Sexual dimorphism in predisposition to Alzheimer’s disease

3.1. Sex and genetics interaction in AD: focus on ApoE and BDNF

3.2. Sex-specific brain structure changes and biomarkers in AD

One explanation for the sexual dimorphism in AD risk and severity centers on the divergent changes in brain structures that men and women have in response to disease or disease-causing insults. It is posited that women may be more sensitive to pathologic agents for AD and experience more rapid structural loss when compared to their male counterparts, who are posited to have a larger cognitive reserve based on increased brain volume (Mielke et al., 2014). Many examples of this are apparent in the literature.

As part of the Alzheimer’s Disease Neuroimaging Initiative, a study examined 1-year atrophy rates using 3D-tensorebased MRI morphometry in 1368 MRI scans (144 subjects with AD, 338 subjects with MCI, and 202 controls), it was discovered that annual atrophy rates were faster in women (Hua et al., 2010). A follow-up study further showed that women with MCI had greater atrophy in numerous brain regions and a steeper decline in certain cognitive tasks compared to men, as well as greater atrophy in numerous regions once an AD diagnosis had been made (Holland et al., 2013). In the Minimal Interval Resonance Imaging in AD study, women with probable AD demonstrated significantly faster rates of hippocampal atrophy than their male counterparts (Ardekani et al., 2016). On a molecular level, another study used discovery driven quantitative proteomics to detect modulation of several redox proteins in the temporal lobe of patients with combined AD and cerebrovascular disease (AD + CVD subjects) and observed sex-specific alterations in the white matter and mitochondrial proteomes in women. Female patients also displayed downregulation of certain ATP synthase and cytochrome oxidase subunits, suggesting increased severity of mitochondrial impairment in women with AD + CVD (Gallart-Palau et al., 2016). These studies suggest that when degeneration begins, it may affect women more rapidly than men, resulting in the observed increase in pathophysiology.

In concert, a clinicopathological study observed a large sex difference in the association of AD pathology with the odds of AD dementia in men and women. Specifically, it was reported that each additional unit of AD pathology was associated with a nearly 3-fold increase in the odds of AD dementia in men compared with a more than 22-fold increase in the odds of AD dementia in women (Barnes et al., 2005), again suggesting that women might possibly be more sensitive to the pathologic agents that cause AD. In agreement, a recent study reported that women with higher CSF Aβ−42 and total tau levels, indicative of worse pathologic changes, showed more rapid hippocampal atrophy and cognitive decline than men, furthering the notion that divergent responses in structural integrity to pathologic insults are based on sex in AD (Koran et al., 2017).

Although the current studies seem to be in agreement in regard to divergent structural responses to pro-AD agents, it is difficult to conclude whether the observed sex difference is driven by differences in the etiology and pathophysiology of the AD cascade or the late-life risk factors of poor brain aging. Furthermore, although these structural changes are readily observed, the underlying molecular mechanism leading to this divergence is still far from clear.

3.3. Impact of differing sex hormones in AD

Changes in sex steroid hormones (i.e. estrogen and progesterone in females and testosterone in males) have been associated with AD onset and progression, representing an endocrinological explanation for AD risk (Grimm et al., 2016). Among these sex hormones, estrogen has received the most attention in clinical and preclinical investigations. Although much of the body’s estrogen is synthesized in the ovaries, a significant amount is also produced in the CNS, where 1 of the 3 isoforms of estrogen, E2 and its receptor (ER), has been observed in high levels in limbic system regions, such as the hypothalamus and hippocampus (Cui et al., 2013). E2 has been shown to mediate sex-specific behaviors, regulate synaptic plasticity, and promote neural survival in a number of studies (Green and Simpkins, 2000). Because of its neuroprotective role, the rapid decrease in endogenous estrogen levels during menopause has been proposed as a trigger for the development of AD in females (Paganini-Hill and Henderson, 1994).

