Two-dimensional sorting-out behaviour (segregation) in mixtures of pulmonary endothelial cell lines derived from congenic strains of mice was examined using dense confluent cultures in which mitosis is rare. Cell MHC type was detected by autoradiographic labelling or by immunofluorescence techniques. For autoradiography cultures of one type were previously labelled with [3H] thymidine and one component of the mixed cell types was prepared from these cultures. Autoradiographs were prepared from the fixed cultures. Counts of contiguous neighbours (labelled or unlabelled around a randomly chosen central labelled cell) were made: these were analysed statistically using a new model for such a system. The results show that segregation of a clustering type took place if the alleles of the D locus in the H-2 complex were mismatched in the mixed strains, but that matching at the IC locus (or some locus to the right of IJ) overrides the effect of D mismatch. After 2 days culture sorting-out was easily detectable when the cells were stained for their histocompatibility antigens and groups of up to 12 cells of the same type were associated together.

A central question in developmental biology concerns the identity of the molecules that provide positional information in pattern formation. In 1976 Edelman proposed that H-2 antigens, which can vary epigenetically, may be involved in specifying position during development. Evidence in support of this theory was obtained by Ostrand-Rosenberg et al. (1977) from in vitro studies using cell lines from 4-day-old mouse blastocysts. Only partial expression of H-2-specificities could be detected on these cells, raising the possibility that incomplete expression or masking of portions of the H-2 molecules could be occurring, which might affect cell positioning during ontogenesis. Changes in adhesiveness do in fact occur during development (Bellairs et al. 1978). In adult tissues, evidence has been found for H-2 restriction of adhesion in vitro. Fibroblasts mismatched at portions of the H-2 complex show reduced adhesiveness as compared to syngeneic combinations (Bartlett & Edidin, 1978). Bone marrow and lymphoid cells also show H-2-restricted adhesion (Zeleny et al. 1978). Contact inhibition of movement in kidney epithelial outgrowths appears to be enhanced in H-2 mismatched combinations (Curtis & Rooney, 1979). In lymphocyte-endothelial interactions mismatching within portions of H-2 leads to increased adhesiveness (Curtis & Renshaw, 1982), a result that is perhaps not surprising in view of the immunological role of these cells. In 1981, Curtis & Davies obtained evidence that suggested an in vivo effect of H-2 antigens. When H-2 molecules derived from T cells were injected into syngeneic mice a dramatic alteration was seen in adhesion and positioning of lymphocytes (Davies & Curtis, 1981; Curtis & Davies, 1981). This work provided the first experimental evidence that implicated MHC molecules in the control of cell adhesion and positioning in vivo.

We have reinvestigated this phenomenon using murine lung endothelial cells in vitro. Cells from congenic mouse strains were labelled with [3H]thymidine then mixed with unlabelled cells from the same or a different strain. Mixtures were allowed to adhere to glass surfaces for 24 h or 48 h and then fixed in formol saline. The final arrangement of cells of each histocompatibility type was visualized after 24 h of culture by autoradiographic processing and statistical analysis, and after 48 h of culture by immunofluorescent staining for histocompatibility antigens.

We report for the first time that cells of the same histotype will sort-out in cell culture according to MHC type. This strongly supports earlier ideas on the role of the MHC in cellular interactions (Bodmer, 1972).

Preparation of lung endothelial cultures

Congenic strains of mice were obtained from OLAC Animal Suppliers (Bicester, UK). The haplotypes of the strain used are as follows: B10.BR (kkkkkkkkk); B10.AKM (kkkkkkkkq); B10.A(3R) (bbbbkdddd); B10.D2 (ddddddddd). Sterile techniques were maintained throughout. Lungs from two mice were cut into 1 mm3 fragments in Hanks’-Hepes Saline (HHS; Cambridge Research Biochemicals, Cambridge, UK) and washed twice in calcium- and magnesium-free Tris-saline (CMF). The pieces were trypsinized (0 · 25% trypsin (Difco) diluted 1 in 5 in Versene, 300 BAEE units ml − 1) for approximately 1h then washed twice in Ham’s F10 saline (Flow Laboratories, Irvine, UK) + 10% heat-inactivated foetal calf serum (Gibco-Bio-cult Ltd, Paisley, UK) with insulin (Sigma, Poole, UK), transferrin (Sigma) and sodium selenite at the levels described by Barnes & Sato (1908). This is termed ITS-serum-F10. The tissue fragments were resuspended in this medium and cultured for 24 h at 37°C. Unattached cells and tissue fragments were carefully removed with a Pasteur pipette and 10 ml of fresh lTS-serum-F10 medium added.

