18 September 2008. Human memory is one of the last great biological mysteries. Scientists know that the hippocampus, an area of the brain that is damaged early in Alzheimer disease, plays a crucial role in laying down and retrieving memories, but it is not clear how hippocampal cells cooperate with each other and with other neural networks to facilitate either process. Two recent studies, one on humans and the other on rats, suggest that for short-term memories, at least, the very same neurons that fire when memories are made fire again when those memories are spontaneously retrieved. The studies make a strong case that reactivation of hippocampal neurons is involved in conscious recall, an idea that has been difficult to address.
The studies appear in the September 4 and 5 online and print versions of Science, respectively. In the former, researchers led by Itzhak Fried, who holds positions at the University of California, Los Angeles, and Tel Aviv University, Israel, reported memory studies in patients undergoing surgery for epilepsy. As part of the procedure, electrodes are implanted deep into the patients’ brains to identify areas that are prone to seizure. These electrodes allowed the researchers to record firing patterns from single and multiple neurons during memory forming and retrieval events. “I think the surprise here is the deep specificity,” said Larry Squire, University of California San Diego and San Diego VA Medical Center. “One would have said, ‘Yes, the hippocampus is going to be active during learning and active again during retrieval,’ but this is saying a lot more than that. It is saying not only is that true but, in detail, that particular neurons that fire during learning fire again [during memory],” Squire told ARF. Squire was not involved in this work, but in the August 28 Neuron he reports another study on human memory, showing that the hippocampus and the perirhinal cortex cooperate to improve memory strength (see below). Both these areas degenerate early in Alzheimer disease.
Fried and colleagues examined hippocampal firing in 13 epilepsy patients. First author Hagar Gelbard-Sagiv and colleagues recorded electrical activity when the volunteers were shown audiovisual clips that lasted 5-10 seconds. Single or multicell units were considered responsive to any given clip if electrical activity in that unit consistently went up in all trials of that same clip. The units did not respond to all 48 clips shown. For example, one unit showed increased firing frequency in response to a clip from the TV show The Simpsons, and also responded, albeit more weakly, to a clip from Seinfeld. The average unit response rate was about 17 percent.
Gelbard-Sagiv and colleagues next took electrical measurements when the volunteers recalled any clip. The researchers found that units that had responded to initial views of a clip also responded when the person recalled that same clip. In fact, the units began firing three seconds before the patients verbally reported that they were recalling that particular clip, and continued to fire for several more seconds. In contrast, neurons that had not responded to specific clips during the initial viewing also did not respond when those clips were recalled.
All told, the data indicates that the same neurons fire during recall as fired when the memory was first made. This held true for neurons in the hippocampus and the entorhinal cortex. It was not true for units in the medial frontal lobe, even though they, too, fired when clips were first viewed. “Our results from conscious human patients, who can spontaneously declare their memories, now directly link free recall and neuronal replay in hippocampus and entorhinal cortex,” write the authors. Previous MRI studies hinted as much, but lacked the spatial resolution to be certain (see, e.g., Polyn et al., 2005; Kahn et al., 2004; Nyberg et al., 2000; and Wheeler et al., 2000).
Are these results applicable to every one of us, or are they an artifact of epilepsy? The authors suggest the latter is highly unlikely because epileptic activity results from the correlated activation of a large number of neighboring neurons. “The neuronal responses reported here were extremely sparse and seen selectively in individual neurons out of dozens of non-responsive neurons that were recorded in their immediate vicinity,” write the authors.
Fried and colleagues note that this sparse firing pattern is reminiscent of the responses of hippocampal “place” cells in rodents. The theory goes that a rodent’s environment is mapped, in Cartesian fashion, onto the neurons of the hippocampus. Specific place cells are thus activated when the animal is in a specific location, such as the start of a maze.
Previous experiments suggest that in animals, too, firing of hippocampal neurons might reflect free recall, not simply a response to some environmental cue. “However, the relationship of such replay in rodents to recall of past navigation events has been merely conjectural,” write Fried and colleagues. Since animals cannot tell us what they are thinking, connecting their firing patterns to free recall has been a challenge until now.
In the second science paper, researchers led by György Buzsáki at Rutgers, The Sate University of New Jersey, Newark, report that neuronal firing patterns that occur during a navigational task in rats also occur in the absence of environmental or body cues. Furthermore, those firing patterns accurately predict what choice the animal will make when next presented with the navigation task, as if the animal is spontaneously planning its next move.
The key to this study was to “freeze” the animals during a delay period in a navigational task. In the task, the animals were allowed into a running maze where they had to make one of two choices when they encountered a T junction—they had to go either left or right. The animals were trained to run the maze again and again while alternating between arms. In the delay period between tasks, the animals were trained to run in a running wheel at a constant speed, facing the same direction. Under this “frozen” condition one might imagine that some environmentally controlled place cells would exhibit sustained activity while others are suppressed—unless firing patterns are not merely cued but are internally generated. The latter is what the researchers found.
