It can be a good thing to forget, especially when learned information becomes useless or overly troubling. Yet the mechanisms by which we forget are a mystery: Is forgetting a passive reversal of learning, or must the brain make an effort to purge memories? A new study on forgetting in fruit flies suggests the latter. In a paper published in the February 19 issue of Cell, Yi Zhong and colleagues at Cold Spring harbor Laboratory, New York, present evidence that forgetting of newly formed memories is an active process that depends on signaling through the small GTP-binding protein Rac. The researchers show that activating Rac activity can cause flies to forget recently learned odor-stimulus pairs more quickly, while inhibiting Rac causes the memory to linger. Rac is well known for its key role in cytoskeletal organization, and the study points to changes in actin polymerization as an important determinant of the persistence, or loss, of memory.

Although the work was done in flies, the results might have some relevance to humans, perhaps even to Alzheimer disease. Previous studies implicate Rac or its downstream partners in forgetting in mice (see ARF related news story on Sananbenesi et al., 2007 and Meng et al., 2002). Alterations in the Rac signaling pathway induced by Aβ oligomers have been tied to synaptic defects (Ma et al., 2008, and see ARF related news story on Zhao et al., 2006), and lack of Rac activation leads to dendrite loss in cells expressing presenilin mutants (see ARF related news story on Inoue et al., 2009).

To try to understand how flies forget, Zhong and colleagues started by screening for mutants with enhanced early memory (memory that can be measured after a single training session and decays within a few hours), reasoning that these flies might have defects in forgetting. After repeatedly finding genes that converged on the Rac signaling pathway, first author Yichun Shuai and coworkers set out to test the role of Rac itself through targeted expression of dominant mutants of either constitutively active or inactive Rac alleles. They found that inhibiting Rac in a subpopulation of mushroom body neurons could prolong the retention of an odor aversion memory for hours to days. On the flip side, increasing Rac activity accelerated memory decay. Rac had no effect on learning, and it modulated forgetting that occurred by passive decay, or by active training on a new odor. In support of a physiological role for Rac, the researchers showed that endogenous Rac is activated in fly neurons after training with a time course that coincides with forgetting.

Rac regulates many downstream effects in neurons, including cytoskeleton organization, and dendritic spine morphogenesis (Tashiro and Yuste, 2004), a process which may form the anatomical basis of memory (see ARF related news story). To investigate whether the rearrangement of the actin cytoskeleton was critical to Rac’s actions, the authors looked at the effects of mutating cofilin, an actin de-polymerizing factor that sits downstream of Rac. In the pathway, Rac activates the kinase PAK, which activates a second kinase, LIMK, which phosphorylates and inactivates cofilin. Shuai and colleagues showed that expression of a constitutively active cofilin mutant inhibited memory decay, similar to the effect of Rac inhibition. In addition, a Rac mutant that was unable to activate the PAK/LIMK pathway did not affect memory decay. The results argue that the Rac effect is specific, goes through PAK/LIMK/cofilin, and may involve actin-based changes in cell or dendrite morphology.

In an accompanying preview, Ronald Davis of the Scripps Research Institute in Jupiter, Florida, writes that Shuai and colleagues “have ushered in a completely new line of research—the cell biology of active forgetting.” Among the questions the study raises is whether Rac plays a role in forgetting only in transient memory, or in other types as well. Davis points out that the effect of Rac inhibition to increase memory in flies by increasing cofilin activity and thus presumably decreasing actin filament formation goes opposite to accumulated data showing that long-term potentiation and memory consolidation in mammals is promoted through actin polymerization. It remains to be seen if this discrepancy arises because of an incomplete understanding of the signaling systems involved, or because different stages of memory require different cytoskeleton arrangements, he writes.

Zhong says his group is very interested in looking at exactly what is happening at the dendritic level in flies during early memory and its reversal. In addition, he says, they are probing whether Rac can reverse long-term memories by targeting constitutively active Rac to the neurons responsible to see if that will erase consolidated memories. “We think that in neurons for long-term memory, there may be no such mechanism, and that is why the memories stay,” he said.

