In recent years, a variety of newly developed targeted anti-cancer therapies have successfully been combined with classic therapies, including the combination of EGFR-inhibition and radiotherapy in HNSCC
. However, these new intensified combinations also induce increased toxicity and (acquired) resistance
[3, 13]. The a priori selection of patients that will respond is thus paramount, and research has focused on the identification of resistance mechanisms using different in vitro models. However, the complex tumor microenvironment can also be a source of therapy resistance. Hypoxia can have a great impact on the behavior of a tumor cell and also on the response to treatment
. Therefore, it is important to determine which proteins are influenced by hypoxia, because these proteins could be potential targets to reduce therapy resistance to a variety of treatments, including radiotherapy and EGFR-inhibition.
In the current study, we found that the expression of EGFR and HER2 in vitro was correlated with the in vivo expression, which indicates that the expression of these tyrosine kinase receptors is largely an intrinsic feature of a tumor cell. The in vitro expression of activated EGFR (pEGFR) was also highly correlated with the in vivo expression, while the levels of pHER2 in vivo were clearly higher and varied more in the xenografts than in vitro. These observations suggest that activation of EGFR is possibly more genetically determined, e.g. by mutations present in the different tumor lines, while the activation of HER2 is more determined by factors in the tumor microenvironment that are not present in standard 2D cell culture. Also, in head and neck cancer patients a strong correlation between EGFR and pEGFR levels has been observed
. This correlation supports our observation that the extend of activated EGFR is predominantly influenced by EGFR overexpression and not by other stimuli in the microenvironment.
In contrast to the tyrosine kinase receptors, no in vitro-in vivo correlation was observed for the activated kinases pAKT, pERK1/2 or pSTAT3, indicating that the expression of these activated kinases is influenced by factors in the tumor microenvironment. Here, we show that hypoxia is an activating stimulus for AKT in vivo. This hypoxia-induced increase in pAKT cannot be explained by hypoxia-induced activation of EGFR as is observed in different in vitro models
[16, 17]. EGFR expression was namely predominantly present in oxygenated areas and there was a relatively low overlap between EGFR and pAKT expression in vivo. Besides these observations in our preclinical models, mismatch in EGFR-pAKT expression and the presence of pAKT in hypoxic regions is also observed in HNSCC patient samples
. EGFR-independent upregulation of pAKT by hypoxia has also been observed in lung cancer cells, whereby activation of AKT was induced via the IGF1R/PI3K/AKT pathway
. Also oxidative stress, which can occur during reoxygenation, has been shown to activate AKT in HNSCC cells
. Hypoxia-induced, EGFR-independent, activation of AKT could thus be an important resistance mechanism in HNSCC patients treated with EGFR-inhibition and radiotherapy. Although EGFR is the most commonly overexpressed tyrosine kinase receptor in head and neck cancer, also other receptors are overexpressed like HER2, HER3, and IL-6 receptor
[20, 21], which could possibly play a role in hypoxia-induced activation of AKT. However, we focused on our major finding that activation of AKT is a characteristic of hypoxic cells in HNSCC and therefore a potential target to specifically kill hypoxic cells. Extensive crosstalk between different growth factor receptors, such as EGFR and MET, has been reported
. These growth factor receptors activate similar pathways, which means that cells that overexpress multiple growth factor receptors can sustain survival signaling even if one of the receptors is blocked
. This is exemplified by the study of Erjala et al., which also used a panel of UT-SCC cell lines, that showed that EGFR or pEGFR levels were not correlated, but pHER2 and HER3 levels were correlated with sensitivity to EGFR-inhibition
. Also downstream signaling molecules like pAKT en pERK1/2 were not correlated with sensitivity for EGFR-inhibition. In our cell lines, we did also not observe that overexpression of pEGFR was consistently linked to overexpression of pSTAT3, pAKT or pERK1/2. Therefore, it is more important to determine activation of the common downstream pathway, which is responsible for cell survival, as this will be a more attractive target to overcome treatment resistance than targeting one specific growth factor receptor. In the HNSCC tumor lines studied, we indeed show that pAKT-inhibition decreases cell survival in hypoxic cells, but not in normoxic cells. Hypoxic cells are resistant to a variety of treatment regimens, including radiotherapy
, and as pAKT signaling is an important cell survival pathway
, targeting of pAKT in hypoxic tumors could be a promising way to significantly improve patient outcome. Additionally, multiple animal studies have shown that MK-2206 also inhibits pAKT in vivo and reduces tumor growth
[25–27]. Moreover, Knowles et al. showed that MK-2206 not only reduced primary tumor size in an orthotopic HNSCC model, but also inhibited HNSCC migration in vitro and reduced the number of lymph node metastases in vivo. Although the effect of AKT-inhibition on HNSCC migration could explain the reduced metastases formation in this study, it is also known that hypoxia can induce a metastatic phenotype
. Killing hypoxic cells via pAKT-inhibition, as we show in this study, could thus potentially reduce not only tumor growth, but also the metastatic potential of the tumor. AKT is also a highly druggable target in the clinic since multiple specific AKT inhibitors, including the inhibitor we used in our study, are already tested in phase I/II clinical trials and are generally well tolerated
Although our data show that the microenvironment can induce the expression of activated kinases and that therefore expression levels in tumors do not correspond with cells in vitro, they do not explain why we see very little variation in total expression between the tumors. A possible reason for this is the technique we used to determine expression levels. Western blot analysis determines the total expression in all tumor cells together. However, using immunohistochemistry we observed that the level of expression of the different proteins varied widely between cells in the tumors under the influence of e.g. hypoxia and this spatial information is totally lost by western blot analysis. Thus, possibly by determining the expression in all cells together, these differences between individual cells level out and result in an ‘average’ level of expression, which differs relatively little between tumors as we observed in this study. One of the main advantages of immunohistochemistry is the possibility to analyze specifically tumor cells. Although we used tumors that had a large fraction of viable tumor cells and a very low amount of stromal cells, we cannot exclude the possibility that the presence of small amounts of normal cells affected our results.