Hyperandrogenism Induces a Proinflammatory TNFα Response to Glucose Ingestion in a Receptor-Dependent Fashion

Abstract

Context:

Hyperandrogenism and inflammation are related in polycystic ovary syndrome (PCOS). Hyperandrogenemia can induce inflammation in reproductive-age women, but the mechanism for this phenomenon is unclear.

Objective:

We examined the in vivo and in vitro effects of hyperandrogenism on mononuclear cell (MNC)-derived androgen receptor (AR) status and TNFα release.

Design:

This study combined a randomized, controlled, double-blind protocol with laboratory-based cell culture experiments.

Setting:

This work was performed in an academic medical center.

Participants:

Lean, healthy, reproductive-age women were treated with 130 mg of dehydroepiandrosterone (DHEA) or placebo (n = 8 subjects each) for 5 days and also provided untreated fasting blood samples (n = 12 subjects) for cell culture experiments.

Main Outcome Measures:

AR mRNA content and TNFα release were measured before and after DHEA administration in the fasting state and 2 hours after glucose ingestion. TNFα release in the fasting state was also measured in cultured MNCs exposed to androgens with or without flutamide preincubation.

Results:

At baseline, subjects receiving DHEA or placebo exhibited no significant difference in androgens and TNFα release from MNCs before and after glucose ingestion. Compared with placebo, DHEA administration raised levels of T, androstenedione, and DHEA sulfate, and increased MNC-derived AR mRNA content and TNFα release in the fasting state and in response to glucose ingestion. Compared with MNC exposure to baseline concentrations of DHEA (175 ng/dL) or T (50 ng/dL), the absolute change in TNFα release increased after exposure to T concentrations of 125 and 250 ng/dL and a DHEA concentration of 1750 ng/dL. Preincubation with flutamide reduced the TNFα response by ≥ 60% across all T concentrations.

Conclusion:

Androgen excess in vivo and in vitro comparable to what is present in PCOS increases TNFα release from MNCs of lean healthy reproductive-age women in a receptor-dependent fashion. Hyperandrogenemia activates and sensitizes MNCs to glucose in this population.

Women with classic polycystic ovary syndrome (PCOS) exhibit hyperandrogenism (1). Circulating T in PCOS is often increased above the normal premenopausal female range (>70 ng/dL) but is considerably lower than the male range (>300 ng/dL), with some variance in these upper limits based on the assay selected for measurement (2, 3). Women with PCOS often exhibit insulin resistance, which increases the risk of developing hyperglycemia (4). The compensatory hyperinsulinemia is thought to promote the hyperandrogenism (5, 6). Hyperglycemia stimulates reactive oxygen species (ROS) generation from peripheral blood mononuclear cells (MNCs). ROS-induced oxidative stress activates nuclear factor κB (NFκB), the cardinal signal of inflammation that promotes transcription of TNFα, a known mediator of insulin resistance (7–9). We have previously shown that in PCOS, there is increased NFκB activation and altered TNFα release from MNCs after glucose ingestion that is independent of obesity and inversely correlated with insulin sensitivity (10, 11). Thus, glucose-stimulated inflammation may induce insulin resistance in PCOS.

In PCOS, fasting and hyperglycemia-induced NFκB activation and TNFα release from MNCs are also highly correlated with circulating androgens (11–13). MNCs of lean, healthy, reproductive-age women are not sensitive to hyperglycemia and do not exhibit a proinflammatory response to glucose ingestion (10, 12, 13). We recently reported that acute androgen administration to this uninflamed population to raise circulating androgens to the range observed in PCOS increased MNC-derived ROS generation, NFκB activation, and TNFα mRNA content in the fasting state, and in response to glucose ingestion (14, 15). Thus, hyperandrogenism is capable of promoting inflammation by activating and sensitizing MNCs to glucose. Previous studies suggest that androgens can mediate inflammation in a receptor-mediated fashion (16, 17). To our knowledge, this mechanism has never been explored in lean, healthy, reproductive-age women.

We performed a substudy that combined a previous double-blinded, placebo-controlled protocol with laboratory-based cell culture experiments to examine the effect of in vivo androgen administration and in vitro androgen exposure on MNC-derived TNFα release and androgen receptor (AR) status. We hypothesized that androgen excess in vivo and in vitro increases AR activity and TNFα release in lean, healthy, reproductive-age women.

