Bad news for DARPA's RAM program: Electrical Stimulation of Entorhinal Region Impairs Memory

The neural machinery that forms new memories is fragile and vulnerable to insults arising from brain injuries, cerebral anoxia, and neurodegenerative diseases such as Alzheimer's. Unlike language, which shows a great deal of plasticity after strokes and other injuries, episodic memory — memory for autobiographical events and contextual details of past experiences — doesn't recover after permanent damage to the hippocampus and surrounding structures.¹ Is it possible to improve memory by directly stimulating specific regions in the medial temporal lobes (MTL), even in damaged or diseased brains?

Restoring Active Memory (RAM) is a DARPA research program that aims to enhance memory encoding and retrieval in military service members who have suffered traumatic brain injuries. The approach is to design an implant, or “memory prosthesis,” that will treat memory loss via electrical stimulation.

The end goal of RAM is to develop and test a wireless, fully implantable neural-interface medical device for human clinical use ... DARPA will support the development of multi-scale computational models with high spatial and temporal resolution that describe how neurons code declarative memories—those well-defined parcels of knowledge that can be consciously recalled and described in words, such as events, times, and places. Researchers will also explore new methods for analysis and decoding of neural signals to understand how targeted stimulation might be applied to help the brain reestablish an ability to encode new memories following brain injury.

Initial Funding

The first RAM awards went to teams led by investigators at University of California Los Angeles (Dr. Itzhak Fried, PI) and University of Pennsylvania (Dr. Michael Kahana, PI). DAPRA's contributions to the White House BRAIN Initiative are known for their ambitious overreach.²  The modest call for proposals requested the following:

Proposers should develop a computational model of human neural and behavioral function underlying declarative memories that can be explicitly recalled.  [sure, no problem!]
. . .

Researchers must propose a method for validating their model by demonstrating that the model can be used to restore declarative memories through neural stimulation (i.e., electrical, optical, chemical, etc). ... Efficacy of the model must be validated by demonstrating that human patients can explicitly retrieve the restored memories after at least 14 days.

Piece of cake, right?

from Fig 1A (Jacobs et al., 2016). Stimulation in the left entorhinal region did not improve memory in this patient.

Maybe not...


Initial Findings

The first step in this noble quest has been to stimulate the MTL in epilepsy patients who have electrodes implanted for another clinical reason: to monitor the location of their seizures. The UCLA group reported that stimulation of the entorhinal region improved spatial memory (Suthana et al., 2012), a finding that predated the RAM program. Let's take a closer look.

Human entorhinal cortex (Schröder et al., 2015)

The entorhinal cortex (EC), located deep within the temporal lobes, projects to the hippocampus, a key region for the formation of episodic memories.³ EC plays an important role in spatial navigation and is famous for containing a spatial map. The EC and its neighbors in the parahippocampal region also receive projections from neocortical association areas, thus serving as a convergence site for cortical input and a distribution center for cortical afferents to the hippocampus.⁴

from Fig. 1 (Eichenbaum, 2000). The anatomy of the hippocampal memory system.


Suthana et al. (2012) began by reviewing all the potential benefits of entorhinal stimulation:

In rodents, electrical stimulation of the perforant pathway, which originates in the entorhinal cortex and projects into the hippocampus, results in long-term potentiation, release of acetylcholine, and resetting of the theta phase, all of which are associated with improved memory. It has also been shown that electrical stimulation can enhance neurogenesis in the hippocampus. Whether direct stimulation of this entorhinal output to the hippocampus enhances learning is not known.

In the study, seven individuals with epilepsy performed a spatial task where they learned destinations within virtual environments. There were four blocks, each containing six different destinations (that repeated across blocks). In half the trials of blocks 1-3, electrical stimulation (at 50 Hz) was delivered to the EC or hippocampus. No stimulation was given in block 4.

Spatial learning was quantified by determining the actual path traveled by the participant, relative to the shortest possible path. This variable was called excess path length, with shorter “excess” path length indicating better performance. Latency to reach a destination was measured as well.

The graph below shows the results averaged across six patients with entorhinal stimulation. During the three “learning” blocks, stimulation made no difference for (A) latency or (C) excess path length. During the identically structured “retention” block (when no stimulation was actually given), there seemed to be a small difference, with shorter latency and smaller excess path length for the destinations that had been learned with stimulation. No differences in performance were found when the hippocampus was stimulated, which is a little odd. Previous studies have shown that direct stimulation of the hippocampus impairs memory.

