Differential adhesive affinity and aggregation of wild-type and E-cadherin null embryonic stem cells
Following up our previous studies (Moore et al., 2009, 2014a), we used mouse embryonic stem (ES) to study cell sorting patterns in aggregates/embryoid bodies. Three ES cell lines, RW4 wild-type (WT), CFG37 GFP-labeled wild-type, and 9j E-cadherin null (E-cadherin (-/-)) cells were used in cell aggregation and sorting experiments. CFG37 ES cells were isolated from blastocysts from transgenic mice expressing GFP-histone H2B driven by the beta-actin promoter (Okabe et al., 1997; Moore et al., 2009, 2014a), and 99% of the cell population was GFP-positive with largely uniform signals in individual cells. Western blot analyses of the cells in standard adherent culture indicated that the 9j ES cells lacked E-cadherin protein, and had slightly elevated N-cadherin, possibly as a result of compensatory expression for the loss of E-cadherin, though N-Cadherin levels were increased in both WT and E-cadherin null cells following differentiation induced by retinoic acid (Fig. 1A). The cultured ES cells remained undifferentiated as indicated by the expression of Oct3/4 and were differentiated following treatment with retinoic acid as indicated by Dab2 induction and Oct3/4 reduction. In comparison, the E-cadherin (-/-) cells aggregated at a lower rate than those of wild-type, as observed under a microscope to observe the clustering of the cells (Fig. 1B). The lower adhesive affinity of the E-cadherin-deficient cells was also demonstrated by using a Coulter counter to measure the progressively declining numbers of particles as the cells clustered (Fig. 1C).
In suspension cultures, the cells formed aggregates but they still showed similar characteristics of cadherin and marker expression as determined by Western blot (Fig. 1D). The aggregates consisting entirely of CFG37 cells showed a uniform GFP signal throughout the whole spheres, while we were able to distinguish GFP-positive and negative cells in spheroids mixed with GFP-labeled or unlabeled cells (Fig. 1E).
Interaction and assembling of embryonic stem cells to form aggregates in suspension culture
In cell sorting experiments, two or more different cell types were first dispersed into single cells, intermixed at a 1:1 ratio, and then placed on non-adhesive plastic dishes to allow cell aggregation. We mixed the CFG37 GFP-labeled WT with unlabeled ES cells, either RW4 WT or the E-cadherin deficient 9j lines, and observed their association and cell sorting within the spheroids formed.
By time-lapse video microscopy, WT-GFP + WT and WT-GFP + E-cad (-/-) cells exhibited rather different characteristics in clustering, sorting, and assembling into spheroids. For the WT-GFP + WT unlabeled cell intermix, suspended individual cells rapidly clustered and associated into spheres by 8 to 12 hours, with GFP-positive cells intermixed. The majority (> 95%) of aggregates contained both GFP-positive and negative cells, though the ratio appeared somewhat variable. Initially, the spheroids formed by collecting single cells around, and subsequently enlarged presumably by cell doubling. Collisions and fusion of neighboring spheroids to form larger aggregates also frequently occurred (Supplemental movie 1).
The heterotypic, WT-GFP + E-cadherin (-/-) aggregates developed in a distinctive manner as observed in all cases. Typically, following initial cell congregation, a small cluster of GFP-positive cells formed, surrounded by a cloud of unlabeled, E-cadherin (-/-) cells. The loosely gathered E-cadherin (-/-) cells appeared to follow the movement of the GFP-positive core. Cohesive aggregates of GFP-positive and -negative spheroids developed slightly slower than the mixture of all wild-type cells, by an approximate 12-hour lag time. In the heterotypic (E-cadherin positive and negative) aggregates that formed, the cells moved dynamically against each other in the spheroids, though GFP-positive cells appeared segregated from the start (Supplemental movie 2). When two spheroids collided and combined, the GFP-positive central cores appeared to fuse, segregated from the GFP-negative, presumably E-cadherin-deficient cells in the periphery (Supplemental movie 3).
Two main diverse sorting patterns in aggregates of undifferentiated embryonic stem cells with high and low adhesive affinity
As we have previously reported (Moore et al., 2009, 2014a), when pluripotent ES cells of wild-type and less adhesive E-cadherin knockout were mixed to form aggregates, the less adhesive E-cadherin (-/-) cells sorted to the outer layer, enveloping the highly adhesive wild-type ES cells. We attributed the cell sorting pattern to the differential adhesive affinity hypothesis proposed by Steinberg (Steinberg, 1962; 1963; Steinberg and Gilbert, 2004). However in the previous studies, we also found that differentiated ES cells sorted to the outer layer to form a polarized endoderm epithelial layer, and concluded that the ability of the differentiated cells to establish apical polarity overcomes differential adhesive affinity to ultimately be positioned peripherally (Moore et al., 2009, 2014a).
