INTRODUCTION

Epidemiological studies highlight differential sex and sex-by-age interactional patterns in the incidence of depression and suicide. The life onset risk for major depression is 1.6-fold increased in women relative to men aged 10–50 years (Kessler et al, 1993), but equalizes between women and men after age 55 years (Bebbington et al, 1998). Suicide attempts are fourfold more lethal in men than women, with the highest suicide rate occurring in men over age 75 years (Szanto et al, 2002). To what extent these patterns relate to brain effects of age-related decline in sex hormones (Feldman et al, 2002; Soules et al, 2001) or sex differences remain unknown.

The serotonin (5HT) 1A and 2A receptors systems in prefrontal, mesiotemporal, and occipital cortices and raphe nucleus (5HT1A only) have been explored as potential mediators of sex, age, and hormone effects on mood, given abnormalities noted in affective disorders (Drevets et al, 2007; Meyer, 2008) and modulation by sex hormones. 5HT acting at 5HT1A somatodendritic autoreceptors on raphe nucleus serotonin neurons negatively regulates downstream 5HT release and 5HT activity at postsynaptic 5HT1A, and 5HT2A receptors have inhibitory and excitatory actions on cortical targets, respectively (Barnes and Sharp, 1999; Sharp et al, 2007), making these receptors critical modulators of global 5HT tone.

Age and sex relationships to 5HT1A receptor binding in healthy adults have been inconsistent, potentially because of limitations in postmortem and positron emission tomography (PET) methods (Arango et al, 1995; Cheetham et al, 1990; Matsubara et al, 1991; Palego et al, 1997; Costes et al, 2005; Jovanovic et al, 2008; Meltzer et al, 2001; Parsey et al, 2002; Rabiner et al, 2002; Stein et al, 2008; Tauscher et al, 2001). Potential sex differences in nonspecific binding (Parsey et al, 2002) and menstrual cycle-related receptor fluctuations (Jovanovic et al, 2006) in [11C]WAY100635 PET studies might have presented confounds. Furthermore, potential inverse relationships between hormones (estradiol; androgens, as mediated by aromatization of testosterone to estradiol) and 5HT1A receptor expression (Osterlund et al, 2000; Osterlund and Hurd, 1998; Pecins-Thompson and Bethea, 1999; Ricci et al, 2006; Simon et al, 1998; Zhang et al, 1999) deserve study in humans to explicate age–sex–5HT relationships.

Although the weight of evidence reveals an absence of sex differences in 5HT2A BPP (Adams et al, 2004; Biver et al, 1996; Frokjaer et al, 2009; Meyer et al, 1999; Rosier et al, 1996), 5HT2A receptor binding was positively associated with increases in estradiol in men and women (Frokjaer et al, 2010; Kugaya et al, 2003; Moses et al, 2000), similar to experimental animals (Cyr et al, 1998; Sumner and Fink, 1995, 1998). Whether the well-described inverse association between age and 5HT2A receptor binding (Meltzer et al, 1998) differs by sex, as suggested by nonlinear 5HT2A receptor decreases that were greater in midlife (ages 20–45 years) relative to older subjects (ages 46–70 years) in a predominantly female sample (Sheline et al, 2002), remains to be studied.

In the current study, we addressed prior limitations and unanswered questions through scanning of a large cohort (n=71) of healthy subjects aged 20–80 years with arterial-based quantification of PET data with assessment of sex, age, and hormone effects on nonspecific tracer factors. We ‘standardized’ women endocrinologically by scanning ovulatory, premenopausal women in their follicular menstrual cycle phase and by confirming hypogondal status in postmenopausal women. Regions of interest included prefrontal, mesiotemporal, and occipital cortices as well as raphe nucleus (5HT1A only). We hypothesized that the abrupt menopausal decline in estradiol would be associated with increases of 5HT1A receptor binding potential (BP) with aging that would be less pronounced in men, given more gradual androgen decreases with aging. We also tested the hypothesis that the age-related decline in 5HT2A was associated with decreased gonadal steroids in late life (Moses et al, 2000), and sought to replicate the curvilinear inverse relationship between age and 5HT2A receptor binding in women (Sheline et al, 2002) and the linear inverse relationship between age and 5HT2A receptor binding in men (Meltzer et al, 1998). We hypothesized that 5HT receptor binding in all brain regions examined would have consistent relationships with age, sex, and steroids as suggested by consistent 5HT receptor genomic expression across brain regions (David et al, 2005).

