Letters

7 pages
4 views

Common pathway in the medial temporal lobe for storage and recovery of words as revealed by event-related functional MRI

of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Description
Lesion studies have provided compelling evidence that episodic memory is dependent on the integrity of the medial temporal lobe (MTL). This role of the MTL in episodic memory has been supported by several neuroimaging studies during both episodic
Transcript
  Common Pathway in the Medial Temporal Lobe for Storage andRecovery of Words as Revealed by Event-Related Functional MRI Sander M. Daselaar, 1 * Dick J. Veltman, 2 and Menno P. Witter 3 ABSTRACT: Lesion studies have provided compelling evidence thatepisodic memory is dependent on the integrity of the medial temporallobe (MTL). This role of the MTL in episodic memory has been supportedby several neuroimaging studies during both episodic encoding and re-trieval. After two meta-analyses of positron emission tomography (PET)and functional magnetic resonance imaging (fMRI) studies, we investi-gated a possible dissociation within the MTL memory system in relation toencoding and retrieval processes. Based on previous reports that specifi-cally related the function of the MTL in episodic memory to successfulencoding and actual recovery of information, we applied event-relatedfMRI to compare successful encoding of words (ES) directly with success-ful recognition of those same words (RS). Our results did not indicate aclear dissociation between encoding and retrieval activations in the MTL.Instead, a region in the left MTL, covering the parahippocampal cortexand hippocampal formation, which was activated during ES almost com-pletely overlapped with the area that was activated during RS. An addi-tional region in the left anterior MTL, including the entorhinal cortex, wasfound to be activated exclusively during ES. Research has indicated that alarge percentage of cells in this region are particularly sensitive to therelative novelty of stimuli. Our results, therefore, suggest that the para-hippocampal/hippocampal region is involved in the formation and subse-quent reactivation of memory traces, whereas the activity observed in theentorhinal cortex may reflect elementary memory processes related tonovelty detection.  ©  2003 Wiley-Liss, Inc. KEY WORDS: medial temporal lobe; hippocampus; functional MRI;encoding; retrieval INTRODUCTION Lesion studies in both human and nonhuman primates have clearly dem-onstrated that the medial temporal lobe (MTL) is crucial to episodic mem-ory function (Murray et al., 2000; Stefanacci et al., 2000). This role of theMTL in episodic memory has been supported by a large number of neuro-imagingstudies,inwhichincreasedactivationhasbeenreportedintheMTLduring both episodic encoding (Stern et al., 1996; Gabrieli, 1997; Rom-bouts et al., 1997) and retrieval (e.g., Schacter et al., 1996; Nyberg et al.,1996; Daselaar et al., 2001). It has been proposed that different regions of the MTL mediate functionally different processes, each contributing tooverall episodic memory function. Based on meta-analy-sis of positron emission tomography (PET) studies, Lep-age et al. (1998) suggested that the posterior part of theMTLisassociatedmorewithretrievalprocesses,andthatthe anterior part is more involved in the encoding of study material. However, including functional magneticresonance imaging (fMRI) studies in their meta-analysis,SchacterandWagner(1999)reportedananteroposteriorgradient exactly opposite the one proposed by Lepage etal. (1998)In connection with these findings, although yielding mixed results, a number of neuroimaging studies havealso directly compared episodic encoding and retrievaltasks. In line with Schacter and Wagner (1999), Rom-bouts et al. (2001) reported posterior MTL activationduring encoding and anterior activity in association withretrieval attempt using pictures as study items. Using a modified artificial grammar-learning task, Dolan andFletcher(1999)reportedleftanteriorhippocampalactiv-ity in relation to encoding and posterior MTL activity associated with retrieval demands. In a PET study by Schacter et al. (1999), again using visual objects as study items, little indication was provided that encoding andretrieval are preferentially associated with activity in theanterior as opposed to the posterior MTL. Finally, a number of imaging studies did not report any activity inthe MTL comparing encoding and retrieval tasks (Mc-dermottetal.,1999a,b;Kohleretal.,2000;Lepageetal.,2001). Alikelysourcefortheseinconsistentfindingsisthat,inmost of these studies, items that were not recollectedduring the recognition phase, and novel items that werenot presented during the study phase, were also includedinthedataanalysis.Inarecentevent-relatedfMRIstudy, Wagner et al. (1998) reported that activity in the MTLduring encoding is specifically