When first author Eva Pastalkova and colleagues measured electrical activity in the “frozen” rats, rather than finding some constant electrical activity evoked by their environment, they found that the pattern of firing changed over time. “This was already very interesting because it showed that there are self-evolving sequences at work,” said Buzsáki, in an interview with ARF. “But the next thing was even more exciting,” he said. The researchers noted that the firing pattern depended on the outcome of the previous navigational task. If the animal had last chosen the left arm at the T junction, they found one firing pattern, while if the choice had been right, they found a different pattern. Some neurons, for example, almost exclusively fired before either right or left choice. “The final exciting part was that every now and then the animal makes errors, and it turns out that neurons predicted the error 20 seconds earlier,” said Buzsáki. In other words the neurons were already encoding the future plan.
The study indicates that place cells in the rat hippocampus are not merely responding to place. “This is a very nice study. It shows that place cells are not just place cells but are cells that can help construct a whole episode that includes information about the space but goes beyond space,” said Squire. According to Squire, there has been some debate between two different viewpoints. One emphasizes the spatial function of the hippocampus and spatial functions of the neuronal activity as predominant, and the other emphasizes the idea that, even in a rat’s world, cells in the hippocampus are recording information about whole episodes. “This study comes down squarely in support of the latter,” he said.
In his own study, Squire and colleagues were interested in finding out what activity in the brain predicts the success of memory. It is well known that structures in the medial temporal lobe are important for forming memories because patients who sustain damage in those areas of the brain have severe difficulties in making new memories. “The questions of interest these days have to do with the anatomical components of that region of the brain, which includes the hippocampus and adjacent structures of the parahippocampal gyrus and the perirhinal cortex,” said Squire. The issues have been whether these structures carry out similar or contrasting activities, or what kind of function they have in forming memories. “There has been some viewpoint that the perirhinal cortex part of the system is mainly interested in learning single items for later recall, whereas the hippocampus is interested in more exotic things, such as when something happens or where something happens,” said Squire.
To address this issue, first author Yael Shrager and colleagues used functional MRI measurements to correlate brain activation during memory tasks with the strength of subsequent recall. Squire explained that, conventionally, when researchers subtract out brain activity related to objects that are subsequently forgotten from activities related to remembered items, not much is left. “We call that a remember-minus, or forgotten subtraction. When that has been done in the past nothing much was seen,” said Squire.
So the group took a different tack, asking subjects to rate the confidence with which they recall different memories. Fourteen volunteers aged 18-34 were scanned as they studied a list of words, then they were given a memory test outside the scanner. Study participants rated the confidence of their memory on a scale of 1 to 6, 6 being 100 percent certainty. The researchers found that activation of both the hippocampus and the perirhinal cortex are predictive of the confidence, or strength, of memory.
Does this study tell us anything about the proposed different roles for the hippocampus and perirhinal cortex in memory? Squire said that it suggests the two regions are more similar than they are different. “It speaks against the kind of dichotomization that has been talked about and more for the idea that both structures are involved in forming memories for single items. That’s not to say that they are involved in exactly the same job, but to say that they cooperate and that both are involved in forming memories for new items,” he said.
Will any of these studies help researchers get a handle on Alzheimer disease, which attacks the hippocampus early on in the disease process? “The first answer is ‘not in any obvious way,’ but I would like to point out that we know so little about the brain that anything we learn about how it works is helpful,” said Squire. “It was not long ago that we didn’t know exactly where to target treatment and it was even less long ago that we learned there were structures other than the hippocampus that are affected,” he said.
“Why are memories so unreliable in the first place?” said Buzsáki. “If you go to a lecture, for example, your brain will listen to every word but will not remember all the details. Why not?” Buzsáki noted, for example, that there has been some research showing that memory can be improved by electrically mimicking brain waves that are generated during sleep, which is known to help memory consolidation (see Marshall et al., 2006). The best hope may be that any new information we learn about memories may help scientists find interventions that help strengthen their formation or retrieval.—Tom Fagan.
Pastalkova E, Itskov V, Amarasingham A, Buzsáki G. Internally generated cell assembly sequences in the rat hippocampus. Science 2008 September 5; 321:1322-1327. Abstract
Gelbard-Sagiv H, Mukamel R, Harel M, Malach R, Fried I. Internally generated reactivation of single neurons in human hippocampus during free recall. Sciencexpress 2008 September 4. Abstract
Shrager Y, Kirwan CB, Squire LR. Activity in both hippocampus and perirhinal cortex predicts the memory strength of subsequently remembered information. Neuron 2008 August 28; 59: 547-553. Abstract