Circuit-specific remodeling of synaptic co-transmission in stress-responsive networks Authors: Hoyong Park, Hyogyun Kim, Minsung Sim, Doyeon Ham, ChiHye Chung Published: 2026 ABSTRACT ABSTRACTNeurons do not always use a ‘one neurotransmitter-one receptor’ paradigm to communicate at synapses. Some neurons communicate using more than one neurotransmitter, a phenomenon termed co-transmission. Co-transmission can occur through a few distinct organizational modes at synapses, enabling the integration of excitatory, inhibitory, and modulatory signals in various combinations. Increasingly, evidence suggests that these organizational modes are particularly important in neural circuits functioning in stress response. In monoaminergic nuclei, the lateral habenula, and the hippocampus, stress exposure induces shifts in co-transmission via circuit-specific changes to neurotransmitter balance. Here, we review recent advances in our understanding of co-transmission in these circuits, with an emphasis on evidence that distinct modes of small-molecule co-transmission are employed in circuits that contribute to some of the multidimensional aspects of stress. We further discuss potential mechanisms that may recalibrate multi-transmitter signaling during stress-stimulated co-transmission. Currently, evidence suggests that stress affects not only single-transmitter systems but also multi-transmitter communication across neural circuits. Viewing co-transmission as a dynamic substrate of stress-induced neural plasticity, this review provides new insight into the neural basis of adaptive and maladaptive stress responses and may help identify novel therapeutic opportunities for affective disorders, such as major depressive disorder and bipolar disorder.KEYWORDS: Co-transmission, stress, neural circuits, monoaminergic systems, glutamate-GABA co-release Abbreviation basal amygdala basolateral amygdala dopamine transporter dentate gyrus dorsal raphe nucleus entopeduncular nucleus gamma-aminobutyric acid glutamate decarboxylase glutamate decarboxylase 1/2 GABA transporter locus coeruleus lateral habenula median raphe nucleus nucleus accumbens noradrenaline transporter orbitofrontal cortex prefrontal cortex serotonin transporter substantia nigra substantia nigra pars compacta soluble N-ethylmaleimide-sensitive factor attachment protein receptor supramammillary nucleus tyrosine hydroxylase tryptophan hydroxylase 2 vesicular GABA transporter vesicular glutamate transporter vesicular glutamate transporter 2 vesicular glutamate transporter 3 vesicular inhibitory amino acid transporter vesicular monoamine transporter 2 ventral tegmental area Introduction Co-transmission has emerged as a physiologically relevant mode of synaptic communication in which a single neuron can signal through more than one neurotransmitter (Trudeau 2004; Hnasko and Edwards 2012; Nusbaum et al. 2017; Svensson et al. 2019; Ceballos et al. 2024). Unlike the canonical model of single-transmitter signaling, co-transmission encompasses diverse molecular architectures, release modes, and receptor targets (El Mestikawy et al. 2011; Vaaga et al. 2014; Svensson et al. 2019). Rather than representing an exception to classical synaptic organization, co-transmission is recognized as a fundamental feature of neural communication. This diversity enables circuit-specific modes of communication that extend beyond simple excitatory or inhibitory signaling. By combining fast synaptic transmission with slower modulatory components, co-transmission expands the computational repertoire of neural circuits. This expansion is particularly relevant to the study of stress. Stress research has long examined monoaminergic modulatory synapses and forebrain circuits that mediate stress-induced changes in emotional, cognitive, mnemonic, and behavioral responses. Monoaminergic nuclei have received attention since they are closely associated with the pathophysiology and pharmacological treatment of affective disorders, including depression and anxiety disorders (Chaouloff et al. 1999; Pani et al. 2000; Cabib and Puglisi-Allegra 2012). The lateral habenula (LHb) integrates aversive and stress-related information and suppresses the activity of monoaminergic centers, including dopaminergic and serotonergic nuclei (Hikosaka 2010). Indeed, LHb hyperactivity has been repeatedly reported in animal models of depression (B. Li et al. 2011; K. Li et al. 2013; Park et al. 2017; Ryu et al. 2023; Park et al. 2024, 2025), and this finding has been proposed as an important mechanism explaining how monoaminergic output becomes suppressed or imbalanced under stress conditions (Dolzani et al. 2016; Browne et al. 2018). In parallel, studies of the hippocampus, prefrontal cortex (PFC), and amygdala have shown that stress alters synaptic activity in these brain areas to induce emotional and behavioral changes. Chronic stress induces dendritic atrophy and synaptic remodeling in subsets of neurons in the hippocampus and PFC, whereas some amygdala circuits undergo structural changes in a different direction (McEwen et al. 2015; McEwen 2017). It is probably not a coincidence that co-transmission has been prominently documented in these brain areas (Chaouloff et al. 1999; Pani et al. 2000; Cabib and Puglisi-Allegra 2012; Chaudhury et al. 2013; Perez-Caballero et al. 2019). Co-transmission refers to the use of more than one neurotransmitter by a single neuron or neural circuit, making its functional output difficult to explain based on the action of a single transmitter alone (El Mestikawy et al. 2011; Vaaga et al. 2014; Svensson et al. 2019). For example, some neurons in the ventral tegmental area (VTA) and entopeduncular nucleus (EPN) co-release the excitatory (glutamate) and inhibitory neurotransmitter (GABA) onto the LHb (Tritsch et al. 2016; Ceballos et al. 2024). In this case, without measuring the relative contribution of each transmitter and identifying the stimulus conditions under which their release is preferentially recruited, it is challenging to predict whether the net circuit output will be excitatory or inhibitory (Vaaga et al. 2014; Silm et al. 2019). Co-transmission also has important implications in monoaminergic circuits. In these circuits, monoamines such as serotonin, dopamine, and noradrenaline may act in conjunction with glutamate or GABA, enabling both fast synaptic transmission and slow neuromodulation to occur within a single circuit response (El Mestikawy et al. 2011; Sengupta et al. 2017). Therefore, activity changes or physiological responses after stimulation should be interpreted as composite outputs formed by multiple transmitters, rather than as effects attributable to one transmitter type (Trudeau 2004; Tritsch et al. 2012; Qi et al. 2014; Sengupta et al. 2017; Zych and Ford 2022). Stress-induced circuit imbalance should also be examined by identifying which transmitter combination is recruited and which signal dominates the output