J. Anat., Lond. (1962), 96, 2, pp. 287-248 237With 6 plates and 1 text-figurePrinted in Great Britain 44

PLASMA MEMBRANE CONTACTS IN THt CENTh3LNERVOUS SYSTEM - X e

BY A. PETERSDepartment of Anatomy, University of Edinburgh

INTRODUCTION

In low-power electron micrographs a thin, dense line is visible at the surfaces ofcells. This line, the plasma membrane, is about 75-80 A. thick and in many tissues isseparated from the surrounding connective tissue elements by a basement membrane.At high resolution, Robertson (1957) has shown that in potassium permanganate

fixed material this 75 A. thick plasma membrane has a triple-layered structure,consisting of two dark layers, each 25-30 A. wide, separated by a lighter zone of thesame width. A somewhat similar appearance is also obtained from sections ofosmium-fixed material (for example, see Karrer, 1960a, b). Robertson (1959) hasapplied the general term of 'unit membrane' to such triple-layered-structures andhas discussed both the chemical significance of the appearance and its relation tothe cell membrane of classical histology.

In recent years, it has been found, with few exceptions (Karrer, 1960a, b;Robertson, 1959), that when adjacent cells come together their plasma membranesremain separated by a distance of 150-200 A. (see Porter, 1959), but one situationin which the gap between adjacent plasma membrane surfaces is eliminated is in themyelin sheaths of the peripheral (see Robertson, 1959) and central (Peters, 1960b,1961) nervous systems. Here regions of the plasma membrane of the myelin-forming cell come into close contact. Moreover, as shown by Maturana (1960) andPeters (1960a, b, 1961) in the central nervous system, where there is little cytoplasmon the outside of the sheath, contacts may also occur between the outsides ofadjacent sheaths. Since central sheaths are considered to be formed by glial cells,these observations suggested that contacts may be formed by glial cells elsewherein the central nervous system and, as will be shown, this has proved to be the case.The optic nerve has been chosen as the material for this investigation, since as

part of the central nervous system it has the advantage that it can be readilyremoved intact and easily fixed in short lengths without damage. Further, sincethe nerve fibres run in one direction they can be sectioned in any given plane.

MATERIALS AND METHODS

Short lengths of the optic nerves of adult rats, mice and toads (Xenopus laevis,Daudin) and of 7- and 14-day post-natal rats, were fixed for 1 hr. at 40 C. in thechromate-osmium tetroxide mixture of Dalton (1955) after which they werewashed in 10 % alcohol and dehydrated. The tissue was then embedded in Araldite(Glauert & Glauert, 1958) and sectioned. The sections were mounted, without theuse of a supporting film, and stained with either lead acetate (Watson, 1958) or

potassium permanganate (Lawn, 1960) before examination on a Metropolitan-Vickers, EM 6, electron microscope. For staining of the plasma membranes thebest results were obtained with potassium permanganate.To obtain measurements of the spacings and widths of plasma membranes and

their component layers, negatives were examined with a travelling microscope andwith a travelling photo-microdensitometer attached to a pen recorder, which alsogave useful information about the relative intensities of the various layers and lines.

DESCRIPTION

The plasma membrane of the cell appears in electron micrographs as a triple-layered structure consisting of two limiting dark layers separated by a lighter layer,and throughout the following account the dark layer adjacent to the cytoplasm willbe referred to as the cytoplasmic dark layer, and that on the outside of the membraneas the surface dark layer of the plasma membrane. In Robertson's (1957) potassiumpermanganate fixed preparations the two dark layers have approximately the sameelectron density, but in the present osmium fixed material, stained by eitherpermanganate or lead, the appearance is somewhat different in that the cytoplasmicdark layer is more intense than the surface dark layer (P1. 1, fig. 1; P1. 6, fig. 11, P).The overall dimensions of the membranes appear to be similar to those given byRobertson (1957) and other workers (for example, see Karrer, 1960a, b and Mercer& Schaffer, 1960), that is 70-80 A., although some difficulty is encountered inobtaining exact measurements, since the layers of the membrane do not appear insections as discrete lines, but they have indefinite boundaries (PI. 1, fig. 1; PI. 6,fig. 11, P). This problem of obtaining overall measurements also arises in respectof the sites where plasma membranes come into contact and for this reason the mostaccurate measurements are made from the centres, or mid-lines, of the layers, whichcan be identified as definite peaks in the micro-photodensitometer recordings.Although overall measurements must be more of an approximation, they have beenmade as far as possible for comparison with the measurements of other workers.

