Associations of plasma sex‐related hormone and protein levels and alcohol dependence

Abstract

Background and Aims

Sex‐related hormones and proteins may underlie sex differences in alcohol use disorder characteristics and consequences. Previous reports suggest steroid sex hormones may influence alcohol consumption behaviors while proteins that regulate their circulation levels are rarely investigated. Following up on our earlier study of individual sex‐related hormones' associations with alcohol dependence (AD), this study measured the associations between the combinations of sex‐related hormones and proteins and AD in a larger sample.

Design

Sex‐stratified case‐control comparison and clustering of plasma sex‐related hormone and protein levels, plus a genome‐wide association study on the AD‐associated hormone and protein combinations.

Setting

Addiction treatment programs in the United States.

Participants

Four hundred treatment‐seeking recently abstained AD patients (median = 24.5 days) and 388 age‐and‐sex‐matched controls from a community biobank (male:female ≈ 2:1).

Measurements

Plasma levels of total testosterone, estrone (E1), estradiol (E2), follicle‐stimulating hormone (FSH), luteinizing hormone (LH), sex hormone binding globulin (SHBG) and albumin.

Findings

In males, we found higher E2 [β = 0.13, P _unadjusted < 0.001, false discovery rate (FDR) = 0.005], progesterone (β = 0.01, P _unadjusted = 0.025, FDR = 0.054), LH (β = 0.12, P _unadjusted = 0.017, FDR = 0.044) and albumin (β = 0.09, P _unadjusted < 0.001, FDR = 0.005) in AD patients than controls; in females, we observed lower E1 (β = −0.24, P _unadjusted = 0.024, FDR = 0.113) and progesterone (β = −1.86, P _unadjusted = 0.026, FDR = 0.113) and higher albumin (β = 0.18, P _unadjusted < 0.001, FDR = 0.001) in AD patients than controls. Three principal components (PCs), reflecting hormone signatures, were statistically significantly associated with AD in each sex: PC1 (TT, E2 and SHBG; β = −0.15, P = 0.016), PC3 (E1, progesterone and albumin; β = 0.28, P = 0.007) and PC4 (SHBG and albumin; β = 0.34, P = 0.001) in males; PC3 (progesterone and albumin; β = −0.38, P = 0.009), PC4 (TT and albumin; β = −0.49, P = 0.001) and PC5 (TT and progesterone; β = 0.54, P = 0.004) in females. Statistically significant SNP × group associations were found in females between PC4 and polymorphisms in RP51004I9.1 (top SNP: rs6082693; P = 5.41E‐12) and in males between PC3 and a genomic locus that contained SLC35E4, DUSP18 and OSBP2 (top SNP: rs13053277; P = 1.77E‐07).

Conclusions

Three sex‐related hormone and protein signatures appear to be associated with alcohol dependence in each sex. Peripheral levels of sex‐related hormones and proteins, as well as their combinations, could be altered in people with alcohol dependence who have abstained for a few weeks compared with controls, and such differences may be sex‐specific.

INTRODUCTION

Sex differences are well‐known in alcohol use pattern, the development of alcohol use disorder (AUD), and the consequences of problematic alcohol use [1, 2, 3, 4]. Both psycho–socio‐cultural and biological factors contribute to these differences, but certain sex differences such as alcohol pharmacokinetics and vulnerability to the neurotoxic effects of alcohol are mainly biologically determined [1]. Steroid sex hormones (testosterone, estrogens and progesterone) have been shown to influence alcohol consumption behaviors in animal models and to be associated with alcohol use in humans and AUD risk later in life, with some degree of sex specificity [see review by Erol et al. [5]. For instance, increased testosterone levels are generally associated with higher alcohol use and AUD risk in males, whereas higher estrogen levels are generally linked to increased alcohol intake in females [5]. Conversely, chronic alcohol intake has also been shown to affect steroid sex hormone levels. This includes reduced testosterone and progesterone levels, but increased estradiol levels in males and increased estradiol in females. These changes may be attributed to disruptions in the hypothalamic–pituitary‐gonadal axis [see review by Rachdaoui and Sarkar [6]]. We also observed altered levels of sex hormones in individuals with AUD history, including higher testosterone and estradiol levels in males and higher testosterone levels in females [7]. Together, this evidence supports the role of steroid sex hormones in regulating sex differences in alcohol use behaviors and the development of AUD. Furthermore, the dynamics between these hormones and chronic alcohol use complicate their role in sustaining alcohol consumption and alcohol‐related medical conditions.

