Available online at www.sciencedirect.com

ScienceDirect The development of the hippocampal neural representation of space Thomas J Wills1 and Francesca Cacucci2 The hippocampal formation (HF) contains a neural representation of the environment, based on the activity of several classes of neurons whose firing is tuned to an animal’s position and orientation in space. Recently, work has begun on understanding when and how this neural map of space emerges during development. Different classes of spatially tuned neurons emerge at different ages, some of them very early during development, before animals have started exploring their environment. The developmental timeline thus far uncovered has yielded insights into both the mechanisms of the ontogeny of the neural code for space, as well as how this system functions in the adult. Addresses 1 Department of Cell and Developmental Biology, University College London, Gower Street, London WC1E 6BT, UK 2 Department of Neuroscience, Physiology and Pharmacology, University College London, Gower Street, London WC1E 6BT, UK Corresponding authors: Wills, Thomas J ([email protected]) and Cacucci, Francesca ([email protected])

Current Opinion in Neurobiology 2014, 24:111–119 This review comes from a themed issue on Neural maps Edited by David Fitzpatrick and Nachum Ulanovsky

0959-4388/$ – see front matter, Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.conb.2013.09.006

Introduction The rodent hippocampal formation (HF) contains several classes of neurons whose firing is tuned to an animal’s position and orientation in space, and which together constitute a neural ‘cognitive map’ [1] of the environment, by means of which an animal can locate itself and navigate to a goal. Four of these cell types, and their anatomical localisations, are illustrated in Figure 1: place cells, which code for a unique position in space [2]; head direction (HD) cells, which code for the current head direction of the animal [3]; grid cells, which fire in regularly repeating locations and may provide a distance metric [4] and boundary vector/border cells, which respond to barriers to movement [5,6]. Despite an intense research effort towards understanding how this neural code operates in the adult rat, until recently very little was known about how this neural map of space is assembled during development. During www.sciencedirect.com

the first three weeks of post-natal life pronounced developmental changes take place in the rat HF at the cellular and physiological level [7–9]. Similarly, a repertoire of spatial behaviours emerges between two and three weeks of age: rats develop the capability for independent movement at around post-natal day 13 (P13); followed by a drive to actively explore beyond their nest at P15–19; and from P20 onwards show learning on hippocampus-dependent spatial memory tasks (for a review, see [10]). This article will review what is known about the development of the hippocampal neural code for space, focusing on work that ourselves and others have done mapping the emergence of spatially responsive neurons in the HF of immature rats, as well as the development of hippocampal network synchrony and oscillations. We will highlight what we have learned about the principles of hippocampal development, and how development has informed us about the functioning of this system in the adult.

HD cells: the earliest spatial signal to emerge The most striking aspect of the development of spatially tuned firing is the very early emergence of adult-like HD signalling. By P16, sub-populations of neurons in both pre-subiculum and the medial entorhinal cortex (mEC) are tuned to fire in a tight range of heading angles, as in the adult [11,12], see Figure 2a. The preferred direction of firing is stable between two sessions in the recording arena (separated by approx. 15 min), indicating that the HD system is capable of maintaining a directional ‘reference memory’ across at least this time span. HD cells are adult-like before rats have extensively explored their environment [10], suggesting that some aspects of their development might be experience-independent. Consistent with this hypothesis, there is no correlation between the amount of experience in the recording arena and the quality of HD cell firing, and furthermore, HD cells appear to fire in an adult-like fashion from the very first exposure to the recording arena onwards, see Figure 2b [11]. The early maturation of the HD system, in comparison to the other spatially tuned cells investigated so far (place and grid cells, see below) also suggests that it may be a ‘primary’ spatial signal, not dependent on the input of any other hippocampal spatial cells for its development, possibly reflecting the origin of the HD signal outside the HF (see Figure 1b). This would also be consistent with the Current Opinion in Neurobiology 2014, 24:111–119

112 Neural maps

The development of grid cell firing

Figure 1

(a)

9.2

14.9

17.0

4.1

Key

(b)

Head Direction Cells Place Cells Grid Cells BVC/Border Cells

mEC Para DG CA3 CA1

Pre

ADN LMN DTN

Sub Current Opinion in Neurobiology

Spatially tuned neurons and their anatomical localisation in the hippocampal formation (HF). (a) Examples of (from left to right): a place cell; a head direction (HD) cell; a grid cell; a boundary vector cell(BVC). For HD cells, the polar plot represents the firing rate (action potentials/ seconds of dwell time) for each directional heading, the peak firing rate is shown at the top-left corner. For all other cell types, the false-colour firing rate map represents an overhead view of the recording arena. When the rat visited those positions shown in hot colours, the firing rate was high. Peak rate (action potentials/seconds of dwell time) is shown in the top-left corner. Grid cell example thanks to Caswell Barry, BVC to Colin Lever. (b) Schematic of anatomical structures discussed in this review, showing the major connections and direction of information flow between them. See [64] for an extensive anatomical description of the HF. Colours represent the localisation of spatially tuned neurons: green, HD cells; red, place cells; blue, grid cells; brown, boundary vector/ border cells. Those structures within the green dashed-line ellipse lie outside of the HF, but represent the principal pathway for the generation of the HD cell signalling and its transmission to the HF, see [65]. Abbreviations: ADN, antero-dorsal nucleus of the thalamus; CA, cornu ammonis; DG, dentate gyrus; DTN, dorsal tegmental nucleus; mEC, medial entorhinal cortex; LMN, lateral mammillary nucleus; Para, parasubiculum; Pre, pre-subiculum; Sub, subiculum.

