The derivation of osteogenic cells from hESCs would be a significant advance from several perspectives. Not only would genetically manipulated, hESC-derived osteogenic cells serve as a continuous source of cells to study bone diseases, but also an infinite resource of osteogenic precursors for bone regeneration could be generated provided histocompatibility issues could be overcome. Even if hESCs are never used in therapeutic tissue engineering and regenerative medicine, the findings and discoveries made using hESC models will most likely apply to an individual's own induced pluripotent stem (IPS) cells [24
], even though the 2 different pluripotent cell types may not be totally identical. In order for these prospects to become a reality, however, a 2-fold challenge must be overcome: (1) how to compel hESCs or IPS cells to reproducibly differentiate along the osteogenic pathway, and (2) how to reliably eliminate cells capable of teratoma formation from constructs destined for in vivo implantation. None of these problems has been solved, and no studies to date promise an easy solution.
The most achievable of these goals has been in vitro osteogenesis by hESC progeny. Upregulation of osteogenic markers and formation of mineralized nodules by hESC-derived cells, either following embryoid body formation or by avoiding this step, has been achieved by several groups using 2 major approaches: culturing hESC descendants in medium with FBS plus osteogenic supplements and/or co-culturing them with human or mouse osteogenic cells [2
]. However, it was shown previously that patterns of osteogenic differentiation observed in artificial conditions in vitro usually do not correlate with, and cannot foretell, the way that the same cells differentiate upon in vivo transplantation. The degree of osteogenic differentiation of rat and human BMSCs in vitro inversely correlated with the amount of bone formed by these cells in vivo [12
]. Several human and rat BMSC populations that demonstrated strong osteogenesis in vitro formed no bone following in vivo transplantation [14
]. On the contrary, rabbit BMSC strains that displayed adipocytic differentiation in vitro formed good bone in diffusion chambers in vivo [11
]. Human muscle fibroblasts were able to undergo osteogenic differentiation in vitro but not in vivo [17
]. More generally, many human fibroblast-like cell strains derived from various organs that have never shown the ability to form bone in vivo, were able to undergo osteogenesis in vitro [16
]. Taken together, these data strongly suggest that markers and/or phenotypes forced upon cells by artificial in vitro stimuli have little relevance to the genuine differentiation potential of these cells. In addition, in vitro assays are not useful in revealing the tumorigenic potential of hESC-derived cells.
Conversely, the use of transplantation assays under defined experimental conditions [21
] has become a valuable standard for delineating the osteogenic potential of cell populations. Not only does it analyze the physiological osteogenic function in vivo, but it also provides the most convincing result, the formation of histologically proven bone tissue [27
]. These assays are more physiologically relevant than commonly used in vitro differentiation assays. Having said this, obviously, studies in immunocompromised mice are only a start and cannot fully predict how hESC-derived cells will behave in a human environment. In the current study, using the in vivo transplantation assay, histologically proven bone of human origin was formed 8, 12, and 16 weeks post-transplantation by hESC-derived strains generated through several regimes of in vitro differentiation; this bone constituted the first demonstration of in vivo osteogenic capacity of HSF-6-derived cells. Bone of score 2 formed in the transplants of strains H and R was considerably more extensive than any bone, formed in vivo by several hESC lines, described thus far [3
]. The new bone demonstrated numerous osteocytes embedded in lamellar, well-mineralized matrix, as can be judged by parallel organization of collagen bundles revealed under polarized light, and by intense green fluorescence revealed under ultraviolet light; such fluorescence was shown earlier to selectively distinguish mineralized bone in sections stained with eosin-containing dyes [28
]. These results imply that long-term cultivation of hESC-derived strains in differentiating conditions employed in this study, including several weeks in postconfluent, multilayered cultures, followed by multiple passages and transplantation in HA/TCP vehicles, may, indeed, lead to the creation of clinically relevant bone tissue by nudging a subset of hESC descendants toward the osteogenic pathway.
