Cortical layers serve multiple purposes: some serve more local microcircuits while others act as interconnecting loops (thalamo-cortical) or large-scale networks (bottom-up) that have a multifunctional role.

Being part of the brain’s connectome (Hilgetag et al., 2016) including the entirety of brain’s connections, cortical layers serve local microcircuits processing sensory or motor information, as well as the interconnected loops (thalamo-cortical) or large-scale networks: (i) bottom-up, sensory-to-motor; (ii) top-down, cognitive-to-motor; and (iii) inter-hemispheric, for coordination of various behaviors, serving a multifunctional purpose (Miller and Phelps, 2010; Makarova et al., 2011; Georgopoulos, 2015).

Indeed, neurons from different cortical layers/columns communicate to each other via synchronized firing (Romo et al., 2003; Opris et al., 2013) across cortical layers, quantified by cross-correlation histograms (CCHs). Also, neuronal integration has been studied using “noise and signal correlations” or LFP-spike interactions (Bosman and Aboitiz, 2015).

As seen in Figure 1, the neurons in prefrontal cortical layers L2/3 and L5 change firing patterns during the presentation of the matching target (together with the non-matching distractors), resulting in inter-laminar CCHs changes (increases the number of coincident spikes) that may represent trans-laminar integration of information (Foxworthy et al., 2013; Opris et al., 2013). Information flows through cortical layers in a feed-forward manner (Bastos et al., 2015), going from layer 4, to layers 2/3 and onwards (Bastos et al., 2012). Another view maintains that cortical layers can have distinct inputs that activate them, triggering spikes when the integration input bypasses a threshold (Opris and Casanova, 2014).

The integrative hypothesis posits that integration of perceptual signals from supra-granular layers with the behavioral signals in the infra-granular layers takes place in the prefrontal cortical inter-laminar microcircuits (Opris et al., 2011, 2012a,b, 2013; Takeuchi et al., 2011).

Layer 1: Cross-modality modulation is achieved through layer L1 cell-mediated inhibitory and disinhibitory circuits (Ibrahim et al., 2016). Neurons in layer L1 form “canonical neuronal circuits” to control information processes in both supra- and infra-granular cortical layers (Lee et al., 2015). Projections of neurons to layer L1 from all cortical inputs has been shown to make a significant contribution to the integration process throughout the neocortex (Mitchell and Cauller, 2001).

Layer 2/3: This layer provides the integration of perceptual stimuli by inter-areal bottom up connections and the integration of perception with action by trans-laminar connections and top-down cortico-subcortical connections (Opris et al., 2011, 2012a,b, 2013).

Layer 3: This layer provides inter-hemispheric callosal connections (Georgopoulos, 2015).

Layer 4: The primary visual cortex receives orientation- and direction-tuned inputs from thalamus in layer 4 (Sun et al., 2016).

Layer 5: Pyramidal neurons in this layer integrate inputs from many sources and distribute outputs to cortical and subcortical structures (Kim et al., 2015). Cortical-subcortical neurons project to subcortical structures in the thalamo-cortical loop, integrate information and generate responses more relevant to movement control, while cortico-cortical neurons are more important in visual perception.

Layer 6: This layer projects to the thalamus (Opris and Casanova, 2014).

It is obvious that cortical neurons that interconnect to each other across layers play an important role in the emergence of brain function. To demonstrate this we show in Figure 1 an ideal example that depicts simultaneously recorded cells with differential firing in adjacent minicolums and inter-laminar layers during the match epoch of the DMS task (Figure 1A). Functional integration may be pursued based on: (i) inter-laminar; and/or (ii) inter-columnar neuronal interactions. The cross-correlograms (Figure 1C) that quantify these interactions between cells provide crucial evidence for the integration process (Opris et al., 2012b).

Figure 1C shows an example of inter-laminar interaction between two cell pairs with individual firing rates displayed in raster/peri-event histograms (PEHs) during the epoch between matching image presentation (match phase onset) and offset of the matching “target” (0.0 s ± 2.0 s). Cell pairs in this study (Opris et al., 2012b) were recorded on “appropriate sets of adjacent pads” (corresponding to Minicolumns 1 and 2) on the biomorphic multielectrode arrays (MEAs), shown in the illustration of both trans-laminar cell pairs in L2/3 and L5/6 (Figure 1C). Neurons in both, upper and lower layers showed “significant increases” in mean firing rate (p < 0.001) in the “post”-match phase (0.0 s to 2.0 s) associated with target selection (Figure 1A) and the decision making phase of the task (Opris et al., 2012b), relative to the “pre”-match phase (−2.0 s to 0.0 s). An important observation highlighted with this type of recording configuration was that layer L2/3 neurons in the same minicolumn, exhibited higher firing rates in the “post”-match epoch (0.0 s to +2.0 s; Figure 1C, upper raster/PEHs) than neurons in L5 of the same, over the same time period (Figure 1C, lower raster/PEHs).

