AR, apoE, and cognitive function

Abstract

Reduced androgen levels in aged men and women might be risk factors for age-related cognitive decline and Alzheimer’s disease (AD). Ongoing clinical trials are designed to evaluate the potential benefit of estrogen in women and of testosterone in men. In this review, we discuss the potential beneficial effects of androgens and androgen receptors (ARs) in males and females. In addition, we discuss the hypothesis that AR interacts with apolipoprotein (apoE)4, encoded by ε4 and a risk factor for age-related cognitive decline and AD, and the potential consequences of this interaction.

Introduction

Increasing age and female sex are risk factors for developing mild cognitive impairments (MCI) and Alzheimer’s disease (AD) (Farrer, Cupples, Haines, Hyman, Kukull, Mayeux, Myers, Pericak-Vance, Risch, and van Duijn, 1997; Fratiglioni, Viitanen, von Strauss, Tontodonati, Herlitz, and Winblad, 1997; Gao, Hendrie, Hall, and Hui, 1998; Katzman, Aronson, Fuld, Kawas, Brown, Morgenstern, Frishman, Gidez, Eder, and Ooi, 1989; for review (Raber, Huang, and Ashford, 2004)). Therefore, evaluating risk factors and therapeutic strategies for these age-related impairments become increasingly important. We will discuss the potential roles of androgens and androgen receptors (ARs) in cognitive function in humans and mice and potential interactions of androgens and ARs with apolipoprotein E.

Sex differences in spatial learning and memory

There are sex differences in spatial learning and memory in humans (Berger-Sweeney, Arnold, Gabeau, and Mills, 1995; Reinisch, Ziemba-Davis, and Sanders, 1991; Roof and Havens, 1992) and rodents (Beatty, 1979; van Haaren, van Hest, and Heinsbroek, 1990). Spatial learning and memory is relevant, as it is impaired in Alzheimer’s disease (AD). Some, but not all (Bucci, Chiba, and Gallagher, 1995), studies of spatial learning and memory in rodents have shown that males learn more quickly than females and exhibit superior performance in a variety of mazes (Bucci et al., 1995; Einon, 1980; Frye, 1995; Joseph and Gallagher, 1980; Luine, Renner, and McEwen, 1986; McEwen, 1988; McNemar and Stone, 1932; Means and Dent, 1991; Roof, 1993). Spatial learning and memory have been attributed to the hippocampus (Kesner, Bolland, and Dakis, 1993; Olton, Branch, and Best, 1978) and sex differences in hippocampal structure may contribute to sex differences in spatial learning and memory (McEwen, Alves, Bulloch, and Weiland, 1997). Different strategies used by males and females to solve spatial tasks may contribute to these sex differences. Females tend to rely more on local cues and landmarks, while males rely more on the spatial relationship between fixed points. Rodent lesion and pharmacological studies support that females may be more susceptible to spatial memory impairments than males (Hörtnagl, Hansen, Kindel, Schneider, Tamer, and Hanin, 1993; Kolb and Cioe, 1996).

Androgens, the hippocampus, and sex differences in learning and memory

Sex steroids, which cause sex differences in brain organization (organizational effect) (Beatty, 1979) and in behaviors in adulthood (activational effect) (Joseph, Hess, and Birecree, 1978), might contribute to sex differences in spatial learning and memory (Joseph et al., 1978).

Androgens might contribute to the sex differences in spatial learning and memory. In adulthood, testosterone enhances spatial memory (Alexander, Packard, and Hines, 1994; Flood and Roberts, 1988; Ishunina, Fisser, and Swaab, 2002; Janowsky, Oviatt, and Orwoll, 1994; Raber, Bongers, LeFevour, Buttini, and Mucke, 2002; Vazquez-Pereyra, Rivas-Arancibia, and Castillo, 1995) and androgens also enhance both short-term and long-term emotional learning and memory (Vazquez-Pereyra et al., 1995). In addition, post-training administration of androgens to ovariectomized rats enhanced spatial and emotional learning and memory (Frye and Lacey, 2001). However, in some studies androgens impaired spatial learning and memory (Galea, Kavaliers, Ossenkopp, and Hampson, 1995; Gouchie and Kimura, 1991; Goudsmit, Van De Poll, and Swaab, 1990; Hampson, 1995; Naghdi, Nafisy, and Majlesi, 2001).

