Published on

Published in: Health & Medicine, Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

  1. 1. REPORTS gotten experiences, with remembered experi- Building Memories: ences being associated with a greater posi- tive-going response over frontal and parietal Remembering and Forgetting of regions (4). However, ERP studies are char- acterized by limited spatial resolution. Thus, Verbal Experiences as Predicted the precise functional neuroanatomic encod- ing differences that predict whether a partic- ular verbal experience will be remembered or by Brain Activity forgotten are currently unknown. A second unanswered question concerns Anthony D. Wagner,* Daniel L. Schacter, Michael Rotte,† the exact roles of medial temporal structures Wilma Koutstaal, Anat Maril, Anders M. Dale, Bruce R. Rosen, in memory encoding. Lesion studies in hu- mans and other species indicate that medial Randy L. Buckner temporal regions are essential for the process- ing of experiences such that they can be A fundamental question about human memory is why some experiences are remembered at a later time (5). However, remembered whereas others are forgotten. Brain activation during word en- modulated medial temporal activation has coding was measured using blocked and event-related functional magnetic been notably absent in neuroimaging studies resonance imaging to examine how neural activation differs for subsequently that systematically varied the nature of cog- remembered and subsequently forgotten experiences. Results revealed that the nitive operations engaged during encoding ability to later remember a verbal experience is predicted by the magnitude of (2). Rather, parahippocampal gyrus, a sub- activation in left prefrontal and temporal cortices during that experience. These component of the medial temporal memory findings provide direct evidence that left prefrontal and temporal regions jointly system, has been indirectly implicated in promote memory formation for verbalizable events. memory encoding because parahippocampal activation is greater during the processing of Memory encoding refers to the processes by trial-by-trial comparison between specific en- novel stimuli relative to familiar stimuli (6 ). which an experience is transformed into an coding trials that lead to subsequent remem- These results raise the possibility that para- enduring memory trace. Psychological stud- bering and those that lead to subsequent for- hippocampal contributions to encoding may ies have shown that the memorability of an getting. Results from event-related potential be restricted to novelty detection processes. experience is influenced greatly by the cog- (ERP) studies, which allow for trial-by-trial To address these issues, the neural corre- nitive operations engaged during initial en- analysis, suggest that the neural signature lates of incidental word encoding were exam- coding of that experience, with semantic pro- during verbal encoding differs for subse- ined in two whole-brain functional magnetic cessing leading to superior memorability rel- quently remembered and subsequently for- resonance imaging (fMRI) studies. One ex- ative to nonsemantic processing (1). Func- tional neuroimaging studies have implicated left prefrontal cortex in verbal encoding: left Fig. 1. Statistical activation prefrontal activation is greater during seman- maps are shown for the blocked-design and event-re- tic relative to nonsemantic encoding (2), and lated data. Images are trans- left prefrontal participation decreases and verse sections for the data memorization is impaired when semantic en- averaged across subjects. The coding operations are disrupted (3). These left hemisphere of the brain studies have all relied on blocked experimen- corresponds to the left side tal designs, where trials from each encoding of the image. (A) In the blocked-design experiment, condition are presented sequentially, insepa- greater activation during rable from each other during the functional word processing relative to scan. While blocked designs allow compari- fixation was noted in the son between encoding conditions that yield, posterior and dorsal extent of on average, higher or lower levels of subse- left inferior frontal gyrus (A: quent recollection, they do not allow a direct 34, 6, 34 and 43, 6, 31; BA 44/6), right inferior fron- tal gyrus (B: 37, 6, 34; BA A. D. Wagner and M. Rotte, Massachusetts General 44/6), left lateral parietal Hospital NMR Center, Harvard Medical School, cortex (C: 28, 68, 43; BA Charlestown, MA 02129, USA, and Department of 7), anterior and ventral left Psychology, Harvard University, Cambridge, MA inferior frontal gyrus (D: 02138, USA. D. L. Schacter, W. Koutstaal, A. Maril, 46, 34, 15 and 43, 28, 12; Department of Psychology, Harvard University, Cam- BA 45/47), bilateral frontal bridge, MA 02138, USA. A. M. Dale and B. R. Rosen, operculum (E: left 31, 19, Massachusetts General Hospital NMR Center, Harvard 12, and 40, 25, 3; right 34, 19, 6; BA 47), left middle temporal gyrus (F: 59, 43, 3; BA 21), Medical School, Charlestown, MA 02129, USA. R. L. bilateral visual cortex (G: BA 17/18/19), parahippocampal gyrus near fusiform gyrus (H: 31, 43, Buckner, Massachusetts General Hospital NMR Cen- 18; BA 36/37/35), and fusiform gyrus (I: 37, 58, 15; BA 37). Other regions included the ter, Harvard Medical School, Charlestown, MA 02129, hippocampus ( 37, 15, 15), supplementary motor area (0, 6, 62; BA 6), medial superior frontal USA, and Department of Psychology, Washington gyrus ( 3, 6, 50; BA 6), and right lateral cerebellum. (B) Regions demonstrating greater activation University, St. Louis, MO 63130, USA. during semantic relative to nonsemantic processing included left frontal (A: 43, 9, 34 and 43, *To whom correspondence should be addressed. E- 13, 28; D: 40, 22, 21 and 40, 31, 12; E: 28, 22, 6), parahippocampal (H: 34, 40, 12), and mail: adwagner@nmr.mgh.harvard.edu fusiform (I: 43, 58, 9) cortices. (C) In the event-related study, comparison of word processing †Permanent address: Otto v. Guericke University, trials to fixation trials revealed many of the same regions noted in the blocked-design experiment. Neurology II, Magdeburg D-39120, Germany. Complete listings of stereotaxic coordinates are available from the author upon request.1188 21 AUGUST 1998 VOL 281 SCIENCE www.sciencemag.org
  2. 2. REPORTSperiment used blocked-design procedures to Posterior LIFG Fig. 2. Statistical activationinvestigate how systematic manipulation of maps encompassing frontal A regions that demonstrate athe encoding task affects prefrontal and me- greater response during thedial temporal activation, whereas the other encoding of words later re-used newly developed event-related proce- membered (high confidencedures (7) that allow direct comparison be- hit trials) relative to wordstween specific encoding trials that result in later forgotten (miss trials).subsequent remembering and forgetting. In Displayed at the left are trans-the blocked-design experiment, activation verse and coronal sections through the activation foci forduring performance of a semantic processing the event-related data aver-task (deciding if a word is abstract or con- aged across subjects. Greater Anterior LIFGcrete) was compared to that during a nonse- activation was noted in themantic processing task (deciding if a word is B posterior and dorsal extent ofprinted in upper- or lowercase letters). left inferior frontal gyrusTwelve normal, right-handed subjects were (LIFG) bordering precentral gy- rus (A: 50, 9, 34; BA 44/6),scanned while performing alternating task- the anterior and ventral extentblocks consisting of semantic processing, of LIFG (B: 50, 25, 12; BAnonsemantic processing, and visual fixation 45/47), and the left frontal(8, 9). The novelty of the words in the se- operculum (C: 31, 22, 6; BAmantic and nonsemantic blocks was equiva- 47). Time courses were derivedlent. Behaviorally, reaction times (RTs) were for each condition within a Frontal Operculum three-dimensional region sur-longer for semantic (873 ms) relative to non- rounding the peak voxel andsemantic (539 ms) decisions. Subsequent C reflect raw mean signal chang-memory was superior following semantic es. Regions were defined, using(85% recognized) than following nonseman- an automated algorithm thattic (47% recognized) processing (10). identified all contiguous voxels Many brain regions demonstrated signifi- within 12 mm of the peak that reached the significance level.cantly greater activation during word pro-cessing relative to visual fixation (Fig. 1)(11). These activations likely reflect process-es associated with memory encoding and alsomore general processes associated with stim-ulus perception and response generation. Toidentify regions that demonstrate differential Fig. 3. Activation maps andactivation during encoding conditions that the corresponding time cours-yield higher relative to lower subsequent es from temporal regions arememory, we directly compared the semantic A/B shown for the trial comparisonand nonsemantic processing conditions. Re- of remembered (greater re- sponse) to forgotten (lesser re-gions demonstrating greater activation during sponse) words. Temporal focisemantic processing included several areas in included a region ( 31, 46,left prefrontal cortex, as well as left parahip- 12) that encompassed para-pocampal and fusiform gyri (Fig. 1). Al- hippocampal gyrus (A: BA 36/though these results indicate that temporal 37/35) and the more medialand prefrontal processes influence the encod- extent of fusiform gyrus (B: BA 37), and a region that encom-ing of verbal experiences, they do not directly passed the lateral extent ofspecify the encoding differences that predict fusiform gyrus and portions ofwhether a specific experience will be later inferior temporal gyrus (C:remembered or forgotten. C 43, 55, 9; BA 37). Other In a second experiment, event-related f MRI regions, including visual (L.,was used while participants performed a single left) and motor (R., right) cor- tices, did not show modulatedincidental encoding task. The objective was to activation across remembereddetermine whether trial-by-trial differences in and forgotten trials.encoding activation predict subsequent memoryfor experiences even when the processing taskwas held constant. Thirteen normal, right-hand-ed subjects underwent six fMRI scans, eachconsisting of word and fixation events present-ed in a continuous series of 120 rapidly inter-mixed trials (12). During word trials, subjectsmade a semantic decision (“abstract or con-crete?”). Following the encoding scans, memo-ry for the words was assessed by a recognitiontest. Subjects indicated whether they recog-nized each test word as studied, reporting theirconfidence (high or low) when they recognized www.sciencemag.org SCIENCE VOL 281 21 AUGUST 1998 1189
  3. 3. REPORTS the word (13). Behavioral results indicated that temporal regions are engaged during the ex- 7. A. M. Dale and R. L. Buckner, Hum. Brain Mapp. 5, 329 subjects discriminated between previously stud- (1997); R. L. Buckner et al., Neuron 20, 285 (1998). perience. Although modulated parahip- 8. Informed consent was obtained from all subjects ied and unstudied words when responding with pocampal activation has not been noted in (seven men, five women, aged 18 to 29 years). Echo high confidence, but not when responding with many studies, our experiments demonstrate planar and conventional imaging was performed on a low confidence (14, 15). that left parahippocampal gyrus is more ac- 1.5-T GE Signa scanner with an ANMR upgrade. Im- aging procedures [see R. L. Buckner, W. Koutstaal, The f MRI data were analyzed by catego- tive during the encoding of verbal experienc- D. L. Schacter, A. D. Wagner, B. R. Rosen, Neuroimage rizing encoding trials based on whether the es that are later remembered relative to those 7, 151 (1998)] included collection of structural im- word was subsequently remembered or for- later forgotten, even though these two classes ages [radio frequency–spoiled GRASS (gradient-re- called acquisition in the steady state) sequence, 60- gotten on the postscan memory test. There of experiences were equally novel within the slice sagittal, 2.8-mm thickness] and echo planar were four trial types: high confidence hits, context of the experiment [see also (20)]. functional images sensitive to blood-oxygen level– low confidence hits, misses, and fixation. These results indicate that, although medial dependent contrast (118 sequential whole-brain ac- quisitions, 16 slices each, 3.125-mm in-plane resolu- Word processing relative to fixation resulted temporal regions are sensitive to stimulus tion, 7-mm thickness, skip 1 mm; T2*-weighted in greater activation in many brain regions, novelty (21), the role of parahippocampal asymmetric spin-echo sequence: TR 2 s, TE 50 replicating most of the regions noted in the gyrus in memory encoding extends beyond ms, 180° offset –25 ms). 9. During each of four scans, blocks were ordered: non- blocked-design study (Fig. 1). Importantly, novelty detection and encompasses more semantic (40 s), fixation (24 s), semantic (40 s), the event-related design also permitted iden- general encoding mechanisms. Parahip- fixation, nonsemantic, fixation, semantic. A brief (8 s) tification of regions that demonstrate differ- pocampal gyrus is the principal neocortical fixation block began each scan. During semantic and ential activation during the encoding of input pathway to the hippocampal region nonsemantic blocks, 20 words were visually present- ed: 10 abstract and 10 concrete nouns; half in upper- words subsequently remembered and those (22), and thus it is suitably situated to play an case and half in lowercase letters. Each word was subsequently forgotten. When comparing important role in memory formation. presented for 1 s followed by 1 s of fixation between high confidence hits to misses, greater acti- Parahippocampal and prefrontal regions words. Subjects responded by left-handed key press. During fixation blocks, a cross-hair (“ ”) was pre- vation was noted in multiple left prefrontal may act interdependently to promote the en- sented for the entire duration. regions (Fig. 2) and left parahippocampal and coding of event attributes important for con- 10. Alpha was set to P 0.05 for all behavioral analyses. fusiform gyri (Fig. 3) (16, 17). This pattern scious remembrance. Verbal experiences Response latencies differed across tasks [F(1,11) 89.97]. Memory was assessed in the scanner 20 to was independently present and significant for may be more memorable when semantic and 40 m later, using a yes-no recognition procedure (8). these regions when comparing high confi- phonological attributes of the experience are Subsequent memory differed across tasks [F(1,11) dence hits to misses within each of the word extensively processed via participation of left 69.50]. 11. Functional runs were averaged within each subject, types (abstract or concrete). The subsequent prefrontal regions (2, 23). Left prefrontal re- transformed into stereotaxic atlas space, and aver- memory effect was rather specific: other re- gions may serve to organize these attributes aged across subjects (8). Activation maps were con- gions active during word processing relative in working memory, with this information structed using a nonparametric Kolmogorov-Smirnov statistic to compare (i) word processing (semantic to fixation failed to demonstrate greater acti- serving as input to parahippocampal gyrus and nonsemantic) to fixation and (ii) semantic to vation during high confidence hits relative to and the medial temporal memory system nonsemantic processing. Peak activations were iden- misses (Fig. 3). (24). A specific experience may elicit the tified by selecting local statistical activation maxima Our results specify how the neural signa- recruitment of these processes to a greater or that were P 0.001 within clusters of five contigu- ous significant voxels. These criteria minimize false ture during encoding differs for events sub- lesser extent because of variable task de- positives, as verified using the logic of control func- sequently remembered and events subse- mands, shifts in subjects’ strategies, charac- tional runs [E. Zarahn, G. K. Aguirre, M. D’Esposito, quently forgotten. When task demands were teristics of target items, or attentional modu- Neuroimage 5, 179 (1997)]. 