Marker succession during the development of keratinocytes from cultured human embryonic stem cells

Abstract

Human embryonic stem cells injected into scid mice produce nodules containing differentiated somatic tissues. From the trypsinized cells of such a nodule, we have recovered keratinocytes that can be grown in cell culture. The method of recovery is sensitive enough to detect small numbers of keratinocytes formed in the nodule, but for purposes of analysis, it is preferable to study the development of the entire keratinocyte lineage in culture. The principle of our analysis is the successive appearance of markers, including transcription factors with considerable specificity for the keratinocyte (p63 and basonuclin) and differentiation markers characteristic of its final state (keratin 14 and involucrin). We have determined the order of marker succession during the time- and migration-dependent development of keratinocytes from single embryoid bodies in cell culture. Of the markers we have examined, p63 was the earliest to appear in the keratinocyte lineage. The successive accumulation of later markers provides increasing certainty of emergence of the definitive keratinocyte.

Human embryos can for obvious reasons not be used experimentally to study early development. Embryos of other mammalian species, especially the mouse, have been used a great deal and much has been learned from studying them. But the increasing structural complexity of the embryo after implantation imposes great difficulties for the study of the consecutive changes that lead from stem cells to definitive somatic cell types.

Embryonic stem (ES) cells of the mouse, first grown in cell culture by Evans and Kaufman (1) and Martin (2) give rise to many forms of differentiation (3). One such example, from the laboratory of F. Watt, is the formation of keratin-containing cells from murine ES cells (4, 5). Later, more advanced differentiation of keratinocytes was obtained (6), but no definitive keratinocytes were isolated in these studies even though keratinocytes of a murine teratoma had earlier been serially cultivated (7). One reason may be that, on serial cultivation, rodent somatic cells tend to convert quickly into established (“immortal”) cell lines with altered properties. This behavior is not a problem for human somatic cell types (such as fibroblasts and keratinocytes), for these are stable in culture and very rarely develop spontaneously into established cell lines.

Human ES cells were first cultivated by Thomson and coworkers (8) and, like murine ES cells (9), they differentiate into many cell types (8, 10–12). Methods of influencing or directing differentiation to certain cell types have been described for murine ES cells (13), for human ES cells (14, 15), for human embryonic germ cells (16), and for developing embryos (17).

When transplanted to scid mice, human ES cells have been shown to give rise to respiratory and gut epithelium, bone, cartilage, smooth and striated muscle, ganglia, renal structures, and stratified squamous epithelium with hair follicles (10, 11, 18). This finding demonstrates that keratinocytes are generated from human ES cells in the absence of embryonic implantation and the orderly sequence of fetal development. But the analysis of the process is more accessible to study in culture under conditions in which migrating and differentiating cells originating from ES cells have an essentially 2D (“monolayer”) structure. To study the development of the keratinocyte, we have attempted to standardize the starting conditions by beginning with an identifiable origin of the stem cells, a single embryoid body generated from suspended ES cells in culture (9, 14, 19). When the embryoid body is allowed to attach to a surface, cell migration from its perimeter is accompanied by differentiation along the keratinocyte lineage.

Transcription Factors Used as Markers

p63 is a transcription factor whose gene and transcripts were first fully described by F. McKeon and coworkers and whose expression they showed to be specific for keratinocytes and related epithelial cell types (20). Disruption of the gene results in failure of development of the epidermis, all other stratified squamous epithelia, and a few related epithelia such as mammary, sebaceous and lacrimal gland, prostatic, urothelial, and cervical (20–24). p63 is thought to be necessary for the maintenance of stem cell precursors (21, 25) and is present in all growing cells of keratinocyte colonies with high growth potential (holoclones, ref. 26). In the human, even heterozygous mutations in p63 produce developmental defects of ectodermal structures (27–30).

Basonuclin is a transcription factor containing three separated pairs of zinc fingers (31, 32). It is present in basal cells of the epidermis and other squamous epithelia (33). In rapidly growing cultured keratinocytes (33, 34) and in squamous tumors (35), it is concentrated in cell nuclei, but in the normal epidermis or in keratinocytes cultured under conditions not optimal for cell growth, it may be cytoplasmic (34). Nuclear localization of basonuclin may result from the absence of phosphorylation of Ser-541, located immediately C-terminal of the nuclear localization signal (36).