AR signaling

Androgens such as dihydrotestosterone (DHT) initiate many of their effects by binding to AR (Barley, Ginsberg, Greenstein, MacLusky, and Thomas, 1975). The human AR gene contains 8 exons encoding a 110 kDa member of the nuclear receptor superfamily (Simental, Sar, and Wilson, 1992). The first exon encodes the N-terminal domain containing the major transactivation function (AF-1), which interacts with the glutamine-rich region of steroid receptor coactivator-1 (SRC-1) (Robyr, Wolffe, and Wahli, 2000). SRC-1 in turn can interact with the global activator CBP/p300, which can directly interact with ARs and together with SRC-1 synergistically induce AR transactivation (McKenna, Lanz, and O'Malley, 1999). CBP/p300. The second and third exons encode the DNA binding domain (DBD), which enable ARs to bind the regulatory part of target genes (Umesono and Evans, 1989). Exons 4–8 encode the ligand binding domain (LBD) and a minor transactivation function (AF-2). AF-2 also interacts with SRC-1 (Umesono and Evans, 1989). The activity of SRC-1 is regulated by phosphorylation by mitogen-activated protein kinase (MAPK), which is stimulated by androgens. In the absence of hormone, the LBD of AR prevents the transactivation of AF-1 and the required three-dimensional structures for SRC-1 interaction at AF-2. Hormone binding enables AF-1 and AF-2 to associate with a multi-subunit complex of coactivator proteins. This complex is proposed to associate additional proteins required for interaction of AR to RNA polymerase II. AR can also be activated in a ligand-independent fashion. For example, growth factors and Interleukin-6 (IL-6) activate ARs in the absence of androgen. Point mutations at different sites in exons 2–8 have been reported for partial and complete forms of androgen insensitivity. In human prostate cancer cells, the ligand-independent activation of ARs by IL-6 involves phosphorylation of SRC-1 by MAPK (Ueda, Mawki, Bruchovsky, and Sadar, 2002). However, it should be emphasized that SRC-1 phosphorylation requires the presence of IL-6 and by itself is not sufficient for ligand-independent transactivation of ARs (Ueda et al., 2002).

AR not bound to hormone is localized in the cytoplasm as a complex with heat-shock proteins and immunophilins. When androgens bind, AR changes its confirmation, dissociates from the complex, forms homodimers, and unmasks its nuclear localization signal. This signal can then bind importins, which transport the hormone-bound AR into the nucleus. ARs can modulate gene transcription by binding to specific androgen response elements in the DNA, but are also able to trans-activate and trans-repress without directly interacting with specific DNA elements. The calcium-binding protein calreticulin is able to dissociate AR from the DNA by competing for the DBD of the AR and may export AR back to the cytoplasm. Hormone-dissociated AR moves back to the cytoplasm ready for another nuclear translocation. In addition to nuclear receptor pathways (genomic effects), androgens might also mediate their effects by mechanisms not involving nuclear receptors (nongenomic effects).

ARs in cognitive areas in the brain

Increasing evidence supports an important role for androgens and ARs in hippocampal function. The levels of AR mRNA and AR binding in the hypothalamus and hippocampus of male rats are similar (Burgess and Handa, 1993; Kerr, Allore, Beck, and Handa, 1995). In addition, androgens increase spine-synapse density in the CA1 of ovariectomized female rodents (Leranth, Hajszan, and MacLusky, 2004; Woolley, Gould, Frankfurt, and McEwen, 1990). While there are no absolute sexual differences in AR mRNA (Simerly, Chang, Muramatsu, and Swanson, 1990) and AR binding (Handa, Reid, and Resko, 1986) in adult rat brain, differences in circulating androgen levels could modulate AR function in a sex-dependent fashion. Neonatal castration of males or prenatal treatment of DHT to females reverses the sex difference in CA3 pyramidal cell layer volume and neuronal cells soma sizes, which correlates with water maze performance in adulthood (Isgor and Sengelaub, 1998). Circulating androgens may also be important in sustaining neuron synapses in CA3 later in life. The decline in synaptic density in CA3 neurons in aged male rats is associated with age-related deficits in spatial ability (Geinisman, Detoledo-Morrell, Morrell, and Heller, 1995). Further, DHT selectively modulates the induction of the mRNA for the cellular immediate early gene c-fos in the CA1 region of the hippocampus following exposure to a novel open field for 20 min (Kerr, Beck, and Handa, 1996). Finally, DHT attenuates the binding of the N-methyl-D-aspartate (NMDA) receptor antagonist MK-801 in CA1 (Kus, Handa, Hautman, and Beitz, 1995) and modulates NMDA-mediated depolarization in CA1 pyramidal cells (Pouliot, Handa, and Beck, 1996).

Apolipoprotein E (apoE) and AR

ApoE plays an important role in the metabolism and redistribution of lipoproteins and cholesterol. The three major human apoE isoforms are encoded by distinct alleles (ε2, ε3, and ε4). They differ in having cysteine (Cys) or arginine (Arg) at positions 112 and 158 and differ in metabolic properties (Mahley and Huang, 1999). ApoE2 (Cys-112 and Cys-158) binds defectively to low-density lipoprotein (LDL) receptors. ApoE3 (Cys-112, and Arg-158) binds normally to LDL receptors and is associated with normal lipid metabolism. ApoE4 (Arg-112, Arg-158) binds normally to LDL receptors but is associated with elevated cholesterol levels. While apoE3 shows a lipoprotein preference for high-density lipoprotein (HDL), apoE4 show a lipoprotein preference for very low-density lipoprotein (VLDL).

