Structural development and dorsoventral maturation of the medial entorhinal cortex

We investigated the structural development of superficial-layers of medial entorhinal cortex and parasubiculum in rats. The grid-layout and cholinergic-innervation of calbindin-positive pyramidal-cells in layer-2 emerged around birth while reelin-positive stellate-cells were scattered throughout development. Layer-3 and parasubiculum neurons had a transient calbindin-expression, which declined with age. Early postnatally, layer-2 pyramidal but not stellate-cells co-localized with doublecortin – a marker of immature neurons – suggesting delayed functional-maturation of pyramidal-cells. Three observations indicated a dorsal-to-ventral maturation of entorhinal cortex and parasubiculum: (i) calbindin-expression in layer-3 neurons decreased progressively from dorsal-to-ventral, (ii) doublecortin in layer-2 calbindin-positive-patches disappeared dorsally before ventrally, and (iii) wolframin-expression emerged earlier in dorsal than ventral parasubiculum. The early appearance of calbindin-pyramidal-grid-organization in layer-2 suggests that this pattern is instructed by genetic information rather than experience. Superficial-layer-microcircuits mature earlier in dorsal entorhinal cortex, where small spatial-scales are represented. Maturation of ventral-entorhinal-microcircuits – representing larger spatial-scales – follows later around the onset of exploratory behavior. DOI: http://dx.doi.org/10.7554/eLife.13343.001


Introduction
The representation of space in the rodent brain has been investigated in detail. The functional development of spatial response properties has also been investigated in the cortico-hippocampal system (Ainge and Langston, 2012;Wills et al., 2014), with studies suggesting the early emergence of head-directional selectivity (Tan et al., 2015;Bjerknes et al., 2015), border representation (Bjerknes et al., 2014) and place cell firing, but a delayed maturation of grid cell discharges (Wills et al., 2010;Langston et al., 2010).
Even though there is information on the emergence of functional spatial properties in the hippocampal formation, remarkably little is known about the structural development of the microcircuits which bring about these properties. To understand this, we investigated the development of the architecture of the medial entorhinal cortex (MEC) and parasubiculum (PaS), two key structures in the cortico-hippocampal system.
In adult animals, layer 2 of MEC contains two types of principal cells, stellate and pyramidal cells (Alonso and Klink, 1993;Germroth et al., 1989). Stellate and pyramidal neurons are distinct in their intrinsic conductance (Alonso and Llinás, 1989;Klink and Alonso, 1997), immunoreactivity (Varga et al., 2010), projections (Lingenhö hl and Finch, 1991;Canto and Witter, 2012) and inhibitory inputs (Varga et al., 2010). Pyramidal neurons in layer 2 of MEC can be identified by calbindinimmuno-reactivity (Varga et al., 2010) and are clustered in patches across various mammalian species (Fujimaru and Kosaka, 1996;Ray et al., 2014;Naumann et al., 2016), while stellate cells can be identified by reelin-immuno-reactivity (Varga et al., 2010) and a lack of structural periodicity . In rodents, the grid-like arrangement of pyramidal cell patches is aligned to cholinergic inputs Naumann et al., 2016). Functionally, about a third of all cells in layer 2 exhibit spatial tuning with grid, border, irregular and head-directional discharges being present .
Neurons in layer 3 of MEC are characterized by rather homogenous in vitro intrinsic and in vivo spatiotemporal properties (Tang et al., 2015). A majority of cells exhibit a lack of spatial modulation, and the remaining are mainly dominated by irregular spatial responses (Tang et al., 2015) with a fraction also exhibiting grid, border and head-directional responses (Boccara et al., 2010).
The parasubiculum is a long and narrow structure flanking the dorsal and medial extremities of MEC (Video 1). The superficial parasubiculum, corresponding to layer 1 of MEC is divided into large clusters, while the deeper part, corresponding to layers 2 and 3 of MEC, is rather homogenous (Tang et al., 2016). In terms of functional tuning of cells, a majority of the cells of PaS show spatially tuned responses, and include grid, border, head-directional and irregular spatial cells (Boccara et al., 2010;Tang et al., 2016).
Here we investigate the emergence of the periodic pyramidal-cell patch pattern in layer 2 of MEC, as well as the development of cellular markers that characterize the architecture of adult MEC and PaS. The results indicate an early emergence of pyramidal cell organization, a delayed maturation of pyramidal but not stellate cells and a dorsal-to-ventral maturation of MEC circuits.