SUBJECTS AND METHODS

Subjects

5HT receptor data in relation to amygdala activity to negative emotional faces within a subset of this cohort have been previously published (Fisher et al, 2006, 2009). Subjects provided written informed consent as approved by the University of Pittsburgh Biomedical Institutional Review Board. Healthy subjects screened by the Structured Clinical Interview for DSM-IV (First et al, 1998) and the Hamilton Scale for Depression (Hamilton, 1960) were recruited in a sex-balanced fashion from age 18 through 80 years. To optimize scan acquisition in women during times of low circulating steroids, we excluded perimenopausal women, whose hormone concentrations are less predictable. Premenopausal women were younger than age 50 years and were scanned after confirmation of an ovulatory cycle (21-day progesterone ⩾25 nmol/l) and between days 2–9 of a subsequent menstrual cycle. Postmenopausal women were ⩾50 years of age with estradiol concentrations <40 pg/ml, FSH >30 IU/l, and >1 year of amenorrhea. Premenopausal women were free from oral contraceptives for 3 months or depo-medroxyprogesterone acetate for 1 year, were more than 6 months post-breastfeeding, and more than 2 years postpartum. Postmenopausal women were free from hormone treatment in the preceding 2 years and had <5 years of lifetime hormone therapy. Although prior hysterectomy or oopherectomy were excluded initially, we ultimately included 3 women with remote hysterectomy/oopherectomy because it was so common in women over 60 years.

Subjects had no personal history of a major Axis I disorder and no family history of a mood or psychotic disorder. Subjects were excluded if they had medical or neurological illnesses likely to affect cerebral physiology or anatomy, gross abnormalities of brain structure evident by magnetic resonance images (MRI), suicidal intent, substance abuse within 1 year, lifetime history of substance dependence (other than nicotine), eating disorders, or exposure to medications likely to alter cerebral physiology. After consent and screening of 176 subjects, 115 were eligible, of whom 79 were scanned and 70 and 62 produced usable WAY100635 and altanserin data, respectively (Supplementary Figure A).

Endogenous hormones were characterized through menstrual cycle charting and scan day measurement (0945–1215 hours) of thyroid-stimulating hormone (TSH), free thyroxine (fT4), dehydroepiandrosterone sulfate (DHEAS), and sex hormone-binding globulin (SHBG) in all subjects and estradiol and progesterone in women and total testosterone (TT) in men. Free androgen index (FAI) was calculated as (total testosterone (nmol/l)÷sex hormone-binding globulin (nmol/l)) × 100. Estradiol, progesterone, TT, fT4 (Coat-A-Count, Siemens, Los Angeles, CA), and DHEAS (Diagnostic Systems Laboratories, Webster, TX) were measured by radioimmunoassay. TSH and SHBG were measured by time-resolved fluroimmunometric assay (DELFIA, Perkin Elmer, Boston, MA). Intra- and inter-assay coefficients of variation (CVs) for each assay were <10%. All specimens were analyzed in duplicate and in the same assay run to reduce variability. Log-base 10 transformation was used to induce a normal distribution for use in models of receptor binding.