Several proteins regulate the biosynthesis of peripheral steroid sex hormones and their bioavailability in the circulation, including pituitary hormones, follicle‐stimulating hormone (FSH) and luteinizing hormone (LH) and blood proteins albumin and sex hormone binding globulin (SHBG). Briefly, gonadotropin‐releasing hormone (GnRH) is released from the hypothalamus. On reaching the anterior pituitary, GnRH stimulates the secretion of FSH and LH into the blood circulation. At the gonads, FSH and LH stimulate gamete production and maturation as well as steroid sex hormone production. These hormones in turn inhibit hypothalamic GnRH secretion, forming an inhibitory feedback loop. In blood, the majority of steroid sex hormones are protein‐bound, with only a small fraction remaining free to interact with receptors or diffuse into cells. Therefore, the ability of steroid sex hormones to elicit their actions also depends on the levels of steroid‐binding proteins in blood including albumin, SHBG (binds to testosterone and estradiol) and corticosterone binding globulin (binds to progesterone). In addition to steroid sex hormones, the levels of these regulatory proteins were found to differ between individuals with AUD and controls and they varied at different phases after drinking cessation [7, 8, 9, 10, 11, 12]. Therefore, the alteration in peripheral steroid sex hormone levels observed in individuals with AUD may result from upstream signaling of their biosynthesis, and their downstream functions may be affected by their bioavailability, justifying studying these molecules alongside steroid sex hormones. However, most studies to date examined the relationships between steroid sex hormones and their regulatory proteins independently (or in ratios) with AUD. Given their inter‐relatedness, analyzing these hormones and proteins together using multivariate methods may provide deeper insights into their dysregulations in AUD.

Using a population‐based biobank resource, the United Kingdom (UK) Biobank, we recently reported sex‐specific differences in the plasma levels of testosterone, estradiol, SHBG and albumin between individuals with and without a lifetime history of alcohol dependence (AD) [7]. Males with AD history had higher total testosterone and estradiol levels, but lower bioavailable testosterone and estradiol, whereas females with AD history had higher total testosterone levels (adjusted for menopause). Both sexes had higher SHBG and lower albumin levels in the AD group [7]. However, because UK Biobank did not measure progesterone, FSH and LH levels, their associations with AD could not be investigated. In a smaller sample, but with extensive AUD‐related phenotypes and more hormones and proteins measured (testosterone, estradiol, estrone, progesterone, FSH, LH and SHBG), we also found higher plasma FSH levels in males with AD compared to non‐AD males [9]. These results support sex‐specific dysregulation of peripheral steroid sex hormones and their regulators in individuals with AUD. Because the hormones and proteins measured are interdependent biologically and functionally, examining them in combinations may provide additional information on alcohol‐induced dysregulation of steroid sex hormones.

The aims of this study are to (1) replicate and extend our Ho et al. [9] findings on the associations between plasma levels of sex‐related hormones (testosterone, estrogens and progesterone) and proteins (FSH, LH, SHBG and albumin) and AD using a sample well‐phenotyped for AUD research, which is approximately 10 times larger than the prior study; (2) investigate the associations of plasma sex‐related hormones and proteins in combination (‘profiles’) with AD; and (3) explored the genetic associations of AD‐related sex‐related hormone and protein combinations. To address these aims, we first used linear regression to examine the associations of individual sex‐related hormones and proteins with AD, and in a complementary approach, we used principal component analysis to define combinations of sex‐related hormones and proteins and then examined their associations with AD. Finally, we performed a genome‐wide association study (GWAS) to identify genetic variants associated with the sex‐related hormone and protein combinations.

METHODS

Subjects

Through a previous open‐label clinical trial searching for biomarkers of acamprosate treatment response [13], we recruited over 420 treatment‐seeking individuals with AD admitted to residential or outpatient treatment programs at Mayo Clinic and Mayo Health System sites in southern Minnesota between 2008 and 2012. All participants had provided informed consent. All subjects fulfilled the criteria for the Diagnostic and Statistical Manual of Mental Disorders 4th edition Text Revision (DSM‐IV‐TR) AD and were diagnosed by board‐certified addiction psychiatrists. AD subjects were interviewed by a trained study coordinator using the Psychiatric Research Interview for Substance and Mental Disorder [14] to capture DSM‐IV diagnoses of substance use and other psychiatric disorders. Alcohol consumption pattern 90 days before the interview was recorded by Timeline Followback Calendar [15]. This study only included AD subjects who had consented for future use of their data and biospecimens and had ≥1 mL plasma sample collected before acamprosate initiation available (n = 400). Subjects using hormone replacement therapy were excluded, while females using oral contraceptives around blood draw were noted. The original trial was approved by Mayo Clinic Institutional Research Board (IRB 07–007204). This study, which used the trial data and biospecimens, was also approved under IRB 11–003714. This study was funded through a research grant from the National Institutes of Health (R21AA026875), although not otherwise pre‐registered, and the results should be considered exploratory.

For non‐AD controls, we initially selected 400 participants from the Mayo Clinic Biobank recruited between 2009 and 2016 [16], all of whom self‐reported never being diagnosed with alcoholism. They were individually matched to the 400 AD subjects by sex, age (difference within 4 years), race and the current use of birth control (females only). Participants who were on hormone replacement therapy or without 1 mL plasma sample available were excluded. Average alcohol consumption [by Alcohol Use Disorders Identification Test‐Concise (AUDIT‐C); [17]] and history of anxiety and depression were self‐reported in the biobank questionnaires. We subsequently performed a medical record review, which led to the exclusion of 12 selected participants because of a history of AD [see International Classification of Diseases (ICD) codes excluded in Table S1] or pregnancy at the time of blood collection.

Hormone and protein measurements

Plasma samples were extracted from peripheral venous blood. 1 ml plasma was used for the measurement of total testosterone (TT), estrogens [estrone (E1) and estradiol (E2)], progesterone, FSH, LH, SHBG and albumin. Method details for measuring these sex‐related hormones and proteins were identical to our previous report [9] and are described in Supplementary Methods. The measurements were performed by the Mayo Clinic Immunochemical Core Laboratory and their staff was blinded to the group identity and sex of the samples.

