As an organism navigates through an environment, its ability to associate relevance to specific contexts is crucial for survival. Mammals have evolved a system to form a cognitive spatial map of the environment relying on the hippocampus, a limbic structure required for the formation of new episodic memories (Smith and Bulkin, 2014). The hippocampus consists of excitatory neurons forming a trisynaptic loop, with local and hilar interneurons providing feedforward and feedback inhibition (Amaral and Lavenex, 2006). In addition, neuromodulation from two main sources influences hippocampal processing: the medial septal cholinergic and GABAergic projections that influence hippocampal theta rhythms (Bland and Oddie, 2001), and catecholaminergic projections from the midbrain and the locus coeruleus (LC; Madison and Nicoll, 1982; Otmakhova and Lisman, 1999). The association of events to contexts is facilitated via plasticity of synaptic connections, as first proposed by Hebb (1949). These synaptic changes are bidirectional and depend upon specific neural activity and protein synthesis (Abel et al., 1997). Importantly, not all events and contexts have similar relevance, and as such, neuromodulators such as catecholamines can enhance synaptic plasticity associated with an event, linking the synaptic change to the memory of the event (Frey and Frey, 2008; Shohamy and Adcock, 2010; Lisman et al., 2011).

Experimental evidence has shown that, as with other brain structures, changes in hippocampal synaptic plasticity and hippocampus-dependent learning are influenced by dopaminergic neurotransmission at these synapses. There are two main sources of dopamine (DA), one arising from the LC, and the other from the ventral tegmental area/substantia nigra pars compacta (VTA/SNc) of the midbrain (Duszkiewicz et al., 2019). The primary neurotransmitter of the LC, however, is norepinephrine (NE), which is released at a 10-fold concentration to that of DA (Kempadoo et al., 2016). By comparison, the VTA and midbrain DA projections have been shown with previous methods to have very sparse innervation of the hippocampus (McNamara et al., 2014; Kempadoo et al., 2016; Takeuchi et al., 2016). A confounding factor is that the hippocampus does not show significant expression of the DA transporter, which is the crucial anatomical marker of DA projections in other brain regions (Borgkvist et al., 2012). A goal of this review is to provide the most recent evidence in support of a direct anatomical projection from dopaminergic VTA neurons to the hippocampus. Then, DA neuromodulation of hippocampal synapses will be discussed within the framework of spike timing dependent plasticity (STDP). Finally, the relevance of dopaminergic neurotransmission to learning and memory will be reviewed.

The existence of direct dopaminergic innervation in the rodent hippocampus has been historically controversial, with this controversy persisting until recently. The low extracellular concentrations of DA in the hippocampal formation, especially in comparison to the striatum, have initially led scientists to believe that the hippocampus was completely devoid of DA innervation (Lindvall and Bjorklund, 1978).

The development and use of anterograde and retrograde fluorescent tracing techniques provided the first strong indications of midbrain dopaminergic innervation of the hippocampus (Gasbarri et al., 1997). The first evidence came from experiments using co-localization analysis of retrogradely transported True Blue (TB) from the hippocampus with tyrosine hydroxylase (TH) expressing cells in midbrain dopaminergic areas (Swanson, 1982). It was shown that following TB injection in different rat hippocampal regions [CA1, CA3, and dentate gyrus (DG)], retrogradely-labeled DA neurons could be identified in the rostral VTA and SNc. These results found further support from studies analyzing TH-positive fibers in the hippocampus of NE-depleted rats (Verney et al., 1985) that demonstrated the existence of DA innervation predominantly in the CA1 and subiculum and to a lesser extent in the CA3 and DG.

Electrolytic and chemical lesions of the VTA/SNc or their hippocampal terminals also indicated direct midbrain dopaminergic innervation of the hippocampus. In these lesion experiments, rats were pretreated with hippocampal injections of dismethylimipramine to prevent degeneration of noradrenergic terminals and then given an injection of the neurotoxicant 6-hydroxydopamine (6-OHDA) into the dorsal hippocampus. This selective monoaminergic toxin caused destruction of midbrain DA afferents, producing a dramatic decrease in DA content in the dorsal hippocampus (Itoh et al., 1984). Moreover, chemical and electrolytic lesions of the VTA and SNc (Scatton et al., 1980) led to a severe reduction of hippocampal DA and 3,4-Dihydroxyphenylacetic acid (DOPAC) concentrations. Taken together, these data suggest that an important portion of hippocampal afferents are dopaminergic and originate from the VTA/SNc regions.

