01-intro.Rmd

\mainmatter

<!-- ```{=latex} -->

<!-- % Trigger ToC creation in LaTeX -->

```{=tex}
\setcounter{tocdepth}{3}
\tableofcontents
```
<!-- ``` -->

# Introduction {#intro}

The evolutionary goal of a species is to survive, which is achieved by adapting to environmental conditions, optimising strategies to live in the specific environment and reproduce [@darwin_origin_1859]. Higher organisms, therefore, have developed structures which are able to help with optimising their behaviour on different levels. With multicellular organisms evolving $570$ million years ago, more complex functions such as mechanosensation (touch, hearing), olfaction based on chemotaxis, and vision developed [@schlosser_short_2018], helping them to learn about their environment. Hence, the encoding of memories evolved and similar mechanisms to achieve memory on the synaptic scale were observed in invertebrates and vertebrates [@glanzman_common_2010]. Information which is accumulated over the course of time -- from being exposed to sensory stimuli -- is stored across different areas of the brain or nervous system [@christophel_distributed_2017; @mesulam_sensation_1998] and can be encoded for longer by protein synthesis and structural changes in synapses and cells [@bourne_nanoscale_2012; @hyden_protein_1968]. One mechanism to induce such change is through organised cellular activity, resulting in an increase of calcium in dendrites and synapses. Hence, synaptic inputs which occur shortly before postsynaptic activity can increase calcium concentrations enough to induce spike time dependent plasticity (STDP) and protein synthesis to stabilise connections [@dan_spike_2006; @hebb_organization_1949]. When different cells strengthen connections based on the same inputs, they become related and act together in the future. Such an association of cells is often referred to as *"cell-assembly"* [@hebb_organization_1949].

For a distribution network of memory to work, the coordination of areas and timing of cells is crucial for memory formation and can be achieved by oscillatory activity [@bressler_episodic_1993; @varela_brainweb_2001]. These oscillations arise in the brain through changes induced by population activity, synaptic inputs with the resulting membrane potential changes and can be measured as local field potentials (LFP) [@herreras_local_2016]. One important higher level function, which does not represent solely processing of sensory inputs, is computation for spatial navigation. Navigation strategies can be observed in several species such as ants [@wehner_desert_2003], drosophila [@kim_idiothetic_2017; @neuser_analysis_2008], bees [@dacke_honeybee_2007], pigeons [@bond_spatial_1981; @olson_characteristics_1983], rodents [@morris_place_1982], bats [@tsoar_large-scale_2011], and dolphins [@arranz_rissos_2018] and is essential for finding food, a partner, or avoiding danger. Over the last decades, the endeavour to understand the *"cognitive map"* [@tolman_cognitive_1948] -- how we animals use the abstract -- or internally generated models of space -- has driven research and revealed that several dimensions of space are represented on a cellular or network level in the brain (place, speed, grid, head-direction, vectors, and borders; for review see @poulter_neurobiology_2018). Coding spatial properties can be achieved by different mechanisms, allowing for a detailed representation of the environment. For instance, place cells fire at a higher rate in discrete positions of the environment which are referred to as place fields [@okeefe_hippocampal_1978; @okeefe_hippocampus_1971; @okeefe_place_1976]. This principle of coding information (in this case, a location), via the number of action potentials over time, is known as rate code. However, this is not the only possible method of transferring information. An additional way an action potential can code space is via the phase code [@mehta_experience-dependent_1997; @okeefe_phase_1993; @skaggs_theta_1996]. The phase code is based on the position of a spike relative to the ongoing LFP and is thought to communicate information about an animal's position more efficiently than a rate code alone [@jensen_hippocampal_1996; @reifenstein_grid_2012]. This has been found to play an important role in several areas of the brain (lateral septum [@tingley_transformation_2018], medial entorhinal cortex [@giocomo_temporal_2007] and the hippocampus [@skaggs_theta_1996]). The major neuronal oscillatory frequency band in mammals involved in the spatial phase code is the $\theta$ band.

