Columnar organization in the midbrain periaqueductal gray: modules for emotional expression?

13 Wiley, J. W., Gross, R. A. and Macdonald, R. L. (1993) J. Neurophysiol. 70, 324-330 14 Hua, X-Y. et al. (1991) J. Pharmacol. Exp. Ther. 258, 243 -248 15 Plummer, M. R., Rittenhouse, A., Kanevsky, M. and Hess, P. (1991) J. Neurosci. 1 I, 2239-2248 16 Foucart, S., Bleakman, D., Bindokas, V. P. and Miller, R. J. (1993) J. Pharmacol. Exp. Ther. 265, 903-909 17 Balasubramaniam, A., Sheriff, S., Rigel, D. F. and Fischer, J. E. (1990) Peptides 1 I, 545-550 18 Grundemar, L., Wahlestedt, C. and Reis, D. J. (1991) J. Pharmacol. Exp. Ther. 258, 633-638 19 Schofield, G. G. and Ikeda, S. R. (1988) Eur. J. Pharmacol. 151,131-134 20 Zidichouski, J. A., Chen, H. and Smith, P. A. (1990) Neurosci. Left. 117, 123-128 21 liles, P. and Regenold, J. T. (1990) Nature 344, 62-63 22 Toth, P. T., Bindokas, V., Bleakman, D., Colmers, W. F. and Miller R. J. (1993) Nature 364, 635-639 23 Betz, W. and Bewick, G. S. (1992) Science 255, 200-203 24 McEnerny, M. W., Snowman, A. M. and Snyder, S. H. (1994) J. Biol. Chem. 269, 5 - 8 25 Xiong, Z., Bolzon, B. and Cheung, D. W. (1993) Pfl6gers Arch. 423,504-510 26 Horn, J. P. (1992) Can. J. Physiol. Pharmacol. (Suppl.) 70, 19-26 27 Hirning, L. D., Fox, A. P. and Miller, R. J. (1990) Brain Res. 532, 120-130 28 Kohler, C., Eriksson, L., Davies, S. and Chan-Palay, V. (1987) Neurosci. LetL 78, 1 - 6 29 Milner, T. A. and Veznedaroglu, E. (1992) Hippocampus 2, 107-126 30 Colmers, W. F., Lukowiak, K. D. and Pittman, Q. J. (1987) J. PhysioL 383,285-299 31 Haas, H. L., Hermann, A., Greene, R. W. and Chan-Palay, V. (1987) J. Comp. Neurol. 257, 208-215 32 Klapstein, G. J. and Colmers, W. F. (1993) Hippocampus 3, 103-112 33 Colmers, W. F., Lukowiak, K. D. and Pittman, Q. J. (1988) J. Neurosci. 8, 3827-3837 34 Bleakman, D., Harrison, N. L., Colmers, W. F. and Miller, R. J.

(1992) Br. J. PharmacoL 107, 334-340 35 McQuiston, A. R. and Colmers, W. F. (1992) Neurosci. Lett. 138, 261-264 36 Monnet, F. P., Debonnel, G. and de Montigny, C. (1990) Eur. J. Pharmacol. 182, 207-208 37 Monnet, F. P., Fournier, A., Debonnel, G. and De Montigny, C. (1992)J. PharmacoL Exp. Ther. 263, 1212-1218 38 Colmers, W. F., Klapstein, G. J., Fournier, A., St-Pierr e, S. and Treherne, K. A. (1991) Br, J. Pharmacol. 102, 41-44 39 Coan, E. J. and Collingridge, G. L. (1985) Neurosci. Lett. 53, 21-26 40 Klapstein, G. J. and Colmers, W. F. (1993) Soc. Neurosci. Abstr. 19, 1869 41 Stasheff, S. A., Bragdon, A. and Wilson, W. A. (1985) Brain Res. 344, 296-302 42 Rizzi, M., Monno, A., Samanin, R., Sperk, G. and Vezzani, A. (1993) Eur. J. Neurosci. 5, 1534-1538 43 Wahlestedt, C. et aL (1990) Brain Res. 507, 65-68 44 Goodman, J. H. and Sloviter, R. S. (1993) Brain Res. 616', 263-272 45 Roman, F. J. etaL (1989) Eur. J. PharmacoL 174, 301-302 46 Tam, S, M. and Mitchell, K. N. (1991) Eur. J. Pharmacol. 193, 121-122 47 Quirion, R., Mount, H., Chaudieu, I., Dumont, Y. and Boksa, P. (1991)in NMDA Receptor Related Agents: Biochemistry, Pharmacology and Behavior (Kamayama, T., Nabeshima, T. and Domino, E. F., eds), pp. 203-210, NPP 48 Bouchard, P., Dumont, Y., Fournier, A., St-Pierre, S. and Quirion, R. (1993) J. Neurosci. 13, 3926-3931 49 Williams, J. T., Colmers, W. F. and Pan, Z. Z. (1988) J. Neurosci. 8, 3499-3506 50 Kombian, S. B. and Colmers, W. F. (1992) J. Neurosci. 12, 1086-1093 51 Livingstone, M. S. and Hubel, D. H. (1981) Nature 291, 554-561 52 Madison, D. V. and Nicoll, R. A. (1982) Nature 299, 636-638 53 Illes, P., Finta, E. and Nieber, K. (1993) Naunyn-Schmied. Arch. PharmacoL 348, 546-548 54 Williams, J. T., Henderson, G. and North, R. A. (1985) Neuroscience 14, 95-101

Acknowledgements We would like to thank our co-authors on our individualand collaborativestudies on neuropeptide Y, mostnotably R.J.Mi//er, V. P. Bindokas, P. Toth, N. Harrison, £ B. Kombian, A. R.McQuiston and G.3. Klapstein,many of whomgenerously supp/iedunpublished data orfigures. Researchin W F.C's laboratory was supportedby the MedicalResearch Councilof Canada. W.E C isa Seholarof the AIberta Heritage Foundation for MedicalResearch.

Columnarorganizationin the midbrainperiaqueductalgray: modulesfor emotionalexpression? R i c h a r d B a n d l e r a n d M i c h a e l T. S h i p l e y

Independent discoveries in several laboratories suggest that the midbrain periaqueductal gray (PAG), the celldense region surrounding the midbrain aqueduct, contains a previously unsuspected degree of anatomical and functional organization. This organization takes the form of longitudinal columns of afferent inputs, output neurons and intrinsic interneurons. Recent evidence suggests: that the important functions that are classically associated with the PAG - defensive reactions, analgesia and autonomic regulation - are integrated by overlapping longitudinal columns of neurons; and that different classes of threatening or nociceptive stimuli trigger distinct co-ordinated patterns of skeletal, autonomic and antinociceptive adjustments by selectively targeting specific PAG columnar circuits. These findings call for a fundamental revision in our concept of the organization of the PAG, and a recognition of the special rolesplayed by different longitudinal PAG columns in co-ordinating distinct strategies for coping with different types of stress, threat and pain.

havioral responses to threatening or stressful stimuli. Traditionally, there was little interchange between these research areas. However, a meeting held during 1990 brought together scientists working on the PAG from a variety of perspectives. A number of important points of consensus emerged from this meeting and a subsequent symposium at the Annual Meeting of the Society for Neuroscience in 1992. First, that the nocicepfive inhibition and behavioral responses elicited from the PAG were best conceptualized as components of co-ordinated reactions important for survival. Second, that functional specificity within the PAG was represented in the form of distinct, longitudinal neuronal columns extending, for varying distances, along the rostrocaudal axis of the PAG 1-3. In this article, the evidence for longitudinal columnar organization within the PAG is reviewed, and the mechanisms by which PAG columns coordinate distinct patterns of behavioral and physiological reactions critical for survival are considered.

