIIK promoter directing transgene expression to the forebrain to yield CamIIK-HSD1, a 10.5 kb DNA fragment was prepared for micro-injection by agarose gel electrophoresis, electroelution and dialysis against 10mM Tris.HCl/0.1mM EDTA (pH7.4) prior to dilution of DNA to a concentration of 1ng/μl. Construct expression was confirmed by transfection into Cos7 cells and 11β-HSD1 activity measured.
Microinjection into the pronuclei of fertilized C57Bl/6xCBA/C3H F1 embryos was performed using standard techniques. G0 offspring were screened by Southern blot hybridisation analysis of tail biopsy genomic DNA digested with BamHI and probed with α32P-dCTP-labeled rat 11β-HSD1 cDNA to reveal diagnostic restriction fragments. Transgenic lines 3615 and 3621, carrying high and low copies of the transgene, respectively, were propagated from independent founder animals. F7 or greater C57Bl/6J backcross male mice were studied throughout. Mice hemizygous for the transgene (referred to as CamIIK-HSD1 or transgenic (Tg) mice) were compared with non-transgenic (wildtype; wt) littermate controls. Animals were fed standard chow and water ad libitum.
The expression pattern of the transgene was determined by immunohistochemistry, using an antibody directed against the HA tag. Mice (terminally anaesthetised with sodium pentabarbitone) were transcardially perfused with 4% paraformaldehyde, the brains postfixed and frozen. Transgene protein was localised on 60μM coronal cryostat sections using a rabbit anti–hemagglutinin (anti-HA) antibody (71-5500, Invitrogen, Zymed Labs, UK) in conjunction with the streptavidin-biotin based peroxidase staining system (Vector Laboratories, Peterborough, UK). Controls included wildtype littermates. To determine the effect of the transgene on 11β-HSD1 activity, various brain regions were homogenised and assayed as described (Sandeep et al., 2004). After this basic characterisation, studies focused on the highest expressing line (3615) which exhibited hippocampal 11β-HSD1 activity 50% above wildtype. Key data were confirmed in a separate line (3621).
Male CamIIK-HSD1 transgenic (Tg) mice and their wildtype (wt) littermates aged 3-6 months, were housed singly for one week prior to study. Basal blood samples were taken by tail-nick at 0800-0900h for circadian nadir and 1900-2000h for peak corticosterone levels. At least one week later, animals were subjected to restraint stress (10 mins) with samples taken before and 10 and 90 mins after stress. Plasma corticosterone was determined by radioimmunoassay as described previously (Paterson et al., 2007).
Male Tg mice and wildtype mice, young (6-9m) and aged (18m) were accustomed to the behavioural room to minimise stress. Mice were tested for affective behaviours in the open field and elevated plus maze (Holmes et al., 2006) and for learning and spatial memory retention (probe test) in the watermaze as described for C57Bl/6J mice above. Data were accumulated by automated video recording and quantified blind to genotype. In addition, conditioned passive avoidance was assessed in a light:dark box (Step-through passive avoidance apparatus; Ugo Basile, Comerio, Italy) using the following protocol; day 1: 5 mins free exploration of the apparatus; day 2: latency to enter the dark compartment, followed by a mild 0.3mA or moderate 0.5mA 3s footshock in the dark compartment, then retest 5h later for latency to enter the dark compartment. Latencies are measured automatically by the integral software following the opening of the door separating the light and dark components. Old (18m) mice (wt or Tg) failed to respond to the same magnitude of shock (0.3mA) as young mice with an increased latency to re-enter the dark compartment, suggesting insensitivity to mild shock and so were given a more intense footshock (0.5mA).
Differences by genotype were analysed by Student's t-test or 1 or 2-way ANOVA with post hoc Tukey's HSD tests, as appropriate. Significance was set at p<0.05. Data are means±SEMs.
Young (6 month, n=10) and aged (24-27m, n=14) male C57BL/6J mice were tested in the reference memory watermaze task (Fig 1a). Aged mice had slower average swim speeds (young 0.25±0.01m/s; old 0.18±0.01m/s; ANOVA, age effect, F(1,22)=17.9, p=0.0003) but showed similar visible platform learning over 4 days of trials (decrease in escape latency, young 18.58±1.20s; old 21.84±3.44s; F(1,22)=0.6, P=0.4), indicating no difference in perception (vision) or motivation with ageing. While young mice were able to acquire the spatial memory task with 5 days of training, aged mice were impaired in their spatial learning performance [escape latency, ANOVA, age effect, F(1,22)=14.8, P=0.0009; path length, F(1,22)=4.2, P=0.05], failing to acquire the task as confirmed in the probe trial [% time in target quadrant of probe test, young 47.46±2.77; aged 25.4±2.7; ANOVA, age effect, F(1,22)=30.8, P<0.001](Fig 1a).
