Possible Dosage Effect of Maternally ExpressedGenes on Visual Recognition Memory inPrader-Willi Syndrome

Beth Joseph,* Mark Egli, James S. Sutcliffe, and Travis ThompsonJohn F. Kennedy Center, Vanderbilt University, Nashville, Tennessee

Seventeen patients with Prader-Willi syn-drome (7 with paternal deletion of chromo-some 15q11-q13 and 10 with maternal unipa-rental disomy [UPD]), and 9 controlsperformed a computerized visual recogni-tion task. A series of color digital photo-graphs were presented; most were pre-sented twice, but the remainder appearedonly once. Photographs presented twicewere separated in their presentation by ei-ther 0, 10, 30, 50 or 100 intervening photo-graphs. Subjects indicated whether eachphotograph had been presented previously.This procedure was implemented twice,once using photographs of foods, and onceusing photographs of nonfood objects. Asthe number of intervening photographs be-tween the first and second presentation in-creased, subjects were less likely to remem-ber having seen the photograph before.Performance by UPD subjects was less af-fected by increasing the number of inter-vening photographs relative to the othertwo groups, suggesting they had superior vi-sual recognition memory. This raises thepossibility of a beneficial effect of havingtwo copies maternally expressed genes onchromosome 15. UBE3A is suggested as apossible candidate for this effect. Am. J.Med. Genet. (Neuropsychiatr. Genet.) 105:71–75, 2001. © 2001 Wiley-Liss, Inc.

KEY WORDS: Prader-Willi syndrome; vi-sual memory; imprinting;UBE3A

INTRODUCTION

Prader-Willi Syndrome (PWS) is a neurodevelop-mental disorder characterized by short stature, intel-lectual disability, and feeding problems in infancy fol-lowed by the onset of hyperphagia in early childhood.Approximately 70% of patients with PWS exhibit com-mon paternal-specific interstitial deletions affectinghuman chromosome 15q11-q13 [reviewed in Ledbetterand Ballabio, 1995; Nicholls et al., 1998]. Most of theremaining cases (∼25%) are caused by maternal unipa-rental disomy (UPD), indicating that PWS results fromdeficiencies of imprinted, paternal-specific gene ex-pression. A third class of defects, termed imprintingmutations, is comparatively infrequent (∼5%) and isoften caused by small, submicroscopic deletions of the15q imprinting center [Sutcliffe et al., 1994; Buiting etal., 1995]. Imprinting mutations similarly result in theloss of imprinted, paternal-specific gene expressionthroughout a 1.5 Mb PWS critical region [Sutcliffe etal., 1994].

Although paternal deficiencies result in PWS, iden-tical maternal deletions, paternal UPD or imprintingmutations give rise to a distinct clinical phenotype inAngelman syndrome (AS), characterized by profoundmental retardation, absent speech, seizures, movementdisorder, and other features [Williams et al., 1995]. Un-like the PWS phenotype that is caused by the defi-ciency of multiple contiguous genes in the PWS im-printed domain, AS can result from loss-of-functionmutations in a single gene, the E6-AP ubiquitin-protein ligase (UBE3A) [Matsuura et al., 1997; Kishinoet al., 1997]. UBE3A is imprinted with maternal-specific expression in brain and biallelic expression inperipheral tissues [reviewed in Jiang et al., 1998b].

Significant interest has arisen in defining potentialphenotypic differences between patients with PWS ofdifferent etiologies. PWS deletion subjects (PWS-DEL)are hemizygous for all genes in the ∼4 Mb deletion in-terval and therefore, have a presumptive decrease inexpression of nonimprinted genes outside of the ∼1.5Mb PWS critical region. PWS-UPD and imprinting mu-tation subjects are chromosomally balanced but wouldbe expected to have increased expression of any im-printed, maternally expressed gene. Therefore, anyphenotypic differences between these groups could be

Grant sponsor: NICHD; Grant number: P01 HD30329.*Correspondence to: Beth Joseph, Ph.D., John F. Kennedy Cen-

ter, Box 156 Peabody College, Vanderbilt University, Nashville,TN, 37203.