Image Acquisition

To provide an anatomical framework for PET data analysis, MRIs were obtained using a 1.5 T Signa Scanner (GE Healthcare, Milwaukee, WI) and a three-dimensional (3D) spoiled gradient recalled sequence. PET scans were acquired on an ECAT HR+ PET scanner (Siemens, Erlangen, Germany) in 3D mode (Meltzer et al, 2004). A 10-min transmission scan was obtained for attenuation correction using rotating 68Ge/68Ga rods. Subjects underwent consecutive 90-min PET acquisitions of both [11C]WAY100635 and [18F]altanserin, separated by 10 min, which allowed for 100 min (five half-lives) between radioligand injections. Radiosyntheses of [carbonyl-11C]WAY100635 (McCarron et al, 1996) and [18F]altanserin (Lemaire et al, 1991; Soloff et al, 2010) were performed as previously described. The [11C]WAY100635 dynamic emission scan (34 frames of increasing length over 90 min) occurred after IV bolus administration of 8.0 to 16.6 mCi (mean±SD=13.9±2.0) of high specific activity [11C]WAY100635 (2.1±1.1 mCi/nmol at the time of injection). The [18F]altanserin dynamic emission scan (22 frames of increasing length over 90 min) occurred after IV bolus administration of 6.3 to 7.8 mCi (mean±SD=7.2±0.3) of high specific activity [18F]altanserin (11.3±17.2 mCi/nmol at time of injection). Arterial blood sampled during scanning was corrected for radiolabeled metabolites to compute the plasma input function for [11C]WAY100635 (Bailer et al, 2005) and [18F]altanserin (Henry et al, 2004) scans. Plasma samples of [11C]WAY100635 and [18F]altanserin were obtained at 2, 10, 30, 60, and 90 min after injection to correct the input function for radiolabeled metabolites. Plasma protein binding of [11C]WAY100635 and [18F]altanserin was measured by ultracentrifugation using the Centrifree membranes to determine the free tracer fraction (fp).

Image Analysis

Brain regions of interest (ROIs) (average of right and left hemispheres) selected a priori based upon regional 5HT1A and 5HT2A receptor abnormalities in depression and suicide and associations with sex, age, and hormones were: amygdala, hippocampus, lateral orbitofrontal cortex, pregenual cortex, subgenual cortex, anterior cingulate gyrus, occipital cortex, and raphe nucleus (encompassing median and dorsal raphe—5HT1A receptor studies only). Because the raphe nuclei cannot be resolved on MR, ROI placement was guided along the z axis by PET scan and was drawn on seven MR planes superior to the interpeduncular cistern (Meltzer et al, 2001). A reference region for assessing nonspecifically bound and free radioligand, [11C]WAY100635 VND, was defined in cerebellar gray matter using guidelines that excluded the vermis (Parsey et al, 2005) and minimized the spill-in effects from neighboring cortex (Bailer et al, 2005; Drevets et al, 1999; Meltzer et al, 2001, 2004; Price et al, 2002b). The cerebellar white matter was not used because it did not exhibit in vivo kinetics consistent with nondisplaceable uptake in 5HT1A receptor-rich areas. The cerebellum was assumed to have minimal levels of specific [18F]altanserin binding and to provide a reasonable estimate of [18F]altanserin VND, despite overestimation of the nonspecific component as a result of the blood– brain barrier-permeable radiometabolites (Lammertsma and Hume, 1996).

PET images were aligned with MR images using automated image registration (Woods et al, 1993). ROIs were manually traced on the MR image using a modified version of the IDL-based (Interactive Data Language, Boulder, CO) computer program, ROITOOL, of CTI PET Systems (Knoxville, TN) according to published guidelines (Drevets et al, 1999; Meltzer et al, 2001). Regional tissue time-activity concentrations were obtained from the dynamic PET image for each ROI. Logan graphical analysis with generalized linear least squares smoothing (Bailer et al, 2005; Logan et al, 2001; Price et al, 2002a) was applied to the arterial input function and regional tissue time-activity concentrations to derive [11C]WAY100635 and [18F]altanserin distribution volume (VT). To control for the dilutional effect of expanded CSF spaces on brain radioactivity concentrations, we corrected VT for atrophy in all subjects using MR-based correction factors that varied from 0 to 1 (no dilution) (Meltzer et al, 1990, 1999). The [11C]WAY100635 and [18F]altanserin binding potential measures (BPP and BPND) were derived according to traditional relationships: BPP=VT–VND and BPND=VT/VND–1 (Innis et al, 2007), where VT is regional distribution volume of tracer and VND is cerebellar reference tissue volume of distribution. We derived both BPP and BPND as both measures have been used in prior 5HT1A and 5HT2A receptor studies. BPP relies upon fp in plasma whereas BPND relies upon free and nonspecific binding in the cerebellar reference tissue (VND). Because our study results are largely consistent for the two BP outcome measures, tabulated and graphical data are presented herein for BPP and in Supplementary Tables for BPND.