under a given stress condition. Neurons use a wide range of signaling molecules, from classical neurotransmitters, including gaseous messengers to neuropeptides. However, stress changes the activity level and firing patterns of neural circuits, thus synaptic vesicle-mediated co-transmission deserves separate attention as a form of co-transmission that is directly shaped by neural activity. Gaseous messengers and neuropeptides differ from small-molecule neurotransmitters in their synthesis, storage, release conditions, and diffusion properties. In this review, we focus on co-transmission of small-molecule neurotransmitters that are loaded into synaptic vesicles (Tritsch et al. 2012, 2016; Root et al. 2018) and released through vesicular exocytosis by neural activity (Salio et al. 2006; Nusbaum et al. 2017; Hökfelt et al. 2018). We summarize the current studies of small-molecule neurotransmitter co-transmission in stress-related neural circuits and discuss changes in transmitter recruitment and synaptic output upon stress exposure. Principles of synaptic co-transmission Synaptic vesicle-mediated co-transmission can be organized in three broad modes: (1) co-packaging within the same vesicle, (2) segregation into distinct vesicle pools within a terminal, and (3) spatial separation across boutons or axonal branches (El Mestikawy et al. 2011; Vaaga et al. 2014; Tritsch et al. 2016; Nusbaum et al. 2017). In the first mode, different neurotransmitters are loaded into the same synaptic vesicle and released together during stimulation. In this case, the synaptic vesicle membrane contains transporters corresponding to each neurotransmitter (Gasnier 2000; El Mestikawy et al. 2011). This mode has been mainly identified in glutamate-GABA co-transmission, where the vesicular glutamate transporter (VGLUT) and the vesicular inhibitory amino acid transporter/vesicular GABA transporter (VIAAT/VGAT) coexist within the same synaptic vesicle. Using immunoisolation of synaptic vesicles, substantial VGAT vesicle population was detected together with VGLUT2. Additionally, post embedding double-immunogold electron microscopy further revealed the co-expression of VGLUT1, VGLUT2, and VGAT within individual synaptic vesicles in the adult rat hippocampal CA3 region and cerebellum (Zander et al. 2010). In rat ventral striatum, VGLUT2 was detected in the vesicular monoamine transporter (VMAT) 2 synaptic vesicles and VMAT2 was detected in VGLUT2 vesicles, implying their colocalization in single synaptic vesicle of dopaminergic neurons (Hnasko et al. 2010). Some studies have further reported that VMAT uses extracellularly taken-up GABA for vesicular packaging (Tritsch et al. 2012, 2014). Optogenetically evoked GABAergic transmission from dopaminergic axons was blocked following conditional deletion or pharmacological inhibition of VMAT2, indicating that vesicular GABA release depends on this transporter (Tritsch et al. 2012). Interestingly, these neurons lack detectable GAD65 and GAD67 and instead acquire GABA from the extracellular space through plasma membrane GABA transporters for vesicular release (Tritsch et al. 2014). Although GABA and monoamines have not yet been directly visualized within the same vesicle, shared use of VMAT provides a plausible molecular route for monoamine-GABA co-packaging. In striatal GABAergic autapses expressing VGLUT3, Mixed AMPAR- and GABAR-mediated miniature currents and Sr²+-evoked asynchronous quantal events showed that glutamate and GABA were co-released through individual vesicular fusion events (Zimmermann et al. 2015). More directly, repeated minimal stimulation of individual EPN somatostatin expressing boutons exhibited coincident successes and failures of glutamatergic and GABAergic synaptic responses, together with a strong trial-by-trial correlation between their amplitudes (S. A. Kim et al. 2022). Together, these findings indicate that, The co-packaged transmitters leave the terminal together, share release probability and timing, and respond to presynaptic calcium entry and release machinery as a coupled signal. The resulting synaptic output depends strongly on the chemical ratio of transmitters contained within each vesicle (El Mestikawy et al. 2011; Vaaga et al. 2014). Co-packaging also changes transmitter storage itself, as interactions between vesicular transporters may increase vesicular filling efficiency (Hnasko et al. 2010; El Mestikawy et al. 2011). Vesicular transporters use an electrochemical proton gradient to package neurotransmitters into synaptic vesicles (Gasnier 2000). Interactions between different vesicular transporters increase the vesicular ΔpH and increase transmitter packaging efficiency, producing vesicular filling synergy (Hnasko et al. 2010). Specifically, glutamate enhanced monoamine uptake by VMAT2 in synaptic vesicles from the ventral striatum. However, vesicles lacking VGLUT2 did not show this effect. Consistent with this functional interaction, selective deletion of VGLUT2 from dopaminergic neurons reduced the amount of dopamine stored in synaptic vesicles, as well as stimulus-evoked dopamine release. Under some conditions, different transporters may also interfere with one another and limit the rate or amount of vesicular packaging (El Mestikawy et al. 2011). The second and third modes are organized by packaging different neurotransmitters into separate vesicles (Vaaga et al. 2014; Tritsch et al. 2016; Nusbaum et al. 2017). In the second mode, distinct synaptic vesicles containing different neurotransmitters coexist within a single release system, such as a single synaptic bouton. In this case, the two types of synaptic vesicles are segregated into separate pools (Silm et al. 2019; Zych and Ford 2022). dopamine and GABA release from nigrostriatal axons differed in paired-pulse ratio and extracellular Ca²+ sensitivity. Moreover, both of releases are modulated by different mechanisms by presynaptic D2 or GABAB receptors, respectively (Zych and Ford 2022). Since synaptic vesicle release depends on local calcium concentration, it is influenced by vesicle–calcium channel distance, channel conductance, and vesicular calcium sensor sensitivity (Takács et al. 2018; Zych and Ford 2022; Hirai et al. 2024). GABAergic and cholinergic transmission from septohippocampal terminals preferentially depended on P/Q-type and N-type calcium channels, respectively (Takács et al. 2018). Dopamine and glutamate release from midbrain dopaminergic axons also exhibited in their distinct sensitivity to calcium channel blockers, indicating preferential coupling of dopamine release to P/Q-type channels and glutamate release to N-type channels (Silm et al. 2019). In addition, Glutamatergic and GABAergic transmission at supramammillary nucleus (SuM) → dentate gyrus (DG) synapses also differed in extracellular Ca²+ sensitivity and sensitivity to EGTA-AM, indicating different spatial coupling between their respective vesicle pools and Ca²+ channels and the two transmitters to be recruited under different conditions. The third mode separates transmitters at the level of boutons or axonal branches. In monoaminergic circuits with branched axons, different release sites may use different transmitter outputs depending on the projection target (El Mestikawy et al. 2011; Soiza-Reilly and Commons 2011; Vaaga et al. 2014; Zhang et al. 2015; Tritsch et al. 2016; Voisin et al. 2016). Immunolabeling revealed that VGLUT3 and SERT were concentrated in largely distinct varicosities of serotonergic axons (Voisin et al. 2016). Similarly, dopaminergic axons formed largely segregated dopamine- and glutamate-associated release sites (Fortin et al. 2019). Postsynaptic receptors are also positioned near the transmitter released from each bouton, which gives this mode a spatially organized postsynaptic counterpart. The mechanisms that direct each neurotransmitter to a specific release site remain unclear. Still, this arrangement highlights a key feature of segregated co-transmission: when transmitters occupy separate vesicle pools or distinct boutons, neuronal firing patterns and presynaptic release machinery shape how strongly each signal is recruited (Vaaga et al. 2014; Tritsch et al. 2016; Nusbaum et al. 2017). This contrasts with co-packaging, where transmitters are tied to the same vesicle fusion event. When co-transmission relies on separate vesicle pools or distinct boutons, transmitter recruitment is no longer fixed to a single vesicle fusion event. Instead, the neural activity pattern becomes an important determinant of which transmitter is preferentially released. The contrast between large dense-core vesicles and small synaptic vesicles illustrates this general principle: small synaptic vesicles are clustered near presynaptic active zones and undergo rapid exocytosis in response to brief stimulation or low-frequency activity. Large dense-core vesicles are commonly located farther from active zones and generally require sustained or high-frequency activity that produces a substantial increase in intracellular Ca²+ (Walch-Solimena et al. 1993; Salio et al. 2006; Nusbaum et al. 2017). Similar activity-dependent differences have been observed in co-transmitting circuits. For example, a study of serotonergic inputs to the amygdala found that increasing the stimulation frequency suppressed glutamate release while enhancing 5-HT transmission (Sengupta et al. 2017). Thus, co-transmitting circuits may also use firing patterns as a way to shift transmitter weighting. Monoaminergic co-transmission in stress-responsive circuits Monoaminergic circuits were traditionally viewed as diffuse modulatory systems. A relatively small number of neurons gives rise to widely branched axons that innervate broad cortical and limbic areas (Saper 1987; Gil et al. 1997; Avery and Krichmar 2017). Serotonergic signaling arises mainly from the dorsal raphe nucleus (DRN) and median raphe nucleus (MRN), dopaminergic signaling from the VTA and substantia nigra (SN), and noradrenergic signaling from the locus coeruleus (LC). For this reason, monoamines have long been described as signals that diffuse through volume transmission and regulate broad excitability states across the brain (Özçete et al. 2024). Recent studies extend this view by showing that monoaminergic systems also generate pathway-defined signals (Figure 1). Some monoaminergic neuronal subpopulations use additional small-molecule transmitters, including glutamate and GABA, through shared vesicular mechanisms or segregated vesicle pools (Hnasko and Edwards 2012; Vaaga et al. 2014; Tritsch et al. 2016). These co-transmission mechanisms shift the focus from total monoamine output to transmitter composition within defined projection pathways. Projection-specific monoaminergic co-transmission and LC-derived catecholamine coupling(A) DRN serotonergic projections to the OFC/PFC, basal amygdala, and VTA contain serotonin–glutamate co-transmission and are associated with stress coping, anxiety-related regulation, and motivational control. (B) VTA dopaminergic projections to the NAc and PFC contain projection-dependent glutamatergic or GABAergic signals that shape reinforcement, motivational valence, cognitive control, and aversive processing. (C) LC projections to the PFC, hippocampus, and BLA show dopamine–noradrenaline coupling through shared catecholamine biosynthesis and NET-mediated transporter regulation. Serotonergic circuits Serotonergic neurons are concentrated in the DRN and MRN and are commonly marked by Pet1/FEV, tryptophan hydroxylase 2 (Tph2), and the serotonin transporter (SERT). Although it has long been thought that they are a relatively homogeneous population, single-cell transcriptomic studies have divided these neurons into multiple subtypes, including populations that co-express VGLUT3 (Fremeau et al. 2002; Gras et al. 2002; Ren et al. 2019; Okaty et al. 2020). SERT and VGLUT3 are co-expressed in subpopulations of the SN, VTA, and raphe nuclei, but not in the cortex, hippocampus, or brainstem (Fremeau et al. 2002; Gras et al. 2002). Interestingly, VMAT2 and VGLUT3 co-expressed in the projection areas such are the hippocampus (Amilhon et al. 2010). Deletion of VGLUT3 not modify serotonin (5-HT) outflow in the raphe nucleus, however it increases 5-HT release at its target area, hippocampus (Amilhon et al. 2010). Single-cell transcriptomics studies further revealed that other serotonergic neurons in the raphe nucleus express glutamate decarboxylase 1/2 (GAD1/2), the enzymes responsible for GABA synthesis, which raises the possibility of serotonin-GABA co-transmission (Ren et al. 2019; Okaty et al. 2020), however, the functional evidence for serotonin-GABA co-transmission is limited. Serotonin-glutamate co-transmission involves not only the co-release of two neurotransmitters at the same synapses but also functional interactions between release of each neurotransmitter. In a subset of serotonergic terminals in the hippocampus and cortex, VGLUT3 co-localizes with VMAT2. Deletion of VGLUT3 does not alter serotonin (5-HT) outflow in the raphe nucleus; however, it increases 5-HT release in target areas such as the hippocampus (Amilhon et al. 2010). Consistent with this, hippocampal and cortical synaptic vesicles accumulate more 5-HT in the presence of glutamate, whereas this is not the case in the striatum, where most of VMAT2 is confined to VGLUT3-negative dopaminergic terminals (Amilhon et al. 2010). Loss of this synergy under VGLUT3-deficient conditions further supports the idea that VGLUT3-mediated glutamate uptake enhances vesicular serotonin loading and release through vesicular filling synergy (Amilhon et al. 2010). Beyond vesicular interactions, the spatial organization of serotonin and glutamate signaling also differs depending on the level of synaptic organization being examined. At the ultrastructural level, SERT and VGLUT3 occupy segregated domains within