After a few days confluent cultures were obtained. These were washed twice with CMF, then 3ml trypsin—Versene was added. The trypsin used was at 300 BAEE units ml − 1. The cells that detached after 2 min were collected separately and fresh trypsin was added. After 2 min endothelial-like cells started to detach and by 7 min most of them were in suspension; these were collected and grown to confluence, when the trypsinization process was repeated. After three such transfers cells of a predominantly endothelial morphology could be collected. Further categorization of the endothelial-like fraction included transmission electron microscopy (Mr C. Mucci, Department of Cell Biology’, University of Glasgow; Dr J. Anderson and Dr P. Toner, E.M. Unit, Glasgow Royal Infirmary) showed that the internal organization of the cells was typical of endothelial-like cells. Intermediate filament proteins from these cells were prepared using the method described by Brown et al. (1976) and Franke et al. (1979) and studied using SDS-PAGE techniques as described by Laemmh (1970) (Dr J. Edwards and Mr A. Hart, Department of Cell Biology, University of Glasgow). The lack of cytokeratins in the cells allowed us to discount an epithelial origin. The presence of factor VIII antigen was confirmed by standard immunofluorescence techniques using rabbit anti-human factor VIII (Nordic Immunologicals, Maidenhead, UK) and sheep antirabbit immunoglobulin (1g) (Nordic Immunologicals) conjugated to FITC. Angiotensin-converting enzyme was identified in the culture using spectroscopy’ according to the method of Cushman & Cheung (1971).

Preparation of endothelial cells for mixed cultures

Confluent cultures of endothelia were washed twice with HHS medium then trypsinized with trypsin—Versene for 7 min. The cells were collected using a Pasteur pipette and the trypsin neutralized with medium containing 3% foetal calf serum. The cells were then washed twice with HHS and resuspended with a Pasteur pipette. The suspension was vortexed to ensure a single cell suspension, divided into two and recentrifuged. One of these cell preparations was resuspended in ITS-serum-F10. The other was resuspended in the same medium containing 5 μ Ci ml−1 of [3H]thymidine (TRA-61; Radiochemical Centre, Amersham, UK) at a concentration of 5× 106 per 10ml. Both cultures were incubated at 37°C for 24 h in 25 ml tissue culture flasks and then the labelled cultures were given a ‘cold chase’ by resuspending the cells in three changes of F10 at 37°C over a 3-h period. Cells from both cultures were removed with trypsin-EDTA and washed three times with the ITS-serum-F10. Finally, the cells were resuspended in the same medium at a concentration of l×105 viable cells per ml using a Pasteur pipette and vortexing for 1 min. Care was taken to ensure that no clumps of cells were added to the cultures, by examination of the cell suspensions under the phase-contrast microscope before use and re-examination of the cell mixtures after preparation. Cells of different MHC type were mixed together in approximately 50:50 proportions and were added to prepared clean glass coverslips. Cultures were kept at 37°C for 24 h to allow cells to settle and migrate on the glass surface.

Preparation of cultures for autoradiography

After 24 h the cultures were fixed in formol saline and dipped in a solution containing 5g gelatin and 0 · 5 g chrome alum dissolved in 11 of distilled water. Further autoradiographic processing was earned out by Rogers’ (1969) method. Coverslips coated with G5 emulsion (Ilford Nuclear Research, Mobberley, Cheshire, UK) were left in the dark for 21 – 28 days and then developed with Ilford Phenisol, washed in water and fixed in 30% Amfix (May & Baker) for 5 min. After staining in Giemsa stain the coverslips were mounted on slides and examined.