related to subsequent rec-ollection of the study material. Furthermore, in anotherevent-related fMRI study, Daselaar et al. (2001) foundthatactivityintheMTLduringretrievalwasdistinctively associated with correct recognition of study items as op-posedtocorrectrejectionofitemsthatwerenotobservedat study. Finally, Buckner et al. (2001) provided indica-tionsthatprocessessupportingepisodicencodingalsoacton novel items, which are presented during the retrievalphase of episodic memory tasks. Hence, these findingsindicatethatevidenceforafunctionaldissociationwithinthe MTL memory system regarding encoding and re- 1 Center for Cognitive Neuroscience, Duke University, Durham, NorthCarolina;  2 Department of Psychiatry, Vrije Universiteit Medical Center,Amsterdam, The Netherlands;  3 Department of Anatomy, Vrije Univer-siteit Medical Center, Amsterdam, The Netherlands Grant sponsor: Dutch Organization for Scientific Research (NWO).*Correspondence to: Sander M. Daselaar, Center for Cognitive Neuro-science, Duke University, Box 90999, LSRC Bldg., Rm B243N, Durham,NC 27708. E-mail: daselaar@duke.eduAccepted for publication 26 February 2003DOI 10.1002/hipo.10158 HIPPOCAMPUS 14:163–169 (2004) ©  2003 WILEY-LISS, INC.  trieval should be obtained by directly comparing the processesassociated with successful encoding of study items with the pro-cesses that support the actual recovery of those same study items.This issue was recognized only in the PET study by Schacter et al.(1999), so that studied and novel items were specifically separated.However,inthisstudy,thetrialswerepresentedinstimulusblocksthat consisted entirely of either studied or novel items, giving riseto obvious methodological problems.Inthepresentstudy,weusedevent-relatedfMRIwitharandompresentation of trials to compare successful encoding directly withsuccessful recognition of words within the same study and in thesame group of subjects. Furthermore, focusing on the MTL mem-ory system, we investigated a possible dissociation between encod-ing and retrieval processes. In addition, we wanted to assess possi-ble commonalities between these processes as well, based on thelack of anatomical or electrophysiological support for a dissocia-tion (Witter et al., 2000a;Suzuki and Eichenbaum 2000) and theinconsistent results obtained in the imaging studies on this topic. MATERIALS AND METHODS Subjects Twenty right-handed men aged 30–35 participated in thestudy. None of the subjects was taking psychoactive medication,and none reported any neurological or psychiatric impairment onageneralhealthquestionnaire.Thesubjects’informedconsentwasobtained according to the declaration of Helsinki, and the study  was approved by the ethical committee of the Vrije UniversiteitMedical Center.For the purpose of this study, subjects were seen twice at theoutpatient clinic. During the first visit, subjects practiced the en-coding task that was to be performed in the MR scanner (seebelow). During the second visit, within the following week, thesubjects performed an episodic memory task, while functionalMRI data were obtained. The memory task consisted of an encod-ing phase, a 10-min retention interval, and a retrieval phase. Magnetic Resonance Procedures Imaging was done on a 1.5-T Siemens Sonata (Siemens, Erlan-gen, Germany) scanner using a standard circularly polarized headcoil. Stimuli were generated by a Pentium PC and were projectedon a screen at the back end of the scanner table. The projectedimage was seen through a mirror mounted above the participant’shead. Two magnet-compatible four-key response boxes were usedto record the subject’s performance and reaction times. The sub- ject’s head was immobilized using foam pads to reduce motionartifact, and earplugs were used to reduce scanner noise. For eachsubject,aseriesofechoplanarimages(EPI)wereobtainedsensitiveto BOLD contrast, entailing a T2*-weighted gradient echo se-quence (repetition time  2.1 s; echo time  60 ms; flip angle  90°) consisting of transversal whole-brain acquisitions (20 slices,3  3-mm 2 in-plane resolution; 5-mm slice thickness; 1-mm in-terslice gap). Behavioral Task Stimuli consisted of 233 nouns ranging from 3 to 12 letterstaken from a standard Dutch dictionary; 71 of these words wereusedasdummytrials,andtheremaining162wordswererandomly assignedtobeusedeitherastargetsordistracters,yieldingdifferentstimulus sets for each individual subject (see below).During the encoding phase, we implemented an incidental-learning paradigm, in which no reference is made to the test phaseduring the study. Subjects were instructed to indicate whether a noun presented on the screen gave them either a pleasant feeling (bypressingaleft-handbutton),oranunpleasantfeeling(bypress-ing a right-hand button). Whenever a word was considered emo-tionally neutral, the subjects were instructed to make associationsto that word until either a pleasant or