a substantial fraction of serotonergic axon terminals. In contrast, at the level of axonal varicosity, VGLUT3 and SERT show considerable co-localization across several mouse forebrain regions (Voisin et al. 2016; Belmer et al. 2019). Together, these results suggest that both terminal microstructure and projection target shape how serotonin and glutamate are arranged and potentially recruited during synaptic transmission. DRN-derived serotonin-glutamate co-transmission is engaged across multiple projection targets, where it exerts pathway-specific effects on neural activity and behaviors. In cortical projections, the DRN–orbitofrontal cortex/prefrontal cortex (OFC/PFC) pathway contains a substantial population of VGLUT3 expressing serotonergic neurons and provides fast glutamatergic signaling in addition to slower serotonergic modulation (Ren et al. 2018). Conditional deletion of Tph2 from DRN SERT-positive neurons projecting to the OFC increased anxiety-related behavior and immobility in the forced swim test, whereas activation of this pathway increased escape behavior (Ren et al. 2018). The DRN→ basal amygdala (BA) pathway is organized differently from cortical projections. The BA, a key region involved in anxiety and defensive behavior, receives DRN-derived inputs that co-release serotonin and glutamate. In this pathway, transmitter recruitment depends on neuronal activity. Low-frequency stimulation predominantly evokes glutamate-mediated signaling, whereas high-frequency stimulation preferentially recruits serotonergic signaling, and the postsynaptic response further varies according to the target cell type (Sengupta et al. 2017). Serotonin directly suppresses principal cells, whereas glutamate recruits local interneurons to inhibit principal cell activity indirectly. This allows the DRN–BA pathway to regulate circuit output in a manner that depends on both firing frequency and postsynaptic cell identity (Sengupta et al. 2017). The DRN→ VTA pathway represents another projection in which serotonin–glutamate co-transmission contributes to circuit function. DRN inputs form excitatory synapses onto VTA dopamine neurons, increase dopamine release in the nucleus accumbens (NAc), and promote reward-related behavior (Qi et al. 2014). Pharmacological blockade of either glutamate or serotonin receptors attenuates these effects, indicating that both transmitters contribute to the functional output of this pathway (Wang et al. 2019). These pathways show that the same neurotransmitter pair can be recruited in different ways depending on the projections target. In cortical circuits, serotonin–glutamate co-transmission supports rapid modulation of cortical activity (Ren et al. 2018), whereas in the BA and VTA it contributes to anxiety-related behaviors and motivational processing through distinct synaptic mechanisms (Qi et al. 2014; Sengupta et al. 2017) (Figure 1A). Most studies to date, however, have focused on establishing the presence and functional significance of co-transmission under basal conditions. Whether stress alters the relative recruitment of serotonin and glutamate remains largely unknown. Dopaminergic circuits Dopaminergic signaling is mediated primarily by neurons in the SN and VTA. Although the nigrostriatal pathway has classically been associated with motor control, accumulating evidence also links it to but recent studies also connect this circuit to behavioral and cognitive adaptations (B. Kim et al. 2024; T. Y. Kim and Lee 2025). By contrast, VTA dopaminergic neurons form mesocorticolimbic projections, that regulate motivation, affective state, and stress adaptation (Nestler and Carlezon 2006; Cao et al. 2010; Mohebi et al. 2019). Not all dopaminergic neurons rely exclusively on dopamine signaling, as some dopaminergic neurons have been shown to co-release glutamate or GABA together with dopamine (Root et al. 2014; Yoo et al. 2016; Root et al. 2020). In the VTA, a subset of dopaminergic neurons is commonly identified by tyrosine hydroxylase (TH) or the dopamine transporter (DAT) and also express VGLUT2, providing molecular basis for dopamine-glutamate co-transmission. At the vesicular level, VGLUT2-dependent glutamate uptake enhances VMAT2-mediated dopamine storage through vesicular filling synergy by promoting vesicular acidification and increasing quantal dopamine release (Hnasko et al. 2010; Aguilar et al. 2017). The spatial organization of dopamine–glutamate co-transmission is also heterogeneous. High-resolution anatomical studies show that dopaminergic axons contain spatially distinct release sites for glutamatergic and dopaminergic signals, indicating that the two transmitters did not originate from the same vesicle or even the same release site (Zhang et al. 2015). A separate subset of dopaminergic neurons has been shown to co-release GABA through a non-canonical mechanism (Yamaguchi et al. 2011; Phillips et al. 2022). Through this mechanism, extracellular GABA is taken up into dopaminergic terminals and then loaded into vesicles by VMAT2 (Tritsch et al. 2012, 2014; Melani and Tritsch 2022). Dopaminergic axon terminals in both the VTA and SN express GABA transporter (GAT), and GAT inhibition increases extracellular GABA levels, supporting the presence of a GABA uptake mechanism in dopaminergic terminals of the VTA and SN (Zandy and Gonzales 2018). These findings suggest that VTA dopaminergic neurons can employ a mechanism similar to that described in the SN, in which extracellular GABA is taken up and packaged into synaptic vesicles via VMAT2. Dopaminergic co-transmission is especially evident in pathways to the NAc, striatum, and PFC, where transmitter composition and release properties differ across targets. The VTA→ NAc pathway provides one of the best-characterized examples of dopamine–glutamate co-transmission. A subset of VTA neurons projecting to the NAc releases both transmitters (Tecuapetla et al. 2010; Warlow et al. 2024), allowing fast glutamatergic excitation to operate alongside dopaminergic modulation. Optogenetic manipulations combined with conditional knockout approaches have revealed that glutamate primarily mediates rapid excitatory transmission underlying positive reinforcement, whereas dopamine contributes more prominently to value processing during aversive behaviors (Zell et al. 2020; Warlow et al. 2024). The SN→striatum pathway illustrates a different transmitter combination. Here, dopaminergic output is coupled to a GABAergic output element (Tritsch et al. 2012, 2014; Melani and Tritsch 2022). Despite their shared dependence on VMAT, the two transmitters are subject to distinct presynaptic regulation, allowing their modulatory and inhibitory signals to be partially separated during co-transmission (Zych and Ford 2022). Co-released GABA also acts on axonal GABAA receptors to rapidly suppress dopamine release, forming a fast autoregulatory feedback loop (Patel et al. 2024). In this pathway, GABA may provide