Basically, it looks as if the participants were not learning at all without EC stimulation. But the benefits of stimulation were quite modest (p=.03 for both measures), and the error bars were large for non-stimulation trials. Will these findings replicate in a larger sample of patients?


New Findings

Jacobs et al. (2016) tested 49 patients across seven different hospitals and found that 50 Hz electrical stimulation of the entorhinal region during encoding impaired memory in both spatial and verbal tasks. The effects were modest (and not always significant), yet surprising in light of the results from Suthana et al. (2012):

Across all patients and both tasks, entorhinal stimulation impaired memory accuracy (as measured by MS) by an average of 9% (permutation p < 0.02; t[15] = 2.3, p < 0.02). Entorhinal stimulation impaired memory in both the spatial task (permutation p = 0.03; t[5] = 1.7, p = 0.08) and the verbal task (permutation p = 0.09; t[9] = 1.49, p < 0.09).

You can get an idea of the individual variability in the spatial task below, where ^†p < 0.1 and ^∗p < 0.05 (one-sided rank-sum test).

The impairments appeared to be more robust with hippocampal stimulation, in contrast to the lack of effect in Suthana et al.:

Stimulation in the hippocampus significantly impaired performance by 8% overall across both tasks (permutation p = 0.002; t[42] = 2.97, p < 0.003). This impairment was present separately in both the spatial task (permutation p < 0.05; t[22] = 1.94, p < 0.05) and the verbal task (permutation p < 0.001; t[19] = 2.3, p < 0.02).

You might notice from the df above that not all patients had electrodes located in the regions of interest: 28 subjects for hippocampus (43 sites) and only 12 subjects for entorhinal (16 sites).


Nothing is Ever Simple

Why the discrepancy between studies?? Jacobs et al. (2016) discussed some potential differences: number of participants, number of independent observations (i.e., greater statistical power in their study), a better test of MTL-based spatial memory, and duration of stimulation (fixed at 10 seconds per trial vs. longer and variable). They also ran a simulation of Suthana et al.'s statistical methods using similar data and reported that “an effect at least as big as the 64% EPL reduction they observed is found in 19% of randomly shuffled data” (meaning that the result is not statistically significant).

What does this mean for the RAM of the future? In an extensive review of the brain stimulation literature, Kim et al. (2016)...

...tentatively suggest that stimulating multiple memory nodes in concert could enhance cognitive processes supporting memory.

Thus, the stimulation studies published so far make the point that for effective modulation of memory performance to be achieved, a network perspective rather than a purely focal stimulation approach should be considered. Declarative memory relies on a distributed network of multiple neocortical and medial temporal regions that serve cohesive roles in memory processes...

ADDENDUM (Dec 24 2016): DARPA has responded, and they're still bullish on closed-loop stimulation for memory restoration. 

@neuroghetto While this particular stimulation protocol resulted in temporary memory impairment, other protocols show promising results.

— DARPA (@DARPA) December 23, 2016

@AdamJShriver To be clear, although that specific protocol resulted in temporary impairment, we're still bullish on closed-loop stimulation.

— DARPA (@DARPA) December 23, 2016

One promise of this technology is that when you forget where you went for lunch on Thursday, and what you ate, and where you sat, and what you wore, your implant will kick in and retrieve the memories for you. 

And in response to a reader question, an extended quote from the Kim et al. (2016) network approach is in the comments below.


Footnotes

¹ For an extreme example, see Patient H.M.

² see these posts by The Neurocritic: A Tale of Two BRAINS: #BRAINI and DARPA's SUBNETS and DARPA allocates $70 million for improving deep brain stimulation technology.

³ Synaptic connections in the hippocampus and entorhinal cortex.

from Fig. 1 of Dobrunz (1998). Lateral perforant path (dotted green) and medial perforant path (solid green) provide inputs from the entorhinal cortex to the dentate gyrus of the hippocampus. Perforant path axons form synapses onto dentate granule cells (lateral in yellow, medial in red). Axons from the CA3 region of hippocampus form synapses onto cells in CA1 (purple).


⁴ Functional overview of the extended hippocampal-diencephalic memory system.

Fig. 1 (Kessels & Kopelman, 2012).