12. Subjects were six men and seven women (aged 18 to held constant across trials, similar regions lations. Regardless of the source of this vari- 35 years). Three additional subjects were excluded were engaged during the encoding of both ability, greater recruitment of left prefrontal because of excessively poor performance. Imaging remembered and forgotten words. However, and temporal processes will tend to produce procedures were similar to experiment one with the exception that imaging was performed using an echo the magnitude of activation differed across more memorable verbal experiences. planar T2*-weighted gradient echo sequence (3.0-T, remembered and forgotten experiences in an- 128 images, TR 2 s, TE 30 ms, flip angle 90°). atomically specific brain regions. These ef- During each scan, 40 abstract word trials, 40 concrete References and Notes word trials, and 40 fixation trials were rapidly inter- fects cannot be attributed to differences in 1. F. I. M. Craik and R. S. Lockhart, J. Verbal Learn. Verbal mixed, with each trial lasting 2 s. For fixation trials, performance accuracy during encoding be- Behav. 11, 671 (1972). the fixation point remained on the screen for the cause accuracy was comparable for high con- 2. S. E. Petersen, P. T. Fox, M. I. Posner, M. Mintun, M. E. entire 2 s. For word trials, the word was presented for Raichle, Nature 331, 585 (1988); S. Kapur et al., Proc. 750 ms followed by 1250 ms of fixation. Abstract, fidence hits and misses. One possible inter- Natl. Acad. Sci. U.S.A. 91, 2008 (1994); J. D. E. Gabrieli concrete, and fixation trials were pseudo-randomly pretation is that the present modulations re- et al., Psychol. Sci. 7, 278 (1996). For reviews, see L. intermixed with counterbalancing (each trial type flect time-on-task or duty-cycle effects (18), Nyberg, R. Cabeza, E. Tulving, Psychonomic Bull. Rev. 3, followed every other trial type equally often). 135 (1996); R. L. Buckner and W. Koutstaal, Proc. Natl. 13. Approximately 20 min later, subjects were adminis- such that subsequently remembered experi- Acad. Sci. U.S.A. 95, 891 (1998). tered a memory test consisting of 480 studied and ences are those that merely happened to be 3. P. C. Fletcher et al., Brain 118, 401 (1995); C. L. Grady 480 unstudied words. Words were presented individ- processed for a longer duration during learn- et al., Science 269, 218 (1995). ually with self-paced timing. Subjects responded ing. To examine the possible contribution of 4. T. F. Sanquist, J. W. Rohrbaugh, K. Syndulko, D. B. “high confidence studied,” “low confidence studied,” Lindsley, Psychophysiology 17, 568 (1980); E. Halgren or “new.” time-on-task, the event-related data were re- and M. E. Smith, Hum. Neurobiol. 6, 129 (1987); K. A. 14. An Item Type Response interaction [F(1,12) analyzed after matching the encoding RTs for Paller, M. Kutas, A. R. Mayes, Electroencephalogr. 22.97] revealed that studied items were endorsed as high confidence hit and miss trials. Even Clin. Neurophysiol. 67, 360 (1987); M. Fabiani and E. “high confidence studied” more frequently than were Donchin, J. Exp. Psychol. Learn. Mem. Cogn. 21, 224 unstudied items [52% and 7%, respectively; F(1,12) when RTs were matched, left prefrontal and (1995). 57.04], whereas studied and unstudied items were temporal regions still demonstrated signifi- 5. W. B. Scoville and B. Milner, J. Neurol. Neurosurg. similarly endorsed as “low confidence studied” [24% cantly greater activation during the encoding Psychiatry 20, 11 (1957); L. R. Squire, Psychol. Rev. and 20%; (F 1.0)]. The low confidence response 99, 195 (1992); N. J. Cohen and H. Eichenbaum, class likely reflects subject guessing and does not of items subsequently remembered than dur- Memory, Amnesia, and the Hippocampal System (MIT differentiate between encountered and novel stimuli. ing the encoding of items forgotten (19). Press, Cambridge, MA, 1993). 15. Encoding task performance was analyzed based on Our studies, together with previous results 6. C. E. Stern et al., Proc. Natl. Acad. Sci. U.S.A. 93, whether the words were subsequently remembered (2), suggest that what makes a verbal expe- 8660 (1996); E. Tulving, H. J. Markowitsch, F. I. M. with high confidence (“high confidence hits”), low con- Craik, R. Habib, S. Houle, Cereb. Cortex 6, 71 (1996); fidence (“low confidence hits”), or were forgotten rience memorable partially depends on the J. D. E. Gabrieli, J. B. Brewer, J. E. Desmond, G. H. (“misses”). Accuracy during encoding was comparable extent to which left prefrontal and medial Glover, Science 276, 264 (1997). for high confidence hits (88% correct), low confidence1190 21 AUGUST 1998 VOL 281 SCIENCE www.sciencemag.org