Since contacts between plasma membranes are formed so extensively within themyelin sheath, it is appropriate to consider the nature of these contacts beforedescribing those which occur elsewhere in the central nervous system, even thoughthe contacts within the sheath are probably unique in that they are formed bydifferent areas of the plasma membrane of the same cell. In the myelin sheath, theouter surfaces of two regions of the plasma membrane of the myelin-forming cellcome together within the internal mesaxon (PI. 1, fig. 1, M) with the resulting for-mation of the intraperiod line (PI. 1, fig. 1, I) of the sheath, and in the region ofcontact only five layers are visible, three parallel dark layers separated from eachother by intervening light layers. Karrer (1960 a, b) has referred to this five-layeredstructure as a 'quintuple-layered cell interconnexion', but since the word inter-connexion could have functional implications, the term 'quintuple-layered unit'will be used throughout the following account. It can be seen in P1. 1, fig. 1, thatthe limiting dark layers of this quintuple-layered unit are continuous with thecytoplasmic dark layers of the contributing plasma membranes, as are the lightlayers, but where the surface dark layers of the two membranes come together theylose their separate identity and give rise to the intraperiod line, or intermediate

238 A. Peters

Plasma membrane contacts in the central nervous system 239dark layer. The intraperiod line is 20-25 A. thick. This situation is indicateddiagrammatically in Text-fig. 1C. The overall dimension of the quintuple-layeredmesaxon is approximately 150 A., as might be expected from a close contact of twoplasma membranes, each 75 A. thick, and the distance from centre to centre of eachof the limiting, or cytoplasmic, dark layers is 120-125 A.

This quintuple-layered unit of the mesaxon is equivalent to one myelin lamella,since during the later stages in the formation of the sheath, loss of cytoplasm frombetween the turns of the, initially, loosely spiralled mesaxon, allows adjacent turnsto come together, with the resulting formation of compact myelin (Peters, 1960b,1961). When this occurs, as can be seen at the point where the internal mesaxon(PI. 1, fig. 1, M) approaches the inside of the compact myelin of PI. 1, fig. 1, thecytoplasmic dark layers (PI. 1, fig. 1, L) of adjacent quintuple-layered units cometogether to form the major dense, or period, line (PI. 1, fig. 1, D) which is about30 A. thick and throughout the sheath alternates with the intraperiod line.The distance from centre to centre of adjacent major dense lines of the sheath is

referred to as the periodicity or lamellar spacing of the sheath and in central sheathsthe spacing is 115-125 A. The major dense line is the outer line of the sheath andterminates on the outside of the sheath as the pair of plasma membranes separateand become continuous with each other around the tongue of cytoplasm (PI. 1,fig. 1, T1, T2, and T3), while the intraperiod line terminates as the plasma membranesurrounding the tongue of cytoplasm, or tongue process, turns off from the outsideof the sheath (PI. 1, fig. 1, X). Although cytoplasm is in most cases confined to thetongue and to the inside of the sheath, it is sometimes present between the turnsof the spiralled lamellae (PI. 1, fig. 1, Z) when, as would be expected, it is accom-modated by a separation of the layers of the sheath along the major dense line.One other feature of the central myelin sheath worthy of consideration in this

context is the presence of radially arranged thickenings of the intraperiod line(PI. 1, fig. 1; PI. 3, fig. 4, R). These radially arranged thickenings usually extendthroughout the width of the sheath and although they are frequently present inthe mouse and the rat, they are less common in the toad. This feature of the sheathwill be described in more detail elsewhere.Not only are contacts formed between the adjacent layers ofthe plasmamembrane

of the sheath, but they are also formed when the outer surfaces of two adjacentsheaths come together and when the outer surface of the bounding membrane ofthe tongue process comes into contact with either an adjacent sheath or tongueprocess (P1. 1, fig. 1; PI. 2, fig. 2.). In these situations an intermediate line isformed, since the outer surface of a plasma membrane is exposed on the outside ofthe sheath and tongue process. The spacing of the contact is similar to that presentbetween membranes of the sheath.

There are two types of glial cell in the optic nerves examined, oligodendrocytesand fibrous astrocytes (Peters, 1960b; Gaze & Peters, 1961). The former have beendescribed by Schultz, Maynard & Pease (1957) and by Farquhar & Hartmann (1957)and have a rather dark cytoplasm with a well-formed endoplasmic reticulum (PI. 2,fig. 3, 01; PI. 3, fig. 5, Ol and 012). These cells have few processes, which when theyare observed can be traced for only short distances. Their appearance contrastswith that of the fibrous astrocytes which have a less electron-dense cytoplasm

(PI. 2, fig. 2; PI. 8, fig. 4, As) containing a fibrous component, which is especiallyprominent in the processes of the astrocytes (PI. 3, fig. 4; PI. 5, figs. 8 and 9, AsP;PI. 4, fig. 6, AsP1 and AsP2). These processes extend throughout the nerve betweenthe nerve fibres (Gaze & Peters, 1961; Peters, 1960b; Maturana, 1960), and manyreach the surface of the nerve where they spread out to form an outer layer (PI. 6,fig. 10, AsP) which is separated from the pial collagen by a basement membrane(P1. 6, fig. 10). In the rat and mouse optic nerves, where capillaries are present,processes also form end-feet on capillaries (P1. 5, figs. 8 and 9, AsP). The structureof the fibrous astrocytes is somewhat different in the toad from that in the rat andmouse, for while in the former the entire cytoplasm of the cell has a watery appearanceand contains the fibrous component, in the mouse and rat, there are regions of thecell, adjacent to the nucleus (P1. 2, fig. 2; PI. 4, fig. 6, ANu), that are free of thefibrous component; in these regions a well-defined endoplasmic reticulum is present(PI. 3, fig. 4, ER).The outer surfaces of the myelin sheaths and of the membranes surrounding the