Bioavailable and free testosterone levels (BioT and FT, respectively) and bioavailable and free estradiol levels (BioE2 and FE2, respectively) were calculated as described previously [7].

Comparisons between alcohol‐dependent and control groups

Demographic and clinical characteristics were compared between groups using χ² or Fisher's exact test (if more than 50% of cells with lower than 5% frequency) for categorical variables and Student's t test for continuous variables.

Analyses were stratified by sex. Associations between each plasma sex‐related hormone/protein level and case–control status were examined using multivariable linear regression models adjusted for age and body mass index (BMI) for males, and age, BMI and menopause status (estimated as above 51 years [18]) for females. Hormone and protein data were natural log transformed for analysis because of their skewed distributions. Progesterone was treated as a continuous variable with a detection limit of 0.1 ng/mL, and a binary classification was also used for values below this limit (detected vs. not detected). Uncorrected P‐values are presented without multiple testing adjustment because of non‐independence across hormone and protein comparisons, allowing readers to implement their own P‐value threshold for statistical significance. A Bonferroni‐adjusted significance level of P < 0.00625 is noted, which corresponds to eight independent tests (for the eight measured hormones and proteins) at α = 0.05.

Associations of sex‐related hormone and protein combinations with alcohol dependence

Because the biological relationships among the measured steroid sex hormones and related proteins were known (e.g. E2 is derived from testosterone, SHBG and albumin bind to testosterone/E2), and these hormones and proteins are highly correlated statistically, we aimed to identify sex‐specific combinations of the measured sex‐related hormones and proteins associated with AD using a multivariate method. We used principal component analysis (PCA) to identify independent hormone/protein signatures in males and females separately; each independent principal component (PC) defines a different signature comprised of a linear combination of correlated hormones and/or proteins. Factor loadings above 0.7 have been suggested to indicate a strong relationship between the factor and the variable, while between 0.4 and 0.7 indicate a moderate relationship, and below 0.4 a poor relationship [19]. Therefore, we selected a threshold of factor loadings ≥|0.4|for a hormone/protein to be considered important for the corresponding PCs. The resulting PCs were entered into a sex‐stratified logistic regression model with adjustment of age and BMI to determine their associations with AD. PCs identified in males are denoted by the prefix ‘M’ (i.e. MPC) while those identified in females are denoted by prefix ‘F’ (i.e. FPC).

Genetic associations of sex‐related hormone and protein profiles

We performed GWAS on the hormone/protein profiles significantly associated with AD in each sex. Genome‐wide genotyping had been conducted on the AD cases previously [20]. Controls were genotyped by Infinium OmniExpressExome BeadChip Array (Illumina, San Diego, CA) at the Mayo Clinic Genome Analysis Core. Details on quality control and imputation are described in Supplementary Materials. For each SNP, linear regression models were fit to assess the effect of single nucleotide polymorphism (SNP) on the AD‐associated PC's factor score (i.e. eigenvector), adjusted for sex‐specific covariates. For males, the models were adjusted for age, BMI and ancestry PCs, and for females, the models were also adjusted for menopause status and oral contraceptive use. Additional GWAS were conducted with the SNP × AD group interaction term on the AD‐associated PC's factor score. This interaction term, which assessed differential associations between AD cases and controls, was of primary interest. Genome‐wide significance threshold was set at P < 5E−08. Candidate SNP annotation, functional mapping and gene‐set analysis (by MAGMA) was performed in FUMA [21]. Using GWAS summary statistics, MAGMA maps SNPs to genes to obtain gene‐level association results, and tests whether genes in a gene set (17 023 canonical pathways from MsigDB v2023.hHs [22]), are more strongly associated with the phenotype of interest than other genes not in the gene set [23]. A gene set with Bonferroni‐corrected P < 0.05 was considered as significantly associated.

RESULTS

Characteristics of AD subjects and controls

Table 1 shows the comparisons of subject characteristics between the AD and control groups. AD and control subjects did not differ in the distributions of sex, age and race (P‐values are close to 1), reflecting subject matching was successful. Approximately 24% of female subjects were 51 years old and above (classified as postmenopausal in subsequent analyses) and less than 10% of subjects were non‐White. Nearly 65% of AD subjects (n = 251) were current smokers compared to only 2.6% in controls (n = 10).

TABLE 1.

Subject characteristics.