In the brain, apoE has been implicated in development, regeneration, neurite outgrowth, and neuroprotection (Mahley, 1988). The effects of androgens on spine-synapse density (Leranth et al., 2004; Woolley et al., 1990) described above and the effects of apoE on neurite outgrowth might be one pathway involving both apoE and AR. Following injury, apoE in the nervous system has been implicated in efforts to restore neuronal function by remodeling (Laskowitz, Horsburgh, and Roses, 1998; Poirier, 1994; Weisgraber, Pitas, and Mahley, 1994), which may involve effects of apoE on cholesterol and phospholipid metabolism (Boyles, Zoellner, Anderson, Kosik, Pitas, Weisgraber, Hui, Mahley, Gebicke-Haerter, Ignatius, and Shooter, 1989; Mahley, 1988), on lipid efflux, on its cellular trafficking (Ji, Pitas, and Mahley, 1998; Nathan, Chang, Bellosta, Brisch, Ge, Mahley, and Pitas, 1995), on microtubule function. For example, glia, a major source of apoE (Boyles et al. 1985; Pitas et al., 1987), might recycle cholesterol from neuronal membranes by packaging it with apoE for neuronal uptake to promote neurite outgrowth (Boyles et al., 1989; Laskowitz et al., 1998; Poirier, Baccichet, Dea, and Gauthier, 1993). The ability to remodel neurons is apoE isoform-dependent, as it is impaired in AD patients with an ε4 allele who show reduced dendritic remodeling of pyramidal and subcortical neurons, while AD patients with two ε4 alleles show a shift towards proximal branching and lack a relationship between dendritic growth in response to neuronal loss (Arendt, Bruckner, Gertz, and Marcova, 1998; Arendt, Schindler, Brückner, Eschrich, Bigl, Zedlick, and Marcova, 1997). The differential ability of apoE isoforms to remodel neurons by supporting neurite outgrowth (either modulating the length of the longest neurite or the total length of all neurites of a neuron) has been reported in Neuro2A cells (Bellosta, Mahley, Sanan, Murata, Newland, Taylor, and Pitas, 1995), dorsal root ganglion cells (Nathan, Bellosta, Sanan, Weisgraber, Mahley, and Pitas, 1994), primary cortical neurons (Nathan et al., 2002), and in mouse organotypic hippocampal slice cultures (Teter et al., 1999, 2002) using recombinant apoE in the presence or absence of lipoproteins, apoE derived from transfected or apoE derived from human apoE transgenic mice. These effects of apoE on neurite outgrowth are mediated by binding of apoE to the low-density lipoprotein receptor-related protein (Holtzman et al., 1995; Nathan et al, 2002). While apoE3 increased neurite outgrowth, in most, but not all, studies apoE4 either had no effect or decreased neurite outgrowth, as compared to no apoE, and its effect dominated over that of apoE3 when both were present (Holtzman, Pitas, Kilbridge, Nathan, Mahley, Bu, and Schwartz, 1995). The lack of an effect of apoE4, as compared to no apoE, and a dominant effect of apoE4 over apoE3 was also seen in vivo when age-dependent reductions in synaptophysin-immunoreactive presynaptic terminals were quantified in apoE transgenic mice (Buttini et al., 2000) and when neuritic sprouting was quantified in organotypic hippocampal slice cultures of apoE transgenic mice (Teter, Xu, Gilbert, Roses, Galasko, and Cole, 1999). The differential effects of apoE isoforms on neurite outgrowth might be due to a stronger interaction of apoE4 than apoE3 with AR, resulting in reduced AR-mediated neuritic sprouting in the presence of apoE4. However, it should be pointed out that at this moment it is not clear whether interaction of apoE4 with AR would be harmful because as a result of this interaction less AR is available for AR-mediated signaling (loss of AR function) and therefore this interaction interferes with the regular AR function or whether this interaction would be harmful as the apoE4-AR complex causes detrimental effects by itself (gain of AR misfunction).