Results
We first investigated development of brain size and thickness of layers of the MEC (Figure 1) by observing rats at E18, P0, P4, P8, P12, P16, P20, P24 and adults (>P42). The majority of the brain development takes place within the first few weeks postnatally (Figure 1a), with the brain size increasing 1000% from 0.12 ± 0.00 g at E18 (mean ± SD; n=3) to 1.23 ± 0.07 g at P12 (n=5). Subsequently, the growth plateaus to~25% with the brain weighing 1.71 ± 0.08 g at P24 (n=6) and having a weight of 2.11 ± 0.14 g in adults (n=9) (Figure 1b). The superficial layers (layers 1-3) of the MEC (Figure 1c) double in thickness during this early postnatal period from 243 ± 35 mm at P0 (mean ± eLife digest Many animals, from rats to humans, need to navigate their environments to find food or shelter. This ability relies on a kind of memory known as spatial memory, which provides a map of the outside world within the animal's brain. Specifically, cells in a part of the brain called the medial entorhinal cortex act like the grids present on a map, and are known as grid cells. Other cells in this region represent boundaries in the environment and are known as border cells. These cells and other cells connect to each other to make the spatial memory circuit.
Previous research had reported that the grid cells were not present in the very early stages of an animal's life. It was also not clear how the different cell types involved in spatial memory develop after birth. Ray and Brecht have now studied rats and found that certain characteristic structures in the circuit are present at birth. For example, cells that were most likely to become grid cells, were already laid out in a grid, indicating that this layout is instructed by genetic information rather than experience.
Ray and Brecht also found that the cells that most likely become grid cells matured later than the cells that most likely become border cells. Further analysis then revealed that the circuits in the top part of the medial entorhinal cortex, which represents nearby areas, matured earlier than those in the bottom part of this region, which represent farther areas. These findings could therefore explain why rats explore nearby areas earlier in life before going on to explore further away areas at later stages.
More work is needed to characterize other components of the neural circuits involved in spatial memory to provide a complete understanding of how these memories are formed. Future experiments could also ask if encouraging young rats to explore a wider area can cause the circuits to mature more quickly. SD; n=21, 4 rats) to 652 ± 50 mm at P12 (n=24, 4 rats). A similar increase is also observed in the deeper layers (layers 4-6) from 167 ± 21 mm at P0 (n=21, 4 rats) to 329 ± 54 mm at P12 (n=24, 4 rats).The overall thickness plateaus around this point to 981 ± 81 mm at P12 (n=24, 4 rats) and remains at 882 ± 78 mm in adults (n=24, 4 rats) ( Figure 1d). Proportionally, the thickness of the layers remains similar during development, with layer 2 accounting for~20% and layers 3 and 5/ 6 each accounting for~30% of the MEC. Layers 1 and 4 are the thinnest at about 10% and 5% of the total thickness respectively ( Figure 1d).
We next investigated the microcircuit organization of superficial layers of MEC. Calbindin, a calcium binding protein, is selectively expressed in layer 2 pyramidal cells (Varga et al., 2010;Fujimaru and Kosaka, 1996), which form a gridlike arrangement in adult animals . Concurrently, reelin, an extracellular matrix protein, is selectively expressed in stellate cells in layer 2 of MEC, which are scattered throughout  layer 2. To visualize the development of entorhinal microcircuits we first prepared tangential sections (see our video animation on preparing tangential sections, Video 1) through layer 2 of medial entorhinal cortex and stained for calbindin-immunoreactivity. From the earliest postnatal stages, calbindin+ neurons in the MEC exhibited clustering, forming patches at P0 (Figure 2a). The calbindin+ patches at P0 exhibited a grid-like (Figure 2a Figure 3a). However, the calbindin+ patches in the MEC did not exhibit clustering of their dendrites, as previously described in adults  at E18 and P0 (Figure 3a,b). Some dendritic clustering could be observed at P4 (Figure 3c), while from P8 (Figure 3d-h) the dendritic clustering of calbindin+ pyramidal neurons was similar to that in adults. In layer 3 of the MEC, we observed a transient presence of calbindin expression. The number of calbindin+ neurons in layer 3 declined progressively from prenatal stages to P20 (Figure 3a-g), where it attained adult-like levels with rarely any calbindin+ neurons in layer 3 ( Figure 3h). Quantitatively, calbindin+ neuronal density (calbindin+ neurons per mm 2 ) decreased from 955 ± 315 (mean ± SD; count refers to n=3776 neurons in 8 rats) in P4-P8 rats to 333 ± 99 (n=2104 neurons, 8 rats) in P12-P16 rats to 141 ± 56 (n=828 neurons, 7 rats) in adults ( Figure 3i).
