Social bonds form the foundation of our daily lives, and are now widely acknowledged to have far-reaching effects on health and psychological well-being (Uchino, 2006; Cacioppo et al., 2015; Valtorta et al., 2016). While these social relationships may take many forms including friendships, parent-offspring relationships, and status or power relationships, as adults our closest relationship is usually with a romantic partner. Pair bonds are forms of attachment relationships (Hazan and Shaver, 1987); first studied in mothers and offspring (Bowlby, 1969; Ainsworth et al., 1978), these attachment relationships include a strong preference for the familiar partner, distress upon separation from the partner and the ability of the partner to buffer the individual against stress (Mason and Mendoza, 1998).

Titi monkeys (Callicebus cupreus), socially monogamous New World monkeys, form strong adult pair bonds. They show a behavioral preference for their partner over a stranger (Carp et al., 2016); their partner can buffer their stress response to novelty (Hennessy et al., 1995; Mendoza et al., 2000); and they display distress upon separation from their partner, which is not reduced by the presence of another animal (Mendoza and Mason, 1986a,b; Mendoza, 1991; Laugero et al., 2011; Ragen et al., 2012). Distress is displayed both physiologically by increased cortisol concentrations and behaviorally by contact calls for the mate and increased locomotion (Mendoza et al., 2000; Ragen et al., 2012). Separation thus presents a challenge to the pair bond; and while impaired negative feedback appears to keep cortisol concentrations high for an extended period (Mendoza et al., 2000), titi monkeys will also form new pair bonds if their original mate is gone (Van Belle et al., 2016).

The physiological consequences of separation from a pair mate include the increased cortisol concentrations described above; however, changes in other hormones and in neural substrates are less studied. Oxytocin (OT) and vasopressin (AVP) are nine-amino acid peptides synthesized in the paraventricular (PVN) and supraoptic (SON) nuclei of the hypothalamus (Zingg, 2002). They are intimately involved in pair bonding in prairie voles (Microtus ochrogaster), a socially monogamous rodent (Carter, 1998; Young et al., 2005; Gobrogge and Wang, 2015; Johnson et al., 2016; Numan and Young, 2016), and have been implicated in studies of adult human romantic relationships (Gordon et al., 2008; Walum et al., 2008, 2012; Scheele et al., 2012, 2013; Schneiderman et al., 2014). In human studies, OT and AVP are almost exclusively measured in blood (or other peripheral fluids). Plasma OT and AVP may or may not be reflective of brain activity, depending on context and timing of the blood sample (Landgraf and Neumann, 2004; Freeman et al., 2016a); cerebrospinal fluid (CSF) concentrations are generally understood to be more reflective of central nervous system levels (Born et al., 2002; Landgraf and Neumann, 2004).

Cortisol is a metabolic hormone, and it (as well as other metabolic hormones such as glucose and insulin) may change in response to challenging social circumstances (Mendoza, 2017). These challenging social circumstances could include negative events such as loss of close adult female kin due to predation which can lead to elevated circulating cortisol for several weeks (Engh et al., 2006). Social situations we think of as positive may also be challenging metabolically. For instance, male prairie voles lose significant amounts of weight when caring for their first litter, demonstrated by drops in fat depots and circulating leptin (Campbell et al., 2009), as well as plasma insulin (Conley, 2012). In socially monogamous California mice, male body mass decreased when paired with a breeding female, and increased during their pair mate’s pregnancies (Saltzman et al., 2015). Male tamarins and marmosets gain weight during their mates’ pregnancies (Ziegler et al., 2006; Sánchez et al., 2008) and lose weight during periods of infant care (Sánchez et al., 1999). Social dynamics that present psychological and metabolic challenges—the absence of social partners, greater social responsibility in the form of pair bond maintenance or parenting behavior, or the presence of unexpected individuals—can all influence the physiological regulation of the individual.

Involuntary separation from an attachment figure, if it continues, may cause significant distress (Sun et al., 2014). In male prairie voles, chronic separation from the pair mate resulted in increased plasma corticosterone and increased OT, AVP and corticotrophin-releasing hormone (CRH) in the PVN of the hypothalamus; however, plasma OT and AVP did not change (Bosch et al., 2009; McNeal et al., 2014; Sun et al., 2014). This differs from human studies in which challenges to the pair bond or grief at the loss of a loved one, appear to result in sex-specific changes in plasma OT and AVP (Taylor et al., 2000, 2010). Human studies on the neural correlates of grief have implicated the cingulate cortex, the thalamus and the brainstem (Gundel et al., 2003; Kersting et al., 2009), as well as reward areas such as the nucleus accumbens (NAcc; O’Connor et al., 2008).

