How to teach this paper: ‘Coordination of entorhinal-hippocampal ensemble activity during associative learning,’ by Igarashi et al. (2014)

Kei Igarashi and his colleagues established an important foundation in memory research: the premise that brain regions oscillate together to form synaptic connections and, ultimately, memories.

Photograph of two hands drawing overlapping red and blue waveforms on a chalkboard.
Doing the wave: The team sought to understand what precisely was happening in the brain when brain oscillations synchronize, and how it might enable animals to remember.
Photograph by Margeaux Walter

Our brain activity is a constantly evolving mix of waves at different frequencies, depending on what we’re up to. We ramp up our alpha waves with meditation, increase the gamma gain for difficult cognitive work and settle into slow delta waves while we’re sleeping. Researchers have been able to observe brain oscillations for more than 100 years, but what do these actually do?

Many brain wave theories have cycled in popularity over the past century. Some researchers claim that brain waves are central to synchronizing the activity of single neurons or helping the brain process information. Others purport that they are simply the neural exhaust of a bioelectric organ.

About a decade ago, Kei Igarashi, Laura Colgin and Nobel-Prize-winning scientists May-Britt and Edvard Moser set out to understand whether brain waves had anything to do with helping us consolidate memories. Building on other research suggesting that very fast gamma oscillations and slower theta oscillations might interact to form memories, they looked for cold, hard evidence that this was indeed the case. In doing so, they established an important foundation in memory research: the premise that brain regions oscillate together to form synaptic connections and, ultimately, memories.

PAPER OVERVIEW

Declarative memory allows us to consciously recall facts we have learned—that the Panama Canal was completed in 1914, for example—or specific episodes in our lives—that you visited Panama when you were 13 years old. It’s also one of many different functions that research has tied to brain oscillations.

When we first form and when we retrieve these kinds of memories, parts of our cortex and hippocampus vibe at two particular frequencies: a low, cool theta rhythm alongside a quicker gamma oscillation. Multiple studies across rats and humans have shown that when theta and gamma frequencies ebb and flow together within the hippocampus, we remember more. Earlier work in the Moser Lab also showed that fast gamma alone could synchronize different brain regions but did not directly tie this observation to learning and memory.

Until the 2014 paper, memory researchers had not observed changes in synchronization between brain regions during behavior and did not know how oscillations related to changes in single neurons. Igarashi et al. sought to understand what precisely was happening in the brain when brain waves synchronize, and how it might enable animals to remember.

This paper tackles the mechanisms of memory, making it a great addition to most neuroscience, psychology or cognitive science classes. Though the methods aren’t too technical (especially by today’s standards), a few approaches, such as population vector and spectral analyses, might be worth unpacking in a class focused on neural data analysis. For everyone else, you can simply explain the why of these methods without digging into the how.

PRE-READING

Rats can’t tell you what they learned in history class, so Igarashi et al. focused on associative learning, a kind of learning in which we make connections—both metaphorical and neural—between different stimuli. One nice feature of the paper is that almost all of the figures share this same basic experimental configuration: The researchers trained animals to associate specific locations and smells while recording from multiple regions of the hippocampus. This setup of using tetrode recordings in freely moving animals should be readily understandable to most undergraduates in psychology or neuroscience.

The analysis of the electrophysiology data is a little tricker. This paper is built on the premise that you can record a noisy voltage trace over time from a brain and use an algorithm known as a Fourier transform to break that trace down into individual sinusoidal frequencies, akin to how a complicated musical recording can be decomposed into pure tones. In other words, the paper views data primarily in the frequency domain rather than in the time domain.

Students don’t need to know the mechanics of a Fourier transform, but they should understand why it is being used and maybe even predict what time series data would look like in the frequency domain. (See “Additional activities” below.) A common way to visualize the decomposition of recordings into frequencies is with a spectrogram, a special kind of heatmap that shows the value of a particular metric (such as the power of a brain wave) for different frequencies (on the y-axis) across time (on the x-axis).