an additional inhibitory component that complements dopaminergic modulation of striatal output. A distinct functional profile emerges in the VTA→PFC pathway. Here, glutamate is expected to mediate rapid synaptic transmission, whereas dopamine provides slower neuromodulatory signaling. Recent projection-resolved studies suggest that VTA-derived glutamatergic output to the PFC contributes to cognitive control and pain-related behavior (M. Li and Yang 2024). So far, these studies indicate that the functional consequences of dopaminergic co-transmission depend strongly on projection target, transmitter combination, and local synaptic organization (Figure 1B). Although the molecular mechanisms have become increasingly well defined, how stress reshapes transmitter recruitment within these pathways remains largely unexplored. Noradrenergic circuits: LC-derived catecholamine coupling Although the LC has long been regarded as the brain's principal source of diffuse noradrenergic modulation, accumulating evidence supports a more projection-specific organization of LC circuits. For example, LC neurons innervating the PFC are closely associated with aversive and anxiety-related states, whereas spinal cord projecting neurons contribute to functions such as analgesia (Hirschberg et al. 2017). Molecular and physiological heterogeneity among LC neurons also supports this pathway-specific organization (Chandler et al. 2014). Dopamine and noradrenaline are closely linked through a shared catecholamine biosynthetic pathway, with dopamine serving as the immediate precursor of noradrenaline (Nagatsu 2006). Beyond this biochemical relationship, extracellular dynamics of the two catecholamines are also coupled. In LC target regions, norepinephrine transporters (NET) contribute to the clearance of extracellular dopamine as well as noradrenaline, LC-derived signaling may regulate the two transmitters in parallel rather than independently (Devoto et al. 2005). However, direct evidence for dopamine-noradrenaline co-packaging within the same vesicles or for segregation into distinct vesicle pools remains lacking. The LC→PFC pathway provides a representative example of projection-specific noradrenergic signaling. LC neurons projecting to the PFC are preferentially recruited during aversive and anxiety-related states (Chandler et al. 2014; Hirschberg et al. 2017). Under heightened arousal such as threat exposure, LC input sustains PFC neuronal activity in response to threat cues and supports adaptive behavioral responses (Duan et al. 2025). Although these functions have mainly been explained through noradrenaline signaling, the finding that extracellular dopamine in the mPFC is regulated primarily by NET rather than DAT suggests that LC terminals may also contribute to cortical dopamine signaling (Devoto et al. 2004). Thus, catecholamine signaling in this pathway may show a coordinated regulation of dopamine and noradrenaline rather than purely noradrenergic transmission. The LC→hippocampus pathway provides another example in which LC-derived dopamine signaling extends beyond conventional noradrenergic modulation. Hippocampal dopamine signaling is essential for synaptic plasticity and memory formation (Kempadoo et al. 2016). Neuronal tracing study shows that the hippocampus receives denser dopaminergic innervation from the LC than from the midbrain. Consistent with these studies, LC projections to the hippocampus contribute to hippocampal plasticity and spatial memory (Kempadoo et al. 2016). These findings suggest that LC supports both noradrenaline signaling and dopamine-related plasticity within the hippocampus. The LC→basolateral amygdala (BLA) pathway is associated with negative emotional processes, including anxiety-like behavior, aversive learning, and stress-related memory (McCall et al. 2017; Tan et al. 2026). Although these LC projection-mediated processes have largely been explained by β-adrenergic receptor-dependent regulation of BLA activity, the presence of dopaminergic receptors raises the possibility that catecholamine signaling in this circuit may involve coordinated actions of both transmitters (Miyajima et al. 2010). Direct synaptic evidence for LC-derived dopamine release in the BLA, however, remained limited. Collectively, these observations suggest that the LC provides projection-dependent catecholamine signals in the brain rather than a uniform noradrenergic tone (Figure 1C). Whether stress dynamically alters the relative contribution of dopamine and noradrenaline across individual LC projections remains an important question for future studies. Glutamate–GABA co-transmission in the LHb and hippocampus The previous section focused on co-transmission in monoaminergic circuits. Co-transmission has also been identified in non-monoaminergic regions that play important roles in stress processing (Ceballos et al. 2024). Glutamate and GABA mediate the major excitatory and inhibitory synaptic signals in the brain, making their co-transmission directly relevant to the regulation of excitation-inhibition balance at the level of individual synaptic inputs. This is particularly important because disruption of this balance has been associated with multiple forms of brain dysfunction. For example, administering bicuculline via subcutaneous injection during postnatal days 7–11 induces autism like behaviors and E/I imbalance in hippocampus in adult mice (Shin et al. 2025). Glutamate-GABA co-transmission represents an input specific mechanism that determines the overall circuit output. In the following sections, we summarize current findings on glutamate-GABA co-transmission in the LHb and hippocampus, and discuss their implications under basal physiological conditions (Figure 2). Glutamate–GABA co-transmission in the LHb and hippocampus. Lateral habenula The LHb integrates aversive information and negative motivational states by activating their neural activities. This activation inhibits the activity of downstream monoaminergic nuclei, including the VTA and raphe nuclei (Matsumoto and Hikosaka 2007; Hikosaka 2010). Since the LHb contains few local inhibitory interneurons, its excitation-inhibition balance depends strongly on the transmitter output of afferent projections. This organization makes the LHb particularly well suited for examining how glutamate-GABA co-transmission influences circuit function (Figure 2). The best-studied input is the EPN. The EPN →LHb pathway was initially considered as inhibitory, but later studies showed that EPN neurons co-release both glutamate and GABA onto LHb neurons (Shabel et al. 2014; Root et al. 2018). Functionally, this pathway excites LHb neurons, and its excitatory effect is enhanced by treatment with a GABA receptor antagonist. Both VGLUT2 and GAD67 are co-localized at its synapses. Whether VGLUT2 expressing or GABA expressing neurons in the EPN → LHb pathway, both showed glutamate receptor- and GABA receptor-dependent current (Shabel