## \(\theta\) oscillations

Theta ($\theta$) oscillations range from $4$ to $12\ Hz$ [@buzsaki_theta_2002] and were observed for the first time in the hippocampus and septum of rats and cats by @jung_method_1938, as well as later by @green_hippocampal_1954 in rabbits and monkeys. Exposure of rats to different stimuli led to different oscillatory responses, including oscillations in the $\theta$ range in the hippocampal area, and also in the medial septum. $\theta$ oscillations can also be found in humans, where they are associated with memory and spatial navigation [@arnolds_spectral_1980; @okeefe_theta_1999; @sarnthein_synchronization_1998]. Mechanistically, $\theta$ oscillations can be differentiated into two different types: atropine-sensitive, "slow" $\theta$ ($4-7\ Hz$) associated with rest or anaesthesia, and atropine non-sensitive $\theta$ which are slightly faster ($7-12\ Hz$) and occur during running and exploration [@kramis_two_1975]. This has led researchers to hypothesise that these oscillations are the product of two distinct networks and/or cell types.

<!-- The notion of synaptically different pathways suggests that multiple synaptic inputs contribute to distinct network activity during behaviour. -->

(ref:ThetaGammaShortCaption) Schematic of $\theta$ oscillation with nested $\gamma$ in LFP
(ref:ThetaGammaCaption) **Schematic of** $\theta$ **oscillation with nested** $\gamma$ **in LFP.** A scheme of the local field potential (LFP) (top) with the $\theta$ frequency band between 4-12Hz (middle) and nested $\gamma$ oscillations (30-90Hz) at the peak of the $\theta$ cycle (bottom).

```{r ThetaGammaScheme, fig.cap = '(ref:ThetaGammaCaption)', fig.scap='(ref:ThetaGammaShortCaption)', echo=FALSE, fig.width=7, fig.asp=0.4, fig.show='hold', fig.pos='h'}
FrequencyBandScheme <- readRDS(file = "Data/FrequencyBandScheme.rds")
FrequencyBandScheme
```

\noindent
An association of $\theta$ oscillations with spatial navigation was first reported by @vanderwolf_hippocampal_1969. Freely moving rats show synchronised hippocampal activity reflected in the LFP during exploration [@winson_patterns_1974]. Similarly, these oscillations can be found in mice and also humans: in humans they occur with lower frequency [@ekstrom_human_2005; @miller_lateralized_2018; @watrous_comparative_2013] and as burst activity whereas they are found to be constantly occurring in rodents during movement [@ekstrom_human_2005; @watrous_comparative_2013]. Following the early observations in rodents, there has been a consensus among investigators that oscillations are likely associated with navigation in behaving rodents [@okeefe_hippocampus_1971; @vanderwolf_hippocampal_1969].

(ref:quoteKD) --- Footnote by J.O'Keefe & J.Dostrovsky regarding behaviours involving movement and rest, 1971

(ref:wholeQuoteKD) \* Vanderwolf has reported (and we agree) that the former behaviours are associated with $\theta$ activity in the hippocampal EEG, while, during the latter behaviours, the EEG shows irregular slow waves.

> (ref:wholeQuoteKD)
>
> `r tufte::quote_footer('(ref:quoteKD)')`

\noindent
Cortical structures can be considered as *"current generators"* -- regions which generate an LFP response through population activity or synaptic inputs -- but as far as we know, they are not able to generate $\theta$ rhythm by themselves *in-vivo*. In contrast, the *"rhythm generator"* of $\theta$ is believed to be in deeper non-cortical regions such as the medial septum [@petsche_significance_1962] which projects to a range of cortical areas including the hippocampal formation. It is not completely proven whether some other areas have the potential to generate $\theta$ locally. Although there is no direct evidence for local rhythm generation *in-vivo*, evidence from *in-vitro* studies suggests that connectivity between glutamatergic cells and inhibitory interneurons in combination with membrane properties of cells could be sufficient to generate $\theta$ in the hippocampus and parasubiculum [@glasgow_local_2007; @goutagny_self-generated_2009]. Local interneurons of the hippocampus and the cortex do play an important role in coordinating $\theta$ activity and are a primary target of septal projections [@freund_gaba-containing_1988; @gonzalez-sulser_gabaergic_2014; @stark_inhibition-induced_2013]. Even though the hippocampus has the capability to generate $\theta$ independently of inputs from the isocortex, synchronisation is coordinated by interacting with the parahippocampal network via a feedback loop through the entorhinal cortex [@bragin_gamma_1995; @kocsis_interdependence_1999]. Pyramidal cells of the hippocampus use $\theta$ cycles for the previously mentioned phase code to represent spatial information more accurately [@okeefe_phase_1993; @skaggs_theta_1996].

