Early memory phases, including immediate and short-term memory, are defined by unique molecular, cellular and system processes. By comparison little is known about late memory phases, including remote memory. Here, we devised a novel genetic screen that combines the advantages of both forward and reverse genetic approaches, and identified two new genes required for remote memory: integrin beta2 and steryl-O-acyl transferase 1. Importantly, our studies also demonstrated the feasibility of this approach since two mutants were found out of a total pool of 54 tested. It is estimated that there are 10,000 other mutant mice available from a number of international repositories (http://www.mmrrc.org/
), which would suggest that our screen could reveal a considerable number of additional genes required for remote memory. Additionally, we also uncovered evidence for genetic modifiers for remote memory mutations. Altogether, these genetic tools will undoubtedly be critical for the study of the molecular, cellular and systems underpinnings of this process.
The search for genes involved in memory began with a screen in fruit flies, ultimately determining that the cAMP signaling pathway is an integral component of memory formation 
. Since then studies in Drosphila 
, aplysia 
and rodents 
implicated a number of other genes in memory. To date, most of these studies have focused on mechanisms of memory acquisition and protein synthesis-dependent consolidation – events that are completed within the first few hours following acquisition. However, additional processing/maintenance of memory occurs after this initial period. For example, hippocampal memories (i.e. contextual conditioning) become gradually cortically-based 
. The results presented here demonstrate that integrin beta2 and steryl-O-acyl transferase 1 are required for processes that occur well beyond the initial acquisition and initial consolidation of memory, since the mutations that we identified affect 7-, but not 1-day memory.
Mice with knockout mutations, point mutations, chromosomal deletions and constitutive or regulated overexpression of transgenes were randomly selected and subjected to our primary screen. Our reverse genetic screen is not meant to substitute for detailed analysis of targeted mutations. Instead, it is similar to forward genetic approaches, including those that used the chemical mutagen ENU in searches for novel learning and memory loci 
. In both approaches, no a priori assumptions are made about the nature of the genes to be discovered or the mechanisms that connect them to the relevant phenotypes; both forward and reverse genetic screens begin with primary tests designed to weed out unrelated mutants, followed by a more extensive analysis of candidate mutants with a set of secondary tests. However, key differences confer each approach with advantages and disadvantages. For example, reverse genetic screens are limited by the mutants available for screening, while forward genetic screens can theoretically scan every gene in the genome. In contrast, the process of gene identification is far more laborious in forward genetic screens since a causative (usually point) mutation must be identified amongst the entire genome.
In our screen, we used contextual fear conditioning to identify genetic alterations in mice that affect remote memory. In addition to the individual mutant strains tested, our database includes a large number of wild type mice that provided an accurate description of normal distributions of the behavioral measures included in our screen. This wild type data set was useful in choosing which mutants to pursue in our secondary screen. For example, a considerable percentage of wild type mice (6–16%) in each of the genetic backgrounds failed to freeze in the 7-day test; therefore, in a group of eight mice with normal remote memory, only one or two mice would be expected to have a z-score outside of +/−1 standard deviation. A mutant strain for which the group of eight mice have a mean z-score outside of +/−1 is an ideal candidate for further analysis. Thus, we used these z-score limits as the thresholds to select strains to follow up in our secondary screens. This control data set also demonstrates the critical importance of genetic background in behavioral phenotypes, a point that has been highlighted in numerous previous studies 
. Indeed, our results emphasize this important fact, since not only did we find differences in performance between various wild type genetic backgrounds, we also saw dramatic phenotypic differences in the mutant strains carrying the same mutation but in a different genetic background as analyzed in detailed secondary screens. As we changed the genetic background of these two strains, their associated remote memory phenotypes also changed. This will provide an invaluable opportunity to clone loci that modify the function of the two genes identified in our screen.
To increase the specificity of the primary screen, we included two key controls: immediate and short-term memory tests. These controls allowed us to eliminate mutants with performance deficits (i.e. unable to show freezing responses) since such mutant strains would have resulted in deficits at all 3 time points tested. In our primary screen, four potential remote memory mutants were identified. Subsequent testing showed that three of these mutants had reproducible deficits, but one strain, H2-Dma−/−, had normal remote memory. The reason for this false positive is unknown but could be due to a number of factors, including the small number of mice tested in the primary screen. Consequently, data from primary screens is always verified and further explored in subsequent tests.