References

Jacobs, J., Miller, J., Lee, S., Coffey, T., Watrous, A., Sperling, M., Sharan, A., Worrell, G., Berry, B., Lega, B., Jobst, B., Davis, K., Gross, R., Sheth, S., Ezzyat, Y., Das, S., Stein, J., Gorniak, R., Kahana, M., & Rizzuto, D. (2016). Direct Electrical Stimulation of the Human Entorhinal Region and Hippocampus Impairs Memory. Neuron, 92 (5), 983-990. DOI: 10.1016/j.neuron.2016.10.062

Kim, K., Ekstrom, A., & Tandon, N. (2016). A network approach for modulating memory processes via direct and indirect brain stimulation: Toward a causal approach for the neural basis of memory. Neurobiology of Learning and Memory, 134, 162-177. DOI: 10.1016/j.nlm.2016.04.001

Suthana, N., Haneef, Z., Stern, J., Mukamel, R., Behnke, E., Knowlton, B., & Fried, I. (2012). Memory Enhancement and Deep-Brain Stimulation of the Entorhinal Area. New England Journal of Medicine, 366 (6), 502-510. DOI: 10.1056/NEJMoa1107212


Further Reading

Restoring Active Memory Program Poised to Launch (July 9, 2014)

DARPA has selected two universities to initially lead the agency’s Restoring Active Memory (RAM) program, which aims to develop and test wireless, implantable “neuroprosthetics” that can help servicemembers, veterans, and others overcome memory deficits incurred as a result of traumatic brain injury (TBI) or disease.

UCLA and Penn will each head a multidisciplinary team to develop and test electronic interfaces that can sense memory deficits caused by injury and attempt to restore normal function. Under the terms of separate cooperative agreements with DARPA, UCLA will receive up to $15 million and Penn will receive up to $22.5 million over four years...
. . .

Unique to the UCLA team’s approach is a focus on the portion of the brain known as the entorhinal area. UCLA researchers previously demonstrated that human memory could be facilitated by stimulating that region, which is known to be involved in learning and memory. Considered the entrance to the hippocampus—which helps form and store memories—the entorhinal area plays a crucial role in transforming daily experience into lasting memories. Data collected during the first year of the project from patients already implanted with brain electrodes as part of their treatment for epilepsy will be used to develop a computational model of the hippocampal-entorhinal system that can then be used to test memory restoration in patients.
. . .

The Penn team’s approach is based on an understanding that memory is the result of complex interactions among widespread brain regions. Researchers will study neurosurgical patients who have electrodes implanted in multiple areas of their brains for the treatment of various neurological conditions. By recording neural activity from these electrodes as patients play computer-based memory games, the researchers will measure “biomarkers” of successful memory function—patterns of activity that accompany the successful formation of new memories and the successful retrieval of old ones. Researchers could then use those models and a novel neural stimulation and monitoring system ... to restore brain memory function.

DARPA Project Starts Building Human Memory Prosthetics (August 27, 2014)

“They’re trying to do 20 years of research in 4 years,” says Michael Kahana in a tone that’s a mixture of excitement and disbelief. Kahana, director of the Computational Memory Lab at the University of Pennsylvania, is mulling over the tall order from the U.S. Defense Advanced Research Projects Agency (DARPA). In the next four years, he and other researchers are charged with understanding the neuroscience of memory and then building a prosthetic memory device that’s ready for implantation in a human brain.

Work Begins on Brain Stimulator to Correct Memory (April 3, 2015)

If the Penn team is able to identify markers of memory formation, it will try to influence them by stimulating the brain with low doses of electricity. The goal is to test whether it’s possible to coax the brain’s circuitry into whatever state represents a specific patient’s best possible memory function.

Kahana, who is director of the university’s Computational Memory Lab, says it’s too soon to say whether the idea will work. “We want the brain to exhibit a certain pattern of electrical activity,” he says. “It’s a big leap [to say] we can somehow nudge it into that state by giving it a little push.”

Targeted Electrical Stimulation of the Brain Shows Promise as a Memory Aid (September 11, 2015)

. . .

Just over one year into the effort, the novel approach to facilitating memory formation and recall has already been tested in a few dozen human volunteers, said program manager Justin Sanchez. ...

The study aims to give researchers the ability to “read” the neural processes involved in memory formation and retrieval, and even predict when a volunteer is about to make an error in recall. The implanted electrodes also provide a means of sending signals to specific groups of neurons, with the goal of influencing the accuracy of recall.

Initial results indicate that it is indeed possible to capture and interpret key signals or “neural codes” coming from the human brain during memory encoding and retrieval, and improve recall by providing targeted electrical stimulation of the brain.

Top image (from Penn): Illustration showing placement of deep brain electrodes in an epilepsy patient.