Statistical Analysis

We used multivariate general linear regression modeling in which we tested the main effects of age, age2, sex, and an age-by-sex interaction on atrophy correction factors, fp, VND, BPND, and BPP, for all regions of interest, separately for 5HT1A and 2A receptor binding. 5HT1A receptor BP was evaluated in two separate regression models given different physiology of postsynaptic from presynaptic (raphe) regions, whereas 5HT2A receptor BP was evaluated in a single regression model. In all models, age was centered to limit nonessential multicollinearity that is induced with creation of the interaction term and higher-order effects that are a function of age. We present tabulated regression coefficients (β) (Tables 2 and 3) as centered age ‘per decade’ (β divided by 10) to aid interpretation because regression coefficient values were small. We followed the same procedure in sex-stratified groups, in which log10 (estradiol) and log10 (free androgen index) were added as covariates to the models. Our threshold for significance was p<0.05 for each multivariate six-region regression as well as regressions for cerebellar VT and raphe BP, which were control and secondary ROIs, respectively. We set α=0.008 (0.05/6) to adjust for multiple post hoc exploratory univariate tests by region. Linear mixed modeling was used in the analysis of age and sex effects on the metabolism of parent radioligand over the course of scanning with denominator degrees of freedom approximated based on the Satterthwaite method.

We evaluated the effects of potential confounding variables by examining the relationship of sex, age, and BP with subject characteristics (Table 1) using linear regression and contingency table analysis with χ2 tests of independence. Variables that were at least marginally associated with the dependent measure (p<0.10; body mass index (BMI), hormone concentrations (TSH, free T4, and DHEAS), and injected radioligand mass) were then added together to the regression models. All of these covariates were ultimately dropped from the final model because of absence of significant association with receptor BP.

Table 1 Sample Characteristics (Mean±SD)
Full size table

RESULTS

Demographics

Of the 71 subjects with usable [11C]WAY100635 or [18F]altanserin scan data, there were 37 men and 34 women participants with age that ranged from 20.1 to 80.6 years (Table 1). The sample was largely Caucasian (83%) and well educated with 62% having received a college or postgraduate degree. Depressive and anxiety scale scores were not different on the basis of sex or age.

Reproductive Hormone Data

Three postmenopausal women (ages 67–80 years) had past hysterectomy and/or oopherectomy between ages 40–42 years. Estradiol concentration was higher in the postmenopausal intact women (7.40 vs 4.00 ng/ml). Mean scan day estradiol concentration among all women participants was low (20.4±29.5 pg/ml), as was progesterone in premenopausal women (0.5±0.2 ng/ml), indications of accurate scan timing during the early follicular phase or menopause (Table 1). FAI in men was 4.7±2.1. As expected, estradiol and FAI concentrations were inversely correlated with age, but were not thought to cause multicollinearity in our regressions because correlations were <0.8. Pearson bivariate correlations and traditional multicollinearity statistics were examined in univariate models without observed multicollinearity.

Potentially Confounding PET Variables: Atrophy Correction, Radioligand CER VT, Free Fraction, and Metabolism

Atrophy-correction factors across all ROIs decreased with aging as expected (increasing atrophy) (mean β <−0.05 per decade, p<0.001). Atrophy-correction factors were increased in women relative to men (mean β=0.014 per decade, p=0.004). There were no significant associations for age or sex, or age-by-sex interactions on [11C]WAY100635 and [18F]Altanserin injected mass or on Altanserin metabolism over the scan duration. There was a significant sex-by-age interaction (p=0.013) on the free fraction (fp) of [11C]WAY100635 in plasma, with an age-associated increase in men and analogous decrease in women, that did not impact relationships of age, sex, and hormone concentration with WAY100635 BPP (Supplementary Table B). Altanserin fp, low as described previously (Haugbol et al, 2007), was not related to age or sex variables.

In the full sample of men and women, 5HT1A cerebellar VT trended toward a decline with age (β=−0.03 per decade of life, p=0.07), that plateaued at later ages, as described by a significant age2 association with VT (β <0.01 per decade, p=0.01). This later-age plateauing was also noted for women (β=0.01 per decade, p=0.002; Table 2).

Table 2 Model Results for Age, Sex, and Steroid Hormone Effects on 5HT1A Receptor BPP
Full size table

5HT2A receptor cerebellar VT increased with age in the full sample (β=0.08 per decade, p=0.002), as driven by this relationship in women (β=0.16 per decade, p=0.001). In women, estradiol concentration was also associated with increased cerebellar VT (β=0.30; p=0.03; Table 3).

