Laminar-specific cortico-cortical loops in mouse visual cortex

Abstract

Many theories propose recurrent interactions across the cortical hierarchy, but it is unclear if cortical circuits are selectively wired to implement looped computations. Using subcellular channelrhodopsin-2-assisted circuit mapping in mouse visual cortex, we compared feedforward (FF) or feedback (FB) cortico-cortical input to cells projecting back to the input source (looped neurons) with cells projecting to a different cortical or subcortical area (non-looped neurons). Despite having different laminar innervation patterns, FF and FB afferents showed similar cell-type selectivity, making stronger connections with looped neurons versus nonlooped neurons in layer (L) 5 and L6, but not in L2/3. FB inputs preferentially innervated the apical tufts of looped L5 neurons, but not their perisomatic dendrites. Our results reveal that interareal cortical connections are selectively wired into monosynaptic excitatory loops involving L6 and the apical dendrites of L5 neurons, supporting a role of these circuit elements in hierarchical recurrent computations. Results Neurons with different projection patterns are intermingled in visual areasPrimary visual cortex (V1), the lowest-order area of mouse visual cortex, sends FF projections to the higher-order lateromedial (LM) and anteromedial (AM) visual areas, which in turn reciprocate with FB connections [17][18][19] . Using dual injections of retrograde tracers, we measured the laminar distribution of different projection neurons in V1 and in LM (Figure 1). In each experiment, we compared the laminar distribution of LM-or V1-projecting IT neurons with either IT neurons projecting to AM, PT neurons projecting to the superior colliculus (SC), or CT neurons projecting to the dorsal lateral geniculate nucleus (dLGN) or to the lateral posterior nucleus (LP) of the thalamus. In each case, the different projection neurons were closely intermingled ( Figure 1A,B). In both V1 and LM, IT neurons were distributed across all layers except L1, including L4 19,20 , indicating that FF and FB projections originate from neurons spanning most of the cortical depth. In contrast, PT and CT neurons were confined to L5 and L5/6, respectively, as previously described 4 (Figure 1A,B). In both V1 and LM, we found double-labeled IT neurons in L2-6 after injecting tracers in two different cortical areas, indicating that subpopulations of ascending and descending projection neurons have diverging axons innervating more

A direct interareal feedback-to-feedforward circuit in primate visual cortex

“…This is in contrast with results from mouse V1, where about 80-88% of FF projection neurons project to one or two higher visual areas 19,52 , but only about 50% of their monosynaptic FB contacts arise from the same areas to which they project 19 . Moreover, recent evidence suggests that in mouse a bias to form area-specific monosynaptic excitatory FF-FB loops may be limited to deep layer neurons 20 , while our results in monkey demonstrate area-selective FB-to-FF contacts in both V1 superficial and deep layers. These findings strongly support the existence of area-, and thus, stream-specific, FF-FB loops in primate cortex.…”

“…It is also unclear whether these cortico-cortical loops occur via direct monosynaptic contacts between FF and FB projection neurons, or indirectly via local excitatory or inhibitory neurons, or both. Recent studies have shown that in mouse primary visual cortex (V1), only a fraction of cortical projection neurons form area-specific monosynaptic FF-FB loops 19 , and that these loops may only occur between deep layer neurons 20 . It remains unknown whether similar rules of cortico-cortical connectivity apply to higher mammals.…”

Relationship between input connectivity, morphology and orientation tuning of layer 2/3 pyramidal cells in mouse visual cortex

“…The apical and basal dendritic trees of L2/3 PCs have been shown to play distinct roles in sensory processing. Feed-forward inputs from L2/3 and L4 influence orientation preference of L2/3 PCs (Ko et al 2011;Lee et al 2016), most likely targeting basal dendrites (Young et al 2019;Feldmeyer, Lübke, and Sakmann 2006). In contrast, cortico-cortical feedback inputs (Nassi, Lomber, and Born 2013;Wang et al 2007;Smith et al 2013) as well as orientation-tuned thalamocortical inputs (Chen et al 2013;Roth et al 2016) are likely to shape orientation selectivity via the apical dendrite.…”

Temporal learning of bottom-up connections via spatially nonspecific top-down inputs

“…If selective attention is directed to NPA, the synchronous activity of NPA is amplified, and Conn NPA, rather than Conn PA, become stronger. In contrast, the growth of Conn PA accelerates if selective attention is directed to PA. Interestingly, an earlier study proposed that top-down inputs impinging onto deep layers are target-specific [45]. Thus, selective attention may regulate this learning process via stimulation of deep layers.…”