interaction of CD45 with its ligand may in- Nature 382, 555 (1996).
Nature 382, 555 (1996).
duce its dilnerization and in turn regulate the
activity of Lck. In the absence of ligand, both
wild-type and mutant CD45 molecules are
catalytically active monomers. In the pres-
ence of a CD45 ligand, both wild-type and
mutant CD45 may dimerize, with different
consequences for Lck activity. In cells ex-
pressing wild-type CD45, the catalytic site
of each molecule would be blocked bv the
wedge containing glutamate 624 from the
partner molecule, inhibiting CD45 phos-
phatase activity. Consequently, Lck would
remain in the phosphorylated, inactive
conformation, and TCR signals would be
inhibited. In E624R-mutant CD45 inole-
cules, the wedge is altered so that the cat-
alytic sites are not occluded in the l~gand-
induced dimer. CD45 phosphatase activity
W O L I be
~ ~ retained and maintain Lck in its
active conformation.
'We chose to mutate glutamate 624 of
CD45 because it is analogous to aspartate 228
w~thin the putative inhibitory wedge of
RPTPa (8). Aspartate 228 of one monomer
contacts the mobile loo^ in the active site of
the opposlng monomer through a hydrogen
bond between the side chain carbosvl moietv
of aspartate 228 and a backbone amihe of thk
loop. This ~nteraction,along with other con-
tacts, would preclude the necessary movement
of the loop upon substrate b~nding,rendering
the phosphatase inactive. Mutation of gluta-
mate 624 of CD45 presumably disrupts the
analoeous
" interaction in CD45 dimers, there-
by allowing the mob~leloop to change con-
formation upon substrate binding, resulting in
an active CD45 phosphatase.
Ligand-induced dimerization plays a n es-
sential role in the regulation of receptor
tyrosine kinases, leading to autophosphoryl-
ation and activation of protein tyrosine ki-
nase activity (15). Ligand-induced dimer-
ization may also play a n essential role in the
regulation of RPTPs. However, instead of
leading to activation, dimerization of
RPTPs results in inhibition.
REFERENCES AND NOTES
1. R. J. Mourey and J. E. D~xon.Curr. Opin. Genet. Dev.
4. 31 (1994): B. G. Neel and N. K. Tonks. Curr Opin.
Cell Biol. 9 , 193 (1997).
2 K. K~sh~hara et a/.. Cell 74, 143 (1993); K. F. Byih et cerebral blood flow in extrastriate areas were significantly correlated with decreases in the
a/., J. Exp. Med. 183, 1707 (1996).
3. J. T. Pingei and M. L. Thomas, Cell 58, 1055 (1989); striate cortex. Extrastriate activity was also associated with concomitant activation of
I. S. Trowbridge and M. L. Thomas, Annu. Rev, lm-
munol. 12, 85 (1994).
4. G. A. Koretzky, J P~cus,M. L. Thomas, A. We~ss,
Nature 346, 66 (1990).
5 A. C. Chan. D. M. Desa~.A. Wess. Annu. Rev, lm-
munob 12. 555 (1994).
6. H. L. Ostergaard e t a / , ,Proc. Natl. Acad. Sci. U.S.A.
86, 8959 (1989): E. D. McFarland et a/., /bid. 90,
1402 (1993): M. S~eh,J. B. Bolen. A. Weiss. EMBO J.
12, 315 (1993);T. R Hurley, R. Hyman. B M. Sefton,
Mol, Cell. Biol, 13, 1651 (1993).
7. D. M. Desai, J. Sap, J Schlessnger, A. We~ss,Cell
73. 541 (1993).
8. A. M. B~lwes.J den Hertog, T. Hunter, J. P. Noel,
,iencemag.org
jugated goat antbody to mouse IgG. Cells were ana-
9. The EGFR-CD45 ch~mera(cons~st~ng of resldues 1 lyzed on a FACScan (Becton-D~ck~nson).
through 646 of the human EGFR and res~dues580 19. The ~ n t r a c e u a free
r Ca2'- concentrat~onwas mea-
through 1281 of human CD45, numbered after the s ~ g - sured w t h the calcum-senstve dye n d o - l as de-
n a sequence from sof form ABC) has been descrbed scr~bed(16).Cells (5 x l o 6 per m~lll~ter) were treated
(7).The E624A and E624R mutatons were ntroduced wlth ant~bodyto CD3 (mAb 235 at a 1 :3000 dlut~on
Into the w~ld-PqpeEGFR-CD45 chmera by ol~gonucle- of asc~tes)or EGF (100 ng/ml).
otde-based mutageness usng the polymerase cha~n 20. Cells were hatvested, washed t w ~ c ew ~ t hphosphate-
react~onand conf~rmed by nucleot~de sequencing buffered sal~ne(PBS), and resuspended at 2 x 1O8
H45.01 (4) cells (a CD45-def~cientder~vativeof the per m i i t e r ~nPBS. Cells were Incubated at 37°C for
HPB.ALL T cell n e ) were transfected w~thplasm~ds 15 min, For each sample, 2 x 1O7 cells were used,
encodng the EGFR-CD45 mutant ch~mer~c molecules. and an equal volume of st~mulusn PBS warmed to
Subsequent I m ~ t ~ nd~lut~on
g and seecton In geneticn- 37°C was added for the ndlcated t ~ m eso that f~nal
contalnng medum (2 mg/ml) y~eldedH45XLE624A.5 cond~tonswere as follows: cells, 1 x 1O8 per m~ll~lter;
(expressng the EGFR-CD45/E624A mutant chmera) 235 asc~tes(mAb to CD3) at 1 :500 d~lution;and EGF,
and H45XLE624R.3 (express~ng the EGFR-CD45/ 100 ng/ml. Cells were sed~mentedn a m~crofugeand
E624R mutant ch~mera).H45XL2 (expresslng the ysed ~n200 pI of y s ~ sbuffer [ I % Trlton X-100, 150
EGFR-CD45 wid-type chmera) was der~vedin a s m a r mM NaCl, and 10 mM tr~s(pH 8.0), supplemented
manner (7). Multiple clones expressng each of the chi- wlth protease and phosphatase inhib~tors as de-
meric molecules were solated and analyzed. scribed ( I l ) ] .Lysates were Incubated at 4°C for 30