In addition to spatial navigation, an increase of $\theta$ band coherence between the hippocampus and cortical structures can be observed during other memory tasks, where the increase in coherence is hypothesised to improve learning and memory formation by synchronising synaptic inputs [@bourne_nanoscale_2012; @huerta_bidirectional_1995; @markram_regulation_1997]. A possible mechanism could be a coordinated depolarisation of dendrites through time-locked synaptic inputs, which can lead to the strengthening of connections via STDP and establish cell-assemblies [@buzsaki_two-stage_1989; @huang_theta_2005]. Such a synchronisation has been shown to be crucial for working memory where $\theta$ oscillations are involved in synchronising activity on the cell level [@jensen_frontal_2002; @lee_phase_2005] and across different areas [@liebe_theta_2012; @sarnthein_synchronization_1998], and in spatial memory where an absence of $\theta$ will result in memory deficits [@winson_loss_1978].

A proposed mechanism of how $\theta$ oscillations guide memory formation is by modulating faster $\gamma$ oscillations ($30-90\ Hz$) which are nested in $\theta$ cycles (Figure \@ref(fig:ThetaGammaScheme)) during the encoding and memory retrieval phases. Since local firing was found to be elevated during $\gamma$ [@bragin_gamma_1995; @colgin_frequency_2009; @soltesz_low-_1993], @huerta_bidirectional_1995 proposed that, when timed at the peak, nested action potential bursts during $\theta$ cycles can result in enhanced long term potentiation and, thereby, stabilise connections. Furthermore, a phenomenon referred to as "replay" -- the reactivation of cell assemblies actively coordinated during a previous task -- was found to be strongly linked to phases of $\theta$ cycles, where an increase in phase locking during $\theta$ oscillations was associated with higher performance of working memory [@fuentemilla_theta-coupled_2010; @lee_phase_2005]. An important implication of these findings is that disruption of $\theta$ can lead to a disruption in memory formation; interfering with $\theta$ generated in the medial septum by lesioning indeed impairs performance in spatial and working memory tasks [@gerashchenko_effects_2001; @poucet_septum_1990].

Today, more than 80 years after the first observation of $\theta$, its exact purpose is still not entirely understood, nor are important interactions between different cortical and subcortical areas sufficiently known.

## Medial Septum

The medial septum (MS) is a structure located in the forebrain (Figure \@ref(fig:MS-Scheme)) and, due to its rhythmic activity, is known to be a central $\theta$ generator [@petsche_significance_1962]. Because of its widespread connectivity and the oscillatory properties of medial septal cells, it is important for entraining $\theta$ and synchronising activity in different areas of the brain [@andersen_septo-hippocampal_1979; @brucke_beeinflussung_1959; @gogolak_firing_1968; @green_hippocampal_1954; @petsche_significance_1962; @rawlins_septo-hippocampal_1979]. Depending on the brain state, the majority of cells in the MS show a pronounced rhythmic firing pattern linked to hippocampal $\theta$ [@dragoi_interactions_1999; @serafin_rhythmic_1996].

(ref:MSSchemeShortCaption) Schematic of medial septum
(ref:MSSchemeCaption) **Schematic of medial septum.** Sagittal (A) and coronal (B) section of the mouse brain. Blue shows the location of the MS and dark grey fibre bundles. The dashed line in (A) indicates the position of the section of (B). Modified from Allen Brain Atlas [@sunkin_allen_2013].

```{r MS-Scheme, fig.cap = '(ref:MSSchemeCaption)', fig.scap='(ref:MSSchemeShortCaption)', fig.pos='H', fig.align='center', echo=FALSE, fig.height=3, fig.asp=0.4}
sagMS  <- grImport2::pictureGrob(grImport2::readPicture(rawToChar(rsvg::rsvg_svg("Figures/Schemes/MS_sagittal_clean.svg"))))
coronalMS  <- grImport2::pictureGrob(grImport2::readPicture(rawToChar(rsvg::rsvg_svg("Figures/Schemes/MS_coronal_clean.svg"))))
wrap_elements(gridExtra::arrangeGrob(sagMS), clip = T) + wrap_elements(gridExtra::arrangeGrob(coronalMS), clip=T) + plot_layout(widths = c(3.2,2)) &
  plot_annotation(tag_levels = "A") &
  theme(plot.tag = element_text(size=16*7.5/5.9))
```