et al. 2014). These results provide a mixed excitatory–inhibitory input whose net effect depends on the balance between glutamatergic and GABAergic transmission. This same-vesicle co-release was subsequently demonstrated at somatostatin-positive EPN inputs to the LHb, making this pathway one of the clearest examples of glutamate–GABA co-release at identified central synapses (S. A. Kim et al. 2022). The VTA→ LHb pathway also provides glutamate-GABA co-transmitting input to the LHb (Root et al. 2020). Although the VTA is often described as a dopaminergic center, it contains neurons that express VGLUT2, neurons that express VGAT, and neurons that express both. A subset of VTA neurons co-expresses VGLUT and VGAT, and strongly projects to the LHb. At these terminals, VGLUT and VGAT are detected together, and optical stimulation evokes both glutamatergic excitatory currents and GABAergic inhibitory currents in LHb neurons. These observations provide direct evidence for glutamate-GABA co-transmission in the VTA→ LHb pathway (Root et al. 2020). VTA glutamate-GABA co-transmitting neural population have unique responsive properties to cued reward experiments compared to other populations in the VTA. Its activity significantly decreases when an expected sucrose is omitted and increases most strongly when an unexpected sucrose is delivered while other populations do not. The pattern suggests that co-transmitting neural population is sensitive to mismatches between expected and actual outcomes, rather than reward prediction cues alone (Root et al. 2020). Thus, in addition to regulating the excitation-inhibition balance of the LHb, the VTA→ LHb pathway conveys information related to reward and aversive outcomes. Glutamate-GABA co-transmission in the LHb makes it difficult to classify afferent inputs as purely excitatory or inhibitory. Instead, the functional output of individual inputs reflects the relative contribution of glutamatergic and GABAergic signaling. In this circuit context, stress-related shifts in LHb output are therefore likely to depend, at least in part, on changes in this transmitter balance. Hippocampus The hippocampus is another key region involved in stress-related brain function, particularly in contextual processing and behavioral adaptation. Glutamate-GABA co-transmission has been identified both in long-range afferents and local hippocampal circuits. A representative long-range input is the SuM → DG pathway. SuM neurons directly innervate DG granule cells and local GABAergic interneurons, and evoke both glutamatergic excitation and GABAergic inhibition (Hashimotodani et al. 2018). SuM input strongly activates local interneurons but not granule cell firing. Instead, it increases granule cell responsiveness when paired with performant path input at a −10 ms interval, indicating that the pathway modulates how incoming information is integrated within the DG rather than simply driving excitation or inhibition (Hashimotodani et al. 2018). Glutamate-GABA co-transmission also occurs within local hippocampal circuits. VGLUT3-expressing interneurons co-release glutamate and GABA onto pyramidal neurons (Fasano et al. 2017; Pelkey et al. 2020). Although inhibitory responses predominate under basal conditions, eIPSC is enhanced when mGluR antagonist treatment in WT mice while VGLUT3 KO mice do not. These indicates that the glutamatergic component contributes to inhibit hippocampal inhibitory signaling through metabotropic glutamate receptor signaling (Fasano et al. 2017). Together, these findings indicate that the both long-range and local co-transmitting circuits contribute to information processing by regulating excitation-inhibition balance in the hippocampus. Another direct evidence linking stress to glutamate-GABA co-transmission in these circuits remains limited, the established roles of the DG in contextual discrimination and stress-related information processing make this an important area for future investigation (Leutgeb et al. 2007; Yassa and Stark 2011; Jeong et al. 2024). Emerging principles of stress-related co-transmission remodeling The studies reviewed above suggest that stress influences co-transmitting circuits through mechanisms beyond simple increase or decrease in neural activity. In circuits with co-transmission, stress may also bias synaptic output toward one transmitter signal over another. Current evidence points to three, not mutually exclusive, levels at which stress may reshape co-transmission (Figure 3). First, stress may alter the molecular machinery that produces neurotransmitters and loads them into vesicles, thereby changing the relative contribution of co-transmitted signals. Second, stress may modify neuronal firing patterns and the activity-dependent recruitment of co-transmitted signals. Third, stress may remodel presynaptic release machinery, changing the release efficiency of each transmitter. Mechanisms of stress-induced co-transmission remodeling (A) Stress alters the expression of synthetic enzymes and vesicular transporters (e.g. VGLUT, VGAT, GAD), thereby changing the composition and relative abundance of co-released neurotransmitters, and biasing synaptic output. (B) Stress-induced changes in neuronal firing patterns influence the selective recruitment of vesicle pools with distinct release properties (release thresholds and calcium sensitivity), allowing dynamic regulation of favoring transmitter output without altering transmitter identity. (C) Stress may reorganize the presynaptic release architecture, including the spatial coupling between synaptic vesicles and voltage-gated Ca²+ channels, enabling fine-tuned control of transmitter release within a shared synaptic terminal. Transmitter identity shifts Stress may change the relative contribution of co-transmitted signals by altering the molecular machinery that controls transmitter availability and vesicular loading. These changes do not necessarily indicate a change in neuronal identity. Rather, they reflect changes in the relative availability of co-transmitted signals within a defined circuit. Neurotransmitter-synthesizing enzymes and vesicular transporters are central to this process as they determine which neurotransmitters are synthesized, packaged, and released (Figure 3). This mechanism has been reported in serotonergic circuits. Following chronic social-defeat stress, immunohistochemistry and in situ hybridization revealed a decrease of VGLUT3-expressing neurons in the ventral DRN of both susceptible and resilient rats, without a reduction in the total number of neurons. Susceptible rats additionally showed higher TPH2 expression and reduced VGLUT3–TPH2 co-expression (Prakash et al. 2020). In another study, strong foot shock reduced VGLUT3 and induced GAD67 expression within serotonergic neurons of the DRN. Mild foot shock that induced conditioned fear did not alter transmitter identity (H. Q. Li et al. 2024). These changes shifted the