In this study, we examined the neural and hormonal correlates of both acute and chronic separation from the pair mate in adult male titi monkeys. Our overall hypothesis was that pair bonded males would show short and long term adaptions to separation in neural and hormonal systems engaged in emotion, stress, and motivation; and that a reunion with the familiar pair mate would have different significance than an encounter with a new female. We included a number of neural areas based on human and animal studies (see “Materials and Methods” Section). We predicted that males would display increased plasma cortisol and plasma AVP in both long and short separations, as well as activation of areas associated with “social pain” such as the cingulate cortex (Eisenberger, 2015). We predicted the involvement of reward pathways, such as the NAcc and ventral pallidum (VP), both because they have been implicated in human and vole studies of grief (O’Connor et al., 2008; Bosch et al., 2016), and because they are responsive to aversive as well as appetitive circumstances (Roitman et al., 2005; Saga et al., 2016; Soares-Cunha et al., 2016). We examined areas that either produce or have receptors for OT and AVP in this species [PVN; SON of the hypothalamus; lateral septum (LS)]. We also examined the correlates of reunion with the partner vs. a female stranger, and predicted that we would see responses to the mate but not the stranger, especially in PVN, SON and LS. Finally, we also hypothesized that parenting in a monogamous primate can alter the salience of an unfamiliar female as a potential new mate. Based on evidence from prairie voles that pair bonding may be facilitated by reproduction (Resendez et al., 2016), we compared responses to reunion as a function of whether or not the pair had produced offspring together prior to the separation, predicting increased neural uptake in fathers in the PVN, SON and LS.

Male titi monkeys showed widespread effects of separation and reunion on brain FDG uptake and central and peripheral hormones. In many cases, the neural changes appear to be potentially attributable to the reactions of the OT and AVP systems, or to opioids acting on κ opioid receptors. Observing the patterns of change in FDG uptake and hormone concentrations aids our understanding of these challenges to the pair bond (summarized in Figure 7). Notably, in many of our models, the order in which we performed our PET scans was statistically significant. This argues for continued attention to careful counter-balancing of conditions whenever possible, as well as inclusion of order effects in statistical models.

Short-term separation from the pair mate resulted in a multitude of neural and hormonal changes (Figure 8). FDG uptake was lower than baseline in many regions of interest, particularly those in which OT and AVP are synthesized (PVN) or which have OT or AVP receptors (LS, PAG, Cerebellum) in the titi monkey brain (Ragen and Bales, 2013; Freeman et al., 2014). This is somewhat counter-intuitive, as we might expect increases in some of these areas. However, reductions in PVN uptake during short separations may have been associated with the production of CRH, activation of the hypothalamo-pituitary-adrenal axis and the subsequent release of cortisol, which was elevated during short separations. Cortisol may have then exerted negative feedback on CRH production, leading to the depression in glucose uptake in the PVN. Plasma cortisol and insulin, which were also elevated, can be associated with both metabolic arousal/readiness and psychological stress (Gold, 2015; Mendoza, 2017).

The LS also showed a decrease in FDG uptake during short-term separation. The LS in titi monkeys is a nexus for many neurotransmitter systems; it has receptors for OT (Freeman et al., 2014), both μ and κ opioid receptors (Ragen et al., 2015a), and most likely dopamine receptors as well (Sheehan et al., 2004). It also exerts an inhibitory influence on the NAcc via GABAergic afferents (Sheehan et al., 2004); whereas the NAcc exerts an inhibitory influence on the VP (Numan and Young, 2016). One possibility is that decreased FDG uptake in the LS resulted in lowered inhibition in the NAcc, which then resulted in increased inhibition in the VP (and lower FDG uptake there). One problem with this hypothesis is that in this study, NAcc FDG uptake did not change during short separations; it is possible that our timeline was not optimal for detecting this change.

As stated above, there were no changes in FDG uptake in the dopaminergic NAcc, which has AVPR1a receptors in titi monkeys, and which has been implicated in human and vole studies of grief (O’Connor et al., 2008; Bosch et al., 2016). Short-term separation, however, did result in a reduction in FDG uptake in the closely related, also dopamine-innervated VP, which does not have either OXTR or AVPR1a in titis (Freeman et al., 2014). Neither do titi monkeys have μ opioid receptors in VP; however, they do have κ opioid receptors (Ragen et al., 2015a). κ opioid receptors are associated with negative affect and the dysphoria of social separation in vole and titi monkeys (Resendez et al., 2012; Resendez and Aragona, 2013; Ragen et al., 2015b). κ opioid receptors are also present in titi monkey LS (Ragen et al., 2015a), although they have not been examined for titi monkey midbrain or cerebellum. These commonalities suggest a role for dopamine and κ opioid receptors in the response to short-term separation between pair mates in titi monkeys. Future studies should include measurement of opioid peptides (such as dynorphin), in CSF.

Finally, CSF OT concentrations were also elevated with short-term separation. Prairie voles that undergo a social challenge (isolation) have also shown central elevations in OT (Grippo et al., 2007; Sun et al., 2014), and humans that are in a distressed pair relationship showed increased plasma OT (Taylor et al., 2010). Taken together with the results of neural glucose uptake, we infer that male titi monkeys undergoing an acute separation from their pair mate are releasing OT, which then is binding to OTR in the LS, and AVPR1a in the PAG and cerebellum, since OT and AVP can bind to each other’s receptors (Barberis and Tribollet, 1996). OT in the PAG is known to lead to the release of opioids in response to pain (Ge et al., 2002; Yang et al., 2011); perhaps the response is similar for “social pain”. This release of OT may be a coping strategy to deal with the pain of separation while increasing the motivation to interact socially to locate the mate, or perhaps find a new mate.