\noindent
To understand differences between cells in the MS, a substantial amount of work has been put into further identifying diverse electrophysiological properties and different cellular markers in septal neurons [@luttgen_chemical_2004; @serafin_rhythmic_1996; @sotty_distinct_2003; @unal_synaptic_2015]. Several important proteins expressed by septal cells, such as hyperpolarisation-activated cyclic nucleotide-gated non-selective cation (HCN) channels or serotonin receptors, can change oscillatory behaviour and be modulated by different network conditions, depending on the brain state [@aznar_5-ht1a_2003; @colom_characterization_2005; @gaspar_somatostatin_1987; @koenig_spatial_2011; @luttgen_5-ht1a_2005; @varga_presence_2008].

GABAergic cells in the septum are known to be associated with the hippocampal $\theta$ and fire in bursts at different parts of the $\theta$ cycle [@serafin_rhythmic_1996], whereas cholinergic cells fire less rhythmically and exhibit slower firing rates [@pascale_simon_firing_2006]. The MS is known to project to the hippocampal and parahippocampal formation, including such target areas as the MEC and the PaS [@alonso_study_1984; @gaykema_cortical_1990; @raisman_connexions_1966; @van_groen_connections_1990]. These projections appear to depend on the neurotransmitter identity and preferentially target specific cell types [@desikan_target_2018; @fuchs_local_2016; @fuhrmann_locomotion_2015; @gonzalez-sulser_gabaergic_2014; @unal_synaptic_2015], leading to synchronicity and potential coordination of network activity across different brain regions. Interfering with this important synchronisation has decisive consequences for learning and memory [@chrobak_intraseptal_1989; @mizumori_reversible_1990; @winson_loss_1978]. For example, it has been shown that deactivation of the MS leads to a reduction in exploratory behaviour and impairs working memory [@brandon_reduction_2011; @koenig_spatial_2011; @lee_effects_1988; @poucet_septum_1990], making it an integral part of the memory system. Underlining the importance of the MS in hippocampal $\theta$ is the onset of cell firing in the MS, which occurs shortly before the onset of hippocampal $\theta$, suggesting that *in-vivo* the MS is a central part of $\theta$ rhythm generation [@bland_mechanisms_1999].

In this study, I will focus on GABAergic and cholinergic projections of the MS since they represent the major projecting cell types of the MS [@amaral_analysis_1985; @unal_synaptic_2015]. Both groups have been investigated in the past with a major focus on projections to the hippocampus or MEC. However, it should also be noted that the glutamatergic cells of the MS can project to other regions [@fuhrmann_locomotion_2015; @justus_glutamatergic_2017]. Early evidence indicated that other regions of the parahippocampal area are also innervated by the MS and could play important roles in $\theta$ generation [@swanson_connections_1979]. Yet, MS interactions with other regions of the parahippocampal network are rarely investigated even though they heavily project to the medial entorhinal cortex [@canto_all_2012]. Thus, I hypothesise that anatomically confirmed projections target specific cell types in the parasubiculum to entrain $\theta$ in the parahippocampal formation.

### GABAergic projections

The majority of the MS is made up of GABAergic cells [@freund_gabaergic_1989] which can be divided into several sub-classes [@hangya_gabaergic_2009] and are generally thought to fire rhythmically in $\theta$ bursts. Two major groups were identified by @joshi_behavior-dependent_2017. One was referred to as "Teevra" cells, a parvalbumin positive (PV^+^) cluster firing in strong bursts and targeting PV^+^ and cholecystokinin positive (CCK^+^) cells in the cornu ammonis 3 (CA3). However, Teevra cells were not found to project to the parahippocampal network, suggesting that they play a role in modulation of the hippocampus [@joshi_behavior-dependent_2017]. "Komal" cells, a second group of PV^+^ cells in the MS, fire preferentially at the peak of the $\theta$ phase, whereas Teevra cells are locked at the trough. A subset of Komal cells -- referred to as "Orchid" cells -- projects to the presubiculum and entorhinal cortex, but the majority of Komal projections have not yet been identified [@viney_shared_2018]. @varga_presence_2008 showed that PV^+^ cells expressing HCN channels are more likely to fire at the trough of hippocampal $\theta$, as Teevra cells, and are more likely to exhibit bursty behaviour. Other reports suggesting the involvement of HCN come from observations by @xu_hippocampal_2004, where injecting an HCN blocker into the MS lead to reduction of hippocampal $\theta$, further underlining innate cell properties contributing to the oscillatory behaviour of the MS.