molecular transmitter profile of serotonergic neurons toward reduced glutamatergic and relatively enhanced GABAergic signaling. In stressed animals, this shift was required for fear generalization, preventing the gain of GAD67 or restoring VGLUT3 expression in serotonergic neurons suppressed generalized fear. In addition, antidepressant treatment suppresses both the VGLUT3-to-GAD67 switch and fear generalization (H. Q. Li et al. 2024). A similar form of remodeling has been described in glutamate-GABA co-transmitting pathways to the LHb. The EPN→ LHb pathway normally co-releases glutamate and GABA, but depression-like or withdrawal-related conditions reduce the GABAergic contribution, shifting the balance toward greater net excitation without increasing glutamatergic transmission (Shabel et al. 2014; Meye et al. 2016). In congenitally helpless rats, optogenetically evoked EPN→ LHb synaptic responses showed a lower GABAAR/AMPAR current ratio than responses recorded in control rats, indicating a selective reduction in the inhibitory component of co-transmission (Shabel et al. 2014). Similarly, cocaine withdrawal weakened GABAergic transmission from EPN terminals while leaving glutamatergic transmission intact. Reduced presynaptic VGAT expression impaired GABA loading into synaptic vesicles. It weakened GABAergic transmission. Restoring VGAT expression in EPN→ LHb terminals rescued GABAergic transmission, alleviated withdrawal-associated aversion, and prevented stress-induced reinstatement of cocaine seeking (Meye et al. 2016). Antidepressant treatment reversed this shift by strengthening GABAergic transmission and restoring glutamate-GABA balance in this pathway (Shabel et al. 2014). These observations suggest that pathological states alter the relative contribution of transmitter signals that already coexist within the circuit rather than establishing a new transmitter phenotype. Given that excessive LHb activity is closely associated with aversive processing and suppression of downstream monoaminergic nuclei, this glutamatergic bias may represent an important axis of stress-related circuit dysfunction. These findings suggest that stress can change the output of co-transmitting circuits by altering the relative contribution of transmitters that are already present, without fundamentally changing neuronal identity. Activity-dependent transmitter recruitment Stress changes the activity patterns of many neural circuits. In VTA-related circuits, stress alters phasic or burst firing of dopamine neurons, and these changes are associated with stress susceptibility and motivational behavior. In serotonergic and hippocampal circuits, stress also modifies neuronal firing and input processing. Co-transmitted signals can differ in their release requirements, thus changes in firing pattern may alter their relative contribution without requiring stable changes in gene expression or cellular identity. The DRN→ BA pathway illustrates this possibility. Uncontrollable stress increases the activity of DRN serotonergic neurons and serotonin release (Grahn et al. 1999; Maier and Watkins 2005). Within the DRN→BA pathway, serotonergic axons are recruited during both fear-memory formation and retrieval, and optogenetic manipulation showed that their activity facilitates fear expression and impairs fear extinction (Sengupta and Holmes 2019). This pathway co-releases both serotonin and glutamate, with the relative contribution of each transmitter varying according to stimulation frequency. Low-frequency stimulation preferentially evokes fast glutamatergic transmission, whereas high-frequency stimulation recruits slow serotonin receptor-mediated modulation (Sengupta et al. 2017). Changes in DRN firing under stress may therefore shift the relative contribution of these two signaling molecules. Stress-dependent changes in neuronal activity may also influence dopamine–glutamate co-transmission in VTA circuits. Repeated social-defeat stress produced a persistent increase in burst firing among VTA dopamine neurons, which remained three weeks after the final stress exposure (Razzoli et al. 2011). Moreover, optogenetically inducing phasic, but not tonic, firing in VTA dopamine neurons promoted social avoidance and reduced sucrose preference in mice exposed to subthreshold defeat and converted previously resilient mice to a susceptible phenotype (Chaudhury et al. 2013). A subset of VTA dopamine neurons release glutamate together with dopamine. Dopamine acts as a relatively slow modulatory signal, whereas glutamate provides fast excitatory synaptic transmission (Hnasko et al. 2010; Zhang et al. 2015; Zell et al. 2020). In the VTA→NAc pathway, stimulation across a broad frequency range evoked fast glutamatergic responses. In contrast, endogenously released dopamine contributed detectably only during 20-Hz burst stimulation. Under this condition, dopamine-receptor blockade weakened the later glutamatergic responses in the train, whereas DAT inhibition enhanced them (Chuhma et al. 2009; Broussard 2012). These findings raise the possibility that stress modifies not only dopamine release but also the contribution of co-released glutamate. Activity-dependent regulation may also occur at the level of vesicle recycling. Synaptic vesicles are continuously retrieved and reused after release, and this process influences transmitter content and release properties. In midbrain dopamine neurons, dopamine- and glutamate-containing vesicles follow distinct recycling routes. AP-3 supports the formation of dopamine-containing vesicles, whereas glutamatergic vesicles are recycled through an AP-3-independent pathway. This finding suggest that neuronal activity may differentially recruit and replenish dopamine- and glutamate-containing vesicle populations (Silm et al. 2019). Repeated firing during stress may thus bias subsequent transmitter release toward one vesicle population over another. Although this hypothesis is supported by physiological studies, direct evidence under stress conditions remain unclear. Determining whether stress-dependent changes in neuronal activity alter transmitter recruitment within co-transmitting circuits will be an important direction for future studies. Presynaptic release-site regulation Presynaptic release-site regulation represents another mechanism through which stress may influence the balance between co-transmitted signals. Vesicle docking and priming, the size of the readily releasable pool, SNARE-related exocytotic machinery, synapsin-dependent vesicle mobilization, and the spatial coupling between vesicles and voltage-gated Ca2+ channels all contribute to transmitter release efficiency. Stress alters presynaptic release machinery even in circuits that are not primarily discussed as co-transmitting systems. Acute stress increases depolarization-evoked glutamate release in the PFC and frontal cortex through corticosterone-dependent accumulation of presynaptic SNARE