Gonadal hormones are critical modulators of motivated behavior. Estrogens, progestins, and androgens control the expression of a range of naturally motivated behaviors (Beatty, 1979; Fahrbach and Pfaff, 1982; Pfaff, 1982), as well as influencing the acquisition, escalation, and relapse behaviors associated with addiction (Anker and Carroll, 2011; Becker et al., 2012). Therefore, identifying the neurobiological mechanisms by which hormones influence the reward circuitry of the brain is key to understanding their underlying impact on motivated behaviors, and to suitably tailor treatments in both men and women suffering from addiction (Becker et al., 2012).

One way that hormones can influence brain regions associated with reward is through modulating structural plasticity. In the nucleus accumbens (NAc), a primary component of the mesolimbic dopamine system, changes in dendritic spine density and morphology of medium spiny neurons (MSNs) occur in response to both natural rewards and drugs of abuse (Pitchers et al., 2010; Russo et al., 2010; Staffend and Meisel, 2012; Staffend et al., 2014). The structural changes induced by drugs of abuse in this region are incredibly persistent and have been found to correlate with aspects of drug-mediated behavior (Li et al., 2004; Robinson and Kolb, 2004), suggesting that they may be functionally relevant to long-lasting changes in synaptic signaling and behavior. Recent evidence has shown that sex steroid hormones also modulate spine density in the NAc. Our own lab has found that estradiol decreases MSN spine density in the NAc core (NAcC) of ovariectomized female rodents (Staffend et al., 2011; Peterson et al., 2014). Other work has shown that high doses of testosterone given to intact male rats in a model of anabolic-androgenic steroid abuse decreased spine density of MSNs, though this was found in the NAc shell (NAcSh) (Wallin-Miller et al., 2016).

These findings provide initial evidence that sex steroid hormones may differentially affect spine plasticity between the two subregions of the NAc. However, further study of androgen action in males is required to determine if this phenomenon exists outside of models of steroid abuse. These data also show a striking similarity in that both estrogens and androgens decrease spine density in the NAc, an unusual effect considering that both classes of hormone typically increase the density of spines in other regions of the brain (Woolley and McEwen, 1992; Leranth et al., 2003; Hajszan et al., 2007; de Castilhos et al., 2008; Christensen et al., 2011; Khan et al., 2013). Previous work in our lab has developed a model of the molecular mechanisms underlying estradiol-induced spine decreases in the female NAcC, finding that these changes occur through activation of mGluR5 and endocannabinoid signaling (Peterson et al., 2014, 2016). This mGluR5 signaling mechanism was also found to be critical for estradiol-driven potentiation of psychostimulant-induced behaviors in female rodents (Martinez et al., 2014, 2016). Whether hormone-induced NAc spine decreases and behavioral effects in males are also driven by mGluR5 is unknown. As mGluR5 activity has previously been associated with behavioral responses to drugs of abuse, and is considered a potential therapeutic target for addiction (Olive, 2009; Cleva and Olive, 2012), it is particularly important to understand how sex steroid hormones may work with this receptor to influence mesolimbic regions. Therefore, the aims of this study were to explore the effects and mechanism of androgen action on spine plasticity in the male NAc, with specific interest on the action of mGluR5.

To determine if androgen activity affects MSN spine density in the NAc, we treated castrated male rats with DHT, a potent agonist of the androgen receptor (AR). Spine density, head diameter, and length were measured from confocal images of dendritic segments of NAcC and NAcSh MSNs (Figure 1) 24 h after administration of DHT or oil control, a time point at which hormone-mediated spine changes often peak (Woolley, 1998). Systemic DHT treatment had no effect on spine density in the NAcC (Figure 2A), but did significantly reduce spine density in the NAcSh [Figure 2C, t(9) = 2.53, p = 0.032]. DHT had no effect on either spine length or head diameter in either subregion (Figures 2B,D). Although DHT cannot be aromatized into estradiol, a metabolite of DHT, 3β-androstanediol, is active at estrogen receptor β (ERβ) (Pak et al., 2005). Thus, to determine whether ERβ activation mediates DHT-induced spine plasticity in the male NAcSh, the ERβ agonist DPN was systemically administered to castrated males, and spine density and spine morphology were measured 24 h later. Systemic DPN had no effect on NAcSh spine density or morphology (Figure 3), suggesting that observed DHT effects are mediated by activation of ARs.