Precise “functional connections” between single units (cells) within each minicolumn were provided by cross-correlation histograms (CCHs), represented for individual L2/3 and L5 cell pairs recorded on vertically positioned electrode pads (Figure 1B) of the MEA (Opris et al., 2011, 2012b, 2013; Takeuchi et al., 2011). Normalized synchronized firing (shown by CCHs) for both minicolumns were shown in Figure 1C for cell firing in the displayed PEHs: (i) “pre”-match epoch (−2.0 s to 0.0 s, Pre, black curve); or (ii) “post”-match epoch (0.0 s to +2.0 s, Post, green) for the same cell pairs. Although both CCHs depict a “significantly correlated firing” (p < 0.001; t-test), the differences in max correlation for both neuron pairs suggest that inter-laminar firing was more synchronized in the “post”-match epoch (0.0–2.0 s) than in the “pre”-match epoch (p < 0.001; t-test).

The demonstration of functional connections between individual cells within same cortical layer and different minicolumns was provided by CCHs (Opris et al., 2012a) plotted for individual L2/3-vs.-L2/3 and L5-vs.-L5 cell pairs, recorded on horizontally positioned electrode pads on the MEA. Normalized correlograms for cell pairs are presented in Figure 1C. Although both correlograms (L2/3-vs.-L2/3 and L5-vs.-L5) show significantly correlated firing (Post vs. Pre: p < 0.001), the differences in peak correlation for both cell pairs indicate that intra-laminar inter-columnar firing was more synchronized in layer 2/3 than in layer 5 (p < 0.001). This supports the idea of laminar integration in supra-granular layers (Petro and Muckli, 2017).

Cortical-subcortical integration of sensory and motor signals occurs during cortico-striatal interactions funneling signals in the cortico-thalamic loops (McFarland and Haber, 2002). Two recent results reported by Opris et al. (2013) and Santos et al. (2014) have demonstrated differential (pre)frontal cortico-striatal interaction during the planning and preparation for movement in monkeys.

Neurons in Layer 3 of fronto-parietal cortices provide important interhemispheric callosal connections necessary for the coordination of movement (Georgopoulos, 2015). Obviously, inter-hemispheric integration is playing a key role such as coordination.

One way to test whether trans-laminar integration takes place in cortical minicolumns is to compare the inter-laminar firing in correct vs. error trials. A trial was considered correct when the animal responded adequately to each instruction of the DMS task’s event sequence (shown in Figure 1A), receiving a drop of juice as reward, or an error trial, when the animal failed to obey one or more instructions, and was not rewarded (Opris et al., 2012b). Experiments by Opris et al. (2012b) compared correlated firing of pairs of cells from layers L2/3 and L5 during the match epoch (dealing with target selection) under correct vs. error trials. Figure 2A shows that cell pairs (in layers L2/3 and L5) that exhibited increased firing during the match epoch in correct trials (left), reduced their firing on error trials, with the inappropriate target being selected (right). The trend of increased (respectively decreased) correlated firing in correct (respectively error) trials was present across entire subpopulation of prefrontal cell pairs, as shown in Figure 2A. Moreover, the significant increase (decrease) in the mean CCH peaks (p < 0.001), provides evidence for the enhanced (lack of) inter-laminar correlated firing between L2/3 and L5 in correct (error) trials, as shown in Figure 2B. Taken together, these unique simultaneous recordings in prefrontal cortical layers (and minicolumns) of rhesus macaques, during in vivo performance of a cognitive task, provide evidence for prefrontal cortical inter-laminar integration during correct trials and lack of integration during error trials.

During the perception-to-action cycle, when the brain plans a movement, the executive mechanism coordinates the interactions between the environment and the perceptual/sensory-motor/executive systems (Fuster, 2000; Opris et al., 2011, 2013). On top of the executive hierarchy, the prefrontal microcircuits are assumed to “bind perceptual and executive control” functional signals to coordinate goal-driven behavior (Figure 3A). Here, we highlight the recorded neuronal “firing simultaneously in prefrontal cortical layers and the caudate-putamen” of rhesus monkeys (see Figure 3B), trained in a spatial-vs.-object version of the visual match-to-sample task (Opris et al., 2013). During the perception and executive selection epochs, cell firing in the prefrontal layers and caudate-putamen exhibited “preferences for the same location” on spatial-trials, but not on object-trials.

When the perceptual-executive circuit is activated by the stimulation of prefrontal infra-granular-layers with electrical stimuli “patterns” obtained from supra-granular-layers, Opris et al. (2013) could replicate spatial preferences (similar to neural tuning) in percent correct performance, on spatial trials (Figure 3C). These results show that inter-laminar prefrontal microcircuits play causal roles in the perception-to-action cycle (Mahan and Georgopoulos, 2013; Opris et al., 2013).