Among multiple medium compositions tested, no single composition was found to ensure consistent bone formation by HSF-6-derived strains. The same was true for other variable culture parameters, such as total time in monolayer culture, time intervals between consecutive passages, and percentage of carbon dioxide in the gas phase of the cultures. Yet, some concoctions, such as KO-DMEM
Dex/AscP (media B and H), promoted bone formation more often than others. Most individual components tested, such as hPL, KOSR, bFGF, ITS+
, as well as medium conditioned by the osteogenic cells, hBMSCs, failed to promote in vivo bone formation by HSF-6-derived cells. The choice of FBS lot and the use of heat-inactivated versus not inactivated FBS did not seem to make any difference either. At this point, it is not clear why KO-DMEM-based media are more favorable to osteogenic differentiation of HSF-6-derived strains than αMEM-based media; it is possible that lower cell proliferation rates observed in KO-DMEM-based media may support a better differentiation process. As far as the stimulating effect of Dex/AscP is concerned, these osteogenic supplements have been long known to accelerate in vitro osteogenic differentiation of cells already committed to osteogenesis. More recently, Dex, either alone or in combination with AscP, was shown to stimulate in vivo bone formation by cells with low osteogenic potential, as well as by osteogenic cells transplanted with less than optimal vehicles [30
]. Since Dex alone was shown to decrease collagen synthesis [34
], AscP may have increased bone formation by stimulating the secretion of extracellular matrix proteins [35
While the use of carpet cultures failed to increase overall incidence of bone formation, the most extensive bone of score 2 was formed by 2 strains exclusively employing the carpet culture technique; in standard, noncarpet, transplants, far less extensive bone of score 1 was formed, at best. To generate carpet cultures, strains H and R, following 4 passages, and 52 to 67 days, respectively, in monolayer cultures, were additionally cultured for 25 and 23 days on the surfaces of HA/TCP particles before transplantation. Whereas the carpet technique employed in this article is a novel one, a number of earlier studies, in an attempt to increase bone formation, used co-cultivation of osteogenic cells with various scaffolds as a prelude to in vivo transplantation, not just in an experimental [30
], but also in a clinical settings [37
]. Our data suggest that co-culturing some HSF-6-derived strains with HA/TCP increases their bone-forming potential and may propel hESC-derived cells toward osteogenesis. In addition, cartilage formation was observed in a few carpet transplants, probably due to the fact that the carpet cultures are rather dense, and most likely, the interior is less exposed to nutrients and oxygen than the periphery.
More generally, HA/TCP particles were used in this study because it had been convincingly demonstrated that among various groups of scaffolds, calcium phosphate ceramics, and, in particular, synthetic, biphasic calcium phosphates (consisting of various proportions of HA and TCP), are by far superior in promoting in vivo bone formation by osteogenic cell populations [22
]. For supporting bone formation, several characteristics of HA/TCP vehicles are important, such as biocompatibility, Ca/P ratio, surface roughness, macroporosity, and interconnectivity of pores [42
]. Apparently, HA/TCP particles manufactured by Zimmer and used in this study possess the right combination of qualities: they promote extensive bone formation by both human and mouse BMSCs [22
]. At the same time, HA/TCP scaffolds are thought to be mostly osteoconductive rather than osteoinductive [44
]; that is, they promote bone formation by cells already put through a differentiative scheme (primed to osteogenesis), such as BMSCs, but do not induce bone formation by nonosteogenic cells, such as skin fibroblasts [22
]. This is most likely the reason why undifferentiated HSF-6 cells did not form bone when transplanted alongside HA/TCP. Yet, some HSF-6-derived strains that showed no progress along the osteogenic line (were ALP-
-negative) formed bone when transplanted with HA/TCP. This suggests that HA/TCP may be slightly osteoinductive, depending on the stage of cell differentiation, and can nudge toward bone formation cells pre-committed to osteogenesis. Moreover, for some HSF-6-derived strains, such as strain H, a longer contact with HA/TCP brought about by carpet cultures may further increase the osteoinductive influence. Taken together, our results suggest that the use of HA/TCP scaffolds, possibly in conjunction with the carpet technique, and further refinement of the culture conditions may improve osteogenic differentiation by hESC-derived cells.