Characteristics	Males			Females
	Controls	AD	P	Controls	AD	P
n	257	268		131	132
Age (mean y, SD)	41.9 (11.4)	41.9 (11.4)	0.960 a	41.8 (11.9)	41.8 (11.9)	0.996 a
>51 y (n, %)	56 (20.9)	55 (20.5)	0.915 b	33 (24.8)	32 (24.1)	0.887 b
Oral contraceptives (n, %)	–	–	–	9 (6.8)	9 (6.8)	1.000 b
Hormone replacement therapy (n, %)	0	0	–	0	0	–
Race (n, %)
White	238 (92.6)	248 (92.5)	1.000 c	122 (93.2)	123 (93.2)	1.000 c
Black	4 (1.6)	4 (1.5)		0	0
Asian	1 (0.4)	1 (0.4)		0	0
Native American	1 (0.4)	1 (0.4)		0	0
Other	6 (2.3)	6 (2.2)		2 (1.5)	2 (1.5)
Unknown	7 (2.7)	8 (3.0)		7 (5.3)	7 (5.3)
Body mass index (kg/m2, SD) d	29.4 (5.9)	28.8 (5.5)	0.205 a	27.5 (6.7)	28.1 (6.2)	0.522 a
Current smokers (n, %) e	9 (3.5)	169 (64.5)	<0.001 c	1 (0.8)	82 (64.1)	<0.001 c

Abbreviation: AD, alcohol dependence.

^a Student's t test.

^b χ² test.

^c Exact test.

^d No. of subjects excluded because of missingness: 6 male controls, 17 male cases, 7 female controls, 9 female cases.

^e No. of subjects excluded because of missingness: 6 male cases, 4 female cases.

Alcohol consumption characteristics and the rates of mood comorbidities are listed in Table 2. Because of the differences in the data collection tools mentioned in the method section, these variables were not statistically compared between groups. However, AD subjects reported much greater alcohol consumption via timeline followback (TLFB) than controls using AUDIT‐C, and AD subjects reported higher rates of anxiety disorders and major depression as captured by Psychiatric Research Interview for Substance and Mental Disorders (PRISM) compared to self‐report anxiety and depression in controls.

TABLE 2.

Alcohol consumption and mood comorbidities in controls and AD subjects.

	Males	Females
Controls
n	257	131
Past 12 mo drinking frequency (n, %) a
Every day or nearly every day	24 (9.6)	11 (9.1)
2–3 days/week	55 (22.0)	19 (15.7)
2–4 times/mo	67 (26.8)	24 (19.8)
<1 time/mo	51 (20.4)	24 (19.8)
Not in last y	53 (21.2)	43 (35.5)
Past 12 mo drinking amount per drinking day (n, %) a
0–2 drinks	157 (80.9)	76 (100.0)
3–4 drinks	37 (19.1)	–
5–6 drinks	–	–
7–9 drinks	–	–
10 or more drinks	–	–
Past 12 mo had 6 or more drinks on one occasion (n, %) a
Never	113 (58.3)	56 (77.8)
Less than monthly	64 (33.0)	15 (20.8)
Monthly	11 (5.7)	1 (1.4)
Weekly	6 (3.1)	–
Daily or almost daily	–	–
AUDIT‐C score (mean; SD)	2.3 (1.9)	1.6 (1.6)
Self‐report history of anxiety (n, %)	40 (16.2)	29 (23.8)
Self‐report history of depression (n, %)	44 (17.8)	27 (22.1)
AD subjects
n	268	132
Days between last drink and blood draw (mean day, SD)	25.2 (15.9)	23.2 (16.3)
Past 90 days drinking pattern (mean, SD)
Total standard drinks	635 (538)	432 (422)
Total standard drinking days	46.6 (27.5)	43.8 (26.4)
Average standard drinks per drinking day	13.2 (8.4)	9.6 (6.2)
Anxiety disorders (n, %) b	72 (28.2)	51 (40.2)
Major depression (n,%) b	47 (18.4)	45 (35.4)

Abbreviations: AD, alcohol‐dependent; AUDIT‐C, Alcohol Use Disorders Identification Test‐Concise.

^a Based on self‐report AUDIT‐C questions in the Mayo Clinic Biobank questionnaire.

^b Captured by Psychiatric Research Interview for Substance and Mental Disorders (PRISM).

Plasma sex‐related hormone and protein differences between AD and control groups

Comparisons of plasma sex‐related hormone and protein levels between AD and control subjects are shown in Table 3. In males, AD subjects showed significantly higher E2 [P < 0.001, false discovery rate (FDR) = 0.005] and albumin levels (P < 0.001, FDR = 0.005) as well as higher progesterone (P = 0.025, FDR = 0.054) and LH levels (P = 0.017, FDR = 0.044) compared to controls with age and BMI adjustments. Progesterone was also detected in a larger proportion of AD subjects than in the controls (21.1% vs. 12.8%; P = 0.017, FDR = 0.044). BioE2 levels were also higher in AD subjects than controls (P = 0.015, FDR = 0.044). In females, albumin levels were significantly higher (P < 0.001, FDR = 0.001), while E1 (P = 0.024, FDR = 0.113) and progesterone levels (P = 0.026, FDR = 0.113) were lower in AD subjects than the controls after adjusting for age, BMI and menopausal status. In both males and females, although TT levels were relatively higher in AD subjects, their differences between groups were not significantly different (P > 0.05). FSH and SHBG levels were also not different between groups in both sexes (P > 0.50).

TABLE 3.

Plasma levels of sex‐related hormones/proteins by sex and AD groups.