Compared with ε2 and ε3, ε4 increases the risk of developing Alzheimer’s disease (AD) (Corder, Robertson, Lannfelt, Bogdanovic, Eggertsen, Wilkins, and Hall, 1998; Corder, Saunders, Pericak-Vance, and Roses, 1995; Strittmatter, Saunders, Schmechel, Pericak-Vance, Engchild, Salvesen, and Roses, 1993). ApoE is associated with the pathological hallmarks of AD. The severity of AD pathology is influenced by ApoE genotype. ApoE genotype also has a detrimental impact on AD pathology when it occurs in patients with other conditions, including progressive supranuclear palsy (PSP) (Tsuboi, Joseph, Cookson, and Dickson, 2003) and Down Syndrome (DS) (Del Bo, Comi, Bresolin, Castelli, Conti, Degiuli, Ausenda, and Scarlato, 1997). While the ε4 allele frequency is similar in control subjects with no neurodegenerative condition and in PSP patients with minimal or no AD pathology, it is significantly higher in PSP patients who have concomitant AD pathology or the pathological signs of normal aging. The detrimental effects of ε4 are not limited to AD. Compared with ε2 and ε3, ε4 also increases the risk of developing cognitive impairments following neurotrauma (Nathoo, Chetty, van Dellen, and Barnett, 2003), ischemia (Guangda, Bangshun, Xiujian, and Yangzhong, 1999; Roses and Saunders, 1997), cardio-pulmonary bypass surgery (Tardiff, Newman, Saunders, Strittmatter, Blumenthal, White, Croughwell, Davis, Roses, and Reves, 1997), and human immunodeficiency virus (HIV) infection (Corder et al., 1998), as well as cognitive impairments that occur with normal aging (Howieson, Camicioli, Quinn, Silbert, Care, Moore, Dame, Sexton, and Kaye, 2003) and in the context of Parkinson Disease (Tang, Xie, Xu, Lin, and Ren, 2002; Zareparsi, Camicioli, Sexton, Bird, Swanson, Kaye, Nutt, and Payami, 2002). In multiple sclerosis patients, ε4 is associated with a worsened disease progression (Enzinger, Ropele, Strasser-Fuchs, Kapeller, Schmidt, Poltrum, Schmidt, Hartung, and Fazekas, 2003; Fazekas, Strasser-Fuchs, Kolleger, Berger, Kristoferitsch, Schmidt, Enzinger, Schiefermeier, Schwarz, Kornek, Reindl, Huber, Grass, Wimmer, Vass, Pfeiffer, Hartung, and Schmidt, 2001; Masterman, Zhang, Hellgren, Salter, Anvret, Lilius, Lannfelt, and Hillert, 2002; Schmidt, Barcellos, DeSombre, Rimmler, Lincoln, Bucher, Saunders, Lai, Martin, Vance, Oksenberg, Hauser, Pericak-Vance, Haines, and Group, 2002). In contrast, apoE4 might be protective against liver damage caused by the hepatitis C virus (Wozniak, Itzhaki, and Faragher, 2002). These results indicate that apoE affects fundamental biological processes not unique to AD.

Various mechanisms have been proposed to mediate the differential effects of human apoE isoforms on brain function. We hypothesize that apoE4 has detrimental effects on androgen- and AR-mediated pathways.

Mice deficient in mouse apoE (Apoe^−/−) with or without transgenically expressing human apoE isoforms in brain under the control of neuron-specific or astrocyte-specific promoters and mice expressing human apoE isoforms under the control of the endogenous mouse apoE promoter have been used to define the physiological and pathological role of apoE. In brain, Apoe^−/− mice show no obvious alterations in brain development but show age-dependent structural and functional alterations. Expression of apoE4 in Apoe^−/− mice leads to age- and sex-dependent impairments in spatial learning and memory (Raber, Wong, Buttini, Orth, Bellosta, Pitas, Mahley, and Mucke, 1998; Raber, Wong, Yu, Buttini, Mahley, Pitas, and Mucke, 2000). At 6 months of age, apoE4 female showed impairments in hippocampus-dependent spatial learning and memory in the water maze. These impairments were observed in mice that express apoE4 in neurons (Raber et al., 1998; Raber et al., 2000) or astrocytes (Van Meer, Acevedo, and Raber, 2007) and are therefore independent of the cellular source of apoE. They require apoE4 and are not seen in Apoe^−/− female mice, consistent with a pathogenic gain of function of apoE4. Effects of apoE4 on AR function might contribute to these cognitive impairments. Treatment of 6-month apoE4 female mice with testosterone or dihydrotestosterone antagonized these impairments (Raber et al., 2002). Further support for a role of AR in these effects comes from recent studies showing protective effects of two tissue selective androgen receptor modulators (SARMs, AC-262536 or SARM AC-264184) against detrimental effects of 6-month-old apoE4 on water maze performance (Acevedo, Gardell, Bradley, Piu, and Raber, 2008). The treatments were administered subcutaneously daily, starting one week prior to cognitive testing. These data demonstrate that, like testosterone and DHT, SARMs, antagonize cognitive deficits in 6-month-old apoE4 female mice. SARMs are tissue specific and do not have side effects associated with the use of conventional androgens. Effects of SARMs outside the brains have been reported. For example, in rats SARMs restored levator ani muscle mass to levels expressed in intact controls (Sun, Robl, Wang, Huang, Kuhns, Lupisella, Beehler, Golla, Sleph, Seethala, Fura, Krystek, An, Malley, Sack, Salvati, Grover, Ostrowski, and Hamann, 2006), muscle strength and body composition and prevented bone loss in orchidectomized animals (Gao, Reiser, Coss, Phelps, Kearbey, Miller, and Dalton, 2005), prostate tumor growth, and orchidectomy-induced bone loss (Allan, Lai, Sbriscia, Linton, Haynes-Johnson, Bhattacharjee, Dodds, Fiordeliso, Lanter, Sui, and Lundeen, 2007). Our data indicate that SARMs are also promising for developing treatments in the brain.