In further reiterating the cell mixing and sorting experiments, however, we now found unexpected cell sorting patterns that diverged from the previously established conclusion (Moore et al., 2009; 2014a). In some cases, the wild-type ES cells were found at the outer layer with the less adhesive E-cadherin null ES cells positioned in the interior (Fig. 2A, lower panel), in addition to the typical patterns (Fig. 2A, upper panel) reported previously. Here, immunostaining of E-cadherin was used to identify E-cadherin-positive and negative cells. For the mixture of wild-type and E-cadherin null cells (WT(GFP) + E-cad (-/-)), two representative contradictory examples are present: one showed that a shell of E-cadherin-positive cells enveloped E-cadherin negative cells (presumably 9j); the other showed a pattern in which E-cadherin-positive cells were centrally located surrounded by E-cadherin negative cells (Fig. 2A). The aggregates were produced by mixing undifferentiated wild-type and E-cadherin null ES cells and cultured for 2-4 days. In such a time frame, a negligible number of the ES cells underwent differentiation, which commonly initiates after day 4-5 of aggregation, as we have previously documented (Capo-chichi et al., 2005; Rula et al., 2007).
To clarify the surprising observations, we further investigated the cell sorting patterns by mixing GFP-labeled cells with unlabeled cells to form aggregates and by performing live cell imaging and histology analyses. We compared immunostaining with the endogenous GFP signal of the labeled cells, and found that both E-cadherin immunostaining and GFP signal were equivalent and could distinguish E-cadherin wild-type and null cells (Fig. 2B). In these aggregates that were thought to be generated with a similar procedure, several sorting patterns were observed and documented (Fig. 2B). The mixing of WT and WT-GFP cells produced a largely random intercalated pattern, but mixing of WT-GFP and E-cadherin deficient ES cells generally resulted in a segregated configuration (Fig. 2B). As shown in 3 representative examples, the E-cadherin and GFP-positive cells sorted either to the center or periphery in individual aggregates. Moreover, in the third example, both surface and internally localized E-cadherin and GFP-positive cells were also present simultaneously in the same spheroids (Fig. 2B, right panel). From observations in 8 independent experiments, each of the three sorting patterns shown (Fig. 2B) could be found in the range from 10% to 70% among all the aggregates, indicating high inter-experiment variation in the resulted sorting patterns. We now realized that the variability of the sorting patterns was caused by the dynamic transition of the cell aggregates at the moment when the experiments were conducted and completed, and a slight difference in cell aggregation time can produce a large variation in cell sorting result.
Initial sorting and subsequent maturation of aggregates of embryonic stem cells with differential adhesive affinity
Following multiple repetitions of cell sorting experiments with intermixing of WT and E-cadherin null ES cells, we concluded that the highly adhesive WT ES cells unequivocally sorted initially to the interior of the cell aggregates as reported previously (Moore et al., 2009; 2014a); however, upon subsequent maturation of the aggregates the E-cadherin-positive cells then localized to the surface. In a standardized protocol followed in the lab with precise cell density and mixing speed, we consistently observed that at an earlier time course (12 hours) when cell aggregates were relatively small, the highly adhesive GFP-positive WT cells clustered in the center and were surrounded by unlabeled E-cadherin-deficient ES cells (Fig. 3A). By 24 hours, both central and peripheral sorting patterns for GFP-positive cells were present (Fig. 3A). Finally, after 48 hours in culture, the majority of GFP-positive cells localized as a shell enveloping the GFP-negative, presumably the E-cadherin-deficient, ES cells (Fig. 3A). Representative examples of optically sectioned 24-hour (Fig. 3B) and 48-hour (Fig. 3C) individual spheroids were analyzed, comparing heterotypic intermixing of WT-GFP plus WT controls with WT-GFP plus E-cadherin-deficient ES cells (Supplemental movie 4-7). The confocal sectioning of the aggregates provided visualization of the 3-dimensional distribution of GFP-positive cells within the spheroids (Supplemental movie 4-7). The cell aggregates at 12 and 24-hour time points appeared to harbor a rough surface, and the spheroid became progressively larger and rounder, with a smoother edge by 48 hours (Fig. 3).
