3.1 Hippocampal function
Several different functional roles have been attributed to the hippocampus. Penfield and Erickson studied the involvement of the hippocampus in olfaction (Penfield, 1942). In 1937, James Wanceslas Papez (1883-1958) proposed that the hippocampus was involved in emotion processing and proposed the “Papez circuit” (Papez, 1995). Although the identification of the functional role of the hippocampus took researchers occupied for decades (we will focus on this topic in the following chapters), the study of the anatomy took its own way. The first anatomist who described the hippocampus was Julius Caesar Aranzi (1530-1589) in 1564. Aranzi named it hippocampus because of its similarity to the tropical fish Seahorse (Per Andersen, 2007).
Today, one of the most studied function of the HPC is spatial memory and representation of the environment. Edward Tolman (1948) first advocate the idea of the existence of a cognitive map in the brain able to represent the space (Tolman, 1948).
Twenty years later, O’Keefe and Dostrovsky in 1971 first described place cells in the CA1 field of the anterior dorsal hippocampus of rats. They demonstrated, by combining electrophysiological recordings with animal behavior, that cells in the hippocampus of the rat increased their firing activity (preferential firing) when the animal was in a specific location of the environment (its place field), they named these cells place cells (O'Keefe, 1976). This discovery conveyed O’Keefe, together with May-Britt and Edvard Moser, the Nobel Prize in Medicine in 2014. There is also evidence of the existence of place cells in humans (Ekstrom et al., 2003). Different place cells have different place fields, which have been reported to be stable across recording sessions in the same environment. As reported by different studies, place cells are principal cells of CA1, CA3 and DG but there
is also evidence for cells having similar functions in the subiculum (Kjelstrup et al., 2008;
Sharp and Green, 1994; Sharp, 1997). One characteristic of place cells is that their position in the dorsoventral axis determines the size of the place field they code for. Their place field increases from the dorsal to the ventral axis. Another property of place cells is that variations in the environment can influence their firing rates (Anderson and Jeffery, 2003).
More recently, Marianne Flynn together with colleagues in the laboratory of May-Britt and Edvard Moser (Fyhn et al., 2004) described another very interesting characteristic for spatial representation in the brain, the grid cells in the medial entorhinal cortex (MEC).
Grid cells were lately observed also in the presubiculum and parasubiculum (Boccara et al., 2010). Grid cells, like place cells, fire in a specific location of the environment but they present multiple firing fields organized in hexagonal grid-like configuration. Importantly, while place cells may differentially codes multiple contexts, grid cells maintain the positional relationships between distinct environments. Some properties of grid cells have been described: the scale between the firing fields, (distance), the orientation of grid axes according to direction references and the spatial phase. Moreover, grid cells also show differential tuning of fields’ size along the dorsoventral axis of the MEC like place cells.
Altogether, these evidences show that the HPC has a functional role in spatial navigation, representation of the environment and the formation of cognitive maps.
3.2 The role of the hippocampus in memory
As seen in chapter 1.2, clinical studies have shown evidence that the HPC is a core structure for episodic memory. Today, the HPC is known to play fundamental roles in the formation, retention, recall and extinction of episodic memories (Per Andersen, 2007). To precisely define the role of HPC in memory, researchers have used animal models with a huge diversity of approaches which include: experimental memory tasks together with lesional studies (Aggleton and Pearce, 2001), the use of toxins (Jarrard, 1989) and drugs (McGaugh and Izquierdo, 2000) to alter cells activity and genetic approaches (Tonegawa et al., 1995).
These studies led researchers to propose models explaining hippocampal function. We have models suggesting that sensory information from the entorhinal cortex to the dentate gyrus is encoded into CA3 auto-association networks and can be subsequently retrieved via Shaffer collateral activation of CA3-CA1 synapses (Rolls and Kesner, 2006; Nakashiba et al., 2008; Wang and Morris, 2010). Precisely, the learning-induced memory trace is thought to be encoded by the occurrence of plasticity in CA3-CA3 auto association circuits
and that correct retrieval of this memory trace is dependent on plasticity in CA3-CA1 synapses (Rolls and Kesner, 2006; Gruart et al., 2006; Whitlock et al., 2006; Pastalkova et al., 2006). The sensory information coming from the entorhinal cortex reaches the DG and undergo a process called pattern separation (described below) to produce sparse representations that are encoded in CA3 (Leutgeb et al., 2007). Direct entorhinal projection to the CA3 and CA1 are thought to be necessary to provide precise ongoing sensory inputs (cues) able to modulate the acquisition and/or retrieval of the memory trace via activation of Shaffer collaterals and back-projections from CA1 to the neocortex (Levy et al., 1998; Remondes and Schuman, 2004).