Counting of sections and statistical analysis of results

All coverslip cultures were prepared in duplicate and were read ‘blind’ by two observers. Nuclei labelled with [3H]thymidine appeared black and granular. Counts were made of the number of labelled and unlabelled contiguous neighbours around at least 150 labelled or unlabelled cells chosen by semi-random selection of the centre cell. If the population has six nearest neighbours, the expected number of contiguous neighbours of each type for a population containing the proportion (p) of labelled cells is given by:

(1 – p)6 for 0 labelled neighbours

6(1 – p)5 for 1 labelled neighbour

15(1 – p)4p2 for 2 labelled neighbours and so on by binomial proportions to: p6 for 6 labelled neighbours.

The counts of contiguous neighbours, around both labelled and unlabelled nuclei provide the best estimate of the proportion of labelled cells in the population since at least 1800 cells plus the centre cell have been scored. If there is no sorting-out the counts of labelled cells around labelled centre cells and of labelled cells around unlabelled centre cells should be almost identical. Indeed, large discrepancies from this expectation would be reasons for suspecting non-random distributions.

On the null hypothesis that there is no sorting-out, the expected number of labelled nuclei is calculated from the proportion of labelled cells in the population and the equations given above. If the alternative estimate of the proportion of unlabelled cells is used a very minor error will be introduced, since the number of labelled nuclei around a labelled centre nucleus should be one seventh smaller (for 6 contiguous neighbours) than the number of labelled cells around an unlabelled centre cell.

Observation showed that the average number of contiguous neighbours was between 5 · 5 and 5 · 8 for different cultures. Thus the six-contiguous neighbour model was used. Contiguous neighbour scores were tested for a significant degree of departure from random expectation using chisquared tests. The null hypothesis is that if no preferred adhesion existed the number of labelled or unlabelled contiguous neighbours would be given by the equation above.

Despite the fact that equal numbers of labelled and unlabelled cells were added to the cultures, and that the percentage labelling of the labelled cells was, in all cases, greater than 95%, in some combinations the final amount of labelled cells was much less than 50% in the mixed cultures. This may be due to allogeneic effects on adhesion (see Curtis & Rooney, 1979). The percentage of labelled cells present was determined for each combination and the expected values of contiguous neighbours calculated from that. Significant departure of the distributions from random was accepted if P< 0 · 01.

Detection of histocompatibility antigens on cells

A monoclonal antibody method including the application of fluorescently labelled protein A was used. Cultures to be examined for the display of histocompatibility antigens were washed twice with phosphate-buffered saline (PBS) cooled to 4°C and kept at this temperature until examination. After a further wash with PBS after 10 min the cultures were then incubated with 25 μ l per cm2 of culture surface, of monoclonal antibody 11.4.1 (anti-k at K) or alloantibody C57.B1/10 anti-Bio.D2 (Searle Laboratories, High Wycombe, UK) for 2h. The monoclonal antibodies were obtained from the supernatant of a cell line of the same title kindly supplied by Dr M. Edidin, Johns Hopkins University. After 2h of incubation the cultures were washed thrice with PBS. After a further 2 h of incubation with either Texas-Red-labelled sheep anti-mouse IgG or protein A – Texas Red conjugate (both from Molecular Probes Inc, suppliers Amersham International, Amersham, UK; batch numbers 14 and 00, respectively, at 25 μ l per cm2 of culture surface) the cultures were washed with PBS.

The cultures were examined by epilkimination with 546 nm light, the fluorescence being visualized in the red region of the spectrum. Since the intensity of fluorescence was low it was viewed through an ISIT video camera system. This consisted of a Falcon video camera type LTC 1162 and control unit type CCU 1595 made by Custom Camera Designs (Wells, Somerset, UK). The pictures were displayed on a monitor and photographed with a 35 mm camera.

We took great care to ensure that the cell suspensions were composed of single cells. The cells were seeded out at a density close to confluence so that the whole culture would be occupied with cells and therefore cell division would be minimized. This was found to be so (see Table 1). The cultures were grown for up to 24h during which time cell division is fairly inappreciable under these high-density conditions (see Table 1). However, the main evidence that neither of these errors has been made in the experiments comes from examination of the control cultures. There is no evidence of a departure of the distribution of the labelled and unlabelled nuclei from the normal when these are of the same MHC type (Tables 2, 3).