	Males							Females
	CTL N	AD N	CTL Mean (SD)	AD Mean (SD)	β a	P a	FDR	CTL N	AD N	CTL Mean (SD)	AD Mean (SD)	β b	P b	FDR
TT (ng/dL)	249	250	379.2 (187.3)	404.9 (192.9)	21.90 (15.73)	0.164 c	0.213	124	122	20.2 (19.2)	21.0 (10.9)	0.13 (0.08)	0.075	0.217
E1 (pg/mL) ▼	247	249	43.8 (18.7)	45.0 (16.2)	0.05 (0.03)	0.109	0.157	122	121	137.4 (364.7)	73.2 (61.0)	−0.24 (0.11)	0.024	0.113
E2 (pg/mL) ▼	247	249	23.3 (12.0)	25.7 (10.3)	0.13 (0.04)	<0.001*	0.005	122	121	192.6 (669.8)	83.5 (96.5)	−0.25 (0.19)	0.184	0.342
PROG (ng/mL)	250	251	0.122 (0.063)	0.135 (0.075)	0.01 (0.01)	0.025	0.054	124	123	3.837 (8.439)	1.998 (4.551)	−1.86 (0.84)	0.026 d	0.113
FSH (IU/L) ▼	248	251	6.2 (7.9)	5.6 (4.8)	0.03 (0.06)	0.629	0.681	119	123	30.0 (37.3)	30.2 (38.2)	0.04 (0.12)	0.771 e	0.835
LH (IU/L)	249	251	4.6 (3.4)	5.0 (2.8)	0.12 (0.05)	0.017 f	0.044	123	123	19.1 (21.1)	18.9 (18.6)	−0.10 (0.14)	0.477	0.620
SHBG (nmol/L)	250	251	21.3 (10.5)	22.0 (11.5)	0.00 (0.05)	0.995 g	0.995	124	123	51.1 (53.5)	45.0 (29.2)	−0.01 (0.07)	0.852	0.852
Albumin (g/dL)	251	251	4.5 (0.4)	4.6 (0.3)	0.09 (0.03)	<0.001*, h	0.005	124	123	4.3 (0.3)	4.5 (0.3)	0.18 (0.04)	<0.001*, h	0.001
PROG Detected	250	251	32 (12.8%)	53 (21.1%)	0.60 (0.25)	0.017	0.044	124	123	46 (37.1%)	53 (43.1%)	0.29 (0.28)	0.303	0.492
FT (pmol/L) ▲	249	231	330 (149)	347 (141)	12.7 (12.4)	0.305	0.360	124	118	12.9 (11.5)	13.5 (7.5)	0.40 (1.15)	0.731	0.835
BioT (pmol/L) ▲	249	231	9360 (4324)	10 079 (4190)	599 (356)	0.092	0.150	124	118	352 (294)	386 (215)	27.9 (30.6)	0.361	0.521
FE2 (pmol/L) ▲	246	229	2.5 (1.3)	2.6 (1.1)	0.19 (0.11)	0.072	0.134	122	117	11.6 (29.0)	7.4 (8.6)	−4.6 (2.7)	0.096	0.217
BioE2 (pmol/L) ▲	246	229	72.3 (37.6)	78.9 (31.2)	7.5 (3.1)	0.015	0.044	122	117	300 (593)	214 (245)	−94.0 (57.2)	0.100	0.217

Note: Means (SD) of hormone/protein levels are based on the unadjusted models. BMI were missing in some subjects, therefore, sample sizes could be smaller than the total number of subjects.

Abbreviations: AD, alcohol dependence; BioT/E2, bioavailable testosterone/estradiol; CTL, controls; E1, estrone; E2, estradiol; FDR, false discovery rate; FT/E2, free testosterone/estradiol; PROG, progesterone; SHBG, sex hormone binding globulin; TT, total testosterone.

^▼ Log values of these hormones were used for statistical analysis.

^▲ Refer to Ho et al. [7] for calculation details.

^a Male model with adjustment = hormone ~ AD + age + BMI.

^b Female model with adjustment = hormone ~ AD + age + BMI + menopause.

^c When current smoking status was included in the model: β (SE) = −13.95(20.56), P = 0.498.

^d When current smoking status was included in the model: β (SE) = −2.58(1.23), P = 0.037.

^e When current smoking status was included in the model: β (SE) = 3.77(5.29), P = 0.477.

^f When current smoking status was included in the model: β (SE) = −0.18(0.37), P = 0.619.

^g When current smoking status was included in the model: β (SE) = −1.67(1.20), P = 0.167.

^h When time of blood draw was included in the model: β (SE) = 0.08(0.03), P = 0.0021 for males and β (SE) = 0.15(0.04), P = 0.0002 for females.

^* P < 0.05÷8 tests (for the 8 measured hormones and proteins) = 0.00625 (Bonferroni‐adjusted statistically significant threshold).

We also conducted a sensitivity analysis on the time of blood collection and found it to be significantly associated with albumin levels only in both sexes among controls (Table S2). However, even after adjusting for the time of blood collection, albumin levels remained significantly different between AD subjects and controls with similar effect sizes (Table 3 footnotes). Regarding current smoking status, because of the extremely low frequency of current smokers among controls (Table 1), we did not adjust for smoking in the primary analyses, and instead, conducted sensitivity analysis. Among AD cases, current smoking status was significantly associated with higher TT, LH and SHBG levels in males and higher PROG and lower FSH levels in females (Table S3). However, the statistical significance of the associations of these hormones/proteins with AD unchanged after adjusting for current smoking status (Table 3 footnotes).