10. R. Majet and A. Wess, data not shown. mln, followed by centr~fugationat 13,000gfor 10 mln.
11. D M . Desa~,J. Sap, 0 S~lvenno~nen, J. Schless- Nnety percent of the ysate was subjected to Immu-
nger, A. Wess, EMBO J. 13, 4002 (1994). noprecip~tat~onw ~ t h polyclonal rabb~t antibody to
12. A. Wess and D. R. Lttman, Cell 76, 263 (1994). ZAP-70 [I598 (17)] and prote~nA-Sepharose beads
13. A. Takeda. J. J. Wu, A. L. Maizel, J Biol. Chem 267, for 2 hours at 4"C, after wh~chmmune complexes
16651 (1992). were washed, Immune complexes and 10% of the
14. J. A. Ledbetter, G. L. Scheven, F. M. Uckun, J. B. untreated ysate were resolved separately by SDS-
mboden, J lmmunol 146,1577 (1991): J . Matve, G polyacylam~degel electrophores~sand transferred to
Rimon, P. Tatham, S Cockcroft, Eur J lmmunol 21 , polyinyl~denefluoride membranes (M~ll~pore). lmmu-
195 (1991), R. S. M t i e r et a/.,J. lmmunol. 153, 84 nobottng was done with mAb to phosphotyros~ne
(1994). (4G10; Upstate B~otechnology) or anti-phospho-
15 J. Schessinger and A. Ullrch, Neuron 9, 383 (1992). MAPK (New England Bioabs), followed by vsual~za-
16. G. Grynk~ew~cz, M. Poenie, R. Y . Ts~en,J Biol. t ~ o nby enhanced chem~lum~nescence (Amersham).
Chem. 260, 3440 (1985). The blots were str~ppedand reprobed w ~ t hantbody
17. D. Q~aneta/., J. Exp Med 185, 1253 (1997). to ZAP-70 (1598) or ant~bodyto MAPK (Santa Cruz
18. Cells were staned with a control monoclonal ant~body B~otechnology),respect~vey.
(mAb) [goat ant~bodyto mouse ~mmunoglobul~n G 21. We thank S. M. Fu for mAb 235. G. Seivant for help
(Caltag Laboratories)]: with LA22 (ant~bodyto EGFR; w ~ t he q u ~ b r ~ u bm~ n d ~ nstudes,
g members of the
Upstate Botechnoogy, Lake Pac~d,NYJ for the EGFR- Weiss lab for d~scuss~ons and assstance, and A.
CD45 ch~mera, w~th Hle-1 (Becton-D~ck~nson)for DeFranco for h s cr~t~cal readng of t h ~ smanuscr~pt.
CD45, and w~thLeu4 (antbody to C D ~ EBecton-D~ck-
; Supported in part by agrantfrom N H (t0A.W ) . R.M.
Inson)for theTCR. Cells were sta~nedat 4°C w~thsatu- IS supported by the NIH M e d c a Scentist Traning
rating concentrations of fluorescein ~soth~ocyanate Program.
(F1TC)~onjugated prlmary ant~body(control, He-1, and
Leu4) or primary ant~body(LA22)followed by FTC-con- 1 1 August 1997; accepted 6 November 1997
Dissociated Pattern of Activity in Visual Cortices
and Their Projections During Human Rapid Eye
Movement Sleep
Allen R. Braun," Thomas J. Balkin, Nancy J. Wesensten,
Fuad Gwadry, Richard E. Carson, Mary Varga, Paul Baldwin,
Gregory Belenky, Peter Herscovitch
Positron emission tomography was used to measure cerebral activity and to evaluate
regional interrelationships within visual cortices and their projections during rapid eye
movement (REM) sleep in human subjects. REM sleep was associated with selective
activation of extrastriate visual cortices, particularly within the ventral processing stream,
and an unexpected attenuation of activity in the primary visual cortex; increases in regional
limbic and paralimbic regions, but with a marked reduction of activity in frontal association
areas including lateral orbital and dorsolateral prefrontal cortices. This pattern suggests a
model for brain mechanisms subserving REM sleep where visual association cortices and
their paralimbic projections may operate as a closed system dissociated from the regions
at either end of the visual hierarchy that mediate interactions with the external world.
S i n c e its discovery in 1953 (1 ), the stage of lnovelnents (REb1 sleep) has been the sub-
sleep characterized by electroencephalo- ject of unremitting scientific investiga-
graphic desynchronization and rapid eye tion. Exceptional interest in REM sleep
SCIENCE VOL.279 2 JANUARY 1998 91
Table 1. Results of SPM contrasts and correlations between REMs malized to a mean of 50). Regions in which rCBF rates are correlated
and rCBF. Regions where rCBF values differ from baseline (REM-wake with REMs (REM Corrs) are tabulated with associated correlation
and REM-SWS) are tabulated with associated Z-scores and magni- coefficients. In each case Talaraich coordinates indicate local maxima
tude (Mag.) of rCBF differences (milliliters per 100 g per minute nor- or minima.

REM-wake REM-SWS REM Cons

Brodman
Regions
no.