tongue processes on their external surfaces form intimate contacts with the plasmamembranes of both types of glial cells (P1. 2, figs. 2, 3; PI. 3, fig. 4). In PI. 2, fig. 3,it can be seen that as the plasma membrane (P) of an oligodendrocyte (01) comesinto contact with the outside of a myelin sheath an intermediate line (I) is formed.A similar situation is found when the outside of a sheath comes into contact withthe plasma membrane (PI. 2, fig. 2, P) of an astrocyte (PI. 2, fig. 2, As) for anintermediate line is again formed (PI. 2, fig. 2, I), while contact between the plasmamembrane of a glial cell and that bounding the tongue process (PI. 2, fig. 2, T) ofthe sheath results in the production of a quintuple-layered unit (PI. 2, fig. 2, Q).In the case of the fibrous astrocytes, however, contacts of this type are mostfrequently formed with the outsides of myelin sheaths when the plasma membraneof the astrocyte overlies a region of the cell in which the cytoplasm is devoid offibrils, for although many sections have been examined only one exception to thishas been observed. This exception is shown in PI. 4, fig. 7, where a plasma mem-brane overlying cytoplasm containing fibrils forms a contact with the plasmamembrane of a tongue process. This will be referred to again later. Elsewhere,when fibrils are present within the cytoplasm of the astrocyte, the overlying plasmamembrane and the outside of a sheath have always been separated by a distanceof 150-200 A. (see PI. 1, fig. 1; PI. 3, fig. 4). Observations made on sites in whichthe cytoplasm of the fibrous astrocyte changes from the fibril-containing to theendoplasmic reticulum-containing type show this point clearly, for while theoutsides of sheaths are in close contact with the latter regions of the cell, a separationoccurs as soon as fibrils are visible within the cytoplasm. Further in the optic nerveof the toad, where the entire cytoplasm appears to be filled with fibrils, no contactshave been observed between the outsides of the sheaths and the plasma membranesof the fibrous astrocytes.

It should be emphasized that contacts are not formed in every situation wherethe plasma membrane of an oligodendrocyte or that overlying the non-fibrousregions of an astrocyte is adjacent to the outside of a sheath, although such contactsare relatively common. The result of such a contact is the addition of an extra layerto the outside of the sheath (P1. 2, figs. 2,3; PI.3, fig. 4) and the spacing from the

240 A. Peter

Plasma membrane contacts in the central nervous system 241centre of the cytoplasmic dark layer of the plasma membrane of the glial cell to thecentre of the outer major dense line of the sheath is 115-120 A., similar to thatwhich has been shown to be typical of spacings within the sheath.A plasma membrane contact may also be formed when either two glial cells or

processes both containing a well-developed endoplasmic reticulum come together.In the optic nerves of young animals it is not always possible to identify suchprocesses with certainty, but in adult nerves plasma membrane contacts have beenfound between the cell bodies of oligodendrocytes (PI. 3, fig. 5). Due to the dis-position of the glial cells, situations in which two oligodendrocytes come togetherare infrequent and it has not been possible to obtain micrographs of sufficient clarityto make exact measurements. An example of this type of contact is shown inPI. 3, fig. 5, where the plasma membranes (P1 and P2) of two oligodendrocytes (01.and 012), which are separated on the right of the micrograph, come into contact (C)towards the left. At one point (Q) the resulting quintuple-layered unit can beobserved.Although the plasma membrane of a fibrous astrocyte overlying cytoplasm

containing fibrils rarely seems to come into contact with the outside of a myelinsheath in the optic nerves of the rat and mouse, it is common to find contactsbetween such membranes and those of adjacent astrocyte processes and cell bodies(P1. 3, fig. 4; PI. 4, fig. 6). These contacts are independent of whether fibrils arepresent (PI. 4, fig. 6), or absent (PI. 3, fig. 4) in the cytoplasm underlying the othermembrane taking part in the contact. Although they do not occur in every situationin which two astrocyte plasma membranes are adjacent to each other (PI. 4, fig. 6),contacts are relatively common and appear to be particularly prevalent betweenastrocyte processes forming both the outer layer of the nerve and the end-feet oncapillaries. In the outer layer of the nerve, where there is often considerableoverlap of processes, the regions of contact may be followed for quite long distances.As before, a quintuple-layered unit (P1. 3, fig. 4; PI. 4, fig. 6, Q) is formed at the siteof contact, but this is somewhat different from those described above in that thecytoplasmic dark layers of the constituent plasma membranes appear to becomethicker and more intense at the site of contact (P1. 4, fig. 6); the apparent thicknessincreases by about 40 A. Furthermore, the distance from centre to centre of each ofthese thickened cytoplasmic dark layers of the quintuple-layered units is 130-135 A.,about 20 A greater than the spacing of quintuple-layered units at other sites. Thiscontrast between adjacent astrocyte plasma membranes is shown diagrammaticallyin Text fig. l B.As mentioned above, in one instance the membrane of a fibrous astrocyte, in a

region of the cell where fibrils are present within the cytoplasm, has been seen tocome into contact with the plasma membrane of a tongue process on the outside ofa sheath. As can be seen in PI. 4, fig. 7, the resulting quintuple-layered unit hasthe same appearance as that between adjacent astrocyte processes (compare withPI. 4, fig. 6), in that the cytoplasmic dark layers of the quintuple-layered unit arethickened and the spacing between their centres is about 130 A.Thus far, the observations have been confined to plasma membrane contacts

between the sheaths and glial elements of the central nervous system. In the ratand mouse optic nerves, however, capillaries are present both within (PI. 5, fig. 8)

