Development of the Thalamocortical Interactions: Past, Present and Future

Abstract

For the past two decades, we have advanced in our understanding on the mechanisms implicated in the formation of brain circuits. The connection between the cortex and thalamus has deserved much attention, as thalamocortical connectivity is crucial for sensory processing and motor learning. Classical dye tracing studies in wild type and knock-out mice initially helped to characterize the developmental progression of this connectivity and revealed key transcription factors involved. With the recent advances of technical tools to specifically label subsets of projecting neurons, knock-down genes individually and/or modify their activity, the field has gained further understanding on the rules operating in thalamocortical circuit formation and plasticity. In this review, I will recapitulate the most relevant discoveries that have been made on this field, from development to early plasticity processes. I will emphasize how the implementation of new tools has helped the field to progress and what I consider to be the perspective and open questions for the future.

Introduction

The thalamocortical system has been widely used as a model to study basic axon guidance mechanisms, to decipher the factors involved in the development of cortical maps, to understand how sensory systems develop, and to study the mechanisms involved in anatomical and functional circuit plasticity upon sensory loss. The reason for this great interest is because the thalamus has unique anatomical and functional specificities. Anatomically, TCAs conform a large ipsilateral projection, which develops embryonically crossing distinct anatomical boundaries. TCAs are highly topographically organized within a tract conveying sensory maps into the cortical territories. And finally, regarding their function, TCAs play a central role in sensory processing and perception as carry information received from the periphery to the specialized cortical areas. The application of novel techniques to this field, such as the use of conditional mouse lines or the in utero electroporation, has greatly contributed over the past years to boost our understanding on the role of TCAs in cortical development and sensory function (Figure 1).

Figure 1. Developmental timeline in the thalamocortical research.

Timeline representing the major discoveries in the field of thalamocortical development over the past 40 years. The major techniques used to achieve those findings are highlighted. The panels below emphasised four key findings that were made in the last few years. (1) The mechanism that allow rostral and intermediate TCAs to topographically arrange depend on interaction of guidance cues and receptors at the corridor cells. (2) The velocity of axonal growth in TCAs is controlled by activity-dependent gene expression. (3) The specification of a cortical area directly depends on the subtype of thalamic input that receives, being the higher-order mode the default signature. Moreover, this thalamic influence depends on neurotransmitter release from thalamic neurons. (4) Prenatal thalamic neurons are able to generate calcium waves of spontaneous activity among sensory nuclei that influence cortical areas size and plasticity.

Here, I will give a perspective of the major discoveries made for the thalamocortical system and then review how technical advances have rapidly helped the field to dissect the role of intrinsic versus extrinsic, mainly thalamic, mechanisms involved in the development of sensory cortical areas.

Axon guidance to the cortex

The ability to trace axons in fixed embryonic tissue offered by the carbocyanine dye technique, pioneered the field by the end of the last century. Using these anterograde and retrograde tracers to label axons and cells, several authors described for the first time the timing and trajectory of the thalamocortical connectivity and associated projections [1–8]. Briefly, thalamic axons exit the thalamus, traverse the pre-thalamus and avoid the hypothalamic region to cross the diencephalic-telencephalic boundary to the subpallium. Once in the subpallium, they cross distinct anatomical territories, interact with other developing axons, and cross the pallial-subpallial boundary to enter the neocortex where they arrive around E15.5 in mice (reviewed in [9–12]).

After the first anatomical description of the pathway followed by the TCAs, a new wave of discoveries was initiated in the late 90’s. By using full knock-out mice for genes expressed in the origin, along the pathway or in the target structure, several transcription factors were identified to be involved in thalamocortical pathfinding such as Emx2, Pax6, Tbr1, Gbx2, Nkx2.1 or Ebf1 [13–15]. Moreover, axon guidance molecules were also implicated. To cross the diencephalic-telencephalic boundary (DTB), TCAs need a Slit-mediated repulsion from the hypothalamus [16–18]. Once in the subpallium, TCAs are attracted toward the corridor, a dynamic region at the mantle of the medial ganglionic eminence (MGE) populated by GABAergic neurons originated at the lateral ganglionic eminence (LGE) [19]. This cellular bridge provides a permissive environment for TCAs to cross the MGE and to arrive to the cortex through the action of members of the Nrg1 family of guidance molecules. The position of these corridor cells at the mantle of the MGE is controlled by the midline repellent Slit2 [20]. It is believed that a fine control of the location of these corridor cells during evolution contributed to the switch from an external to an internal path of the TCAs that is characteristic in mammals [20].