X Y Z Mag. Z-score X Y Z Mag. Z-score X Y Z r
Extrastriate cortex
Fusiform, lingular gyms 19,37 24 -60 -8 4.22 3.10
Ventral lateral occipital-lnferotemporal cortex 19, 37 28 -66 4 2.62 3.30
Lateral occipital-middle temporal cortex 18 32 -72 8 3.37 3.14
Dorsal lateral occipital cortex 18,19 34 -82 12 3.64 3.10
Striate cortex
Calcarine cortex 17

Paralimbic cortex
Parahippocampalgyms- hippocampus 37 24 -20 -16 2.50 3.81
Prefrontal cortex
Lateral orbital cortex 10,11,47 38 38 -12 -5.10 -3.63
Dorsolateral prefrontal cortex 9,10,46 32 44 20 -2.78 - 7.98
'Left hemisphere only. t P < 0.001. SP < 0.01 (otherwise P < 0.05).

has been sustained in part because it is the teria. Moreover, the functional anatomy o f tices and their projections during REM
stage during which intense visual imagery- the visual system during R E M sleep is of sleep, compared w i t h both wakefulness and
laden dream activity is most prevalent (2). significant interest and may provide clues stage 3-4 sleep [slow wave sleep (SWS)],
However, the neural mechanisms that un- about the nature of this enigmatic "third and evaluated the regional cerebral corre-
derlie generation o f these images remain state" of consciousness (5). lates o f the REMs that characterize this
uncharacterized. W e used positron emission tomography sleep stage.
Although visual imagery (formation o f (PET) and H2150 t o measure regional ce- T e n healthy male volunteers (6) under-
an internally generated percept) and visual rebral blood flow ( C B F ) w i t h i n visual cor- went sleep deprivation and restriction pro-
perception (representation o f an externally
generated stimulus) are behaviorally dis-
tinct. the extent to which this distinction is
predicated o n activation of different por-
tions of the visual svstem is unclear. creases and decreases in
I t is n o t known, for example, whether rCBF during REM sleep
images can be evoked at higher levels o f the compared with postsleep
visual hierarchy-extrastriate cortices in- wakefulness (A) and SWS
cluding V4, V5, inferotemporal and occip- (B)are illustratedas well as A REM - Wake
itoparietal processing pathways, and their positive and negative cor-
relations between rCBF
anterior projections-or whether this pro-
values and REM density
cess requires the contribution of early visual (C). In (A) and (B) stage-
areas-V1 and V 2 regions of the striate specific changes in rCBF
cortex-that are normally involved in pri- were evaluated for AN-
mary visual perception. COVA corrected data sets
Previous functional imaging studies, per- [see (20)]; the resulting B REM- SWS
haps because they utilized different visual SPM (Z) maps are dis-
imagery-eliciting tasks during wakefulness, played on a standardized
have produced contradictory results (3, 4). magnetic resonance im-
Functional imaging during REM sleep af- aging (MRI) scan, which
was transformed linearly
fords a singular opportunity for study, be-
into the same stereotaxic - -75
cause it is characterized by naturally occur- (Talairach) space as the
ring visual imagery-laden mentation and is SPM (Z) data. Planes of C Correlatian RE& vs. rCBF
defined by well-established, measurable cri- section are located at - 10
mm (left), 0 mm (middle),
A. R. Braun, F. Gwadry, M. Varga, Language Section,
Voice Speech and Language Branch, National Institute and + I 0 mm (right) rela-
on Deafness and Other Communication Disorders, Na- tive to the anterior com-
tional lnstitutes of Health, Bethesda, MD, USA. missural-posterior commissural line. Values are Z-scores representingthe significance level of changes
T. J. Balkin, N. J. Wesensten, G. Belenky, ~epartmentof in normalized rCBF in each voxel; the range of scores is coded in the accompanying color table, with
Neurobiology and Behavior, Division of Neuropsychiatry, blue-white designatingZ-scores of -4.0 and below and with dark red designatingz-scoresof +4.0 and
Walter Reed Army Institute of Research, Washington,
DC, USA. above. In (C) rapid conjugate eye movements recorded during the 180 s after the start of the scan were
R. E. Carson, P. Baldwin, P. Herscovitch, PET Imaging summed and correlated with normalized rCBF images on a pixel-by-pixel basis. The map illustrates
Section, Clinical Center, Building 10, Room 1C401, Na- these correlation coefficients displayedon the standardizedMRI scan at the same planes of section. The
tional Institutes of Health, Bethesda, MD 20892, USA. range of coefficients is coded in the accompanying color table, with blue-white designating coefficients
'To whom correspondence should be addressed. E-mail: of -0.75 and below and with dark red designating coefficients of +0.75 and above. Location of local
abraun@pop.nidcd.nih.gov minima and maxima for Z-scores and correlation coefficients are summarized in Table 1.

SCIENCE VOL. 279 2 JANUARY 1998 www.sciencemag.org

cedures (7) before the scanning session ( 8 ) .
Scans were perfortnecl during polysomno-
graphically defined uninterrupted sleep
stages 3-4 anLl during REM sleep ( 9 ) . A
waking study a-as performed at the end of
the sleep period, after at least 15 mill of
continuous wakef~~lness.
Within-subject between-stage contrasts
(REM-wake; REM-SWS) were analyred hv
statistical parametric mapping (SPh'l) ( 1 0),
and correlations Pet\veen the freiluency of
REMs (REM dens it^) and rCBF during
REh'l sleep were examined (1 1 ). In addi-
tion, regional data were extracted from the
PET iinages ancl analy:ed by covariance
techniili~es(principal conlponent analyses
and regional intercorrelations) to assess
f~rnctionalconnections bet~veenregions of
the brain ( 12).
