This Week in The Journal

Intracortical Connections Refine Orientation Tuning in V1

Logan Chariker, Robert Shapley, and Lai-Sang Young

Most neurons in primary visual cortex (V1) respond preferentially to lines of a particular orientation. Such orientation selectivity is thought to arise from the convergence of inputs from lateral geniculate nucleus (LGN) neurons whose receptive fields lie in a line. Although much evidence supports this hypothesis, it cannot easily account for some features of orientation selectivity, including the selectivity of complex cells, which receive few, if any, direct inputs from LGN. Therefore, most models of V1 orientation selectivity include contributions from recurrent intracortical connections.

Columns of neurons in V1 have similar orientation preferences, and these columns are organized into hypercolumns, in which all orientations are represented. Evidence suggests that each hypercolumn receives direct input from only ∼10 LGN neurons. Can such sparse connectivity produce the full range of orientation preferences detected in V1? To answer this, Chariker et al. created a large-scale computational model of V1 layer 4Cα (L4Cα). Each hypercolumn contained ∼3000 excitatory and ∼1000 inhibitory neurons. Each excitatory neuron received input from ∼200 excitatory and ∼100 inhibitory neurons in the same layer, as well as feedback from ∼50 L6 neurons. Additionally, each L4Cα neuron received 0–6 inputs from LGN.

When the model network was given simulated visual input, L4Cα neurons exhibited many properties observed in vivo, including a broad range of firing rates, higher rates in complex cells than in simple cells, tuning curves of varied widths, and a full range of orientation preferences. Furthermore, when stimuli with the preferred orientation—but not the orthogonal orientation—were presented, gamma-frequency oscillations emerged in the L4Cα population. The simulations indicated that interactions between L4Cα neurons greatly sharpened the orientation tuning of simple cells, and generated orientation selectivity even in model cells that received no direct input from LGN.

These results show that sparse, broadly tuned inputs from LGN, when paired with recurrent excitation and inhibition in V1, can generate narrow tuning curves with a full range of orientation preferences in L4Cα. The model makes several predictions, including: orientation selectivity is weak in LGN cells, intracortical connections sharpen orientation selectivity, and orientation preferences vary within local populations in L4Cα. These predictions should be examined experimentally to further test the validity of the model.

Cognitive Function Varies across the Breathing Cycle

Christina Zelano, Heidi Jiang, Guangyu Zhou, Nikita Arora, Stephan Schuele, et al.

(see pages 12448–12467)

Activity in the olfactory bulb and piriform cortex oscillates in phase with respiration, and this is thought to facilitate odor processing. Oscillations in olfactory regions might be expected to drive similar oscillations in downstream areas, and there is evidence that this occurs in rodents, where respiration-linked oscillations are present in barrel cortex and hippocampus. Whether respiration drives oscillations in human cortex, and whether this affects cognitive behavior, has been unknown. Zelano et al. addressed both of these questions.

Mean spectral power (pseudocolor) at delta frequencies is highest (red) in the amygdala during the inspiration phase (peaks in black line) of nasal respiration (top). Power changes are diminished when patients breathe through the mouth (bottom). See Zelano et al. for details.

Intracortical electroencephalographic recordings in epilepsy patients revealed that the local field potential in piriform cortex oscillated with breathing when patients breathed through their noses. In addition, the power of delta- (0.5–4 Hz) and theta-range (4–8 Hz) oscillations increased in piriform cortex during the inspiratory phase of breathing. Increases in delta and theta power also increased in the amygdala and hippocampus during inspiration, but oscillations at the breathing frequency (0–0.6 Hz) were not reliably detected in these areas. Interestingly, the effects of breathing on neural oscillations depended on nasal airflow: the effects were attenuated when patients breathed through their mouths.

Experiments in healthy volunteers suggested that increases in oscillatory power during inhalation might affect cognitive performance. When asked to distinguish expressions conveying fear or surprise—a task that is thought to depend on the amygdala—participants responded more quickly to fear-expressing faces during inhalation than during exhalation. In a hippocampus-dependent object-recognition test, participants responded more accurately to pictures presented during inhalation than to pictures presented during exhalation. Notably, in both behavioral tasks, the effect of respiratory phase was only significant when participants breathed through their noses, and cognitive performance was worse during mouth breathing than during nasal respiration. Finally, in the single epilepsy patient that performed the fear recognition task, delta power in the amygdala during inhalation was inversely correlated with reaction time.

Together, these data reveal that breathing affects brain oscillations and cognitive function in humans, even on tasks that are not obviously linked to olfaction. More practically, they suggest that when faced with a cognitive task, it is wise to keep your mouth closed and inhale deeply.