However, only a few studies have shown a direct link between estrogen and AD. One clinical study from the Mayo Clinic, the Cohort Study of Oophorectomy and Aging, showed an almost 2-fold increase in dementia risk in women who underwent bilateral oophorectomy before menopause (Rocca et al., 2011). In the Cache County Memory Study, women who used hormone replacement therapy in the perimenopausal period had fewer cases of AD later on life (Zandi et al., 2002). Another study found that surgical menopause was associated with a faster rate of cognitive decline and more pHF-tau pathology (Bove et al., 2014). In support of this, 1 preclinical study demonstrated that in young APPswe mice with ovariectomies, absolute Aβ levels remain similar to controls, but the Aβ/sAPPα ratio significantly increases due to decreased APP. However, supplemental estrogen administered to these mice paradoxically reduced Aβ levels but did not alter the total levels of APP. These results suggest that estrogen might enhance α-secretase activity leading to increased formation of sAPPα, although the uncoupling between APP and Aβ metabolism with ovarectomy and subsequent estrogen replacement complicates this conclusion (Levin-Allerhand and Smith, 2002). Another study showed that ovariectomies in APP transgenic mice (Tg2576) increased the levels of Aβ levels in the brain and that estrogen replacement therapy blocked this effect (Zheng et al., 2002). However, the same experiments in Tg2576+PS1 double mutant mice demonstrated a much smaller effect, suggesting that differences in transgenes or strains might mediate the relative impact of estrogen in these models (Wang et al., 2003).

The downstream mechanisms of estrogen’s effect on AD neuropathology are still poorly understood, and hypotheses centering on transcriptional regulation and downstream kinase cascades have both been presented (Hwang et al., 2016; Sarkar et al., 2015). One probable mechanism whereby estrogen is neuroprotective in AD is through shifting APP processing toward the nonamyloidogenic pathway via α-secretase through activation of extracellular-regulated kinase 1 and 2 (ERK1 and 2) and protein kinase C (Cordey et al., 2003; Wang et al., 2016). Furthermore, estrogen may act through activation of the MARK/ERK pathway to reduce Aβ levels, potentially through BACE1 inhibition and increased Aβ clearance (Tamagno et al., 2009).

But although the preclinical evidence seems to suggest a protective role for estrogen in AD pathogenesis, clinical trials concerning estrogen replacement therapy (ERT) have resulted in inconclusive data, with some even showing decreased cognitive function with estrogen-only therapies (Asthana et al., 2001; Henderson et al., 2000; Mulnard et al., 2000; Shumaker et al., 2003; Tang et al., 1996; Wang et al., 2000). The results of these studies are surprising, although it is possible that ERT confers other risks that interact with AD pathogenesis to worsen cognition, such as those regarding cardiovascular health. Furthermore, it remains unclear if estrogen + progesterone or progesterone alone may have different neuroprotective properties in postmenopausal women compared to those who took ERT. Finally, some preclinical studies have suggested that estrogen may be an important factor in preventing AD during certain “critical windows,” which is supported by association data in the Cache County Memory Study (Zandi et al., 2002). One preclinical study suggested that ERT had a higher likelihood of being neuroprotective in the perimenopausal but not postmenopausal period (Christensen and Pike, 2017). Intriguingly, another study noticed an increase in Aβ and poorer working memory in female AD mice compared to males that could be reversed by supplementing neonatal females with testosterone propionate, resulting in brain masculinization (Carroll et al., 2011) and suggesting that early exposure to sex hormones during development may have a large impact on later life AD risk. Still, whatever the reason for the disconnect between preclinical and clinical studies, the role of estrogen in AD pathogenesis is likely to be more complex than originally perceived.

Interestingly, studies in human patients and animal models suggest that androgen deprivation may also represent a risk factor for AD pathogenesis in males (Pike et al., 2009; Rosario et al., 2010; Verdile et al., 2014). One study demonstrated that brain testosterone levels were inversely correlated to soluble Aβ concentration in men (Rosario et al., 2011). Another preclinical study indicated that increased testosterone in aged male 3xTg-AD mice was correlated with reduced Aβ plaque pathology (Overk et al., 2013), suggesting a role for testosterone in amyloid processing or clearance. These data were supported by findings where testosterone altered APP processing (Goodenough et al., 2000). Despite these interesting data points, there still remains only a few studies to date that have investigated androgens’ role in AD pathogenesis, and so closer investigations are likely warranted.