complexes. Corticosterone also promotes synapsin I phosphorylation and increases the trafficking of glutamatergic vesicles into the readily releasable pool, thereby preparing presynaptic terminals for enhanced release (Musazzi et al. 2010; Treccani et al. 2014). Chronic stress also changes the expression of genes and proteins related to synaptic vesicle exocytosis in the hippocampus and PFC. Chronic restraint stress increased the expression of the vesicular SNARE protein VAMP2 without changing the synaptic-density (Gao et al. 2006; Müller et al. 2011). Together, these findings indicate that chronic stress produces sustained, brain-region-specific remodeling of presynaptic release machinery (Figure 3). In co-transmitting synapses, transmitter-specific release may depend on distinct presynaptic architectures. In the SuM→ hippocampus pathway, glutamate and GABA are released from distinct vesicle populations. Both forms of transmission depend on P/Q-type Ca²+ channels, but GABA release is more tightly coupled to Ca²+ influx. Their release properties also differ. Glutamatergic transmission shows frequency-dependent depression, whereas GABAergic transmission remains relatively stable across stimulation frequencies (Hirai et al. 2024). Such presynaptic differences are likely to change release probability, presynaptic Ca²+ coupling, and responses to different stimulation frequency. Stress-dependent changes in vesicle recycling, vesicle-pool organization, or the spatial relationship between release sites and calcium channels could therefore alter not only total amount of transmitter release but also the relative contribution of individual co-transmitted signal. Direct evidence is still limited. At present, stress is well established to remodel presynaptic release machinery, and co-transmitting synapses rely on may organize different transmitters-specific through distinct release architecture. The remaining task is to determine whether stress directly rearranges these presynaptic structures in co-transmitting circuits remains an important question for future investigation. Technical limitations and conclusion Current evidence for stress-related co-transmission remains uneven across experimental levels. Many studies rely on electrophysiological approaches in rodent ex vivo slice preparations. Although these approaches have been instrumental in defining vesicle organization, release dynamics, and receptor-mediated responses with high resolution, they cannot fully capture the physiological conditions associated with stress in living animals (Calhoon and Tye 2015). Direct measurements of how the relative release of co-transmitted signals changes in defined projection pathways during stress remain scarce. This limitation complicates interpretation of molecular findings. Stress changes in the expression of Tph2, VGLUT3, GAD, VGAT, VMAT2, or related transporters may suggest co-transmission remodeling, but these changes alone do not demonstrate causal link to stress-related behavior (Nusbaum et al. 2017; Silm et al. 2019). They may contribute to pathological remodeling, reflect compensatory responses, or just represent the current state of the circuit. Future studies should measure transmitter release in defined projection pathways at specific stages of stress exposure. Genetically encoded sensors for glutamate, dopamine, serotonin, and GABA, combined with projection-resolved labeling in behaving animals, provide opportunities to observe changing co-transmission in vivo (Marvin et al. 2013; Patriarchi et al. 2018; Sun et al. 2018; Marvin et al. 2019; Wan et al. 2021). A recent study using high-performance GRAB sensor also showed neurotransmitter release kinetics with two-photon microscopy in live fruit fly brain (Xia and Li 2025). The in vivo sensors enable us to observe co-transmission change during stress exposure. These approaches will be particularly informative when paired with pathway specific manipulation of co-transmission-related molecules. For example, conditional deletion or restoration of VGLUT2, VGLUT3, VGAT, VMAT2, or related enzymes in defined projections could directly test whether changes in transmitter-output balance contribute to stress susceptibility or resilience. Stress paradigms should also be considered individually rather than as interchangeable models. Stress paradigms are categorized by duration and modality that each induce distinct forms of circuit adaptation (de Kloet et al. 2005; Roozendaal et al. 2009; Popoli et al. 2012; Chaudhury et al. 2013). Acute, subchronic and chronic stress may modulate how synaptic function is altered. Acute stress may preferentially influence rapid activity-dependent transmitter recruitment and glucocorticoid-sensitive presynaptic release regulation, whereas prolonged stress is more likely to involve longer-lasting changes in transmitter-related gene expression, vesicular transporter levels, and release machinery. In the same way, stress modalities such as foot shock social defeat, neuropathic pain, and withdrawal-related states regulate synapse in different ways depending on how these affect the brain. Another issue is the relationship between synapse-specific co-transmission and volume transmission, particularly in monoaminergic synapses. Although this review focuses on vesicle-mediated synaptic release, monoamines also signal extrasynaptic diffusion (Agnati et al. 2010). Vesicle-based co-transmission and diffuse neuromodulatory signaling co-exist within same monoamine co-transmission circuit. Catecholaminergic pathway adds further complexity because extracellular transmitter levels are determined by transporter-mediated uptake. Distinguishing and understanding these signaling modes will be essential for interpreting how stress alters circuit function and behavior (Agnati et al. 2010; Liu et al. 2018). Stress-related circuit dysfunction cannot always explained by changes in the amount of single neurotransmitter (Hnasko and Edwards 2012; Shabel et al. 2014). The studies reviewed here indicate that stress can also alter the relative contribution of co-transmitted signals within circuits. Evidence is strongest for changes in transmitter-output balance associated with altered transmitter identity, whereas activity-dependent transmitter recruitment and presynaptic release-site regulation remain less well understood. Determining how these mechanisms operate across different stress conditions and projection-defined circuits will be important for understanding how stress reshapes synaptic communication and behavioral adaptation. Funding Statement This work was supported by the National Research Foundation of Korea [grant number RS-2025-00520820] and by the Konkuk University Researcher Fund in 2024. Disclosure statement No potential conflict of interest was reported by the author(s). Declaration of generative AI use. The authors report generative AI was not used in their research or preparation of this manuscript. References