Almost a decade later, studies by Gasbarri et al. (1994a,b, 1996) provided a rigorous analysis of the topographical distribution of midbrain DA neurons and their projections to the different regions of the hippocampal formation of the rat. Their data showed a direct innervation of the hippocampus from the VTA, SNc, and retrorubral field (RRF) that was more prominent at CA1 and subiculum in the ventral part of the hippocampus as shown previously (Verney et al., 1985). By combining retrograde tracing using Fluoro Gold (FG) and TH immunostaining, they quantified that only a part (15%–18%) of this innervation is actually dopaminergic. Importantly, these dopaminergic projections did not produce any collaterals to other regions.

Despite all this evidence, the extent of hippocampal innervation by midbrain dopaminergic sources in the rat brain remained underwhelming. Techniques using genetic mouse models and cell-type specific viral tracing approaches allowed more recent studies to analyze this innervation in the mouse brain. These endeavors were mainly hindered by the inefficiency of retrograde transport from dopaminergic terminals (Tervo et al., 2016). Some of these studies have attempted to map midbrain dopaminergic projections to the dorsal CA1 (dCA1) hippocampal field using axon-targeted channelrhodopsin (ChR2) viral methodologies (McNamara et al., 2014; Kempadoo et al., 2016; Takeuchi et al., 2016). All of these studies have provided similar results that showed sparse innervation with only a limited number of dopaminergic fibers present in the dorsal hippocampus and the dCA1 region in particular. Since ChR2 is not an anterograde tracer, it is likely that these results underestimated the amount of hippocampal innervation that stems from midbrain dopaminergic sources. In fact, several large-scale input-output tracing efforts (Beier et al., 2015; Lerner et al., 2015; Poulin et al., 2018) have omitted the hippocampus from their analysis mainly due to the sparse dopaminergic innervation in this area. These discrepancies suggested the existence of alternative sources of DA in the rodent hippocampus, namely the classic NE area, the LC (Devoto et al., 2001; Smith and Greene, 2012). LC neurons synthesize DA as a precursor of NE, and they have been shown to co-release both neurotransmitters in the hippocampus, acting as an alternative source of DA (Smith and Greene, 2012; Kempadoo et al., 2016; Takeuchi et al., 2016). These data refueled the controversy regarding the source of DA in the rodent hippocampus and created a narrative that all physiologically relevant DA functionality comes almost entirely from the LC.

To address the source of DA innervation, two recent studies used a sensitive anterograde viral tracing methodology to examine the distribution of direct midbrain dopaminergic projections to the dCA1 (Broussard et al., 2016; Tsetsenis et al., 2021). This method utilizes an adeno-associated virus (AAV) to express the synaptic vesicle protein synaptophysin fused with a fluorescent protein exclusively in midbrain dopaminergic neurons of the VTA/SNc (Figure 1A). Because synaptophysin is expressed multiple times in the many synaptic vesicles in each axonal terminal, the localization of synaptophysin fusion proteins results in an enriched fluorescent signal in DA neuronal terminals enhancing the sensitivity of detection (Wiedenmann and Franke, 1985). Using synaptophysin fluorescence, prominent punctate fluorescent terminals were identified along the pyramidal cell layer (PCL) in the dCA1 (Broussard et al., 2016; Tsetsenis et al., 2021). Additionally, sparse fiber tracts were observed in stratum oriens (SO) and stratum radiatum (SR) in agreement with what was previously shown with ChR2 expression. These labeled puncta and fibers were also immuno-positive for TH, confirming their midbrain dopaminergic identity.

To differentiate the VTA/SNc and LC sources further, Tsetsenis et al. (2021) generated conditional knockout mice in which TH was specifically ablated from all neurons expressing DA beta-hydroxylase (Dbh), which catalyzes the conversion of DA to NE. Thus, in these mice, TH is removed from noradrenergic neurons in the LC (TH^LC KO), which is the source of noradrenergic input to the dorsal hippocampus. The procedure leaves midbrain dopaminergic fibers and terminals as the sole source of the TH signal in the hippocampus. TH immunofluorescence was significantly reduced in the dCA1 region of TH^LC KO mice, particularly in the SO and SR, but a substantial signal remained in the PCL. The remaining TH signal accounted for one-third of the TH signal of control mice. This smaller signal strongly resembled the VTA/SNc innervation to the dCA1 seen using the synaptophysin anterograde tracing approach, indicating that it represents the midbrain dopaminergic input.