Subjects and Methods

Participants

Sixteen lean, healthy women between 20 and 40 years of age were recruited for study participation. This is the same well-characterized cohort from our previous study on hyperandrogenism and inflammation (14). Some largely descriptive data from these participants included herein were presented in the previous publication. All participants had a normal body mass index (18 to 25 kg/m²) along with regular menstrual cycles lasting 25 to 35 days and a luteal range serum progesterone level consistent with ovulation (>5 ng/mL). Participants lacked skin manifestations of androgen excess or polycystic ovaries on ultrasound and had normal serum androgen levels.

None of the participants had diabetes or an inflammatory illness. Tobacco smokers were excluded, and no participant took any medication that could impact carbohydrate metabolism or immune function for at least 6 weeks before entering the study. None of the participants exercised on a regular basis for a minimum of 6 months before study participation. All participants provided written informed consent according to Institutional Review Board guidelines for the protection of human subjects.

Study design

An electronically generated predefined randomization schedule was used by a research pharmacist for block assignment of participants to receive 130 mg/d orally of micronized dehydroepiandrosterone (DHEA) (Spectrum Chemical and Laboratory Products) (n = 8) or identical placebo (n = 8) at study entry. Both preparations were administered at 9 pm daily for 5 days. Study personnel and participants were blinded to the treatment assignment. A 75-g oral glucose tolerance test (OGTT) was administered to all participants beginning at 9 am after a 10- to 12-hour overnight fast before and after treatment. The pretreatment OGTT was performed after the onset of menstruation (d 5 through 8) on the morning before beginning DHEA or placebo, and the post-treatment OGTT was performed on the morning after completing the 5 days of assigned treatment.

During each OGTT, blood samples were collected from an antecubital vein while fasting and at 10, 20, 30, 60, 90, 120, and 180 minutes after ingestion of a standard 75-g glucose load. Androgen trough levels were measured from the fasting samples. Insulin sensitivity was derived from the glucose and insulin levels obtained during each OGTT (IS_OGTT) using the Matsuda Index formula: 10 000 divided by the square root of (fasting glucose × fasting insulin) × (mean glucose × mean insulin) (18). All participants underwent dual energy absorptiometry to assess body composition on the day of the initial OGTT as previously described (14).

Real-time PCR

AR mRNA content was quantified by real-time PCR in MNCs isolated from blood samples obtained while fasting and 120 minutes after glucose ingestion during the pre- and post-treatment OGTT. Total RNA was isolated from MNCs (approximately 20 mg) using an RNAeasy kit (QIAGEN). Total RNA (1 mg) was treated with DNase (Life Technologies, Inc) and reverse-transcribed using TaqMan reverse transcription reagents (PE Biosystems) and oligo(dT) primers. Primer sequences for the AR (GenBank accession no. M_34233.1) were selected using PRIMER EXPRESS software (PE Biosystems). The AR primers were coamplified with a group of rRNA primers for housekeeping genes (ie, β-actin, ubiquitin C, and ribosomal protein L13a). The sequences of these primers are shown in Table 1. A reference value was calculated from the housekeeping gene signals with qbase PLUS software (Biogazelle) and used to normalize against differences in RNA isolation and degradation, and in efficiencies of reverse transcription and PCRs that remained constant with DHEA treatment.

Table 1.

Primers Used in Real-Time PCR

Gene	Sense Primers (5′→3′)	Anti-Sense Primers (5′→3′)
Androgen receptor	CAAAAGAGCCGCTGAAGGGAAACA	TTCTTCAGCTTCCGGGCTCCCAGA
β-Actin	TGACTGACTACCTCATGAAGATCC	CCATCTCTTGCTCGAAGTCCAG
Ubiquitin C	ACTACAACATCCAGAAAGAGTCCA	CCAGTCAGGGTCTTCACGAAG
Ribosomal Protein L13A	AACAAGTTGAAGTACCTGGCTTTC	TGGTTTTGTGGGGCAGCATA

Cell culture

MNCs were isolated from blood samples obtained while fasting and 120 minutes after glucose ingestion during the pre- and post-treatment OGTT and cultured for 24 hours as previously described (13, 19). Each culture well was seeded with an MNC concentration of 2.5 × 10⁶ cells/mL. MNCs obtained from untreated fasting blood samples of 12 participants were supplemented with varying concentrations of DHEA (175, 875, and 1750 ng/dL; Spectrum Chemical and Laboratory Products) or T (50, 125, and 250 ng/dL; Sigma-Aldrich) for 24 hours with or without a 1-hour preincubation with flutamide (20 mg/mL; Sigma-Aldrich), an AR antagonist. MNCs treated only with vehicle (0.1% dimethylsulfoxide) were used as control. Culture supernatants were collected and stored at −80°C until assayed for TNFα.