Compared to age-matched wild-type and Apoe^−/− mice, expression of apoE4, but not apoE3, reduces cytosolic AR binding to androgens in the neocortex of female and male mice (Raber et al., 2002). As there was no sex differences in cytosolic AR binding in any mouse genotype and female mice have lower circulating androgen levels than males, apoE4 female mice might be more susceptible to the effects of reduced AR binding on cognitive function. ApoE4 males were only relatively protected. When apoE3 and apoE4 male mice were treated with the AR antagonist hydroxyflutamide (Malgor, Valsecia, Verges, and De Markowsky, 1998), spatial learning and memory impairments were seen only in apoE4 male mice (Raber et al., 2002). Male apoE3 and apoE4 mice have similar plasma testosterone levels, suggesting that differences in plasma testosterone are not required for the differential effects of hydroxyflutamide on cognitive performance in male apoE3 and apoE4 mice. In contrast to the situation with mice, in humans, low serum testosterone concentrations and the interaction of testosterone with apoE4, might each increase one's risk of developing AD (Hogervorst, Williams, Budge, Barnetson, Combrinck, and Smith, 2001).

Some of the biological effects of testosterone require aromatization of testosterone to estrogen. Such effects might be partially or completely mediated by ERs and, in the presence of apoE4, increased ER function compensate for reduced AR function. ApoE gene expression is upregulated by 17β-estradiol (Srivastava, Srivastava, and Averna, 1997). Thus in the context of apoE4, stimulating AR-related pathway might be a valuable therapeutic target, as ARs might be more responsive to androgens (Raber et al., 2002) than ERs are to estrogens (Yaffe, Haan, Byers, Tangen, and Kuller, 2000).

The lower circulating concentrations of endogenous androgens in females than males might contribute to their increased susceptibility to the detrimental effects of apoE4 on cognitive function. To test this hypothesis, the influence of chronic reduced circulating androgens by castration on the cognitive performance of 3-4-month-old Apoe^−/−, apoE3, and apoE4 male mice was assessed (Pfankuch, Rizk, Olsen, Poage, and Raber, 2005).

When exploratory activity in the open field was assessed, there were no effects of genotype, treatment, or genotype x treatment interactions that could influence performance in cognitive tests.

The mice were also tested for object recognition. After habituation to an open field, the mice were trained in three consecutive trials and than tested in two consecutive trials with a 5–min inter-trial interval. For both the training and testing sessions, three objects were placed in the open field, and the animal was allowed to explore for 10 min. All objects were only used once and replicas were used in subsequent trials. Five min after the training trials, the animals were tested for recognition of the novel location of one of the familiar objects. Five min after the novel location test, the animals were tested for novel object recognition. The time spent exploring each object during the training and testing sessions was recorded by an observer. There was no genotype difference in the total time the mice spent exploring the three objects over the five trials. When a familiar object was moved to a novel location, sham Apoe^−/−, apoE3, and apoE4 mice spent significantly more time exploring it in the novel location (Table 1). However, while castrated Apoe^−/− and apoE3 male mice spent significantly more time exploring the familiar object in the novel location, castrated apoE4 male mice did not (Table 1). Five min after the novel location trial, mice were tested for novel object recognition. While there was no genotype difference in novel object recognition, there was a significant effect of castration (Table 1); castrated Apoe^−/−, but not apoE3 or apoE4, mice spent less time exploring the novel object than genotype-matched sham mice. The object recognition data support that apoE4 increases the susceptibility for hippocampal dysfunction, as following hippocampal lesions novel location recognition is impaired but novel object recognition is not.

Table 1.

Object recognition of sham and castrated Apoe^−/−, apoE3, and apoE4 male mice¹

	Old Location	Novel Location	Novel Object
Sham Apoe−/−	32 ± 5	49 ± 4*	74 ± 4
Castrated Apoe−/−	32 ± 7	58 ± 5**	45 ± 10
Sham apoE3	35 ± 2	47 ± 3*	69 ± 5
Castrated apoE3	28 ± 8	46 ± 7*	55 ± 11
Sham apoE4	28 ± 5	47 ± 4**	58 ± 7
Castrated apoE4	35 ± 3	39 ± 3	61 ± 5

^*p < 0.05 versus old location.

^**p < 0.01 versus old location.

¹Pfankuch, Rizk, Olsen, Poage, and Raber, 2005

In the visible sessions of the water maze test, all groups showed comparable swim speeds and there were no effects of genotype, treatment, or genotype x treatment interactions. All groups learned to locate the visible and hidden platform locations and there were no effects of genotype, treatment, or genotype x treatment interactions on time to locate the visible or hidden platform or on cumulative distance to the visible or hidden platform. However, genotype and treatment effects were identified in the probe trials (platform removed) following the second (Probe 1) and third day (Probe 2) of hidden platform training. While already in Probe 1 sham-treated Apoe^−/− and apoE3 male mice spent more time in the target quadrant than in any other quadrant, sham-treated apoE4 mice did not (Table 2). Compared to sham genotype-matched controls, castration impaired spatial memory retention in Probe 1 in Apoe^−/−, but not apoE3 or apoE4, mice. In Probe 2, all groups showed spatial memory retention.

Table 2.

Spatial memory retention of sham and castrated Apoe^−/−, apoE3, and apoE4 male mice in the water maze probe trial¹

	Left	Target	Right	Opposite
Sham Apoe−/−	19 ± 5	49 ± 6**	16 ± 5	17 ± 5
Castrated Apoe−/−	20 ± 6	28 ± 5	34 ± 5	18 ± 4
Sham apoE3	16 ± 3	48 ± 8*	22 ± 5	14 ± 4
Castrated apoE3	16 ± 4	48 ± 9*	20 ± 8	16 ± 6
Sham apoE4	30 ± 7	42 ± 7	17 ± 5	12 ± 3
Castrated apoE4	21 ± 6	49 ± 7**	16 ± 6	14 ± 3

^*p < 0.05 versus any other quadrant.