an unpleasant feeling hadbeen induced. Two different trial types were distinguished: 81study words, and 54 baseline items.During a baseline trial, subjects were presented with a cue topress either a left-hand button (“  left”) or a right-hand but-ton (“right  ”). Our choice for this baseline condition wasbased on a study carried out by Stark and Squire (2001), in whichless MTL activity was observed during baseline tasks requiring a nonmnemonic decision and a subsequent motor response thanduring an unconstrained, pure rest condition. Stark and Squire(2001) argued that such a rest condition may not be an optimalzero baseline condition, because it promotes all kinds of reflectiveprocesses,includingmnemonicmentalprocesses,andthatbaselineconditions that require a nonmnemonic decision are likely to ap-proach zero more closely.Before each experiment, the trials were randomly intermixedinto 27 blocks, consisting of three study words and two baselineitems. In this way, the encoding trials could be presented at a maximum of six times in succession, and the baseline items at a maximumoffourtimesinsuccession.Thiswasdonetoensurethatthe experimental frequency would be high enough to be able tofilteroutlow-frequencydriftsintheMRsignal.Presentationofthestimuliwasself-paced,althoughatimelimitof5swasemployedincase of nonresponses. The interstimulus interval (ISI), taken fromstimulus offset to the onset of the next stimulus, was 2 s. During theISI,theword“NEXT…”appearedatthebottomofthescreen.The number of functional scans was fixed to 245. Whenever thesubjects had completed the encoding trials, the remaining scantimewasfilledupwithadditionalbaselineitemsanddummytrials.For the purpose of the present study, only the scans that wereobtainedduringtheencodingtrialswereincludedinthedataanal-ysis; consequently the number of encoding phase scans variedacross subjects, ranging from 187 to 245 scans.Duringtheretentioninterval,astructuralscanwasmadeinvolv-ing a coronal inversion recovery prepared 3-D gradient echo, T1- weighted sequence (MPRAGE: inversion time: 300 ms; TR   15ms, TE  7 ms, flip angle  8°).The retrieval phase (verbal recognition) consisted of three dif-ferent trial types: 81 targets (study words), 81 distracters (novel words), and 54 baseline items (press left/press right). For the re-trieval trials, subjects were instructed to indicate whether they hadpreviously seen the word presented during the encoding task (left- 164  DASELAAR ET AL.  handed press) or not (right-handed press). The stimulation para-digm was similar to the one employed in the encoding task. Trials were randomly presented in a self-paced fashion in 27 stimulusblocks, consisting of three targets, three distracters, and two base-lineitems.Thetrialswereprecededbyfivedummytrialsconsisting of three words and two baseline items, because pilot experimentshad shown that many errors were made in the first few trials of thetask. The duration of the recognition task was fixed to 285 func-tional scans; therefore, the number of recognition trials completedvaried across subjects. Data Analysis Data were analyzed using SPM99 (Wellcome Department of Cognitive Neurology, http://www.fil.ion.ucl.ac.uk/spm). Afterthe first three volumes were discarded, time series were correctedfor differences in slice acquisition times and were realigned. Next,the EPI volumes were spatially normalized into approximate Ta-lairach and Tournoux space, defined by a standard SPM EPI tem-plate, and resliced to a resolution of 3  3  3-mm voxels (Ta-lairach and Tournoux, 1988). The data were spatially smoothedusing a Gaussian kernel of 8 mm.The event-related fMRI analysis was based on the assumptionthat individual hemodynamic responses summate in a practically linear fashion over time (Dale and Buckner, 1997; Buckner,1998). Evoked hemodynamic responses to event types were mod-eled as delta functions convolved with a synthetic hemodynamicresponse function in the context of the general linear model (Jo-sephs et al., 1997). For each subject, successfully encoded items(i.e., correctly recognized afterward) together with baseline itemsthat were presented during the encoding phase, as well as success-fully retrieved items together with baseline items presented during the retrieval phase, were modeled as separate sessions within a fixed-effects model, adding up to 38 sessions over 19 subjects intotal. We chose a fixed-effects analysis in the present study becausethis type of model permits conjunction analyses to assess the over-lap of encoding and retrieval processes in the MTL, as well asencoding/retrieval interactions. Conjunction analyses are prob-lematic within a random effects model, because for that type of model independent observations are assumed, whereas each sub- ject’s encoding/retrieval pair of images is not independent.Encodingsuccess(ES)wasdefinedassuccessfullyencodeditemsminus encoding