The fundamental nature of memory has eluded philosophers and scientists for over two millennia. As early as 350 B.C., Aristotle conjectured that the mind was like a blank wax tablet, or tabula rasa, imprinted with one’s experiences but only made decipherable by associations between the distinct ideas etched into the tablet . As modern neuroscience emerged in the 20th century, doctors learned more about how memories were stored by studying patients with memory impairments, such as Henry Molaison (known as H.M.), who was unable to form long-term memories after the removal of a part of the brain called the hippocampus . At the same time, scientists explored how memories are formed by training animals to perform specific behaviors in the lab. These studies suggested that memory is physically represented in specific parts of the brain, and gave rise to the theory that discrete neural networks or even individual brain cells could make up a physical memory trace, or engram.
Recent advances in neuroscience and biotechnology allow researchers to find and manipulate memory engrams in mice and rats. Using a technology called optogenetics that allows immediate and precise control over individual brain cells, researchers have developed methods to selectively activate a memory , engineer a false memory , and strengthen or weaken a specific memory  by delivering pulses of light to brain cells.
Controlling brain cells with light
Optogenetics, which draws knowledge from optics, genetics, and neuroscience, is one of the most transformative technologies to emerge in recent years. In order to understand how optogenetics works, it is first necessary to know how brain cells, or neurons, function (Figure 1).
Figure 1 ~ a) Neurons normally open ion channels in response to neurotransmitters. If enough ion channels open, the cell membrane depolarizes and the neuron fires. b) Light-sensitive ion channels such as channelrhodopsin allow passage of ions when the cell is exposed to specific colors of light.
Almost all cells in the body maintain a charge gradient (an imbalance in the concentration of positively and negatively charged ions) across their cell membranes. This gradient creates a polarized electric potential, or voltage. Neurons are one of the few cell types that are electrically excitable, meaning that when enough ions cross the membrane to reduce the ion imbalance to a specific threshold, the neuron fires. As a result, the cell rapidly opens ion channels and allows ions to rush inside that reverse the charge gradient.
When a neuron fires, the electrical signal is propagated down a long cellular extension called an axon to the axon terminal, where a neuron chemically communicates with its neighbor by releasing chemicals called neurotransmitters into a small space between the two cells, called the synapse. The neurotransmitters consequently cause ion channels to open in the neighboring neuron and the process repeats itself, transmitting the signal down a chain of neurons.
Optogenetics uses special ion channels that open and allow ion passage upon exposure to light of a specific color. These channels are not present in most multicellular organisms, but can be added to these animals’ cells using genetic engineering. Unlike previous techniques of stimulating neurons, which used electrodes that indiscriminately shocked neurons into firing, optogenetics allows precise activation of the specific cell types into which the ion channels are introduced. Channelrhodopsin (ChR), which depolarizes cells when they are exposed to blue light, is the most popular of these ion channels among neuroscientists .
Optogenetics as a tool to activate and inactivate memories
In 2012, work from Nobel laureate Susumu Tonegawa’s lab at MIT harnessed optogenetics paired with innovative mouse genetic techniques to identify and manipulate the neurons that were activated when mice were trained to associate a distinctive environment, or context, with a mild foot shock . This type of training, called fear conditioning, causes mice to display a stereotypical fear behavior—freezing—whenever they are returned to the context in which the conditioning occurred. In Tonegawa’s experiments, the cells that were activated during the mice’s fear response were selectively labeled with ChR.
When those mice were moved to a new environment not associated with the shock, they distinguished between environments and did not freeze in fear as they had before. But when the specific cells that were previously activated by fear training were caused to fire by exposure to blue light, the mice in that new environment froze in fear. This suggests that activation of the cells representing an engram, even in an entirely different context, could cause mice to recall the fear memory.
One year later, the Tonegawa lab took this idea a step further by artificially installing a false memory using similar methods . The researchers first labeled hippocampal cells that fired repeatedly when the mice explored a novel context (Context A), and later selectively activated those cells with blue light while the mice received foot shocks in a different context (Context B). When mice were placed in Context A again, they froze in fear even though they had not physically been trained to fear that context. However, mice did not freeze when they were placed in a new and unique Context C. The researchers concluded that they had identified the memory engram-bearing cells for Context A, and by activating them while mice received foot shocks, the mice learned to associate Context A and foot shocks. However, the researchers did not identify how the cells encoding these two experiences – the cells that were activated by exploration in Context A and those that were activated by foot shocks in Context B – could have connected to make a new false memory.