Long-term separation resulted in fewer changes than short-term separations, in FDG uptake or hormone concentrations, when compared to baselines. CSF OT, as well as plasma insulin, remained elevated; again probably associated with motivation to travel and engage in social interactions. In addition, whole brain FDG uptake was lower. This represents the reverse of a shift that occurs in whole brain FDG uptake at the formation of the pair bond, when they increase significantly and remain high (Bales et al., 2007; Maninger et al., unpublished data). We have found that this increase occurs between 2–7 days post-pairing (Maninger et al., unpublished data). The decrease in whole brain FDG uptake in this study occurs along a similar timeline, 2–14 days post-pairing, indicating a similar timeline for whole brain metabolism to adjust to a drastic change in social circumstance. Finally, with long separations we see a reduction in glucose uptake in the CeA, which has dense AVPR1a binding in titi monkeys and has been associated with passive stress-coping in other species (Ebner et al., 2005; Freeman et al., 2014).

In our time*condition models, reunion with the pair mate involved a decrease in glucose uptake in the MeA, CeA, SON and PVN when compared to baseline. The MeA typically inhibits other areas involved in social behavior, so a reduction in uptake there may indicate that these other areas could escape inhibition. The CeA did not really change when compared to the long-term separation level, so this is probably not a specific reaction to the pair-mate. Changes in SON and PVN were mediated by fatherhood (see below). Males also experienced a reduction in plasma AVP, an increase in CSF and plasma OT and an increase in insulin. Reunion with the pair mate was our only condition to show a coordinated increase in both CSF and plasma OT. Plasma OT is probably most synchronized with CSF OT in the case of a potent social stimulus (Landgraf and Neumann, 2004; Feldman et al., 2011; Kenkel et al., 2012). Reductions in plasma AVP were mostly present in fathers (Figure 6).

Fatherhood mediated the neural and hormonal effects observed following reunion with the pair mate and exposure to a female stranger. Upon reunion with the pair mate, fathers experienced a reduction in FDG uptake in the PCC when compared to non-fathers. Fathers also displayed higher cortisol but lower plasma AVP and plasma glucose, than did non-fathers. In a recent study in which human males chose a dance partner vs. a non-social choice situation, the PCC was specifically activated by partner choice (Yokoyama et al., 2016). This may suggest that non-fathers are more open to a new partner than fathers. Prairie vole males whose mates do not become pregnant in a short time-frame show less aggression towards novel females (Resendez et al., 2016).

The possibility that pair bonds are weaker in males that have not yet reproduced with their mates was also supported by neural differences during the stranger encounter; when encountering a stranger female, fathers displayed lower FDG uptake in the SON and PVN. Non-fathers that have been separated from their mate for an extended period are possibly initiating the formation of a new pair bond with the stranger female. In other words, since non-fathers may not be as bonded to their absent pair mate, when they are exposed to a stranger female there is greater activity in the SON and PVN, which might result in a release of neuropeptides to encourage the formation of a bond with the new female. While we have studied the neurobiology and physiology of pairing in male titi monkeys (Bales et al., 2007), we have not yet studied the changes that occur within minutes or hours of pair formation or the initiation of parenting—which may prove to be a particularly sensitive period for neurobiological cascades and subsequent regulation. The observed mediating effect of fatherhood on the neuroenergetics and neurobiological cascades within this primate taxon contributes to the growing body of literature around the biology of fatherhood (Fernandez-Duque et al., 2009; Bales et al., 2011; Gettler, 2014; Saltzman and Ziegler, 2014; Bales and Saltzman, 2016).

Overall, our results suggest that mechanisms for maintenance of the pair bond in socially monogamous titi monkeys may rely on similar systems as they do in prairie voles. Regions of the mesolimbocortical system, such as the VP, areas of the extended amygdala that facilitate social memory, such as the LS and areas that produce OT and AVP (PVN and SON) seem to be of particular importance for the maintenance of pair bonds in both the rodent and primate species; avoidance of separation being one behavior that leads to sustained bonding. The specifics, however, probably differ according to species, as do distributions of the receptors for OT and AVP (Freeman and Young, 2016). The regional similarities but mechanistic differences are expected in the context of convergent evolution by which selection acts on similar structures and processes in ways unique within particular lineages. Studies in humans should take into account emerging information regarding the distribution of these receptors in humans (Freeman et al., 2016b). Future studies should also use neurochemically specific PET tracers, where available, to more closely identify the specific peptide and monoamine systems involved in various aspects of attachment. Finally, the multitude of neural and hormonal changes associated with separation and reunion with the pair mate argue for the salience and impact of these manipulations as social stressors.

The research was designed by KLB, SPM, WAM, KH and NM. Research was carried out by KH, NM, BJR, RHL, MRJ, SPM, MLF-P and KLB. SRC was responsible for imaging data. CM and EF performed data analysis. KLB, NM and KH wrote the manuscript, which was edited by all authors.

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.