Plasma and serum measurements

Plasma glucose, insulin, and TNFα along with serum T, androstenedione, and DHEA sulfate (DHEA-S) were measured as previously described (14). Samples from each participant were measured in duplicate in the same assay upon study completion. The interassay and intraassay coefficients of variation for all assays were no greater than 7.4 and 12%, respectively.

Statistics

Data were analyzed using the StatView software package (SAS Institute). The primary endpoint was change from baseline in AR mRNA content and TNFα release between DHEA and placebo groups. Secondary outcomes were also evaluated as change from baseline within group. All values were initially examined graphically for departure from normality, and the natural logarithm transformation was applied as needed. Unpaired (DHEA vs placebo) and paired (before vs after treatment) Student's t tests were performed. For each participant, absolute change from baseline was calculated to assess treatment effects on TNFα release. However, percentage change from baseline was used to assess treatment effects on AR mRNA content in view of intersubject variability. Differences in the MNC-derived TNFα response among the different in vitro androgen conditions within groups were analyzed using repeated measures ANOVA, followed by Tukey's post hoc tests. Pearson linear regression was employed for correlation analyses using the method of least squares. Data are presented as mean ± SE, and results with a two-tailed α-level of 0.05 were considered to be significant.

Results

Age and body composition

The placebo group and the DHEA-treated group were similar in age (28 ± 2 vs 28 ± 3 y; P = .89), body mass index (23.0 ± 0.5 vs 22.1 ± 0.4 kg/m²; P = .22), percentage total body fat (35.8 ± 1.9 vs 31.0 ± 2.7; P = .17), percentage truncal fat (33.2 ± 2.3 vs 29.0 ± 2.9; P = .28), and the ratio of truncal fat to leg fat (0.97 ± 0.08 vs 0.90 ± 0.03; P = .47).

Serum androgen levels and insulin sensitivity

Before DHEA or placebo administration, the DHEA-treated group and the placebo group exhibited similar serum levels of T, androstenedione, and DHEA-S (Table 2). DHEA administration significantly (P < .002) raised all three androgen levels compared with placebo. IS_OGTT was similar in both groups before and after treatment.

Table 2.

Serum Androgen Levels and Insulin Sensitivity of Subjects

	Placebo (n = 8)		DHEA (n = 8)
	Before	After	Before	After
T, ng/dL	42 ± 4	45 ± 4	40 ± 5	123 ± 9a
Androstenedione, ng/mL	1.2 ± 1.0	1.5 ± 1.0	1.2 ± 0.4	2.2 ± 0.1a
DHEA-S, μg/dL	138 ± 16	147 ± 17	134 ± 24	589 ± 40a
ISOGTT	10.6 ± 0.69	8.2 ± 1.2	10.6 ± 0.9	10.6 ± 1.0

Values are expressed as means ± SE.

^aPlacebo vs DHEA after treatment, P < .002.

Response to in vivo androgen administration

The change from baseline in AR mRNA content and TNFα release from MNCs obtained while fasting was significantly (P < .04) higher after DHEA compared with placebo (Figure 1, A and B). Before DHEA or placebo administration, the change from baseline in AR mRNA content and TNFα release decreased and was similar in both groups after glucose ingestion. After DHEA or placebo administration, the change from baseline in AR mRNA content and TNFα release decreased once again after glucose ingestion in the placebo group, but increased in the DHEA-treated group and was significantly (P < .04) different between groups. The increased TNFα response after glucose ingestion in the DHEA-treated group in particular approached statistical significance (P = .05).

Figure 1.

Comparison between groups of the change from baseline in AR mRNA content (%) (A) and TNFα release (Δ) (B) from MNCs for fasting samples before and after (before vs after) DHEA or placebo administration (left); and for fasting and 2-hour postglucose ingestion samples for each OGTT (before, 0 vs 2; after, 0 vs 2) as a measure of the response to glucose ingestion before and after DHEA or placebo administration (right). After DHEA administration, AR mRNA content and TNFα release from MNCs were significantly higher compared with placebo in the fasting state (*, AR mRNA content, P < .03; TNFα release, P < .04), and in response to glucose ingestion (†, AR mRNA content, P < .01; TNFα release, P < .04).

The within-group analysis revealed a significant increase in the change from baseline in glucose-challenged TNFα release from MNCs after DHEA administration (−8.3 ± 3.8 vs 6.3 ± 2.7; P < .04) but no change after placebo. However, there was no significant change from baseline in glucose-challenged AR mRNA content after administration of DHEA or placebo. All results remained the same after age adjustment.