In the DMS task in Figure 1A the distractor images are those images that represent a different object than the sample image. As shown previously (Opris et al., 2011, 2012a,b, 2013), a major factor influencing the selection of the matching target in the match phase of the DMS task was the number of distractor images presented with the “sample” image in a given trial (see Figure 4). The increase in task difficulty via increasing the number of distractors, allowed the animal to make a sizeable amount of error trials for comparison to the correct trials (Opris et al., 2012a). In Figure 4A is shown a gradual “decrease in cell pair firing in layers L2/3 and L5” as a function of the number of images presented during the matching phase. Consistent with our prior results (Opris et al., 2012a,b), neuronal population mean firing rates in layers L2/3 and L5 (Figure 4B) confirmed this decrease as a function of the number of distractors in the match phase (p < 0.001). Moreover, this decrease was also expressed in terms of “correlated firing” between L2/3-L5 cell pairs, as shown in Figures 4C,D, in which CCHs on trials with few (2 or 3) distracter images exhibited significantly higher correlations than on trials with more (6 or 7) distracter images (p < 0.01). The decrease in inter-laminar correlated firing is consistent with a decrease in integration caused by the increase in the number of distractor targets. Likely the decrease in task performance is due to an increase in the “cognitive work load” of the task (Kelley and Lavie, 2011; Opris et al., 2011).

To test causal relationship to columnar inter-laminar integration, the recorded firing of neurons was examined via a nonlinear multi-input multi-output (MIMO) model, which extracted and characterized multi-laminar firing patterns during performance in the DMS task (Hampson et al., 2012). Figure 5 provides causal evidence for the integration of natural signal flowing through the PFC microcircuits with the injected signal via stimulation. Minicolumnar neuronal firing of the PFC neurons was recorded from rhesus macaques trained in a DMS task, via biomorphic MEAs that provided signals from neurons in prefrontal cortical layers 2/3 and 5 (Opris et al., 2015a). The MIMO device sent patterns of electrical stimuli via stimulation to prefrontal cortical layer 5, during columnar “inter-laminar integration” at the time of target selection (Opris, 2013). Such stimulation improved (augmented) normal task performance significantly (Figure 5D). The executive control circuit was facilitated by applying stimulation patterns in the prefrontal cortical infra-granular-layers similar to the patterns coming from the supra-granular-layers that produced enhanced spatial preference in performance percentage, similar to neural tuning. Thus, inter-laminar prefrontal microcircuits may play causative roles in the cognitive perception-to-action cycle (Hampson et al., 2012; Opris, 2013). Such causal relations to integration of cognitive signals provide direct evidence that cortical minicolumns may represent the turbo-engines of the brain (Jones, 2000; Jones and Rakic, 2010; Opris et al., 2013; Chapman and Mudar, 2014).

Figure 6 illustrates modular integration across hierarchical levels. Executive control has been defined as the brain’s ability “to control thought and action” by coordinating “multiple systems and mechanisms across multiple brain areas” to pursue a goal (Miller and Phelps, 2010). Examples of executive control functions include: attention, working memory, decision making, intention or motor plan and behavioral inhibition (Fuster and Bressler, 2012). Thus, the ability of the brain to exercise executive control over behavior relies upon the integration of the multiple microcircuitries (sensory, motor, reward), loops (thalamo-cortical) and large scale networks (bottom-up or top-down) with distributive encoding organized in a hierarchical manner.

The hierarchy of brain functions was introduced by Joaquin Fuster based on Hughlings Jackson’s assumption that the neocortex is the climax of the nervous system and it controls (activates and/or inhibits) the functions of lower levels. However, cortical disease led to two sets of findings: “negative” signs and symptoms from “loss of the controlling cortex” and “positive” ones from the “emergence of the lower center” (Fuster, 1990; York and Steinberg, 2011). This implied an anatomical and physiological hierarchy of centers within the brain, with higher ones activating or suppressing the function of lower ones (Fuster, 1990; York and Steinberg, 2011).

Our focus is on the hierarchy of neural circuitry underlying the executive control of behavior (Figure 6) spanning from the frontal and parietal cortices, subcortical structures like the basal ganglia and thalamus, the brainstem and the spinal cord. The hierarchical integration of various stimuli (Hirabayashi et al., 2013a,b) in the executive function, may follow a bottom-up integration of visual information (Felleman and Van Essen, 1991) while the coordination of movement kinematics is performed in a top-down manner according to a prior intention/plan (Noga and Opris, 2017). At the top of the executive hierarchy are the frontal (premotor) and parietal (motor) cortical microcircuits, interconnected within thalamo-cortical loops through cortico-striatal projections and further in the brainstem to the mesencephalic locomotor region (MLR) and the central pattern generators in the spinal cord (Noga and Opris, 2017).