To study these questions, we have chosen to perform a spontaneous differentiation of HM-1 mouse ES cells in different experimental conditions for 4 days. We compared classical 2D cell culture vs 3D culture in hanging drops with 2 standardized spheroid sizes. To cover a large experimental range, we compared small spheroids from hanging drops with 15 cells starting population (3D small) vs large spheroids from hanging drops with 2000 cells starting population (3D large). These populations were exposed to normoxia and hypoxia (physoxia) of 3.5% oxygen tension. The average diameters of the spheroids after 4 days of spontaneous differentiation were as follows: The small spheroids with 15 cells starting population reached a diameter of 92 μm ± 23 μm (mean ± SD) in normoxia and 73 μm ± 26 μm in hypoxia, respectively. The large spheroids with 2000 cells starting population reached a diameter of 406 μm ± 15 μm in normoxia but were much smaller with 174 μm ± 52 μm in hypoxia, respectively.
Afterward, the genes Bmp4, a marker for mesenchymal differentiation, Nes, and Map2, markers for ectodermal/neural differentiation, were analyzed. Bmp4, a member of the transforming growth factor β (TGF-β) superfamily, was termed bone morphogenic protein due to its discovery from bone extracts. But the contribution of BMPs to vertebrate development has been shown to be so extensive that several researchers have suggested that the name “body morphogenic proteins” would better describe their significance.
Furthermore, Bmp4 is involved in cancer stem cell biology. Looking at the Bmp4 gene expression in this setup, there is a very strong differential gene regulation already in 2D culture when comparing normoxia vs hypoxia. Upon spontaneous differentiation, there is an upregulation of a 2-fold change in normoxia, but a strong downregulation to 0.06-fold in hypoxia (Fig. 1A). In 2D culture, due to the flat cell arrangement, there will be no significant oxygen gradient in the stem cell mass. It thus can be assumed that all cells in hypoxic 2D experience the same oxygen concentration of about 3.5%. Thus, it is tempting to assume that hypoxia generally downregulates Bmp4 gene expression. However, looking at small 3D spheroids cultured in hypoxia, there is no downregulation like in 2D. The gene expression stays almost the same as in the maintenance culture with a relative expression change of 0.8-fold (Fig. 1A). If the gradual lowering of oxygen tension was the trigger to downregulate Bmp4 expression like in 2D, the gene expression of Bmp4 would be expected to be the same or even lower in hypoxic 3D with small spheroids.
In non-vascularized and therefore non-perfused spheroids, an oxygen gradient is built up within the spheroid due to the limited diffusion of O2 into the tissue. Limitation of diffusion distance of oxygen is one of the fundamental reasons why humans and most animals possess a cardiovascular system.
A 3.5% ambient oxygen tension will be experienced by the outermost cell layer and oxygen concentration towards the spheroid center will decrease gradually. By definition, a spheroid always has a lower oxygen tension in the spheroid core than the ambient oxygen tension. It can thus be suggested that biochemical cues provoked by small spheroid formation have an opposing effect on hypoxia-driven downregulation of Bmp4. In the normoxic 3D situation with small spheroids, there is no significant gene expression difference compared to 2D, as in both culture conditions the Bmp4 expression is about double as much compared to the maintenance culture (Fig. 1A). However, if hypoxia would be the driver for downregulation of Bmp4 expression, spheroids with an inherent oxygen gradient should express less Bmp4 than 2D cultures. But this is not the case.
Here, further points have to be discussed. Some evidence suggests that the oxygen decrease in relation to spheroid size is not linear and that in spheroids with a diameter lower than 100 μm oxygen diffusion would not be limited substantially. If this was the case, then in the hypoxic situation with small spheroids, no difference in gene expression should occur. Only in the normoxic situation with small spheroids, this data could explain the lack of difference compared to the 2D situation, because the lowering of the oxygen concentration does not suffice to reach hypoxia in small spheroids up to 100 μm. However, this topic still needs further investigation as a measurement of oxygen tension in spheroids is inherently difficult.
In this respect, it is important to note that normal cell culture conditions are unphysiological in terms of oxygen concentration, as most cells in a mammalian organism never face 21% O2. When oxygen concentration becomes such a crucial parameter for the fidelity of in vitro data with stem cells, consequently ambient oxygen concentration would need to be controlled much better in the experimental setup. Furthermore, it is also important to note that in vitro cells are supplied with oxygen defined by the ambient oxygen tension and limiting diffusion into the tissue, whereas in vivo, oxygen is provided through the vascular system bound to hemoglobin to bring oxygen in close proximity to the cells, depending on the degree of tissue vascularization.
This is important for medical applications like islet cell transplantation to cure Diabetes, where islet spheroids are being transplanted into the comparably hypoxic environment of the liver. In the first 2 weeks, islets are being supplied with oxygen by diffusion only until revascularization has been completed. Hence, spheroid size is also crucially important in applied clinical reality as the sheer cell survival and thus the success of the transplantation depends on the oxygenation that can be provided by diffusion only. As the global scientific community is working relentlessly to develop islet cells from stem cells for the cure of Diabetes, correct spheroid size is now of double importance: correct cellular differentiation and clinical cellular survival.
The strong downregulation of Bmp4 to 0.01-fold of the large 3D cell spheroids in normoxia (Fig. 1A) can be likely explained by severe hypoxia caused by the large spheroid size. But as the small 3D spheroids behave opposite with no Bmp4 downregulation even under hypoxia, there has to be another mechanism of Bmp4 gene regulation counteracting the effects mediated by ambient hypoxia in 2D. The strong downregulation of the large spheroids in hypoxia would fit again into the concept of Bmp4 downregulation caused by hypoxia alone, but here it also could be a metabolic breakdown effect with a drastic reduction of cellular activity due to putative anoxia in the spheroid center. Skiles has similarly analyzed the effect of ambient oxygen tension, spheroid size, and HIF-1α expression. He found that the rise of hypoxia, independent of whether it was caused by spheroid size or ambient hypoxia or the cumulation of both, led to a linear upregulation of HIF-1α expression and corresponding rise in vascular endothelial growth factor (VEGF) secretion. When hypoxia reached a critical point, he found VEGF secretion to drop again, arguing that the entire metabolism would collapse and the adaptive function of HIF-1α would not be functional anymore. Based on this data, he proposed to independently control both spheroid size and ambient oxygen tension, a statement that we strongly agree with. But in contrast to his data, our observations demonstrate a partial disconnection of Bmp4 regulation influenced by the spheroid size and ambient oxygen tension, which would suggest that spheroid size also affects gene expression by alternative pathways than by oxygen-dependent ones. Such alternative regulatory mechanisms could be responsible for Bmp4’s role in fate decisions in a dependence of geometrically defined culture conditions, which is shown in a remarkable paper of the Brivanlou group. They detected mechanisms of pattern formation and self-organization which were clearly related to the size of stem cell colonies. The colonies were cultured in 2D colony culture and even there, a size effect of the stem cell colony could be shown. More deliberately tuned than classical morphogen gradients, they describe a mechanism of solute-dependent inhibitory mechanisms that sheds light on the underlying mechanisms of how diffusion distance influences cell fate.
Our experimental setup does not allow a deeper understanding of the regulatory interdependencies on a mechanistic level. For this, follow-up studies would be needed targeting a deeper understanding of the separate influence of spheroid size vs hypoxia. There, viability and functionality of cells in relation to spheroid sizes as well as individual gene expression on a single cellular level in relation to the spatial position within a spheroid should be analyzed.
Putting these observations into a larger context, it gets clear that if spheroid size and oxygen tension lead to such deliberate fate decisions, we would need to incorporate such knowledge into stem cell differentiation protocols. Fate decision checkpoints would need to be activated not only by soluble agents but additionally by spatiotemporal control. How this relates to our observed behavior of Bmp4 regulation will need further research.