Table 1.

Percentage mitotic figures in cultures

Percentage mitotic figures in cultures
Percentage mitotic figures in cultures
Table 2.

Contiguous neighbour measurements: counts of contiguous neighbours in cultures at 24 h A. B10AKM* mixed with B10AKM. Centre cell labelled

Contiguous neighbour measurements: counts of contiguous neighbours in cultures at 24 h A. B10AKM* mixed with B10AKM. Centre cell labelled
Contiguous neighbour measurements: counts of contiguous neighbours in cultures at 24 h A. B10AKM* mixed with B10AKM. Centre cell labelled
Table 3.

Summary of contiguous neighbour measurements in cultures at 24 h

Summary of contiguous neighbour measurements in cultures at 24 h
Summary of contiguous neighbour measurements in cultures at 24 h

The results of contiguous neighbour analyses are shown in Tables 2 and 3. Table 2 presents two examples of actual analyses to illustrate the basic type of data obtained. For reasons of brevity such data for other combinations are presented in condensed form in Table 3. They show that some of the allogeneic combinations have nearest neighbour distributions that differ significantly from random expectations.

The lack of evidence for segregation in the control mixtures suggests that pseudo-sorting-out due either to mitosis or to incomplete dissociation of the cells is unlikely to have happened. A further source of possible error is that the population proportions have been measured incorrectly. Since at most 5% of the cells that are supposed to be labelled may in fact be unlabelled, we also calculated chi-squared values for the contiguous neighbour analyses on the assumption that there were 2 · 5% more labelled cells present than measured. These results (not shown) did not alter the pattern of significant and non-significant results. Thus, where the chi-squared value given in Table 3 is significant (6d.f. at the 1% level), we conclude that sorting-out has taken place.

Pseudo-sorting-out might arise from cell division or from differential plating out efficiencies. The first possibility is obvious and has been examined experimentally (see above). The second possibility would arise from the fact that, if there was a limited amount of substrate space available for cell attachment, the type of cell that adheres more slowly might find insufficient substrate available for complete attachment. This might lead to an emphasis of local non-randomness. Attachment rates were examined (see Table 4) and no difference was found for the two types separately, though some mutual inhibition may occur when they are mixed.

Table 4.

Attachment rates of different strain types in mixtures and unmixed cultures

Attachment rates of different strain types in mixtures and unmixed cultures
Attachment rates of different strain types in mixtures and unmixed cultures

Segregation after 2 and 3 days of culture visualized by immunofluorescence

While the nearest neighbour analysis system is ideally suited to detection of segregation at a level where it is not immediately clear by subjective examination, a simpler method could be applied to the situation after 2 days of culture. By this time segregation became very clear (see Figs 1 – 4) and was detected by the clustering together of groups of from four to twelve cells of one histocompatibility type. Very few isolated cells of one type were found surrounded by cells of the other type. Since the histocompatibility type was detected by immunofluorescent labelling, and since no cross-reaction of the monoclonal antibodies took place, we can be reasonably certain that cell types are correctly identified. Incidentally, these findings demonstrate that class I histocompatibility antigens are still being displayed by the cells at this time in culture.

Figs 1 – 4.

Mixed culture of B10.D2 and B10.AKM endothelia cultured for 3 days showing segregation of strain types by MHC type. Figs 1, 2. Cultures labelled with 11.4.1 monoclonal antibody against k haplotype at K locus (positive for B10.AKM) visualized by reaction with protein A – Texas Red conjugate and illuminated with green light, the red fluorescence being detected with an ISIT camera. The video display was photographed with an ASA 400 film on an Olympus OM-2 35 mm camera. The same areas visualized by bright-held microscopy at very low light level are shown in Figs 3 and 4, matching Figs 1 and 2, respectively. Note the segregated areas of B10.AKM (fluorescent) and B10.D2 (non-fluorescent). Bar, 50 μ m (Figs 1 – 4).

Figs 1 – 4.