A closer analysis of the co-localization of the immature neuronal marker doublecortin with calbin-din+ pyramidal cells and reelin+ stellate cells (Figure 7a-c) revealed doublecortin to be mostly colocalized with calbindin+ rather than reelin+ neurons (Figure 7d). Spatial cross-correlations between doublecortin and either calbindin or reelin (Figure 7e; n=8 rats from ages P8 -P20) from tripleimmunostained calbindin, reelin and doublecortin regions of layer 2 of the MEC revealed a greater overlap of doublecortin with calbindin (0.54 ± 0.10) than with reelin (0.08 ± 0.13). This difference in the Pearson's cross correlation coefficient was significant at p=0.0009 (Mann-Whitney two tailed). Third, wolframin expression, a marker which co-localizes with calbindin+ pyramidal neurons in layer 2 of MEC in adult rodents (Kitamura et al., 2014), develops from dorsal to ventral in layer 2 medial entorhinal cortex and parasubiculum ( Figure 8). Specifically, wolframin expression starts to appear in the dorsal MEC and the dorsal PaS shortly after birth ( Figure 8a) and is present only in the dorsal~10% of the PaS. It extends progressively more ventrally (Figure 8b) and covers~40% at P8 and~75% at P12 of PaS. At P20 it is expressed throughout the full extent of medial entorhinal cortex and the parasubiculum (Figure 8c).

Discussion
Neurogenesis in the medial entorhinal cortex is completed prior to E18 (Bayer, 1980a;1980b), and at this time the basic laminar organization of medial entorhinal cortex is already evident. While the basic structure of medial entorhinal cortex appears early, we observe massive developmental The clustering of layer 2 MEC calbindin+ neurons into patches is also an early developmental event, and key aspects of the grid-layout of calbindin+ neurons are already present at birth. This observation indicates that the periodic structure of patches is a result of genetic signaling rather than spatial experience. Periodic patterns are ubiquitous in nature, and several chemical patterning systems have been explained on the basis of interaction between dynamical systems (Turing, 1952). Since it has been suggested that the grid layout of calbindin+ neurons is functionally relevant for grid cell activity , it would be interesting to investigate, whether genetic manipulations would result in changes of layout periodicity and have functional effects. The dendritic clustering of calbindin+ pyramidal neurons is similar to dendritic development in the neocortex (Petit et al., 1988) and is established by the end of the first postnatal week. The cholinergic innervation of the calbindin+ patches was present by P4 in line with other long-range connectivity patterns in the MEC (O'Reilly et al., 2015), which are also established early in development.
Reelin is an important protein in cortical layer development (D'Arcangelo et al., 1995) and in the early stages of postnatal development we see the strongest reelin expression in layer 1, where reelin secreting Cajal-Retzius cells are involved in radial neuronal migration (Pesold et al., 1998). Stellate cells in layer 2 of MEC, which can be visualized by reelin-immunoreactivity (Varga et al., 2010), were scattered  throughout postnatal development.
Layer 3 of the MEC features a complementary transition of calbindin+ and reelin+ neurons during the first couple of postnatal weeks. While the density of reelin+ neurons increases, there is a concurrent decline in calbindin+ neuronal density in layer 3 of MEC, though part of the calbindin+ neuronal density decline can be attributed to the increasing brain size. Taken together with the presence of radial neuronal migration promoting Cajal-Retzius cells in layer 1 during this period, it would be An interesting observation is the presence of clusters of neurons in the parasubiculum, which transiently express calbindin in early postnatal stages, and subsequently express wolframin. Transient expression of calbindin has been observed in early postnatal development in the neocortex (Hogan and Berman, 1993) and midbrain regions (Liu and Graybiel, 1992), but its functional significance remains largely unknown. Our data show, however, that at early developmental stages the parasubiculum and medial entorhinal cortex share a similar organization in calbindin+ patches. Additionally, the expression of wolframin in the parasubiculum persists in adults, while calbindin+ neurons in MEC layer 2 also exhibit wolframin (Kitamura et al., 2014) from the end of the first postnatal week. Current studies generally focus on cell-type specific investigations using proteins expressed by these cells. However, investigations to study the specific roles of these proteins (Li et al., 1995) might provide interesting insights towards understanding the finer differences in the functionalities exhibited by these cells. For instance, calbindin is a calcium buffer, and reduces the concentration of intracellular calcium (Mattson et al., 1991), while wolframin is implicated in increasing intracellular calcium levels (Osman et al., 2003). With the medial entorhinal cortex and parasubiculum having many similarities in their spatial discharge properties Boccara et al., 2010;Tang et al., 2016), a structure-function comparison of the wolframin+/transiently-calbindin+ neurons in the parasubiculum and the wolframin+/ permanently-calbindin+ neurons in the medial entorhinal cortex would be worthwhile.