AD‐associated sex‐related hormone and protein profiles

The PCA results are shown in Figure 1 and the scree plot is shown in Figure S1. In both males and females, the first eight PCs explained more than 80% of the variance in hormone/protein levels. In males, MPC1, MPC3 and MPC4 were significantly associated with AD (P < 0.05). Although in females, FPC3, FPC4 and FPC5 were significantly associated with AD (P < 0.05). PC2, PC6, PC7 and PC8 were not associated with AD in either sex (P > 0.05).

FIGURE 1.

Principal components (PC) of plasma sex‐related hormone and protein levels in (a) males and (b) females. The factor score of each hormone/protein in a PC is located in each box. Factor scores with absolute values ≥0.4 are bolded. The effect size and P‐values of the associations between each PC and alcohol dependence are listed at the bottom of each PC. (c) The sex‐related hormone and protein combinations (in terms of PCs) associated with alcohol dependence. The directions of association of individual hormones and proteins in each PC are indicated by upward and downward arrows for positive and negative associations accordingly.

In males, MPC1 explained 28% of the variance and TT, E1, E2 and SHBG all contributed substantially to MPC1 [i.e. abs (loadings) > 0.4] in the same direction. PC1 was significantly associated with AD (P = 0.016), corresponding to relatively higher levels of TT, E1, E2 and SBHG in cases. In females, FPC1 explained 37% of the variance, with major contributors also being E1, E2 and SHBG, but not TT. The association of FPC1 with AD was marginally significant in females (P = 0.072), corresponding to lower values of E1, E2 and SHBG in cases.

Interestingly, in both males and females, FSH and LH were significant contributors to PC2 (explained 22% of variance in both sexes) with loadings in the same direction, although neither MPC2 nor FPC2 was significantly associated with AD (P > 0.05).

In both sexes, progesterone and albumin contributed substantially to PC3. E1 also contributed substantially in males, but in the opposite direction of progesterone and albumin. MPC3 and FPC3 were significantly associated with AD in both males and females (P = 0.007 and 0.009, respectively), corresponding to higher values of progesterone and albumin in cases, along with lower values of E1 in males.

In males, SHBG and albumin contributed substantially to MPC4 in opposite directions, with the combination of high values of albumin and low values of SHBG significantly more common in cases compared to controls (P < 0.001). In females, TT and albumin contributed substantially to FPC4 in the same direction, with the combination of low values of TT and low values of albumin significantly more common in controls compared to cases (P = 0.001).

In males, progesterone and albumin contributed substantially to MPC5 in opposite directions, but MPC5 was not significantly associated with AD (P > 0.05). In females, TT and progesterone contributed substantially to FPC5 in the same direction, and the combination of higher values of TT and progesterone was significantly more common in controls compared to cases (P = 0.004).

Genetic associations of sex‐related hormone and protein profiles

We further explored potential genetic associations of the sex‐related hormone and protein combinations identified by PCA that were significantly associated with AD: MPC1, MPC3 and MPC4 in males and FPC3, FPC4 and FPC5 in females. GWAS results are presented in Figures S2–S7 and Tables S4–S9. Because these PCs were associated with AD and hence, were expected to have different SNPs associated with them depending on subject group, we focused on the regression models with the SNP × AD group interaction term, which may indicate differential associations between AD cases and controls.

We detected one genome‐wide significant association in females between FPC4 and a locus on chromosome 20 (Figure S6C). Five SNPs at this locus passed the genome‐wide significance threshold [rs6082693, rs199768, rs199770, rs1204627 and rs199773; P = 5.41E−12 to 4.79E−8; minor allele frequency (MAF) = 0.060–0.110] (Table S9). These SNPs are located close to or in the RP5‐1004I9.1 gene, encoding long intergenic non‐coding RNA (lincRNA) (Figure S8). FPC4 scores are lower among rs6082693 A allele carriers in controls, which is not observed in AD subjects (Figure S9A).

In males, we found one marginal genome‐wide significant association between MPC3 and a locus in chromosome 22 (Figure S3C). The top SNP (rs13053277; P = 1.77E–07; MAF = 0.086) and multiple SNPs at high linkage disequilibrium with it are located in a genomic risk locus that covers solute carrier family 35 member E4 (SLC35E4), dual specificity phosphatase 18 (DUSP18) and oxysterol binding protein 2 (OSBP2); another peak with the lead SNP rs35482156 (P = 4.27E–07; MAF = 0.115) is also located in OSBP2 (Figure S10). These SNPs represent expression quantitative trait loci (eQTL) for SLC35E4, DUSP18 and OSBP2 in various tissues (Figure S12 and Table S10).

No significant gene‐level associations or significantly associated gene sets were detected in all SNP × AD GWASs.

DISCUSSION

This study examined associations between AD and signatures of multiple inter‐related plasma sex hormones and proteins, in the largest sample of carefully phenotyped males and females with AD compared to age‐and‐sex‐matched controls. We observed significantly higher plasma E2 and albumin in AD males and higher plasma albumin in AD females after adjusting for menopause status. Using PCA, we identified three sex‐related hormone and protein signatures associated with AD in each sex. Subsequent sex‐specific GWASs revealed genomic loci that may contribute to these signatures. These findings offer novel insights into steroid sex hormone dysregulations in individuals with AD.