^**p < 0.01 versus nay other quadrant.

¹Pfankuch, Rizk, Olsen, Poage, and Raber, 2005

Next, we tested emotional learning and memory using the passive avoidance test. On day 1, apoE4 mice required significantly more training than Apoe^−/− and apoE3 mice (Table 3), indicating that the apoE4-induced cognitive impairments are not limited to spatial learning and memory. Although the mice were all trained to criterion, this might not be required to see comparable memory retention 24 later. Twenty-four hours later, all groups showed passive avoidance memory retention and there were no effects of genotype or treatment or genotype x treatment interactions (Table 3).

Table 3.

Passive avoidance learning and memory of sham and castrated Apoe^−/−, apoE3, and apoE4 male mice¹

	Trials to Criterion	Memory Retention Day 2
Sham Apoe−/−	3.0 ± 0.5	261 ± 20
Castrated Apoe−/−	2.6 ± 0.4	247 ± 28
Sham apoE3	2.8 ± 0.2	288 ± 8
Castrated apoE3	2.8 ± 0.3	244 ± 30
Sham apoE4	3.8 ± 0.4	289 ± 8
Castrated apoE4	3.9 ± 0.4	300 ± 0

¹Pfankuch, Rizk, Olsen, Poage, and Raber, 2005

These data show that castration impaired novel location recognition in apoE4, but not apoE3 or Apoe^−/−, mice. In contrast, castration impaired novel object recognition and spatial memory retention in the water maze in Apoe^−/−, but not apoE3 or apoE4, mice. While apoE4 mice required more trials than apoE3 or Apoe^−/− mice to reach the criterion during passive avoidance training, castration did not modulate passive avoidance learning or memory retention in any genotype. These data support that while circulating androgen levels might have differential roles in specific cognitive tests, they protect Apoe^−/− male mice against the detrimental effects of apoE4 on novel location recognition.

Castration, which lowers androgen levels and androgen receptor levels, had no detrimental effect on the hidden water maze learning curve or spatial memory retention of apoE4 mice. In contrast, acute androgen receptor blockade with hydroxyflutamide caused striking impairments in the ability of apoE4 male mice to locate the hidden platform. Chronically reduced circulating androgen levels might be more appropriate than acutely reduce circulating androgen levels to test the hypothesis that lower circulating concentrations of endogenous androgens in females than males might contribute to their increased susceptibility to the detrimental effects of apoE4. However, it is possible that in apoE4 mice the time period of 3 months between the castration and behavioral testing allowed for compensatory changes not present following acute androgen receptor blockade and that these changes protected against behavioral deficits in the water maze. Alternatively, in the presence of apoE4, lowering the androgen receptor levels might be beneficial for spatial learning and memory in the water maze. In support of this possibility, castrated, but not sham-castrated, apoE4 mice showed spatial memory retention in the first probe trial.

To further explore the role of AR in effects of apoE4 on brain function we crossed apoE4 mice with mice lacking functional ARs. Mutant mice with a naturally occurring defect in the AR gene (testicular feminization mutant or tfm) are available on the C57/BL6J background(Couse and Korach, 1998; Tabibnia, Cooke, and Breedlove, 1999). A single point mutation in the N-terminal region of the AR gene results in a premature stop codon and the expression of nonfunctional truncated androgen receptors (tfm-AR). Because the trait is X-linked, only males are androgen insensitive. The lack of functional androgen receptors results in complete infertility of the tfm male mice. Female tfm mice are heterozygous, carrying one wild-type and one tfm copy of the androgen receptor. Therefore, tfm male, but not female, mice are suitable for conclusively determining the role of AR in the effects of apoE4 on cognition. Tfm male mice have lower testosterone levels than controls (Jones, Pugh, Hall, Channer, and Jones, 2003), so there is no concern that due to the lack of functional AR testosterone and estradiol levels are elevated. To determine the role of AR in the cognitive effects of apoE4, we backcrossed tfm mice onto the Apoe^−/− background to eliminate mouse apoE and crossed the tfm/Apoe^−/− female mice with apoE4 transgenic male mice and the sex (Mroz, Carrel, and Hunt, 1999), presence of apoE4 (Raber et al., 2002), and tfm mutation (Rizk-Jackson, Robertson, and Raber, 2007) of the offspring was determined. The presence of the tfm mutation was determined by DNA sequencing and by confirming the tfm male phenotype (feminine exterior). Only male mice were behaviorally tested. We behaviorally compared Apoe^−/−, apoE4, tfm, and tfm/apoE4 male mice. As no effort was made to equalize circulating androgen levels across the groups, tfm and tfm/apoE4 male mice would be expected to have lower circulating androgen levels than the Apoe^−/− and apoE4 male mice (Jones, Pugh, Hall, Channer, and Jones, 2003).