Greg Cole of the University of California at Los Angeles has studied the Rac/PAK/LIMK pathway in AD. He is not surprised that Rac, a protein that regulates multiple pathways related to synaptic plasticity in neurons, can affect memory, including memory decay rates. “The idea that increased Rac activity can accelerate memory decay is consistent with some Aβ effects where Aβ activates Rac,” he wrote in an e-mail to ARF. However, Cole continues, “I see Rac and downstream PAKs as central to too many processes to be good candidates for direct inhibition. Since Rac is normally inhibited by ser71 phosphorylation by Akt, my lab has looked at treatments like DHA and cocktails that can increase this natural inhibition in a more modulatory way. Other ways of increasing this may include exercise to increase of BDNF or other methods of improving insulin/neurotrophic factor coupling to Akt, which is uncoupled by IRS hyperphosphorylation in AD neurons. What we need is restoration of normal controls on Rac activity, including trophic factor control of inhibitory Akt input and less activation by Aβ.” (See further comment and citations below.)

Zhong reports that his lab is also starting to look at how Rac might affect memory in the context of Alzheimer’s, using their own fly model of Aβ pathology (see ARF related news story on Iijima et al., 2004).—Pat McCaffrey


  1. This fly paper catches genetic evidence that Rac inhibition slows memory decay, constitutively increased Rac activation accelerates memory decay, and that cofilin hyperactivation gives rise to the same phenotype as seen with Rac inhibition. The authors conclude that the “Rac-regulated PAK/LIMK/cofilin pathway might be critical in influencing memory decay.” The specificity for Rac activation defects in an active forgetting process relevant to stronger longer-term memory with repetitive learning is novel and interesting. To the extent that these observations can be generalized to mammals, they may relate to the acute and chronic soluble Aβ oligomer-induced dysregulation of Rac/PAK/LIMK1/cofilin signaling (Zhao et al 2006., Ma et al., 2008, Gureviciene et al., and other refs) with LTP deficits and enhanced LTD (Li et al., 2009) and synapse loss (Freir et al., 2010). Conversely, memory consolidation is also impaired along with enhanced LTP and reduced LTP when PAK is selectively genetically inhibited in forebrain (Hayashi et al., 2004).


    . Soluble oligomers of amyloid Beta protein facilitate hippocampal long-term depression by disrupting neuronal glutamate uptake. Neuron. 2009 Jun 25;62(6):788-801. PubMed.

    . p21-activated kinase-aberrant activation and translocation in Alzheimer disease pathogenesis. J Biol Chem. 2008 May 16;283(20):14132-43. PubMed.

    . Role of p21-activated kinase pathway defects in the cognitive deficits of Alzheimer disease. Nat Neurosci. 2006 Feb;9(2):234-42. PubMed.

    . Altered cortical synaptic morphology and impaired memory consolidation in forebrain- specific dominant-negative PAK transgenic mice. Neuron. 2004 Jun 10;42(5):773-87. PubMed.

    . Normal induction but accelerated decay of LTP in APP + PS1 transgenic mice. Neurobiol Dis. 2004 Mar;15(2):188-95. PubMed.

    . Synthetic amyloid-beta oligomers impair long-term memory independently of cellular prion protein. Proc Natl Acad Sci U S A. 2010 Feb 2;107(5):2295-300. PubMed.

    . Abeta oligomers inhibit synapse remodelling necessary for memory consolidation. Neurobiol Aging. 2010 Jan 22; PubMed.

  2. This paper has interesting implications for AD research. Early memory without consolidation is labile and prone to forgetting. It is generally believed that forgetting occurs either passively along time or is induced by irrelevant information (interference-induced). Whereas the molecular mechanisms underlying memory acquisition and consolidation have been extensively studied, much less is known about what happens during memory decay.

    In this paper by Yi Zhong and colleagues, forgetting in Drosophila is reported to be mediated by Rac, a key member of Rho family of GTPases required for multiple processes such as cytoskeletal remodeling, transcription, and vesicle trafficking. By genetically manipulating the activity of Rac in flies, the authors found that actin polymerization mediated by the Rac-PAK-LIMK-cofilin signaling pathway contributes to both passive and interference-induced memory decay. Upon expression of dominant-negative Rac, the memory decay of the flies (after a single training session in olfactory aversive conditioning) slowed down significantly. In contrast, forgetting was enhanced by constitutively active Rac.

    It has been proposed that forgetting could be beneficial for an individual’s survival under a changing environment, in which the initially acquired memory becomes “inappropriate” and needs to be replaced by new memory in accordance with the current environment. One would therefore anticipate that Rac could be crucial for acquiring conflicting memory by elimination of the pre-existing “irrelevant memory.” To test this hypothesis, the authors employed the reversal learning paradigm, in which flies were first trained by pairing electric shock with one odor and then exposed to a second odor without a shock. They were then retrained by switching the two odors. Remarkably, reversal learning was impaired by expression of dominant-negative Rac, but was enhanced upon expression of constitutively active Rac. Moreover, repetitive training, which results in the formation of long-lasting memory, was found to reduce the level of active Rac. In contrast, activated Rac was increased after reversal learning, which involves forgetting memory from the first learning event. This study, to my mind, provides compelling evidence that Rac activity is critical to memory decay.

    In a number of neurodegenerative diseases, especially Alzheimer’s (AD), patients have difficulty acquiring new memories and suffer loss of pre-existing memories. It is, therefore, interesting to explore if increased Rac activity might contribute to the memory loss observed in AD patients, and if Rac inhibition might be beneficial to alleviate their symptoms. A recent study found that Cdk5, a serine/threonine kinase that regulates the Rac-PAK signaling pathway and has been implicated in AD (Cheung et al., 2006), is involved in extinction of contextual fear memories (Sananbenesi et al., 2007). In this study, the authors reported that increased PAK activity is associated with fear extinction, and that Rac1 regulates membrane localization of the Cdk5 activator p35 and subsequently PAK activity. Notably, administration of either a Cdk5 inhibitor (butyrolactone I) or a Rac inhibitor (NSC 23760) facilitated fear memory extinction.

    The precise role of Rac, therefore, appears to be different in various types of memory loss, which might be explained by the distinct mechanisms underlying each of these processes (e.g., passive memory loss over time versus active-learning fear extinction). Nonetheless, it is generally believed that accumulation of the β amyloid (Aβ) peptide contributes considerably to the pathogenesis of AD, and Aβ peptide has been shown to increase actin polymerization via Rac1 (Mendoza-Naranjo et al., 2007), suggesting that Rac1 might indeed be involved in memory loss during AD pathogenesis.


    . Synaptic roles of Cdk5: implications in higher cognitive functions and neurodegenerative diseases. Neuron. 2006 Apr 6;50(1):13-8. PubMed.

    . Abeta1-42 stimulates actin polymerization in hippocampal neurons through Rac1 and Cdc42 Rho GTPases. J Cell Sci. 2007 Jan 15;120(Pt 2):279-88. PubMed.

    . A hippocampal Cdk5 pathway regulates extinction of contextual fear. Nat Neurosci. 2007 Aug;10(8):1012-9. PubMed.

    View all comments by Nancy Ip
  3. A distinct feature of the nervous system is the intricate network of synaptic connections among its neurons. The changes in the strength and efficacy of existing synapses, as well as remodeling of connectivity through the loss and gain of synapses in the neuronal network, are believed to be the basis of learning and memory in the brain. Interestingly, long-term potentiation has been associated with an increase in spine formation and spine head growth, whereas long-term depression has been associated with spine shrinkage and retraction (1). The morphology of dendritic spines is known to change in response to several factors, including learning, age, hormones, and disease conditions (2). In addition to their morphological plasticity, spine-like protrusions also display rapid motility, changing shape and size in a matter of seconds to minutes. This morphological plasticity suggests that long-term memory might be encoded by alterations in spiny structures and associated synaptic contacts. Collectively, these events are critically important in synaptogenesis, in modulating of existing synapses, as well as in long-term synaptic plasticity.

    Although the molecular mechanisms that underlie these morphological changes are not clearly understood, emerging evidence supports at least two important signaling pathways that have been linked to dendrite spine formation and Alzheimer disease etiology: 1) cAMP-dependent activation of PKA has been shown to be critical for the maintenance of the late phase of long-term potentiation, and downstream phosphorylation of CREB has been linked to formation of new spines. Interestingly, it has been shown that Aβ inhibits the PKA/CREB pathway (3). 2) The Rho family of small GTPases, well-known regulators of the actin cytoskeleton, profoundly influences spine formation. Among the members of this family, Rac1, Cdc42, Rnd1, and Ras promote spine formation and growth, whereas Rap and RhoA induce shrinkage and loss of spines. P21-activated kinase (PAK) is a downstream signaling effector of the Rho/Rac family of small GTPases and has been shown to be associated with spine formation and memory consolidation (4). A role of PAK in cognitive deficits of Alzheimer disease has also been reported (5).

    This recent paper by Shuai and colleagues (6) suggests that the act of forgetting might also be linked to activation of the Rac pathway, using a simplistic model of olfactory learning in the fruit fly Drosophila. With the help of genetic manipulation, the authors were able to distinguish changes in Rac activity during passive memory decay, interference learning, and reversal learning, which are three different forms of forgetting events. The authors demonstrate that reversal learning, which erases memory of the first learning event, increased the level of activated Rac. On the other hand, they show that multiple training trials suppress the level of Rac activation, suggesting that the Rac-dependent forgetting events might be overridden by mechanisms that facilitate memory formation. Their results on passive memory decay also implied that forgetting might be an intrinsic characteristic of initially acquired memory, with training inducing memory formation as well as a forgetting event, but in a different time scale. Therefore, they conclude that forgetting mechanisms are also likely favored through an enhancement of learning interference. In the Drosophila olfactory memory model, it appears that cAMP/PKA and Rac/PAK-dependent memory acquisition and forgetting events are independent, as suggested by this group and others (6,7). In a more complex system, as it has been proposed in the mammals, it seems that memory consolidation might mechanistically require both pathways (4,8,9). As demonstrated by several groups, Rac signaling cascade in the brain is directly linked to an increase of spine formation through subsequent activation of PAK leading to F-actin polymerization and changes in membrane morphology. Besides the known involvement of cAMP/PKA/CREB activation cascade, Rac/PAK-dependent cellular events appear to be also intimately associated with the process of memory consolidation, at least in rodents.

    It is very exciting to think that perhaps similar cellular pathways as the one described in the fly system may be relevant to human disorders associated with memory dysfunction. One of the known hallmarks of Alzheimer disease is that patients forget recent events, hence are unable to consolidate their new memory. In our lab, we have shown that lack of presenilin function or expression in cortical neurons produced an increase of Rac/PAK cascade activation, which was also associated with an increase of spine-like protrusions (10). Mutations in presenilin 1 (PS1) and presenilin 2 (PS2), respectively, cause autosomal-dominant early onset familial Alzheimer disease. PS1 is an essential component of the γ-secretase complex, the enzyme responsible for intramembraneous cleavage of amyloid precursor protein to generate toxic β-amyloid peptides.

    Are these signaling events meaningful in the context of Alzheimer’s? Perhaps. Recent studies support the idea that familial Alzheimer disease-linked mutations in PS1 might cause a partial loss of function (11,12). It still remains to be determined whether Rac/PAK signaling is altered in neurons expressing familial Alzheimer disease-linked PS1 variants. If this is the case, one might want to consider the possibility that changes in Rac/PAK signaling in neurons might represent one of the earliest cellular dysfunctions that is relevant to cognitive decline in Alzheimer’s.


    . Molecular mechanisms of dendritic spine morphogenesis. Curr Opin Neurobiol. 2006 Feb;16(1):95-101. PubMed.

    . Overview on the structure, composition, function, development, and plasticity of hippocampal dendritic spines. Hippocampus. 2000;10(5):501-11. PubMed.

    . Amyloid beta -peptide inhibition of the PKA/CREB pathway and long-term potentiation: reversibility by drugs that enhance cAMP signaling. Proc Natl Acad Sci U S A. 2002 Oct 1;99(20):13217-21. PubMed.

    . Altered cortical synaptic morphology and impaired memory consolidation in forebrain- specific dominant-negative PAK transgenic mice. Neuron. 2004 Jun 10;42(5):773-87. PubMed.

    . Role of p21-activated kinase pathway defects in the cognitive deficits of Alzheimer disease. Nat Neurosci. 2006 Feb;9(2):234-42. PubMed.

    . Forgetting is regulated through Rac activity in Drosophila. Cell. 2010 Feb 19;140(4):579-89. PubMed.

    . Olfactory memory formation in Drosophila: from molecular to systems neuroscience. Annu Rev Neurosci. 2005;28:275-302. PubMed.

    . Hippocampal LTP is accompanied by enhanced F-actin content within the dendritic spine that is essential for late LTP maintenance in vivo. Neuron. 2003 May 8;38(3):447-60. PubMed.

    . The molecular biology of memory storage: a dialogue between genes and synapses. Science. 2001 Nov 2;294(5544):1030-8. PubMed.

    . Steady-state increase of cAMP-response element binding protein, Rac, and PAK signaling in presenilin-deficient neurons. J Neurochem. 2008 Mar;104(6):1637-48. PubMed.

    . The presenilin hypothesis of Alzheimer's disease: evidence for a loss-of-function pathogenic mechanism. Proc Natl Acad Sci U S A. 2007 Jan 9;104(2):403-9. PubMed.

    . When loss is gain: reduced presenilin proteolytic function leads to increased Abeta42/Abeta40. Talking Point on the role of presenilin mutations in Alzheimer disease. EMBO Rep. 2007 Feb;8(2):136-40. PubMed.

    View all comments by Angele Parent

Make a Comment

To make a comment you must login or register.


News Citations

  1. Memories—Familiar Kinase, Cdk5, Limits Fear Extinction
  2. AD Pathology—Loss of Kinase Sends Synapses PAKing
  3. γ-Secretase Drives Spine Formation Via Novel Substrate
  4. Persistence of Dendrites Leads to Lifelong Memories
  5. What the Fly Forgot—Aβ Expression in <i>Drosophila</i>

Paper Citations

  1. . A hippocampal Cdk5 pathway regulates extinction of contextual fear. Nat Neurosci. 2007 Aug;10(8):1012-9. PubMed.
  2. . Abnormal spine morphology and enhanced LTP in LIMK-1 knockout mice. Neuron. 2002 Jul 3;35(1):121-33. PubMed.
  3. . p21-activated kinase-aberrant activation and translocation in Alzheimer disease pathogenesis. J Biol Chem. 2008 May 16;283(20):14132-43. PubMed.
  4. . Role of p21-activated kinase pathway defects in the cognitive deficits of Alzheimer disease. Nat Neurosci. 2006 Feb;9(2):234-42. PubMed.
  5. . Synaptic activity prompts gamma-secretase-mediated cleavage of EphA4 and dendritic spine formation. J Cell Biol. 2009 May 4;185(3):551-64. PubMed.
  6. . Regulation of dendritic spine motility and stability by Rac1 and Rho kinase: evidence for two forms of spine motility. Mol Cell Neurosci. 2004 Jul;26(3):429-40. PubMed.
  7. . Dissecting the pathological effects of human Abeta40 and Abeta42 in Drosophila: a potential model for Alzheimer's disease. Proc Natl Acad Sci U S A. 2004 Apr 27;101(17):6623-8. PubMed.

Further Reading

Primary Papers

  1. . Forgetting is regulated through Rac activity in Drosophila. Cell. 2010 Feb 19;140(4):579-89. PubMed.
  2. . Rac in the act of forgetting. Cell. 2010 Feb 19;140(4):456-8. PubMed.