A dorsal-to-ventral development profile was observed in the superficial layers of the MEC and parasubiculum. This conclusion was suggested by the progressive disappearance of the calbindin expression in layer 3 from dorsal to ventral; the progressive disappearance of doublecortin expression in layer 2 and parasubiculum from dorsal to ventral; and the progressive appearance of the wolframin expression in superficial layer 2 of MEC and parasubiculum from dorsal to ventral. Homing behavior in rats, as well as spontaneous exploratory behavior develops around the end of second postnatal week (Wills et al., 2014;Bulut and Altman, 1974) while spontaneous exploration of larger environments outside the nest emerge towards the end of the third postnatal week (Wills et al., 2014). This is coincident with the timeline of maturation of calbindin+ patches in the dorsal and ventral MEC respectively. Since the dorsal MEC represents smaller spatial scales and the ventral MEC progressively larger scales (Hafting et al., 2005;Stensola et al., 2012), these data may indicate that the rat's navigational system matures from small to large scales. Early eyelid opening experiments have indicated an accelerated development of spatial exploratory behaviour (Kenny and Turkewitz, 1986;Foreman and Altaha, 1991), and similar experiments might provide insights into whether early behavioral development is accompanied by an accelerated development of the microcircuit underlying spatial navigation.
The higher co-localization of doublecortin with calbindin+ pyramidal cells than reelin+ stellate cells, supports further the dichotomy of structure-function relationships exhibited by these two cell types Tang et al., 2014). Grid and border cells have been implicated to be largely specific to pyramidal and stellate cells respectively and the delayed structural maturation of pyramidal cells might reflect the delayed functional maturation of grid cells (Wills et al., 2010;Langston et al., 2010), with the converse being applicable to stellate and border cells (Bjerknes et al., 2014). The divergent projection patterns of pyramidal and stellate cells, with the former projecting to CA1 (Kitamura et al., 2014) and contralateral MEC (Varga et al., 2010) and the latter to dentate gyrus (Varga et al., 2010;Ray et al., 2014) and deep layers of MEC (Sürmeli et al., 2015), have differing theoretical interpretations in spatial information processing.
The same sets of neurons, which correspond to grid and border cells , have also been implicated to be differentially involved in temporal association memory (Kitamura et al., 2014) and contextual memory (Kitamura et al., 2015) respectively. An underlying differential structural maturation timeline of the microcircuit governing these processes may also translate into a differential functional maturation profile of these memories.
We conclude that the structural maturation of medial entorhinal cortex can be coarsely divided into an early appearance of the calbindin+ neuron patches and a progressive cell-type specific refinement of the cellular structure, which proceeds along the dorsal to ventral axis.

Materials and methods
All experimental procedures were performed according to the German guidelines on animal welfare under the supervision of local ethics committees (LaGeSo) under the permit T0106-14.

Brain tissue preparation
Male and female Wistar rats (n=83) from E18 to P24 and adults (>P42) were used in the study. The ages were accurate to ± 1 day. Animals were anaesthetized by isoflurane, and then euthanized by an intraperitoneal injection of 20% urethane. They were then perfused transcardially with first 0.9% phosphate buffered saline solution, followed by 4% formaldehyde, from paraformaldehyde, in 0.1 M phosphate buffer (PFA). For prenatal animals, pregnant rats at E18 were perfused in the aforesaid manner and the E18 animals were then extracted from the uterus. Subsequently, brains were removed from the skull and postfixed in PFA overnight. Brains were then transferred to 10% sucrose solution for one night and subsequently immersed in 30% sucrose solution for at least one night for cryoprotection. The brains were embedded in Jung Tissue Freezing Medium (Leica Microsystems Nussloch, Germany), and subsequently mounted on the freezing microtome (Leica 2035 Biocut) to obtain 60 mm thick sagittal sections or tangential sections parallel to the pia.
Tangential sections of the medial entorhinal cortex were obtained by separating the entorhinal cortex from the remaining hemisphere by a cut parallel to the surface of the medial entorhinal cortex (Video 1). For subsequent sectioning the surface of the entorhinal cortex was attached to the block face of the microtome.
Immunohistochemical stainings were performed according to standard procedures. Briefly, brain sections were pre-incubated in a blocking solution containing 0.1 M PBS, 2% Bovine Serum Albumin where, r is the cross-correlation between the monochromatic images f1 and f2 without smoothing. n is the number of pixels in the image. The Pearson's cross-correlation coefficient can vary from -1 (anti-correlated) through 0 (un-correlated) to 1 (correlated).
For analysis of dorso-ventral variation in overlap between doublecortin with calbindin, two regions of the same size were selected from a section double-stained for calbindin and doublecortin. One region was selected from the dorsal half of the section and another from the ventral half and the regions were represented as pairs. Where, due to section damage, it was not possible to obtain regions from both dorsal and ventral parts, the data was presented as unpaired.
For analysis of variation in overlap between doublecortin and calbindin/reelin, comparisons were performed between the same regions from a section triple stained for calbindin, reelin and doublecortin.