RichardBandleris at the Dept of Anatomy and Histology, Universityof Sydney, Sydney, New South Wales2006, Australia, and Michael T. Shipleyis at the Oept of Anatomy, University of Maryland, 655 WestBaltimore Street, Baltimore,MD 21201-1559, USA.

During the past two decades, two largely independent Historical perspective themes have dominated research on the PAG, namely Early studies reported that defensive or 'aversive' inhibition of nociception, and the integration of be- reactions, hypertension and, later, antinociception, TINS, Vol. 17, No. 9, 1994

© 1994, Elsevier ScienceLtd

379

Lateral PAG: defensive behavior hypertension tachycardia non-opioid analgesia

a highly defensive animal. Remarkably similar patterns of heightened defensive reactMty were produced by injections of low doses of KA in homologous regions of the PAG of Confrontational defense the rat Ls'~. Moreover, the kind of regional blood flow: defensive strategy produced was limbs found to be dependent on the viscera region of the PAG excited (Fig. 1 ). face f Excitation of neurons within the Rostral Flight intermediate third of the lateral PAG 1if regional blood flow: PAG produced a confrontational limbs defensive reaction, characterized viscera t, by facing and backing away from face ; Intermediate a "threatening' stimulus, whereas PAG excitation of neurons in the caudal third of the bteral PAG produced Ventrolateral PAG: an avoidance or flight reaction. quiescence Overall, the defensive reactions elicited from the lateral PAG were hyporeactivity strikingly similar to the natural hypotension Caudal reactions of a rat or cat when bradycardia PAG threatened or attacked. In contrast, opioid analgesia microinjection of EAAs nmde in the caudal, ventrolateral PAG of Fig. 1. Iflustratlon of the lateral and ventrolateral neuronal columns within the rostral, the inter- the rat or cat, elicited a different mediate (two sections) and the caudal periaqueductal gray (PAG). The colored outlines indicate the reaction, namely a cessation of position of the dorsolateral (pink) and dorsomedial (blue) neuronal columns. Injections of excitatory ongoing spontaneous activity (that amino acids (EAAs) within the lateral and ventrolateral PAG column emit fundamentally opposite alterations in sensory responsiveness, and somatic and autonomic adjustments. Injections of EAA is. quiescence), and a profound made within the intermediate, lateral PAG (dark blue) elicit a confrontational defensive reaction, hyporeactivity, the animal neither tachycardia, and hypertension associated with decreased blood flow to limbs and viscera, and orienting nor responding to its increased blood flow to extracranial vascular beds. Injections of EAA made within the caudal, lateral environment ~sm. The quiescent, PAG (light blue) produce flight, tachycardia, and hypertension associated w~th decreased blood hyporeactive pattern of behavior flow to viscera/and extracranial vascular beds and increased blood flow to hmbs. In contrast, elicited r-ore the ventrolateml PA(I injections of EAA made within the ventrolateral PAG @reen) cause cessation of all spontaneous is similar to the reactions of an activity (that is, quiescence), a decreased responsiveness to the enwronment, hypotension and animal subsequent to injury, or bradycardia. The lateral and ventrolateral PAG also mediate different types of analgesia. after defeat in a social encounter. In addition to these behavioral effects, opposite could be elicited by electrical stimulation at sites in the PAG and adjacent midbrain tegmentum (for review, patterns of cardiovascular changes were elicited from see Ref. 4). Although studies of defensive and lateral and ventrolateral PA(; regions. Activation of cardiovascular functions usually focussed on the sites lateral to the aqueduct produced increased dorsal and lateral regions of the PAC-, whereas arterial pressure and tachycardia; activation of sites analgesic studies usually stimulated sites in the ventrolateral to the aqueduct produced decreased ventral PAG, this was not the result of any systematic arterial pressure and bradvcardia ~ I:;. Additional search for regional differences. Indeed, since at the study in the cat ill'' and rat > revealed that PAGtime the PAG was generally considered to be a elicited pressor and depressor reactions were not due primitive reticular structure, with little anatomical to a generalized change in peripheral vascular resistorganization, there was no incentive to search for ance. Rather. there was evidence of striking regional regional differences. In addition, any regional differ- alterations in peripheral vasoconstrictor tone such ences would probably have been obscured by the use that blood flow was redirected to vascular beds with of electrical stimulation which affects both cell bodies potentially the highest metabolic demand (Fig. 1). For and fibers of passage. Not surprisingly, there was example, from caudal, lateral PAG sites at which flight little appreciation of functional micro-organization was elicited, the pressor response was associated with a pattern of increased blood flow to the limbs and within the PAG. decreased flow to viscera and the face. In contrast, at Discovery of functional columns intermediate, lateral PAG sites at which a confronSomatic and autonomic integration. Intracerebral tational defense reaction was elicited, the pressor microinjection of excitatory amino acids (EAAs) is response was associated with increased blood flow to now a widely used technique for exciting neurons but the face, but decreased blood flow to the limbs and not fibers of passage. Although the technique is not viscera. Neither of these cardiovascular patterns was without its limitations ~''~, its application to the study of secondary to the elicited somatic changes, as the the PAG revealed striking functional differences in same patterns were observed in the paralyzed adjacent PAG regions. Injections of glutamate, aspar- animall~ 1~. Similar behaviorally coupled patterns of tale, ,,l,-homocysteate or weak doses of kainic acid blood flow, associated with different defensive be(KA), made in the region of the PAG lateral to the haviors that were produced naturally (in the freely aqueduct Lr.s changed a normally calm, placid cat into moving cat), were described many years earlier in a 380

TINS, VoL 17, No. 9, 1994

series of elegant studies by Mancia and colleagues lr' 18. Taken together, these results point out the need to correct the still often cited, but clearly erroneous view, that a single cardiovascular pattern characterizes all defensive behavior from the early-alerting stage onwards. Antinociception. Since the early reports that either electrical stimulation or microinjections of EAAs into the PAG elicited antinociception, a growing body of evidence indicates that distinct types of antinociception are mediated by the lateral and ventrolateral PAG columns. Besson and colleagues 1~)evaluated both the antinociception and behavioral reactions produced by threshold electrical stimulation within the PAG, and found that although antinociception elicited from the lateral PAG was invariably associated with strong 'emotional reactions' (vocalization, running and jumping), such emotional reactions never accompanied the antinociception elicited from the ventrolateral PAG. More recently, it was found that antinociception elicited by microinjections of EAA into lateral PAG was associated always with hypertension and tachycardia, whereas antinociception produced by activation of the ventrolateral PAG was associated with hypotension and bradycardia 1''. In addition, the ventrolateral PAG is the only PAG region at which injections of low doses of morphine produces antinociception e° ee. Thus, a non-opioid-mediated antinociception, associated with active defensive behaviors, hypertension and tachycardia, is elicited from the lateral PAG; whereas, an opioid-mediated antinociception, associated with quiescence, hyporeactivity, hypotension and bradycardia, is elicited from the ventrolateral PAG (Fig. 1). This latter pattern, opioid-mediated analgesia coupled to a behavioral reaction of quiescence and hyporeactivity, has also been reported following defeat in a social encounter 2:~.