  4. 4. REPORTS hits (88% correct), and misses (89% correct) (F 1.0). 19: 31, 77, 34). The superior occipital and medial 20. J. B. Brewer, Z. Zhao, J. E. Desmond, G. H. Glover, Semantic decision RTs declined across trial types superior frontal activations can be seen in Fig. 2. Two J. D. E. Gabrieli, Science 281, 1185 (1998); G. Fernan- [F(2,24) 9.26]: RTs were longer for high confidence regions demonstrated less activation for high confi- dez et al., J. Neurosci. 18, 1841 (1998). hits (1000 ms) compared to low confidence hits (966 dence hits relative to misses: precuneus (BA 31: 3, 43, 21. R. J. Dolan and P. C. Fletcher, Nature 388, 582 (1997). ms) [F(1,12) 5.40], which were in turn longer com- 40) and left middle frontal gyrus (BA 9: 12, 31, 34). 22. W. A. Suzuki and D. G. Amaral, J. Comp. Neurol. 350, pared to misses (936 ms) [F(1,12) 3.91, P 0.06]. No regions demonstrated greater activation for low 497 (1994).16. The procedures for selective averaging and statistical confidence hits relative to misses, which is in accord 23. J. Jonides et al., Nature 363, 623 (1993); J. A. Fiez et map generation for rapidly intermixed trials are de- with the behavioral data indicating that these two trial al., J. Neurosci. 16, 808 (1996); R. L. Buckner, Psy- scribed elsewhere (7). Statistical activation maps types likely did not mnemonically differ. chonomic Bull. Rev. 3, 149 (1996); A. D. Wagner, J. E. were constructed based on the differences between 18. M. D’Esposito et al., Neuroimage 6, 113 (1997). A Desmond, J. B. Demb, G. H. Glover, J. D. E. Gabrieli, J. trial types using a t-statistic (7). Fixation trial events similar interpretation may be applicable to the re- Cogn. Neurosci. 9, 714 (1997). were subtracted from the word trial events (collaps- sults from the blocked-design experiment. However, 24. M. Moscovitch, J. Cogn. Neurosci. 4, 257 (1992). ing across subsequent memory). Miss trial events and greater left prefrontal activation during semantic 25. Supported by grants from the National Institute on high confidence hit events were subtracted from each processing has been noted even when the nonseman- Aging (AG08441 and AG05778), the National Insti- other, as were those for miss trial events and low tic processing task has a longer duty cycle [ J. B. Demb confidence hit events. Clusters of five or more voxels et al., J. Neurosci. 15, 5870 (1995)]. tute on Deafness and Other Communication Disor- exceeding a statistical threshold of P 0.001 were ders (DC03245-02), the Human Frontiers Science 19. RTs were matched as follows. First, the median RT considered significant foci of activation (7). Program, and the Deutsche Forschungsgemeinschaft across all trial types was determined for each (SFB426). We thank Y. and W. Jacobson and two17. In addition, modest but significantly greater activation subject. Trials with response latencies that fell anonymous referees for helpful comments on an for high confidence hits relative to misses was noted in below the median RT were selected and sorted earlier version of this manuscript, and C. Brenner and a more ventral extent of left inferior frontal gyrus based on subsequent memory. Selection of trials in C. Racine for assistance with data collection. [Brodmann’s area (BA) 47: 34, 31, 3], left precentral this manner resulted in matched RTs for the high gyrus (BA 6: 31, 0, 56), medial superior frontal gyrus confidence hit (852 ms) and miss (839 ms) trial (BA 8: 3, 28, 43), and left superior occipital gyrus (BA types [F(1,12) 2.32, P 0.15)]. 