Response to in vitro androgen exposure

Compared with MNC exposure to baseline concentrations of DHEA (175 ng/dL), the change from baseline in TNFα release remained unaltered after exposure to a DHEA concentration of 875 ng/dL but increased significantly (P < .0001) after exposure to 1750 ng/dL (Figure 2A). Compared with MNC exposure to baseline concentrations of T (50 ng/dL), the change from baseline in TNFα release increased progressively and significantly (P < .002) after exposure to T concentrations of 125 and 250 ng/dL, respectively.

Figure 2.

A, TNFα release from MNCs after exposure to varying physiological concentrations of T and DHEA. Compared with MNC exposure to baseline concentrations of DHEA (175 ng/dL) or T (50 ng/dL), the absolute change (Δ) in TNFα release increased after exposure to T concentrations of 125 ng/dL (P < .007) and 250 ng/dL (P < .002), and a DHEA concentration of 1750 ng/dL (P < .0001). B, TNFα release from MNCs after exposure to varying physiological concentrations of T preceded by preincubation with the AR antagonist flutamide. Compared with a maximum T-stimulated baseline (100%) after MNC exposure to T concentrations of 50, 125, and 250 ng/dL, there was a significant (P < .0009) ≥60% reduction in TNFα release in response to flutamide preincubation, to a level that was similar to vehicle alone or flutamide within vehicle.

Compared with the amount of TNFα released in the presence of T alone, there was a significant (P < .0009) ≥ 60% reduction in TNFα release in response to flutamide preincubation across all T concentrations to a level that was similar to vehicle alone or flutamide within vehicle (Figure 2B).

Correlations

Serum T and DHEA-S levels after DHEA or placebo administration were positively correlated with the change from baseline in MNC-derived AR mRNA content and TNFα release in the fasting state for the combined groups (Table 3). However, the correlation between serum DHEA-S and fasting AR mRNA content only approached statistical significance (P = .08). These correlations were not significant for the individual groups but were similar in magnitude in the DHEA-treated group.

Table 3.

Pearson Correlations for the Combined Groups of Circulating Androgen Levels After DHEA or Placebo Administration vs Change From Baseline of TNFα Release and AR RNA Content in the Fasting State Signifying the Treatment Response to DHEA and Placebo, and During Post-Treatment Glucose Challenge Signifying the Post-Treatment Response to Glucose Ingestion

	AR RNA Content (% Change)	TNFα Release (Absolute Change)
Treatment response to DHEA and placebo
T, ng/dL
r	0.492	0.587
P	0.044a	0.017a
Androstenedione, ng/mL
r	0.340	0.392
P	0.197	0.134
DHEA-S, μg/dL
r	0.453	0.515
P	0.078	0.041a
Post-treatment response to glucose ingestion
T, ng/dL
r	0.656	0.422
P	0.006a	0.104
Androstenedione, ng/mL
r	0.304	0.384
P	0.243	0.142
DHEA-S, μg/dL
r	0.598	0.457
P	0.015a	0.048a

^aP < .05.

After DHEA or placebo administration, serum T was positively correlated with the change from baseline in MNC-derived AR mRNA content in response to glucose ingestion, and serum DHEA-S was positively correlated with the glucose-stimulated AR mRNA and TNFα responses for the combined groups (Table 2). Serum androstenedione was also positively correlated with the glucose-stimulated TNFα response in the DHEA-treated group (r = 0.88; P < .004).

After DHEA or placebo administration, the area under the curve for glucose excursion during the OGTT was positively correlated with the glucose-stimulated TNFα response (r = 0.50; P < .05), and the glucose-stimulated AR mRNA and TNFα responses were positively correlated with each other (r = 0.52; P < .05). Measures of body composition were not correlated with AR mRNA content, TNFα release, or insulin sensitivity in the fasting state, or in response to glucose ingestion.

Discussion

Our data clearly show for the first time that the inflammatory response from MNCs of lean, reproductive-age women induced by hyperandrogenism is mediated through the AR. Five days of oral androgen administration increased fasting AR mRNA content and TNFα release, with further increases in response to glucose ingestion. Post-treatment androgen levels are positively associated with glucose-stimulated changes in AR mRNA content and TNFα release in this lean cohort with normal adiposity. Furthermore, MNC-derived TNFα release is increased after in vitro androgen exposure and suppressed when pretreated with an AR antagonist. These results corroborate our previous reports (14, 15) to bolster the concept that, in PCOS, hyperandrogenism may be the progenitor of nutrient-induced inflammation and is independent of obesity or excess abdominal adiposity.