Table 3 Model Results for Age, Sex, and Steroid Hormone Effects on 5HT2A Receptor BPP
Full size table

5HT1A Receptor PET Data

There was no significant age-by-sex interaction for 5HT1A receptor BP (see Table 2). 5HT1A receptor BP was reduced in women relative to men in postsynaptic prefrontal (LOF, SGC, PRE) and occipital (OCC) regions in the range of 16–19% (BPND) and 9–18% (BPP) (post hoc p= 0.03–0.10; Figure 1), but did not survive the post hoc regional correction for multiple testing (p<0.008). 5HT1A receptor BP was nonsignificantly lower in women relative to men in amygdala and raphe regions as well. Raphe 5HT1A receptor BP was inversely related to age or age2 (4.5% decrease per decade; p=0.05).

Figure 1
figure 1

Reduced cortical 5HT1A receptor binding in women relative to men. Postsynaptic 5HT1A receptor BPP was reduced in women relative to men in lateral orbitofrontal (9.3%; p=0.06), subgenual (10.5%, p=0.06), pregenual (12.4%, p=0.03), and occipital cortex (18.0%; p=0.03). The BPND and BPP results were entirely consistent (Supplementary Table A). Of note is one consistently extreme value in the male sample. Sensitivity analysis excluding this subject did not impact these results. This pattern of reduced 5HT1A receptor BPP in women could contribute to the greater risk for affective disorders in women relative to men!. *Extreme outliers, °outlier.

PowerPoint slide

Full size image

In the sex-stratified model for postsynaptic 5HT1A receptor BP in men, there was a small inverse relationship between BP and age (p=0.05) and FAI (p=0.06), accounting for only 1–9% of the variance in BP. These results remained at the borderline of significance in a sensitivity analysis that excluded an older male outlier with low BP values. Large magnitude inverse relationships between hippocampal BP and age (β=−0.74; p=0.09) and log10 (FAI; β=−5.56; p=0.19) were not significant. The inverse relationship between 5HT1A receptor BP and FAI appeared to be driven by high receptor BP in several men <50 years old with low FAI values (Supplementary Figure B1). Presynaptic raphe 5HT1A receptor BP tended to be inversely related to age in men (β=−0.26; p=0.096), but did not survive the sensitivity analysis.

For women, increased postsynaptic 5HT1A receptor BP was associated with increasing age (p=0.005; Figure 2) and estradiol concentration (p=0.004; Supplementary Figure B2; note that slope is small). These associations were small to medium, accounting for 12–21% of the variance in BP in these regions. The post hoc regional analyses showed greatest associations in lateral orbitofrontal (p=0.02), occipital (p=0.02–0.4), and pregenual cortices (0.06–0.08). Figure 2 displays how small the sex-related differences are for age–5HT1A receptor associations. The positive relationship between 5HT1A receptor BP and estradiol appeared to be driven by low receptor binding in several premenopausal women with low estradiol values (Supplementary Figure B2). Neither reproductive hormones nor age were significantly related to raphe BP.

Figure 2
figure 2

Age–5HT1A receptor BPP relationships in men and women. There are small differences in sex–age relationships to 5HT1A receptor BP in LOF. For women, increased postsynaptic 5HT1A receptor BP was associated with increasing age (p=0.005). In men, there was a tendency for increasing age to be associated with stable or decreasing BP.

PowerPoint slide

Full size image

5HT2A Receptor PET Data

There was no significant age-by-sex or age2-by-sex association with 5HT2A receptor BP (see Table 3). 5HT2A receptor BP decreased significantly with age (p<0.001) as expected, which persisted in the sex-stratified model for men and women. There was a very small, positive, curvilinear relationship of age to 5HT2A BP in the full sample and in women (mean β <0.001 per decade), but not in men (Figure 3). The significance of the divergence of model fits for men and women is diminished by the lower number of data points at the extremes of age for men (only 1 man <23 years relative to 3 women) and women (only 2 women over age 70 years relative to 4 men) and absence of data for several age categories in the women (27–38 and 70–78). There was a trend for women to have higher 5HT2A receptor BPP relative to men in amygdala (p=0.02) and pregenual cortex (p=0.6), which was not confirmed in the BPND model (Supplementary Table C). Neither estradiol nor FAI was significantly associated with 5HT2A receptor BP.