observation that lesions to [13], or temporary inactivation of [14] the hippocampus (where place cells are located) have little impact on HD cell firing in adults. The early emergence of HD cells may also have an important adaptive function. Adult rodents can use the integration of internal cues derived from selfmotion (or ‘path integration’) to return to a home base in the absence of landmark cues [15], and the accuracy of this behaviour is correlated to HD cell responses in adult rats [16]. Pre-weanling rats also use path integration to return to their littermates during their first exploratory forays in life [17], therefore accurate HD cell signalling may be a pre-requisite in order for pups to commence exploration and return successfully to their nest. Current Opinion in Neurobiology 2014, 24:111–119

The adult mEC contains both HD and grid cells, potentially encoding the two different spatial parameters (direction and distance) thought to be necessary to support path integration [18]. It is noteworthy, therefore, that there is a marked delay between the developmental emergence of these two cell types: the very earliest age at which stable, adult-like grid cell firing is consistently observed is P20 [11,12], see Figure 2c, roughly a week after adult-like HD cells can be recorded. This corresponds to an age when exploratory behaviour is present, and animals are starting to show learning in hippocampus-dependent spatial memory tasks [10]. Before this age, some spatially localised firing can be observed in the mEC, including some multi-peaked firing fields. However, this spatial firing is not arranged with the regularity and hexagonal symmetry of adult grid cells, and, just as importantly, the location of the firing fields is not stable, either within or between recording sessions [11,12], and therefore could not support path integration. Other adult-like characteristics of mEC firing, such as the existence of ‘conjunctive’ (combined grid and HD tuning) cells, and speed-modulation of grid cell firing [18], also emerge at around P20 [19], suggesting a relatively rapid emergence of an adult-like grid cell network.

The development of grid network connectivity Adult grid cells are arranged in functional ‘modules’ [4,20,21]: within each module all grid cells share the same wavelength and orientation, and the relative spatial phases of the grid fields remains fixed, even though the absolute position of their firing peaks (phase) can change [22]. This coherent spatial structure is indicative of grid cells (or their inputs) being recurrently connected with a pattern of excitation and inhibition whereby cells excite those other cells with similar spatial phases, and inhibit those with dissimilar phases. This finding has served as important supporting evidence in favour of continuous attractor models of grid cell network structure [23,24,25,26]. A recent study found no evidence of direct excitatory connections between mEC layer II stellate cells (putative grid cells), but did find disynaptic inhibitory connections between them: such connections could create a network with a continuous attractor structure [27]. When does this network connectivity emerge during development? It appears to be present concurrently with the first stable grid cells: simultaneously recorded grid cells share the same wavelength and orientation already at P20 [11]; furthermore, when pups are transferred between different recording arenas, absolute spatial phases change, but all relative spatial phases are preserved, as in the adult [19,22], see Figure 2d,e. In light of this, it is also interesting that increased network synchrony in the mEC in vitro emerges from P22 onwards, roughly coincident www.sciencedirect.com

Development of the neural representation of space Wills and Cacucci 113

with the emergence of grid cells in vivo [12], see also [27]. These data suggest that a stable grid cell network emerges during development as a coherent unit, and that recurrent connectivity might be a pre-requisite of stable grid firing. A further test of whether adult-like recurrent connectivity exists in young pups would be whether the relative spatial offsets between grid cells’ firing are preserved even when individual grid fields are deformed and less symmetric, such as when rats encounter a novel environment for the first time [28], as is the case in adult rats [29]. Interestingly, whilst recurrent connectivity in the grid cell circuit was traditionally a feature of continuous attractor models of grid cell firing, several groups working on the other major class of grid cell model (‘oscillatory interference’ models, see ‘Oscillations and timing’, below) have also incorporated this type of recurrent connectivity into their models [30,31; Bush & Burgess, abstract P2-D118 at British Neuroscience Association 2013]. Importantly, whilst the existence of recurrent connectivity between grid cells is consistent with continuous attractor models, it still does not speak to the existence of the additional, necessary component of these models, namely a set of asymmetrically weighted connections matching the animal’s speed and direction of movement which shift neural activity in line with changes in the animal’s position, see [24], for example, for further details. Modelling the development of the grid cell network has so far received less attention than modelling the adult state of the system. It has been proposed that recurrent connectivity could be created intrinsically, through synchronous waves of coordinated neural activity [24]. However, this assumes a topographical mapping between grid cell phase and anatomical position at some point in development, for which there is currently no evidence. Other attempts at modelling the emergence of grid firing posit that the required network connectivity could be learnt during the initial exploration of the environment, and therefore be experience-dependent. In these models, the learning process depends on some pre-existing spatially tuned input, either place cells [32; Widloski & Fiete, abstract 100.4/KKK20 in Soc. Neuro. Abstr 2010], or repeating unidirectional bands [33], see also [34]. It should be noted that an analogous model of HD cell development has been proposed, according to which the connections in a continuous attractor network are learnt using an instructive input from stable visual landmarks [35]. In the adult mEC, a topographic mapping exists between the anatomical position of grid cells along the dorso-ventral axis, and the scale of space that is encoded by their grid fields; grid cells belonging to modules with short wavelengths lie towards at the dorsal tip of the mEC, and those with long wavelengths lie more www.sciencedirect.com