phase baseline items, and retrieval success (RS) assuccessfully retrieved items minus retrieval phase baseline items.The main effects of ES and RS, and the differences and conjunc-tion between these main effects, were tested for significance em-ploying a threshold of   P   0.05. Although we focused specifically on the MTL, we applied a whole-brain false discovery rate (FDR)correction for multiple comparisons, which takes into account theexpectedproportionoffalsepositivesamongsuprathresholdvoxels(Genovese et al., 2002). The difference maps for ES versus RS andRSversusESweremaskedinclusively( P   0.005,uncorrected)by the main effects of ES and RS, respectively, to minimize contribu-tionsofpossibledeactivationsoftheoppositemaineffects(RSandES, respectively). RESULTS Because of technical difficulties, data from one subject wereexcluded from the analyses. Behavioral Performance The subjects performed highly on the episodic memory task  with accuracy scores of 90.3% (SD  6.9) for the studied itemsand 92.0% (SD    8.3) for the novel items. Response times forsuccessfully encoded items (1.45 s; SD    0.41) differed signifi-cantly (t[19]    4.29;  P     0.000) from the response times forsuccessfully retrieved items (1.16 s; SD  0.30). A similar differ-ence(t[19]  4.29; P   0.000)wasfoundcomparingtheresponsetimes for the baseline items presented during encoding (0.81 s;SD  0.13) with those presented during retrieval (0.58 s; SD  0.10). This finding indicates that the subjects put in more effortduring the retrieval task than during the incidental encoding task.No significant correlation was found between response times onthe encoding trials and test score. Imaging Results In agreement with previous studies on verbal episodic memory,both ES (Fig. 1a) and RS (Fig. 1b) yielded activity in the left, butnot in the right, MTL (Daselaar et al., 2001; Wagner et al., 1998).Interestingly, considerable overlap was found between the regionof activity during ES and RS, as indicated by the conjunctionanalysis (Fig. 1c). The region of overlap covered the parahip-pocampal cortex and hippocampus proper. However, we foundindicationsfordifferentialMTLactivityinrelationtoESandRSas well:activityinthemostanteriorpartoftheMTL,whichincludedtheentorhinalcortex,wasseenforthemaineffectofES,butnotforRS. This observation was statistically confirmed by a direct com-parison of ES and RS (Fig. 1d). No difference was found in theMTL contrasting RS versus ES. The main effects, overlap, anddifferences between ES and RS are summarized in Table 1 andFigure 1e,f. Figure 2 indicates time courses for the effects of inter-est, i.e., the areas of conjunction (Fig. 1c) and ES  RS (Fig. 1d). DISCUSSION Thecentralaimofthepresentstudywastoinvestigateapossibledissociation within the MTL memory system with regard to en-coding and retrieval processes. To this end, we applied event-related fMRI to compare successful encoding of words (ES) di-rectly with the successful recovery of those same words (RS), using a shared baseline condition, which consisted of left versus rightmotor decisions. Our choice for this baseline condition was basedonfindingsreportedbyStarkandSquire(2001)(seeMaterialsandMethods), which indicated that an unconstrained rest conditionmaynotbeanoptimalzerobaselinecondition,becauseitgivesriseto all kinds of reflective, including mnemonic, mental processes,and that baseline conditions that require a nonmnemonic decisionare likely to approach zero more closely. __________________  ENCODING AND RETRIEVAL PROCESSES IN THE MEDIAL TEMPORAL LOBE  165  Our results did not indicate a rostrocaudal distribution of en-coding and retrieval activation, neither as described by Schacterand Wagner (1999) nor as described by Lepage et al. (1998).Instead, the region in the left MTL, which included the parahip-pocampal gyrus extending into the hippocampal formation, that was activated during ES almost completely overlapped with theareathatwasactivatedduringRS.Accordingly,theseresultsdonotindicate that episodic retrieval involves memory processes that aredifferent in nature, and therefore can be dissociated from, episodicencodingprocesses.Asidefromthefactthattheexperimentalfind-ingsconcerningadissociationbetweenencodingandretrievalhavebeen quite contradictory, there is also no anatomical or electro-physiological support for this concept. The finding of largely over-lappingencodingandretrievalactivationsintheMTLratherseemsto indicate that episodic retrieval involves reactivation of memory traces that have initially been formed in the MTL during the en-coding phase of episodic memory. This finding is in line with a large body of anatomical and electrophysiological evidence indi-cating that encoding, storage, and retrieval are most likely medi-ated by neurons intermingled within the parahippocampal regionand hippocampal formation (Suzuki and Eichenbaum, 2000), to-gether forming networks involved in both encoding and recall(Wilson and McNaughton, 1994). An additional region in the anterior part of the MTL, whichincluded the entorhinal cortex, was found to be activated exclu-sively during ES. The anterior MTL was previously implicated in FIGURE 1. Statistical parametric maps (SPMs) of encoding and retrieval-related comparisons ( P  < 0.05, false discovery rate [FDR]-corrected) overlaying sagittal (a–e) and coronal (f) sections of the left medial temporal lobe. a) ES, encoding success; b) RS, retrieval suc-cess; c) conjunction ES and RS; d) ES-RS; e) aggregate image of comparisons: ES, yellow; RS, green; Conjunction, magenta; ES  RS,blue; f) see “e)”. 