Shea Lab Journal Club: Monitoring large populations of locus coeruleus neurons reveals the non-global nature of the norepinephrine neuromodulatory system

Monitoring large populations of locus coeruleus neurons reveals the non-global nature of the norepinephrine neuromodulatory system (2017) Totah et al.

This is the latest installment of  an ongoing series of capsule summaries of the Shea Lab journal club meetings. We are doing this to actively provoke open discussion of papers we read that are either published in a traditional journal or BioRxiv.


The brainstem nucleus locus coeruleus (LC) is nearly the sole source of the neuromodulator noradrenaline (NA) for the forebrain. The relatively small number of cells in locus coeruleus collectively broadly and diffusely innervate most of the forebrain. This projection pattern and the seemingly homogenous cellular composition of LC has inspired the conventional view that LC output carries a unitary signal, or some sort of global variable, to all brain regions.

Some recent anatomical data is leading researchers to reexamine that model. For example, this paper from the Waterhouse lab and this one  from my colleagues Justus Kebschull and Tony Zador show that there is heterogeneity in LC, and that individual cells may target a more restricted brain location than was previously appreciated. This potentially provides an anatomical substrate for LC to differentially modulate target brain structures. However, this would still depend on cells targeting those structure carrying distinct functional activity patterns.

The fundamental question that this paper tries to answer is: “To what extent do different LC neurons carry correlated or distinct information?” This is a very important question that has implications for past and ongoing work in my lab involving LC and NA, so with great interest, I selected the bioRxiv preprint for discussion in a recent Shea lab journal club. The paper is very complex and contains a lot of analysis, but my intention is not to write a comprehensive review. Here I just want to comment on some what I see as some important limitations of the study as written. Also, since this paper is in the preprint stage, maybe my thoughts will be useful to the authors moving forward.

In any neural system, there are always processes occurring in parallel at many time scales. For example, very rapid firing rate fluctuations may closely track the fine temporal structure of a vibratory tactile stimulus. Or auditory system neurons may “phase lock” to the periodic structure of sound. On the other hand, some neurons vary their spiking output dramatically across the circadian cycle of sleep and wakefulness, a much slower pattern. This contrast is also found in the multiplexing of firing rate fluctuations at long and short time scales in LC neurons.

For decades, researchers have recognized that LC neuronal spiking patterns are composed of slowly varying “tonic” spiking patterns that are punctuated by “bursts” of spikes at irregular intervals. Tonic firing changes occur over minutes or more, and individual phasic bursts last several hundred milliseconds. I would argue that these observations point to a range of time scales over which LC firing is likely to change with respect to behavior. As a result, I was disappointed that the authors of this paper performed the majority of their analyses at time scales that are shorter than that.

In many ways, the analysis in this paper makes sense. When one is looking at neuronal correlations, typically they are seen on a spike to spike basis. If cell B’s spikes are reliably occurring with a short delay after each of  cell A’s spikes, then that shows that A and B are correlated. The presence of gap junctions between some cells in LC makes these correlations largely expected. At this timescale, the authors found fewer correlations than one might have thought. The authors also looked for correlations over longer 200 ms windows, and found surprisingly few.

So why did this paper fall short of fulfilling my initial enthusiasm? It’s because the authors for the most part didn’t examine the relationship between firing in different cells over the timescale that I suspect is most closely related to behavior. To me, the relevant questions are: What is the relationship between phasic bursts across cells? Do they occur in a synchronous, coordinated manner? As tonic firing rates rise and fall over minutes within a single cell, does firing in other cells rise and fall in a coordinated manner? The authors quantified correlated firing over a maximum window of 200 ms, which is barely enough to encompass one of the phasic bursts. They did examine longer timescales, but only periodic activity; that analysis was not sensitive to slow, aperiodic fluctuations. So, the analysis demonstrates a surprising lack of spike to spike correlations, but they didn’t answer any of the questions above.

I imagine they may have the data that could answer these questions, and I would be very interested to know the results. To best answer to these issues however, one needs to fully explore a broad range of firing rates. That would be best achieved in an awake animal experiencing a range of brain states, but these experiments were performed in anesthetized rats. I don’t know whether they saw large swings in tonic firing rates and frequent phasic bursts under these conditions. In the spirit of constructive feedback, since this is a preprint, I recommend to the authors to describe how firing rates and bursts are coordinated over longer times. Otherwise, congratulations to them on a nice piece of work!