Differentiation Markers

K14 is a keratin of the basal cells of all stratified squamous epithelia (37–39).

Involucrin is a protein precursor of the cross-linked envelope that forms late in the terminal differentiation of the keratinocyte (40). In contrast to K14, involucrin is made only in suprabasal cells (41, 42).

Methods

The human ES cell line H9, which is used in all these experiments, was derived at the University of Wisconsin by D. A. Thomson and coworkers (8). For preparation of feeders, fibroblasts of 13-day mouse embryos (PMEF-H) treated with Mitomycin C were purchased from Specialty Media (Phillipsburg, NJ), and 3T3-J2 cells were as reported in refs. 43 and 44. Marker proteins detected by specific antibodies were as follows: Oct4 (Santa Cruz Biotechnology), p63 [with the 4A4 monoclonal antibody (20), provided by F. McKeon and A. Yang], basonuclin (36), involucrin (Biomedical Technologies, Stoughton, MA), and K14 (Chemicon). Staining of Western blots of keratinocyte extract with the K14 antiserum revealed a single strong band corresponding to the expected molecular weight of 50,000. Extracts of cultured H9 cells submitted to Western blot analysis and staining with a pan keratin antibody revealed no labeled bands with a molecular weight <81,000.

Culture medium for growing ES cells was as described (14), without lymphocyte-inhibitory factor. For experiments on detection and isolation of keratinocytes, including experiments on attached embryoid bodies, we used FAD medium (44, 45) with or without subsequently added irradiated 3T3-J2 supporting cells.

Results

Demonstration That the Culture System Supports the Multiplication of Keratinocytes Generated from Human ES Cells Injected into scid Mice. Two months after injection of 10⁷ H9 cells into scid mice, the resulting nodules were trypsin-disaggregated, and the cells were inoculated into dishes containing supporting irradiated 3T3 cells and FAD medium. Under phase microscopy, colonies with the morphology of keratinocytes were identified (Fig. 1A). Staining the same colony for p63, basonuclin, and K14 showed that all three markers were present (Fig. 1B).

Fig. 1.

Part of a keratinocyte colony formed with 3T3 support from cells of an ES cell-produced nodule in a scid mouse. A large nodule resulting from ES cells injected into the leg muscle of a scid mouse was excised, minced, and trypsin-disaggregated. Cells from the second trypsinization were plated on 3T3 feeders and fed with FAD medium. Eleven days later, a colony with morphology typical of keratinocytes was seen under phase microscopy (A). The colony was fixed and stained for p63 (red), basonuclin (blue), and K14 (green), and the resulting stains were merged (B). Nuclear p63, basonuclin, and cytoplasmic K14 are evident. In separate images for each staining, the two transcription factors are detectable in almost all cells, although not with identical intensity. K14 is present everywhere but is particularly marked at the expanding perimeter of the colony.

One such colony was allowed to grow for 20 days. This period of time allowed stratification and terminal differentiation. Immunostaining of such a colony for involucrin (Fig. 2) showed that most regions of the colony contained K14 (Fig. 2 A) and involucrin (Fig. 2B), but some regions contained only K14, suggesting the absence of stratification in those regions. These experiments show conclusively that the colonies whose founding cell originated from nodules formed in scid mice are keratinocytes, although they may not be identical with keratinocytes cultured from epidermis of postnatal humans.

Fig. 2.

Presence of involucrin in a keratinocyte colony derived as in Fig. 1. A primary colony formed with 3T3 support was fixed and stained on day 20. Each cell containing p63 (red) also contains K14 (green) (A). Most regions of the colony contain involucrin (blue) (B). In squamous epithelium, K14 synthesized in the basal layer persists in the suprabasal layers (60) but not in cornified cells. In comparing A and B it is clear that cells brightly stained for K14 do not contain appreciable involucrin (arrowheads), whereas cells containing involucrin are faintly stained for K14 (arrows). Presumably, complete destruction of K14 has not yet taken place.

From serial photographs of several different living colonies and cell counts carried out on enlarged prints, we obtained cell-doubling times of 16–22 h, up to a colony size of 1,000 cells. We made secondary and tertiary subcultures of such colonies, but in these experiments we did not obtain holoclones with persistent high growth potential.