In addition to the corridor cells, other cells provide guidance information to TCAs. Neurons located at the prethalamic, perireticular and internal capsule regions, with early projections to the thalamus, also influence TCA pathfinding at the DTB and at subpallial levels [21–23]. In fact, in mutant mice in which internal capsule projecting cells are displaced, TCAs do not cross the DTB [15]. In addition, the role of striatal cells is gaining importance in the guidance of TCAs at the subpallium, as many of their projections are located in the vicinity of TCAs. The modification of the position of these striatal cells or their elimination directly leads to consistent TCAs pathfinding errors [24]. Moreover, these cells might influence TCAs trajectory indirectly by modifying the position of corridor cells in a Frizzled-dependent manner [25].

The guidance of TCAs is also influenced by early pioneer cortical cells located in the subplate, a deep transient cortical cell population that projects to the thalamus [3]. It was proposed more than 15-years ago that the correct formation of the thalamo-cortical reciprocal connectivity relies on the early interaction between subplate axons and TCAs, the so called ‘handshake hypothesis’ formulated by Molnar and Blakemore [26]. By performing tracing studies the authors showed the intimate anatomical apposition of cortical axons and TCAs at the pallial subpallial boundary (PSPB) suggesting that these axons help each other’s guidance [22, 27]. However, this hypothesis remained for decades under debate as other studies showed that these axons only interdigitate in some portions of their trajectory or even repel each other in vitro [28–30].

After those years, the field entered in the era of analysing mutant mice in which the thalamocortical connectivity was affected by the lack of specific transcription factors. However, those studies did not clarify whether the handshake hypothesis was right, as results were overall contradictory. Some of the mutant mice did show deficits in TCAs pathfinding while in others, either the development of corticofugal axons was normal or both corticofugal axons and TCAs were severely affected [15, 31–34]. Conditional mutagenesis recently shed light into this issue and supported the subapallial interaction between both axonal tracts as a pre-requisite for their correct development. Specific genetic elimination of corticofugal axons without affecting the thalamus or any other subpallial structure leads to a scenario in which TCAs do not progress through the PSPB and thus, do not reach the cortex [35]. Actually, the opposite genetic experiment was done in parallel and showed that TCAs are also required in vivo to guide corticothalamic axons into the corridor and thalamus [36], altogether supporting the handshake hypothesis at least in a particular stage and region at the subpallium.

Acquiring Topography

One of the most important aspects for the correct function of the thalamocortical system is that its connectivity must be well arranged topographically. TCAs from the distinct principal sensory-modality nuclei target primarily modality-matched cortical areas. The dorsolateral geniculate nucleus (dLGN) projects to the primary visual cortex (V1), the ventromedial posterior (VPM) to the somatosensory area (S1) and the ventro-medial geniculate nucleus (vMG) to the primary auditory area (A1). This topography is acquired embryonically and at two consecutive subpallial structures, the corridor and the striatum [10, 11, 37]. Gradients of axon guidance molecules such as Netrin-1, Slit1, or Ephrin-A5 are expressed in register in both structures and play a role in the fine topographical positioning of TCAs [38–42]. However, a specific chemoattractive response is achieved at the corridor. By combining in vitro chemotaxis assays with the analysis of TCAs defects in specific mutant mice, it was determined that the rostral positioning of TCAs is achieved at the corridor by a modulation of Netrin-1 responsiveness by Slit1 [41] (Figure 1). Rostral TCAs specifically express a novel Robo1 co-receptor, FLRT3, which is a member of the Fibronectine leucine-rich repeat transmembrane proteins. By a series of in vitro biochemical assays, it was shown that in the presence of Slit1, both Robo1 and FLRT3 receptors are required to induce Netrin-1 attraction, via the upregulation of DCC receptor at the plasma membrane [42]. Thus, TCAs expressing FLRT3 can modulate Netrin-1 responsiveness in a context-dependent manner in developing TCAs.

Anyhow, it is more than likely that the specific expression of axon guidance receptors in the distinct TCAs subpopulations is genetically determined by unique transcription factors alone or in combination. For example, Lhx2, a transcription factor of the LIM homeodomain family is expressed by all early postmitotic thalamic neurons but downregulated in a region-specific manner in a late postmitotic stage [43]. While rostral and intermediate thalamic neurons switch off Lhx2, this transcription factor remains active only thalamic neurons from the auditory nucleus [43]. When Lhx2 was ectopically upregulated in rostral or intermediate thalamic nuclei, TCAs from those territories fail to reach their corresponding cortical areas and abnormally acquire an auditory fate targeting A1 [43]. These experiments suggest that there must be a transcriptional code that determines the fate of thalamocortical neurons from the distinct sensory nuclei. In fact, the development of the genome-wide analysis has helped to test this possibility through an unbiased approach. Using a genetic dual labelling method in mice, Gezelius and colleagues purified distinct sensory thalamic neurons and revealed their transcriptome profiling. Genes such as Sp9, Hs6st2, Crabp2 or Cck, were found to be expressed at specific sensory nuclei during embryonic development [44]. Complementary to this, other studies using new generation sequencing revealed the transcriptional profile of the distinct sensory thalamic nuclei at early postnatal stages [45].

The final topography of TCAs is also influenced by intracortical mechanisms. After several decades of intense debate, it is well accepted that the positional identity of cortical areas occurs independently of the thalamus, as generally precedes the thalamocortical innervation [46]. However, as we will recapitulate in the next sections, TCAs influence the final anatomical and functional organization of cortical territories at perinatal-early postnatal stages. Anyhow, intracortical mechanisms also guide TCAs to specific cortical territories influencing their final topographical disposition. The adaptation of the in utero electroporation technique as a successful gene delivery system in the mouse embryo allowed to produce gene manipulation in a time- and region-specific manner [47–49]. When the regional identity of the subplate cells was shifted out of register to the respective cortical plate by the in utero overexpression of Fgf8, TCAs tracked this shift by producing ectopic arbors in the “matching” subplate area [50]. These results strongly suggested the existence of positional information in the subplate that guides TCAs to the correct cortical territory, and matched previous data describing that TCAs must interact with subplate cells to recognize cortical areas [27, 51–54]. Specific elimination of subplate cells in S1 of the cat, showed that corresponding TCAs neither enter in S1 nor in any remaining sensory area with the intact subplate. Therefore, subplate cells are an essential cortical check-point for the spatial guidance of TCAs. It is likely that this positional information might depend on activity-based competition among TCAs, as thalamic axons interact with subplate cells through functional synapses [52, 53, 55–57]. Anyhow, TCAs have a remarkable capacity to reorganize themselves as was shown to occur in the Semaphorin6A mutant mice. In this mouse, visual TCAs are profoundly derailed embryonically provoking a topographical shift in the targeting of somatosensory axons to V1 [58]. Postnatally, the misrouted dLGN axons reach the V1 through alternative routes and re-establish a normal pattern of thalamocortical connectivity [58].

Controlling axon growth and branching

As aforementioned, TCAs are guided toward their correct targets by the interaction with axon guidance molecules and /or intermediate targets. Arriving on time to meet specific target cells is indispensable to achieve a correct functional circuit. Spontaneous activity, which is present in almost all the brain structures at embryonic stages, has been shown to influence early developmental processes such as neurotransmitter specification, axon pathfinding or axonal targeting [59–64]. Spontaneous activity in the form of calcium activity in developing neurons and axon growth cones has gained much attention in the last years. The development of novel techniques for imaging intracellular calcium ion concentration has helped to identify calcium as a second messenger inside the cell that mediates a wide spectrum of cellular functions. Calcium spikes are produced at distinct frequencies by embryonic thalamic neurons and in fact, is in this parameter where it is encoded the transcriptional regulation of genes that modulate TCAs growth rate (Figure 1). When TCAs traverse the subpallium their growth rate is increased and decelerates when they reach to the cortex [65]. This change in axonal growth rate is controlled by a developmental decrease in the frequency of calcium spikes at the cell soma [65]. The activity-dependent regulation of TCAs growth occurs via the transcriptional regulation Robo1 and Dcc genes. Robo1 acts as a brake while Dcc functions as an accelerator for growth [65, 66].

Spontaneous neural activity is also influential for the process of axonal branch formation in the thalamocortical system. Pioneer work by Herrmann and Shatz in the mid 90’s showed that when the sodium channel blocker tetrodotoxin (TTX) was infused into the kitten visual cortex fewer branches of visual TCAs were formed in V1[67]. Such activity-dependent branch formation was corroborated later using organotypic co-cultures assays first [68–70], and subsequently by specific genetic manipulation in vivo [71–74].

Influencing cortical territories, area identity and size

Although the patterning of cortical areas is grossly achieved by intrinsic cortical mechanisms, thalamocortical input has a critical role in shaping and delineating cortical areas later in development, both anatomically and functionally. As mentioned before, each cortical area is reciprocally connected with a distinct subset of thalamic neurons which are allocated in well-defined nuclei. Primary cortical areas have typical cytoarchitectonical features, especially S1 and V1, and thus, have been extensively used to understand the role of thalamocortical input in shaping cortical differentiation. Within the sensory areas, there is a topographical representation of the sensory periphery on the cortical surface. This point-to-point representation on the cortical field is due to the fact that neighbouring neurons respond to the activation of neighbouring peripheral receptors.

