Imaging the activity of neuronal populations: when spikes don't flash and flashes don't spike



Calcium imaging has become a powerful tool to observe the activity of entire neuronal networks. The neuronal populations can be stained via bath application or multi-cell bolus loading of AM-esters of calcium-sensitive dyes, or via genetic approaches. Two-photon laser scan imaging, possibly in combination with novel types of optical fibres, fast 3D scanning and temporal deconvolution algorithms, allows the observation of stained brain regions in vivo (e.g. Ohki et al. 2005). These recent developments open a window into brain function between the single cell level and large scales covered by non-invasive techniques such as EEG and MRI, and are thus greeted with enthusiasm by neuro-theorists and experimentalists alike (Grinvald, 2005).

The most pertinent data these techniques yield are the somatic fluorescence levels of several dozen to hundreds of neurons, with a temporal resolution of up to about 10–20 Hz. Even at these rather high scanning speeds a decent signal-to-noise ratio and thus easy detection of fast rises in fluorescence can be achieved because the spatial resolution is usually still high enough to provide a sufficient number of pixels (or even voxels, in 3D scanning) per neuronal soma. Such fast rises in somatic neuronal fluorescence and thus in turn in calcium levels are interpreted as tight correlates of somatic sodium action potentials (APs) and thus mirror neuronal output. This correlation was established in a seminal study by Kerr et al. (2005) on neocortical networks in vivo, where the authors simultaneously imaged and recorded from layer 2/3 pyramidal neurons. Not only would fast transients always relate to the occurrence of spikes and vice versa, but the size of the calcium transients was also found to scale linearly with the number of APs in a train or burst. Therefore, this type of imaging allows the resolution of the spatiotemporal patterns of spiking activity in a group of cortical neurons.

However, the one-to-one correlation between fast calcium signals and AP firing may not be as tight for neuronal cell types other than cortical pyramidal cells. Here the study by Lin et al. (2007) in this issue of The Journal of Physiology comes into play. The authors have performed calcium imaging of populations of olfactory bulb neurons in a nose–brain preparation of the Xenopus tadpole, also in combination with on-cell and whole-cell recordings of individual bulbar neurons. Their data show that there are distinct activation patterns within the two main populations of subglomerular bulb neurons, mitral cells and granule cells, both spontaneously and in response to sensory stimulation. Of particular interest in the above context is the finding that the occurrence of granule cell APs as established by the electrophysiological recordings is often not well correlated with increases in fluorescence and vice versa. On the other hand, mitral cells do show such a tight correlation, serving as a control for the authors' findings in granule cells. Thus the authors provide convincing evidence that the causal relation between APs and fluorescence increases does not necessarily hold for all cell types and therefore is an assumption that needs to be validated in any newly investigated system.

As to the mechanisms that underlie this decorrelation, the granule cells of the olfactory bulb lack an axon but nevertheless feature regular sodium spikes that cause somatic calcium transients. However, granule cells have been shown to carry also other types of calcium signals and Lin et al. (2007) present some preliminary evidence for a role of low-threshold voltage-dependent calcium channels and NMDA receptor-mediated somatic calcium entry. Thus any neuron types that feature calcium signals independent of sodium spikes, e.g. low-threshold spikes, NMDA spikes or calcium waves, may produce such ‘false positives’ if the AP-independent signal can propagate to the soma. Candidates are for example thalamic neurons, certain inhibitory interneurons or cerebellar purkinje cells. This issue might be resolved by imaging at yet higher temporal resolution since the calcium signals produced by AP-independent mechanisms rise more slowly than those due to sodium APs. In any case, complex spike phenomena where AP-mediated and AP-independent calcium signals mix will preclude a proper resolution of the number of APs per calcium transient. The reverse finding of granule cell APs without detectable transients, i.e. ‘false negatives’, is potentially more troublesome and also deserves further investigation.

Aside from this general aspect, the study by Lin et al. (2007) is one of the first to characterize population activity in the subglomerular layers of the olfactory bulb in response to odour input (see also Yaksi & Friedrich, 2006). Since it is becoming more and more evident that the olfactory system operates in a very similar way across phylae, these results are likely to be relevant also for our understanding of mammalian olfactory function. For example, the authors report striking suppressive responses with respect to the calcium level in mitral cells. While the mechanisms of these phenomena await further clarification, the authors' system would also be well-suited to test the hypothesis that granule cell response latencies are odour specific (Kapoor & Urban, 2006).