Using this detection system, we recovered keratinocyte colonies at a frequency of about 1 per 10⁴ cells plated. This sensitivity is evidently greater than that afforded by histological sections stained with hematoxylin/eosin, because in serial sections of the nodules we were unable to positively identify any stratified squamous epithelium among the differentiated tissues formed.

Differentiation from Cultured ES Cells: The First Step Marked by the Loss of Oct4. ES cells are known to contain the germ-line-specific nuclear transcription factor Oct4, a member of the POU family (46–51). To study the early development of the keratinocyte lineage beginning with cells containing Oct4, we found it convenient to deposit, in the middle of a 60-mm tissue culture dish, a single embryoid body previously prepared from aggregated ES cells. The embryoid body quickly attached to the surface, and centrifugal cell migration began from its perimeter. Five days later, the embryoid body still contained Oct4 (Fig. 3). In the zone of migration, Oct4 began to disappear quite close to the embryoid body and was absent from nearly all cells located close to the migration front. We now consider how migrating cells develop along a keratinocyte lineage.

Fig. 3.

Disappearance of Oct4 from cells migrating out of an embryoid body. A single embryoid body was deposited on a tissue culture dish. Five days later, the culture was fixed and stained for Oct4. Arrows in all photographs indicate direction of migration. The cells of the embryoid body (left) stain brightly for nuclear Oct4 (red). A few cells located in the migration zone close to the embryoid body have retained detectable Oct4, but, beyond this, very few of the 4′,6-diamidino-2-phenylindole-stained nuclei (blue) contain even a trace of Oct4.

p63: An Early Marker of the Keratinocyte Lineage. In 5-day cultures of an embryoid body, we detected in the nearby migrating region, a few cells with nuclei containing p63. At 15 days we found large clusters of such cells, located at some distance from the embryoid body (Fig. 4). Only 5.8% of the p63-containing cells also possessed K14. No K14 could be identified in cells not containing p63.

Fig. 4.

Appearance of p63 and K14 in cells migrating from an attached embryoid body inoculated 15 days previously. The embryoid body is located to the left outside the photograph. Of 517 p63-containing cells, only 30 also contained K14 (green). Because K14 is a cytoplasmic protein, it extends beyond the corresponding p63-containing nucleus. No K14-containing cell lacked p63. Cells lacking both markers are revealed by 4′,6-diamidino-2-phenylindole staining for DNA (blue). Because no supporting feeders are present, those cells, although they lack distinctive morphology, must belong to nonkeratinocyte human lineages.

The Order of Appearance of Basonuclin and K14 in p63-Containing Cells. After 27 days of migration, p63-containing cells were abundant close to the migration front (Fig. 5A). Of 114 such cells, 50 (or 44%) also contained K14. These cells often contained basonuclin as well (compare Fig. 5 B and C).

Fig. 5.

Cells of the keratinocyte lineage close to the migration front. After migration from an embryoid body for 27 days, the migration front (A) shows numerous cells with K14 (green), p63 (red), and basonuclin (blue). At higher power (B and C), the presence of cytoplasmic K14 (B) can be correlated with the presence of purple staining nuclear basonuclin (C, white arrowheads). Other nuclei contain only p63 (white arrows).

In the part of the migration zone located close to the embryoid body (far behind the front), cells containing p63, even when numerous, did not contain basonuclin. In this region, we observed a few cells that definitely contained basonuclin, but no p63. These might be primordial germ cells, which are known to contain basonuclin (52–54) and which have recently been shown to develop in cultured embryoid bodies (55). It was only in cells that had advanced further from the embryoid body that basonuclin appeared in cells of the keratinocyte lineage, because these cells also contained p63.

The relation between the times of appearance of basonuclin and K14 in the keratinocyte lineage was examined in the middle region of the migration zone by scoring those cells containing p63 and K14 but no basonuclin and those cells containing p63 and basonuclin but no K14. In four experiments, we found 167 cells in the first category and 26 in the second. This finding indicates that, in 86% of the total, the appearance of K14 preceded that of basonuclin. We conclude that after the appearance of p63, K14 and basonuclin appear nearly simultaneously; in general, K14 is detected earlier.

Formation of Involucrin by Cells Migrating from an Embryoid Body. After a 13-day period of cell migration from an embryoid body, the cells were trypsinized and inoculated onto a culture containing a feeder layer of 3T3 cells and fed with FAD medium. Some slowly growing keratinocyte colonies developed. Like keratinocytes cultured from the nodules formed in scid mice, these colonies were found to contain cells possessing involucrin by immunostaining (Fig. 6).