In summary, while an endocrinological basis of AD risk is likely to exist, the significance of these declining sex hormones with age is relatively unclear. For instance, although estrogen seems to be neuroprotective in rodent models, ERT in humans was unable to improve cognition or delay AD onset. Regardless of the impact of sex hormones on AD, it is very likely that subtle second messenger system alterations between sexes underlie and likely unify many of the observed increases in AD risk for women when viewed at the genetic (APOE, BDNF), structural, or hormonal levels. In agreement with this, it has been observed that sexually dimorphic second messenger signaling is evident for certain GPCRs and may increase AD risk in a sex-specific manner (Bangasser et al., 2017). One of the most salient examples of this is discussed in Section 3.5.

3.4. Sexual dimorphism in immune function and the link with AD

While Aβ and hyperphosphorylated-tau are considered to be the pathogenic intermediates underlying AD, their interaction with the immune system has garnered a lot of attention. Recently, many immune function genes that are expressed in microglia have been identified in GWAS studies investigating genetic causes of AD risk, including TREM2 (Guerreiro et al., 2013; Wang et al., 2015), CR1 (Lambert et al., 2009), CLU (Harold et al., 2009; Lambert et al., 2009), and CD33 (Bradshaw et al., 2013; Griciuc et al., 2013). In addition, a recent gene-network analysis implicated immune and microglial dysfunction as the leading perturbed pathway in AD in human patients, a result which was specific to AD and not found in other neurodegenerative disorders, such as Huntington’s Disease (Zhang et al., 2013). In line with this, it has been found that elevation of certain cytokines in plasma are correlated with dementia development or conversion from MCI to AD, and imaging studies have found that increased microglial activation is a better predictor of cognitive decline than Aβ load (Buchhave et al., 2010; Cagnin et al., 2001; Edison et al., 2008; Engelhart et al., 2004; Holmes et al., 2009; Jin et al., 2008; Laurin et al., 2009; Patel et al., 2005).

Microglia play an active role in both Aβ regulation and immunologic response (Heneka et al., 2015). As evidence accumulates showing that persistent immune dysregulation affects AD development, numerous studies have demonstrated specific mechanisms whereby microglia may mediate neural degeneration and cognitive dysfunction (Hong et al., 2016; Ji et al., 2013; Lenz and McCarthy, 2015). Of these mechanisms, microglial regulation of synapses may be the most important for AD. It has been shown that the degree of synaptic degeneration is a much better predictor of cognitive decline in AD than Aβ levels, as degeneration of these synaptic structures syncs up more closely with symptom development and precedes neuronal loss (DeKosky and Scheff, 1990; Hong et al., 2016; Lue et al., 1996; Mucke and Selkoe, 2012; Terry et al., 1991). These findings are in line with the hypothesis that chronic Aβ toxicity primes microglia, causing them to respond more readily to local environmental changes with cytokine release and enhanced pruning of synapses (Heneka et al., 2015; Holmes, 2013; Perry and Holmes, 2014).

Given the role for the immune system in AD development, it is important to note that females have a propensity toward greater immune reactions than men. First, males tend to develop infections more readily than females (Klein, 2000), and females tend to have a more robust immune response in the form of greater antibody production and a greater propensity toward TH1 responses (Amadori et al., 1995; Hanamsagar and Bilbo, 2016; Kivity and Ehrenfeld, 2010; Kovacs et al., 2002). In addition, estrogen generally enhances the immune system at low doses while androgens tend to be more immunosuppressive (Hanamsagar and Bilbo, 2016; McClelland and Smith, 2011). At the level of the brain, males tend to have more microglia early in development while females tend to have many more microglia from early adulthood onward (Lenz and McCarthy, 2015; Schwarz et al., 2012).