242 A. Petersand outside (P1. 6, fig. 10) the nerve. These capillaries examined outside the opticnerve lie in the collagen of the pia mater, external to the outer layer of astrocyteprocesses (PI. 6, fig. 10, AsP). In both situations, the walls of the capillaries are onecell thick and show no fenestrations of either the endothelial cells or the surroundingbasement membrane (P1. 5, fig. 9; PI. 6, fig. 10, B).Within the optic nerve (PI. 5, figs. 8, 9), the outer plasma membranes of the

endothelial cells of the capillaries are separated from the surrounding glial elements(PI. 5, fig. 9, AsP) by a distance of from 400 to 1000 A., this gap being occupied bya thick basement membrane (PI. 5, fig. 9, B), which in the regions of greatestseparation (PI. 5, fig. 9, Z) appears to split into two layers, one layer being appliedto the outside of the endothelial cell and the other to the adjacent glial element.Most of the area around the capillaries is occupied by the end-feet of fibrousastrocytes (P1. 5, figs. 8 and 9, AsP) and as pointed out earlier, their plasma mem-branes are often seen to come into contact at some point as they run together.Gray (1961) has also described contacts between the plasma membranes of clearglial cell processes abutting on capillaries in the cerebral cortex. A somewhatparallel relationship between endothelial cells and astrocyte processes also occursin respect of the pial capillaries (P1. 6, fig. 10), for here, although there may be awide gap between the outside of the capillary and the surface of the nerve, theendothelial cells are always separated from the nervous elements by the layer ofastrocyte processes on the outside of the nerve (P1. 6, fig. 10, AsP).An apparently constant feature of the capillaries in both sites is that where two

endothelial cells abut or overlap, their plasma membranes come into contact(P1. 5, fig. 9; PI. 6, fig. 10, C) to form a quintuple-layered unit (P1. 6, fig. 11, Q).Where the area of overlap between cells is small the area of contact usually occupiesthe whole length of the approximation (P1. 6, fig. 10, C), but where the overlap isextensive, several areas of contact may occur, the membranes coming into contactand separating again at intervals. These quintuple-layered units have the sameappearance as those occurring both in the sheath and between oligodendrocytesfor the spacing from centre to centre of the cytoplasmic dark layers of the unit is120-125 A. and there is no thickening of the cytoplasmic dark layers. However,one distinctive feature of the endothelial cell contacts is that the cytoplasm im-mediately adjacent to the quintuple-layered unit always appears to be darkened,giving the same appearance as that which occurs at terminal bars (Yamada, 1955).

DISCUSSION

Since the myelin sheaths of the central nervous system are formed by glial cells,it is apparent that with the exception of those between endothelial cells, all of thecontacts described in this account involve the plasma membranes of glial cells. Inthe present study, no contacts have been observed in which the axolemma takespart, although Robertson (1959) states that they occur between young developingnerve fibres. Further with the exception of the major dense line of the sheath, allthe contacts are between the outer surfaces of plasma membranes and a quintuple-layered unit always results; this is also true of the myelin sheath in its early stagesof development, when cytoplasm is present between the turns of the spiralledmesaxon (Peters, 1960b).

Plasma membrane contacts in the central nervous system 243The most simple form of the quintuple-layered unit is that formed by plasma

membranes other than those overlying the fibril-containing cytoplasm of thefibrous astrocyte. The overall dimension of such a simple unit is about 150 A.(Text-fig. 1 A). However, while the cytoplasmic dark layer of each of the constituentplasma membranes has been observed to be in continuity with those of thequintuple-layered unit, the surface dark layers appear to lose their separate identityat the point of contact and come together to form the single intermediate darklayer, or in the case of the mesaxon, the intraperiod line. The resulting intermediate

0 0C S -75 A C S -75 A. C I 1 I M S

SC

OA~~~~~~~

SC,

O' ~~~~~Mesaxon-'_150A' 160A

Tongue

A B C ok

Text-figure 1. Diagrammatic representation of the structure of plasma membrane contactsoccurring in the central nervous system. The dimensions of the structures and their relativeintensities, as they appear in electron micrographs, are indicated. Text-fig. IA shows the simpleform of contact which occurs between endothelial cells and between oligodendrocytes, whileText-fig. 1B indicates the type of contact which results when one, or both, of the plasmamembranes belongs to a fibrous astrocyte in a fibril-containing region of the cell. The natureof the contacts present in the myelin sheath is shown in Text-fig. 1 C. For further explana-tion see text. (C, cytoplasmic dark layer; S, surface dark layer; I, intermediate layer ofquintuple-layered unit or intraperiod line of sheath; M, major dense line of sheath.)