In an attempt to identify early indicators of subsequent bone formation, we compared the transcription levels of several genes associated with osteogenesis in 4 HSF-6-derived strains, 3 of which demonstrated contrasting patterns of bone formation. Strain B formed bone in 2 out of 3 regular transplants but not in carpet transplants. Strain H formed bone in 2 out of 3 carpet transplants, and in one of these transplants, the most extensive bone of score 2 was observed; this strain, however, formed no bone in regular transplants. Strain N formed no bone in any transplants and was, thus, chosen as a negative control for bone formation. Expression of 4 widely accepted osteogenic markers was compared in these strains, as well as in undifferentiated HSF-6 cells and in strongly osteogenic hBMSCs.
Collagen type I is widely distributed in most connective tissues and is also the most abundant bone protein expressed during the osteoblast proliferation and extracellular matrix biosynthesis [49
]. Collagen type I
was expressed by strains B, H, N, and P (at similar levels by 3 latter strains, and at slightly higher level by strain B), and, at significantly higher level, by hBMSCs. The transcription factor Runx2 induces osteogenic commitment by promoting expression of osteoblast-specific genes, and is an accepted early marker of osteogenic commitment [48
was expressed at low, and similar, levels by strains B, H, N, and P and, at slightly higher level, by hBMSCs. ALP is the most commonly used marker of osteogenic differentiation downstream of Runx2; it provides inorganic phosphate to promote mineralization, and also facilitates mineralization by hydrolyzing its inhibitor [49
]. Among the strains studied, ALP
was expressed by undifferentiated hESCs (in agreement with some previous observations [3
] and contrary to other [2
]), as well as by hBMSCs; it was indistinguishable in all of HSF-6-derived strains, osteogenic or not. BSP is a protein found almost exclusively in mineralized tissues where it is important for nucleation of HA crystal formation; as such, BSP is considered a very late marker of osteogenic differentiation [56
]. Among the strains tested, BSP
was expressed only by osteogenically committed hBMSCs. Taken together, these 4 markers of osteogenesis, Coll I
, and BSP
, failed to distinguish between the osteogenic and nonosteogenic HSF-6-derived strains. Recently, similar findings were reported for osteogenic and nonosteogenic strains of hBMSCs: no positive correlation was found between the osteogenic potential of the strains and expression of Runx2, ALP, collagen type I, osteopontin
, and BSP
]. Apparently, many of our HSF-6-derived strains contained cells committed to mesenchymal differentiation (expressed collagen type I
) and primed for osteogenesis (expressed Runx2
). However, even the best bone-forming strains, B and H, showed no further progression along the osteogenic line of differentiation, and did not spontaneously achieve the stages of osteogenic maturation where ALP
are expressed. To proceed toward osteogenesis, these strains required an additional nudge accomplished by the slight osteoinductive effect of HA/TCP.
Chondrogenic differentiation is closely related to, and sometimes precedes, osteogenesis. The following 3 markers were thus included into our gene expression assay: collagen type II and aggrecan, the most abundant cartilage proteins considered classical markers of chondrogenesis, and collagen type X, which is a marker of late-stage chondrocyte hypertrophy associated with endochondral ossification [59
]. All of these markers were expressed at either low or very low levels by the strains under study, and none of them could distinguish between the osteogenic and nonosteogenic HSF-6-derived strains.
Previously, a small number of publications described in vivo bone formation by cells, derived from several hESC lines, that had been either pretreated with osteogenic supplements [5
] or co-cultured with bone-derived cells [19
]. Four to 11 weeks post-transplantation, small, mineralized, von Kossa–positive deposits were observed that were designated “bone” despite absence [5
] or very poor bone histology [19
], and no proof of donor origin. In the most convincing demonstration of bone formation by hESC descendants published to date [3
], osteogenic differentiation was induced by culturing BG01-derived cells in medium with osteogenic supplements followed by infection with a GFP-carrying retroviral vector and selection of cells positive for GFP and ALP. The hESC-derived strain thus generated was transplanted into calvarial defects of immunocompromised mice where 6 weeks later, small nodules of histologically convincing bone were observed. Unfortunately, Arpornmaeklong and co-authors [3
] did not report whether bone nodules were formed in all recipients, in some, or just in a single one, suggesting that, in their hands, bone formation was sporadic at best, as it was in our study. In that article, the new bone was formed only in the periphery of the circular calvarial defect, close to the recipient's bone surrounding the defect. While human origin of at least some osteocytes was shown by immunostaining, the authors concluded that “human cells participate in the regeneration of bone,” leaving room for parallel bone formation by the osteoblasts of recipient origin. In this regard, transplantation into a heterotopic site where transplanted cells would not be influenced by host skeletal tissues seems to better suit the purpose of analyzing the true differentiation potential of cells under study.
