Heterogenic expression of stem cell markers in patient-derived glioblastoma spheroid cultures exposed to long-term hypoxia

Aim: To investigate the time profile of hypoxia and stem cell markers in glioblastoma spheroids of known molecular subtype. Materials & methods: Patient-derived glioblastoma spheroids were cultured up to 7 days in either 2% or 21% oxygen. Levels of proliferation (Ki-67), hypoxia (HIF-1α, CA9 and VEGF) and stem cell markers (CD133, nestin and musashi-1) were investigated by immunohistochemistry. Results: Hypoxia markers as well as CD133 and partially nestin increased in long-term hypoxia. The proliferation rate and spheroid size were highest in normoxia. Conclusion: We found differences in hypoxia and stem cell marker profiles between the patient-derived glioblastoma cultures. This heterogeneity should be taken into consideration in development of future therapeutic strategies.

Nestin R&D Systems Inc., MN, USA 196908 1 + 3000 mEnVision Podocytes in human glomeruli [58] Musashi-1 MBL International, MA, USA 14H1 1 + 400 mEnVision Human colon cells [59] Immunohistochemistry Immunostaining was performed using a Dako Autostainer Universal Staining System (Dako A/S, Glostrup, Denmark). All reagents were obtained from Dako A/S, Denmark, and used as described by manufacturer's protocol. For more details, see previous publications [48,50,51]. Applied antibodies are listed in Table 1. Omission of primary antibodies served as negative controls as well as controls for nonspecific staining induced by the detection systems alone. Reactions were evaluated digitally by measuring the staining intensities of the spheroids using the VIS software (Visiopharm, Hørsholm, Denmark) [46,47,52]. Briefly, 20 randomly selected spheroids were marked as regions of interest. Dependent on the cellular location of the given protein, software classifiers for either membrane staining, nuclear staining or cytoplasmic staining were trained. Mean staining intensity and positive area fractions were calculated by the software within each regions of interest. Total immunopositivity was estimated as mean staining intensity × positive area fraction. The software classifiers were trained to identify specific staining reactions only. Potential background staining was below the cutoff level.

Statistics
For comparison of staining intensity one-way analysis of variance with Bonferroni correction was used. Unless otherwise stated data are shown as mean ± SEM, n = 20 and statistical significance */ § p < 0.05, **/ § § p < 0.01 and ***/ § § § p < 0.001. Statistical analyses were performed using Statistical Package for the Social Sciences software (SPSS, IBM Corporation, NY, USA).

Spheroid area
No observable changes in spheroid area were seen in hypoxia during the 7 days. On the contrary, spheroid area was significantly increased after 7 days in normoxia; in T111, the significant increase was already seen after 48 h ( Figure 1).

Expression of Ki-67
The Ki-67 level was estimated to be in the 60-90% range in the four spheroid cultures. A slight to moderate reduction of Ki-67-expressing cells was seen after 7 days in T78, T87 and T111. In T78 and T111 growing in hypoxia, the percentage of Ki-67-positive cells decreased from around 80-60% after 7 days ( Figure 2M & P). In T78 and T111, the percentage of Ki-67-positive cells was lower in hypoxia compared with normoxia, especially after 7 days in hypoxia (p < 0.05 and 0.001; Figure

Hypoxia-induced markers
HIF-1α showed a hypoxia-induced upregulation in all spheroid cultures being significant at almost all time points ( Figure 3). The protein was rapidly induced in hypoxia, while it remained absent in normoxia until day 7, where a low expression was found.
In general, only low levels of CA9 were observed in our spheroid cultures. However, higher levels were observed in hypoxia compared with normoxia. In T78 and T87, only a faint CA9 expression was observed ( Figure 4M   (p < 0.001; Figure 4A-L, N & P). Toward the late time points, an increase in the protein expression was also seen in normoxia ( Figure 4N & P). The expression of VEGF was analyzed in spheroids cultured for either 6 h or 7 days in hypoxia and normoxia, respectively. In general, the protein expression was higher in hypoxia compared with normoxia for both time points. Moreover, a higher expression was generally seen after 7 days compared with 6 h for both normoxic and hypoxic spheroid cultures. Variations between the spheroid cultures were still seen. Compared with the other spheroid cultures, T86 showed a high protein expression after 6 h and the increase toward 7 days was relatively small ( Figure 5F). Generally, VEGF was expressed at a low level in T87 ( Figure 5G). In T111 ( Figure 5A-D & H), the protein was rapidly induced by hypoxia but it still increased over time.

Expression of stem cell markers
The expression of the three stem cell markers CD133, nestin and musashi-1 was investigated in spheroids cultured for 6 h and 7 days in hypoxia and normoxia, respectively. In general, the expression patterns varied widely between spheroid cultures, with no systematic hypoxia-induced upregulation of stem cell markers.
There was no significant hypoxia-induced regulation of musashi-1 ( Figure 6Q-X). The musashi-1 level increased from 6 h to 7 days in T87 in normoxia only.

Discussion
In the present study, we aimed to elucidate differences in expression levels of hypoxia-and stem cell-related markers in glioblastoma spheroids of known molecular subtype. In summary, we found the expression of HIF-1α, CA9 and VEGF to increase in response to long-term hypoxia (7 days). Hypoxic spheroid cultures were generally smaller and had a lower proliferation rates compared with normoxic cultures. Finally, the stem cell markers CD133 and nestin were to some degree upregulated after 7 days in hypoxia.
Spheroids grown in normoxia for 7 days were significantly bigger than the corresponding spheroids grown in hypoxia. This was most pronounced for T78 and T111 and in line with the Ki-67 level being significantly reduced in hypoxia in these cultures at day 7. However, in general the proliferation only decreased modestly when the cells were exposed to long-term hypoxia. The prevention of a pronounced decrease in the proliferation could be explained by a general change in glucose metabolism in the tumor cells resulting in some proliferation, although the environment is hypoxic. Currently, it is well known that hypoxia plays multiple roles in cancer; however, the molecular changes at the cellular level are not fully elucidated. As a response to the hypoxic microenvironment, tumor cells have been shown to alter their glucose metabolism by diverting the glucose flux into the pentose phosphate pathway, where no oxygen is needed for the process (review in [60]). Because of this, cancer cells gain a proliferative advantage due to the limitations of oxidative damage by reactive oxygen species [28]. The cancer cells evade apoptosis, obtain genomic instability and unlimited proliferative potential [29]. HIF-1α was expressed in all spheroid cultures at all time points after exposure to hypoxia. In T86 and T87, there was a tendency for a late increase in the protein level. In T78 and T111, HIF-1α showed the highest expression after 6 h. The rapid induction of HIF-1α seen in this study supports previous results of HIF-1α to be induced after 2 [3,16] or 24 h of hypoxia [15,23]. The great variation in HIF-1α expression and different reaction patterns to hypoxia observed in our study could be explained by the pronounced tumor heterogeneity, which might be associated to different resistance mechanisms in tumors. Also the existence of four molecular subtypes described in the introduction might influence the varying HIF-1α expression. In addition, the spheroid cultures formed a variety of spheroids of different sizes varying between the spheroid cultures. Large spheroids will become more hypoxic explaining the variations observed among the cultures. For example, the T111 spheroids were small and the area remained constant after a few days of culturing. The decreasing HIF-1α expression observed in this spheroid culture after 6 h may partly be explained by the lack of increase in spheroid size. HIF-1α is upregulated as an acute response to hypoxia but since the spheroids stop increasing in size, a steady state in the microenvironment seems to be reached and the acute response decreases.
Relatively high differences in CA9 expression were seen between spheroid cultures, with the highest expression detected in T86 and T111. This confirms previous findings showing hypoxia-induced expression of CA9 after 6 h [61]. The protein CA9 is known to regulate the cellular pH and the expression was therefore expected to increase in all spheroid cultures. The upregulation of CA9 is commonly explained by upregulation of HIF-1α [24][25][26], but the CA9 expression and the HIF-1α expression do not necessarily follow each other. CA9 is upregulated as a response to the acidic environment due to the shift in glucose metabolism, but it has multiple functions. CA9 has been shown to be involved in cellular mechanisms such as proliferation, which may explain the variations observed between the spheroid cultures and why CA9 tends to increase under normoxic conditions for the T111 spheroid culture. Whether there is an association between the CA9 expression and the molecular subtype still needs to be elucidated in future studies. The late upregulation and the incipient expression in normoxic samples at the late time points (72 h and 7 days) are presumably due to an increasing spheroid size and thus a tendency of the central parts to become more hypoxic.
A distinct hypoxia-induced upregulation of VEGF was seen in all spheroid cultures, which are in line with other studies [15,23]. At both early and late time points, the expression was higher in hypoxia compared with normoxia. Moreover, the protein was upregulated over time in the individual spheroid cultures. The upregulation seen in normoxic cultures is most likely associated with spheroids being larger and partly hypoxic after 7 days. Though all spheroid cultures showed a hypoxia-induced upregulation, big variations between the cultures were seen. This might be explained by our spheroid cultures being of different molecular subtypes.
Hypoxia-induced upregulation of VEGF is generally thought to be induced by HIF-1α, which may also explain the increase in our setting. Although the expression of VEGF is mainly regulated by HIF-1α/STAT3 other regulators such as the protein tyrosine kinase Src has been identified [62].
Hypoxia has previously been demonstrated to promote the self-renewal of cancer stem cells and maintain the undifferentiated phenotype. Therefore, a clear tendency of hypoxia-induced upregulation of stem cell markers was expected, especially after 7 days [3,23,50,63]. All spheroid cultures expressed the stem cell markers CD133, nestin and musashi-1, but no systematic patterns were observed. For the stem cell marker CD133, the expression level increased significantly from 6 h to 7 days in T78, T87 and T111 in hypoxia. This is in line with previous studies demonstrating a hypoxia-induced increase in the expression of CD133 [3,18,50,64]. The expression of nestin also increased in hypoxia over time in T86 and T87. Though nestin has only been sparsely investigated in the context of hypoxia, these findings confirm previous studies carried out on both glioblastomas by our group [50] and on cochlear stem cells [65]. Regarding musashi-1, the highest expression was detected in T78, T86 and T87, whereas only a faint expression was observed in T111. No hypoxia-induced changes were observed in any spheroid cultures. To our knowledge, musashi-1 has not previously been investigated in the context of hypoxia, but recent studies suggest it as a possible prognostic marker in glioblastomas [45,52]. Although long-term increases in expression of stem cell markers were found, no markers showed the same initial expression level and increase over time in all four spheroid cultures. This may partly be explained by different subtypes: T78 and T111 are of mesenchymal subtype, T86 is of classical subtype and T87 is of proneural subtype [21]. Using The Cancer Genome Atlas database, Zarkoob et al. found a strong correlation between the mesenchymal and neural glioblastoma subtypes and CD133 [66]. However, Brown et al. found the CD133 protein to be enriched in the proneural subtype [18]. In the present study, we found no significant variations between spheroid cultures of the three molecular subtypes, but inclusion of more stem cell markers will be needed to fully investigate this. SOX-2 would be a marker of high interest in such studies. In line with the hypoxia-induced upregulation of CD133 and nestin found in the present study, SOX-2 has previously been reported to be upregulated in response to hypoxia [67,68] and SOX-2 might therefore play an important role in hypoxia-induced biology of glioblastoma. Moreover, due to the pronounced tumor heterogeneity future studies should include more spheroid cultures representing each glioblastoma subtype in order to elucidate the subtype-dependent stem cell-related profile. Nevertheless, the potential subtype-dependent role of stem cell biology and association to the expression of stem cell markers may suggest that future investigations of stem cell-related mechanisms and hypoxia should be investigated to identify the responsible and targetable mechanisms.
Overall, the present study has shown that there are differences in hypoxic-, stem cell-and proliferation-marker profiles between the glioblastoma subtypes but also within each subtype making it an important but complex aspect to address in future treatment strategies. Moreover, it should be taken into consideration that it might be difficult to predict treatment response based on each subtype, since several studies have indicated that multiple molecular subtypes can be found in the same tumor underlining the importance of these hypoxic profiles [69,70].

Conclusion
The present study has shown differences in hypoxia-and stem cell-marker profiles between different glioblastoma cultures. We have demonstrated heterogenous expression patterns in both short-and long-term perspectives. This should be taken into consideration in future glioblastoma research and in the development of new therapeutic strategies. Especially, regarding the anti-VEGF treatment leading to a more hypoxic microenvironment, this may be of critical importance.

Future perspective
Treatment of glioblastomas remains a challenge. Targeting of, for example, VEGF as well as the stem cell aspects of the tumor biology has until now only been of limited success. We believe that a key to successful therapeutic strategies is a better knowledge and understanding of the heterogeneity of the tumor microenvironment. This may reveal different resistance mechanisms, which are associated with escape of some tumor cell populations, while other tumor cell populations are efficiently killed by the targeted therapy. Taking the tumor heterogeneity into account including critical microenvironmental features like hypoxia might therefore bring us closer to a better understanding of the complex biology of glioblastomas. Working in the era of precision medicine, this may suggest that novel therapeutics strategies should focus on microenvironmental aspects being critical in at least a part of glioblastomas. This may step-by-step increase the future survival for patients with glioblastomas.

Acknowledgements
The excellent laboratory work of H Wohlleben and TD Højgaard is gratefully acknowledged. financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Financial & competing interests disclosure
No writing assistance was utilized in the production of this manuscript.

Ethical conduct of research
The authors state that they have obtained appropriate institutional review board approval or have followed the principles outlined in the Declaration of Helsinki for all human or animal experimental investigations. In addition, for investigations involving human subjects, informed consent has been obtained from the participants involved.
future science group www.futuremedicine.com