Fig. 2. Rostra/ (A) to caudal (F) series of coronal sections through the periaqueducta/ gray (PAG) of the rat showing output neurons labeled retrograde/)/ after a large injection of wheat-germ agglutinin (WGA)-HRP into the ventral medulla. Note that PAG output neurons cluster into groups. At intermediate levels (C and D) there are two distinct clusters in the dorsomedial and lateral PAG. These comprise the dorsomedial and lateral neuronal columns. At the transition between intermediate and caudal thirds of the PAG (E) a third cluster, just ventrolateral to the lateral column, appears. This indicates the start of the ventrolateral PAG neuronal column. See also Fig. 48 in ReL 25.

Discovery of anatomical columns

Taken together, the studies reviewed above suggest that distinct integrated patterns of behavioral, autonomic and antinociceptive adjustments are co-ordinated by longitudinal neuronal columns situated lateral and ventrolateral to the aqueduct. Recent anatomical experiments provide strong evidence for the existence of these and other longitudinal columns within the PAG. I'AG outputs to brain stem and spinal cord. Although the PAG has few direct descending projections to the spinal cord, retrograde and anterograde tracing studies indicate that PAG regions lateral, ventrolateral and also dorsomedial to the aqueduct, contain many neurons that project to the ventromedial and ventrolateral medullarY-'_,7. However, a wedge-shaped, dorsolateral PAG region lacks medullary projecting output neurons (Figs 2C-F). These findings suggest that what has been traditionally called TINS, Vol. 17, No. 9, 1994

?

~i~ ~~

01 Fig. 3.5raining for the enzyme NADPH diaphorase in the dorsolateral penaqueductal gray (PAG). Note that except for the dorsal raphO nucleus (panel D), NADPH-positive neurons are selectively localized to the dorsolateral PAG column. Compare, for example, B, C and D with Figs 2C, D and E. These data provide strong evidence for the existence of an immunohistochemica/ly distinct dorsolateral PAG column. Abbreviations: Dk, nucleus of Darkschewltsch; DR, dorsal raph6 nucleus," EW, Edinger-Westphal nucleus, III, ocu/omotor nucleus. See also Fig. 1 in ReL 10. 381

nucleus

.=rnent

Lumbar enlargement

Nucleus of the solitary

Cervical enlargement

Fig. 4. Spinal and medullary afferent input to the lateral and ventrolateral periaqueductal gray (PAG) columns. The lateral PAG column (blue) receives a topographically organized spinal input: lumbar enlargement to caudal, lateral PAG; cervical enlargement to caudal intermediate, lateral PAG; and laminar spinal trigeminal nucleus to rostral, intermediate, lateral PAG. Afferents to the lateral PAG arise from both superficial (lamina I) and deep laminae of the dorsal horn/spinal trigeminal nucleus. The ventrolateral PAG column (green) receives spinal input from both cervical and lumbar enlargements (without any obvious somatotopic organization). This spinal input arises from the intermediate spinal gray matter, as well as from superficial and deep laminae of the dorsal horn. Although the ventrolateral PAG neuronal column does not receive an input from the spinal trigeminal nucleus, there is a strong projection from the nucleus of the solitary tract.

the 'dorsal PAG' (a collective term for regions dorsal and lateral to the aqueduct), consists of three anatomically distinct longitudinal columns: dorsomedial and lateral columns, each containing large numbers of medullary projecting output neurons; and a dorsolateral column, containing output neurons that project to the cuneiform nucleus and the periabducens region of the rostral dorsomedial ports, but not to the medulla 2s. The dorsolateral PAG column can also be readily distinguished neurochemically from the adjacent dorsomedial and lateral PAG columns, for example, by an absence of staining for cytochrome oxidase29, but an intense staining for AChE (Ref. 30) or NADPH-diaphorasezl (Fig. 3). These findings leave little doubt that the so-called 'dorsal PAG' consists of three anatomically and neurochemically distinct longitudinal columns of neurons. Although previous anatomical tracing studies reported that the lateral and ventrolateral columns projected to the same regions in the medulla, recent anterograde tracing experiments, using more precise techniques, for example, labeling with Phaseolus vulgaris leucoagglutinin (PHA-L), indicate that the medullary projections of the lateral and ventrolateral 382

PAG might differ more than that suggested initially. For example, it has been reported that the ventrolateral (but not the lateral) PAG column sends a discrete projection to the periambigual region of the lateral medulla32. The periambigual region contains the majority of vagal preganglionic neurons that project to the heart, and the projection from the ventrolateral PAG terminated selectively among periambigual neurons labeled by injection of a retrograde tracer into the pericardium of the heart. The bradycardia elicited by excitation of the ventrolateral PAG might be mediated, then, by a selective projection from the ventrolateral PAG to vagal cardiac motorneurons. Brain stem and spinal inputs to the PAG. Although the PAG has long been recognized as a major somaticand visceral-afferent recipient zone, it is only recently that the extent to which ascending spinal and medullary afferents 'respect' PAG longitudinal columnar boundaries has been appreciated. For example, in monkey, cat and rat, the lateral column of the PAG is targeted precisely by somatotopically organized afterents arising from lumbar and cervical enlargements of the spinal cord and the laminar spinal trigeminal TINS, Vol. 17, No. 9, 1994

A

13

d

Fig. 5. (A) Four approximately equivalent rostrocaudal levels through the periaqueductal gray (PAG) showing the terminal fields of afferents from four specific medial cortical areas. Rostral is at the top and caudal is at the bottom. Note that inputs from each medial cortical field terminate as one to two discrete columns extending along the rostrocaudal axis of the PAG. Illustrated from left to right are PAG inputs derived from precentral medial cortex (PRO; anterior cingulate cortex, pars dorsalis (ACd); prelimbic cortex (PL); and infralimbic cortex (IL). (B) Four approximately equivalent rostrocaudal levels through the PAG showing the terminal fields of afferents from the medial preoptic area (MPO) (left); central nucleus of amygdala (CNA) (right). Inputs from these two subcortical nuclei terminate as distinct and at some levels, spatially complementary longitudinal columns. Adapted from Ref. 3.

nucleus (that is, the lumbar enlargement to the the PAG were published more than ten years ago 37'35, caudal lateral PAG, cervical enlargement and spinal there has been little general appreciation of these trigeminal nucleus, to progressively more rostral observations. In view of the emerging functional and parts of the lateral PAG33-35). Afferents arising from anatomical evidence of columnar organization, it is the upper cervical spinal cord (C1-C3), that terminate important to consider if forebrain connections of the along the rostrocaudal extent of the lateral column, PAG also exhibit columnar organization. appear to be the one exception to this somatotopically Cortical inputs to the PAG. Earlier tract tracing organized pattern 36. The ventrolateral column of the studies revealed modest projections to the PAG from PAG is also targeted by lumbar and cervical afferents, medial prefrontal and insular cortex 39-41. Retrograde but without any apparent somatotopic organization. ,tracing studies, using newer, more sensitive techThe ventrolateral column also receives a projection niques, confirmed these observations but demonfrom the nucleus of the solitary tract 31 (NTS). strated that cortical inputs to the PAG were far more In summary (Fig. 4), somatotopically organized substantial, and arose from a greater number of spinal and spinal trigeminal inputs target the lateral cortical fields than previously reported 3. In the rat, neuronal column, the PAG region that co-ordinates neurons projecting to the PAG are present in a active defensive behaviors, the direction of which continuous band which extends from the frontal to depends, in part, on a detailed picture of the tactile almost the caudal pole of the hemisphere, including a environment. In contrast, the ventrolateral neuronal number of well-defined medial and lateral cortical column, the PAG region that co-ordinates a reaction fields. Medially, substantial numbers of PAG-projecting of quiescence, hypotension and a decreased respon- neurons are present in the infralimbic cortex, presiveness to the environment, receives a convergent limbic cortex, anterior cingulate cortex and the somatic and visceral input. Neither the dorsomedial precentral medial cortex. Laterally, neurons projectnor the dorsolateral PAG columns appear to receive ing to the PAG are located in anterior and posterior any comparable, direct somatic or visceral afferent insular cortices and perirhinal cortex. In all of these input. cortical fields, the PAG-projecting neurons were Forebrain connections of the PAG. Because studies mostly confined to layer V and appeared to be of analgesia, autonomic regulation and defense domi- pyramidal cells. Anterograde tracing experiments nated PAG research during the past two decades, further revealed other important features of cortical anatomical investigations of PAG circuitry focussed inputs to the PAG. Namely, that each cortical field largely on descending projections from the PAG to projects densely and focally to restricted parts of medullary regions that regulate these functions. the PAG, and terminates as one to two longitudinal Although survey studies of forebrain afferent input to columns along the rostrocaudal axis of the PAG TINS, Vol. 17, No. 9, 1994

383

Pressor

Depressor

Saline

Fig. 6. Series of coronal sections through the periaqueductal gray (PAG) illustrating the distnbution of FOSlike immunoreactive neurons (dots) foflowing a I h intravenous infusion (flow rate = 0.8 + 0 . 2 m l h -1) of l-phenylephrine hydrochloride (3-51~g m1-1), sodium nitroprusside (3-5f~gml-~), or physiological saline (control), in the chloralose-anaesthetized rat. Phenylephrine induced a 40-55% increase in mean arterial pressure, and nitroprusside induced a 20-35% decrease in mean arterial pressure. Rostral is at the top and caudal is at the bottom. (M.T.S., unpublished observations.) A similar pattern of selective FOS-labeling within the caudal, ventrolateral PAG has been reported following a 20-35% decrease in mean arterial pressure produced by intravenous infusion of sodium nitroprusside in the conscious rabbit (see Figs 1 and 7 in Ref. 64).

(medial cortical field projections illustrated in Fig. 5A). • Taken together, the cortical projections to the PAG constitute a set of partially overlapping but largely complementary, rostrocaudally oriented columns of cortical inputs spanning the longitudinal axis of the PAG. The organization of these projections is consistent with the hypothesis that discrete cortical inputs preferentially target discrete longitudinal columns of PAG neurons and can, thereby, selectively trigger or modulate the somatic, autonomic and antinociceptive functions co-ordinated within the PAG. In this regard, 384

it is noteworthy that most of the cortical fields projecting to the PAG have been implicated in many of the functions classically associated with the PAG, such as emotion, autonomic regulation and antinociception x42-44. A key issue for current research is the degree to which these cortical (and other 'input') columns articulate with the lateral, ventrolateral, dorsolateral and dorsomedial columns of output neurons defined by retrograde labeling. Experiments to simultaneously 'label' input and output columns and to compare these with 'functional' columns are an important priority. Subcortical forebrain connections of the PAG. The existence of robust columnar cortical inputs to the PAG raises the possibility that there are a number of discretely organized channels by which integrated cortical activity might directly influence the neural operations of the PAG, a structure usually presumed to be somewhat remote from direct cortical or cognitive influence. Notwithstanding the novelty of these extensive and organized cortical inputs, it would be a mistake to ignore the massive subcortical forebrain projections to the PAG, because many of these subcortical projections terminate more densely in PAG than do cortical inputs. The analysis of forebrain subcortical inputs is still in its infancy, but the principle of columnar organization appears to be as characteristic of these inputs as those from the cortex, brainstem and spinal cord. Only two subcortical inputs to the PAG have been studied in any detail - those from the medial preoptic area (MPO) and the central nucleus of the amygdala (CNA). Medial pre@tic area. The MPO is a complex region consisting of several distinct nuclei lateral to the walls of the third ventricle at the base of the telencephalon 15. The MPO has been strongly implicated in neuroendocrine (gonadal steroid) regulation, sexual behavior, thermal regulation and sleep 46'17. The MPO is also one of the most strikingly sexual dimorphic structures in the brain 4~'~'~. Retrograde tracing experiments showed that neurons projecting to the PAG are present in all of the nuclear groups of the MPO except for the medial subdivision of the medial preoptic nucleus. Anterograde labeling experiments usng PHA-L and wheat-germ agglutinin (WGA)-HRP demonstrate that the projection from the MPO to the PAG has a striking degree of columnar specificity throughout the entire rostrocaudal extent of the PAG (Fig. 5B, left panels). At rostral levels, the projection is largely restricted to the dorsomedial and dorsolateral columns of the PAG. However, further caudally the projection becomes distinctly bi-columnar with dense labeling in terminal fields in the dorsomedial and lateral PAG columns and much lighter labeling in dorsolateral and ventrolateral columns (Fig. 5B, left, bottom two panels). Tracer injection into the MPO also produced retrograde labeling of neurons in the PAG, and furthermore, the reciprocal pathway revealed a degree of columnar organization, that is, many ventrolateral PAG cells, but few lateral PAG cells, projected to the MPO. Central nucleus of the amygdala. The CNA is a nodal point for intra-amygdalar circuits, and receives extensive afferents from structures that include certain of the cortical fields projecting to the PAG specifically anterior insular cortex and a rostral component of posterior insular cortex "~° (M. T. TINS, VoL 17, No. 9, 1994

A 2.2

Fig. 7. (Right.) Series of coronal sections through the caudal one-half of the periaqueductal gray (PAG); rostra/ is to the left, caudal is to the nght. The approximate boundaries of the lateral and ventrolateral PAG are indicated in row 3. For each of the noxious manipulations the distribution of Fos-like immunoreactive neurons is illustrated by filled circles. Labeling of Fos was bilateral, although for the purpose of illustration only one side has been plotted. All experiments were carried out in halothane-anesthetized rats. Each of the deep noxious manipulations (rows 5-9) elicited a s~gnificant increase in the number of Fos-positive neurons in the ventrolateral PAG. In contrast, the cutaneous noxious manipulation (radiant heat) increased expression of fos predominantly in the lateral PAG (row 2). Note that the ventrolateral PAG region from which quiescence and hyporeactivity were elicited in the freely moving rat (row 4), is the same PAG region in which there was a dramatic increase in the number of Fos-/ike immunoreactive neurons following each of the deep noxious manipulations. Modified from Ref. 69.