Hippocampal 11β-HSD1 mRNA levels were significantly increased with ageing selectively in CA3 pyramidal cells (F(1,22)=5.6, P<0.05, Fig 1b). Examination of individual animals revealed that 11β-HSD1 mRNA in CA3 cells correlated with spatial learning and with spatial memory retention [path length to platform on day 5, r=0.41, F(1,22)=4.3, P<0.05; probe test, r=0.44, F(1,22)=5.2, P<0.05; Fig 1c] such that higher 11β-HSD1 mRNA associated with poorer learning and memory. Neither MR nor GR mRNA expression was altered with age in any hippocampal subregion (Fig.S1, Supplementary Material online). In the parietal cortex, 11β-HSD1 mRNA 11β-HSD1 mRNA expression was significantly higher in layer V of aged compared with young mice (F(1,20)=4.4, P<0.05) and correlated negatively with spatial learning (path lengths on day 5, r=0.47, F(1,18)= 5.2, P<0.05) Fig 1c, but not spatial memory retention (probe test) (P=0.3). Plasma corticosterone levels were significantly increased in aged C57BL/6J mice [young, 3.3±12.5 nM; aged, 147.2±18.8 nM, F(1,22)=6.9, P<0.05], but did not correlate with watermaze performance (path length on day 5, P=0.6; probe, P=0.3).
To determine whether the increased brain 11β-HSD1 causes learning and memory deficits, forebrain-specific CamIIK-HSD1 transgenic (Tg) mice (3 lines) and wild-type (wt) littermates were generated. Tg mice exhibited transgene protein in the forebrain, notably in cortex and hippocampus showing an increase in 11β-HSD1 activity of 125% and 50% in these regions, respectively (Fig 2a).
Young (6-9m) Tg mice learned to find and escape onto the hidden watermaze platform as well as wt (Fig 2b; genotype effect [F(1,15)=1.46, p=0.37]. Whilst old (18m) Tg mice performed the cued version of the watermaze as well as wt with comparable swim speeds (wt 0.18±0.01m/s; Tg 0.18±0.02m/s), they were slower to learn the hidden platform location; specifically wt mice, but not Tg mice, significantly improved between the first and second day of training (F(1,14)=3.8, p<0.01; Fig 2b).
In the conditioned passive avoidance test, young Tg and wt mice had similarly increased latencies to re-enter the dark compartment at 5h re-test following footshock [genotype effect, F(1,14)=0.29, p=0.08; Fig 2c]. While old wt mice responded with an increased latency to enter the dark compartment, old Tg mice failed to respond to this conditioning stimulus [genotype effect, F(1,14)=9.8, p=0.01; Fig 2c]. This impaired learning was not due to anxiety-related behaviours (Table S1, Supplementary Material online) or altered basal, circadian or stress-induced glucocorticoid levels (Table S2, Supplementary Material online), mirroring normal HPA axis function in 11β-HSD1−/− mice on this background (Yau et al., 2007).
The key findings in this study are that 11β-HSD1 expression is increased in the hippocampus and cortex with ageing in mice; levels of 11β-HSD1 mRNA in CA3 and cortical layer V pyramidal cells correlates with cognitive function in the watermaze. Modest transgenic overexpression of 11β-HSD1 in the forebrain produces accelerated cognitive dysfunction with ageing, providing the first evidence that forebrain 11β-HSD1 levels alone are sufficient to alter cognitive behaviour in ageing mice, an effect not due to affective dysfunction or to altered plasma glucocorticoid levels.
Increased 11β-HSD1 mRNA expression with age in C57BL/6J mice was confined to hippocampal CA3 and cortical layer V pyramidal cells; interestingly both regions are particularly sensitive to age-related damage (Casu et al., 2002; Mueller et al., 2008). Specifically in CA3, excess glucocorticoids and chronic restraint stress cause dendritic atrophy, a structural deterioration that makes neurons more vulnerable to excitotoxins (Conrad et al., 2007) and impairs spatial memory (Luine et al., 1994; Conrad et al., 1996; Sunanda et al., 2000). Moreover, spatial memory in the watermaze depends critically on the integrity of hippocampal CA3 subfield (Steffenach et al., 2002) and on plasticity-related mRNA transcripts within the hippocampal CA3 subregion (Haberman et al., 2008). Parietal cortex is also involved in spatial learning (Save and Poucet, 2009).