166  DASELAAR ET AL.  the successful encoding of words. Using in vivo electrophysiolog-ical recordings on epilepsy patients, Fernandez et al. (1999) foundthat encoding-related responses in the anterior perirhinal cortex  were greater for subsequently recollected versus forgotten words.These results were replicated in a recent study by Strange et al.(2002), in which high-resolution fMRI was used. Furthermore, inthese studies, as in our investigation, hippocampal activity during the encoding phase was found to predict successful retrieval. Thisfinding suggests that the anterior part of the MTL and the moreposteriorly located parahippocampal/hippocampal region operateon a functionally related basis. However, we found that these re-gions may make separate contributions to verbal encoding, sinceoverlapping activity in relation to ES and RS was observed in theparahippocampal/hippocampal region, but not in the anteriorMTL.XiangandBrown(1998)reportedthatalargepercentageof cells located in the anterior part of the MTL, including the ento-rhinal cortex, in monkeys is particularly sensitive to the relativenovelty of stimuli. In this respect, it is possible that the additionalactivity observed in the anterior part of the MTL during encoding success reflected the engagement of novelty detection processes.Even though corresponding activity was found in the MTL inrelationtoencodingandretrievalinthepresentstudy,thequestionremains what might account for the rostrocaudal encoding/re-trieval distribution reported by Lepage et al. (1998) and by Schacter and Wagner (1999). Apart from considerable differencesin experimental design, one obvious explanation is that differenttypes of study material were used in the imaging studies, which were included in their meta-analyses. To make any inferences re-gardingencodingandretrievalprocessesbasedonthesemeta-anal-yses, one must assume that the relative location of activity withinthe MTL is dependent solely on the type of memory process and isnot dependent on the type of study material. However, there issubstantial evidence that different regions of the MTL mediatestimulus-selectiveprocessingsuchthatobjectsmaybeprocessedby networks that differ to some extent from those mediating wordprocessing or the processing of faces (Leube et al., 2001). Suchproposed topological differences are to some extent in line withknown anatomical and electrophysiological data, which indicateorganizational differences between various MTL regions. A goodexampleisthefindingthatobjectmemoryisspecificallysupportedby normal perirhinal functioning (Buckley and Gaffan,1997,1998), whereas spatial information from the dorsal visualstream reaches the MTL predominantly through more posteriorportions of the parahippocampal gyrus (Witter et al., 2000b), where a large percentage of cells do not differentiate between pre-sentation of novel and familiar stimuli (Riches et al., 1991), butresponds to location or spatial stimuli (Rolls and O’Mara, 1995;Vann et al., 2000). Hence, the differential encoding/retrieval dis-tributions reported by Lepage et al. (1998) and by Schacter and FIGURE 2. Plots of encoding success (ES) and retrieval success(RS) indicate event-related data for the local maxima of the areas of conjunction (Figs. 1c, 2a, c; [x,y,z]: 27,  96,5) and ES  RS (Figs. 1d,2b, d; [x,y,z]: 36,  87,7) in terms of the percentage signal changeaveraged across subjects (relative to the global mean over voxels and scans) against peristimulus time (seconds), adjusted for confounds,and binned every 2 s. TABLE 1. Maxima of Regions Showing Significant (P  < 0.05, FDR-Corrected) BOLD Signal Rises inthe Left MTL ComparisonTalairach coordinatesZ-value Regionx,y,z (mm)Encoding success   baseline (ES)   15   15   14 4.14 HF/PHGRetrieval success   baseline (RS)   21   24   16 2.43 HF/PHG  21   24   6 2.56 HF/PHGConjunction ES and RS   21   24   16 3.86 HF/PHGES  RS   18   15   14 3.02 ERCMTL, medial temporal lobe; FDR, false discovery rate; HF, hippocampal formation; PHG,parahippocampal gyrus; ERC, entorhinal cortex. __________________  ENCODING AND RETRIEVAL PROCESSES IN THE MEDIAL TEMPORAL LOBE  167
Related Documents
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks