Clonal analysis of chimaeric patterns in aortic endothelium

A carbohydrate-based polymorphism recognized by the lectin Dolichos biflorus agglutinin (DBA) provides a histochemical marker of chimaerism between inbred mouse strains (Ponder, Festing & Wilkinson, 1985a). The polymorphism is confined to intestinal epithelium and vascular endothelium. We have used the DBA marker system for studying clonal development of crypts of Lieberkühn, and for visualizing cell migration pathways in the intestinal epithelium of aggregation chimaeras (Schmidt, Wilkinson & Ponder, 1984, 1985b,c; Ponder et al. 1985b; Schmidt, Garbutt, Wilkinson & Ponder, 1985a). Here, we report the results of the application of the DBA marker to the quantitative clonal analysis of aortic endothelium.

The aorta is formed in situ from ‘blood islands’ -clusters of cells developing in the splanchnic mesoderm. Cells in the peripheral zone of blood islands flatten to become the endothelium, a special type of epithelium, while cells in the central zone become primitive blood corpuscles (Jenkinson, 1925). Adjacent blood islands coalesce to form tubes. During subsequent development the paired dorsal aortas, which run along both sides of the notochord, fuse to form a single descending aorta and eventually acquire outer walls of connective tissue. The endothelium, separated from the surrounding outer layers by basal lamina, lines the aorta as a monolayer of exceedingly thin cells which are greatly elongated in the long axis (Thorgeirsson & Robertson, 1978).

In the adult, endothelial cells retain a capacity for cell division and movement: spreading, migration and subsequent proliferation of endothelial cells occurs at the sites of mechanical injuries or lesions (Hauenschild, 1980). Cell turnover in the normal adult endothelium is relatively slow: on average, 0 ·3 –1 ·5 cells in a hundred are replaced per day in the rat aorta. Marked regional heterogeneities in [³H]thymidine uptake exist however, and it was suggested that heavily labelled regions might represent growth centres, comparable to intestinal crypts, from which outward migration of cells occurs (Schwartz & Benditt, 1973). Our chimaeric system would reveal any highly ordered arrangement of proliferative compartments and of cell migration pathways such as occur in the intestinal epithelium (Ponder et al. 1985b; Schmidt et al. 1985b).

The primary objective of the present study was to define a clonal pattern characteristic of the aortic endothelium, and thereby to gain insight into the clonal history of the development of the endothelium using the quantitative methods of clonal analyses developed by us (Schmidt et al. 1985a, c). Two types of clones are distinguished in the analysis. A ‘coherent clone’ is a group of cells which are descended from a common progenitor cell through previous divisions and which have stayed together as a coherent group. A ‘descendent clone’, by contrast, is a group of cells related by descent from a common progenitor, but which may have separated by cell mixing and are therefore no longer contiguous. In chimaeras with balanced proportions of components, individual clones may be obscured by aggregation into larger patches, but in chimaeras with unbalanced proportions, each patch may be regarded as a single coherent clone (West, 1975; Whitten, 1978), and clusters of patches may be recognized, which can be interpreted as single descendent clones (Schmidt et al. 1985c).

Three DDK ⟷C57BL/6JLac (B6) and one RIII-ro (RIII) ⟷B6 mouse aggregation chimaera were constructed at the MRC Laboratories, Carshalton, according to methods described by Mintz (1971a). The ages of the mice and the relative proportions of the parental components (for method see below) are given in Table 1. In order to minimize the effect of patch aggregation (see Schmidt et al. 1985a,c for discussion), we restricted our analysis to mice with highly unbalanced proportions. The aortas chosen were the most unbalanced out of 25 chimaeric preparations examined.

Animals were killed by ether overdose. Large segments of thoracic aortas (cf. Table 6) were dissected from the chimaeras, transferred to wax-based Petri dishes containing PBS, cleared of adventitious loose connective tissue and fat, and cut open dorsally (using dissecting scissors, Prof. Kinmoth’s Scissors, Macarthy’s Surgical Ltd, Dagenham, Essex, and forceps, MicroSurgery, Serr. 4″, Macarthy’s Surgical Ltd). Any remaining blood was removed by gentle pipetting. The preparations were pinned out and slightly stretched to approximately the in situ length. They were then fixed and stored in 10% formol saline.

Aortic endothelia were stained with Dolichos biflorus agglutinin (DBA)-peroxidase conjugate. The conjugates were prepared by the periodate method (for details see Ponder & Wilkinson, 1983).

Following fixation the preparations were incubated for 30 min in 0 ·1 % phenylhydrazine HC1 in PBS pH7 ·3 to block endogenous peroxidase, washed in PBS, transferred to PBS containing 0 ·5% bovine serum albumin (PBS-BSA) (30min), and incubated overnight in DBA-per-oxidase conjugate 1:75 in PBS-BSA. The peroxidase was subsequently demonstrated using 3′3′ diaminobenzidine (DAB) as a substrate yielding a brown reaction product (10 mg DBA (Sigma) + 20ml 0 ·1M-Tris buffer, pH7 ·3 + 40 μl 30% H₂O₂). Following incubation for 5 min, preparations were washed in PBS. To ensure that the endothelial layers were undamaged, cell nuclei were stained with Mayer’s haemalum (3min), after preparations had been scored and photographed (see below).

DBA serves as a strain-specific marker because endothelial cells of DDK or RIII but not B6 origin express DBA-binding sites (Ponder & Wilkinson, 1983; Ponder et al. 1985a).

Numbers of DDK (chimaeras 45,46) or RIII (chimaera 49) cells per patch were counted (patch size). Chimaera 47 was not analysed as the DBA-negative (i.e. unstained) cells of the B6 minority component could not be counted following our protocol. Enumeration of the numbers of DBA-positive cells was possible as areas containing cell nuclei appeared lighter than the surrounding stained cytoplasm. Scoring was carried out before counterstaining with Mayer’s haemalum (see above) which unavoidably also stained the nuclei in the relatively thick intimal layers underlying the endothelium.

Using the method of maximum likelihood, geometric distributions were fitted to the observed patch-size-frequency data and tested for goodness of fit by means of χ²-tests (Ross, 1980).

The percentage contribution of the minority component (Table 1) was calculated by the line interception method (Aherne & Dunhill, 1982). The entire preparation was scanned under ×80 magnification (10/0 ·25 objectives; Leitz Laborlux 12, Wetzlar) and the number of interceptions of a 10 ×10 eyepiece graticule overlying cells of the minority component was scored. This number, divided by the total number of points sampled (400 –500 for each aorta), ×100, yielded an estimate of the percentage of the minority component.,

The method, which is an exploratory statistical technique (Greig-Smith, 1983), has been described in detail elsewhere (Schmidt et al. 1985c). The principle of the analysis is to examine the distribution of numbers of objects (in the present case, mosaic patches) scored by means of a grid. If the objects are randomly distributed the numbers of objects per square óf the grid will conform to a Poisson distribution (see below); if they are clustered or regularly spaced, the distribution will depart from this expectation. By repeating the analysis for larger blocks obtained by combining adjacent grid squares, the occurrence of pattern at different scales can be sought (cf. Fig. 5).

The mathematical analysis is based on the analysis of variance (ANOVA). This analysis is performed for each block size. The ANOVA establishes the variability in the numbers of patches found between the sample blocks at a particular block size. In the Greig-Smith method the ANOVA does not, however, involve computation of the components of variance in the usual sense (cf. Bailey, 1978); instead, mean square/mean ratios (Index of Distribution [IOD]) are calculated separately for each block size as described in Schmidt et al. (1985c; Table 1). A set of IODS obtained from replicate tissue samples should conform to a χ²-distribution if patches are randomly (Poisson) distributed. IOD values lower than expected indicate regular spacing, while higher IODs indicate clustering. A distinction between x- and y-directions can be made, allowing one to test for differences of pattern in the longitudinal and transverse directions.

Evidence of pattern is sought as follows: IOD ratios of successive block sizes are expected to fluctuate around 1, the expected range depending on the degrees of freedom at the particular block size (Table 4). If IODs for one block size are consistently larger than the IODs for the immediately smaller block size, the IOD ratios will be consistently larger than 1 and, in addition, some values may be significantly high. In that case, clustering or, if an overall non-random distribution has already been established, an increase in the intensity of pattern may be inferred at about the order of the larger block size. While the initial search for consistent increases of IODs in the data involves a subjective assessment (Table 2), the validity of that interpretation is subsequently tested using the F-statistic (Table 4).

The positions of discrete patches of the minority component on photomontages of the aortas were mapped by recording the coordinates of patch centres (judged visually) with the aid of a digitizing tablet (MOP Videoplan, Kontron). The thoracic aortas were divided into proximal and distal halves, except for aorta 46 for which only a short region of midthoracic aorta was obtained. A ‘form factor’ (cell length/width = 3 ·32) was used to correct for the longitudinal shapes of cells. This was determined using a MOP Videoplan, on 50 randomly chosen endothelial cells (random number tables, Lindley & Scott, 1984: cells were identified by nuclear staining with Meyer’s haemalum, and outlines of cell bodies were visualized as dark-brown lines by treatment with 0 ·2 % silver nitrate and subsequent exposure to light (Pool, Sanders & Florey, 1958)). A 16 ×16 grid was used in the mathematical analysis, which yielded a mean number of approximately one patch per quadrat (Fig. 5; Schmidt et al. 1985c for further explanation).

Aortas were photographed on a Leitz Orthoplan microscope with a plate camera attachment using Ilford FP4, 4″ ×5″, 125 ASA film.

Examples of the general appearance of patches are given in Figs 1 and 2. Patches were often conspicuously longitudinally arranged, i.e. in the direction of the blood flow, but the longitudinal shape of endothelial cells (cf. Materials and Methods) will have exaggerated the shape of patches. Indeed, when the elongated cell shape was corrected for in the Greig-Smith analysis (see Materials and Methods) no consistent difference between the x- and y-directions over all seven samples were found, which implies that the longitudinal appearance of patches may be a visual deception. IOD values were averaged (Table 2) when splits in the x- and y-directions were performed (at block sizes 2, 8 and 128 -cf. Schmidt et al. 1985c).

The fragmentary nature of the patches was reflected in the patch-size-frequency distributions which revealed a preponderance of singletons. The overall concave shapes of the distributions fitted a geometric model (Fig. 3).

The analysis revealed an overall non-random distribution of patches as 53 out of 56 IOD values were greater than 1 (Table 2). This conclusion was validated by comparing the IODS with the χ²-distribution for each of the eight scales examined (Table 3).

In order to detect an increase of clustering at a particular scale it was necessary to identify consistencies of increases of IODs. Subjective assessment of the values given in Table 2 suggested two such increases: one at block size 2, and a further one at block size 32. An objective assessment of the strength of the evidence for our conclusion was provided by using the F-statistic. The ratios of successive IODs were determined and compared with the F-distribution by examining complete sets of replicate samples (similar to the χ²-procedure, cf. Table 3). The results (Table 4) indicate that a definite increase in the strength of pattern occurred at block size 2, as all F-values at the 2:1 level of comparison were above 1 and, moreover, three of them were significant. A further increase is probable at block size 32, as six out of seven IOD ratios at 32:8 were above 1; one of them was significant. There were no consistent trends at other block sizes as successive IOD ratios were fairly evenly spread above and below 1. The lack of further change in pattern justifies combining block sizes 2 to 16, and 32 to 128 for a summarizing graphical representation (Fig. 4). The F-test of the summarized data (Table 5) presents strong evidence for a strengthening in the pattern at block sizes 2 and 32.

Having demonstrated a non-random arrangement of patches (Tables 2, 3) and having detected two scales at which clustering of patches occurred, we proceeded to identify the pattern in the preparations of aortic endothelium. An example of this procedure is given in Fig. 5. It is reasonable to suggest that the patches which clustered at block size 2 were related by origin from a single source, representing ‘descendent clones’ of DDK cells; further clustering at the larger block size 32 was thus indicative of a further level of ‘descendent clones’. On the basis of this hypothesis we were able to calculate the total numbers of large-scale descendent clones for each preparation of thoracic aortic endothelium (Table 6). An interpretation of the different scales of clustering in terms of clonal development is given below.

The two levels of clonal pattern are difficult to interpret on the basis of well-defined growth centres and outward directions of cell migration in the adult tissue such as might be inferred from the findings of Schwartz & Benditt (1973) (see Introduction). Cell turnover in normal adult aortic endothelium is low and it seems probable that the patterns which we observe reflect overwhelmingly the effects of cell proliferation and mingling during the growth and development of the tissue.