Mixed culture of B10.D2 and B10.AKM endothelia cultured for 3 days showing segregation of strain types by MHC type. Figs 1, 2. Cultures labelled with 11.4.1 monoclonal antibody against k haplotype at K locus (positive for B10.AKM) visualized by reaction with protein A – Texas Red conjugate and illuminated with green light, the red fluorescence being detected with an ISIT camera. The video display was photographed with an ASA 400 film on an Olympus OM-2 35 mm camera. The same areas visualized by bright-held microscopy at very low light level are shown in Figs 3 and 4, matching Figs 1 and 2, respectively. Note the segregated areas of B10.AKM (fluorescent) and B10.D2 (non-fluorescent). Bar, 50 μ m (Figs 1 – 4).

Segregation of cells according to MHC type

The results show clearly that a number of the combinations of cells of differing MHC type show segregation into areas predominantly of one or other type. The sorting-out is not large-scale, or clear after 24 h. By 48 h extensive segregation has taken place. The type of segregation seen at 24 h can be classed as ‘clustering’ segregation, rather than ‘total’ (see Smith, 1982, for discussion).

We report that the segregation of cells of like tissue but different histocompatibility type can take place in culture. If the two cell strains used are mismatched at D and IC loci the cultures are non-random after 24 h of culture and very clearly so by 48 h. The cells in control cultures and in mixtures mismatched at other loci are still random in arrangement after this time period. The sorting-out is of the two-dimensional type reported for mixed tissue types by Garrod & Steinberg (1973).

All the control mixtures, i.e. a labelled cell strain with unlabelled cells of the same strain type, showed no sorting-out. This result demonstrates that pseudo-segregation has not arisen either from division of cells or from incomplete dissociation of the cells. Those combinations in which the haplotypes of all loci are mismatched, i.e. B10.D2 with B10.BR or B10.D2 with B10.AKM, show sorting-out. The combination B10.AKM with B10.BR shows sorting-out, illustrating the importance of a single locus mismatch at D in cell interactions (see Curtis & Rooney, 1978; Curtis & Renshaw, 1982; Curtis, 1978, 1982, for adhesive reactions in which D mismatch plays an important role). The combination B110.A(3R) is matched with B10.D2 at the right-hand end of the MHC and this includes loci IE, G, S and D. This combination showed no sortingout, which is consistent with the interpretation that the D mismatch is one of the requirements for sorting out. The combination BIO.A(3R) with B10.BR is matched only at locus IC and yet does not sort-out. Though doubts can be held about the identity of the IC locus in genetic terms (see Murphy, 1981; Klein, 1982; Steinmetz, 1985), it appears from these results and also from those reported by Curtis & Rooney (unpublished data) that IC matching is sufficient to prevent sorting-out or to minimize cellular abreactions. The strain types of endothelial lines available to us to date do not permit exploration of the possibility that other loci may also be of importance in these reactions.

In the context of these findings, the experiments reported by McClay & Gooding (1978) are open to another interpretation. They examined the collection of cells to preformed aggregates of the same and different MHC types. On the basis of their results they concluded that the MHC may not be involved in sorting-out. It should be noted that in their experiment sorting-out did not, and indeed could not, occur, since one type was collected on the outside of an aggregate, i.e. they were collected in a sorted-out pattern. Their experiment was not a test of sorting-out, even though the aggregates collected cells of different MHC type as well as they collected cells of the same type.

It has been proposed that differences in adhesion (see Steinberg, 1978; Garrod & Steinberg, 1973; and discussed by Curtis, 1978), lie at the basis of cell segregation phenomena. Though this seems likely no rigorous proof of it has yet been obtained.

Other groups have reported that MHC mismatching on cells in vitro leads to a reduction in adhesive interactions (Bartlett & Edidin, 1978; Curtis & Rooney, 1979). The exception to these findings appears to be the case where immunological interactions favour increased binding of allogeneic combinations (Curtis & Renshaw, 1982). It is interesting to note that the same loci were reported to be important in control of adhesive interactions, namely D and IC, which we have found to be important in segregation. Circumstantial evidence in favour of the idea that adhesion is being altered by histoincompatible reactions in these experiments comes from the facts that the percentage of cell types found in mixtures can be appreciably different from the 50:50 mixtures used, and from the fact that short-term plating-out efficiency is reduced in allogeneic mixtures (see Tables 2, 4). These results have led to some speculation on the possible significance of MHC involvement in cell adhesions in embryogenesis. Although most adult tissues express class 1 products, the evidence suggests that these are only partially expressed during ontogenesis (Ostrand-Rosenberg et al. 1977). Edelman (1976) proposed that H-2 products could be involved in specifying position during development. One way in which this idea can be envisaged is by the control of expression of the gene product. Adult tissues show a variable adhesivity when expressing a complete set of MHC products, which could indicate that the products are not themselves responsible for the adhesive interactions but instead might control tissue-specific adhesive mechanisms. Thus the incomplete or masked MHC products found in embryonic cells could indicate that the controlling mechanism is being regulated at the level of the gene or at the cell surface, which could result in differences in adhesivity between embryonic cell compartments. According to this hypothesis, cell migration and positioning would be under MHC control. Evidence in support of this hypothesis has already been obtained in vivo, where adult lymphocyte migration and positioning are at least partially controlled by MHC product release.

Monoclonal antibodies to MHC products may have limited use in screening for incomplete, partial or masked products on embryonic cells. The lack of the specific epitope could give rise to a false interpretation of a negative result. Initially, it will be important to check whether embryonic cells express MHC molecules on their surfaces, perhaps at low levels or possibly even as fragments in solution. Matters for further investigation include: (1) whether further sorting-out would occur over longer time periods; and (2) whether differences in the level of MHC antigen surface density might occur between the strain types causing quantitative differences in adhesion (e.g. Harris & Gill, 1986; Williams et al. 1980). However, this seems unlikely because the significant differences in segregation or non-segregation follow MHC match/mismatch rules rather than any simple hierarchy.

Finally, it should be noted that the two-dimensional sorting-out seen in this work apparently is a close parallel to events in vivo, since endothelia grow two-dimensionally in the body.

We thank the Scottish Hospitals Endowment Research Trust for a grant and Professor Robert Goudie for inspiration.