3.2.1 Pattern separation and completion as mechanisms to store and retrieve memories The HF has the extraordinary ability to perform processes called pattern separation and completion. Pattern separation and completion are processes thought to occur in the trisynaptic pathway, in the DG, CA3 and CA1. Pattern separation is a process allowing the separation of partially overlapping patterns of activity. The outcome will be that retrieval of one pattern is separated from other patterns, providing specificity to memory encoding.
Two similar or partially overlapping patterns of neuronal activity arriving from the entorhinal cortex produce very different output patterns in the DG (Deng et al., 2010; Leutgeb et al., 2007). In the CA1 areas, this pattern separation is displayed by place cells (Moser et al., 2015). How is this possible and why? One of the possible explanations can be found in the anatomy of the HPC. First, it is known that the DG has much more cells (5-10 times more;
(Amaral et al., 2007)) than its afferent EC and its downstream target CA3. This peculiarity confers the DG the physical properties to perform as a separator. Second, the DG has a very sparse coding pattern, where only few sparse cells code for a particular information (Houser, 2007). The main reason for such pattern separation is essentially the ability to discriminate different memories (Leutgeb et al., 2007; Deng et al., 2010). On the other hand, we have the process of pattern completion. Pattern completion is thought to occur mainly in the CA3 via auto-association network as described above and basically consist in the ability to re-activate a network of neurons storing a particular memory by the reactivation of only a set of neurons via Shaffer collaterals activating CA1 cells.
An evidence of the role of DG and CA3 in pattern separation is the study of Leutgeb et al. (2007) where they performed in vivo electrophysiological recording of DG and CA3 principal cells of rats running in two different boxes (squared or round). They investigate DG and CA3 pattern separation during changes in spatial and temporal coincidence. Mice were trained in a square maze that was then transformed into a round maze (flexible walls)
and were recorded during this environmental transition. Recordings showed a progressive change in firing properties of place cells in DG and CA3 during context transformation.
They revealed the recruitment of new cell populations in CA3 but not in the DG when the environments differ over time. These results have brought to light a dual mechanism for pattern separation where DG and CA3 differentially represent spatial environments. The DG can de-correlate the EC information while CA3 is able to use dedicated neuronal representation for each environment.
3.2.2 Trisynaptic pathway vs direct pathway
Sensory information to the HPC can either pass through the trisynaptic partway or via a direct pathway relying EC layer III to CA1. A question that arises is to know whether memory consolidation through the trisynaptic pathway specifically require CA3 integrity or it can be accomplished via the direct pathway. To test this hypothesis, Nakashiba and his colleagues (2009) developed a CA3-TeTX transgenic mouse where specifically CA3 output can be controlled (Nakashiba et al., 2009). Mice were trained in the contextual fear-conditioning paradigm and CA3 output was blocked post-training. They performed in vivo electrophysiological recordings of CA1 cells during CA3 output blockade after fear conditioning of mice. They found that: post-training blockade of CA3 output impairs the consolidation of the contextual fear memory; CA3 output blockade reduced intrinsic frequencies of CA1 ripples and reduced reactivation of CA1 experience-coding cells. This study reveals that memory consolidation require CA3 integrity in the trisynaptic pathway.
3.3 Hippocampal neurogenesis and memory
Adult neurogenesis has been firstly described by Joseph Altman in 1965 in the hippocampus of rodents and subsequently described also in the DG of humans (Altman and Das, 1965; Cameron and McKay, 2001; Spalding et al., 2013; Bergmann et al., 2015).
Adult neurogenesis in the hippocampus generates granule cells in the DG from precursor cells in the subgranular zone (SZ, between the hilus and granule cell layer). New born granule cells in the adult brain are incorporated into preexisting circuits known to contribute to structural plasticity of the DG (axons and dendritic rearrangements, and synapotgenesis) (Hastings and Gould, 1999; Kaplan and Hinds, 1977; Markakis and Gage, 1999). Adult neurogenesis is believed to play major roles in processes such as pattern separation and completion and in forgetting (Nakashiba et al., 2012; Frankland et al., 2013). Nakashiba et al. demonstrated that ablation of adult GCs neurogenesis impaired pattern separation during contextual memory discrimination tasks and showed
some deficits in rapid pattern completion (Nakashiba et al., 2012). Frankland et al. recently proposed a model for the role of neurogenesis in forgetting (Frankland et al., 2013). In their review, they suggest that adult GCs neurogenesis are implicated in a process aiming at clearing memories in the hippocampus. As new born GCs integrate the already existing circuits, their connectivity will bring new information that will interfere and compete with the preexisting one. This hippocampal circuit remodeling is thought to interfere with the process of pattern completion, occurring during memory recall, of pre-existing stored information and thus induce forgetting.