Looking at Map2, a spontaneous upregulation upon differentiation in 2D culture happens both in normoxia and hypoxia. In normoxia, Map2 is upregulated 3.7-fold whereas in hypoxia it is upregulated significantly lower with a 2.3-fold change, respectively (Fig. 1B). Also, here, 2D culture in hypoxia leads to an oxygen tension of 3.5% of virtually every cell. When cultured as a 3D spheroid, however, oxygen tension is supposedly lower than in 2D culture. If hypoxia alone was the trigger for the stem cells to express less Map2 than in normoxia like in 2D, small hypoxic 3D spheroids should therefore express less Map2 than in 2D hypoxia. However, taking into account that diffusion distance might not be linear as discussed above, within small hypoxic spheroids hypoxia could theoretically be not significantly higher than in hypoxic 2D. In this case, the regulatory behavior of Map2 in small 3D spheroids should be the same as in 2D. Unexpectedly, small hypoxic spheroids express significantly more Map2 than in hypoxia 2D (3.7 vs 2.3-fold upregulation, respectively; Fig. 1B). Also, here, the same discussion as in Bmp4 applies, and further research is needed on how much the lowering of oxygen tension in relation to spheroid size applies and which conclusions can be drawn. When cultured as large spheroids, the upregulation of Map2 in the normoxic cultivation is highest with a 7-fold increase compared to baseline (Fig. 1B). However, in these large spheroids, there is by definition significant hypoxia present. It is very difficult to directly measure oxygen concentration within a spheroid without destroying the spheroid architecture. It is however recognized that any multicellular spheroid larger than 200 μm experiences relevant hypoxia. In our study, the normoxic spheroids have reached 406 μm on average, therefore chronic hypoxia, possibly even core anoxia must be present. Thus, although experiencing much more hypoxia than in small spheroids, why is Map2 significantly upregulated (7 vs 4.7-fold, respectively)? The opposing, hypoxia-independent regulatory mechanisms activated by spheroid size must therefore be much stronger than the effects of hypoxia alone. The significantly lower upregulation of large spheroids cultivated in hypoxia of 2.2-fold change (Fig. 1B) could be a similar effect as observed in the work of Skiles, wherefrom a certain low oxygen concentration on the entire regulatory mechanism breaks down as homeostasis cannot be maintained anymore.
For the observed differences of Map2 expression in relation to hypoxia with spheroid size, the literature gives only very scarce data. Mostly studied in hypoxic brain damage models, some reaction on the hypoxic insult yields in colocalization of Map2 expression with HIF-1α, but the mechanism of this relationship remains blurry. To address this question more thoroughly, detailed studies on cell survival in relation to spheroid size and oxygen tension would be needed.
Looking at the Nes expression, in 2D the upregulation is a 4.4-fold change at normoxic conditions and a 3.2-fold change under hypoxia, respectively (Fig. 1C). In small 3D spheroids, there is an upregulation of 4.4-fold in normoxia and 6.7-fold in hypoxia, respectively (Fig. 1C). This upregulation of small spheroids in hypoxia is significant compared to hypoxic 2D, again indicating other, hypoxia-independent mechanisms for Nes gene regulation as hypoxic conditions in 2D and small spheroid are likely to be similar. In large spheroids cultivated in normoxia, the upregulation is 8.3-fold vs. large spheroids in hypoxia with blunted upregulation of 3.7-fold (Fig. 1C). Regarding the lower Nes expression in the large hypoxic spheroids, like described above in the other examples, most probably physiological regulatory mechanisms were broken down due to significant hypoxia up to anoxia, and regulation for this experimental condition has to be interpreted with caution.
The influence of HIF-1α and hypoxia on Nes expression was already described on mesenchymal stem cells, wherein 2D cultures, hypoxia was leading to a HIF-1α dependent Nes overexpression on the RNA and protein level. However, the mechanisms of this regulation are not clear. The authors here suggest indirect regulation by VEGF activation. To the best of our knowledge, Nes does not have functional hypoxia response elements (HRE’s) in the promoter region, which would be needed for direct activation by HIF-1α.