Comparison of REM sleep anLl waking
scans (Fig. 1'4 anil Table 1) revealed focal
activations of the extrastriate (fusiform, in-
ferotemporal, and lateral occipital) cortices
during REM sleep, tnanifest in cluster, of
significant spatial extent in hot11 hemi-
spheres (13). These clusters diLl not encorn-
nass the striate cortices icalcarine cortex and
contiguous portions of the cuneus, as Llelim-
ited in the Talairach atlas) where rCBF was
unchanged compared with n7akefulness. KO
hemispheral differences in the magnitude of
REM-wake REM-SWS
Contrast
Fig. 2. Results of principal component analyses
depcting the first unrotated component for REM-
wake and REM-SWS. Princpa component ana-
yss was carred out on a difference matrix gener-
ated for each contrast. Values represent loadings
for each of the 20 regons n which weghts ex-
ceeded 0.25 In absolute value. Extrastrate areas:
crces, fusform-~nferotemporal:triangles, lateral
occipital. Strate areas: squares. The frst cotnpo-
nent for REM-wake and REM-SWS contrasts ac-
counted for 80.896 and 78.2'0 of the total vari-
ance, respectively
a-n~w.sciencetnag.org* SCIENCE
rCBF activations were detected.
Because the use of resting" scans as hase-
line may introdlrce a possible confound
(, 1 4)-failure
, to detect differences in the
primary visual cortex coulci 1.e due to spon-
taneous production of visual imagery during
vvaking scans acciuired at rest-we contrast-
ed REM sleen n i t h S W S , a state in ~ v h i c h
spontaneo~rsvisual imagery is unlikely to
have occurreii ( 1 5 ) .
This analysis not only confirlned the
ilissociation of activity in striate a n ~ extra-
l trolateral (orbital) and dorsolateral prefron-
striate cortices but also sho\\.ed that the
striate cortex n.as cleactivated during REbI
sleep compareil n ~ i t hS W S (Fig. 1B and
Table 1 ) . Focal Lleactivation of the primary
visual cortex n.as ilelimited in a single clus-
ter of significant spatial extent (16). Fur-
thermore, rCBF in the striate cortex was
lower illrring REM sleep than 'luring S\VS
even 1v11en absolute values corrected for
partial pressure of C O , (PCO?) were evalu-
ateci (although ilifferences diLl not reach
statistical significance) (1 7).
In contrast. rCBF in extrastriate cortices
of the ventral processing stream-fusiform,
inferotemporal, and ventral lateral occipital
cortices-was significantly higher luring
REM sleep than Lluring SWS [Fig. 1B and
Table 1 (2-scores exceeded threshold in the
left hemisphere only)]; these regions n7ere
includeLl in a larger cluster of significant
activation ( 1 8 ) . rCBF in regions of the
a 0
-8
--
-2 1 0 I 2
Strlate cortex
- ' 2-
-2
-- - 2
Striate cortex
Fig. 3. Correatons between rCBF rates n strate
and extrastrate codices. Values represent stan-
dardzed dfference scores-that I S , increases or
decreases n rCBF durng REM sleep compared
witii wake (A) and SWS (B),Z-transformed in each
Instance, rCBF responses n extrastr~ate (Ta-
airach x = -30. !! = -88, r = -4 in A; x = -31.
y = -58, z = - 8 n B) and striate (x = -1, y =
-80. z = - 12. both A and B) cortices were neg-
atively correlated in both instances: r2 = 0.952.
REM-wake. P < 0.001' I-' = 0.911. REM-SWS,
P < 0.001
\'cJI.. 279
dorsal processing stream was unchanged.
Contrasts of REM sleep with either
\vakefulness or S\XIS (Fig. 1, A and B and
Table 1 ) also revealed significant changes
in regions that represent principal targets of
extrastriate projections, beyond the bound-
aries of t h e visual bystem (1 9). During REM
sleep, activity in the parah~ppocalnpalgyri
anLl the contlgirous portions of the hip-
pc>catnp~rsincreasecl in parallel with the
extrastriate cortices, and rCBF in the ven-
tal cortices was significantly Llecreased.
There alere n o l~etnispl~eral differences in
the magnitude of these changes (2G).
T h e sleep-stage contrast results were
closely paralleleil 13y the spatial distrib~rtion
of correlations lxtween REbI density and
rCBF (Fig. 1 C and Table 1 ) ; REM density
was positively correlateL1 with rCBF in ex-
trastriate cortices in 130th hemisnheres,, nar- L
titularly within ventral regions. In contrast,
REh'l density was negatively correlateil with
rCBF in the striate cortices. I n addition,
REbI densitv was nositivelv correlated rvith
rCBF in hippocampus and in parahip-
pocampal gyri but was negatively correlated
a.it11 rCBF in lateral orhital and Llorsolateral
nrefrontal cortices.
Results of the principal colnponent anal-
ysis are illustrated in Fig. 2. Striate and
extrastriate regions loaded o n the same
component in both instances but with op-
posite signs-that is, across iniliviiluals in-
creases in rCBF in extrastriate regions Llur-
ing REbI sleep (compareil n.it11 either wake-
fulness or S W S ) were associated with con-
comitant Liecreases in the primary visual
cortex. Permutation tests i~xlicateil that
these coinponents n.ere significant in each
instance (I' < 0.05, exact), and the inverse
relationsl~inbetween striate and extrastriate
loadings was stronger for ventral extrastri-
dte regions (21 ). Inverse relationships be-
tween rCBF responses in selecteil regions of
the extrastriate and striate cortices are illus-
trateil in Fig. 3 .
These reslrlts suggest
", that a f ~ ~ n c t i o n a l
ilissociation 17etween activity in the striate
extrastriate visual cortices characterizes,
and may constitute a definitlg feature of,
REbI slee~l.REM sleerl x a s associated a i t h
selective activation of regions that are locat-
ed in higher or "ilownstream" portions of the
visual hierarchy. These incl~rderegions that
may constitlrte h ~ r m a nanalogues of V4 and
5!\ as well as more anterior projections of
ventral object-processing and dorsal spatial-
processing pathways (22). REM-associated
activation in most instances was n o r e robust
in extrastriate areas of the ventral stream. In
contrast, the striate cortices were not acti-
vated, and most analyses suggested that
rCBF in the primary visual cortex may be
significantly attenuated illlring REbI sleep.