Whilst the selective increase in hippocampal 11β-HSD1 mRNA expression were modest (with no changes to MR or GR expression), two lines of evidence suggest this may be sufficient to influence hippocampal functional deficits with ageing. First, 11β-HSD1 mRNA in CA3 cells correlated with age-associated cognitive function. Neither plasma corticosterone levels, which increased with age, nor the hippocampal MR and GR expression correlated with spatial learning and memory. Although correlation does not prove causation, modest transgenic overexpression of 11β-HSD1 in forebrain, including hippocampus, produced hippocampus-associated cognitive deficits in the watermaze and inhibitory avoidance tasks. Indeed, in this model 11β-HSD1 expression is modified solely in the forebrain to generate an accelerated cognitive decline observed at a time when the wildtype controls have unimpaired cognition; this excludes the possibility that other factors modified in ageing may be causal in generating this phenotype. Since we found no evidence for either elevated plasma corticosterone levels (diurnal or with stress) in the transgenic mice we infer that the likely cause of the cognitive deficits is intraneuronal glucocorticoid excess, the reverse of the situation in 11β-HSD1−/− mice, but in this case clearly reflecting direct effects upon the CNS rather than secondary to the metabolic and neuroendocrine changes seen in generalised 11β-HSD1 knockout mice (Kotelevtsev et al., 1997; Harris et al., 2001; Morton et al., 2001).
The mechanisms leading to elevated 11β-HSD1 in specific neurons are uncertain. However, the 11β-HSD1 gene promoter is predominantly directly regulated by C/EBP transcription factors, with induction by C/EBPα and constraint by C/EBPβ, at least in liver (Chapman and Seckl, 2008). C/EBPβ is also important for memory consolidation (Taubenfeld et al., 2001) and intriguingly shows deficient induction in the ageing hippocampus (Monti et al., 2005).
The cognitive deficits were only manifest with ageing in Tg mice thus excluding developmental effects of the transgenic manipulation, also unlikely given that expression from the CAMIIK promoter occurs postnatally when much brain development is complete. There may be several reasons why we did not observe cognitive deficits in young Tg mice: intraneuronal corticosterone concentrations (via 11β-HSD1 action and free corticosterone from blood) may not be high enough in young Tg mice, with lower plasma corticosterone levels than aged mice, to impact on cognitive function; small cognitive impairments may indeed exist in young Tg mice but the tasks used were not sufficiently demanding for the detection of such changes; aging per se may be necessary for the emergence of deficits. Indeed, aged mice are more sensitive to stress-induced spatial memory impairments than young mice (Buchanan et al., 2008); the time-related accumulation of myriad deficits that underpin ‘allostatic load’ through life may also be pertinent (McEwen, 2007) interacting with increased intracellular GC exposure to initiate premature cognitive deficits. Taken alongside the rise of 11β-HSD1 in the hippocampus and layer V of the cortex of aged cognitively-impaired mice, these data suggests that cognitive decline with ageing may reflect intraneuronal “Cushing's disease of the brain”, tipping the balance from the beneficial adaptive effects of acute rises in glucocorticoids to the deleterious consequences of chronically elevated levels. This forms intriguing parallels with obesity and metabolic syndrome which associate with excess 11β-HSD1 selectively in adipose tissue (Chapman and Seckl, 2008). The data also afford a rationale for developing 11β-HSD1 inhibitors for therapy of age-related cognitive disorders (Webster et al., 2007). Indeed early proof-of-concept studies suggest efficacy in humans (Sandeep et al., 2004).
This work was supported by a Wellcome Trust project grant (WT070207; MCH, JLP, JJM, JRS), a MRC project grant (G0501596; JLWY, JRS) and a Wellcome Trust programme grant (WT083184; JRS, JJM). JLWY holds a Research Councils UK Academic Fellowship. MCH, JLWY and JRS are members of The University of Edinburgh Centre for Cognitive Aging and Cognitive Epidemiology, part of the cross council Lifelong Health and Wellbeing Initiative (G0700704/84698). Funding from the BBSRC, EPSRC, ESRC and MRC is gratefully acknowledged.