15 May 1998; accepted 20 July 1998 Prevention of Allogeneic Fetal in maternal uterus or embryonic tissues. These findings are consistent with the known Rejection by Tryptophan expression of IDO in human syncytiotropho- blast (6). Pregnant mice (n 8 to 32) carrying Catabolism syngeneic or allogeneic concepti were treated with 1-methyl-tryptophan or with David H. Munn, Min Zhou, John T. Attwood, Igor Bondarev, placebo, beginning at 4.5 dpc (10). Concep- Simon J. Conway, Brendan Marshall, Corrie Brown, ti were examined macroscopically and his- Andrew L. Mellor* tologically at various times during gesta- tion (11). At 6.5 dpc, mice from all treat- In 1953 Medawar pointed out that survival of the genetically disparate (allo- ment groups carried normal numbers of geneic) mammalian conceptus contradicts the laws of tissue transplantation. concepti and embryonic development was Rapid T cell–induced rejection of all allogeneic concepti occurred when pregnant normal (Table 1). At 7.5 to 8.5 dpc, the mice were treated with a pharmacologic inhibitor of indoleamine 2,3-dioxy- mean number of allogeneic concepti in fe- genase (IDO), a tryptophan-catabolizing enzyme expressed by trophoblasts and males treated with IDO inhibitor was re- macrophages. Thus, by catabolizing tryptophan, the mammalian conceptus duced significantly (P 0.01) and exten- suppresses T cell activity and defends itself against rejection. sive hemorrhaging surrounded most of those that remained (Fig. 2A). However, atMedawar (1) considered three mechanisms response to interferon- and other signals 7.5 dpc, most remaining allogeneic concep-that might explain the immunological par- from activating T cells, inhibit T cell pro- ti were developmentally normal (Fig. 2C),adox of fetal survival: (i) anatomic separa- liferation in vitro by rapidly consuming with rare embryos showing signs of degen-tion of mother and fetus, (ii) antigenic im- tryptophan (4, 5); some tissue macrophages eration. By 8.5 (Fig. 2F) to 9.5 dpc, all alloge-maturity of the fetus, and (iii) immunologic may use this immunosuppressive mecha- neic embryos showed signs of inflammation“inertness” (tolerance) of the mother. In nism in vivo. Because IDO is also ex- and progressive deterioration (12). After 9.5view of evidence that the entire repertoire pressed by human syncytiotrophoblast cells dpc, no allogeneic concepti remained in anyof maternal T cells specific for paternally (6 ) and systemic tryptophan concentration mice treated with IDO inhibitor. In contrast, theinherited major histocompatibility complex falls during normal pregnancy (7), we for- mean number of syngeneic concepti and the(MHC) class I alloantigens is transiently mulated the hypothesis that IDO expression developmental status of embryos were not af-affected and tolerized during pregnancy (2, at the maternal-fetal interface is necessary3), the first two mechanisms cannot explain to prevent immunological rejection of fetal 6.5 7.5 9.5 dpcfetal allograft survival, and attention has allografts. To test this hypothesis, we ex- s a s a s afocused on the third mechanism. Certain posed pregnant mice (carrying syngeneic ormacrophages, induced to express IDO in allogeneic fetuses) to 1-methyl-tryptophan, IDO a pharmacologic agent that inhibits IDOD. H. Munn, M. Zhou, J. T. Attwood, I. Bondarev, S. J. enzyme activity (8). ActinConway, B. Marshall, A. L. Mellor, Programs in Molec- First, IDO transcription during pregnancyular Immunology and Developmental Biology, Insti- (9) was assessed in females mated with CBAtute of Molecular Medicine and Genetics, and Depart- (syngeneic) or C57BL/6 (B6, allogeneic) Fig. 1. Analysis of IDO transcription duringment of Pediatrics, Medical College of Georgia (MCG), murine gestation. PCR products were generated1120 15th Street, Augusta, GA 30912, USA. C. Brown, males (Fig. 1). IDO transcripts were detected by RT-PCR amplification from RNA samplesDepartment of Pathology, University of Georgia, Ath- from 7.5 to 9.5 days post coitus (dpc) in all prepared from pooled syngeneic (s) or alloge-ens, GA 30602, USA. concepti but were not detected at 6.5 dpc. At neic (a) concepti, at the gestation times indi-*To whom correspondence should be addressed. E- later gestation times (10.5 and 13.5 dpc), IDO cated (9). Results are representative of threemail: mellor@immag.mcg.edu transcripts were detected in placenta but not separate experiments. www.sciencemag.org SCIENCE VOL 281 21 AUGUST 1998 1191