In the object recognition test, Apoe^−/−, apoE4, and tfm mice spent significantly more time exploring the familiar object in the novel location. In contrast, tfm/apoE4 mice did not. Five min after the novel location trial, mice were tested for novel object recognition. All groups of mice spent more time exploring the novel object than the familiar objects, but there were no group differences. In the water maze test, all groups learned to locate the visible and hidden platform and there were no group differences in ability to locate the visible or hidden platform. However, group differences were identified in the probe trial following the second day of hidden platform training. While Apoe^−/− and apoE4 mice spent more time in the target quadrant than any other quadrant, tfm and tfm/apoE4 mice did not. Following one additional day of hidden platform training, tfm mice still did not show spatial memory retention. In contrast, all other groups showed spatial memory retention in the probe trial following the third day of hidden platform training.

ApoE4 lowers androgen binding to cytosolic AR (Raber et al., 2002). Rather than decreasing cytosolic AR levels, apoE4 might reduce androgen binding to cytosolic AR by interacting with AR. Our preliminary data indicate that mouse apoE colocalizes with AR in astrocytes. In addition, co-immunoprecipitation on extracts of cortex and hippocampus with an anti-apoE antibody and probing the resultant pull-downs with an anti-AR antibody reveals AR in pull downs of cortical and hippocampal extracts from NSE-apoE4 and NSE-apoE3, but not Apoe^−/−, mice. In cortical extracts, more AR was pulled down from extracts of NSE-apoE4 than NSE-apoE3 mice. Finally, our preliminary data indicate that GST-apoE proteins pull down ARs from cortical extracts of Apoe^−/− mice and that GST-apoE4 might pull down more ARs than GST-apoE3 or GST-apoE2. Castration lowers androgen and AR levels (Apostolinas, Rajendren, Dobrjansky, and Gibson, 1999; Beatty, 1979; Kerr et al., 1995; Lu, McKenna, Cologer-Clifford, Nau, and Simon, 1998) and this interaction might also contribute to the lower sensitivity of AR-mediated signaling in testosterone-treated stimulated peritoneal macrophages of castrated apoE4 than apoE3 transgenic male mice (Brown, Colton, and Vitek, 2003). The impairments of tfm/apoE4, but not apoE4, mice in novel location recognition and spatial memory retention in the first water maze probe trial indicate that AR might protect against the detrimental effects of apE4 by interacting with it. Alternatively, AR might antagonize the effects of apoE4 on cognitive function through AR-mediated signaling.

Potential effects of apoE-AR interactions on anti-inflammatory effects of androgens

AR expression in microglia increases after injury, suggesting a role for AR in the immune response (Garcia-Ovejero, Veiga, Garcia-Segura, and Doncarlos, 2002). Consistent with such a role, DHT dose-dependently decreases production of nitric oxide and TNF-α (Brown, Xu, Okhubo, Vitek, and Colton, 2007). Strikingly, this protective effect of DHT was significantly lower in primary cultures of microglia from apoE4 than apoE3 mice. This was not due to potential differences in AR mRNA or protein expression. Potential apoE4-AR interactions could account for this result. Consistent with this hypothesis, DHT-mediated AR signal transduction was less robust in microglia from apoE4 than apoE3 mice. The DHT inhibition of LPS-induced activation of the MAPK pathway (including phosphorylation of p38 MAPK and p54/p56 Janus kinase) was much less profound in microglia from apoE4 than apoE3 mice (Brown et al., 2007). MAPK might be important for the AR-apoE interactions. As described above in the section on AR signaling, the activity of SRC-1 is regulated by phosphorylation of MAPK by androgens. In contrast to androgens, apoE reduced phosphorylation of MAPK in human embryonic kidney cells (McGarvey, Nguyen, and Malkowicz, 2005) and in response to serum growth factors in rat embryonic fibroblasts(Ho, Deckelbaum, Chen, Vogel, and Talmage, 2001) and.

ApoE4 might also have detrimental effects on anti-inflammatory effects of androgen and AR function outside the brain. Peritoneal macrophages, isolated from transgenic mice that expressed either the human apoE3 or apoE4 under the control of the mouse ApoE promoter and stimulated with interferon-γ either alone or in combination with synthetic double-stranded RNA (Poly I:C) or lipolysaccharide, from male apoE4 homozygous mice showed a greater release of the inflammatory mediators nitric oxide and tumor necrosis factor-α (TNF-α) than those from male apoE3 homozygous mice (Brown et al., 2003; Brown, Wright, Colton, Sullivan, Laskowitz, and Vitek, 2002). This effect was sex-dependent, as it was not seen in peritoneal macrophages from female apoE3 and apoE4 homozygous mice, and involved androgens, consistent with the increase in nitric oxide synthase (NOS) activity in the brains of rats following castration and decrease of NOS activity following administration of the dihydrotestosterone (Sing, Pervin, Shryne, Gorski, and Chaudhuri, 2000). After IFN-γ + LPS stimulation, TNF-α release was higher in peritoneal macrophages from castrated apoE3 homozygous mice than in macrophages from sham-treated or castrated + testosterone-treated apoE3 homozygous mice; these effects were not seen in stimulated peritoneal macrophages from castrated and castrated + testosterone-treated apoE4 homozygous mice (Brown et al., 2003). A potential stronger interaction between apoE4 and AR than apoE3 and AR might require a higher androgen dose to see a protective effect of AR in the context of apoE4 than apoE3. In that case, a lower dose on androgen might be able to displace apoE3, but not apoE4, from the AR complex. Such effects of apoE on peripheral AR functions might have consequences for the brain. Increased levels of inflammatory proteins have been found in the plasma and brains of patients with dementia. Plasma levels of inflammatory proteins are increased before clinical onset of dementia and high levels of a1-antichymotrypsin and IL-6 were associated with increased risk to develop dementia (Engelhart, Geerlings, Meijer, Kiliaan, Ruitenberg, van Swieten, Stijnen, Hofman, Witteman, and Breteler, 2004). Importantly, peripheral inflammation was shown to activate brain microglia and produce chronically elevated pro-inflammatory cytokines (Qin, Wu, Block, Liu, Breese, Hong, Knapp, and Crews, 2007).