GABAergic low-rhythmic-firing neurons (LRN), on the other hand, were recently shown to project to several areas of the hippocampus, make local connections in the MS and receive inhibitory feedback during hippocampal sharp-waves [@salib_gabaergic_2019]. @salib_gabaergic_2019 postulated that, indicated by molecular markers, PV^+^ LRNs might be more involved in MS local connectivity opposed to MS PV^-^ LRNs which target the dentate gyrus (DG) and CA3.

We know from other studies that targets of PV^+^ cells in the hippocampus or MEC are mainly interneurons which partially express PV themselves and, in the case of the MEC, are mainly located in layer II or III [@freund_gabaergic_1989; @fuchs_local_2016; @gonzalez-sulser_gabaergic_2014; @unal_synaptic_2015]. This target specificity suggests local disinhibition as a common mechanism of PV^+^ projections. Calbindin (CB) positive interneurons -- a separate group of GABAergic MS neurons -- have been shown to project to the MEC targeting low threshold interneurons in mice [@fuchs_local_2016]. However, it was also reported by @unal_synaptic_2015 that $15\%$ of CB^+^ cells in the rat MS express choline acetyltransferase (ChAT) -- an enzyme to synthesise acetylcholine -- suggesting that the same molecular markers might be expressed by distinct cell types with different neurotransmitters. Interestingly enough, to this day no cortical layer I interneurons have been reported to receive inputs from PV^+^ septal projections, highlighting the focus of research on layer II and III of cortical areas. Recently, several studies have provided more evidence that strong PV^+^ projections innervate the parasubiculum, indicating possible inhibition of local networks [@tang_functional_2016; @unal_synaptic_2015].

However, it is not clear whether these fibres have functional connections and to what extent they preferentially target parasubicular cells as compared to the better-documented MEC, or if they elevate disinhibition. This is an important question which could have implications for communication along the parahippocampal axis (e.g. increased synchronicity) and $\theta$ generation.

### Cholinergic projections

Cholinergic cells (ChAT^+^) in the MS are thought to be slowly firing [@manseau_hippocamposeptal_2008] and project to the hippocampus and MEC [@desikan_target_2018; @frotscher_cholinergic_1985; @lamour_septo-hippocampal_1984; @unal_synaptic_2015]. Similar to the PV^+^ cells in the MS, ChAT^+^ cells project mainly to distinct areas. For example, only a few cells (less than $2\%$) will project to the CA1 and the MEC simultaneously [@unal_synaptic_2015]. The majority of projection neurons, however, will specifically target one of the two areas. Cholinergic projections can work in different ways due to the properties of acetylcholine, which can slowly inhibit, slowly excite by binding muscarinic receptors or quickly excite a cell via nicotinic receptor binding. All these mechanisms have been shown to be present in cholinergic projections from the MS to the hippocampus or MEC.

In the entorhinal cortex, the majority of cells receiving nicotinic inputs were superficial layer I interneurons and layer II putative 5HT~3~R^+^ cells [@desikan_target_2018]. Hyperpolarisation, due to probable activation of type II muscarinic receptors, was observed in pyramidal cells, stellate cells and their intermediates as well as putative PV^+^ interneurons [@desikan_target_2018]. A similar pattern was observed in the lateral entorhinal cortex (LEC). It was also reported that cholinergic projections to the DG can inhibit granule cells by activating astrocytes which, in turn, will excite hilar interneurons and mossy cells through glutamate [@pabst_astrocyte_2016]. This suggests that an activation of cholinergic cells can result in complex interactions in the network. Indeed, when interfering with cholinergic septal cells using a chemogenetic approach to increase cell firing, $\theta$ frequency is reduced in the MEC but grid cell firing was not affected [@carpenter_modulating_2017]. In comparison, optogenetic activation of cholinergic MS cells leads to a reduced power, including in the $\theta$ frequency band and a slight increase in coherence in the hippocampus, both of which also seem to be state dependent [@vandecasteele_optogenetic_2014]. Taken together, these experiments show that cholinergic projections can have different targets and consequences in the parahippocampal formation. Therefore, it is important to collect more cell-specific data on how cholinergic MS cells interact with elements of the parahippocampal network involved in $\theta$ generation.