This spine density decrease in the NAcSh is reminiscent of NAcC spine decreases observed in female animals 24 h following the administration of estradiol. Further study of the mechanism of this estradiol-induced effect has shown that it is dependent on the activation of mGluR5 (Peterson et al., 2014). To determine if mGluR5 signaling is required for the DHT-induced spine decrease in males, we looked at the effect of DHT treatment on NAcSh spine density in animals pretreated with the mGluR5 antagonist, MPEP, or vehicle control. We found that the effect of DHT was dependent on mGluR5 activation [Figure 4A, hormone by drug interaction: F(1,14) = 21.97, p < 0.001], with animals that received MPEP prior to DHT exhibiting a higher spine density than those that were vehicle treated [F(1,14) = 25.69, p < 0.001]. Again, this plasticity appears to be restricted to changes in spine density, as there was no influence of treatment condition on spine length or head diameter (Figure 4B).

The restriction of DHT-induced spine changes to the NAcSh suggests an interesting sex difference in how primary gonadal hormones and mGluR5 influence NAc structure. In female rodents, estradiol/mGluR5 interactions only drive spine decreases in the NAcC (Peterson et al., 2014), despite the ability of both systemic and local mGluR5 activation to decrease spine density throughout the NAc (Gross et al., 2016). To see if this sex difference extends to mGluR5 signaling independent of hormonal manipulation, the mGluR5 agonist, CHPG, was infused into the NAc of intact males, and spines were analyzed 24 h later. Compared to infusion of saline, CHPG reduced spine density in the NAcSh [Figure 5C, main effect of CHPG: F(1,26) = 14.09, p = 0.010] and had no effect on spines in the NAcC (Figure 5A). Thus, activation of mGluR5 throughout the NAc produces the same subregion-specific spine changes as systemic DHT. In addition to assessing the effect of CHPG in this region, the role of endocannabinoid signaling in this process was tested. Activation of mGluR5 in the NAc is known to lead to the synthesis and release of endocannabinoids (Alger and Kim, 2011), and this mechanism has previously been implicated in estradiol-induced spine changes in females (Peterson et al., 2016). Endocannabinoid signaling also appears to mediate mGluR5 effects on spine plasticity in the NAcSh of male rats, as systemic pretreatment with the CB1R inverse agonist AM251 influenced the effect of CHPG on spines [drug by drug interaction: F(1,26) = 11.38, p = 0.019]. Specifically, AM251 pretreatment abolished the spine density decreases in the NAcSh following intra-accumbal CHPG [Figure 5C, F(1,26) = 7.50, p = 0.011]. Neither CHPG nor AM251 influenced head diameter or spine length in either region (Figures 5B,D).

Administration of exogenous androgens is known to be rewarding and produces CPP (Alexander et al., 1994). Infusion of testosterone or its metabolites specifically into the NAcSh is sufficient to produce this effect (Frye et al., 2002), suggesting one behavioral outcome of androgen-induced plasticity in this region. To determine if mGluR5 signaling is required for androgen-induced CPP, we looked at the effect of blocking mGluR5 activity prior to DHT injection. In this CPP model, animals were given i.p. injections of either vehicle or MPEP 30 min prior to conditioning with DHT (Figure 6A). All animals initially preferred the Aspen/gray compartment over the corncob/white, a preference that remains stable in animals injected with saline on both sides of the conditioning apparatus during the behavioral procedure (Figure 6B). Regardless of MPEP pretreatment, the amount of time spent in the non-preferred compartment during the 10 min posttest period was significantly increased after conditioning with DHT [Figure 6C; t(7) = 3.027, p = 0.019 for saline group; t(6) = 2.627, p = 0.039 for MPEP group], while the amount of time spent in the initially preferred (vehicle paired) compartment was decreased [t(7) = 5.776, p < 0.001 for the saline group; t(6) = 5.206, p < 0.002 for the MPEP group]. These data suggest that systemic blockage of mGluR5 signaling was not sufficient to influence the acquisition of DHT-induced CPP.