Received 18 February 2000; Accepted 17 July 2000

American Journal of Medical Genetics (Neuropsychiatric Genetics) 105:71–75 (2001)

© 2001 Wiley-Liss, Inc.

ascribed to these differences in gene expression in theregion.

Several marked phenotypic differences have beenfound between the PWS groups. In general, people withmaternal UPD appear to be less severely affected thanpeople with deletions. UPD patients have a facial ap-pearance that is “less typical of PWS,” higher IQ scores,a lower pain threshold, and tendency to skin-pick lessoften than deletion patients [Cassidy et al., 1997;Dykens et al., 1999]. Deletion patients self-injure alarger number of body sites than those with UPD [Sy-mons et al., 1999], have higher scores on the ChildBehavior Checklist and the Yale-Brown ObsessiveCompulsive Scale [Dykens et al., 1999], and are morelikely to be hypopigmented [Butler et al., 1986; Lai etal., 1993; Gillessen-Kaesbach et al., 1995]. Althoughdifferences in hypopigmentation are due to involve-ment of the OCA2 (or P) gene [Brilliant et al., 1994],the etiology of other phenotypic differences remains tobe determined.

The present study examined visual short-termmemory in people with PWS and either the chromo-some 15 deletion or maternal UPD. Visual recognitionmemory has been reported as a relative cognitiveweakness in children with PWS [Warren and Hunt,1981]. Whether this ability is equally impaired in bothdeletion and maternal UPD patients is unknown, be-cause no distinction was made between the PWSgroups in their report. The present study represented asystematic replication and extension of a study re-ported in Warren and Hunt. Their studies used picturerecognition tasks in children with PWS, children withundifferentiated mental retardation, and typically de-veloping children. The present study examined wheth-er the difficulty with picture recognition previously ob-served in young children with PWS continues intoadolescence and adulthood. In addition, the role of cuetype was examined by having the participants performthe picture recognition task with two sets of pictures(food or nonfood items). Finally, the effect of genotypeon visual recognition memory performance was exam-ined by including participants with PWS having thechromosome 15 deletion or maternal UPD.

MATERIALS AND METHODSSubjects

Three groups of subjects participated: three malesand four females with PWS and paternal deletion of15q11-q13 (PWS-DEL), five men and five women withPWS and maternal UPD (PWS-UPD), and three menand six women selected to be similar to the subjectswith PWS in chronological and mental ages, bodymass, and IQ. So that group differences in recognitionmemory measures would not be confounded by groupdifferences in IQ, subjects in each group were selectedon the basis of having similar IQ scores from a largersample of subjects participating in the John F.Kennedy Center’s Prader-Willi Syndrome programproject. Tukey’s Studentized Range (HSD) tests re-vealed no statistically significant differences betweenthe three groups in terms of age (PWS-DEL mean 425.4 yr, SD 4 8.0; PWS-UPD mean 4 24.3 yr, SD 4

6.1; Non-PWS mean 4 30.6 yr, SD 4 13.4), Full ScaleIQ (PWS-DEL mean 4 64.7, SD 4 10.0; PWS-UPDmean 4 65.7, SD 4 5.6; Non-PWS mean 4 68.0, SD 410.4), and Body Mass Index (PWS-DEL mean 4 30.9,SD 4 7.1; PWS-UPD mean 4 32.0, SD 4 8.9; Non-PWS mean 4 46.6, SD 4 9.0).

Apparatus

Experimental stimuli were color stock photographsof food or of commonplace household and office objects(e.g., paper clips) or animals presented on a laptop com-puter with a color display (Toshiba Tecra). Two experi-menters seated near the participant recorded the ver-bal responses on a data sheet.