layer is quite thin, about 20 A., which is much less than would be expected from asimple contact between adjacent surface dark layers, each 20-30 A. thick. However,it must be remembered that the resolution of the electron microscope is only about10-15 A., so that although these layers have an apparent width of about 20 A., theirtrue width may be much less.Where radial thickenings of the intraperiod line are present within the sheath,

this type of contact is modified. In transverse sections each radial thickeningextends for 60-90 A. along the intraperiod line, which increases in width by about20 A. The lamellar spacing of the sheath is not changed by the presence of these

thickenings, however, which encroach upon the lighter layers of the sheath, betweenthe major dense and intraperiod lines.Another variation is shown by the quintuple-layered unit formed between

endothelial cells, where there is an increase in intensity of cytoplasm on each sideof the area of contact. The appearance is similar to that obtained at the sites ofterminal bars (Yamada, 1955) and has been observed by Maynard, Schultz & Pease(1957) between endothelial cells in the cerebral cortex, while Bennett, Luft &Hampton (1959) have observed a similar appearance in capillaries in other regions.Whether quintuple-layered units are present between endothelial cells in otherparts of the body in association with terminal bars requires investigation, althoughHama (1960) shows that in the blood vessels of the earthworm, where terminal barsare present, the plasma membranes do not appear to come into contact.The second type of quintuple-layered unit, a unit involving at least one plasma

membrane of an astrocyte in a region of the cell where the cytoplasm containsfibrils, differs from the simple form described above in that the cytoplasmic darklayers become thickened. The thickness of each cytoplasmic dark layer becomesabout 40 A., as compared with about 25 A. in areas where the membranes areseparate (Text-fig. 1B). Since the distance from centre to centre of each of thecytoplasmic dark layers is 130-135 A and the overall dimensions are 160-165 A.,some 10-15 A. more than the first type of quintuple unit, it seems likely that this isagain a true contact and that the differences in dimensions are to be accounted forby the thickening of the cytoplasmic dark layers.The two types of cell contact formed by the plasma membrane of the fibrous

astrocyte are associated with the presence or absence of fibrils in that region of thecell, and indicate that the properties and perhaps structure of the plasma membraneare influenced by the underlying cytoplasm. The fibril-containing cytoplasm appearsto be over-riding in its influence in determining the form of a contact, for when theplasma membranes from these areas of the cell come into contact either with anastrocyte plasma membrane overlying cytoplasm from which fibrils are absent, orwith the membrane surrounding a tongue process of a sheath, both cytoplasmicdark layers of the resulting quintuple-layered unit become thickened. On the otherhand, contact between the plasma membrane of an astrocyte, in a region of thecells where fibrils are absent, and the membrane of either a tongue process or anyother part of the surface of a sheath does not lead to a thickening. No variationshave been found in respect of different areas of the plasma membrane of oligoden-drocytes, but here the structure of the cytoplasm appears to be relatively constantthroughout the cell.

Quintuple-layered contacts between the outer surfaces of plasma membraneshave been reported by Gray (1961) in the cerebral cortex of the rat, while Robertson(1961) has shown contacts between endothelial cells in the developing spinal cordof the mouse and states that the plasma membranes of axons may come intocontact in early development. Outside the nervous system, Karrer (1960a, b) hasmade a careful study of the quintuple-layered contacts formed between cells in theepithelium of the cervix uteri and between cardiac muscle cells in the thoracic andpulmonary veins, while Muir (1961) has shown contacts to be present between cellsof the small intestine at the site of the terminal bar. In all these sites the overall

244 A. Peters

Plasma membrane contacts in the central nervous system 245dimensions of the quintuple unit appear to be about 150 A., but it is of interest thatRosenbluth & Palay (1961) have found that in the sheaths of ganglion cells fromthe eighth nerve ganglion of the goldfish, a quintuple unit may become condensedto produce a unit which is only 100 A. thick. When this happens, the intermediateline disappears, so that a triple-layered unit is produced, which although slightlythicker, has an appearance similar to that of a single plasma membrane.The only site where the cytoplasmic surfaces of plasma membranes have been

observed to come together is within the myelin sheath during the formation of themajor dense line (Text-fig. 1 C). This line does not appear to be the result of a contactof the cytoplasmic surfaces of the membranes forming the mesaxon, but a fusion,or interlocking, of the cytoplasmic dark layers of the adjacent membrane. This issuggested by the fact that the distance from centre to centre of adjacent majordense lines of the sheath is 115-125 A., which is the same distance as that betweenthe centres of the cytoplasmic dark layers of the quintuple-layered mesaxon, andnot greater as would be expected from a simple apposition of adjacent layers of thespiralled mesaxon. Such an interlocking of the cytoplasmic dark layers is alsosuggested by the appearance of the major dense line, which although it is notappreciably thicker than each of the separate dark layers coming together in itsformation, is much more intense, indicating an increased concentration of osmio-philic material. This increase in intensity can be readily appreciated from densito-meter recordings and is apparent when the intensity of the major dense line iscompared with that of the cytoplasmic dark layers which exist adjacent to boththe cytoplasm on the inside of the sheath and the cytoplasm within the tongueprocess. This is indicated in Text-fig. 1 C.