ventrally (though there can be considerable anatomical overlap between different modules) [4,20]. Little is currently known about which anatomical features underlie this modular functional arrangement, or how they emerge during development [20]. During development, almost all grid cells recorded between P20 and P30 were of a similar, short, wavelength (approximately 30 cm) [11,12]. It is tempting to speculate that short wavelength grid modules develop first, allowing the animal to map out small-scale space at the beginning of its independent life, and that the neural code for larger spaces develops later, as rats become capable of moving over larger distances.

Place cells The maturation of the CA1 place cell network follows a slow and gradual trajectory. Spatially localised, stable place fields exist as early as P16, but the quality of spatial tuning and stability continue to improve until much later in life, at least P45 [11,12,36,37], see Figure 2f. This developmental programme contrasts with that of HD and grid cells, which appear to emerge in adult-like form relatively abruptly (within a few days). However, it is important to stress that, at any given age, the CA1 region will contain a heterogeneous set of place cells at different developmental stages [36], and that fully adult-like place cells can be seen as early as P16 [11], Figure 2g. Similarly to HD and grid cells, the maturity of place cell firing is not correlated with the amount of experience of the recording arena [11], suggesting that the differences between cells are due to intrinsic factors rather than experiencedependent learning. The existence of adult-like place responses as early as P16, in the absence of grid cell firing in the mEC, was the first demonstration that grid cells are not a necessary input for place cell firing (see also [38,39] for further evidence in adult rats). The question then arises as to what are the inputs driving early place cell firing. One intriguing possibility is that this input is provided by cells which fire in response to the presence of boundaries at a specific distance and direction from the animal, such as border cells in mEC [5] or boundary vector cells in the subiculum [6]. Preliminary evidence indicates that such responses can be recorded early during development, at least as early at P17 in both mEC and subiculum [Bjerknes et al., abstract 702.09/ DDD25 in Soc. Neurosci. Abstr 2012; Cacucci et al., abstract in Soc. Neurosci. Abstr 2013]. If boundaryresponsive cells are providing the main input to place cells early in development, it should be possible to detect differences in place cell firing near to boundaries, as opposed to in the middle of an open field, in young pups. In support of this hypothesis, preliminary evidence indicates that in pre-weanling pups, the spatial stability of place field position is correlated with Current Opinion in Neurobiology 2014, 24:111–119

114 Neural maps

Figure 2

(a)

Trial 1

Trial 2

3.0

5.2

2.1

1.7

Trial 1 6.7

7.8

12.3

12.8

P16, first exposure to recording arena.

(b)

Trial 2

0.37

0.68

1.69

2.3

2.1

2.6

5.1 0.52

P16

(c)

Adult

Trial 1

1.1

0.51

1.3

4.4

Trial 2

Trial 1

2.1

Trial 2

6.7

7.4

7.5

7.7

Trial 1

Trial 2

4.7

9.0

4.2 3.3

5.5

8.4

0.9

P16-17

P20-21

(d)

Recording Arena ‘A’

Recording Arena ‘B’

5.5

2.1

6.3

6.4

4.5

4.8

3.2

(e)

Adult

Recording Arena ‘A’

Recording Arena ‘B’

Cell 1

Cell 2 Cell 1

Cell 3

Cell 2

Cell 3

P22

(f)

Trial 1

(g) Trial 1

Trial 2

Trial 1

Trial 2

Trial 1

Trial 2

1.4

5.7

2.1

1.7

9.2

10.2

1.6

1.1

3.5

3.9

11.6

13.1

2.6

1.8

2.1

6.6

3.9

5.7

6.2

2.1

2.5

2.3

6.7

Cell 1

Cell 2

P16P17 1.7

Trial 2

1.5

11.1

Cell 3

P16-17

P20-21

Adult

Cell 4

Current Opinion in Neurobiology

The development of spatially responsive firing in rat HF neurons. (a) Three HD cells recorded from mEC at P16 (left), and from adults (right). Each row represents one cell, each column a separate recording trial (separated by 15 min). Note the similarity between P16 and adult HD cells, in terms of directional selectivity and the stability of the preferred direction of firing. (b) A HD cell recorded at P16, during the first exposure to the recording arena. Current Opinion in Neurobiology 2014, 24:111–119

www.sciencedirect.com

Development of the neural representation of space Wills and Cacucci 115

the distance of the field from the nearest wall of the recording arena [Cacucci et al., abstract in Soc. Neurosci. Abstr 2013]. The boundaries of an environment may therefore provide a foundational input for hippocampal place cells, consistent with the boundary vector cell model [40], which originally proposed that environment boundaries are a critical determinant of place cell firing.