Shipley and M. Ennis, unpublished observations). The CNA has been implicated in the regulation of defensive behaviors and antinociception 5L5~, and has been shown also to be an important component in circuits that mediate the autonomic responses accompanying conditioned emotional (fear) reactions 5x54. It had been known for some time that the CNA projects to the PAG :~7'5~ but the organization of its connections with the PAG, in light of PAG columnar organization, was not established until recently 56. Retrograde labeling experiments show that the projection to the PAG originates mainly in the medial subdivision of the CNA. Anterograde tracing from the CNA demonstrates robust anterograde labeling throughout the rostrocaudal extent of the PAG. As was the case for other afferents to the PAG, the CNA projection to the PAG exhibits a marked topographic columnar specificity. At rostral levels, the projection is moderate and most labeled fibers aggregate along the wall of the aqueduct [Fig. 5B (right), top two panels]. Beginning at the level of the occulomotor complex, the projection density increases markedly and forms dense dorsomedial and lateral-ventrolateral input columns, separated by the dorsolateral PAG which contained only moderate labeling [Fig. 5B (right), third panel]. Caudally, there is heavy labeling in the lateral and ventrolateral, but neither the dorsomedial or dorsolateral PAG columns. Thus, the principle of longitudinal columns of terminal labeling is a key feature of inputs to the PAG from the CNA. As was the case for the MPO, retrograde tracer injections in the CNA also produced labeling of neurons in the PAG, particularly in its rostral half. Thus, there exist two-way projections from the PAG and the CNA, although these projections do not seem to have point-to-point reciprocity. Ascending pathways. Previous research ~7'57-6°, as well as preliminary results from our own laboratories, indicates that the PAG has extensive projections to midline, intralaminar and reticular thalamic nuclei, the hypothalamus and the basal forebrain (for example, the CNA and MPO). However, with the exception of the CNA and the MPO, these ascending projections are poorly characterized, and it is premature at this time to draw any conclusion regarding their columnar organization. TINS, Vol. 17, No. 9, 1994

A 1.7

A 1.2

A O.7

/-.:.\ ,-I I/"-X.~, ..! i/~ .:~ ", .:: ;

('\

t

;" "

,

,f

".

A 0.2

.) °

Anesthetized control

~k2, '

• "~'~.l~

i

. bll

,'

•

~-."-~,

Radiant heat

skin

7 ~

• Lateral PAG [ ] Ventrolateral PAG

I'

Sites eliciting quiescence and

)

t

I'

'~,

N

N ~

hyporeactivity following kainic

m

acid microinjection

/

/~.\,

°\

,,~ ..'.'

..,,,

"" \

J

\\/.%.2/

. " 1 # . - #1

j

•

'\

..[''1

Intravenous

5-HT

{'t

,. &

Intraperitoneal acetic acid

. .~ \.~" /

/

,

.o.°

• ...~

!

,

.>< /

O..o.• ..;. t/

"..':.'.

%% :

:".\\

:~

--

-;=',7" t ,\

.~

]~

Kneejoint kaolin and carrageenan

(Type IV)

S

.:..4

j

, \

•.%

/

" °• ,

" "

// "k

Triceps surae carrageenan

(Type IV)

( -

".;:.; to-~

i

(

Deep neck muscle formalin

R e l a t i o n s of a n a t o m i c a l and functional c o l u m n s : c o i n c i d e n c e or c o n g r u e n c e ?

Thus, compelling evidence has been provided for three manifestations of columnar organization in the PAG: (1) discrete functions (for example, defensive strategies, behaviorally coupled cardiovascular changes and antinociception) are represented along longitudinal neuronal columns that span a considerable extent of the rostrocaudal axis of the PAG; (2) 385

A

B

fined to the input terminal column are preferentially influenced by selective activation of the MPO. 400 Injections of PHA-L were made 400 1 into the MPO to anterogradely label MPO axons, and terminals in 300" 300' the PAG. Ten days later, microwires were used to stimulate the E .E MPO-PAG pathway for 45-90 rain i= 200" I'200 to induce production of the immediate-early gene protein, FOS, 100' 100 in the PAG. The animals were processed histologically by a double-labeling technique to enable 0 0 simultaneous visualization of MPOacid acid acid acid PAG terminals and FOS. The results obtained showed that there C D 500- Social behavior 500" Defensive behavior was a remarkable anatomical overlap of the MPO terminals and FOSexpressing neurons in the PAG. 400' 400 The tight registration of input terminal columns and activated target cells suggests that the primary 300 300 action of input columns is to inE fluence the PAG target cells in E I-200 200 equally discrete columns. Furthermore, when MPO stimulationinduced FOS expression was com100" 100" bined with retrograde labeling of PAG neurons projecting to the 0' 0' rostral ventral medulla, it was acid acid acid acid found that a significant proportion of the PAG neurons activated by Experimental group Experimental group stimulation of the MPO were outFig, 8. Histogram showing for each manipulation the time spent in each of the following behavioral put neurons that send direct procategories: (A) quiescence: hyporeactive and immobile; (B) non-social behavior: cage exploration jections to the medulla. Taken and self grooming; (C) social behavior." investigation or grooming of partner; and (D) defensive behavior: defensive alerting, defensive uprights, backing or forward locomotion away from partner together, these results support the and reactive immobility (freezing). The manipulations were: control no noxious manipulation; hypothesis that the anatomical deep, bilateral injection O.05ml, 5% formalin into deep dorsal neck muscles; acetic acid, i.p. registration of input and output injection O.5mL 3.5% acetic acid; kainic acid, microinjection of 40pmol of kainic acid into columns provides the neural subventrolateral periaqueductal gray. For detailed description of methods and results see Refs 9, 10 strate for functional columnar and 69. Modified from Ref. 69. organization. A sexual column in the PAG? medullary projecting PAG output neurons are organ- Further investigation of the MPO-PAG pathway ized as discrete longitudinal anatomical columns; and provides evidence for another potential 'anatomical(3) ascending and descending afferents to the PAG functional column' in the PAG. The PAG has long have terminal fields that are co-extensive with longi- been associated with sexual behavior (that is, tudinal columns as defined in (1) and (2). lordosis63), although less is known about this than the Registration of input and ou~ut columns. It would defensive, autonomic and antinociceptive functions of seem logical that inputs preferentially target, and the PAG. The recent development of antibodies to influence, the PAG neurons lying within the afferent gonadal steroid receptors has led to a mapping of terminal column. Indeed, Golgi studies of the these receptors in the brain. Neurons expressing PAG 61'62 suggest that the dendritic arbors of many estrogen receptors (ER) are preferentially located in PAG cells ramify largely within the confines of a single brain structures that are implicated in gonadal steroid longitudinal column. However, some PAG neurons function and reproductive behavior, and one of the have dendritic arbors that ramify extensively within largest populations of ER neurons is in the MPO. the PAG, raising the possibility that the discrete Also, the PAG has a large number of neurons that afferent input columns might become 'diffused' in the express ER (mostly in the dorsomedial and lateral PAG by acting on dendrites of neurons distributed columns of the PAG). These are the same columns across multiple PAG neuronal columns. Therefore, it that are targeted by inputs from the MPO. A recent is important to determine the degree to which study shows that PHA-L-labeled axons and terminals columnar inputs preferentially influence PAG neurons from the MPO overlap to a remarkable extent with restricted to the afferent column. the ER neurons in the PAG (M. T. Shipley, T. Rizvi, Recent studies in one of our laboratories (M. T. S.) A. Murphy and M. Ennis, unpublished observations). have used a novel experimental strategy to address It was further shown by retrograde labeling experthis question. These studies indicate that, at least for iments that many of these PAG-ER neurons have the MPO, PAG neurons whose cell bodies are con- direct projections to the medulla. Taken together with 500"