Although preliminary evidence suggested that androgens can modulate spine density in the NAc (Wallin-Miller et al., 2016), this effect had yet to be thoroughly characterized and the molecular mechanisms and behavioral outcomes of this plasticity were unknown. Here we used DHT, a potent agonist of ARs, in castrated rats to find evidence that AR activation selectively decreases MSN spine density in the NAcSh. The restriction of spine decreases to the NAcSh seen here is in line with previously observed testosterone-induced spine decreases in a rat model of anabolic steroid abuse, suggesting that this subregion is a specific target for androgen-induced structural remodeling (Wallin-Miller et al., 2016). Furthermore, we provide evidence that this occurs through a novel mechanism: DHT activation of mGluR5 signaling. These findings have implications both for the role of androgen signaling in motivated behavior and for the mechanisms of androgen action throughout the brain.

Steroid hormone regulation of mGluR signaling has thus far been focused on the study of estradiol (Micevych and Mermelstein, 2008). In multiple brain regions it has been found that surface localized estrogen receptors ERα and ERβ couple to and regulate the activity of various mGluR subtypes, exerting effects on structural plasticity, synaptic physiology, and behavior (Micevych and Mermelstein, 2008; Christensen et al., 2011; Boulware et al., 2013; Martinez et al., 2014, 2016; Peterson et al., 2014). Palmitoylation and interaction with caveolin are also molecular requirements for the coupling of ERs to mGluRs (Boulware et al., 2007; Meitzen et al., 2013). There is reason to think that this signaling model might generalize to other steroid hormones. ARs are also shown to localize to the cell membrane, a process that, similar to ERs, is dependent on the post-translational modification S-palmitoylation and interaction with integral membrane caveolin proteins (Tabori et al., 2005; Pedram et al., 2007; Bennett et al., 2009). Use of DHT, which cannot be aromatized into estradiol, and demonstration that ERβ activation, which can occur via DHT metabolites, has no effect on spine plasticity in the male NAcSh suggests that the observed effects are AR dependent. Additionally, the ability of local microinjection of CHPG into the NAc to produce the same structural effects of systemic DHT suggests that DHT activation of mGluR5 may be occurring site-specifically. The fact these local mGluR5-induced spine changes are mediated by endocannabinoid signaling further supports a common mechanism for both estradiol and DHT-induced spine changes (Peterson et al., 2016). We are keenly interested in the exact nature of this putative AR/mGluR5 interaction in the NAcSh and to determine if other androgen-mediated effects depend on mGluR signaling.

Our findings that both DHT and mGluR5 mediated spine plasticity are limited to the NAcSh uncovers interesting sex differences in both NAc mGluR5 signaling and its regulation by primary gonadal hormones. In females, mGluR5 activation decreases spine density in both the NAcC and NAcSh (Gross et al., 2016). The restriction of mGluR5-induced spine decreases to the NAcSh in males, therefore, indicates intrinsic organizational differences regarding how mGluR5 signaling influences this region between the sexes. This is in line with broader evidence that both sex differences in hormonal activation and intrinsic organization exist in the reward circuitry of the brain, with differences favoring greater responsivity in females to drugs of abuse (Hu et al., 2003; Becker and Hu, 2008). In the case of mGluR5 structural plasticity, estradiol seems to take advantage of the intrinsic sex difference, driving mGluR5 spine decreases only in the NAcC (Peterson et al., 2014). These structural changes are likely to be functionally relevant as dendritic spine changes specifically in the NAcC are correlated with the expression of psychostimulant-induced locomotor sensitization and in females, estradiol-enhancement of psychostimulant sensitization depends on the activation of mGluR5 (Li et al., 2004; Martinez et al., 2014). Conversely, the lack of androgen/mGluR5 activity in the NAcC may explain the more minimal effect of circulating gonadal hormones on psychostimulant-induced behaviors in males (Hu et al., 2003; Jackson et al., 2005).