Oscillations and spike timing The rodent HF exhibits several prominent synchronised and rhythmic signals, which are related to the concurrent behaviour of the animal [41]. As discussed below, these signals all develop early, before animals are capable of self-locomotion, suggesting that they may be a pre-requisite for the development of spatially tuned firing. The earliest adult-like activity in the local field potential (LFP) are large, slow deflections which reverse polarity across the CA1 pyramidal cell layer, seen from P2 onwards [42,43]. These are consistent with the Large Irregular Activity (LIA) state of the adult hippocampus, seen during rest and immobility [41], and may also correspond to the Giant Depolarising Potentials [44] seen in immature hippocampal slices in vitro [42]. These large, synchronous activity bursts are thought to be required for the development of network connectivity in the HF [45]. The 140–200 Hz fast oscillations (‘ripples’), which accompany LIA sharp waves in the CA1 layer [1,46], emerge later, at around P7 [43]. In the adult, movement and exploratory behaviour are accompanied by a 7–12 Hz (theta) oscillation in the hippocampal LFP [41], which is first observed during development at P8 [47,48]. The frequency of theta is lower in the youngest pups (5 Hz), and gradually increases to adult levels [11,12,47,48]. Gamma oscillations (30–100 Hz), which co-occur with theta in the adult HF,are first observed at P2, and increase sharply in power after P8 [48]. Immature rats spend a large proportion of their time in a behavioural state known as ‘active sleep’ characterised by twitches of the limbs, head and eyes [49]. Active sleep muscle twitches are associated with neuronal firing in the hippocampus,

and with the first bouts of theta and gamma oscillations between P2 and P8 [48], see Figure 3a, highlighting the importance of peripheral input from the sensorymotor system in HF development [50]. The firing of adult place cells is modulated at theta frequency, but at a slightly higher frequency than the ongoing LFP theta, with the result that the timing of spikes, relative to LFP theta phase, changes as animals traverse each place field (‘phase precession’) [51]. As the theta phase of spiking correlates better with distance travelled through the field than with other variables, it may provide a temporal code for space, in addition to the rate code provided by the place field [52]. Theta-band modulation and phase precession are features of grid cell, as well as place cell, firing in rodents, and these phenomena originally inspired the ‘oscillatory interference’ models of spatial firing [30,31,53,54]. The common theme of these models is that the frequency difference between two theta-band oscillators, one of whose frequency is modulated by velocity, serves to track the distance travelled by the animal and thereby generate spatially stable firing fields. During development, both grid and place cell firing are consistent with these models: both are modulated within the theta band, at a slightly higher frequency than the LFP theta, at the earliest ages at which these spatial responses can be observed [11,12], see Figure 3b,c. Phase precession appears at least as early as P17 in place cells [11,12], indicating that this temporal aspect of spatial coding, in the hippocampus, is adult-like during earliest exploration (see Figure 3d). Intriguingly, recent evidence shows that, during laboratory foraging tasks, some species of bats show very little theta modulation in the HF, both in the LFP and in neuronal spiking. They nevertheless display the same spatially tuned neurons as those described in rodents [55,56], at odds with the idea that these responses would be generated via theta-band oscillators (but see [57]). A cross-species, comparative approach may therefore be fruitful to fully understand the development of spatial coding in mammals.

(Figure 2 Legend Continued ) The central firing rate map shows all data from the recording trial, the four firing rate maps at the corners show the directional firing filtered by which quadrant of the recording arena (in 2D space) the rat was occupying. As for adult HD cells, the cell always codes for the heading direction of the animal, regardless of 2D position. (c) mEC cells recorded at P16–17, P20–21 and from adult rats. Each row shows a cell, each pair of columns a trial. For each trial, the left column shows the firing rate map, the right column a spatial auto-correlogram of the firing rate map, which highlights hexagonal grid structure. Note the absence of stable and regularly symmetrical grid firing fields at P16–17. Note also that, for grid cells only, adult recording arenas are larger than those for immature rats: rate maps are shown to scale. (d, e) Simultaneously recorded grid cells have a coherent spatial structure, which is maintained in all environments. (d), left column, shows three grid cells simultaneously recorded at P22. All cells share the same wavelength (spacing between fields) and orientation. In a different environment (right column, ‘Recording Arena B’), the absolute position of the grid fields changes, but the relative offsets between grid fields remains the same, as highlighted in the cartoon version of the same data (e). Coloured circles indicate correspondence between cells in (d) and (e). (f) Three place cells recorded at P16–17, P20–21 and from adult rats. Note the gradual improvement in the specificity of spatial tuning, and the stability of the place field position. (g) Four examples of adult-like place cells, recorded at P16–17. All data adapted from [11] except (d), adapted from [19]. www.sciencedirect.com

Current Opinion in Neurobiology 2014, 24:111–119

116 Neural maps

Figure 3

P8

(a)

CA1, P17

(d)

theta

12.4 (4-14 Hz)

Cell 1

(20-100 Hz) 100 µV

EMG

θ-phase (˚)

gamma 300 200 100 0

0

25

50

75

100

125

150

125

150

Position (cm) 500 ms

3.9 (b)

Place cell, CA1, P17

Cell 2

0µV −150µV

2 sec

θ-phase (˚)

150µV 300 200 100 0

0

25

50

75

100

Position (cm) (c)

Grid cell, mEC, P20 150µV 0µV −150µV

2 sec Current Opinion in Neurobiology

The development of theta-band oscillations and modulation of spike timing in the rat HF. (a) Hippocampal LFP recorded from a P8 pup, showing examples of theta (top) and gamma (middle) band oscillations, occurring concurrently with nuchal muscle EMG activity (bottom). Adapted from [48] with permission. (b, c) Theta-band LFP oscillations and spike-time modulation at the earliest appearance of place cells (b; P17) and grid cells (c; P20). The black ticks show spike times, the blue traces above show simultaneously recorded LFP. Note that spiking is modulated at approximately the same frequency as LFP theta oscillations but spikes systematically move to an earlier phase of the LFP theta oscillation during a burst of firing. (d) Phase precession in two place cells recorded in a linear running track at P17. For each cell, the top row shows a false-colour firing rate map, the red arrow shows the direction of running (data include runs of one direction only). The scatter plot, bottom, shows the relationship between the rat’s position when spiking occurred (x-axis) and LFP theta phase (y-axis). Note that spiking systematically shifts to an earlier theta phase as the rat runs through the field. Adapted from [11].

Conclusions The development of neural systems relies on a variety of different mechanisms, including molecular cues, intrinsic neural activity and environmental experience [58– 61]. Studying the developmental timeline of spatially tuned neurons (Figure 4) has shown which spatial responses occur before the first active exploration of the environment, and may therefore not require this experience (e.g. HD cells) and those which instead emerge later and may therefore require experience-dependent learning (e.g. grid cells). Further work will be required to understand how these different mechanisms interact during development. Current Opinion in Neurobiology 2014, 24:111–119

Studying the development of the hippocampus has also yielded important insights into the organisation of spatial firing in the adult rat. One clear example is the finding that adult-like place cell firing precedes the emergence of grid cell firing during development by at least one week, demonstrating that other types of input are sufficient to support place cell responses during this developmental window. Further studies are required to substantiate the intriguing first indications that place fields may be built on the foundations of responses to the borders of the environment. If this were the case, it would represent a striking confirmation, and extension, of the boundary vector cell theory, which was proposed even before the www.sciencedirect.com

Development of the neural representation of space Wills and Cacucci 117

Figure 4

(a)

P0 Sensory Motor

Righting

P4

P12

P8

Whiskers Moving

P20

P24

P28

Ears open Eyes open Crawling

Pivoting

P16

Exploratory Behaviour Hippocampusdependent Behaviour

Walking Open field exploration Watermaze T-maze T-maze Hidden Delayed Spontaneous Platform Alternation Alternation

(b)

P16

P20

P24

P28

Head Direction Cells Grid Cells Place Cells Current Opinion in Neurobiology

Schematic summary of the development of spatial behaviour (a) and spatially tuned firing in the HF, in the rat (b). In (a), the black dots represent the approximate age of emergence of behavioural or sensory-motor function. For further detail on the development of spatial behaviour, see [10]. The coloured bars in (b) represent the maturation of spatially responsive neurons: HD cells are fully mature by P16; grid cells emerge abruptly after P20; place cells mature gradually throughout this period, and are still not adult-like by P28. Summarising data from [11,12,36].

experimental description of border and boundary vector responses in the adult HF [40]. The developmental pattern of grid cell emergence also speaks to the ongoing debate regarding how the grid cell signal is generated in adults. At present, the developmental data argue for the importance of elements from both the continuous attractor and oscillatory interference classes of models (recurrent connectivity and theta-band modulation of spiking, respectively), and is consistent with a current generation of models that include both of these elements [26,30,31]. We would hope that the experimental study of the development of spatially tuned neurons in the HF will also encourage theoreticians to incorporate biologically plausible developmental mechanisms into their models. A striking aspect of the development of the neural map of space is the dissociation between the developmental programmes of different cell types. Grid cell networks emerge abruptly (over a small number of days), in an allor-none fashion [11,12,19]. CA1 place cells, by contrast, exhibit a protracted development (over the course of weeks), and different cells mature at different ages [11,12,36]. The underlying reasons for this dissociation are not currently understood. Most physiological and anatomical characteristics of the CA1 are mature by three weeks of age [7,8,62,63]; therefore there is no obvious candidate mechanism to explain the protracted functional development of place cell firing. Furthermore, little is known of the anatomical and physiological www.sciencedirect.com

development of the parahippocampal regions where grid cells are found. One speculative hypothesis would be that abrupt or gradual developmental programmes are determined by the network architectures underlying these systems in the adult brain: the recurrent attractor network of the grid cells might necessarily form in an ‘all-or-none’, abrupt fashion, whereas in the CA1, the lack of extensive recurrent connections means that different place cells can mature independently, at different ages. Finally, it is also tempting to speculate on the functional links between the development of spatial firing and spatial behaviour: as discussed, the emergence of adult-like HD firing precedes the onset of active exploration, possibly supporting directional path integration during the crucial first forays outside the nest; whilst it is also intriguing that the onset of hippocampus-dependent spatial learning coincides with the emergence of grid cells (for further discussion, see [10]). We believe that the studies outlined in this review provide the first steps towards an integrated understanding of the development of spatial cognition, from the neural circuit to the behavioural level, and will inspire further work in this field.

Acknowledgements The authors thank Neil Burgess and Caswell Barry for helpful comments on the manuscript, and wish to acknowledge funding from the following sources: The Royal Society (University Research Fellowship to TW); the European Research Council (grant no. 282091, ‘DEVSPACE’, to FC); the Biotechnology and Biological Sciences Research Council, UK (grant no. BB/ I021221/1 to FC). Current Opinion in Neurobiology 2014, 24:111–119

118 Neural maps

References and recommended reading Papers of particular interest, published within the period of review, have been highlighted as:  of special interest  of outstanding interest

19. Wills TJ, Barry C, Cacucci F: The abrupt development of adultlike grid cell firing in the medial entorhinal cortex. Front Neural  Circuits 2012, 6:21. A more detailed examination of the emergence of grid cells, showing that important characteristics of adult-like firing emerge at P20–P22. Also includes data from mEC cells as they make the developmental transition to becoming grid cells. 20. Stensola H, Stensola T, Solstad T, Froland K, Moser MB, Moser EI: The entorhinal grid map is discretized. Nature 2012, 492:72-78.

1.

O’Keefe J, Nadel L: The Hippocampus as a Cognitive Map. Oxford, UK: Oxford University Press; 1978, .

2.

O’Keefe J, Dostrovsky J: The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely-moving rat. Brain Res 1971, 34:171-175.

3.

Taube JS, Muller RU, Ranck JB Jr: Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. J Neurosci 1990, 10:420-435.

4.

Hafting T, Fyhn M, Molden S, Moser MB, Moser EI: Microstructure of a spatial map in the entorhinal cortex. Nature 2005, 436:801-806.

5.

Solstad T, Boccara CN, Kropff E, Moser MB, Moser EI: Representation of geometric borders in the entorhinal cortex. Science 2008, 322:1865-1868.

6.

Lever C, Burton S, Jeewajee A, O’Keefe J, Burgess N: Boundary vector cells in the subiculum of the hippocampal formation. J Neurosci 2009, 29:9771-9777.

24. McNaughton BL, Battaglia FP, Jensen O, Moser EI, Moser MB:  Path integration and the neural basis of the ‘cognitive map’. Nat Rev Neurosci 2006, 7:663-678. Comprehensive review of the functional principles of continuous attractor networks, as they apply to grid and HD cell firing. Proposes a mechanism for grid cell network development based on intrinsic activity early in development.

7.

Dumas TC: Late postnatal maturation of excitatory synaptic transmission permits adult-like expression of hippocampaldependent behaviors. Hippocampus 2005, 15:562-578.

25. Burak Y, Fiete IR: Accurate path integration in continuous attractor network models of grid cells. PLoS Comput Biol 2009, 5:e1000291.

8.

Danglot L, Triller A, Marty S: The development of hippocampal interneurons in rodents. Hippocampus 2006, 16:1032-1060.

9.

Leinekugel X: Developmental patterns and plasticities: the hippocampal model. J Physiol Paris 2003, 97:27-37.

26. Navratilova Z, Giocomo LM, Fellous JM, Hasselmo ME, McNaughton BL: Phase precession and variable spatial scaling in a periodic attractor map model of medial entorhinal grid cells with realistic after-spike dynamics. Hippocampus 2011, 22(4):772-789.