386

OtJin~cnnc~

500 t

Non-social behavior

TINS, Vol. 17, No. 9, 1994

the findings that the MPO input preferentially activates PAG neurons in the MPO input columns, including neurons projecting to the medulla, it is reasonable to suggest that there might be a functional MPO-PAG-medullary circuit that engages preautonomic and, perhaps, somatic pre-motor neurons involved in spinal sexual reflexes or sexual behaviors, or both. The relation of this potential 'sex column' to other functional columns in the PAG is not known but is an important issue for future study. Functional-anatomical mapping of cardiovascularrelated neurons in the PAG. One of the earliest and strongest indications of functional columns in the PAG came from experiments showing that discrete activation of the lateral column produced pressor responses while activation of the ventrolateral column elicited depressor responses 1'11'~2. In view of this, it seemed not unreasonable to suspect that neurons in the PAG might respond selectively to peripheral manipulations that induced changes in blood pressure. To investigate this, the expression of the product of the immediate-early gene, FOS, was used to map the locations of PAG neurons activated by peripherally acting pressor (phenylephrine) and depressor (nitroprusside) drugs. As shown in Fig. 6, the results were quite dramatic. Raising pressure peripherally induced expression of FOS in neurons in the lateral and dorsolateral columns of the PAG. In contrast, decreasing pressure by reducing sympathetic vasomotor tone induced FOS expression in neurons restricted largely to the ventrolateral column of the PAG. Animals receiving saline injections of the same volume as the cardiovascular drugs had few FOS-positive cells in the PAG, indicating that changes in blood volume alone had little effect on PAG neurons. These findings indicate that changes, induced peripherally, in resting arterial pressure activate discrete columns of PAG neurons. How these 'pressor and depressor activated' input columns relate to the PAG output neurons, whose activation produces selective changes in arterial pressure, remains to be determined. Deep and cutaneous 'pain' columns in the PAG. 'Pain' not only signals the existence of an injurious stimulus but perhaps, even more importantly, signals the need for behavioral change 65. Many years before this conclusion was drawn, the eminent English physician Sir Thomas Lewis 66, in summarizing extensive clinical observations, concluded that pain arising from deep (muscular, articular and visceral) and superficial tissues produced quite different patterns of behavioral and physiological reactions. The reactions to deep or superficial pain described by Lewis bear a striking resemblance to the reactions produced by excitation of the ventrolateral or lateral columns of the PAG (Table I). The resemblance of the PAG-elicited reactions to those typical of deep and cutaneous pain responses suggests that the ventrolateral and lateral PAG might play important roles in co-ordinating the different responses to deep and cutaneous pain. Spinal afferents carrying deep (somatic and visceral) and cutaneous nociceptive information project to the PAG 67. However, it is not known if deep or cutaneous nociceptive afferents differentially activate the ventrolateral and lateral columns of the PAG. A series of experiments by Keay and colleagues 68'69 studied this question by examining the distribution of FOS-like immunoreactive cells in the PAG of halothane~NS, Vol. 1~ No. 9, 1994

TABLE I. Physiological responses to pain or to PAG excitation

Reaction to pain

Reaction to excitation of the PAG

Deep (i) Hypotension, bradycardia (ii) Quiescence, inactivity, nausea, tonic contractions of trunk musculature, localized hyperalgesia

Ventrolateral (i) Hypotension, bradycardia (ii) Quiescence, hyporeactivity (iii) Opioid-mediated analgesia

Superficial (i) Hypertension, tachycardia (ii) Brisk protective reflexes and strong emotional reactions

Lateral (i) Hypertension, tachycardia (ii) Strong emotional reactions: confrontation, flight (iii) Non-opioid-mediated analgesia

Physiological responseselicited by pain arisingfrom deep (musculoskeletal and visceral) and superficial (cutaneous and mucous membranes of body orifices) structures66. The ventrolateral or lateral PAG was stimulated by excitatory amino acids. Abbreviation: PAG, periaqueductal gray.

anaesthetized rats subjected to a range of noxious stimuli chosen to activate selectively cutaneous, deep somatic or visceral structures. It was found that deep somatic noxious stimulation (injection of algesic substance into muscle or joint) and visceral noxious stimulation (i. p. injection of acetic acid; i.v. injection of 5-HT) preferentially stimulated expression of FOS in the ventrolateral PAG, whereas cutaneous noxious stimulation (radiant heat of the skin) led to expression of FOS predominantly in the lateral PAG (Fig. 7). In a related set of experiments 69, the behavior elicited by microinjection of KA into the ventrolateral PAG was compared with that elicited by several of the deep noxious manipulations. As seen in Fig. 8, the reactions of quiescence and hyporeactivity produced by deep somatic or visceral noxious stimuli were quite similar to those produced by injection of KA into the ventrolateral PAG. In contrast to cutaneous pain, which often can be controlled to some extent, and from which an animal might be able to escape, deep pain is both inescapable and usually impossible to control. A quiescent and hyporeactive reaction, such as that mediated by the ventrolateral PAG, might represent one way for the animal to reduce discomfort and limit interactions which might increase pain (for example, touch, pressure or movement of the painful part of the body). Taken together, the FOS and behavioral data provide strong initial evidence that the lateral and ventrolateral PAG neuronal columns are important regions for co-ordinating different responses to different classes of pain 69. The elucidation of the ascending pathways that selectively convey deep and cutaneous nociceptive information, to ventrolateral and lateral PAG columns, as well as the identification of the specific populations of PAG neurons activated by different nociceptive signals, are important questions which await future study.