Before the initiation of differentiation, HSF-6 cells demonstrated a typical morphology of undifferentiated hESC colonies, showed high expression of markers of pluripotency, Oct4
], and had a normal karyotype that featured no aneuploidy often developed by cultured hESCs [64
]; together with teratoma formation, these features proved the undifferentiated nature of HSF-6 cells [55
]. In differentiating conditions, during prolonged cultivation in various media, HSF-6-derived strains underwent substantial morphological changes. At passage 0, they demonstrated an array of diverse phenotypes, reminiscent of multiple tissue types, as well as the formation of multilayered sheets and complex structures. By passage 3, all strains acquired more uniform, fibroblast-like, yet somewhat variable, morphology. Despite the wide variety of phenotypes observed in the strains at passage 0 and, to a lesser extent, at passage 1, no obvious correlation was observed between the cell morphologies of the strains and the frequencies of either bone or teratoma formation by these strains following several passages and in vivo transplantation. All HSF-6-derived strains analyzed at passages 3 and 4 demonstrated high viability and high attachment capability; no significant chromosomal aberrations occurred in these strains over the extended culture period in differentiating conditions.
For the future therapeutic use of hESCs or IPS cells, prevention of tumor formation represents a goal at least as important as the primary goal of achieving a desired, in our case, osteogenic, differentiation. Teratomas, or tumors consisting of a range of differentiated somatic tissues of all 3 germ layers, are formed by hESCs implanted into immunocompromised mice in vivo [55
]. While diploid, low-passage hESCs form mostly benign teratomas, culture-adapted hESCs can switch to formation of highly malignant teratocarcinomas [55
], an extremely alarming prospect, potentially jeopardizing the entire field of hESC/IPS-based therapy. The formation of teratomas was observed in all transplantation sites tested, and also by hESCs both injected without any vehicle and transplanted in conjunction with various scaffolds [67
]. For the latter reason, we did not investigate new transplantation conditions and used our standard technique that employs HA/TCP-based scaffolds, for transplantation of both undifferentiated HSF-6 cells and differentiated HSF-6-derived strains. Using this approach, teratomas consisting of multiple tissues, but mostly of mesenchymal-type cell condensations, were routinely observed in 12- to 20-week-old transplants of undifferentiated HSF-6 cells.
It was suggested earlier that to eradicate teratoma formation by hESC-derived cells, period of their cultivation in differentiating conditions should be
>2 weeks, thus allowing them to differentiate beyond the pluripotent phenotype [69
]. We cultured HSF-6-derived strains in various differentiating conditions for much longer periods, ranging from 7 to 14.5 weeks. Yet, teratoma-like nodules were formed 12 to 20 weeks post-transplantation by some, but not by other strains of HSF-6 descendants, suggesting that contamination with less differentiated cells persisted in some, but not in all, of the strains. Molecular analysis demonstrated that neither strains B and H, that formed no teratomas, nor strains N and P, that gave rise to either less frequent or ample teratomas, respectively, expressed Oct4
; all these strains, thus, lost their undifferentiated status [63
]. Apparently, the expression of the pluripotency markers was not a prerequisite for teratoma formation. Markers of osteogenic and chondrogenic differentiation analyzed in this study were not predictive of teratoma formation either, as could be expected. In an attempt to further characterize HSF-6-derived strains and to find molecular signs predicting tumor formation, the expression of embryonic germ layer markers [65
] was analyzed. For endoderm, it was α-fetoprotein, which is synthesized by the yolk sac visceral endoderm and by fetal liver, and is highly expressed by teratocarcinomas and liver tumors [71
], and α-1-antitrypsin, a serpin protease inhibitor [73
]. Neither of these genes was expressed by any of the 6 strains studied. For ectoderm, we analyzed the expression of 2 keratins that are filament-forming proteins of epithelial cells: keratin 14 of stratified epithelium and keratin 17 of squamous epithelium [74
], and of neurofilament, the cytoskeletal component of myelinated axons from central and peripheral nervous system [75
]. While neurofilament