Perhaps most interestingly, however, is the fact that about 80% of patients with autoimmune disorders are women (Cooper and Stroehla, 2003; Gleicher and Barad, 2007; Kivity and Ehrenfeld, 2010), (Hanamsagar and Bilbo, 2016), suggesting that women may have more primed immune systems and thus more readily exhibit heightened immune responses. In support of this, one study found that while children of women with autoimmune disorders often developed autoimmune diseases, they were rarely the same disease exhibited by the mother (Gleicher and Barad, 2007). This is an important point, as it may explain why AD is not necessarily associated with other autoimmune diseases, since AD patients may manifest disease due to common immunological priming while expressing divergent pathophysiology to specific insults.

Finally, the interaction between stress and the immune system is different between men and women. Specifically, it was shown that while glucocorticoid signaling, which is a main mediator of the vertebrate stress response, induced apoptotic pathway upregulation in male hepatic cells, female hepatic cells tended to upregulate IL-6-related pathways, which are proinflammatory (Deak et al., 2015). Furthermore, glucocorticoid signaling tends to be higher in females at baseline, and estrogen has been shown to increase glucocorticoid release while androgens decrease it (Gaillard and Spinedi, 1998). Interestingly, when challenged with a lethal dose of the immunogen LPS, the corticosteroid dexamethasone was able to prevent death in male rats at low doses while much higher doses were needed for females (Deak et al., 2015), demonstrating that the female immunosuppressive response to glucocorticoids is less potent than males.

Still, while AD is likely to be affected by immune system function and while ample evidence suggests very different inflammatory responses between sexes, whether these divergent responses lead to a difference in AD susceptibility is unclear. For one, even the role of sex in autoimmune disease pathogenesis is unknown, with hypotheses describing differences in sex hormones, X-inactivation patterns, and gene dosing, and microchimerism all having evidence supporting and diminishing their likelihood as being the underlying mechanism (Gleicher and Barad, 2007; Kivity and Ehrenfeld, 2010; Selmi, 2008). Future studies demonstrating an immune system and sex interaction in AD pathogenesis would be needed, and rigorous hypothesis-based testing would be essential to confirm if any of the mechanisms potentially leading to increased autoimmunity also lead to increased AD risk in women.

3.5. Stress and sex-specific CRF signaling in AD

Clinical evidence has suggested a link between psychosocial stress and AD going back at least 2 decades. Stress-related increases in plasma cortisol levels (Swanwick et al., 1998; Rehman, 2002; Umegaki et al., 2000) and correlations between increased cortisol levels and the severity of cognitive decline (Csernansky et al., 2006; Pedersen et al., 2001) have been reported in AD. These changes in the hypothalamus-pituitary-adrenal (HPA) axis do not appear to be secondary to MDD, as AD patients with and without MDD have higher cerebrospinal fluid cortisol levels compared to controls (Hoogendijk et al., 2006). In support of this, the centrally active stress response receptor corticotropin releasing factor 1 (CRF1) is also increased in the hippocampi of AD patients (Behan et al., 1995, De Souza, 1995), and HPA axis changes have been associated with hippocampal atrophy (O’Brien et al., 1996). Furthermore, many abnormal behaviors, termed Behavioral and Psychosocial Symptoms in Dementia, occur at an alarmingly high rate in AD patients, and at least some of these abnormal behaviors can be attributed to alterations in stress hormone signaling (Reisberg et al., 1987). In line with these human studies, transgenic mouse models have recapitulated at least some of the neuropathological and behavioral changes associated with AD and thus provide an opportunity to investigate how the different biochemical markers underlying the psychosocial stress response and AD pathogenesis interact in vivo (Campbell et al., 2015; Carroll et al., 2011; Cuadrado-Tejedor et al., 2012; Dong et al., 2004, 2008; Green et al., 2006; Jeong et al., 2006; Kang et al., 2007; Rissman et al., 2012).

Women are more susceptible to stress-influenced diseases, and it is possible that sexually dimorphic stress responses could greatly influence the difference between men and women in the rate of AD conversion and progression. Although clinical evidence demonstrating an interaction between sex and stress in AD development remains scarce, an intriguing study demonstrated that women with mild-to-moderate AD have significantly increased levels of cortisol production compared to their male counterparts (Rasmuson et al., 2011), although these data could not establish causation.