Since few examples of contact between the outer surfaces of plasma membraneshave been recorded, it is not possible at this stage to be certain of their significance.In respect of the contacts between the lateral surfaces of cardiac muscle cells in thethoracic and pulmonary veins, Karrer (1960b) has suggested that these may allowfor conduction of excitation between muscle cells. The same suggestion does notappear to be valid for the central nervous system, however, where, in the myelinsheath, the formation of contacts appears to be important in insulating the nervefibre over a great portion of its length, thereby making saltatory conduction possible.It may be that one function of the plasma membrane contacts is to prevent thepassage of substances between cells, so that they must pass through cytoplasm.Such a concept could help to explain the blood-brain and blood-cerebrospinal fluidbarriers, since substances leaving the blood would need to pass through the en-dothelial cells, which at least in the optic nerve and cerebellum (personal observation)are in contact with each other. It is in the capillaries that Davson (1958) suggeststhat the blood-brain barrier resides and this is also considered to be the site byVan Breemen & Clemente (1955), although they regard the capillaries as constitutingonly the first threshold of the barrier, a second threshold being formed by themembranes of the cells adjacent to the blood vessels. In respect of this concept itis interesting that in the optic nerve the processes of the fibrous astrocytes form alayer between the nervous elements and endothelial cells, both inside and sur-rounding the nerve, and that cell contacts exist between the astrocyte processes inboth sites. It is also between the glial elements forming the end-feet on capillaries

that Gray (1961) has described plasma membrane contacts in the cerebral cortex.However, Maynard et al, (1957) find that the astrocytic sheath around capillariesin the cerebral cortex is not complete and estimate that astrocyte processes occupyonly 85 % of the surface. Clearly, although the function of plasma membranecontacts may be to prevent the passage of substances between cells, further in-vestigations are necessary to establish whether contacts always occur in sites wherebarriers to the passage, or free transport, of substances across cell layers exist.

SUMMARY

1. An electron microscope study has been made of the plasma membranecontacts occurring in the optic nerves of rats, mice and toads.

2. The most simple form of contact is found within the mesaxon and between theplasma membranes of either adjacent oligodendrocytes or adjacent endothelialcells, when a quintuple-layered unit results. A similar type of contact also occursbetween the outsides of adjacent myelin sheaths and the outside of a sheath andthe plasma membrane of either an oligodendrocyte or the cell body of an astrocyte.

3. Contacts in which one or both of the membranes are from a fibrous astrocytein a region of the cell where fibrils are present in the underlying cytoplasm have asomewhat different appearance to the above.

4. The dimensions and fine structure of the contacts are described and theirpossible significance is discussed.

I wish to thank Prof. G. J. Romanes for his continued interest during the courseof these studies, which were carried out on an electron microscope on loan from theWellcome Trust.

REFERENCES

BENNETT, H. S., LuFT, J. H. & HAMPTON, J. C. (1959). Morphological classification of vertebrateblood capillaries. Amer. J. Physiol. 196, 381-890.

DALTON, A. J. (1955). A chrome-osmium fixative for electron microscopy. Anat. Rec. 121, 281.DAVSON, H. (1958). Cellular aspects of the electrolytes and water in body fluids. Ciba Foundation

Colloq. on Ageing, 4, 15-32.FARQUHAR, M. G. & HARTMANN, J. F. (1957). Neuroglial structure and relationships as revealed

by electron microscopy. J. Neuropath. 16, 18-39.GAZE, R. M. & PETERS, A. (1961). The development, structure and composition of the optic nerve

of Xenopus laevis (Daudin). Quart. J. exp. Physiol. 46, 299-309.GLAUERT, A. M. & GLAuERT, R. H. (1958). Araldite as an embedding medium for electron micro-

scopy. J. biophys. biochem. Cytol. 4, 191-194.GRAY, E. G. (1961). Ultrastructure of synapses of the cerebral cortex and of certain specializations

of neuroglial membranes. In Electron Microscopy in Anatomy, pp. 54-73. Ed. J. D. Boydet al. London: Edward Arnold.

HAMA, K. (1960). The fine structure of some blood vessels of the earthworm, Eisenia foetida.J. biophys. biochem. Cytol. 7, 717-724.

KARRER, H. E. (1960a). Cell interconnections in normal human cervical epithelium. J. biophys.biochem. Cytol. 7, 181-184.

KARRER, H. E. (1960b). The striated musculature of blood vessels. II. Cell interconnections andcell surface. J. biophys. biochem. Cytol. 8, 135-150.

LAWN, A. M. (1960). The use of potassium permanganate as an electron-dense stain for sectionsof tissue embedded in epoxy-resin. J. biophys. biochem. Cytol. 7, 197-198.

MATURANA, H. (1960). The fine structure of the optic nerve of anurans. An electron microscopestudy. J. biophys. biochem. Cytol. 7, 107-120.

246 A. Peters

Plasma membrane contacts in the central nervous system 247MAYNARD, E. A., SCHULTZ, R. L. & PEASE, D. C. (1957). Electron miscroscopy of the vascular bed

of rat cerebral cortex. Amer. J. Anat. 100, 409-434.MERCER, E. H. & SCHAFFER, B. M. (1960). Electron miscroscopy of solitary and aggregated slime

mould cells. J. biophys. biochem. Cytol. 7, 353-356.MUIR, A. R. (1961). Personal communication.PETERS, A. (1960a). The structure of myelin sheaths in the central nervous system of Xenopus

laevis (Daudin). J. biophys. biochem. Cytol. 7, 121-126.PETERS, A. (1960b). The formation and structure of myelin sheaths in the central nervous system.