Procedures

Each participant experienced two 20-minute ses-sions separated by a 10-minute break. During each ses-sion, one of two sets of photographs, either food photo-graphs or nonfood photographs, were presented. Eachset contained a total of 250 photographs: 100 pairs ofidentical photographs and 50 unique photographs. Thepaired photographs were separated in their presenta-tion by either 0, 10, 30, 50, or 100 intervening photo-graphs, or “lags.” There were 20 pairs tested at each“lag” value. The 50 unique photographs were presentedonly once and were interspersed among the paired pho-tographs. Unique photographs depicted objects verysimilar to objects depicted in the paired photographs.

Participants were informed that they would be view-ing a “slide show.” They were asked to view each slideand say “Yes” out loud if the photograph had appearedpreviously in the set and to say “No” out loud if it hadnot. The participants were given six practice trials inwhich they were shown photographs not used in theexperiment. The practice trials were repeated, if nec-essary, until the participant understood the task.

Each photograph was presented for 4 sec by using aMicrosoft PowerPoint file programmed to simulate thelook and sound of an actual slide presentation. As eachslide was displayed, the experimenter and a researchassistant recorded the participant’s spoken responses.After a 10-minute break, the participant returned tothe room and was exposed to the second set of pictures.

RESULTS

Interobserver reliability scores were computed by di-viding the number of interobserver agreements by thenumber of agreements and disagreements and multi-plying by 100. The mean reliability scores were 99.4%and 99.9% for the food-photograph and the nonfoodphotograph conditions, respectively.

In all three groups, the number of photographs cor-rectly identified as being previously viewed decreasedas lag value increased (F[4, 19] 4 4.44; P < 0.01). ThePWS-UPD group identified the most repeated photo-graphs correctly (94.4%, SD 4 8.4), followed by thenon-PWS group (86.0%, SD 4 15.9) and the PWS-DELgroup (81.7%, SD 4 18.4; F[2,22] 4 3.65; P < 0.04).There was no statistically significant effect of photo-graph type unique to subjects with PWS.

Two types of errors were possible on the picture rec-

72 Joseph et al.

ognition task. One was failing to recognize a previouslyviewed photograph. The second type was incorrectlyidentifying a photograph as being previously seenwhen it was presented for the first time (i.e., false-positive error). Because subjects in the PWS-UPDgroup emitted more false-positive errors than the othersubjects, data from all subjects were reanalyzed in ac-cordance with signal detection theory, a commonmethod of presenting data from recognition experi-ments. A d8 statistic [Swets, 1964] was derived for eachsubject on the basis of the number of target photo-graphs correctly identified and the number of false-positive responses. The upper portion of Figure 1 showsthe mean d8 value for the three subject groups at eachlag value for the food photograph condition. Data forthe nonfood photograph condition are shown in thelower portion of Figure 1.

Performance on memory tasks can be viewed in howwell the subject learned the test material (i.e., “encod-ing”) and the extent to which this learned material isretained over time (i.e., “working memory”). These

variables were summarized by fitting quantitative ex-pressions to the mean d8 values shown in Figure 1.Because there is no theoretical consensus as to thefunction best reflecting memory decay [Wixted andEbbesen, 1991], three commonly used forgetting func-tions, exponential, power, and exponential-power, werefitted to the data by using an iterative Quasi-Newtonmethod [Raner, 1992]. The exponential-power functionof the form:

d8 4 ae−b√L (1)

where d8 is the recognition measure, a is the y-axisintercept, b is the decay parameter, and L is lag value,provided the best overall fit as assessed on the basis ofthe residual sum of squares (i.e., proportion of varianceaccounted for). Eq. 1 has been applied to recognition-memory data in previous studies [e.g., Wickelgren,1972]. The parameter a is said to reflect encoding, i.e.,the degree of learning in memory at the start of theretention period (i.e., L 4 0). As seen in Table I, thisvalue was higher for the non-PWS group, suggestingthese subjects initially learned the test material betterthan those with PWS. The parameter b is the decayrate and reflects the strength of the memory trace overtime; the higher the value, the more rapidly memorydecays. The decay parameter b was lowest for the PWS-UPD subjects under both photograph conditions. Inother words, although subjects with the maternal UPDform of PWS learned the test materials to a similardegree as the PWS-DEL group and to a lesser degreethan the non-PWS group, they retained what they didlearn better than both groups.