To identify the midbrain cell population that directly projects to the dorsal hippocampus, the same study used two viral retrograde strategies (Figures 1B–D) based on canine adenovirus type 2 (CAV2; Soudais et al., 2001) in combination with TH immunohistochemistry and cell-type specific viral and genetic expression of fluorescent markers (Tsetsenis et al., 2021). Injection in the dorsal hippocampus with these retrograde viruses labeled a population of DA neurons near the interface of SNc and lateral VTA. These cells exhibited electrophysiological properties (i.e., firing frequency, I_h current amplitude, input resistance, resting membrane potential) typical of dopaminergic neurons (Tsetsenis et al., 2021). These findings provide strong evidence for the existence of a direct DA innervation in the hippocampus originating from midbrain dopaminergic sources and add important information to the topographical chart of VTA/SNc connectivity, with a focus on a target region that has often been overlooked in DA circuit-mapping analyses. Next, we will review how these two distinct DA sources modulate hippocampal plasticity and learning and memory processes.

Given that anatomical evidence indicates the presence of D1 and D5 receptors in the hippocampus (Savasta et al., 1986; Wamsley et al., 1992) and the anatomical tracing studies highlighted in the previous section, how does dopaminergic transmission in the hippocampus influence information processing in the hippocampus? Early studies of intact synapses from brain slices provided evidence that D1-type receptors facilitate long-term potentiation (LTP) in the CA1 via activation of protein kinase A (PKA; Huang et al., 1995; Kentros et al., 2004; Muzzio et al., 2009). Because D1/D5 receptors are positively coupled to cAMP-PKA, drugs that enhanced this pathway attenuated age-related deficits in memory and LTP (Bach et al., 1999). PKA was also shown to be involved in maintaining the stability of place fields in the CA1. Blockade of PKA had the same effect as blockade of protein synthesis: i.e., it impairs the long-term stability of hippocampal place fields (Rotenberg et al., 2000; Agnihotri et al., 2004; Kentros et al., 2004) and also the ability to retain long-term memories.

The most prominent model proposed to synthesize these findings is the synaptic tag and capture model by Frey and Morris (1997, 1998) and (Viola et al., 2014). This model posits that strong stimulus events induce a synaptic tag comprised of the phosphorylation of receptors or kinases. The second signal, coming from dopamine receptor activation, activates the cAMP/PKA pathway which synergistically act with the synaptic tag to produce plasticity related proteins (O’Carroll and Morris, 2004).

The synaptic tag and capture model has provided a framework to explain how Hebbian processes in the hippocampus can be modified by catecholamines (Lisman et al., 2011). Complementing this conceptual model is evidence that dopamine modifies spike timing dependent plasticity (STDP). STDP is a biologically relevant model for hippocampal plasticity that provides evidence for a Hebbian model of hippocampal circuitry underlying learning and memory. STDP requires the coordinated generation of postsynaptic action potentials and presynaptic stimulation within a limited temporal window (Bi and Poo, 1998, 1999, 2001; Tao et al., 2000). STDP is a refinement on Donald Hebb’s treatise that the correlated activity of hypothetical groups of neurons called “cell assemblies,” a central tenant of which is the importance of temporal order and the spatial specificity ensuring that only activated synapses become modified (Hebb, 1949; Bi and Poo, 2001). Early studies investigated the synaptic modifications of entorhinal input to the DG (Levy and Steward, 1983). Specifically, they tested associativity of the synaptic input to the DG and found that a weak stimulation preceding a strong stimulation by as much as 20 ms, would potentiate the response to the weak stimulation. However, if the strong stimulation preceded the weak stimulation, then the postsynaptic response to the weak stimulation would be depressed.

Postsynaptic spiking is an important event in the synaptic plasticity process (Nicoll and Malenka, 1995; Bi and Poo, 1998, 2001). Postsynaptic neurons fire action potentials, and the active properties of dendrites allow the spike initiating at the axon hillock to back-propagate into the dendrites. This backpropagation provides a precise timing signal to dendritic synapses to play an active role in associative synaptic modification. In the CA1, it was shown that back-propagating spikes coupled with subthreshold stimulation evoked a higher Ca²⁺ influx, potentiating the synapse (Magee and Johnston, 1997). Several channels contribute to the postsynaptic response, including NMDA channels and A-type potassium channels at the postsynaptic site.