!. TAXIJARY 199s 93
 These results were paralleled in t h e
analysis of correlations between rCBF and
REM density ( 2 3 ) , suggesting a physiolog-
ical mechanism that may mediate t h e ef-
fects of REM sleep o n t h e visual cortex. In
h ~ r n a n s ,REMs are associated with cere-
l ~ r a potentials
l that bear strong sinlilarities
t o Pontogeniculooccipital ( P G O ) waves,
which mediate the spontaneous central
excitation of t h e visual cortices during
REh'l sleep in animals ( 2 4 ) . P G O waves
may have a disproportionate effect o n the
excitability of extrastriate cortices. thus
accounting for t h e relat~velys e l e c t ~ v eac-
t ~ v a t i o nof these reglons that was observed
in this study ( 2 5 ) .
REMs also serve as indices of dreamine: 0
,4ltl1ough subjects were not awakened and
debriefed in t h ~ sstudy. , , REM densitv has
been shown to correlate positively wlth the
likelihood of dreamine. u ~ l t hthe intensitv
and bizarreness of dream imagery, and with
the nresence of Inore vlvld and visuallv
active dreams (26). O u r results thus suggest
that the snontaneous eeneration of visual
0
itnages that occurs during REM sleep may
be associated with isolated activation of the
extrastriate cortices ( 2 7 ) .
Lforeover, both principal component
and correlation analyses, which may serve
as a means of describing f~rnctionalconnec-
u
tions between brain regions, demonstrate a
reciprocal relationship between activ~tyin
extrastrlate and strlate cortlces during REbI
sleep, a n observatlon that has not previous-
ly been described in PET studies of visual
f~rnctionduring either \\laking or sleep (28-
30).
Vi'hat mechanism might account for
this? It has been l~ypotl~esized that, because
dreams typically consist of integrated and
coherent v ~ s u a limages, there should be re-
entrant activation of the strlate cortex by
extrastrlate back-projections during REM
sleep, as has been suggested to occur during
waking imagery tasks (31 ). O ~ r reslrltsr sup-
port a n alternative hypothesis. Because 111-
creaseL{ synaptic actlvity in extrastriate ar-
eas is coupled to decreased activity in the
primary visual cortex, firing of the extrastri-
ate back-projections may in fact be attenu-
ated-or acti\,ely ~nhibited-during REM
sleep. Rather than leading to reentrant ac-
tivation of V1 and V2, extrastriate activa-
tion during REbI sleep may be linked to
suppression of activity in pritnary regions
connected to the external environment
(32).
Indeed, REh'l sleep appears to be char-
acterizeL1 by a double dissociation. Activity
in extrastriate regions appears to be recip-
rocally related not only rvith activlty 111 V l
and V2 but also with that in higher order
frontal associat~onareas t o which extrastri-
ate cortices project-dorsolateral prefrontal
94 SCIEKCE
and lateral orbital cortices-l~eteromodal
regions in which visual infor~nation is
bound with that processed in other areas of
the brain (33). A t the same time, however,
we ol~servedconcurrent activation of extra-
striate regions and limbic-related projection
areas-parahippocampa1 cortices and con-
tiguous portions of the hippocampus-and
activity in these nlesial temporal regions
was positi\,ely correlated with R E N density.
Of note, sinlilar results were observeil in the
anterior cingirlate cortex (Fig. 1 ) , to which
the parahippocampal cortices project.
This pattern suggests that pathways that
lnediate the transfer of inforlnation be-
tween visual cortices and the lilnbic system
may be active during REM sleep, but path-
ways that lnediate transfer of \,lsual infor-
Ination to prefrontal association cortices are
not.
Thus, iluring REh'l sleep, the extrastriate
cortices and paraliinbic areas to which they
project may be operating as a closed system,
functionally disconnected from frontal re-
gions in w l ~ l c hthe highest order integration
of visual information takes place. Such a
dissociat~oncould explain many of the ex-
periential features of dreams, including
heightened emotionality, ~mcriticalaccep-
tance of bizarre dreatn content, a dearth of
parallel thoughts or images, temporal d~sorl-
entation, and the absence of reflective
anrareness (34).
REM sleep may represent a state in
which the hrain engineers selective activa-
tlon of a n interoceptive network, which is
dissociated from primary sensory and het-
eroinodal association areas at either end oT
the visual hierarchy that lned~ateinterac-
tions with the external world.
REFERENCES AND NOTES
1. E. Aser~nskvand N. Kle~tnan Science 118 273
(1953).
2 W Denent and N. Kle~tman,J. Exp. Psycho1 53.
339 (1957).
3 Some st-rdes suggest that actlvat on of only the
h~gherlevels of the v~sualprocessing h~erarchyIS
suffcient to evoke and sustan \,suaI n a g e r y [P E.
Roland and B. Guyas, Cereo. Co,qex 5, 79 ( I 995)l.
4 Others [S M. Kosslyn, W. L Thompson, I. J K I T .
N. M. Apert. Nature 378, 496 (1995)l n d c a t e that
contrbuton of the pr n a v v sua processng areas-
p o s s b y actvated through reentrant pathways [D J
F e e n a n and D C. Van Essen, Cereb. Co,qex 1 , I
(l991)]- s also requred.
5. J. A. Hobson. The Dreamlng Bran (Bas~cBooks,
New York. 1988).
6,Informed consent was obta~nedfrom 10 healthy
Inale voI.rnteers (23 6 ? 1.7 years of age. n e a n
z SD; range, 22 to 27 years). E~ghtsubjects were
r~ght-handed,and two were left-handed.
7 There were 8 hours of sleep w t h seectve REM de-
pr~batonon day 1, 2 hours of sleep w ~ t hREM depr-
vaton on day 2; and 15 5 to 23 5 hours of contnu-
ous wakefulness (22.2 = 2.4 hours. mean z SD) on
day 3 Subjects were nontored p o l y s o n ~ o g r a p h -
c a y throughout t h s perod to veVfythe absence of
sleep and f a c t a t e seectve REM depr~baton.
8 PET scans were perforred on a Scand~tron~x
PC2048-156 tolnograph (Uppsaa Sweden) w h c h
\'CTI-. 279 2 JANUARY1998 ~~~ww.scienceinag.org
acqulres I 5 contguous planes. A transnss~onscan
was performed for attenuaton correcton
9 A. Recitsciaffen and A E Kales, A ~Manua!of Stan-
dard~zedTerm/no!ogy. Techn~ouesand Scorfng Sys-
tem for Sleep Stages o f Human Subjects (Govern-
ment Prntlng Offce Wasington DC. 1968). Durlng
a scans, subjects eyes were closed and occluded
by eye patcies, and a n b e n t nose was kept to an
a b s o l ~ ~ nt el n n u m W i e n a target stage was d e n t -
fled. 30 n C of HLiSO were Injected and scans were
nlt~atedautonat~caly For n c u s ~ o nn these anay-
ses, a sleep stage i a d to be un~formlyn a n t a ned
(free of arousals and stage s i f t s ) f r o n 40 s before to
40 s after scan nltlat~on.It was necessary to use
partla sleep-depr~vatonmetiods to i e p ensure t i a t
subjects would sleep In the scanner. Such proce-
dures however, lray affect sleep arch~tecture(that
IS,the reatve anounts and t m n g of the varous
stages withn the sleep pe-~od)To ebauate these
potenta effects, we compared our subjects' sleep
archtecture w t h pubshed normatve data [R L. WII-
lams, I Karacan. C J. Hursch, Electroencephalog-
raphy (EEG) of Human Sleep, Clinical Apphcatlons
( W e y , New York. 197A)l. Except for Increased SWS
t m e durng the f~rstt h r d of the sleep per~od(an
expected consequence of the deprlvat~on proce-
dure), we found no quatatbe or quanttatve abnor-
n i a t ~ e sIn the sleep arch~tectureof our subjectsFur-
thermore the scans selected for analys~swere In
each Instance, assocated w ~ t hsleep samples that
conformed e x p c ~ t and y unalnbguousy to the stan-
dard scorlng cr~ter~a of Rechtschaffen and Kales
throughout the perod of data a c q u s t o n . Therefore
although sleep depr~vatonmay have affected dura-
t o n and t m n g of sleep stages, the procedure pro-
duced no o b s e r ~ a b eefects on the phys~olog~cal
characterstcs of the sleep stages thensebes (that
is. at the e b e of anayss n e s t c r ~ t c ato t h s study).
10 PET scans were reg~steredand n o r n a z e d to a
c o n m o n stereotax~c space [J. Tala~rach and P.
Tournoux. Co-planar Stereotact~cAtlas of the Hu-
man Bran (Theme. New York. 1988)l. and stage-
dependent d~fferencesIn normazed rCBF were an-
alyzed w~:h SP'vl software [K. J. Frston, C. D Fr~th,
P. F. L ~ d d l eR S J Frackow~ak,J. Cereb. Blood
Fioi.:. Metab. 11, 690 (199l)l w ~ t hanayss of covar-
ance (ANCOVA) correcton for global blood flow In
a d d t o n tests of sgnfcance based on the sze of
the activated reglon [K J Fr~ston.K. J. Worsley,
R S J. Frackow~ak.J C Mazz~otta.A. C. Evans,
Hum. Bra111Mapp 1, 21 0 (199A)l were performed a
Z-threshold of 2 33 and a baue of 1W = 2.73 derved
f r o n a n u contrast carred out In the s a n e set of
ndviduals were used. Absolute rCBF rates were
also calculated by usng a measured arter~alInput
funct~onand PC3, balues. Of the 10 subjects In
w h o n REM sleep scans were acqured, 7 had SWS
scans. and 7 had wakng postseep studes Stage
contrasts were performed in a repeated neasures
d e s g n thus, N = 7 for both REM-wakng and REM-
SWS contrasts To evaluate hemspheral effects, d f -
ferences In n o r n a z e d rCBF for each contrast were
conpared between homologous p x e s in right and
left hen~spheresby using a voxe-wse error bar-
ance, dfferences exceedng a Z-threshold of 3 09
were consdered sgnf~cant
II The freq.rency of rapd conjugate eye mobements.
d e n t f e d as p o s t v e or negatve deflect ons on the
electrooculogram (EOG) dur ng the I 8 0 s after the
start of the scan, were computed. Normalzed rCBF
Images were correlated with each subject's REM
counts on a p x e - b y - p x e bass ( N = I 0 for these
analyses) ( F I ~I C ) . Correlation coefflc~ents were
thresholded 2t a level of 0 6 (P < 0 05, ii = 10) for
tabulat~on(Table I )
12. Coordnatesfor 20 b s u a cort c a areas in each h e n -
sphere (7 striate, 6 lateral o c c p ~ t a7, fusform!nfero-
tenpora) were selected by usng the Taarach atlas,
adjusted w ~ t hthe SPM PET tenpate, and used to
extract rCBF v a l ~ ~ ei rso n the stereotaxcay n o r n a -
zed PET n a g e s These balues were n o r n a z e d by
u s ~ c gnean CBF rates for the v~sualcortex as a
whole, rCBF values In homologous regons of r~ght
and left helrspheres were aberaged and used to
generate difference n a t r c e s (REM-SWS and REM-
 wake. N = 7), a p r n c p a conponent anayss was
then a p p e d . S~gn~ficance of the extracted compo-
nent bector was assessed by uslng pernutailon
tests [B. F. J. Manly. Randomization anci Monte
Car10 Methocis in Biology (Chapman & Hall. New
York. 1990)l by randony pernutng the s g n of data
entries 19 t n e s ; Pvaues were derved by determn-
n g tile proportion of pernuted t r a s ~ ~ 1 1hgher
t h frst
egenvaues than the or~g!naldata. S t a b t y of regon-
a oadngs was assessed by the bootstrap proce-
dure [B. Efron. SIAM Monogr 38-1 (1982)l. F Ivalues ~
of the orig~nalload~ngsacross 10.000 bootstrap
samples, expressed as Z-scores, were calculated:
absolute values > 2 4 were consdered stable
13 Cluster 1. eft hemsphere, 8688 m n " P ' (n,., > k).
corrected = 7.90 x 10-4: centered x = 4 9 , y =
6 7 . 9 . z = -5.3; cluster 2. r~ghthemsphere. 21 056
mm3: P ' (n,, > k). corrected = 2.14 X lo-'; cen-
tered x = 21.9 y = 5 1 8. z = 6 8
14. S. M Kosslyn and K. N. Ochsner. Trends Neurosci.
1 7 290 (1994)
15, A Rechtschaffen, P Verdone, J Wheaton, Can.
Psychiatr Assoc. J 8. 409 (1963).
16 17427 n n 3 ; Pi (n, > k ) , corrected = 1 90 X
10--6 centered x = 2.8, y = -83 0. z = 0 4.
;7. Absolute PCO, corrected flow rates were consstent-
y owerdurng REM steepthan durng SWS through-
outthe strate cortex [ I r C B F = - 6 40 ? 5.9 ml!l00
g!mn (nean z SD). x = 4. y = 8 2 , z = 01, and
absolute rCBF rates were consstenty elevated In
extrastriate regons (in the f.rsforn cortex. 1 r C B F =
+8 85 z 2.79 n i l 0 0 g n n . x = -34. y = 5 6 , z =
-8), altiiougii variances In absolute balues were
large.
18 58320 n m 3 ; P ' (n, > k), corrected = 3.55E 1 5 :
centered x = 3 . 9 , y = 0.3 z = 1 2.
19. D. A. Chavs and D. N. Pandya, Trans. Am. Neurol.
Assoc. 99, 192 (1974): W. A. Suzulc and D. G. A n a -
ral. J Cornp Neurol 350. 497 (1994).
20. The frontal eye f~elds(superior dorsolatera prefrontal
and prenotor cortices) were ncompetely sampled In
nany subjects and are not Included in this anayss.
21. The bootstrappng procedures 'ndcated that n o s t
regona oad'ngs were stable, In tile REM-SWS con-
trast, a the fus~form-inrerotenporal reg~ons-but
none of the lateral o c c p ~ t aregons-were assocat-
ed witii Z-scores exceedng tiiresiiod (local n a x -
n u m = 5 4 4 , x = ~ 4 8 y. = -62. z = 0). In the
REM-wake contrast, SIX of the seven fusforn-nfero-
temporal reglons (naximum 4 18, x = ?48, y =
6 2 , z = A), and four of the SIX lateral occip~tal
regons ( ~ r a x m u ~ 3 n58, x = ~ 3 2y, = 8 6 , z = 0)
were stable: n the latter case, a were located In
ventral ratiier tiian dorsal port~onsof the lateral oc-
cipital cortex ( z 5 8 n n ) , In the REM-SWS con-
trast, f~veof the seven striate o a d n g s were stable
(local n l n n u m = -3.79. x = t 4 , y = -82, z = 0).
and In the REM-wake contrast, SIX of seven oadngs
were stable ( r n m u n = -9.89, x = ~ 4y = . -82.z
= 0).
22. S. Zek etal., J Neurosc~.11.641 (1991); D. Watson
etal. Cereb. Co-iex 3, 79 (1993): J V Haxby et a/. ,
J. Neurosci. 14. 6336 (; 994); I. Sereno et a1 . Sci-
ence 268. 889 (1995)
23. These results could reflect the central effects of sac-
c a d c eye novenents per se. Voluntary saccades
durng wakefulness pre\'ousy habe been found to
be assoc~atedw ~ t hreduct~onsn v~sual CBF [T.
Paus. S. Marrett. K. J. Worsey. A. C. Evans. J. Neu-
rophysiol. 74. 21 79 (1995)l. Howeber saccades n
that study were contnuous and r a p d (40 to 140 In
60 s) but were ntermttent and lower In frequency In
our own study [O to 11 In 60 s (5 5 i- 4 0 mean z
SD]. Furthermore, reduct~onsn rCBF were n o r e
w d e y dstrbuted throughout the v s u a cortex, and
no p o s t v e correatons between extrastrate actvty
and saccadc eye novements were obseqed, whcii
suggests that the strate-extrastrate dissocaton ev-
dent In t h s study may be a unique characteristc of
REM sleep.
2A, R W. McCarey. J W Wnkeman. F H Durn. Brain
R e s 274. 359 (1983). C. W, Calla~".!ay,RRLyd~c,H HA.
Baghdoyan, J A Hobson Cell Mol. Neurobiol 7 .
105 (1987)
25. The effects of the PGO wave generator on the oc-
c p ~ t acortex
l do not appear to be n e d a t e d through
the lateral gencuare nucleus [J. A Hobson J. Aex-
ander. C J. Frederlckson Brain Res 14. 607 (1969)]
and therefore are not expected to be exclus~vely
assocated w ~ t hstriate projectons Early studes also
denonstrated tiiat PGO wave amptudes In tile a t -
era assocaton areas of the cat v s u a cortex exceed
tiiose n p r n a r y \,suaI cortces [D. C. Brooks. Exp
Neurol. 22, 603 (196811, and s n g e u n t studes have
shown that the facl~tatoryeffects of PGO waves on
neuronal f~rinarates annear to be nan~festIn extra-
strate but no'f n stratkcortces [T. Kasamatsu and
W R. Adey, Bran Res 55, 323 (1973)l.