One of the first demonstrations of the impact of the peripheral input on the development of sensory cortical areas came from seminal studies in the early 70’s. When the infraorbital nerve, which conveys input from the whiskers to S1 was sectioned at birth, the representation of the injured whiskers in the barrel field was impaired or disappeared producing a concomitant increase in barrels representing the intact whiskers [75–77]. Importantly, manipulation in later stages did not lead to map defects [78] providing strong evidence that an intact periphery, before barrel pattern becomes visible, is a basic condition for a development of a normal map in S1. However, these studies were not dissecting out whether the relevant influencing factor for the development of cortical maps was the periphery itself or the neural activity that flows through the system. Moreover, the changes in cortical maps could be indirect and due to alterations in subcortical territories such as the thalamus.

The development of conditional loss-of-function mouse models in which a specific gene is deleted in cortical or thalamic neurons helped to dissect out the role of the thalamocortical input into the process of cortical map formation. Genetic manipulations of thalamic structures, such as the decrease in size of the dLGN or the ablation of the VPM nuclei, confirmed the importance of the thalamic input as an extrinsic factor essential for the differentiation and establishment of primary versus secondary cortical territories in both, visual and somatosensory systems [79–81]. In the absence of primary (first-order) TCAs, cortical areas will remain in the default differentiation stage, which is that of secondary areas (Figure 1). Thus, TCAs from the first-order nuclei provide specific information to cortical area patterning. This information could involve activity-dependent mechanisms. Indeed, perturbations in the neurotransmitter release of TCAs lead to the reduction or complete abolishment of the barrel-field in the somatosensory system [72–74, 82].

The above-mentioned results show that thalamic input is critical for cortical sensory areas development. However, the time at which TCAs are necessary is less understood. In fact, the thalamic influence over the formation of cortical sensory representations occurs much earlier than thought. By perturbing the embryonic topography of the thalamocortical axons specifically at the basal ganglia, Lokmane and colleagues showed that the ordering of TCAs is required for the transfer of the whisker representation to the barrel field [83]. Somatosensory TCAs target abnormally V1 at embryonic stages and, although they recover and rewire into S1 postnatally, fail to form an anatomical and functional map in the barrel field [83, 84]. Thus, these results strongly suggest that the correct postnatal development of cortical maps may depend on early prenatal events involving a “memory” on the first thalamocortical interactions. The discovery of the intrinsic capability of the thalamus to generate thalamocortical spontaneous waves of activity embryonically might support this view. Thalamic neurons from different sensory modalities engage in a prenatal communication by means of spontaneous calcium waves that propagate to the cortex [85]. Embryonic perturbation of these waves leads to changes in cortical areas size by the specific activity-dependent regulation of genes in thalamocortical neurons (Figure 1). The enlargement of S1 that it is triggered by the perturbation of the thalamic waves or the elimination of the eyes occurs independently of sensory experience [85] supporting again an early role of the thalamocortical communication in cortical areas development.

Although it is believed that the spontaneous activity from the periphery drives the correlated patterns seen in central stations [86–88], the embryonic emergence of the thalamic calcium waves is independent of the peripheral input as thalamic waves persist in an embryonic enucleated mouse model in which retinal input never reached the thalamus [85]. Nevertheless, it is possible that later in life when most of the in vivo recordings have been made, peripheral input imposes or coordinates the pattern of activity in the distinct developing sensory stations, including the thalamus.

Concluding remarks & Future perspective

In this review, I have focused on recapitulating the mechanisms involved in the guidance, targeting and plasticity of the thalamocortical connectivity, specially emphasizing on how the development of specific techniques has helped the field to understand these processes. The rapid progression of techniques such as the in utero electroporation for manipulation of gene function in vivo, the generation of specific Cre-lox mouse lines for controlling spatial and temporal targeting of genes and the incorporation of new generation sequencing for transcriptomic analysis of specific neurons have been crucial. Now, we know the basic rules involved in the establishment of the thalamocortical connectivity, so we are in a privileged position to engage further challenging questions. For example, direct reprogramming of cortical postmitotic neurons has been successfully achieved [89]. We could then test whether thalamocortical neurons could be reprogrammed from one sensory-modality subset into another. Furthermore, endogenous brain cells such the postnatal reactive astrocytes have been reprogrammed into functional neurons in vitro [90, 91] as well as in vivo [92], at least in models of brain injury. Unpublished data from our laboratory suggest that thalamic astrocytes can be reprogrammed into functional thalamic neurons in vitro (A. Herrero and G. López-Bendito, unpublished). If this is also the case in vivo, reprogrammed astrocytes could be used in the future as a strategy to restore sensory circuits in sensory-deprived animal models later in life. Finally, 3D brain organoids offer nowadays an exceptional opportunity to study aspects of human brain development and disease [93]. Sensory processing disorders have been widely associated to neurodevelopmental disorders such as Autism or Schizophrenia [94–96], but the mechanisms involved in their generation remain poorly understood. In this sense, future studies could implicate the generation of brain organoids from patients with Autism Spectrum Disorder to better understand the possible developmental errors in the formation of sensory circuits that may underlie the sensory processing disorders associated.