A repeated measures ANOVA was performed on thedata obtained from the signal detection analysis underboth picture conditions. Evidence for distinct memorydecay functions was supported by a statistically signifi-cant group by lag interaction effects under both thefood photograph condition (F[8, 92] 4 2.56, P < 0.01)and nonfood photograph condition (F[8, 88] 4 2.05, P4 0.05). We then analyzed data from the three possiblepairs of subject groups. If increasing lag value affectedPWS-UPD subjects in a unique way, statistically sig-nificant group by lag interaction effects would be ex-pected only in comparisons involving this group’s data.This was the case under the food photograph condition(Non-PWS vs. PWS-UPD: F[4, 68] 4 3.68; P 4 0.01;PWS-DEL vs. PWS-UPD: F[4, 60] 4 3.87; P < 0.01) andin the non-PWS vs. PWS-UPD comparison for the non-food photograph condition (F[4, 68] 4 2.71; P 4 .05).Group by lag interaction effects were not statisticallysignificant in the three remaining analyses.

Fig. 1. Mean photograph recognition accuracy (d8) as a function of num-ber of photographs intervening between first and second presentation.Data under the food-photograph and nonfood photograph conditions wereobtained during two separate experimental sessions. Curves are exponen-tial-power functions (Eq. 1) fitted by using parameters presented in Ta-ble 1.

TABLE I. Parameter Estimates For Eq. 1 Fitted ToRecognition Performance Data

Non-PWS PWS-DEL PWS-UPD

Food photographsLearning (a) 4.23 ± 0.10 3.56 ± 0.04 3.34 ± 0.10Forgetting (b) 0.052 ± 0.005 0.073 ± 0.002 0.020 ± 0.005R2 0.98 0.99 0.90

Food photographsLearning (a) 4.17 ± 0.08 3.44 ± 0.28 3.66 ± 0.13Forgetting (b) 0.033 ± 0.003 0.034 ± 0.015 0.020 ± 0.006R2 0.97 0.66 0.78

Recognition Memory in Prader-Willi Syndrome 73

DISCUSSION

Picture recognition performance by people in thePWS-DEL group was less accurate than those in thenon-PWS group at each lag value. In addition to repli-cating the Warren and Hunt [1981] study, which didnot identify whether their subjects had the maternalUPD form or deletion form of PWS, we observed that atthe smallest lag values, recognition performance ofsubjects in the PWS-UPD group was comparable wasthose in the PWS-DEL group. At the highest lag val-ues, however, their performance was indistinguishablefrom subjects not having PWS. This observation wasreflected in parameters estimated from fitting Eq. 1 tothe data in Figure 1. The estimated decay parameterwas substantially lower in the PWS-UPD group rela-tive to the PWS-DEL and non-PWS groups, suggestingthat picture recognition by the PWS-UPD group wasmore resistant to the effect of increasing lag thanpeople with a deletion or people not having PWS.

Given an apparent UPD-specific effect, it seems rea-sonable to hypothesize that this effect may be attrib-utable to maternal-specific gene expression of UBE3Aor other maternally expressed genes in the region. Todate, there are two known imprinted, maternally ex-pressed genes within the 15q11-q13 region of rel-evance: the AS gene UBE3A, and another maternallyexpressed transcript termed PIX [Rougeulle et al.,1998], which apparently has no protein-coding poten-tial. UBE3A encodes the E6-AP ubiquitin-protein li-gase, which functions within the pathway of ubiquitin-dependent degradation of cellular proteins. Thisenzyme transfers an activated ubiquitin to target sub-strates, which are subsequently degraded by the pro-teasome [Scheffner, et al. 1995]. Expression studies ofUBE3A reveal that imprinting is restricted to brain. Inmouse, ube3a expression is completely repressed fromthe paternal homolog in hippocampus and cerebellarPurkinje cells, partially repressed in other structures(e.g., cortex) and apparently not at all in others [Al-brecht et al., 1997]. In human, studies have been re-stricted to cortical tissue but show a markedly de-creased level of expression in AS patient autopsymaterial, compared with controls [Rougeulle et al.,1997; Vu and Hoffman, 1997].