<!-- ### Glutamatergic projections  -->

<!-- Glutamergic projections were found in the hippocampus and target -->

<!-- @robinson_optogenetic_2016 showed *in-vitro* that glutamergic cells in the MS target local MS GABAergic interneurons but also some cholinergic interneurons. Measuring hippocampal cells revealed that only a small fraction of cells were functionally connected and which was in line with a very spars projection pattern. @robinson_optogenetic_2016 showed further that upon activation of these fibres *in-vivo* hippocampal $\theta$ could only be induced when stimulated in the MS or stimulating non-glutamergic fibres in the hippocampus. Similarly, @gonzalez-sulser_gabaergic_2014 reported very low connectivity rates for glutamergic responses in the MEC. These studies suggest a more local importance for glutamergic MS cells. In contrast @fuhrmann_locomotion_2015 reported functional connectivity to hippocampal interneurons with an increased connectivity rate compared to @robinson_optogenetic_2016 (4% vs 28%). Additionally they record *in-vivo* hippocampal LFP and stimulate glutamergic cells in the MS using channelrhodopsin. They report that after blocking local glutamergic connectivity in the MS by injecting the glutamate blocker NBQX into the septum optogenetic frequency responses could be strongly reduced, however, it also reduced the correlation between running speed and $\theta$ suggesting a role of glutamergic projections. Additionally, they find that upon light stimulation in the MS mice started to run during the experiment suggesting a role in locomotion similarly to what was reported for cholinergic septal cells [@vandecasteele_optogenetic_2014]. Similar findings by ###Justus in the MEC indicate that glutamergic projections to the MEC are also likelier than expected from previous data [@gonzalez-sulser_gabaergic_2014]. Not only do glutamergic cells project to MEC but imaging of axon terminals reveals a strong positive correlation between speed and fibre activity. However, glutamergic cells in the MS are not uniquely positively correlated with speed but represent a heterogeneous population ###Justus.  -->

## Parasubiculum

The parasubiculum (PaS) is part of the parahippocampal formation (Figure \@ref(fig:PaS-Scheme)), positioned at the posterior part of the brain [@boccara_three-plane_2015], and is found across different species [@ding_comparative_2013]. There is some disagreement as to whether it is a three-layered [@burgalossi_microcircuits_2011; @mulders_neuron_1997; @tang_functional_2016] or six-layered cortex [@boccara_grid_2010; @funahashi_presubicular_1997; @glasgow_local_2007] and it is sometimes plainly divided into superficial and deep layers [@sammons_electrophysiological_2019]. It sits adjacent to the MEC and presubiculum and, therefore, has an important position mediating inputs and outputs of the hippocampal formation [@kohler_intrinsic_1985; @swanson_autoradiographic_1977; @swanson_connections_1979; @van_groen_connections_1990]. In particular, connections from the PaS to the MEC have been found to target layer II, mainly exciting their postsynaptic partner cells including cells from other layers [@caballero-bleda_regional_1993; @canto_all_2012].

It has been shown that the PaS not only has an important role in working memory, but also in spatial memory [@kesner_neural_1998; @liu_excitotoxic_2001; @liu_excitotoxic_2004]. The relatively small size and orientation of the PaS rendered *in-vivo* recordings challenging, but they were recently made possible with the use of more recording sites and better targeting approaches. The resulting recordings underlined its role in the processing of spatial information and showed the presence of different spatially-linked functional cell types, including place cells [@taube_place_1995], grid cells [@boccara_grid_2010; @ebbesen_cell_2016] and head direction cells [@kornienko_non-rhythmic_2018]. One of the prominent features of PaS cells is the presence of strong rhythmic firing at $\theta$ frequency exhibited by pyramidal cells [@ebbesen_cell_2016; @tang_functional_2016]. These cells are known to express a protein called Wolframin ER Transmembrane Glycoprotein (WFS1), represent the largest proportion of cells in the PaS [@kitamura_island_2014; @luuk_distribution_2008; @ramsden_laminar_2015; @ray_grid-layout_2014; @sammons_electrophysiological_2019] and have an intrinsic oscillatory behaviour which can be modulated by acetylcholine [@glasgow_muscarinic_2013; @salkoff_synaptic_2015; @sammons_electrophysiological_2019].