We found significantly higher plasma E2 levels in AD males compared to control males, aligning with previous reports [7] and the notion of male feminization by chronic excessive alcohol use [24, 25]. BioE2 was also higher in AD males, suggesting increased E2 production and bioavailability in these individuals even after approximately 3 weeks of abstinence. Moreover, plasma albumin levels were higher in AD males, with no differences in SHBG levels between groups. These observations contradict the findings in our UK Biobank study where individuals with AD history showed higher SHBG and lower albumin levels [7]. We also expected to observe a similar pattern in our AD subjects because of liver response to excessive alcohol consumption and malnutrition [10, 12, 26]. These discrepancies may be because our younger subjects exhibited hepatic function restoration after abstinence, contrasting to UK Biobank participants who were older adults (age 40–69) recruited from community without restriction on alcohol use before blood draw. This is supported by significant SHBG reductions reported in alcoholic patients after approximately 10 days of hospitalization [10, 12]. Interestingly, we also observed relatively higher LH levels in the AD group, potentially explaining the higher TT, progesterone detection and E2 levels. These findings imply partial restoration in brain‐level regulation of peripheral sex‐related steroid biosynthesis during short‐term abstinence in AD males, warranting further investigation.

In females, similar to males, we observed significantly higher plasma albumin levels in AD subjects compared to controls. However, unlike AD males, who showed elevated plasma E2 levels and a higher percentage of progesterone detection, AD females exhibited relatively lower plasma E1 and progesterone. This supports a sex‐specific association between steroid sex hormones and AUD. The between‐group differences (progesterone and E1) and similarities (TT) suggest that AD females might have an altered estrogen production rate, as progestogens are precursors of androgens and estrogens. Although E2 was not significantly different between groups, E1 can be transformed to E2 reversibly and can also bind to estrogen receptors [27], potentially supplementing insufficient E2 levels and complementing its functions. Furthermore, downregulating peripheral estrogens may discourage alcohol consumption in females, as degrading brain estrogen receptors have been shown to reduce binge‐like ethanol consumption in female mice during diestrus [28] and ethanol‐induced dopaminergic response in the ventral tegmental area was most robust in female mice when peripheral estradiol level was moderate or high [29]. More research is needed on estrogen biosynthesis and estrogen receptor expression during alcohol withdrawal and prolonged abstinence to elucidate estrogens' role in regulating drinking behaviors and moods during AUD recovery.

Importantly, we identified combinations of sex‐related hormones and proteins associated with AD using PCA, allowing us to consider multiple sex hormones and related factors simultaneously in a sex‐specific manner. In males, MPC1 was strongly associated with AD and significantly influenced by TT, E1, E2 and SHBG. This suggests dysregulation of sex‐related steroids and proteins persists in early abstained AD males. Factor loadings indicated enhanced testosterone and estrogen production in AD males, with higher SHBG levels potentially limiting their bioavailability. AD‐associated MPC3 showed notable loadings in progesterone (positive) and E1 (negative), suggesting a higher progesterone‐to‐estrogen ratio and increased albumin levels in AD males, contrasting recent reports [7, 30]. MPC4, also associated with AD, had significant loadings in SHBG and albumin in opposite directions, consistent with population‐based findings [7]. In females, FPC1 was significantly influenced by E1, E2 and SHBG, and was marginally associated with AD, with reductions in these hormones and proteins observed in AD females. These observations aligned with individual hormone/protein comparisons, but contrasted with FPC1 pattern in males. Three PCs (FPC3, FPC4 and FPC5) in females were significantly associated with AD and were predominantly influenced by TT, progesterone and/or albumin, indicating distinct effects independent of FPC1.

We also discovered notable genetic associations of AD with one hormone PC signature for each sex. In females, the top genomic locus associated with hormone FPC4 is located in the lincRNA gene RP5‐1004I9.1, which has unknown function and is expressed exclusively in the testis and minimally in the liver (Figure S11). In males, the top locus associated with MPC3 encompasses three genes: SLC35E4, DUSP18 and OSBP2, which are primarily expressed in the testis but also substantially in other tissues, including the adrenal gland (Figure S12). SLC35E4 is predicted to be an antiporter, DUSP18 is a dual specificity phosphatase involved in signaling pathways, and OSBP2 potentially reduces the cytotoxicity of oxysterols, which can induce inflammation and apoptosis and can be altered by alcohol. SNPs in high linkage disequilibrium with the top SNPs are eQTLs of these genes in various tissues, potentially influencing the hormone‐protein combination observed in AD males by affecting these genes' expression. Further molecular investigations are needed to elucidate the roles of these genes in regulating steroid sex hormones and related proteins under chronic excessive alcohol consumption and during abstinence.

We did not replicate our previous finding of higher plasma FSH levels in AD males compared to controls [9]. Instead, we observed a trend toward higher plasma LH levels in AD males, but the between‐group difference became more insignificant when adjusted additionally for current smoking status. Although subjects were recruited from the same study sites and methods were consistent, this sample was younger (on average 3 years younger in males and 4 years younger in females), with only a quarter older than 51 years compared to nearly half in the original smaller study. One methodological difference is including BMI as a covariate. Although BMI was not significantly different between AD and control groups, heavy drinking is associated with weight gain [31], suggesting our current analysis may be improved by accounting for inter‐individual body composition differences.