26. D. R. Goodenough, H. B. Lews.A. Shapro. LSeser,
J. Ment. D s . 140. 365 (1965), P Verdone,
Percept Mot Skills 20, 1253 (1965): T. Piwk and D.
Foukes, Science 153. 1282 (1966): W Denent and
E. A. Wopert. J. Exp. Psychol. 55. 543 (1958). See
also (2, 74)
27 T i i s appears to be consstent ~ ~ 1 1the t h f n d n g of
Roland and Guyas (3). However, our results may not
be directly conparable w ~ t hstudies of conscious
v s u a Imagery; task-ecited act~\,a?on of the primary
bsual cortex seen by Kosslyn et a/. (4) ~ r g h be t a
s p e c a feature of conscous b s u a mageryor reire\'-
a processes that occur durng the wakng state
28 P. Maquet et a1 [Nature 383, 163 (1996)] d d not s t
rCBF changes in visual cortices; however, they c o n -
pared REM sleep w ~ t i ian amalgan of wake and
non-REM stages, whereas our contrasts were stage
speciflo (for example. REM-SWS) and may be n o r e
sens~t~ve.
29. P. L. Madsen et a1 [J.Cereb. Blood Flow Metab. 11,
502 (1991)l reported REM-assoc~atedIncreases In
rCBF n v~sualassoc~ationcortices: funct~onarela-
tonships between actbty In striate and extrastriate
cortces were not evaluated
30. C. C. Hong et a1 [Sleep 18, 570 (1995)l reported a
Discrete Start Sites for DNA Synthesis
in the Yeast ARSl Origin
Anja-Katrin Bielinsky and Susan A. Gerbi*
Sites of DNA synthesis initiation have been detected at the nucleotide level in a yeast
origin of bidirectional replication with the use of replication initiation point mapping. The
ARSl origin of Saccharomyces cerevisiae showed a transition from discontinuous to
continuous DNA synthesis in an 18-base pair region (nucleotides 828 to 845) from within
element B1 toward B2, adjacent to the binding site for the origin recognition complex,
the putative initiator protein.
An origin of bidirectional D N A replica-
tion is characterized by the transition be-
tween continuous D N A svnthesis (nroceed-
ing in one direction) and discontinuous
synthesis (proceedtng in the opposite direc-
tion). W e have developed replication initi-
ation point (RIP) mapping to deterlnine
this transition in the autonomously repli-
cating sequence (ARS) 1 of the yeast Snc-
charomyces cerevisiae.
ARSl functions as an origin of D N A
replication (ORI) both on a plasmid and in
its normal context on chroinoso~neIV (1).
Department of Molecular B o o g y . Cell B o o g y and B o -
chemstry Dv~sionof B o o g y and Medcine, Brown U n -
verslty Prou~dence.R 0291 2, USA.
-To vi~iioncorrespondence should be addressed E n a ~ l
susan-gerbabrown edu
p o s t v e correlation between REM densty and glu-
cose n e t a b o c rates n lateral occ~p~tal cortex, but
results for p r n a r y bsual cortex were not reported.
31. S. Zek~,A Vision o i the Brain ( B a c k w e , Oxford
1993). See also (3, 4).
32 A pattern of extrasirlate activity n the absence of
striate functon also character~zesc n i c a states as-
soc~atedw ~ t hunusual and often b~zarreperturba-
tions of b~sualawareness, sucii as b n d s ~ g h t[L.
We~skrantz,J L. Barbur. A Sahra~e.Proc. Natl.
Acad. SCI. U.S.A. 92. 61 22 (199511 and confabua-
tory d e n a of bndness [G Goldenberg, W. M u -
bacher, A. Nowak, Neuropsychologia 33, 1373
(1995)l. The features of both syndrones suggest
that synciironous actvaton of prmary and extrastri-
ate cort~cesmay be essent~afor "normal" v~sual
awareness. That such coherence appears to be
breached dur~ngREM sleep IS not nconsstent w ~ t h
the altered awareness that ciiaracterzes this sleep
stage.
33. Decreases In actlvlty In the prefrontal cortex durng
REM sleep have been noted previously by Madsen et
a/. (29) and by Maquet et a1 (28). On the other hand.
Hong et a1 (30) reported pos~t~be correlations be-
tween REM densty and absolute glucose metaboc
rates In the (r~ght)dorsolatera prefrontal cortex:
however, t i i s measure, w h c h reflects phys~olog~cal
responses ober a longer t n e period, may not be
drectly comparable w ~ t hnormalzed rCBF balues
evaluated In t h s study
34. A. Rechtschaffen. Sleep 1 . 97 (1978); J. A. Hobson,
Endeavour 20. 86 (1996).
35. The autiiors w ~ s ito i thank Dr. Alex Martn for his ex-
perise, ns~ght,and c r ~ t c asuggestons, which were
essenta In the preparation of t h s nanuscrpt, and Dr
Scott Sebie for h s vaiuabie i i e p in preparing F I ~1.
1 1 August 1997; accepted 18 Novenber 1997
,4RS1 -containing plasmids respond normal-
ly to the cell cycle, duplicating once per
cycle ( 2 ) , and replication is initiated by the
same cellular protein machinery acting on
chromosomes.
4 R S 1 is composed of subdolnains A, B 1,
B2, and B3 (3). Subdomains A and B l are
recognized by the origin recognition com-
plex ( O R C ) (4), the putative initiator pro-
tein (5) indispensable for ortgin function
(6, 7). Element B2 is easily unlvound D N A
(8) and element B3 is a binding site for the
ARS binding factor I (ABFI) (9).
RIP mapping, described here, has suffi-
cient sensitivity for study of eukaryotic
origins, unlike an earlier method (10). It
allows precise mapping of initiation sites
for D N A synthesis and was applied to a