In our study, the mechanisms of the Nes regulation is not solely attributable to spheroid size or oxygen tension, but most likely driven by opposing effects of either stimuli. The many other putative regulatory influences like the diffusion control mechanism of Bmp4 described by Brivanlou (above) are possible explanations. However, as before, this data shows the importance of controlling both spheroid size and oxygen tension in order to be able to understand the individual effects.
The whole experimental approach was additionally performed with HIF-1α deficient (HIF-1α-/-) stem cells. In comparison to the wild type cells, in HIF-1α deficient cells there is almost no significant regulation observed, and if so, there is a very high standard deviation compared to wild type cells. This happened in any condition tested. The lack of robustly detectable regulation in the absence of HIF-1α in contrast to wild type cells points to a central role of HIF-1α in culture-dependent early differentiation gene expression. This can be best seen in the example of Bmp4, where the strong differential regulation in HIF-1α+/+ cell spheroids is not observed in HIF-1α-/- cells (Fig. 1A, D). This is in line with the literature where Bmp4 is found to act downstream of HIF-1α. But how exactly HIF-1α is mechanistically involved leading to such a strong differential regulation needs further studies.
Interestingly, BMP-4 is induced in hepatocellular carcinoma by hypoxia and promotes tumor progression. This effect could be abolished by transfection of a dominant-negative form of HIF-1α. Striking similarities exist between molecular mechanisms driving embryonic liver development and the progression of hepatocellular carcinoma, making this an example of how stem cell and cancer research share a lot of commonalities.
In summary, the described observations demonstrate that gene regulation of Bmp4, Map2, and Nes in stem cell cultures is dependent not only on ambient oxygen tension but also on the spheroid size.
Relating to cancer research, the knowledge of spheroid biology is especially helpful and has been applied broadly as diffusion gradients play a central role in cancer metabolism as well. Only cell spheroids have allowed us to understand tumor biology better and also their resistance to chemo- and radiotherapy. But the impact of cell spheroids in stem cells and cancer research is much larger than originally thought. With the discovery of cancer stem cells and their identification in various types of solid tumors, there is now a common ground of research in both areas. Cancer stem cells have been detected in the brain, breast, lung, colon, melanoma, and ovarian cancer and the list is expanding rapidly. Based on this data, one of the keys for therapeutic advancement for cancer treatment lies within the understanding of stem cell biology. At the same time, it is well known that the biggest danger in stem cell therapy is tumor formation. Therefore, understanding the mechanisms of fate decisions in stem cells is not only imperative for correct differentiation of stem cells for therapeutic applications, but it also allows to survey this correct differentiation with the knowledge gained from cancer stem cell research, where "triggering events" of stem cells getting out of cell-cycle control, are being studied.
The importance of these principles can be seen in clinics. The cancer patient is usually not killed by the primary tumor. The patient is killed by the metastases. These, in turn, develop because the invasive and metastasizing property of tumor cells is unleashed by hypoxia, and hypoxia is controlled by diffusion distance, and diffusion distance is controlled by the spheroid size, i.e. the size of the avascular portion of a growing tumor. In short: the more hypoxic a tumor, the more aggressive its behavior.
Here, science has a direct legal and regulatory impact. Modern risk management methods, which are a prerequisite for any approval of new therapies, are based on such considerations. If derivations for the future are possible on the basis of reproducible observations, i.e. how often could a possible dangerous event occur, and how serious would this event be, then the integration of these observations is legally obligatory.
The data shown here is only a little hint about the importance of rigid control of cell culture conditions like spheroid size and ambient oxygen tension for cellular differentiation protocols. The interdependent and sometimes opposing regulatory effects are not well understood. Because of the clinical significance, these effects should be investigated in detail.
More and more elements needed for physiological cell differentiation are being uncovered. Recent literature now shows that the understanding of spheroid size control on fate decisions is growing as architectural support of every single spheroid within a micropatterned surface allowed more directed differentiation towards specific cell types for clinical applications. It is getting clear that only a stringent control of all involved biological parameters enforced by suitable and appropriate 3D cell culture platforms will allow the safe differentiation of stem cells for clinical applications.