Limitations

First, this is a cross‐sectional study using plasma samples collected from AD subjects who abstained from alcohol for approximately 3 weeks, so the results might reflect both previous alcohol use and partial recovery. Future longitudinal studies with samples from different stages of alcohol use and abstinence length could provide deeper insights into those relationships. Second, time of blood collection had not been standardized in AD cases nor in controls and fasting was not required. We found the time of blood collection only impacted albumin, but the between‐group findings remained significant even with it included in the model (Table 3 footnotes). Third, potential socio‐economic and lifestyle confounders, such as exercise and diet, were not included as adjustment factors because these data were not systematically collected in both cases and controls. Although we did not adjust for current smoking status in primary analyses because of the low frequency of current smokers among controls, and despite a few measured hormones and proteins being significantly associated with current smoking status in AD subjects, including current smoking status did not alter the lack of statistical significance of their associations with AD (Table 3 footnotes). Fourth, information on menstrual cycle and menopause in AD females was unavailable, limiting the ability to study these factors. We inferred menopause status using the national average menopause age of 51 [18]. Fifth, most participants self‐reported as White, limiting generalizability of our findings to other race/ethnicity groups. Sixth, the progesterone assay had limited sensitivity, with nearly half of the samples below the detection limit, restricting the quantitative analyses such as calculating progesterone‐to‐estrogen ratio. Last, the GWAS sample size was small, limiting statistical power, especially for discovering differential variants between AD cases and controls.

CONCLUSIONS

We demonstrated significant sex‐specific associations of sex‐related hormone and protein signatures with AD in the largest study to date. Although we replicated some findings from existing literature, variations in the duration between last drink and blood draw complicated comparisons across studies. Our GWAS findings, although limited by sample size, suggest a potential role of genetic variations. Together, these findings highlight the need for larger longitudinal studies on sex‐related hormones and proteins during various stages of alcohol use and abstinence lengths to capture the dynamics of hormone fluctuations, as well as studies on the associations of these hormones and proteins with AUD‐related features such as craving and negative mood ratings, which may offer important insights into sex‐specific mechanisms underlying AUD and recovery.

AUTHOR CONTRIBUTIONS

Ada Man‐Choi Ho: Conceptualization (equal); project administration (equal); writing—original draft (lead); writing—review and editing (equal). Jennifer R. Geske: Data curation (lead); formal analysis (equal); writing—review and editing (supporting). Vanessa K. Pazdernik: Formal analysis (equal); visualization (equal); writing—review and editing (supporting). T. Cameron Waller: Formal analysis (equal); writing—review and editing (supporting). Anthony Batzler: Formal analysis (equal); writing—review and editing (equal). Stacey J. Winham: Conceptualization (equal); funding acquisition (lead); supervision (equal); writing—review and editing (equal). Victor M. Karpyak: Conceptualization (equal); funding acquisition (lead); supervision (equal); writing—review and editing (equal).

DECLARATION OF INTERESTS

None.

Supporting information

Figure S1. Scree plot.

Figure S2. Male: GWAS on PC1.

Figure S3. Male: GWAS on PC3.

Figure S4. Male: GWAS on PC4.

Figure S5. Female: GWAS on PC3.

Figure S6. Female: GWAS on PC4.

Figure S7. Female: GWAS on PC5.

Figure S8. Regional plot of top SNP in female PC4 SNP × AD GWAS: rs6082693.

Figure S9. Box plots of PC scores by subject group and by top SNP genotypes.

Figure S10. Regional plot of top SNP in male PC3 SNP × AD GWAS: rs13053277.

Figure S11. RP5‐1004I9.1 gene expression level across tissues.

Figure S12. SLC35E4, DUSP18, and OSBP2 gene expression level across tissues.

Table S1. ICD codes related to alcohol dependence used for control exclusion.

Table S2. Effects of time of blood collection on measured sex‐related hormones and proteins in controls.

Table S3. Effects of current smoking status on measured sex‐related hormones and proteins in AD subjects.

Table S4. Results of SNP*AD interaction GWAS on PC1 in males.

Table S5. Results of SNP*AD interaction GWAS on PC3 in males.

Table S6. Results of SNP*AD interaction GWAS on PC4 in males.

Table S7. Results of SNP*AD interaction GWAS on PC3 in females.

Table S8. Results of SNP*AD interaction GWAS on PC4 in females.

Table S9. Results of SNP*AD interaction GWAS on PC5 in females.

Table S10. Top SNPs of SNP x AD GWAS on PC4 in females which are eQTLs of SLC35E4, DUSP18 and OSBP2.

Ho AM‐C, Geske JR, Pazdernik VK, Waller TC, Batzler A, Winham SJ, et al. Associations of plasma sex‐related hormone and protein levels and alcohol dependence. Addiction. 2025;120(10):2107–2117. 10.1111/add.70104

DATA AVAILABILITY STATEMENT

The data that support the findings of this study will be openly available at dbGaP or a designated NIH‐approved repository. Note that only deidentifiable information will be available. Please contact the corresponding author for details.