J. biophys. biochem. Cytol. 8, 431-446.PETERS, A. (1961). Myelinogenesis in the central nervous system. European Regional Conference

on Electron Microscopy. Delft 1960. Nederlandse Vereniging Voor Electronenmicroscopie.PORTER, K. R. (1959). The biology of myelin. Other membrane-limited structures of cells. In

The Biology of Myelin, pp. 37-58. Ed. S. R. Korey. Paul Hoeber: New York.ROBERTSON, J. D. (1957). New observations on the ultrastructure of the membranes of frog

peripheral nerves. J. biophys. biochem. Cytol-3, 1043-1048.ROBERTSON, J. D. (1959). The ultrastructure of cell membranes and their derivatives. Biochemical

Society Symposium, no. 16, pp. 1-43.ROBERTSON, J. D. (1961). The unit membrane. In Electron Microscopy in Anatomy, pp. 74-99.

Ed. J. D. Boyd et al. London: Edward Arnold.ROSENBLUTH, J. & PALAY, S. L. (1961). The fine structure of nerve cell bodies and their myelin

sheaths in the eighth nerve ganglion of the goldfish. J. biophys. biochem. Cytol. 9, 853-878.SCHULTZ, R. L., MAYNARD, E. A. & PEASE, D. C. (1957). Electron microscopy of neurones and

neuroglia of cerebral cortex and corpus callosum. Amer. J. Anat. 100, 369-408.VAN BREEMEN, V. L. & CLEMENTE, C. D. (1955). Silver deposition in the central nervous system

and the haemotocephalic barrier studied with the electron microscope. J. biophys. biochem.Cytol. 1, 161-166.

WATSON, M. L. (1958). Staining of tissue sections for electron microscopy with heavy metals.J. biophys. biochem. Cytol. 4, 475-478.

YAMADA, E. (1955). The fine structure of the gall bladder epithelium of the mouse. J. biophys.biochem. Cytol. 1, 445-458.

EXPLANATION OF PLATESAll micrographs are of mouse optic nerve.

Key to LetteringANu Astrocyte nucleus I Intraperiod line or intermediate layerAs Astrocyte cell body L Cytoplasmic dark layer of plasma mem-AsP Astrocyte process braneAx Axon M MesaxonB Basement membrane 01 OligodendrocyteC Region of contact between adjacent ONu Oligodendrocyte nucleus

plasma membranes P Plasma membraneCap Capillary Q Quintuple-layered unitD Major dense line R Radial component of sheathEn Endothelial cell T Tongue process of sheathER Endoplasmic reticulum

PLATE 1

Fig. 1. Transverse section of one complete myelin sheath and parts of three others. Within theinternal mesaxon (M) the outer surfaces of two regions of the plasma membrane of themyelin-forming cell come together to form a quintuple-layered unit, the intermediate layer ofwhich is the intraperiod line (I). Throughout the sheath, the intraperiod line alternates with themajor dense line (D) which is formed as the cytoplasmic dark layers (L) of the plasma mem-brane come together. The major dense line (D) terminates on the outside of the sheath as theplasma membranes separate to enclose a tongue of cytoplasm (T1, T2 and T3), while theintraperiod line terminates (X) where the membrane surrounding the tongue process turnsoff from the outside of the sheath. In two regions (Z) cytoplasm is present between thelamellae. The triple-layered structure of the plasma membranes is indicated (P). Note thecontacts between adjacent sheaths and the radial structures (R) present within the sheath.For further explanation see text. x 180,000.

248 A. Peters

PLATE 2

Fig. 2. Contact between the cell body of a fibrous astrocyte (As) and the outside of a myelinsheath. In this region, the cytoplasm of the astrocyte, lying between the nucleus (ANu) andthe plasma membrane (P), is devoid of fibrils. As the plasma membrane comes into contactwith the outside of the sheath, an intermediate line (I) is formed, while contact of the plasmamembrane of the astrocyte (P) and that of the tongue process (T) of the sheath gives rise to aquintuple-layered unit (Q). x 130,000.

Fig. 3. Contact between the plasma membrane (P) of an oligodendrocyte (01) and the outside ofa myelin sheath. At the site of contact an intermediate line (I) is present. x 160,000.

PLATE 3

Fig. 4. Contact between the plasma membrane of the cell body of an astrocyte (As) in the regionwhere the cytoplasm contains endoplasmic reticulum (ER), and the plasma membrane of anastrocyte process (AsP) containing transversely sectioned fibrils. At the site of contact of theseplasma membranes a quintuple-layered unit (Q) is present. A contact (C) has also beenformed between the plasma membrane (P) of the astrocyte cell body (As) and the outside ofa myelin sheath. Note the radially arranged thickenings (R) of the intraperiod line of thesheath in the region adjacent to the tongue process (T). x 100,000.

Fig. 5. Contact between the plasma membranes (P1 and P.) of adjacent oligodendrocyte cellbodies (011 and 01,). On the right the plasma membranes are separated, but towards the leftthey come into contact (C) to form a quintuple-layered unit (Q). Part of the nucleus (ONu) ofone of the oligodendrocytes (012) is visible. x 100,000.

PLATE 4

Fig. 6. Contact between the plasma membranes (P1 and P.) of adjacent fibrous astrocyte processes(AsPL and AsP,). Contact of the plasma membranes results in the formation of a quintuple-layered unit (Q). Note the absence of contact between the plasma membrane of the otherside of the process, AsPs, and that of the astrocyte cell body (As) which contains a nucleus(ANu). x 110,000.