In both males and females it appears that the endocannabinoid system is an important mediator of spine plasticity in the NAc. Carvalho et al. (2016) show that repeated agonism of cannabinoid receptors decreases MSN spine density in the male NAc. In females, spine decreases in the NAcC following repeated administration of estradiol and cocaine are dependent on endocannabinoid signaling (Peterson et al., 2016). Here we observe that endocannabinoid signaling following a single infusion of mGluR5 agonist is necessary for structural plasticity in the male NAcSh. Synthesis and release of endocannabinoids occur following activation of Gq coupled GPCRs, such as mGluR5. In the NAc, endocannabinoid signaling is critical for the induction of mGluR5-dependent LTD at corticostriatal synapses via reduction of presynaptic glutamate release (Robbe et al., 2002; Jung et al., 2005; Alger and Kim, 2011). This suggests that endocannabinoid signaling in the NAc is a mechanism underlying plasticity in both neuronal structure and function. How presynaptic changes in neurotransmitter release result in post-synaptic regulation of dendritic spine structure in this system remains unclear.

Despite the importance of mGluR5 signaling in mediating the structural effects of DHT, a dose of 1 mg/kg MPEP was not able to block the conditioning effects of DHT in a CPP experiment. Although this dose of mGluR5 antagonist was sufficient to abolish NAc spine changes induced by DHT, it is possible that it is unable to block other aspects of mGluR5-mediated plasticity that may contribute to this behavior. However, it is also likely that the neurobiological mechanisms engaged by conditioning with multiple injections of DHT differ from those engaged by the single dose of DHT used in our structural experiments and may not depend on mGluR5. Indeed, across a range of drugs the role of mGluR5 in acquisition of CPP is unclear. Although some studies provide evidence that mGluR5 activity is necessary for ethanol, cocaine, morphine, and nicotine CPP (Popik and Wrobel, 2002; McGeehan and Olive, 2003; Lominac et al., 2006; Yararbas et al., 2010), others have found that blocking mGluR5 activity has no effect on CPP for amphetamine, morphine, nicotine, ethanol, or the cross sensitization of male sexual experience with amphetamine CPP (McGeehan and Olive, 2003; Pitchers et al., 2016), while others still have found that blocking mGluR5 activity may actually potentiate CPP induced by drugs of abuse (van der Kam et al., 2009; Rutten et al., 2010). Differences in the doses of MPEP and behavioral protocols used in these studies may account for their opposing results. However, like other drugs of abuse, there is a relationship between dopaminergic signaling and the conditioned effects of androgens with antagonism of both D1 and D2 dopamine receptors having been shown to abolish testosterone CPP (Schroeder and Packard, 2000). Future work may more directly study this and other molecular mechanisms of androgen reward by using intracranial self-stimulation, as the induction and expression of CPP involves a complex set of learning processes (Huston et al., 2013). Future studies may also turn to other behaviors that involve the NAcSh and are sensitive to androgen signaling, such as social and decision-making behaviors. For instance, the NAc is an important mediator of testosterone-driven intermale aggression (Albert et al., 1989, 1992), and testosterone influences behavior in a risky-decision making task that is known to depend on NAcSh activity (Stopper and Floresco, 2011; Cooper et al., 2014). Additionally, due to the close relationship of the ventral NAcSh and olfactory tubercle (Ikemoto, 2007), behavioral procedures involving olfaction may also be useful in determining the functional relevance of androgen/mGluR5 interactions.

In summary, these findings indicate a role of androgen signaling in promoting structural plasticity in a region critical in the neurocircuitry of reward. Furthermore, we provide evidence of a mechanism of androgen action involving activation of mGluR5 and endocannabinoid signaling, broadening the understanding of how androgens can exert their effects in the nervous system and supporting activation of mGluRs as a conserved mechanism across multiple steroid hormones.

KG, RM, and PM designed the research. KG and KM performed the research and analyzed the data. KG, RM, and PM wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.