ApoE and ARs in humans

ApoE4 also has effects on AR function in the cholinergic basal forebrain, participating in behavioral processes such as attention and memory associated with aging and AD (Ishunina et al., 2002). In the vertical limb of the diagonal band of Broca, a major cholinergic nucleus in the basal forebrain affected in AD, the presence of AD pathology or an ε4 allele negatively correlates with the percentage of AR-positive neurons in women, but not in men.

Altered AR function resulting from a polymorphism in the glutamine (CAG) repeats in exon 1 of the AR gene might increase AD risk; 20 or fewer CAG repeats are associated with increased AD risk in men, but not women, particularly in those lacking the ε4 allele (Lehmann, Butler, and Warden, 2003). As polymorphisms in the CAG repeats correlate with altered AR transcription (Tut, Ghadessi, Trefiro, Pinsky, and Yong, 1997) and translation (Choong, Kemppaines, Zhou, and Wilson, 1996)( the more CAGs the less active AR), these data indicate that enhanced AR function might be beneficial for ε4, but not for non-ε4, carriers. However, it should be emphasized that although testosterone levels are associated with cognitive function in the elderly (see below), the potential effects of polymorphisms in CAG repeats on circulating androgen levels are not clear. Therefore it is possible that 20 or fewer CAG repeats are associated with reduced circulating androgen levels and therefore reduced rather than enhanced AR function.

Sex, apoE4 and salivary testosterone levels in nondemented elderly

Sex-dependent effects of apoE4 on testosterone levels might further contribute to the increased susceptibility of women to age-related cognitive decline. Recently, we reported the effects of apoE4 on cognitive performance and salivary testosterone levels in 116 nondemented elderly women and men, ranging in age from 62 to 92 (mean age ± SEM, 81.60 ± 0.57 years) and recruited from two neighboring retirement communities in Portland, Oregon. For salivary testosterone levels, there was a sex x ε4 interaction, with higher salivary testosterone levels in ε4 carrying than non-ε4 carrying men and lower salivary testosterone levels in ε4 carrying than non-ε4 carrying women (Berteau-Pavy, Park, and Raber, 2007). These data suggest that in ε4 carrying men increased salivary testosterone levels might function as compensatory change to prevent more severe cognitive impairments, as in older men testosterone levels are positively associated with cognitive function (Barrett-Connor, Goodman-Gruen, and Patay, 1999; Cherrier, Asthana, Plymate, Baker, Matsumoto, Peskind, Raskind, Brodkin, Bremner, Petrova, LaTendresse, and Craft, 2001; Cherrier, Matsumoto, Amory, Asthana, Bremner, Peskind, Raskind, and Craft, 2005; Janowsky, Chavez, and Orwoll, 2000; Janowsky et al., 1994; Kenny, Bellantonio, Gruman, Acosta, and Prestwood, 2002; Tan and Pu, 2003; Vermeulen, 2001; Yaffe, Lui, Zmuda, Ferrell, and Cauley, 2002). In men, the difference between immediate and delayed novel image recognition correlated with salivary testosterone levels (r = 0.473, p = 0.015). When the data were grouped by performance level, ε4-carrying men had the highest salivary testosterone levels at each performance level. Similarly, during the hidden session of a virtual reality spatial navigation test (Memory Island), salivary testosterone levels in male ε4 carriers correlated with time to reach the target location (latency) (r = −0.92, p = 0.03), cumulative distance to the target location (r = 0.87, p = 0.06), % trials in which the target was located within 2 min (r = 0.85, p =0.07), and % time spent in the quadrant of the virtual island previously containing the target (r = 0.89, p =0.04).

Summary

Increasing evidence supports a role for androgens and AR in cognitive function. As apoE4 reduces androgen binding to AR, this effect might contribute to increased susceptibility to cognitive impairments in apoE4 female mice and apoE4-carrying women. As androgens can antagonize detrimental effects of apoE4 effects on cognitive function and AR binding and might also protect against other AD-related factors, increased efforts are warranted to further explore the role of AR in health and disease.