One thought on “Shea Lab Journal Club: Monitoring large populations of locus coeruleus neurons reveals the non-global nature of the norepinephrine neuromodulatory system

  1. Thank you for taking time to initiate a discussion on our paper. Your comments bring up some interesting questions and they have also been helpful as we submit the paper. I want to respond to some of your comments and would be happy to continue the discussion.

    Your critique rightly draws attention to the importance of assessing synchrony over those behaviorally-relevant timescales which are associated with tonic and phasic firing patterns. Phasic firing is often defined as a response to a sensory stimulus that occurs 50-100 msec after stimuli (Bouret & Sara 2004; Aston-Jones et al. 1994; Foote et al. 1980; Aston-Jones & Bloom 1981; Usher et al 1999). As you note, phasic activity can also occur spontaneously as bursts. Although we did not study synchrony during spontaneous bursts, we did examine at least one form of phasic activity, which is evoked by sensory stimuli. Specifically, we examined foot shock evoked spike count correlations in 50 ms and 200 ms windows post-stimulus that would capture the behaviorally-relevant phasic firing patterns occurring 50-100 ms after salient sensory stimuli. Given that the phasic response to salient stimuli may underlie reorienting behavior, a primary function of the LC, I believe that we have examined at least one type of phasic activity that is crucial for behavior. I agree that we should look at this data set for spontaneously occurring bursts and examine synchrony among these events.

    Tonic activity, in contrast to phasic activity, as you point out can vary in relation to arousal on the time scale of minutes (Usher et al 1999) or, in the case of circadian rhythms, on a time scale of hours (Aston-Jones et al 2001). However, tonic activity can also be considered as any spiking that is not a burst/phasic, such as the spontaneous activity occurring over 100’s of msec or multiple seconds preceding a stimulus onset. In the behaving monkey, minutes long fluctuations in tonic firing rate were actually associated with correlated tonic firing on a time scale of 200 msec (Usher et al., 1999). Thus, it appears that synchronous minutes-long tonic activity patterns may be apparent at a shorter scale of 200 msec. We chose to assess correlated firing (spike count correlations) at 200 msec because of these prior findings by Usher et al (1999) that were based on a small number of pairs (N=23 pairs). We additionally assessed synchrony at a much larger window of 1 sec and still found extremely low correlations among LC units.

    Many models of the modulatory effect of LC activity on behavior are built upon the oft-assumed propensity for LC synchrony—specifically—on briefer time scales of 100’s of msec or sub-millisecond presumably mediated by gap junctions (Usher et al., 1999). Therefore, I think that updating models of behavior and LC activity requires assessing the degree to which synchrony occurs over the millisecond to seconds timescales that we characterize.

    Finally, it is important to note that LC axons also use synaptic neurotranmission (Miner and Sesack 2007), which suggests that correlated activity on the timescale of 1 msec (mediated by gap junctions) or 10’s to 100’s of msec (mediated by intra- and extra-LC network connections) can matter from the viewpoint of the post-synaptic neuron and may therefore be relevant to behavior.

    All of this being said, I appreciate the importance of the larger timescales. Determining the degree of correlated activity on a minutes-to-hours scale would undoubtedly be important for understanding the role of LC in mediating the well-known effects of arousal on cognition and behavior. But, I am hopeful (based on the observations of Usher et al., 1999) that these shorter epochs (200 msec) that we assessed do actually provide a window into the degree of correlated activity on the minute scale.

    I hesitate to study correlations over minutes because it is not clear to me if we have a way to assess synchrony on this time scale in a meaningful way. Spike count correlation coefficients depend on the time window (Cohen & Kohn 2011) and are typically measured over epochs of less than 1 sec. Using a time bin of multiple minutes might simply max out the correlations. One possibility is to compare spike count correlations in 15 min bins between LC and a comparison data set from a brain region that does not have homogenous activity (e.g., cortex). Even if the very large time window increases correlations, the distributions of pairwise spike count correlation coefficients could be compared between LC and cortex.

    Thank you again for starting an interesting discussion. It has helped us to clarify some ideas in the manuscript.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s