Figure 3
figure 3

Age-related decreases in cortical 5HT2A receptor BPP in men and women. The age-related decline (∼8% per decade in the full sample) in lateral orbitofrontal cortical 5HT2A receptor BPP is representative of all regions of interest. The significance of the distinct model fits for men and women (linear vs curvilinear) is diminished by the lower number of data points at the extremes of age for men (only 1 man <23 years relative to 3 women) and women (only 2 women over age 70 years relative to 4 men) and absence of data for several age categories in the women (27–38 and 70–78 years).

PowerPoint slide

Full size image

DISCUSSION

We found small-to-moderate size associations between 5HT1A/5HT2A receptor BP, sex, age, and reproductive hormones but no age-by-sex interaction relationships. There was a mean 15% reduction in 5HT1A receptor BP in women relative to men in prefrontal and occipital cortex (post hoc p=0.03–0.10). Age and reproductive hormones were associated with increases in 5HT1A receptor BP in women (p=0.004–0.005) and borderline significant decreases in 5HT1A receptor BP in men (p=0.05–0.06). Hormone-5HT1A receptor relationships were driven by extreme 5HT1A receptor data in subjects <50 years of age with low hormone concentrations. The significant decline in 5HT2A receptor BP relative to age (8% per decade) was similar for men and women and was not related to reproductive hormones. Similarity of regression results for 5HT receptor BPND and BPP suggests negligible confounding influences on relationships between age and VND. These reported findings are discussed below for each receptor system separately.

Serotonin-1A Receptor Binding

Presynpatic raphe nucleus 5HT1A receptor BP decreased with age (4.5% decrease per decade) in the full sample (6% variance in BP; p=0.05), which appeared to be driven by trend-level findings in men. The relationship of age with postsynaptic 5HT1A receptor BP tended to be sex dependent, wherein increasing age was associated with increasing BP in women, but stable or decreasing BP in men (Figure 2). These relationships were greatest for women in neocortical regions and for men in hippocampus and raphe. The sex–hormone–5HT1A receptor BP relationships were also sex dependent, wherein increasing FAI was associated with reduced 5HT1A receptor BP in men (most strongly in hippocampus), but increasing estradiol was associated with increased BP in women (primarily neocortex). These hormone–5HT1A receptor relationships were driven by extreme 5HT1A receptor BP values in subjects <50 years of age with low hormone concentrations. The combined age and hormone effects were small in men, accounting for 1–9% of the variance in BP, and small to moderate in women, accounting for 20% of variance in BP. Our finding of 15% neocortical 5HT1A receptor BP decreases in women relative to men (post hoc p=0.03–0.10), which exceeds that of test–retest variability in cortical regions (Parsey et al, 2000), but did not reach significance in the overall regression or regional post hoc tests, suggests that this finding warrants additional study. Notably, we observed consistent 5HT1A receptor reductions in women relative to men in all brain regions (except in hippocampus where BP was equivalent), although the smaller size of amygdala and raphe regions increased 5HT1A receptor BP measurement variability, thus decreasing the significance of differences.

An incremental decline in presynaptic raphe 5HT1A receptor BP with aging, particularly for men, and the trend toward an aging-related decline in postsynaptic 5HT1A receptor BP in men, most prominently in hippocampus, could be contributory to neural mechanisms for late-life depression and suicide given strong associations of 5HT1A receptor deficits in these regions with the pathogenesis of MDD and suicide (Arango et al, 1995, 2001; Blier and DeMontigny, 1987; Bowen et al, 1989; Drevets et al, 1999; Mann, 1999; Meltzer et al, 2004; Parsey et al, 2006; Sargent et al, 2000; Stockmeier et al, 1998). Furthermore, a specific age-related decline in hippocampus in men and women could also contribute to aging-related risk for cognitive disturbance (Sarnyai et al, 2000). Androgens were also associated with reductions in 5HT1A receptor BP, similar to preclinical reports (Ricci et al, 2006; Simon et al, 1998; Zhang et al, 1999), a relationship most robust in men <50 years old. That age and androgen concentration were both associated with 5HT1A receptor reductions, however, argues against a mediating role for male steroid decline in aging-related 5HT1A receptor decreases, given that androgen concentrations decrease with aging. Conversely, it is conceivable that the inverse relationship between androgen concentrations and 5HT1A receptors might mitigate against larger 5HT1A receptor BP decreases in men in late life. In our previous study where we reported an age-associated decrease in postsynaptic 5HT1A receptor BP in men, there were small age-related increases in [11C]WAY100635 VND (Meltzer et al, 2001). It is notable that in the larger sample in the current study, we observed small age-related decreases in [11C]WAY100635 VND (p=0.054), which may explain why a stronger age-related decline was not detected in this study.