10. Wills TJ, Muessig L, Cacucci F: The development of spatial behaviour and the hippocampal neural representation of space. Philos Trans R Soc Lond B Biol Sci 2013. (in press). 11. Wills TJ, Cacucci F, Burgess N, O’Keefe J: Development of the  hippocampal cognitive map in preweanling rats. Science 2010, 328:1573-1576. This study maps the developmental emergence of HD, place and grid cells. By staggering the starting age of testing, the authors were also able to dissociate developmental changes from effects of experience. 12. Langston RF, Ainge JA, Couey JJ, Canto CB, Bjerknes TL,  Witter MP, Moser EI, Moser MB: Development of the spatial representation system in the rat. Science 2010, 328:1576-1580. This study also maps the development of spatially tuned neurons in the HF. The authors also showed that mEC network synchrony, when tested in vitro, develops during the same developmental period. 13. Golob EJ, Taube JS: Head direction cells and episodic spatial information in rats without a hippocampus. Proc Natl Acad Sci U S A 1997, 94:7645-7650. 14. Bonnevie T, Dunn B, Fyhn M, Hafting T, Derdikman D, Kubie JL, Roudi Y, Moser EI, Moser MB: Grid cells require excitatory drive from the hippocampus. Nat Neurosci 2013, 16:309-317. 15. Mittelstaedt H, Mittelstaedt M-L: Homing by path integration. In Avian Navigation. Edited by Papi F, Wallraff HG. New York: Springer; 1982:290-297. 16. Valerio S, Taube JS: Path integration: how the head direction signal maintains and corrects spatial orientation. Nat Neurosci 2012, 15:1445-1453. 17. Loewen I, Wallace DG, Whishaw IQ: The development of spatial  capacity in piloting and dead reckoning by infant rats: use of the huddle as a home base for spatial navigation. Dev Psychobiol 2005, 46:350-361. Demonstrates that rat pups can use path integration to return to their nest, even during the very earliest exploratory forays. 18. Sargolini F, Fyhn M, Hafting T, McNaughton BL, Witter MP, Moser MB, Moser EI: Conjunctive representation of position, direction, and velocity in entorhinal cortex. Science 2006, 312:758-762.

Current Opinion in Neurobiology 2014, 24:111–119

21. Barry C, Hayman R, Burgess N, Jeffery KJ: Experiencedependent rescaling of entorhinal grids. Nat Neurosci 2007, 10:682-684. 22. Fyhn M, Hafting T, Treves A, Moser MB, Moser EI: Hippocampal remapping and grid realignment in entorhinal cortex. Nature 2007, 446:190-194. 23. Fuhs MC, Touretzky DS: A spin glass model of path integration in rat medial entorhinal cortex. J Neurosci 2006, 26:4266-4276.

27. Couey JJ, Witoelar A, Zhang SJ, Zheng K, Ye J, Dunn B,  Czajkowski R, Moser MB, Moser EI, Roudi Y, Witter MP: Recurrent inhibitory circuitry as a mechanism for grid formation. Nat Neurosci 2013, 16:318-324. In vitro study investigating the nature of connectivity between mEC layer II stellate cells (putative grid cells). Authors report the absence of direct excitatory connections between these cells, but demonstrate the existence of disynaptic inhibitory connections, and suggest that these may also support a continuous attractor network. This pattern of connectivity emerges after P16. 28. Barry C, Ginzberg LL, O’Keefe J, Burgess N: Grid cell firing patterns signal environmental novelty by expansion. Proc Natl Acad Sci U S A 2012, 109:17687-17692. 29. Yoon K, Buice MA, Barry C, Hayman R, Burgess N, Fiete IR: Specific evidence of low-dimensional continuous attractor dynamics in grid cells. Nat Neurosci 2013, 16:1077-1084. 30. Zilli EA, Hasselmo ME: Coupled noisy spiking neurons as velocity-controlled oscillators in a model of grid cell spatial firing. J Neurosci 2010, 30:13850-13860. 31. Blair HT, Gupta K, Zhang K: Conversion of a phase- to a ratecoded position signal by a three-stage model of theta cells, grid cells, and place cells. Hippocampus 2008, 18:1239-1255. 32. Welinder PE, Burak Y, Fiete IR: Grid cells: the position code, neural network models of activity, and the problem of learning. Hippocampus 2008, 18:1283-1300. 33. Pilly PK, Grossberg S: Spiking neurons in a hierarchical self organizing map model can learn to develop spatial and temporal properties of entorhinal grid cells and hippocampal place cells. PLoS ONE 2013, 8:e60599. Network model proposing that input from repeating unidirectional bands could combine to form grid cells during development. 34. Krupic J, Burgess N, O’Keefe J: Neural representations of location composed of spatially periodic bands. Science 2012, 337:853-857. 35. Hahnloser RH: Emergence of neural integration in the head direction system by visual supervision. Neuroscience 2003, 120:877-891.