In a series of experiments recorded from the granule cells of the DG, the spike timing dependent protocol was demonstrated in an intact slice to produce LTP and long-term depression (LTD) in specific timing protocols (Yang and Dani, 2014). When D1 or D5 receptors were inhibited either pharmacologically or genetically (with KO mice), the same spike timing protocol failed to induce LTP. Furthermore, the addition of D1/D5 receptor agonist SKF81297 in intact mice would facilitate LTP while using an STDP protocol otherwise typically produces LTD. Mechanistically, this effect on postsynaptic plasticity was dependent upon potassium channels. Mice that lacked the inward rectifying Kv4.2 channel were less likely to show the effect of D1 receptor antagonists, and the use of selective I_A channel blocker 4-AP mimicked the faciliatory effect of D1 receptor agonists. D1 and D5 receptors signaling through adenylyl cyclase and PKA results in activation of mitogen-activated protein kinase (MEK). MEK in turn acts to inactivate I_A channels (see Figure 2). Blocking I_A channels serves to boost back-propagating action potentials from the pyramidal soma, which is the mechanism underlying the coincidence of presynaptic activity and postsynaptic response in the STDP protocol. Thus, in a biological system, dopamine serves to facilitate and widen the temporal window of STDP, which would influence information processing throughout the hippocampus. This evidence provides a model for a mechanism of dopamine modulation of Hebbian plasticity. These mechanistic experiments examined the perforant path to DG circuit, but dopamine had a similar effect on CA1 synapses, namely, an increase in the gain of sensitivity and a loss of temporal contrast of the spike timing mechanism (Zhang et al., 2009). In both the CA1 and the DG, dopaminergic modulation of spike timing dependence required active neurotransmission of NMDA receptors. Taken together, these findings provide evidence of a biological mechanism underlying dopaminergic modulation of synaptic plasticity (Figure 2).

The dopaminergic modulation of STDP provides mechanistic insight into the channels and subunits underlying Hebbian synaptic plasticity, but assessment of measures of plasticity following hippocampus-dependent learning is required in order to provide evidence of the necessity of dopaminergic neurotransmission to learning. One common technique is to measure the AMPA/NMDA (A/N) ratio of hippocampal slices extracted ex vivo from mice that had recently undergone a learning procedure, such as the novel object recognition (NOR) task (Yang et al., 2017). In the NOR task, a mouse sees two objects in an open field, and then in the next trial one object is changed, the novel object. Mice spend more time exploring the novel object rather than the unchanged object. The A/N ratio is a common measure of synaptic plasticity, usually indicating an increase in specific postsynaptic AMPA receptors following a plasticity-inducing learning event. Previous evidence indicated that the DG was crucial in the detection of novelty (Rolls and Kesner, 2006; Kesner and Rolls, 2015), and thus whole cell patch clamp recordings from granule cells of DG neurons were used to measure changes in the A/N ratio of these neurons. While recording, stimulation was passed through the perforant pathway, which is the bundle of axons projecting from the entorhinal cortex to the DG. The NOR task provided evidence of an anatomical dissociation of hippocampal function. In the DG, NOR produces an increase in the A/N ratio, driven specifically by an increase in the AMPA current following the introduction of a novel object compared to mice that had seen identical object pairs (without novelty; Yang et al., 2017). By contrast, in the CA1, no novelty-related plasticity was observed.

In contrast, inhibitory avoidance (IA) footshock training in a distinct context produced plasticity in a different circuit of the hippocampus. In this task, mice are in a well-lighted space, and a door opens to a dark, unlit space. The mice quickly move to the dark space where they are given a footshock. Thus, in subsequent trials, the mice delay entering the dark space when the door opens. IA training produced an increase in the A/N ratio in the synapses arising from the Schaffer collateral/CA1 pathway, but synapses in the perforant pathway remain unchanged (Broussard et al., 2016). This functional dissociation remains consistent with the role of the DG in pattern separation in an environment (Rolls and Kesner, 2006), compared to the contextual discrimination that requires the CA1.

Dopaminergic neurotransmission was necessary for the changes in the A/N ratio observed in both the NOR and IA studies. When animals were injected with a D1-type receptor antagonist, SCH23380, during the learning of the novel-object task, the A/N ratio remained unchanged. These experiments were confirmed with D1- and D5-receptor KO mice, indicating a role for both receptors in novelty-related plasticity. For the IA task, there was an increase in the field excitatory postsynaptic potential (fEPSP) amplitude, and the hippocampal synaptic potentiation was contingent upon intact DAergic neurotransmission. Specifically, the administration of D1 receptor antagonists just prior to and during training blocked the acquisition of long-term memory and the changes in the fEPSP.