For women, the apparent sex-based decrease in neocortical 5HT1A receptor BP, relative to men, appears to be mitigated in later life, as women had age-associated increases in BP. This pattern parallels the epidemiological statistics of greater risk for affective disorders in reproductive-aged women relative to men, which equalizes when women reach menopause (Bebbington et al, 1998; Kessler et al, 1993). A positive relationship between estradiol and 5HT1A receptor BP was unexpected based on animal literature data of 5HT1A receptor density and expression reductions after estradiol administration (Osterlund et al, 2000; Osterlund and Hurd, 1998). Premenopausal women with both low estradiol concentrations and low 5HT1A receptor BP appeared to be driving this association. As with men, in women, age and sex steroids appeared to have opposite effects on 5HT1A receptor BP, which may mitigate against larger age-related changes.

Whether the sex, age, and hormone effects accounting for 9–20% of the variance in 5HT1A receptor BP exert behavioral effects is not known from this study design. Certainly, the complete absence of 5HT1A receptors was associated with high levels of behavioral inhibition and depressive behavior (5HT1A knockout model; Ramboz et al, 1998) and reductions were associated with MDD (Drevets et al, 2000) and anxiety disorders (Lanzenberger et al, 2007). As postsynaptic 5HT1A receptors also serve functions such as gating action potentials of hippocampal CA1 pyramidal cells and stimulating glial release of trophic factors that promote 5HT neuronal outgrowth (Azmitia, 1999), it is conceivable that small age–sex–hormone associations with 5HT1A receptor binding could increase affective disorder risk in vulnerable individuals. Furthermore, extremely high androgen or low estradiol concentrations in young men and women, respectively, appear to have important effects on 5HT1A receptor BP.

The validity of the 5HT1A receptor BP findings is strengthened by our control for potential age–sex effects on plasma fp and nonspecific binding (VND). Although the age-by-sex interaction effect on WAY100635 fp is noteworthy (age-associated increase in men and decrease in women), and replicates a previous trend (Parsey et al, 2002), we confirmed that this did not impact BPP model results when fp was added as a covariate (Supplementary Table B). Furthermore, we confirmed that BPND model results (Supplementary Tables A and C), which do not rely upon fp, were consistent with BPP model results. Similarly, the trend finding of age-associated decreases in [11C]WAY100635 cerebellar VT in the full sample did not influence the model results because of close consistency between BPND and the BPP model results, the latter of which does not rely on free and nonspecific radiotracer binding in brain. We found no sex differences in [11C]WAY100635 VND.

Serotonin-2A Receptor Binding

We found ∼8% decreases in 5HT2A receptor BP per decade of age across the full sample of men and women, similar to that reported previously (Adams et al, 2004). Although we were able to replicate the finding of Sheline et al (2002) of a nonlinear inverse relationship between BP and age in women, lack of data for age extremes (particularly for younger men and older women) limits the ability of these data to support differential age-related patterns of 5HT2A receptor decreases in women vs men (Figure 3).

Endogenous estradiol was not associated with 5HT2A receptor BP in women, which replicates one study in women (Frokjaer et al, 2008), but contrasts with a study in healthy men that described a positive association between estradiol and neocortical 5HT2A receptor BP (Frokjaer et al, 2010). It is noteworthy that the regression coefficient (0.06–0.17) for estradiol in our models of neocortical 5HT2A approximates that found in men (Frokjaer et al, 2010), suggesting we may have been underpowered to detect the significance of an estradiol–5HT2A relationship in our female subgroup. FAI was not associated with 5HT2A receptor BP; however, we did not measure estradiol in men and therefore cannot assess whether aromatization of testosterone to estradiol was related to receptor BP. The trend finding of increased 5HT2A receptor BPP in women is negated by the finding of no sex difference for 5HT2A receptor BPND in this sample and the absence of sex differences in a larger cohort (n=136) (Erritzoe et al, 2009), and other studies (Biver et al, 1996; Frokjaer et al, 2009; Meyer et al, 1999; Rosier et al, 1996).