Fig. 6.

Appearance of involucrin in a stratified keratinocyte colony originating from the migration region of a cultured embryoid body. After 13 days of migration, the cells were trypsinized and inoculated onto 3T3 feeders. Twenty-four days later, a culture containing a colony with the appearance of keratinocytes was fixed and stained. In this merged photograph of the colony, nearly all cells contain p63 (red). Most cells appear to contain K14 (green). Scattered squame-like structures overlying the basal layer contain involucrin (blue).

Effect of Addition of 3T3 Cells to a Culture Containing Cells Migrating from an Embryoid Body. In such experiments, cell migration from an embryoid body was allowed to proceed for 8 days and then lethally irradiated 3T3 cells were added to a density of 2.6 × 10⁴ per cm². After an additional 19 days, nearly all the cells close to the migration front contained p63, K14, and basonuclin (Fig. 7). These conditions therefore resulted in concentration, close to the migration front, of an almost pure population containing the three important markers of the keratinocyte.

Fig. 7.

Concentration of the keratinocyte lineage at the migration front after the addition of 3T3 cells. After allowing migration from an attached embryoid body for 8 days, 2.6 × 10⁴ irradiated 3T3 cells per cm² were added to the culture, and incubation was continued for 19 days. The zone close to the migration front is now nearly completely composed of cells containing p63 (red), K14 (green), and basonuclin (blue).

Discussion

Our results are summarized in Fig. 8.

Fig. 8.

Marker succession in the keratinocyte lineage.

Stages I, II, and III are consecutive. Stage I is defined by the disappearance of Oct4. An interval of time and a degree of cell migration follow before the first marker of the keratinocyte lineage, p63, appears. The presence of p63 in the absence of K14 and basonuclin defines stage II. Such cells are quite numerous early in the process of cell migration. This finding seems consistent with what is known about the appearance of p63 in mouse embryogenesis, for recent studies have found p63 as early as stage E7.5, whereas K5 (partner of K14) does not appear until E12 (F. McKeon, personal communication). It may be postulated that, in human embryogenesis (a slower process than in the mouse), p63-containing cells lacking the later markers should be more numerous.

With further time and migration, K14 and basonuclin appear (stage III). It is curious that basonuclin, which is present in the male and female germ line (20, 52, 53), must disappear soon after fertilization, because we did not detect it in ES cells, and then reappear late in the development of the keratinocyte lineage. Because basonuclin and K14 appear at nearly the same time and K14 is not always the first to be detected, we simply represent the acquisition of both markers as necessary for the transition to stage III. Once p63 has appeared, the other markers of the keratinocyte follow progressively with time and cell migration from the embryoid body. The proportion of p63-containing cells bearing later markers increased from 5.8% to 44% to nearly 100%.

Although cells at stage III have the morphology and staining properties of keratinocytes, we have not yet obtained strains of such cells with the growth potential of postnatal keratinocytes. This finding may mean that the conditions we used to study the development of the keratinocyte lineage in culture (long periods in the absence of 3T3 support), although suitable for analysis of marker succession, are not the best conditions for preservation of growth potential of the cell type.

The keratinocytes identified in our experiments have not been assigned to a particular squamous epithelium (epidermal, oral, esophageal, etc.). The subtype can only be determined by examining differentiation markers of the suprabasal layer where that layer is well developed, as in epithelia grafted to animals. In rodent keratinocytes, such identification may be obscured by metaplastic changes (56, 57), but this complication is less likely in human keratinocytes.

The entire developmental lineage of the keratinocyte can, in principle, be defined by immunostaining for transcription factors known to be components of the keratinocyte (58) or for transcription factors (and their coactivators) that are not yet known in keratinocytes or that might be confined to their precursors. Once the pattern of marker succession has been established, the importance of any given transcription factor can be determined by mRNA ablation by using any of several methods now available.

For selection of other somatic cell types developing from embryoid bodies, different conditions must be used (for example, see refs, 13, 19, and 59). So many differentiation-specific proteins are known that the study of marker succession for different lineages should not be restricted by lack of suitable markers. As in the keratinocyte, transcription factors influencing lineage progression should appear before differentiation markers.