Although studies in humans are limited, preclinical studies have also shown a sex-specific difference in AD pathogenesis under psychosocial stress (Sierksma et al., 2012). For instance, 5-day restraint stress (6 h/d) in 5XFAD transgenic mice leads to increased levels of neurotoxic Aβ42 peptides, β-secretase-cleaved C-terminal fragment (C99), and plaque burden in the hippocampus of female but not male mice at a time point before severe AD-like behavioral deficits are observed. Moreover, sex- and brain region-specific accelerations in β-amyloidosis are accompanied by elevations in BACE1, APP, and phosphorylated EIF2a in response to psychosocial stress (Devi et al., 2010). Also, environmental stress preferentially triggers memory impairments in female but not in male P301L-tau mice (Sotiropoulos et al., 2015). Furthermore, in this model, stress-related increases in caspase-3-truncated tau and insoluble tau have been found to aggregate exclusively in the female hippocampus (Sotiropoulos et al., 2015). Potentially driving this effect, the expression of the molecular chaperones Hsp90, Hsp70, and Hsp105 were altered to favor accumulation of tau aggregates (Sotiropoulos et al., 2015). Although some ambiguity exists between these studies regarding the interaction between stress and sex, it seems likely that both are impacting disease pathogenesis.

As mentioned, the exact stress-response molecules that mediate increased AD pathogenesis are still unknown, though data confirming an important hypothesis implicating central CRF1 signaling in this process is emerging. Increased CRF1 signaling has been associated with multiple stages of APP proteolysis, regulation of Aβ generation, and Aβ-mediated toxicity (Thathiah and De Strooper, 2011; Thathiah et al., 2013), and GPCR signaling in general has been suggested as a novel target for AD drug development (Femminella et al., 2013; Wolfe, 2013). Importantly, CRF overexpression in the forebrain can lead to accumulation of Aβ and tau phosphorylation through CRF1-Gs-PKA signaling, consistent with Gs-PKA signaling’s role in influencing the amyloid production cascade through modulation of α-, β-, and γ-secretases (Park et al., 2015; Robert et al., 2001; Thathiah and De Strooper, 2011; Thathiah et al., 2013; Xu et al., 1996). Specifically, while transient activation of these signaling cascades shifts APP metabolism toward the α-secretase-mediated pathway that results in nonpathogenic amyloids, chronic activation shifts APP metabolism to the β-, γ-, and perhaps also the η-secretase-mediated pathways that result in increased pathogenic Aβ generation (APP-CTF99, A-beta 40, and A-beta 42). (da Cruz e Silva et al., 2009, Willem et al., 2015). In addition, PKA signaling is associated with tau phosphorylation, another molecular pathway that is highly implicated in AD pathogenesis (Blanchard et al., 1994, Sanchez-Mut et al., 2014). Thus, there is much evidence implicating CRF1 signaling as the causative pathway facilitating psychosocial stress’ detrimental effects on AD pathogenesis.

When taken together with the evidence that prolonged CRF signaling through CRF1 results in sexually dimorphic second messenger activation, a potential mechanism begins to take shape: Because Gs-PKA signaling is central to the formation of AD-causative amyloids and phospho-tau, and there exists a bias toward CRF1-Gs-PKA signaling in females, this signaling pathway may provide a functional link between these 2 processes and begins to create a unified, biologically driven explanation for why women develop AD more readily than men. Still, this hypothesis is based on evidence from a combination of isolated experiments that only suggest this mechanism, but rigorous and direct testing of this hypothesis will need to be specifically undertaken to assess whether it is a suitable explanation for what is observed in AD patients. In addition, how CRF1 signaling and peripheral glucocorticoid signaling interact is still unknown in terms of increased AD risk, and experiments specifically targeting one or the other are needed to see if an interaction between the 2 is necessary for the increased risk of AD.

Table 2 summarizes the studies on sex difference in AD with animal models, collected from 2000 to date.