There is no evidence of inflammation in lean, healthy, reproductive-age women before initiation of treatment. TNFα release from MNCs is suppressed in the fasting state and after glucose ingestion before DHEA or placebo administration. The TNFα response to glucose ingestion mimics the results in our previous studies to reinforce the concept that mediators of inflammation are suppressed in lean, healthy young women as the normal postprandial response to glucose ingestion to maximize insulin signaling (10, 12, 20–22). The concomitant suppression of AR mRNA content in the fasting state and in response to glucose ingestion is a novel finding indicative of quiescent androgen action on the immune system in the absence of elevated circulating androgens, even during the metabolically dynamic circumstance after nutrient intake. Thus, it appears that MNCs in our normal study population are strictly regulated against aberrant dietary and hormonal influences.

In contrast, oral androgen administration stimulates an inflammatory response from MNCs in our previously uninflamed study population. The acute rise in serum androgen levels produced by DHEA treatment simultaneously increases AR mRNA content and TNFα release in the fasting state, and in response to glucose ingestion. These key findings highlight the ability of elevated circulating androgens to promote inflammation in a manner similar to what occurs in chronically inflamed, insulin-resistant individuals (23, 24). This effect is corroborated by our previous report of increases in MNC-derived NFκB activation and TNFα mRNA content after DHEA administration (14), along with the direct relationship of androgen levels with AR mRNA content and TNFα release in the current study under the same conditions. The direct relationship between glucose area under the curve during the OGTT and the TNFα response to glucose ingestion after DHEA or placebo administration showcases the impact of androgen-induced MNC sensitization in the promotion of inflammation. Additional support is provided by the direct relationship between glucose-stimulated AR mRNA content and TNFα release after DHEA or placebo administration. Thus, the in vivo data in this substudy corroborate our previous report that hyperandrogenism activates previously resting MNCs in the fasting state and subsequently increases MNC sensitivity to physiological hyperglycemia (14, 15). In PCOS, this increased sensitivity may contribute to the induction of insulin resistance.

Inflammation induced by hyperandrogenism is mediated in a receptor-dependent fashion. Moreover, other investigators have reported a direct proinflammatory effect of androgens on MNCs that requires androgen binding to its receptor (16, 17). Our data in lean, healthy, reproductive-age women are consistent with these previous studies, as evidenced by increased TNFα release after exposure to physiological concentrations of DHEA and T in culture and decreased TNFα release during T exposure when preceded by AR blockade with flutamide. The higher concentration of DHEA required to elicit an increase in TNFα release compared with T is most likely related to its weaker androgenic activity, which is obviated in vivo by peripheral conversion of DHEA to more potent androgens (25, 26). Thus, our in vitro data extend our in vivo findings by demonstrating the mechanism for androgen-induced inflammation in the population of interest.

The acute nature of the androgen-induced inflammatory response may account for the lack of alteration in insulin sensitivity in our lean healthy subjects. Insulin resistance observed in obesity is driven by chronically inflamed excess adipose tissue, particularly when present in the abdomen as a result of sustained NFκB activation and excess adipose tissue-related TNFα production (27–29). Excess abdominal adiposity is also present in almost one-third of lean women with PCOS but is not correlated with insulin sensitivity (12, 30). This suggests that the smaller inflammatory load represented by excess abdominal adiposity in the absence of increased body weight may not contribute substantially to the development of insulin resistance. Nevertheless, our lean study population was also free of excess abdominal adiposity, as indicated by a truncal fat to leg fat ratio below 1.0 (31), and there is no association between measures of abdominal adiposity and TNFα release or insulin sensitivity. It is also unlikely that any previously reported anti-inflammatory, insulin-sensitizing effects of DHEA treatment have prevented a worsening of insulin sensitivity because these effects have been reported mostly in animals in the context of severe immune challenges such as genetically induced obesity, sepsis, trauma, and hemorrhage (32–34). Finally, the modest sample size may be a contributor to the inability to detect an alteration in insulin sensitivity. Further studies are warranted to evaluate whether chronic androgen administration in a larger cohort can adequately alter MNCs to produce any substantial metabolic abnormalities.

In conclusion, androgen excess to the extent present in PCOS promotes inflammation, as evidenced by increases in AR mRNA content and TNFα release from MNCs in the fasting state and in response to glucose ingestion. This effect is mediated in a receptor-dependent fashion as illustrated by decreases in TNFα release during in vitro androgen exposure in the face of AR blockade. Furthermore, these data confirm our recent findings that hyperandrogenism activates and sensitizes MNCs to glucose ingestion in lean, healthy, reproductive-age women (14, 15).