Fig. 7. A contact formed between the plasma membrane of an astrocyte process (AsP), in whichthe cytoplasm contains fibrils, and the plasma membrane of a tongue process (T) on theoutside of a sheath. At the site of contact, the quintuple-layered unit (Q) has the samecharacteristics as that produced between the adjacent astrocyte plasma membrane in Fig. 6.x 120,000.

PLATE 5

Fig. 8. Transverse section of a capillary (Cap) within the optic nerve. This capillary, containinga red blood cell, is completely surrounded by astrocyte processes (AsP) which form end feet,intervening between the endothelial cells (En) and the myelinated nerve fibres. x 20,000.

Fig. 9. Enlarged portion of Fig. 8. The astrocyte process (AsP) contains transversely sectionedfibrils and is separated from the endothelial cells (En1 and En.) by a basement membrane (B),which at the point of widest separation (Z) appears to split into two layers. An extensivearea of contact (C) is present between the plasma membranes of the two endothelial cells(En1 and En,). The cytoplasm on both sides of this contact shows an increased densitysimilar to that which occurs at terminal bars. x 86,000.

PLATE 6

Fig. 10. Part of a capillary (Cap) outside the nerve showing three contacts (C) between adjacentendothelial cells. The walls of such capillaries show no fenestrations and are surrounded by acomplete basement membrane (B). The capillary is separated from the astrocyte processes(AsP) forming the outer layer of the optic nerve by collagen of the pia mater. Note theincreased density of the cytoplasm of both sides of the endothelial cell contacts (C). x 46,000.

Fig. 11. The site of contact between two endothelial cells (En1 and En,). The adjacent plasmamembranes (P1 and Ps), which can be seen to have a triple layered structure, come intocontact to form a quintuple-layered unit (Q). x 160,000.

Journal of Anatomy, Vol. 96, Part 2

U.

Y &.h.

-' - -

p

.Ax

s, r

L

@- a ! In As AdF

A *, ;se

,:';.**:for

'.-S ....

..

PETERS PI.ASMtA MtEMIIBRANE. CONTACTS IN TIlE CENTRAL NERVOUS SYSTEM(Facing p. 248)

Plate 1

3i : #

C'

.11*vd.

'ef

V

I,,f b

..,

Journal of Anatomy, Vol. 96, Part 2 Plate 2

I

,p

II As

ANU'

V4VA :4; ': .,.

:?".. .. s~

Iie

.V '.

,T ..

I: -. ..416

:- I:...I ...

t.

0i

-7.'I ts.

1~I.

3.

3

PETERS PLASNIA MEMBRNIANE CONTACTS IN THE CENTRAL NERVOUS SYSTEMI

.' W.

..'A.

ML

Plate 3

4 ..'Fo

.., .1 ...

tU

*R* X _ ,

' g_ AS~

va' 1 s

-2

'a .

v t ,#R~~~~~~~~-A

QL

PETERS- PI.s--SrA MEMB1H12SRANE CONTACTS IN THIE CENTRAL NERVOUS SYSTEM

Journal of Anatomy, Vol. 96, Part 2

k. 4 {".g, -j. ;az.t, -.

.IV R-

A.

I."1. ..

4 .

.:r..- I.

AT

.~~~~~~lp

_:

-.pg

It

n

.?.l

1:

.6,

4 ,

~.V

.,r

!t

'.4f;; '

?il-N.0

Ps '.o.l N

i

If

WITT

$i.i

,.I

Journal of Anatomy, Vol. 96, Part 2

4i.G..'A

^ ^ x} i^r .1: A; .,

.fi .^;sxs f t ---W ^

tsfSX

*.+ -ffX # 3

o/wf #s- 4

4,.y A]

1.

'r

effW ,

i .

/eX.

As'..?h

PETERS-PLASNIA EMBRANE CONTACTS IN THE CENTRAL NERVOUS SYSTEM

4

Plate 4

i,~Qr,I

*.I

1.

v'k-4.

-W 'j,

v .1IF lk lw 4

IP11

7 l,

Journal of Anatomy, Vol. 96, Part 2

4.. if

.j

rr11sIffI..i:

AsP

8__vtIn^ If

_ .,.

% I :/-.X

Fel

r

1N.,

*. t.Ik--.. I'Df.. .~! s?&

.AP~.a^ S b .e2''kb~~~~~~~~~r~~jFt65I

jW- .

PETERS-PLAS-MA 'MEMBRANE CONTACTS IN THE CENTRAL NERVOUS SYSTEM

Plate 6

1.%jt::

-1

Journal of Anatomy, Vol. 96, Part 2

~~~~~~~~~~~~7 _ .................... Ar g

4 ~~ ~ ~ ~ ~ ~ ~ 4

~~*s 1'" #5IS ;:t$; 4 iAxts D= e.e$Y8@ , ! v, ,; E~i5.se ,,

b ,4rv > - B ¢D~~~~~~~~~w

I .Ww k....

'.EN2 I... '. ..o,,"-.^ - V

i

2.e : EN1.

LUMEN

114'.:-. ,-.

1PETERS-PLASMIA MENIBRANIE CONTACTS IN TlE. CE.,NTRAL NERVOUS SYSTEIM

Plefte 6)