A second measure of plasticity, taken in vivo from freely moving animals, is the fEPSP. The fEPSP measures the postsynaptic responses of a neuronal population in the vicinity of the recording electrode (Clarke et al., 2010). Behavioral experiments can be paired with high or low-frequency stimuli, and it can be determined whether these behaviors contribute to long-term plasticity. For example, an exploration task bidirectionally enhanced synaptic plasticity. That is, if paired with a weak low-frequency stimulus, environment exploration produced LTD, and if paired with a high-frequency stimulation, exploration produced LTP in the CA1. This bidirectional plasticity was once again dependent upon D1/D5 receptor activity (Lemon and Manahan-Vaughan, 2006, 2012).

Long-term potentiation is one mechanism that drugs of abuse utilize to hijack dopaminergic modulation of hippocampal function (Ungless et al., 2001). For example, an injection of biologically relevant doses of nicotine enhances the fEPSP in the DG (Tang and Dani, 2009; Zhang et al., 2010). Local infusions or systemic injections of D1 receptor antagonists into the hippocampus block drug-induced potentiation (Jenson et al., 2015). Other drugs of abuse such as cocaine and methamphetamine also increase plasticity in the hippocampus (Broussard et al., 2012; Jenson et al., 2015). When paired with a low amplitude theta burst (which does not by itself potentiate these synapses), a low dose of methylphenidate increases the population spike amplitude in the DG (Jenson et al., 2015). A low dose of methylphenidate has also been shown to extend the temporal window of STDP in the DG, in a dopamine-dependent manner. The dose that was shown to enhance plasticity also enhanced memory retention in a delayed non-match-to-position task.

The experimental evidence reviewed here strongly supports an important modulatory role for DA in the regulation of long-term plasticity in the hippocampus with direct effects on learning and memory. This neuromodulatory action is largely mediated by activation of D1-type receptors in the dorsal hippocampus, which facilitates LTP and promotes spatial and contextual information encoding and storage (Hansen and Manahan-Vaughan, 2012). A possible mechanism is that DA signaling in the hippocampus lowers the threshold for synaptic plasticity that underlies learning and broadens the timing window between presynaptic and postsynaptic activity (Yang and Dani, 2014). Another possibility is that DA acts as a regulator of different inputs onto dCA1 pyramidal cells (Rosen et al., 2015). These cells receive and integrate information from the CA3 via the Schaffer collateral (SC) pathway and directly from the entorhinal cortex via the temporoammonic (TA) pathway. DA may act to rebalance the relative weights of SC and TA inputs by modulating the intrinsic and synaptic properties of hippocampal GABAergic interneurons, in a similar manner to the effects recently shown for acetylcholine (Palacios-Filardo et al., 2021).

Another level of complexity in deciphering the function of hippocampal DA in regulating cognitive processes is added when one considers the origin of the dopaminergic signal. Recent evidence from studies using optogenetic methods to enhance DA release in the dorsal hippocampus showed that both the VTA/SNc and the LC can release DA in the dCA1 to facilitate specialized aspects of learning (Table 1). How these two signals that act primarily through the same mechanism of D1-type receptor activation can produce diverse behavioral outcomes is still not completely clear. It is possible that VTA/SNc dopaminergic fibers target different synapses or cell populations in the dorsal hippocampus allowing for different types of neuromodulatory action (McNamara and Dupret, 2017). While innervation from the midbrain may be sparser, it can be equally efficient in releasing DA since the extracellular concentration of DA released from LC fibers constitutes only a small fraction of the amount of NE released (Kempadoo et al., 2016). In fact, the amount of NE co-released with DA from LC fibers should not be disregarded as NE can also influence associative memory formation in the dCA1 (Tsetsenis et al., 2022). Moreover, NE has a high affinity and has been shown to act as a potent agonist for D2-type receptors (Lanau et al., 1997; Sanchez-Soto et al., 2016), suggesting that it can activate dopaminergic signaling pathways independently from DA. Also, experiments that rely on LC soma activation should be designed with caution, as it is possible that the observed effects mediated by dopamine receptors in the hippocampus may be due to DA release from VTA/SNc terminals, via activation of LC-VTA inputs (Simon et al., 1979; Mejias-Aponte et al., 2009; Shelkar et al., 2017) and not necessarily attributed to release from noradrenergic terminals. Future research aiming to dissociate the functionality of DA release from these loci would benefit from the development of sensitive techniques for detecting DA in the hippocampal formation as well as discriminating DA from NE release (Dong et al., 2022).

TT and JB wrote the initial draft of the manuscript. JD edited and wrote the final manuscript and contributed with financial support. All authors contributed to the article and approved the submitted version.