The relative location and distribution in the aggregates of GFP-labeled cells was quantitated by an image analytical approached we designed (Fig. 3D). GFP signals within individual aggregates were determined in equal areas of outer ring or inner circle, defined as the region of interest (ROI) (Fig. 3D). The results indicate that the GFP-positive cells relocated to the outer layer by 48 hours in the aggregates composed of WT-GFP and E-cadherin-deficient ES cells (Fig. 3E), though about equal DAPI signals, indication of cell number, were assessed (Fig. 3F). However, this analytical method did not show a distinct, central distribution of the WT-GFP in the 24-hour aggregates mixing with the E-cadherin-deficient ES cells, as the percentage of the GFP signals measured was not significantly lower in the outer ring (Fig. 3E). Although we did observe that the WT-GFP cells were more self-aggregated/associated in the mixtures with the E-cadherin (-/-) cells than with the unlabeled WT cells (Fig. 3B, C), indicating segregation of the two cell types with differential adhesive affinity. We reasoned that this was due to the fact that the E-cadherin and GFP-positive cells located both peripherally and interiorly, but superficially, however the cells were not necessarily at the central area of the spheres. Additionally, the internal to peripheral transition of the GFP-positive high adhesive wildtype ES cells likely initiated in some of the aggregates. We were unable to use this quantitative approach to satisfactorily analyze cell sorting pattern at 12-hour time point because the aggregates were not spherical.
Nevertheless, these observations indicate that the maturation of cell aggregates correlated with the reversion of the initial cell sorting pattern of the mixtures of cells with differential adhesive affinity, the WT-GFP and E-cadherin (-/-) ES cells.
Rapid transition of the sorting patterns
To investigate the transition of cell positioning patterns, we used time-lapse imaging to visualize the reversion of the cell sorting (Fig. 4A). GFP-labeled WT and unlabeled E-cadherin knockout ES cells were intermixed and allowed to coalesce in suspension to form aggregates for 24 hours. The aggregates were analyzed for progressive changes using an enclosed, temperature-regulated epifluorescence microscope system with imaging at 20-minute intervals for additional 24 to 48 hours.
During the early time course, the predicted differential adhesive affinity pattern of heterotypic aggregates was observed with the GFP-expressing, highly adhesive WT cells in the core of the aggregate surrounded by the unlabeled, less adhesive, and peripheral E-cadherin knockout cells (Fig. 4A). Initially, both E-cadherin-positive and -negative cells actively moved against each other, though the segregation of GFP-positive and -negative cells was preserved. At around the 40-hour time point, the GFP-positive cell cluster extended to the outer layer. Subsequently, a layer of the GFP-positive cells formed a partial surface on the spheroid, and the superficially positioned GFP-positive cells appeared to contact and bring additional associated GFP-positive cells to extend the surface shell (Supplemental Movie 8). After reaching the surface, the GFP-positive cells appeared to slow their motion, and the layer of GFP-positive cells was maintained and persisted for at least 8 hours in the recording (Fig. 4A).
As a control, no sorting patterns or bleaching of GFP signals were observed in cell aggregates from mixing of the WT-GFP with WT ES cells (Figure 4B) (Supplemental Movie 9). Thus, the observations using time-lapse imaging indicate that the transition of sorting patterns occurs rapidly, and the surface positioning of the highly adhesive cells is stable.
Formation of polarized apical actin caps on the surface of mature aggregates
The formation of polarity is a possible mechanism for the surface positioning of the cells, as in the case of primitive endoderm positioning on the surface. We first examined the mature (48 hours) aggregates from the mixture of WT-GFP and E-cadherin (-/-) cells for the distribution of the classical polarity markers, ZO-1, aPKC, and Ezrin (Fig. 5A). In these aggregates, the wildtype cells located to the surface, as indicated by immunostaining of E-cadherin. However, no obvious diverged distribution of the classical tight junction associated polarity markers, ZO-1, aPKC, and Ezrin, was observed (Fig. 5A). As positive controls similar to that we reported previously, the polarized distribution of ZO-1 and aPKC was observed in the ES cell aggregates at a later time course, when surface extraembryonic endoderm developed or cavitation to form ectoderm initiated (Meng et al., 2017). Thus, the re-distribution of E-cadherin expressing cells to the outer layer is not associated with formation of the classical tight junction dependent apical polarity of the surface cells.