Students should also be familiar with the concept of coherence, or how much the oscillations in two different brain areas are in sync. Imagine two stadiums that are both doing the wave: If those waves hit similar places in the stadium at the same time, we would think of them as being coherent. In the same way, when two brain areas are coherent, their oscillations hit peaks and troughs around the same time. Igarashi et al. use spectrograms to illustrate changes in power and variations in coherence of waves in different frequencies.

WORKING THROUGH THE FIGURES

Let’s start with the behavior itself. In a classic operant conditioning behavior, rats were trained over a few weeks to associate particular odors, such as the smell of roses, with the location of a food cup.

Figures 1e, 1g and 1h from Coordination of entorhinal-hippocampal ensemble activity during associative learning, by Igarashi et al (2014).
Igarashi et al. Nature 2014

In Figure 1 (above), we see the first evidence for an increase in slow gamma activity (20- to 40-Hz power) in the lateral entorhinal cortex (LEC) and dorsal CA1 (dCA1)—but not in the medial entorhinal cortex (MEC)—shortly after the well-trained animals poke their nose to receive a reward (bright red area in the middle of plots in panels e and g). This uptick in slow gamma was observed at the moment of remembering, suggesting that it is directly tied to the process of memory retrieval. The first figure also distinguishes the roles of slow and fast gamma activity: Fast gamma happens in MEC during exploration, slow gamma happens in LEC during retrieval. In line with these changes in power, the researchers also observed an increase in coherence between the LEC and dCA1 regions (top image in panel h). Not only are these regions increasing their slow gamma activity, but they’re doing so together.

Figures 2a and 2c from Coordination of entorhinal-hippocampal ensemble activity during associative learning, by Igarashi et al (2014).
Igarashi et al. Nature 2014

Figure 2 (above) features one of the most exciting developments in this paper: The uptick in slow gamma coherence occurs only when animals actually learn the task at around 2 weeks (timepoint T4). Brain activity during trials in which the animals make an error (labeled T5e) lacks this coherence. In other words, this coherence is directly related to learning and the ability to correctly remember. The researchers also dig into the specificity of the increase in the slow gamma (20- to 40-Hz) band, emphasizing that the theta (6- to 10-Hz) band doesn’t increase in coherence with learning.

The next figure primarily features control experiments intended to account for changes in the animal’s attention. They want to show that it’s not just that animals are attending to the odor, but that they are remembering the association. Neither the odor alone (non-cued trials) nor a novel odor provoke the same increase in slow gamma power or coherence.

Finally, the researchers show that the increases in 20- to 40-Hz power and coherence correlates with changes in how strongly single neurons in both dCA1 and LEC selectively respond to specific odors. They’re also able to demonstrate that animals actually perform better on the task when there is more selectivity. This suggests that single-cell changes happen alongside changes in the population synchrony. But it’s still a bit of a chicken-versus-egg scenario: It’s hard to know whether neural oscillations drive selectivity in single neurons or vice versa. In the discussion, the authors make an argument for synchrony driving selectivity, given what we know about Hebbian plasticity. Whether this is the case or not could be a great discussion point for your class.

ADDITIONAL ACTIVITIES & VIEWING

Thinking of time series in the frequency domain is always a tough cognitive task, even for advanced undergraduates. Before digging into this paper, I would recommend going through some more simple, contrived examples of how adding up sine waves can create a seemingly noisy signal, like the kind you’d record from the brain. Here’s a notebook where you can explore these kinds of simple signals and demonstrate the kinds of spectrograms they’ll generate.

Because May-Britt and Edvard Moser are Nobel Prize winners, there is quite a bit of media out there about them. This video, produced by their home university, provides a nice overview of their original groundbreaking work in entorhinal cortex and shows a rat with an implant that is similar to the ones used in Igarashi et al.

Get alerts for “How to teach this paper” in your inbox.

This series guides educators and self-learners through recent seminal neuroscience papers.

Explore more from The Transmitter