An increased interest in the PaS has gained traction in recent years, providing the first evidence of MS GABAergic projections to the PaS [@tang_functional_2016; @unal_synaptic_2015]. This further emphasises the prime conditions for the PaS to play a role in the $\theta$ generation as part of the parahippocampal network. Therefore, it is important to understand the extent of MS projections to the PaS and the possible effect it has on $\theta$ modulation in the context of the parahippocampal formation.

(ref:PaSSchemeShortCaption) Schematic of parasubiculum
(ref:PaSSchemeCaption) **Schematic of parasubiculum.** A sagittal (A) and horizontal (B) section. Turquoise shows the location of the PaS, light grey represents ventricular space and dark grey fibre bundles. The dark green in (B) represents the pyramidal cell layer II of the medial entorhinal cortex. The cornu ammonis and the dentate gyrus are marked in light blue. (C) A horizontal section stained with NeuN, a selective neuronal marker to identify neurons (top). A high density of cells can be observed in the dentate gyrus and cornu ammonis 3. The bottom shows the corresponding WFS1 staining. WFS1^+^ pyramidal cells can be seen in the PaS, MECII, and CA1 of the hippocampus. Note also stained fibres in the stratum lacunosum-moleculare, resulting from MEC LII projections.

```{r PaS-Scheme, fig.cap = '(ref:PaSSchemeCaption)', fig.scap='(ref:PaSSchemeShortCaption)', echo=FALSE, fig.width=8, fig.asp=0.4}
horPaS  <- grImport2::pictureGrob(grImport2::readPicture(rawToChar(rsvg::rsvg_svg("Figures/Schemes/PaS_horizontal_clean.svg"))))
sagPaS  <- grImport2::pictureGrob(grImport2::readPicture(rawToChar(rsvg::rsvg_svg("Figures/Schemes/PaS_sagittal_clean.svg"))))
Schematic_PaS_hor_Image_Plot <- readRDS(file = "Data/Schematic_PaS_hor_Image_Plot.rds")
wrap_elements(gridExtra::arrangeGrob(sagPaS), clip = T) + wrap_elements(gridExtra::arrangeGrob(horPaS), clip=T) + Schematic_PaS_hor_Image_Plot + plot_layout(widths = c(3.2,1.5, 1.4)) & plot_annotation(tag_levels = "A") &
    theme(plot.tag = element_text(size=16*8/5.9))
```

## Aim of this study

In my thesis, I intend to characterise functional GABAergic and cholinergic MS projections to the PaS and identify their targets. Both types of projections have been shown to be important for $\theta$ entrainment and modulation in the MEC and hippocampus. However, up to this point it has been unclear which cell types in the PaS receive inputs from the MS and how they are modulated. Based on the literature and anatomical studies, it has been assumed that the PaS is strongly innervated by MS GABAergic projections, which might entrain local $\theta$ oscillations. However, most of the functional studies available were performed in the MEC, which is strongly innervated by the PaS. Whether or not connections in the PaS are similar to the MEC is uncertain. Hence, it is important to understand how the PaS is modulated by MS projections in order to understand the consequences for the parahippocampal formation.

To address these questions, we analysed functional connections in PV-Cre and ChAT-Cre animals from the MS to different cell types of the PaS. To probe whether interneurons were preferentially targeted -- as it was shown in the MEC -- we recorded *in-vitro* cells across all the layers of the PaS and measured synaptic responses induced by optical stimulation. Using the same approach, we screened connections in ChAT-Cre animals to identify cholinergic inputs and their effects on different cell types. Once targets of GABAergic PV^+^ MS projections were identified, we sought out to investigate $\theta$ entrainment in intact networks *in-vivo* using silicon probe recordings and optogenetic stimulation of the MS and projecting fibres.

<!-- ## Spatial Navigation -->

<!-- ### Theta -->