Concluding remarks For more than two decades, research on the PAG has been compartmentalized among workers who have viewed the PAG through conceptual filters set to detect antinociception, autonomic changes, defensive behavior, aversion, fear, vocalization or lordosis. The PAG is central to all of these important functions, but how and why? The conceptual outlook presented here 387

is also based on a filter but a somewhat broader filter that is shaped by our still growing appreciation of columnar organization within the PAG. Our current research is guided by the working hypothesis that the PAG is a structure that plays a pivotal role in coordinating different strategies for dealing with different sets of environmental demands. Clearly, the lateral and ventrolateral PAG columns co-ordinate fundamentally different classes of coping strategies. Thus, the lateral PAG neuronal column mediates what might be classified as active coping strategies, that is, either confrontation or flight from a source of usually escapable threat or pain, whereas the ventrolateral PAG neuronal column mediates a more passive coping strategy that might function to lessen the physiological and emotional impact of an inescapable stressful or painful encounter and help to promote healing and later recovery (see also Refs 8, 12, 13, 70 and 71). To summarize, the PAG lies at a crossroads for a multitude of neural networks that co-ordinate distinct strategies for coping with different classes of threatening, stressful or painful stimuli. Central to such strategies are the rapid, integrated somatic, autonomic and antinociceptive adjustments that are critical to the animal's survival. The organization and regulation of the PAG neural networks that trigger the specific components of such co-ordinated responses involve longitudinally organized PAG columnar circuits. The definition of the neural operations that regulate the moment-to-moment co-ordination of these PAG columns represents a major challenge for future research.

Selected references 1 Bandler, R., Carrive, P. and Zhang, S. P. (1991) Prog. Brain Res. 87, 269-305 2 Bandler, R., Carrive, P. and Depaulis, A. (1991) in The

Acknowledgements Theauthors'research describedm this review wassupported

by grantsto R.B. from the Austrahan NHMRC the National Heart Foundationof Austraha, the Uive and VeraRamaootti Foundationand the H. F. Guggenhelm Foundation, and NIHgrants RO1 NS20643, NS24698, N529365and OC O2588 to M. ZS, We especia//ythank

KevinKeayand M. Matt Enmsfor theJrassistancewith preparationof figures, and for their helpful commentson ear/ierdraftsof the manuscript. Thanks also to GalTAstonJonesand Max Bennett for their commentson the manuscnpt, 388

Midbrain Penaqueductal Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 1-8, Plenum Press 3 Shipley, M. T., Ennis, M., Rizvi, T. A. and Behbehani, M. M. (1991) in The Midbrain Penaqueductal Gray Matter: Func-

tional,

Anatomical

and

Neurochemica/

Organization

(Depaulis, A. and Bandler, R., eds), pp. 417-448, Plenum Press 4 Bandler, R. (1988) in Progress in Psychobiology and Physiological Psychology (Vol. 13) (Epstein, A. and Morrison, A., eds), pp. 67-154, Academic Press 5 Goodchild, A. K., Dampney, R. A. L. and Bandler, R. (1982) J. Neurosci. Methods 6, 351-363 6 Lipski, J., Bellingham, M. C., West, M. J. and Pilowsky, P. (1988) J. Neurosci. Methods 26, 169-179 7 Bandler, R. and Carrive, P. (1988) Brain Res. 439, 95-106 8 Bandler, R. and Depaulis, A. (1991) in The Midbrain Peri-

aqueductal Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 175-198, Plenum Press 9 Depaulis, A., Keay, K. A. and Bandler, R. (1992) Exp. Brain Res. 90, 307-318 10 Depaulis, A., Keay, K. A. and Bandler, R. (1994) Exp. Brain Res. 99, 75-83 11 Carrive, P. (1991) in The Midbrain Periaqueducta/ Gray

Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 67-100, Plenum Press 12 Lovick, T. A. (1991) in The Midbrain Periaqueductal Gray

Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 101-120, Plenum Press 13 Lovick, T. A. (1993) Prog. Neurobio/. 40, 631-644 14 Carrive, P. and Bandler, R. (1991) Brain Res. 541, 206-215 15 Carrive, P. and Bandler, R. (1991) Exp. Brain Res. 84, 599-606

16 Nakai, M. and Maeda, M. 17 Mancia, G., Baccelli, G. Physiol. 223,925-933 18 Mancia, G., Baccelli, G. Physiol. 227, 536-542 19 Besson, J-M., Fardin, V.

(1994) Neuroscience58, 777-783 and Zanchetti, A. (1972) Am. J. and Zanchetti, A. (1974) Am. J.

and Oliveras, J-L. (1991) in The Midbrain Periaqueductal Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and

Bandler, R., eds), pp. 121-138, Plenum Press 20 Lewis, V. A. and Gebhardt, G. F. (1977) Brain Res. 124, 283-303 21 Yaksh, T. L., Yeung, J. C. and Rudy, T. A. (1976) Brain Res. 114, 83-104 22 Morgan, M. M. (1991) in The Midbrain Periaqueducta] Gray

Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 139-150, Plenum Press 23 Miczek, K. A., Thompson, M. L. and Shuster, L. (1982) Science 215, 1520-1522 24 Abols, I. A. and Basbaum, A. I (1981) J. Comp. Neurol. 201, 285-297 25 Holstege, G. (1991) Prog. Brain Res. 87, 307-421 26 Van Bockstaele, E., Aston-Jones, G., Pieribone, V. A., Ennis, M. and Shipley, M. T. (1992)J. Comp. Neurol. 309, 305-327 27 Mantyh, P. W. (1983) J. Neurophysio/. 49, 582-595 28 Redgrave, P. and Dean, P. (1991) in The Midbrain Peri-

aqueductal Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 199-209, Plenum Press 29 Conti, F., Barbaresi, P. and Fabri, M. (1988) Neuroscience 24, 629-633 30 filing, R. B. and Graybiel, A. M. (1986) Neuroscience 18, 373-394 31 Herbert, H. and Saper, C. B. (1992) J. Comp. Neuro/. 315, 244-253 32 Ennis, M. etal. (1991) Soc. NeurosciAbstr. 17, 611 33 Blomqvist, A. and Craig, A. D. (1991) in The Midbrain

Penaqueducta/ Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 345-363, Plenum Press 34 Wiberg, M., Westman, J. and 81omqvist, A. (1987) J. Comp. Neurol. 264, 92-117 35 Yezierski, R. P. (1988) J. Comp. Neurol. 267, 131-146 36 Keay, K. A. and Bandler, R. (1992) NeuroscJ. Left. 139, 143-148 37 Beitz, A. J. (1982) Neuroscience 7, 133-159 38 Mantyh, P. W. (1982) J. Comp. Neurol. 206, 146-158 39 Meller, S. and Dennis, B. (1986) Neuroscience 19, 927-964 40 Neafsey, E. J., Hurley-Guis, K. M. and Arvantis, D. (1986) Brain Res. 377, 261-270 41 Sesack, S. R., Deutch, A. Y., Roth, R. H. and Bunney, B. S. (1989) J. Comp. NeuroL 290, 213-242 42 Cechetto, D. F. and Chen, S. J. (1990) Am. J. Physiol. 258, R245-R255 43 Hardy, S. P. G. and Holmes, D. E. (1988) Exp. Brain Res. 73, 249-255 44 Yasui, Y., Breder, C. D., Saper, C. B. and Cechetto, D. F. (1991) J. Comp. Neuro/. 303, 355-374 45 Simerley, R. B. and Swanson, L. W. (1986) J. Comp. NeuroL 246, 312-341 46 Clemens, J. A., Smalstig, E. B. and Sawyer, C. H. (1976) Endocrinology 99, 728-735 47 Marlsbury, C. W. (1971) PhysioL Behav. 7, 797-805 48 Gorski, R. A, Harlan, R. E., Jacobson, C. D., Shryne, J. E. and Southam, A. M. (1980) J. Comp. Neurol. 193, 529-539 49 Simerly, R. B., Swanson, L. W. and Gorski, R. A. (1984) J. Comp. Neurol. 225, 151-166 50 Price, J. L., Russchen, F. T. and Amaral, D. G. (1987) in Handbook of Chemical Neuroanatomy (Vol. 5) (Bj6rklund, A., H6kfelt, T. and Swanson, L. W., eds), pp. 279-388, Elsevier 51 Frenk, H., McCarty, B. L. and Liebeskind, J. C. (1978) Science 200, 335-337 52 Roozendaal, B., Wiersma, A., Driscoll, P., Koolhaas, J. M. and Bohus, B. (1992) Brain Res. 596, 3 5 - 4 0 53 Davis, M., Rainnie, D. and Cassell, M. (1994) Trends Neurosci. 17, 208-214 54 LeDoux, J. E., Iwata, J., Cicchetti, P. and Reis, D. J. (1988) J. Neurosci. 8, 2517-2529 55 Hopkins, D. A. and Holstege, G. (1978) Exp. Brain Res. 32, 529-547 56 Rizvi, T. A., Ennis, M., Shipley, M. T. and Behbehani, M.