was expressed at very low levels by all strains studied, keratins
were absent from both HSF-6 cells and hBMSCs but were considerably, and variably, expressed by 4 HSF-6-derived strains: their expression was significantly higher in nontumorigenic strains B and H than in teratoma-forming strains N and P. This result may be interpreted as an indication that our HSF-6-derived strains, which are of nonclonal origin, represent mixed cell populations and, in addition to cells exhibiting mesodermal differentiation, can include cells with markers of ectoderm. Surprisingly, the strains that contained higher numbers of ectodermal, keratin-expressing cells demonstrated no tumorigenicity, contrary to those strains that included fewer ectodermic cells, but, apparently, more cells of undifferentiated nature bearing some other, as yet unknown, markers. Obviously, this hypothesis is highly speculative and should be further corroborated by broader and more rigorous data. More generally speaking, it would be very important, indeed, to positively identify those genes whose transcript levels could pick out the hESC-derived strains capable of forming tumors. Without being able to directly eliminate tumorigenic cells from hESC-derived populations, it still may be possible to develop culture conditions unfavorable for proliferation and survival of these cells. Certainly, the effectiveness of such “cleansing” should be supported by repeated in vivo transplantation experiments before even considering therapeutic use of these cell populations. At the moment, for absence of a better way, this has been the approach employed in this article. In our experiments, all teratomas, with a single exception, were formed by the strains generated in the αMEM-based media, the same media that promoted the fastest cell proliferation rates and the highest cell yields. Only a single case of a teratoma-like nodule was observed among all transplants of strains generated in media based on KO-DMEM, the media that, quite coincidently, promoted much slower cell proliferation. Interestingly, no teratomas were formed in Experiment 3 where 50% of all media represented medium conditioned by hBMSCs. Even strains S and U that were generated in entirely αMEM-based media, and that demonstrated very high proliferation rates, formed no teratomas. This observation suggests that media conditioned by hBMSCs expedite hESC differentiation and inhibit teratoma formation. It cannot be ruled out, however, that a longer cultivation of HSF-6 colonies on the MEF layers before the passage 0 (20 days in Experiment 3 vs. 6 days in Experiment 2 and 9 days in Experiment 1) could by itself have decreased the tumorigenic potential of the strains. Previously, no teratomas were observed in 6-week-old transplants of hESC-derived cells [3
]. Similarly, our 8-week-old transplants demonstrated no teratoma formation; all tumors were observed in the transplants harvested between 12 and 20 weeks post-transplantation. Apparently, to reliably rule out tumorigenic potential of hESC descendants, and, in particular, to make these observations clinically relevant, the transplants should be allowed to stay in vivo for longer periods of time. This corresponds to the course taken by others who routinely screened recipient animals for teratomas 3 to 4 months after hESC injection [76
Recently, a method for consistent in vivo bone formation by mouse ESC-derived cells was designed, based on their cultivation on ceramic scaffolds in chondrogenic medium, followed by in vivo implantation of the constructs [77
]. While such an approach did not work for hESCs, this achievement gives hope that a comparable accomplishment with hESCs may be within our reach. The sequence of treatments that can drive dependable osteogenic differentiation of hESCs or IPS cells in vivo still needs to be developed. The current article, not unlike all other studies published to date, has not achieved consistent bone formation by hESC progeny. Yet, it presents the most extensive in vivo bone formation by hESC-derived cells attained thus far; it would be important to verify whether similar results could be achieved by applying our methodology to other strains of hESC, as well as of IPS cells. This approach, based on a relatively simple, although lengthy, technique, without complex steps such as viral infection or cell selection, seems to be a promising avenue for further pursuit of hESC-and IPS-based bone formation. We hope that this article elucidates ways, both to be followed and to be avoided, in the hunt for this important goal.