In a new article published in the July 17, 2014 issue of Nature, a different group of researchers presented evidence for strengthening or weakening of an engineered memory using methods that reinforce or diminish the synaptic connections between neurons . Rather than giving mice a specific context to associate with foot shocks, Nabavi and colleagues optogenetically stimulated the neurons of rats within the auditory cortex and medial geniculate nucleus, regions of the brain involved in interpreting sounds (Figure 2). Following fear conditioning, rats associated the activation of these select auditory neurons with foot shocks, and when optogenetically stimulated, exhibited a fear behavior.
Figure 2 ~ In a recently published experiment that manipulates memories in rats by controlling synapse strength, researchers targeted the auditory cortex and medial geniculate nucleus (MGN) for channelrhodopsin expression. Using blue light, they stimulated the lateral amygdala, a fear memory center to which axons from the aforementioned auditory centers project. By weakening or strengthening synapses in the lateral amygdala with LTD or LTP, respectively, researchers were able to inactivate and then reactivate a fear behavior. Modified from Nabavi et al., 2014.
In order to test whether synaptic connections are important for the maintenance of a memory, the authors artificially harnessed two biological processes in memory formation called long-term potentiation (LTP) and long-term depression (LTD). LTP occurs when signals pass though a pair of neurons many times, causing the synapse to adapt and become strengthened – that is, signal transmission becomes easier. Alternatively, synapses can be weakened and become less likely to transmit a signal through LTD, when a neuron pair transmits weak signals that fail cause firing of the postsynaptic neuron.
When auditory cortex neurons in fear conditioned rats were stimulated with a specific pattern of slow pulses that imitated the natural process of LTD, the fear memory was inactivated and mice failed to respond to auditory cortex stimulation. But when the neurons were subsequently stimulated with a series of fast pulses that imitated LTP, the memory returned and rats displayed the fear behavior once more. Because LTP and LTD respectively strengthen or weaken the connection between neurons at their synapse, the results suggest that the strength of an associative memory can be modulated by control of synapses.
Because of these experiments, we now know that individual memories are stored within specific groups of neurons. Furthermore, as Aristotle theorized, individual memories may have little significance unless associations exist between them. For mice or rats, fear memories involve complex associations between how a foot shock feels and the context in which it took place. A fear memory cannot exist without recollection of either of these aspects, nor can it exist if the two aspects are unlinked.
Scientists have long recognized that synapses allow groups of individual neurons to form intricate circuits, and that these circuits can process more complex information than single neurons alone. Yet it has remained controversial whether synapses themselves are capable of storing information regarding memories. These recent findings suggest that synaptic connections may physically represent the associations so critical to useful memories.
Mark Springel is a first-year graduate student in the Biological and Biomedical Sciences program at Harvard Medical School.
 Aristotle. On the Soul (Book 3, Chapter 4). Cambridge: MIT, 1994. The Internet Classics Archive. Web. 18 July 2014. http://classics.mit.edu
 Stromberg J (2014, Jan 28). A Postmortem of the Most Famous Brain in Neuroscience History. Retrieved from: http://www.smithsonianmag.com/science-nature/postmortem-most-famous-brain-neuroscience-history-180949504/?no-ist
 Liu X, Ramirez S, Pang PT, Puryear CB, Govindarajan A, Deisseroth K, Tonegawa S (2012). Optogenetic stimulation of a hippocampal engram activates fear memory recall. Nature 484: 381-5
 Ramirez S, Liu X, Lin P, Suh J, Pignatelli M, Redondo RL, Ryan TJ, Tonegawa S (2013). Creating a False Memory in the Hippocampus. Science 341: 387-91
 Nabavi S, Fox R, Proulx CD, Lin JY, Tsien RY, Malinow R (2014). Engineering a memory with LTD and LTP. Nature 511: 348-52
 Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005). Millisecond-timescale, genetically targeted optical control of neural activity. Nat Neurosci 8: 1263-8
Implanting a false memory:
Manipulating a synthetic memory through LTP/LTD: