Where a noradrenaline neuron sits in the mouse locus coeruleus predicts what it does

The positional claim needs a boundary, so the counting came first. The group, writing in Nature, labelled the nuclei of noradrenergic neurons through Dbh-Cre, the gene for the rate-limiting enzyme in noradrenaline synthesis. They cleared whole brains and imaged them with selective plane illumination microscopy: 34,998 cells in 8 mice ^[10], about 4,375 pontine noradrenergic cells per animal ^[15]. Of those they placed 1,507, plus or minus 55 across animals, inside the locus coeruleus ^[11], and 367, plus or minus 95, in sub-coeruleus ^[12]. Whether dorsal and ventral mean the same thing in two different labs depends on that boundary reproducing across animals. The locus coeruleus count is the steadier of the two. Its spread is 3.6 percent of the mean, against 26 percent for sub-coeruleus, roughly seven times less relative variation ^[16].

The task was built so that a prediction error is available on most trials. Mice chose between two options that paid probabilistic reward ^[13]. With payoffs uncertain, the animal keeps updating from its own actions rather than waiting on a fixed cue. Activity of the neurons projecting to cerebral cortex correlated selectively with that error term ^[13], and physiological properties and behaviour-related activity varied along the same spatial dimensions as the morphologies and the gene expression ^[14]. A correlation with prediction error does not show that cortex uses the signal to change the next choice.

The nucleus has often been treated as homogeneous ^[9]. That held even after earlier reports that neurons with different somatic shapes have biased axonal projections ^[18], and that subpopulations differ in input-output patterns and in function across regions of the central nervous system ^[19]. This one small structure in the pons supplies noradrenaline to most of the central nervous system ^[6], and it has been tied to arousal, sleep-wake transitions, stress responses, attention and learning ^[8].

Gene expression was graded across the nucleus and correlated with cell-body location ^[3]. A gradient of that kind biases a population without cutting it into types. The relationships reported here are correlations, in mice. Noradrenaline acts on G-protein-coupled receptors on neurons and glia ^[7]. Those receptors sit in the target regions, so how selectively a systemically delivered noradrenergic drug reaches one end of this axis depends on their distribution across those targets. The study did not measure that.