Increased 5HT2A receptor binding is associated with neuroticism, depression, suicide, and eating disorders. Hypotheses abound regarding whether these findings relate to genetic expression of the receptor or regulation relative to central 5HT tone. In the case of aging, the loss of 5HT2A receptors is expected to be a result of aging-related loss of neuropil, as these receptors predominantly are on pyramidal cells. Because of the role of 5-HT2A receptors in stimulating downstream neurons (Barnes and Sharp, 1999; Sharp et al, 2007), their diminution over aging are hypothesized to contribute to aging-related deficits in cognition, sleep, and mood (Meltzer et al, 1998).

Similar to what we noted for [11C]WAY100635, the validity of [18F]altanserin findings is strengthened by our control for potential age–sex effects on plasma free radiotracer fraction (fp) and nonspecific binding (VND). For [18F]altanserin, increasing age was associated with increasing cerebellar VT in the full sample as previously reported (Adams et al, 2004; Eastwood et al, 2001; Meltzer et al, 1998) and in women alone. In women, estradiol concentrations were also associated with increasing cerebellar VT as previously described in men (Frokjaer et al, 2010). Age and estradiol relationships with VND may be contributory, but are not thought to influence the BP outcomes, as the same age and estradiol-associated decreases observed for altanserin BPND were also present for altanserin BPP. The tendency for increased cerebellar VT in women relative to men concurs with one (Adams et al, 2004) but not another study (Erritzoe et al, 2009), and does not appear to impact BP in our study. Some have considered using an alternative reference region, although the pons is fraught with its own problems (see Adams et al, 2004).

In contrast to a prior study (Adams et al, 2004), we found no age–sex relationships to metabolism of [18F]altanserin over the scan duration, which suggests that there were no age–sex discrepancies in blood–brain barrier-permeable radiolabeled metabolites (Smith et al, 1998). Additionally, there were no sex or age effects on altanserin fp, which concurs with absence of sex differences in fp reported previously (Adams et al, 2004). Nonetheless, it is important to note that the plasma-derived Logan BP is expected to underestimate the dual-input compartmental modeling BP value (this model accounts for blood–brain barrier passage of radiometabolites) by about 30%, whereas correlation between the BP values is very high (r∼1.0) (Price et al, 2002a, 2002b). This underestimation reflects overestimation of nonspecific binding in the cerebellar reference region.

Although BMI has been related to 5HT2A receptor BP in other studies (Erritzoe et al, 2009), the effect size in this data set was too small (β=1 × 10−5) to be considered clinically meaningful and therefore was excluded from the model.

CONCLUSIONS

The strengths of this study included: (1) robust and well-validated methods for quantification of atrophy-corrected 5HT1A and 5HT2A receptor BP with careful control for nonspecific effects of age and sex on radioligand free fraction and cerebellar VT, (2) characterization and standardization of reproductive status and associated endogenous hormonal exposures, (3) thorough screening for and exclusion of individuals with personal psychiatric history, and (4) relatively large subject sample by PET standards with a wide age span of men and women, although a higher number of altanserin scans in subjects over age 70 years would have been desirable.

In conclusion, aside from the well-known age effects to decrease 5HT2A receptor BP, at most 9–20% of 5HT1A and 2A receptor BP variance was explained by sex, age, and reproductive hormone effects and there were no age-by-sex interactions with receptor binding. Endocrine standardization minimized confounding introduced by endogenous hormonal fluctuations and reproductive stage and permitted us to detect small effects of sex, age, and endogenous sex steroid exposures upon 5HT1A binding. Although these effect sizes are small, they could potentially lower the threshold for illness in individuals with genetic and/or environmental vulnerabilities for affective disorders. In addition, low hormone concentrations in adults <50 years of age may be associated with more extreme 5HT1A receptor BP values, but remains to be studied further.