In addition to remote memory mutants, our primary screen identified several other potential categories of mutants. For example, we identified 5 mutations that showed normal immediate memory, but abnormal short-term memory, the memory profile we obtained in our primary screen for mice with hippocampal lesions. Although there are a number of possible explanations for this memory profile, it is possible that this class of mutants is enriched for mice with abnormal hippocampal function. In addition to loss of function mutants, our primary screen also detected mutant mice with seemingly enhanced memory in immediate or short-term memory. However, these phenotypes were not the primary target of our screen and therefore, they have not been verified or tested in secondary screens.
To characterize further the memory profile of the remote mutants identified in the primary screen, we examined the time course of memory loss of Itgβ2−/−
mutants (secondary screen). The results demonstrated that both the Itgβ2 and Soat1 strains have intact immediate memory and short-term (30 min or 2 h) as well as intact memory after the protein synthesis-dependent window (tested at one-day), but abnormal remote (seven-day) memory. These data indicate that events involved in the acquisition and protein synthesis-dependent memory consolidation occur normally, ruling out alterations in a number of hippocampal processes ranging from synaptic signaling to transcription/translation. Furthermore, the remote memory deficits are likely to be specific to cortical consolidation of hippocampal–dependent memories since 7-day memory for CTA was normal for both mutants. Accordingly, previous studies suggested that remote memory for contextual fear conditioning and CTA involve different cortical regions 
. Similar to the αCaMKII+/− hetero
zygous mutation, the Itgβ2−/−
mutations may affect physiological processes (i.e. temporal cortex LTP) required for remote memory without disrupting hippocampal-dependent memory formation 
. Alternatively, Itg
could be involved in other processes associated with later phases of memory consolidation, such as synaptic restructuring
Soat1 is a critical enzyme in cholesterol metabolism. The role of cholesterol at the synapse is multifaceted: it is specifically required for synapse formation in neuronal cultures
, for syntaxin I clustering
which is required for synaptic vesicle fusion and it may otherwise influence synaptic vesicles by binding the abundant synaptic vesicle proteins, synaptotagmin I and synaptophysin
. Since cholesterol cannot cross the blood-brain barrier, all CNS cholesterol is from de novo
synthesis requiring intact and active cholesterol metabolic machinery. Furthermore, deficits in cholesterol metabolism are often associated with mental retardation (reviewed in
), suggesting that cholesterol metabolism is crucial for normal brain development and function. Additionally, Soat1 has been associated with aging in microarray studies and with Alzheimer's in genetic studies
. Importantly, the Soat1 mice learnt and had normal 1-day memory suggesting that they had no major developmental defects. The deletion of Soat1 specifically causes memory impairment at remote time points.
mice also showed a remote memory deficit. Integrin β2 (CD18, LFA-1) is localized on leukocytes and, as a heterodimer, binds various ICAMs (IntraCellular Adhesion Molecules). This molecule has been examined in the context of CNS injury. Integrin β2 has a critical role in the phagocytosis of injured neurons
. It is possible that in the absence of injury, Integrin β2 may mediate normal remodeling of neuronal structure important for remote memory. Integrin β2 and ICAM1 are also expressed on activated microglia found in the vicinity of amyloid deposits in Alzheimer's disease. Interestingly, Itg
is located near the breakpoint for trisomy 21 and could thus contribute to Down's Syndrome (
). Taken together, these data indicate a possible contribution of integrin β2 to normal cognitive function via structural remodeling or pruning, two processes that may be required for remote memory.
The results of this study mark the beginning of a systematic genetic dissection of remote memory. We identified two novel remote memory mutants with very diverse cellular roles, Itgβ2−/− and Soat1−/−. Both of these mutants have intact memory acquisition and protein synthesis-dependent consolidation, but showed reproducible deficits in three separate remote memory tests. The approach we developed has the potential to identify many other novel genes since we only screened less than 1% of the current mouse mutant resource. These genes will be valuable tools to elucidate the molecular, cellular and systems processes underlying remote memory.