Initially we observed that cell aggregates of the two contradictory cell sorting configurations exhibited very different F-actin staining patterns (Fig. 2A), and we suspected that the highly adhesive ES cells formed a polarized epithelium to be able to position on the surface. Thus, we further examined the distribution of F-actin in cell aggregates (Fig. 5B, C, D). In mature (48 hours) spheroids derived from wildtype ES cells, the surface was covered with a layer of strong actin staining that consisted of multiple contiguous surface cells, suggesting the formation of a surface epithelium and consolidated apical actin organization (Fig. 5B, arrow). In contrast, cellular actin staining was uniformly distributed around surface cells of E-cadherin (-/-) ES cell aggregates that had not yet had sufficient time to fully develop and compact in the 48-hour incubation time (Fig. 5C, arrowhead). In spheroids composed of mixed WT and E-cadherin deficient cells, an F-actin cap was observed on the surface that was composed of E-cadherin-positive cells (Fig. 5D, arrow), but not on the surface where E-cadherin-deficient cells localized (Fig. 5D, arrowhead), as shown in two examples. Thus, we conclude that the highly adhesive E-cadherin wildtype cells on the surface formed a polarized epithelium (as indicated by the distribution of beta-actin), which may account for the ability of the WT cells to sort to the surface and to envelop the less adhesive E-cadherin null cells.
Polarization of a surface epithelium prior to differentiation in mature ES cell aggregates
Previously, we determined that differentiated ES cells in the aggregations containing undifferentiated cells were able to overcome the force of differential adhesive affinity to position on the surface (Moore et al., 2009). The ability for the differentiated endoderm cells to position on surface was attributed to their propensity to establish an apical polarity facilitated by the Dab2-dependent endocytic trafficking (Moore et al., 2009; 2013).
However, we reasoned that the presently observed cell sorting property of the highly adhesive ES cells to the surface was independent of endoderm differentiation, because extensive differentiation occurs only after 4 or more days of ES cell aggregation (Capo-chichi et al., 2005; Rula et al., 2007). To verify, we designed experiments to determine the relationship between the formation of a spheroid surface actin cap and endoderm differentiation.
In the aggregation of wildtype ES cells, similar to that reported previously, no endoderm differentiation occurred within 24 hours, as indicated by staining for the endoderm marker Dab2 (Yang et al., 2002; 2007) (Fig. 6A). The cells located on the surface exhibited nearly uniform and diffused actin staining around the cell boundary (Fig. 6A, arrowhead). Few Dab2-positive cells were visible even within 48-hour aggregates. In nearly all these 48-hour, mature spheroids, an actin cap had formed on the surface (Fig. 6A, arrow). In rare spheroids in which a surface endoderm epithelium had formed, no actin cap was observed (Fig. 6A, arrowhead). The actin showed a dispersed staining pattern in the differentiated, Dab2-positive endoderm cells on the surface. Thus, the nature of apical polarity of the endoderm cells is distinctive from that of the undifferentiated surface cells signified with of an actin cap.
For aggregates composed of intermixing of wild-type and E-cadherin-deficient ES cells, the actin of the surface cells was also not polarized in 24 hours (Fig. 6B, arrowhead). Most of the 48-hour aggregates showed a partial actin cap though they contained no differentiated cells (Fig. 6B, arrowhead). Based on previous results (Fig. 5D), these cells possessing an actin cap are likely E-cadherin-positive rather than -deficient. In rare spheroids containing Dab2-positive cells either in the interior or on the surface, a partial actin cap and polarized cells were visible on the surface where Dab2 staining was absent (Fig. 6B, arrow). Consistently, the Dab2-positive endoderm epithelial cells positioned on the surface showed a diffuse beta-actin staining pattern (Fig. 6B, arrowhead).
Based on these observations, we conclude that a polarized epithelium forms on the surface of a mature spheroid without undergoing endoderm differentiation. We speculate that this polarity, signified by an actin cap on the apical surface of the adhesive E-cadherin-positive cells, accounts for the ability of the wild-type ES cells to sort to the surface, and to envelop the less adhesive E-cadherin-deficient cells. Thus, a subtle polarity formed by strong adhesion of ES cells assembling an epithelial layer may be able to override differential adhesive affinity for the cells to position to the surface rather than to be engulfed in the interior.