Large-scale (primary) descendent clones included several small-scale (secondary) descendent clones which would be expected if the former were older in developmental terms than the latter. It is therefore reasonable to argue that the small-scale spatial pattern arose late in development but that the large-scale pattern arose during a relatively longer period, and reflects events which took place during early tissue growth. Fragmentation of primary descendent clones into spatially identifiable groups of discrete coherent clones may have resulted from the continued incursion of cells of the other component into primary descendent clones as a result of cell division during tissue growth. There is evidence that in the adult in at least some circumstances (e.g. wounding), endothelial cells can migrate. Cell migration, if relatively restricted, may have contributed to the fragmentation of secondary descendent clones. This interpretation implies that (1) growth is not primarily coherent and (2) cell mingling is not a uniform process throughout development. Furthermore, the finding of non-randomness at all scales suggest that (3) mixing is never complete, even in the early embryo. This conclusion is consistent with observations that some coherent growth occurs already at the pre-implantation stage (Gardner, 1984a), although there may be exceptions for a few tissues (McLaren, 1976; Mullen, 1977: Gardner, 1984b).

Using block size 32 as a guide, we were able to define the approximate outlines or territories, and hence also the total numbers, of primary descendent clones. We assume that each descendent clone occupies a territory of roughly similar size: they would not otherwise have been detected by our analysis. (Note that this does not imply that each must also comprise a similar number of cells: Fig. 1 shows that in this respect descendent clones may be highly variable, as we would have predicted from our previous studies of coherent clone sizes in chimaeric tissues (Schmidt et al. 1985c).) Our data (Table 6) indicate that there are of the order of 5 –20 descendent clones per mm² in the endothelium of the thoracic aorta. This figure reflects the number of patches present at the end of a phase of extensive (but not necessarily complete) mingling which must be presumed to have occurred early in the development of the endothelium. It does not tell us about the number of cells originally allocated, some or all of whose descendents would contribute to the endothelium.

A different spatial pattern with only one scale of descendent clones was found in the clonal analysis of intestinal epithelium (Schmidt et al. 1985c), which suggests that this exploratory statistical method of analysis is able to define a clonal pattern characteristic for a particular tissue. It may therefore provide a quantitative basis for comparison between mosaic patterns found at different developmental stages, and may also be of value in assessing the results of experimental manipulation of tissue development, and possibly the validity of computer models of tissue growth. The analysis was carried out for tissues in which the spatial position and numbers of coherent clones could be confidently inferred from that of observed patches in unbalanced, two-dimensional preparations of epithelia. However, the method should still be able to define a tissue-specific pattern even if discrete patches (and therefore also numbers of descendent clones) cannot be defined, as would be the case if the components of the chimaeras were more nearly balanced. In that case, percent cover per quadrat rather than discrete patches would have to be scored (Greig-Smith, 1983; see also Schmidt et al. 1985c).

Although the establishment of patch-size-frequency distributions and subsequent comparison of the distributions with theoretical models (here, geometric) have allowed us to test the validity of previous statistical methods for estimation of clone size in mosaic tissues (Schmidt et al. 1985a, 1986), such distributions may arise in a variety of ways (e.g. Ross, 1980) and are therefore of limited value in identifying the mechanisms of pattern formation (Schmidt et al. 1986). Here, good fit to geometric (which is in contrast to our previous results for intestinal epithelium) would be consistent with a random spacing of patch-progenitor cells (cf. Schmidt et al. 1985a), but the mathematical analysis of the spatial pattern clearly demonstrated an overall non-random arrangement of patches. The satisfactory description of the data by a geometric model may therefore be due to small scale, local fragmentation of larger patches (see above). The relatively fine grain of mosaic pattern due to overall small-patch sizes, and the fragmentary appearance of patches, implies that for the most part clones do not show coherent growth. Similar patchiness has now been reported by us for tissues derived from endoderm (intestinal epithelium), neuroectoderm (retinal pigment epithelium) and mesoderm (endothelium), which strongly suggests that this mode of tissue growth applies to mammalian epithelia in general.

Lewis, Summerbell & Wolpert (1972) questioned a claim made by Mintz (1971b) that the pattern of mosaic patches identified in chimaeric tissues may be of general developmental significance. We support Mintz’s view and have shown that the use of appropriate mathematical methods in the analysis of mosaic patches in chimaeric epithelia may provide a general method to gain insight into clonal histories during tissue development. Our results suggest that the analysis of patch sizes is likely to be less informative than the analysis of the spatial arrangement of patches.

We thank Professor R. Mead for valuable discussions and advice with the statistical analysis and Mr D. J. Garbutt for carrying out the computer programming. This study was carried out during the tenure by B.A.J.P. of a Career Development Award from the Cancer Research Campaign, and was supported by a programme grant from the Medical Research Council and the Cancer Research Campaign.