PNs represent the majority of neocortical neurons and are distributed in all cortical layers except layer 1. Among PNs, the thick-tufted cells are the most thoroughly studied, provide the cortical output directed to subcortical structures, and can be found in the deep part of layer 5 (layer 5b (Spruston 2008, Ramaswamy and Markram 2015)). Thick-tufted PNs are characterized by a prominent apical dendrite spanning all the way to the pial surface and terminating with a branching apical tuft, whose radium often equals (or exceeds) that of the basal dendrites (Fig. 1). Other classes of PNs are those of layer 2/3 (providing cortico-cortical associative and callosal projections) and those bearing a slender apical dendrite, mainly located in the superficial part of layer 5 (Krieger et al. 2017), see also Fig. 1 in (Shepherd 2013). Modified PNs projecting to the thalamus and claustrum reside in layer 6 (Thomson 2010).
The prototypical PNs, the thick-tufted cells of layer 5, display geometric differences among the basal, oblique, and apical dendritic domains that are clearly recognizable at a first glance (see Fig. 1). In a seminal paper published exactly thirty years ago, Alan Larkman provided a rigorous formal description of such branching pattern differences, pointing out, for instance, that basal dendrites branching points are close to the cell body, whereas intermediate branches of the apical tuft are relatively longer than distal ones (Larkman 1991). It is well known that the dendritic geometry impinges upon the functional properties of neurons (Mainen and Sejnowski 1996). Furthermore, different parts of the dendritic tree of layer 5 PNs are potentially involved in the microcircuitry of different cortical layers. Therefore, it is conceivable that basal and apical dendrites contribute differently to the cortical machinery during physiological cognitive tasks.
Regarding the connections, the basal dendrites receive feedforward input from the receptor surfaces through the thalamic relay. The ascending input is thought to be conveyed indirectly to the dendrites of layer 5 through the canonical cortical circuit (thalamus → layer 4 → layer 2/3 → layer 5 (Thomson and Morris 2002, Lübke and Feldmeyer 2007)). However, layer 5 neurons can be also contacted directly by thalamic afferents (Meyer et al. 2010, Constantinople and Bruno 2013). Conversely, most of the apical tuft of PNs lies in layer 1, where it receives feedback connections from higher cortical areas, conveying input related to attention, context, expectations (Coogan and Burkhalter 1990, Cauller 1995, Cauller et al. 1998). Together, the basal and apical dendritic arborization are in the ideal position to integrate bottom-up and top-down streams of information. The refinement of dendritic recording (Davie et al. 2006) made it possible to ascertain that action potentials can back propagate through the apical dendrite of PNs (Stuart and Sakmann 1994) and to shed light on the interplay between basal and apical dendrites. Using multiple patch-clamp recordings from the soma and the apical dendrite of layer 5 PNs, it has been demonstrated that the coincidence of a back propagated action potential generated at the soma and of an apical dendritic input is able to generate a dendritic calcium spike that, in turn, elicits a burst of somatic spikes (Larkum et al. 1999). This mechanism, originally called backpropagation-activated calcium spike firing (BAC firing) has been considered the electrophysiological basis of the top-down / bottom-up integration operated by a single PN (Larkum 2013). This idea was widened by Bill Phillips and Matthew Larkum, leading to the concept of “apical amplification”, the mean by which the information coming from the external world (bottom-up) is modulated by context-sensitive (top-down) information (Phillips et al. 2016, Phillips 2017). A role of the apical dendrite in cognition and consciousness has been also postulated by (LaBerge 2006) . Recently, it has been demonstrated that apical dendritic potentials can gate sensory perception and that such a modulation depends on contextual information (Takahashi et al. 2020). A dysregulation of context-modulated sensory perception and learning abilities can represent a prominent feature of ID (Alevriadou et al. 2004, Carr et al. 2010, Murray et al. 2019).
As to the distribution of ion channels on the membrane of different dendritic compartments, it has been proposed that the apical tuft and basal dendrites are dominated by NMDA receptors and associated potentials, while a calcium initiation zone, located just beneath the apical tuft, is endowed with voltage gated calcium channels (Nevian et al. 2007, Larkum et al. 2009), Fig. 1). Moreover, the correct degree of coupling between basal and apical compartments would be ensured by hyperpolarization-activated HCN channels, responsible for the Ih current and densely distributed on the apical dendrite of PNs (Phillips et al. 2016). Interestingly, these ion channels are developmentally regulated (Atkinson and Williams 2009) and their dysregulation might be involved in the genesis of neurodevelopmental disorders (see below).
In the last years, a conspicuous line of research focused on the dual basal/apical organization of PNs. Changes in the apical amplification process have been implicated in the pathophysiology of several mental disorders, including schizophrenia (Phillips et al. 2016, Mäki-Marttunen et al. 2019). Furthermore, the integration of two different compartments with feedforward and feedback input seems to be ideally suited to bridge the gap between artificial intelligence and neuroscience, since there are similarities between deep learning algorithms and the functional subdivisions observed in PNs (Guerguiev et al. 2017).
Dendrites of pyramidal neurons are covered with spines, which receive most of synaptic inputs and are thought to play a central role in several functions, from electrical filter/isolation to synaptic and structural plasticity (reviewed in (Yuste 2011, Sala and Segal 2014)). The density and distribution of dendritic spines appear to be differently regulated in apical and basal dendrites of PNs during learning (Knafo et al. 2001), in response to hormones (Gould et al. 1990), and in experimental models of neurological illness (Perez-Cruz et al. 2011).