Can Experimentally "Created" Memories Help Explain Delusions?
July 31, 2013. Shades of Philip K. Dick and Total Recall! Researchers report creating a false memory in mice in a study published July 25 in Science. Though it’s a long way from the memory manipulations of science fiction, the study underscores the malleable nature of memory and may have implications for understanding schizophrenia.
Susumu Tonegawa of the Massachusetts Institute of Technology, Cambridge, and colleagues used optogenetics to artificially activate neurons in the hippocampus, a crucible for new memory formation. They first put the animals in a harmless environment, for which the hippocampus created a neural "memory trace." The researchers then activated that same network of neurons when the mice were in an environment where they received a mild foot shock. When the mice were returned to the harmless environment, they apparently remembered it with fear. The findings suggest that wayward neural signals can shape a memory for something that didn’t happen and support the idea that mistakes in memory formation could provide grist for the delusions and bizarre beliefs of schizophrenia.
“A delusion may just be a pathological memory,” said Carol Tamminga, a schizophrenia researcher at the University of Texas Southwestern Medical Center at Dallas, who was not involved in the study. Tamminga said the study probably has more relevance to post-traumatic stress disorder, but that it may also address how delusions are formed in schizophrenia, which is accompanied by hippocampal malfunctions (see SRF related news story and SRF Live Discussion).
“This paper suggests what the cellular process for that might be and points to the capacity of the hippocampus to play more than one memory at a time, and to combine them in a pathological way,” Tamminga told SRF. Tamminga has proposed that problems in the hippocampus could lead to spurious associations between unrelated things, which could build delusions (Tamminga et al., 2012).
The new study hinged on the ability to reactivate the very neurons involved in laying down a memory in the first place. Last year, Tonegawa’s group first harnessed these neurons by coupling expression of the light-sensitive channel, channel rhodopsin 2 (ChR2), to c-fos, an immediate early gene transcribed in response to neuronal activity (see SRF related news story). When mice explored a new place, the neurons involved in encoding aspects of that environment were labeled with ChR2 and left ready for light-induced reactivation. Last year’s study showed that reactivating these neurons evoked recall for that place, bolstering the idea that activity in that specific group of neurons constituted a memory.
Reactivating these neurons can also contribute to forming a new, albeit mistaken, memory, the new study finds. This result differs somewhat from another finding from last year, by Mark Mayford of The Scripps Research Institute in La Jolla, California, in which this paradigm did not lead to fear responses to the harmless environment but rather to a hybrid memory for both the harmless and harmful environments (see SRF related news story). That study used pharmacogenetic techniques, however, which may have led to more diffuse and longer-lasting activation than the optogenetic methods used by Tonegawa's group.
Remembrance of things false
First authors Steve Ramirez and Xu Liu used a viral vector to introduce the ChR2 gene to the dentate gyrus (DG) of the hippocampus, attached to a c-fos promoter. The goal was to have light-sensitive ChR2 expressed in neurons that became active as memories were being laid down. However, the genetic construct also contained a doxycycline-sensitive element, such that the researchers could keep the whole arrangement under wraps by continuously feeding the mice doxycycline.
To label neurons encoding a harmless environment (“context A”), the researchers took mice off doxycycline and allowed them to explore that environment. Afterwards, they put the mice back on doxycycline to limit ChR2 expression to those neurons. Later, mice underwent fear conditioning in which they were placed in a different environment (“context B”), where they received a mild foot shock. During this time, the researchers also activated the context A neurons in the DG with light so as to bring up a memory of context A. On the following day, the researchers tested the resulting fear memory, as measured by how much time the mice spent “freezing,” presumably in fear, when placed in different environments.
Ramirez, Liu, and colleagues found that the manipulated mice now spent about 25 percent of their time freezing when placed in context A. To make sure that the mice were not simply afraid of any environment now, the researchers put them in yet another environment (“context C”), where they spent less than 5 percent of the time freezing in place. Control mice that had not received a ChR2-containing construct but had gone through the same protocol exhibited this minimal, background level of freezing in both contexts A and C. This suggests that the researchers had been successful in "implanting" a fear memory for context A.
This sign of a memory for something that didn’t happen was specific to manipulations of DG activity, however. When ChR2 expression was induced in a different region of the hippocampal memory machine—CA1—followed by fear conditioning in context B, the mice did not show any increase in freezing in either context A or C. The researchers suggest that this might be due to CA1 neurons capable of responding to both contexts: The researchers found that 30 percent of CA1 neurons were activated by both contexts A and C, whereas only 1 percent of DG neurons showed this overlap. This suggests that coding for place in DG involves discrete sets of neurons, which then morph into a different pattern in CA1.
After all this manipulation, how did the mice react when they were put back in context B, the only place where they had actually experienced a shock? Normally, they would spend a whopping 70 percent of their time freezing in this environment. But the activation of context A encoding neurons seems to have interfered with the creation of a fear memory trace for context B: The experimental mice froze only 30 percent the next time back in that "dangerous" place. This suggests that the memory for context A interfered with the memory for context B.
Reactivating the context A neurons resembled a bona fide memory trace in other ways: Reactivating them in a completely new environment could elicit freezing, as though they induced recall; the number of neurons activated downstream in the amygdala was similar to that found for the genuine context B; and the mice could learn to avoid context A in a conditioned place-avoidance experiment.
Though the findings deal with a specific, simple form of learning, Tamminga said they might have implications for a person with a firmly held complex delusion. “For me, the promise of the study is that we'll soon have real physiological mechanisms that will help us explain what we now think of as bizarre mental events, but which are just mistakes of brain function,” she said.—Michele Solis.
Ramirez S, Liu X, Lin PA, Suh J, Pignatelli M, Redondo RL, Ryan TJ, Tonegawa S. Creating a false memory in the hippocampus. Science. 2013 Jul 26;341(6144):387-91. Abstract