TINS, VoL 1Z No. 9, 1994

(1991) J. Comp. Neurol. 303, 121-131 57 Carstens, E., Leah, J., Lechner, J. and Zimmerman, M. (1990) Neuroscience 35, 609-626 58 Meller, S. T. and Dennis, B. J. (1991) Neuroscience 40, 191-216 59 Reichling, D. B. and Basbaum, A. I. (1992) Neuroscience 42, 183 -200 60 Rinvik, E. and Wiberg, M. (1990) Anat. Embryo/. 181, 577- 584 61 Gioia, M., Tredici, G. and Bianchi, R. (1985) Exp. Brain Res. 58, 318-332 62 Manyth, P. W. (1982) J. Comp. NeuroL 204, 349-363 63 Ogawa, S., Kow, L-M., McCarthy, M. N., Pfaff, D. W. and Schwartz-Giblin, S. (1991)in The Midbrain Periaqueductal

Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp.

211-235, Plenum Press 64 Li, Y. W. and Dampney, R. A. L. (1994) Neuroscience 61, 613-634 65 Wall, P. D. (1979) Pain 6, 253-264 66 Lewis, T. (1942) in Pain, pp. 33-48, Macmillan 67 Yezierski, R. P. (1991) in The Midbrain Penaqueductal Gray Matter: Functional, Anatomical and Neurochemical Organization (Depaulis, A. and Bandler, R., eds), pp. 365-386, Plenum Press 68 Keay, K. A. and Bandler, R. (1993) Neurosci. Left. 154, 23-26 69 Keay, K. A., Clement, C. I., Owler, B., Depaulis, A. and Bandler, R. (1994) Neuroscience 61,727-732 70 Duggan, A. W. (1983) Proc. AusL Physiol. Pharmacol. Soc. 14, 218-240 71 Benus, R. F., 8ohus, B., Koolhaas, J. M. and van Oortmerssen, G. A. (1991) Experientia 47, 1008-1019

Magnetoencephalographyin studiesof human cognitivebrain function

~4V4~ Z

R i s t o N ~ f i t ~ n e n , R i s t o J. I l m o n i e m i a n d K i m m o A l h o

Magnetoencephalography provides a new dimension to the functional imaging of the brain. The cerebral magnetic fields recorded noninvasively enable the accurate determination of locations of cerebral activ@ with an uncompromized time resolution. The first whole-scalp sensor arrays have just recently come into operation, and significant advances are to be expected in both neurophysiological and cognitive studies, as well as in clinical practice. However, although the accuracy of locating isolated sources of brain activity has improved, identification of multiple simultaneous sources can still be a problem. Therefore, attempts are being made to combine magnetoencephalography with other brainimaging methods to improve spatial localization of multiple sources and, simultaneously, to achieve a more complete characterization of different aspects qf brain activi& during cognitive processing. Owing to its good time resolution and considerably better spatial accuracy than that provided by E E G, magnetoencephalography holds great promise as a tool for revealing informationprocessing sequences of the human brain.

scalp :'~, and containing even more than 100 channels7, are now available. Figure 1 illustrates an MEG experiment. The subject is presented with visual-pattern stimuli while an array of SQUID sensors around their head records the variations ifl the magnetic field. The data are subjected to an inverse-problem algorittma to obtain an estimate for the distribution of the activity in the brain, which will then be displayed on cross-sectional anatomical images of the same subject. The characteristics of MEG form a unique combination of noninvasiveness, simple and quick procedures, good accuracy in locating sources, and millisecond-scale time resolution.

Risto Na3t3nen and KimmoA/ho are at the Cognitive Psychophysiology ResearchUnit, Dept of Psychology, PC) Box 11, FIN-O0014 Universityof He/sinkll Finland, and RistoJ. I/moniemiis at the BioMag Laboratory, Medical Engineenng Centre, He/slhki University Central Hospital, FIN-O0290 He/sinki, Finland.

Neuromagnetic fields

It is believed that the magnetic field detectable outside the head is produced by currents initiated at the synapses, and guided postsynaptically by cell structure. Magnetic-field lines encircle the flow path of this primary current, extending outside the skull. Because pyramidal cells are predominantly oriented Electric currents in the brain produce a change in the perpendicular to the cortex, the direction of the magnetic field which can be detected outside the head primary current is also perpendicular to the cortex. with SQUID (superconducting quantum interference Magnetoencephalography is, therefore, most sensidevice) magnetometers. The signals measured can be tive to activity in the fissural cortex, where the used to compute the distribution of cerebral activity as current is oriented parallel to the skull, whereas it a function of time. This method, called magneto- does not detect sources that are oriented exactly encephalography 1 or MEG, is closely related to EEG radially to the skull. Because MEG detects only the in which the electric-field pattern on the scalp is tangential component of the primary current, amplimeasured. The main advantages of MEG compared tude comparison between differently oriented sources with EEG are its superior spatial accuracy and ease of is possible only if source orientations can be estiuse, particularly when a large number of measure- mated, for example, on the basis of magnetic resonment channels are involved. On the other hand, EEG ance imaging (MRI). complements MEG in detecting source components The magnetic field produced by a single postnot detected by MEG. synaptic potential is too weak to be detected outside Magnetic fields generated by the human brain were the head. Instead, what is detected is macroscopic first measured with an induction-coil magnetometer in coherent activity of thousands of neurons. Still, the late 1960s2; the subsequent development of cerebral magnetic fields are so weak that measureSQUID magnetometers a improved the sensitivity so ments are preferably performed inside magnetically that real-time spontaneous activity4 and stimulus- shielded rooms. Noise cancellation is improved by elicited magnetic fields5'6 could be measured. The measuring gradients of the magnetic field instead of early technological difficulties were overcome in the the field itself. In quiet environments, disturbances late 1980s so that instruments covering the whole can be sufficiently compensated for by using elaborate TINS, Vol. 17, No. 9, 1994

© 1994.elsevierScienceLtd

389