The idea that deficient maternal-expression ofUBE3A in the brain has detrimental effects on learn-ing and memory is well established. Loss of maternal-specific expression of UBE3A in AS results in a severephenotype, including profound mental retardation andthe absence of speech [Kishino et al., 1997; Cassidy andSchwartz, 1998], implying that the absence of brainUBE3A has deleterious effects on development. AS hasbeen modeled in mice, by the generation of animalswith a null allele of ube3a [Jiang et al., 1998a]. Mater-nally deficient animals mimic AS with inducible sei-zures and show deficits in context-dependent learningand long-term potentiation.

The present findings raise the possibility that excessmaternal expression of UBE3A, or other maternallyexpressed genes within the deletion interval, may havea beneficial effect on memory and, hence, on someforms of learning. Two alternative possibilities exist: (i)

“leaky” expression of imprinted, paternally expressedgenes from the maternal chromosome in UPD subjectsand (ii) an unknown effect resulting from altered levelsof expression of sense and antisense transcripts at theUBE3A locus, which have been proposed to function incompetition-mediated regulation of UBE3A imprinting[Rougeulle and Lalande, 1998]. Experiments in whichexcess maternal UBE3A expression is induced, such asin mouse models, might allow this hypothesis to betested. Furthermore, although maternal UBE3A ex-pression has been shown in brain structure relevant tomemory function (e.g., the hippocampus), understand-ing the targets of UBE3A and their function might per-mit us to evaluate whether excess expression ofUBE3A in these structures could affect neuronal pro-cesses involved in learning and memory.

REFERENCESAlbrecht U, Sutcliffe JS, Cattanach BM, Beechey CV, Armstrong D, Ei-

chele G, Beaudet AL. 1997. Imprinted expression of the murine Angel-man syndrome gene, Ube3a, in hippocampal and Purkinje neurons.Nat Genet 17:75–78.

Brilliant MH, King R, Francke U, Schuffenhauer S, Meitinger T, GardnerJM, Durham-Pierre D, Nakatsu Y. 1994. The mouse pink-eyed dilutiongene: association with hypopigmentation in Prader-Willi and Angel-man syndromes and with human OCA2. Pigment Cell Res 7:398–402.

Buiting K, Saitoh S, Gross S, Dittrich B, Schwartz S, Nicholls RD, Horst-hemke B. 1995. Inherited microdeletions in the Angelman and Prader-Willi syndromes define an imprinting centre on human chromosome15. Nat Genet 9:395–400.

Butler MG, Meaney FJ, Palmer CG. 1986. Clinical and cytogenetic surveyof 39 individuals with Prader-Labhart-Willi syndrome. Am J MedGenet 23:793–809.

Cassidy SB, Forsythe M, Heeger S, Nicholls RD, Schork N, Benn P,Schwartz S. 1997. Comparison of phenotype between patients withPrader-Willi syndrome due to deletion 15q and uniparental disomy 15.Am J Med Genet 68:433–440.

Cassidy SB, Schwartz S. 1998. Prader-Willi and Angelman syndromes:disorders of genomic imprinting. Medicine 77:140–151.

Dykens EM, Cassidy SB, King BH. 1999. Maladaptive behavior differencesin Prader-Willi syndrome due to paternal deletion versus maternaluniparental disomy. Am J Ment Retard 104:67–77.

Gillessen-Kaesbach G, Robinson W, Lohmann D, Kaya-Westeloh S, Pas-sarge E, Horsthemke B. 1995. Genotype-phenotype correlations in aseries of 167 deletion and non-deletion patients with Prader-Willi syn-drome. Hum Genet 96:638–643.

Jiang YH, Armstrong D, Albrecht U, Atkins CM, Noebels JL, Eichele G,Sweatt JD, Beaudet AL. 1998a. Mutation of the Angelman ubiquitinligase in mice causes increased cytoplasmic P53 and deficits of contex-tual learning and long-term potentiation. Neuron 21:799–811.