Barnes
,
D.
&
Sato
,
G.
(
1980
).
Methods for growth of cultured cells in serum-free medium
.
Analyt. Biochem
.
102
,
255
270
.
Bartlett
,
P. F.
&
Edidin
,
M.
(
1978
).
Effect of the H-2 gene complex on rates of fibroblast intercellular adhesion
.
J. Cell Biol
.
77
,
377
388
.
Bellairs
,
R.
,
Curtis
,
A. S. G.
&
Sanders
,
E. J.
(
1978
).
Cell adhesiveness and embryonic differentiation
.
J. Embryol. exp. Morph
.
46
,
207
213
.
Bodmer
,
W. F.
(
1972
).
Evolutionary significance of the HL-A system
.
Nature, Bond
.
237
,
139
145
.
Brown
,
S.
,
Levinson
,
W.
&
Spudich
,
J. A.
(
1976
).
Cytoskeletal elements of chick embryo fibroblasts revealed by detergent extraction
.
J. supramolec. Struct
.
5
,
119
130
.
Curtis
,
A. S. G.
(
1978
).
Cell-cell recognition: positioning and patterning systems
. In
Cell-Cell Recognition
(ed.
A. S. G.
Curtis
), pp.
51
82
.
Cambridge University Press
.
Curtis
,
A. S. G.
(
1982
).
The genetic basis of cell behaviour in lymphocytes and other cells
. In
Cell Behaviour
(ed.
R.
Bellairs
,
A. S. G.
Curtis
&
G.
Dunn
), pp.
373
393
.
Cambridge University Press
.
Curtis
,
A. S. G.
&
Davies
,
M. D. J.
(
1981
).
H-2D antigens released by thymocytes and cell adhesion
.
J. Immunogenet
.
8
,
367
377
.
Curtis
,
A. S. G.
&
Renshaw
,
R.
(
1982
).
Lymphocyte-endothelial interactions and histocompatibility restriction
. In
In Vivo Immunology
(ed.
P.
Nieuwenhuis
,
A. A.
Van den Broek
&
M. G.
Hanna
), pp.
193
198
.
New York
:
Plenum
.
Curtis
,
A. S. G.
&
R∞ney
,
P.
(
1979
).
H-2 restriction of contact inhibition of epithelial cells
.
Nature, Land
.
281
,
222
223
.
Cushman
,
D. W.
&
Cheung
,
H. S.
(
1971
).
Spectrophotometric assay and properties of the angiotensin-converting enzyme of rabbit lung
.
Biochem. Pharmac
.
20
,
1637
1648
.
Davies
,
M. D. J.
&
Curtis
,
A. S. G.
(
1981
).
The effect of a soluble cell product released by live thymocytes on lymphocyte movements in vivo
.
Thymus
3
,
35
42
.
Edelman
,
G. M.
(
1976
).
Surface modulation in cell recognition and cell growth
.
Science
192
,
218
226
.
Franke
,
W. W.
,
Schmid
,
E.
,
Weber
,
K.
&
Osborn
,
M.
(
1979
).
HeLa cells contain intermediate-sized filaments of the prekeratin type
.
Expl Cell Res
.
118
,
95
109
.
Garrod
,
D. R.
&
Steinberg
,
M. S.
(
1973
).
Tissuespecific sorting-out in relation to contact inhibition of cell movement
.
Nature, Land
.
244
,
568
569
.
Harris
,
H. W.
&
Gill
,
T. J.
III
(
1986
).
Expression of class 1 transplantation antigens
.
Transplantation
42
,
109
117
.
Klein
,
J.
(
1982
).
Evolution and function of the major histocompatibility complex
. In
Histocompatibility Antigens: Structure and Function, Receptors and Recognition, series B
(ed.
P.
Parham
&
J.
Strominger
), vol.
14
, pp.
223
239
.
London
:
Chapman & Hall
.
Laemmli
,
U. K.
(
1970
).
Cleavage of structural proteins during the assembly of the head of bacteriophage T4
.
Nature, Land
.
227
,
680
684
.
McClay
,
D. R.
&
Gooding
,
L. R.
(
1978
).
Involvement of histocompatibility antigens in embryonic cell recognition events
.
Nature, Land
.
274
,
367
368
.
Murphy
,
D. B.
(
1981
).
Genetic fine structure of the H-2 gene complex
. In
The Role of the Major Histocompatibility Complex in Immunobiology
(ed.
M. E.
Dorf
), pp.
1
32
.
New York
:
Wiley
.
Ostrand-Rosenberg
,
S.
,
Hammerberg
,
C.
,
Edidin
,
M.
&
Sherman
,
M. I.
(
1977
).
Expression of histocompatibility-2 antigens on cultured cell lines derived from mouse blastocysts
.
Immunogenetics
4
,
127
136
.
Rogers
,
A. W.
(
1969
).
Techniques of Autoradiography
.
London
:
Elsevier
.
Smith
,
J. V.
(
1982
).
Geometrical and Structural Crystallography
.
New York
:
Wiley
.
Steinberg
,
M. S.
(
1978
).
Specific ligands and the differential adhesion hypothesis; How do they fit together?
In
Specificity of Embryological Interactions, Receptors and Recognition, Series B, no. 4
(ed.
D.
Garrod
), pp.
99
129
.
London
:
Chapman and Hall
.
Steinmetz
,
M.
(
1985
).
Organization of the genes of the H-2 complex
. In
Cell Biology of the Major Histocompatibility Complex
(ed.
B.
Pernis
&
H. J.
Vogel
).
Orlando, NY
:
Academic Press
.
Williams
,
K. A.
,
Hart
,
D. N. J.
,
Fabre
,
J. W.
&
Morris
,
P. J.
(
1980
).
Redistribution and quantitation of HLA-ABC and DR (la) antigens on human kidney and other tissues
.
Transplantation
29
,
274
279
.
Zeleny
,
V.
,
Matousek
,
V.
&
Lengerova
,
A.
(
1978
).
Intracellular adhesiveness of H-2 identical and H-2 disparate cells
.
J. Immunogenet
.
5
,
41
47
.