www.sciencedirect.com

Development of the neural representation of space Wills and Cacucci 119

Network model of HD cell development, proposing that the connectivity underlying continuous attractor networks could be learnt by combining vestibular signals with input from a stable visual landmark. 36. Scott RC, Richard GR, Holmes GL, Lenck-Santini PP:  Maturational dynamics of hippocampal place cells in immature rats. Hippocampus 2010, 21(4):347-353. Shows that place cells continue to develop even in post-weanling rats (P22–P43), with considerable within- and across-animal variability in the maturity of spatial firing. 37. Martin PD, Berthoz A: Development of spatial firing in the hippocampus of young rats. Hippocampus 2002, 12:465-480. 38. Koenig J, Linder AN, Leutgeb JK, Leutgeb S: The spatial periodicity of grid cells is not sustained during reduced theta oscillations. Science 2011, 332:592-595. 39. Zhang SJ, Ye J, Miao C, Tsao A, Cerniauskas I, Ledergerber D, Moser MB, Moser EI: Optogenetic dissection of entorhinal– hippocampal functional connectivity. Science 2013, 340:1232627. 40. Hartley T, Burgess N, Lever C, Cacucci F, O’Keefe J: Modeling place fields in terms of the cortical inputs to the hippocampus. Hippocampus 2000, 10:369-379. 41. Vanderwolf CH: Hippocampal electrical activity and voluntary movement in the rat. Electroencephalogr Clin Neurophysiol 1969, 26:407-418. 42. Leinekugel X, Khazipov R, Cannon R, Hirase H, Ben Ari Y,  Buzsaki G: Correlated bursts of activity in the neonatal hippocampus in vivo. Science 2002, 296:2049-2052. Demonstrates the presence of multi-unit bursts coinciding with slow LFP deflections, consistent with the LIA state in the adult, at P4–6. This activity was dependent on GABA and glutamate, consistent with Giant Depolarizing Potentials (GDPs) observed in vitro. 43. Mohns EJ, Karlsson KA, Blumberg MS: Developmental emergence of transient and persistent hippocampal events and oscillations and their association with infant seizure susceptibility. Eur J Neurosci 2007, 26:2719-2730. 44. Ben-Ari Y, Cherubini E, Corradetti R, Gaiarsa JL: Giant synaptic potentials in immature rat CA3 hippocampal neurones. J Physiol 1989, 416:303-325. 45. Ben-Ari Y: Developing networks play a similar melody. Trends  Neurosci 2001, 24:353-360. Focusing on the HF, this review proposes that network synchrony and the depolarising action of GABA early in development play a critical role in the organisation of neural circuits.

concurrently with bouts of whole-body muscular activity during active sleep. 49. Gramsbergen A, Schwartze P, Prechtl HF: The postnatal development of behavioral states in the rat. Dev Psychobiol 1970, 3:267-280. 50. Mohns EJ, Blumberg MS: Neocortical activation of the  hippocampus during sleep in infant rats. J Neurosci 2010, 30:3438-3449. Shows that muscle twitches during active sleep drive neocortical spindle bursts, which in turn drive neuronal bursting in the hippocampus. Severing cortical input pathways to the hippocampus selectively decreases hippocampal twitch-related bursting. 51. O’Keefe J, Recce ML: Phase relationship between hippocampal place units and the EEG theta rhythm. Hippocampus 1993, 3:317-330. 52. Huxter J, Burgess N, O’Keefe J: Independent rate and temporal coding in hippocampal pyramidal cells. Nature 2003, 425:828832. 53. O’Keefe J, Burgess N: Dual phase and rate coding in hippocampal place cells: theoretical significance and relationship to entorhinal grid cells. Hippocampus 2005, 15:853-866. 54. Burgess N, Barry C, O’Keefe J: An oscillatory interference model of grid cell firing. Hippocampus 2007, 17:801-812. 55. Yartsev MM, Ulanovsky N: Representation of three-dimensional space in the hippocampus of flying bats. Science 2013, 340:367-372. 56. Yartsev MM, Witter MP, Ulanovsky N: Grid cells without theta oscillations in the entorhinal cortex of bats. Nature 2011, 479:103-107. 57. Barry C, Bush D, O’Keefe J, Burgess N: Models of grid cells and theta oscillations. Nature 2012, 488:E1-E2. 58. Katz LC, Shatz CJ: Synaptic activity and the construction of cortical circuits. Science 1996, 274:1133-1138. 59. White LE, Fitzpatrick D: Vision and cortical map development. Neuron 2007, 56:327-338. 60. Inan M, Crair MC: Development of cortical maps: perspectives from the barrel cortex. Neuroscientist 2007, 13:49-61. 61. Cang J, Feldheim DA: Developmental mechanisms of topographic map formation and alignment. Annu Rev Neurosci 2013, 36:51-77.

46. Buzsaki G, Horvath Z, Urioste R, Hetke J, Wise K: High-frequency network oscillation in the hippocampus. Science 1992, 256:1025-1027.

62. Pokorny J, Yamamoto T: Postnatal ontogenesis of hippocampal CA1 area in rats. II. Development of ultrastructure in stratum lacunosum and moleculare. Brain Res Bull 1981, 7:121-130.

47. Leblanc MO, Bland BH: Developmental aspects of hippocampal electrical activity and motor behavior in the rat. Exp Neurol 1979, 66:220-237.

63. Pokorny J, Yamamoto T: Postnatal ontogenesis of hippocampal CA1 area in rats. I. Development of dendritic arborisation in pyramidal neurons. Brain Res Bull 1981, 7:113-120.

48. Mohns EJ, Blumberg MS: Synchronous bursts of neuronal  activity in the developing hippocampus: modulation by active sleep and association with emerging gamma and theta rhythms. J Neurosci 2008, 28:10134-10144. Shows that the first developmental emergence of theta and gamma rhythms between P2 and P12, as well as HF neuronal spiking, occurs

64. van Strien NM, Cappaert NL, Witter MP: The anatomy of memory: an interactive overview of the parahippocampal– hippocampal network. Nat Rev Neurosci 2009, 10:272-282.

www.sciencedirect.com

65. Taube JS: The head direction signal: origins and sensorymotor integration. Annu Rev Neurosci 2007, 30:181-207.

Current Opinion in Neurobiology 2014, 24:111–119