Jiang YH, Tsai T., Bressler J, Beaudet AL. 1998b. Imprinting in Angelmanand Prader-Willi Syndromes. Curr Opin Genet Dev 8:334–342.

Kishino T, Lalande M, Wagstaff J. 1997. UBE3A/E6-AP mutations causeAngelman syndrome. Nat Genet 15:70–73.

Lai LW, Erickson RP, Cassidy SB. 1993. Clinical correlates of chromosome15 deletions and maternal disomy in Prader-Willi syndrome. Am J DisChild 147:1217–1223.

Ledbetter DH, Ballabio A. 1995. Molecular cytogenetics of contiguous genesyndromes: mechanisms and consequences of gene dosage imbalance.In: Scriver CR, Beaudet AL, Sly WS, Valle D, Childs B, Vogetstein B,editors. The metabolic and molecular basis of inherited disease. NewYork: McGraw Hill. p 811–839.

Matsuura T, Sutcliffe JS, Fang P, Galjaard RJ, Jiang YH, Benton, CS,Rommens JM, Beaudet AL. 1997. De novo truncating mutations inE6-AP ubiquitin-protein ligase gene (UBE3A) in Angelman syndrome.Nat Genet 15:74–77.

Nicholls RD, Saitoh S, Horsthemke B. 1998. Imprinting in Prader-Williand Angelman-Syndromes. Trends Genet 14:194–200.

Raner K. 1992. A compiler for math equations. MacTech 8:24.

Rougeulle C, Cardoso C, Fontes M, Colleaux L, Lalande M. 1998. An im-

74 Joseph et al.

printed antisense RNA overlaps UBE3A and a second maternally ex-pressed transcript. Nat Genet 19:15–16.

Rougelle C, Glatt H, Lalande M. 1997. The Angelman syndrome candidategene, UBE3A/E6-AP, is imprinted in brain. Nat Genet 17:14–15.

Rougelle, C, Lalande, M. 1998. Angelman syndrome: how many genes toremain silent? Neurogenetics 1:229–237.

Scheffner, M, Nuber, U, Huibregtse, JM. 1995. Protein ubiquitination in-volving an E1-E2-E3 enzyme ubiquitin thioester cascade. Nature 373:81–3.

Sutcliffe JS, Nakao M, Christian S, Orstavik KH, Tommerup N, LedbetterDH, Beaudet AL. 1994. Deletions of a differentially methylated CpGisland at the SNRPN gene define a putative imprinting control region.Nat Genet 8:52–58.

Swets JA. 1964. Signal detection and recognition by human observers. NewYork: John Wiley & Sons, Inc.

Symons FJ, Butler MG, Sanders MD, Feuer ID, Thompson T. 1999. Self-

injurious behavior and Prader-Willi syndrome: behavioral forms andbody locations. Am J Ment Retard 104:260–269.

Vu, TH, Hoffman AR. 1997. Imprinting of the Angelman syndrome gene,UBE3A, is restricted to brain. Nat Genet 17:12–13.

Warren JL, Hunt E. 1981. Cognitive processing in children with Prader-Willi syndrome. In: Holm V, Sulzbacher S, Pipes P, editors. Prader-Willi syndrome. Baltimore: University Park Press. p 161–177.

Wickelgren WA. 1972. Trace resistance and decay of long-term memory. JMath Psychol 9:418–455.

Williams CA, Angelman H, Clayton-Smith J, Driscoll DJ, Hendrickson JE,Knoll JH, Magenis RE, Schinzel A, Wagstaff J, Whidden EM, Zori RT.1995. Angelman syndrome: consensus for diagnostic criteria. Angel-man Syndrome Foundation. Am J Med Genet 56:237–238.

Wixted JT, Ebbesen EB. 1991. On the form of forgetting. Psychol Sci 2:409–415.

Recognition Memory in Prader-Willi Syndrome 75