HIF-1α is Overexpressed in Odontogenic Keratocyst Suggesting Activation of HIF-1α and NOTCH1 Signaling Pathways

Background: The odontogenic keratocyst (OKC) is an odontogenic cyst that shows aggressive and intriguing biological behavior. It is suggested that a hypoxic environment occurs in OKC, which led us to investigate the immunoexpression and location of hypoxia-inducible factor 1-alpha (HIF-1α) and other hypoxia-related proteins. Methods: Twenty cases of OKC were evaluated for the expression of Notch homolog 1 (NOTCH1), HIF-1α, disintegrin and metalloproteinase domain-containing protein 12 (ADAM-12), and heparin-binding epidermal growth factor-like growth factor (HBEGF) by immunohistochemistry and compared to eight control cases of calcifying odontogenic cystic (COC), orthokeratinized odontogenic cyst (OOC), and normal oral mucosa (OM) in basal and parabasal layers. Results: In OKC, all the proteins tested were expressed significantly higher in both basal (except for NOTCH1 and HBEGF in OOC) and suprabasal epithelial layers compared to controls. Looking at the epithelial layers within OKC, we observed an increased NOTCH1 and HIF-1α expression in parabasal layers. Conclusions: These results suggest that hypoxia occurs more intensively in OKC compared to COC, OM, and OOC. Hypoxia appeared to be stronger in parabasal layers as observed by higher HIF-1α expression in upper cells. Overexpression of NOTCH1, ADAM-12, and HBEGF in OKC was observed, which suggests that microenvironmental hypoxia could potentially regulate the expression of hypoxia-related proteins, and consequently, its clinical and biological behavior.


Introduction
Odontogenic cysts and tumors arise from cells or tissues that are associated with tooth development and commonly located in jaw bones [1]. Among them, the odontogenic keratocyst (OKC), which was

Length, Inflammation, and Immunostaining Assessment
Brightfield images from 10 arbitrary regions were selected from each sample and acquired using an AxioScope microscope equipped with an AxioCam HRC color CCD camera (Carl Zeiss, Oberkochen, Germany), using a 40× objective. The length of the basal layer was measured using the "Freehand line selections" tool from software ImageJ (public domain software developed by Wayne Rasband e NIMH, NIH, Bethesda, MD; http://rsbweb.nih.gov/ij/).
The inflammation score was assessed by counting the total number of inflammatory cells adjacent to the basal layer of the epithelium in each of the 10 images. Analysis of inflammation was carried out in grades: grade 0: no inflammation; grade 1: <15 cells/field; grade 2: 15-50 cells/field; and grade 3: >50 cells/field. The inflammatory score was calculated as the average of all fields examined. Samples of OKC, COC, and OM were divided into two groups according to the inflammation score: Group A: grades 0-2 (mild-to-moderate inflammation) and Group B: grade 3 (intense inflammation) [24].
Images were stacked to RGB and segmented using the H DAB vector of the color deconvolution plug-in of ImageJ. Channel 2 (color 2) was selected and images adjusted using the threshold tool.
Areas stained with diaminobenzidine were identified and measured after selection using the freehand tool. The epithelial area was assessed in two different segments that corresponded to the basal and suprabasal layers. Results were expressed as the average percentage of labelling area (%) of the epithelial layers. Differences in the percentages of stained areas for the basal and suprabasal epithelial layers of OKC, COC, and OM were analysed.

Statistical Analysis
Data obtained from the experiments were analyzed using GraphPad Prism 5 software (GraphPad Software Inc., San Diego, CA, USA). Parametric ANOVA followed by Tukey's post-test was used after analysis of normality to assess differences between the three groups: OKC, COC, and OM. Student's t-test was used to compare basal and parabasal layers of protein expression within each lesion and between the two groups of inflammation scores.

Basal Layer Length and Degree of Inflammation
The basal layer length of the lesions varied between 5.9 mm and 15.7 mm in OKC (mean = 8.8; SD = 1.8), 6.3 mm and 13.9 mm in COC (mean = 8.9; SD = 1.7), 5.2 mm and 17.4 in OOC, and 3.7 mm and 18.3 mm in OM (mean = 8.9; SD = 2.9). Inflammation assessment is detailed in Table 1. There was no difference among inflammation groups in OKC and the other lesions groups could not be analyzed due to the limited number of samples in the grade 3 group.  (Figures 1-4). NOTCH1 exhibited intense cytoplasmic immunostaining in OKC, showing increased expression in the upper layers of parabasal cells adjacent to the cystic lumen ( Figure 1A,B). NOTCH1 nuclear expression was observed in some cells of the upper parabasal layers ( Figure 1A). COC exhibited a weak and selective NOTCH1 expression in a few cells of the basal layer and a membrane staining in ghost cells ( Figure 1C). OM samples showed a weak NOTCH1 expression, especially in the basal and suprabasal layers adjacent to the basal layer. HIF-1α showed a strong immunolabelling in the upper layers of parabasal cells adjacent to the cystic lumen similar to what was observed with NOTCH1 in OKC (Figure 2A,B). Staining was nuclear and cytoplasmic and also observed in central epithelial islands of cystic formation ( Figure 2C). A weak nuclear but not cytoplasmic staining was found in basal layer cells ( Figure 2D). COC exhibited a very weak HIF-1α expression in a few parabasal cells and in selected cells of the basal layer ( Figure  2E). In OM samples, HIF-1α expression appeared stronger than in COC but weaker than in OKC ( Figure 2F). A very weak HIF-1α expression was found in OOC, similarly to what was observed in COC, but with a more evident expression in the basal layer ( Figure 2G). HIF-1α showed a strong immunolabelling in the upper layers of parabasal cells adjacent to the cystic lumen similar to what was observed with NOTCH1 in OKC (Figure 2A,B). Staining was nuclear and cytoplasmic and also observed in central epithelial islands of cystic formation ( Figure 2C). A weak nuclear but not cytoplasmic staining was found in basal layer cells ( Figure 2D). COC exhibited a very weak HIF-1α expression in a few parabasal cells and in selected cells of the basal layer ( Figure 2E). In OM samples, HIF-1α expression appeared stronger than in COC but weaker than in OKC ( Figure 2F). A very weak HIF-1α expression was found in OOC, similarly to what was observed in COC, but with a more evident expression in the basal layer ( Figure 2G). In the center of epithelial islands formed by the odontogenic epithelium, the so-called pupal cysts, intense cytoplasmic (C, arrow) and nuclear (C, arrowhead) staining was seen. Weak nuclear but not cytoplasmic staining was found in basal layer cells (D, arrowheads). COC exhibited a very weak HIF-1α expression in a few parabasal cells. Staining in the basal layer was very weak and in a few selected cells. (E) HIF-1α expression in OM appeared to be stronger than in COC but still weaker than in OKC. The pattern of HIF-1α expression was more similar to that seen in OKC, In the center of epithelial islands formed by the odontogenic epithelium, the so-called pupal cysts, intense cytoplasmic (C, arrow) and nuclear (C, arrowhead) staining was seen. Weak nuclear but not cytoplasmic staining was found in basal layer cells (D, arrowheads). COC exhibited a very weak HIF-1α expression in a few parabasal cells. Staining in the basal layer was very weak and in a few selected cells. (E) HIF-1α expression in OM appeared to be stronger than in COC but still weaker than in OKC. The pattern of HIF-1α expression was more similar to that seen in OKC, where staining intensity increased towards upper cells of the parabasal layers. HIF-1α expression in OOC was very weak, observed in selected basal and suprabasal cells (G). Scale bar: 20 µm.
ADAM-12 and HBEGF showed a similar pattern of expression in each group of lesions (Figures 3 and 4). ADAM-12 and HBEGF were expressed in the cytoplasm of all epithelial strata of OKC ( Figure 3A,B and Figure 4A,B). There was an even distribution in all the epithelial layers for both proteins in OKC. ADAM-12 expression in COC was generally weak, nuclear and cytoplasmic immunolocation was observed in the basal layer, while a less evident cytoplasmic expression was seen in the upper parabasal cells. However, HBEGF expression was predominantly cytoplasmic in COC, with no staining in ghost cells for both proteins. Expression of ADAM-12 and HBEGF in OM samples showed a comparable pattern seen in COC with similar weak staining. In OOC, ADAM-12 expression was very weak in all layers ( Figure 3E), while HBEGF expression was moderate in the basal layer but very weak in parabasal layers ( Figure 4E). where staining intensity increased towards upper cells of the parabasal layers. HIF-1α expression in OOC was very weak, observed in selected basal and suprabasal cells (G). Scale bar: 20 µ m.
ADAM-12 and HBEGF showed a similar pattern of expression in each group of lesions ( Figures  3 and 4). ADAM-12 and HBEGF were expressed in the cytoplasm of all epithelial strata of OKC ( Figure 3A, B and Figure 4A,B). There was an even distribution in all the epithelial layers for both proteins in OKC. ADAM-12 expression in COC was generally weak, nuclear and cytoplasmic immunolocation was observed in the basal layer, while a less evident cytoplasmic expression was seen in the upper parabasal cells. However, HBEGF expression was predominantly cytoplasmic in COC, with no staining in ghost cells for both proteins. Expression of ADAM-12 and HBEGF in OM samples showed a comparable pattern seen in COC with similar weak staining. In OOC, ADAM-12 expression was very weak in all layers ( Figure 3E), while HBEGF expression was moderate in the basal layer but very weak in parabasal layers ( Figure 4E).

Expression of NOTCH1, HIF-1α, ADAM-12, and HBEGF is higher in OKC when compared to COC, OM, and OOC
OKC overexpresses NOTCH1, HIF-1α, ADAM-12, and HBEGF in either basal (except for NOTCH1 and HBEGF in OOC) and parabasal layers when compared to COC, OM, and OOC ( Figure  5A,B). NOTCH1 and HIF-1α were more expressed in parabasal layers of OKC ( Figure 5C). ADAM-12 and HBEGF did not show a statistical difference within their epithelial layers. No differences were observed between the basal and parabasal epithelial layers in COC ( Figure 5D) and OOC ( Figure 5F) for all proteins studied. However, OM samples showed differences only for the expression of ADAM-12, with predominant staining in the basal layer (p > 0.0001) when compared to the parabasal layers ( Figure 5E). 3.3. Expression of NOTCH1, HIF-1α, ADAM-12, and HBEGF is higher in OKC when compared to COC, OM, and OOC OKC overexpresses NOTCH1, HIF-1α, ADAM-12, and HBEGF in either basal (except for NOTCH1 and HBEGF in OOC) and parabasal layers when compared to COC, OM, and OOC ( Figure 5A,B). NOTCH1 and HIF-1α were more expressed in parabasal layers of OKC ( Figure 5C). ADAM-12 and HBEGF did not show a statistical difference within their epithelial layers. No differences were observed between the basal and parabasal epithelial layers in COC ( Figure 5D) and OOC ( Figure 5F) for all proteins studied. However, OM samples showed differences only for the expression of ADAM-12, with predominant staining in the basal layer (p > 0.0001) when compared to the parabasal layers ( Figure 5E). and parabasal (B) layers between OKC, COC, OM, and OOC. OKC overexpressed all the studied protein when compared to COC, OM, and OOC in either basal (A, except for NOTCH1 and HBEGF in OOC) and parabasal (B) layers. Immunoexpression between basal and parabasal layers within the same lesions was also evaluated in OKC (C), COC (D), OM (E), and OOC (F). Differences were only seen for NOTCH1 and HIF-1α in OKC and ADAM-12 in OM. In OKC, NOTCH1, and HIF-1α were more expressed in parabasal layers, while ADAM-12 in OM showed a higher expression in the basal layer. * p < 0.05; ** p < 0.01; *** p < 0.001. Figure 5. Comparison of NOTCH1, HIF-1α, ADAM-12, and HBEGF immunolabelling in basal (A) and parabasal (B) layers between OKC, COC, OM, and OOC. OKC overexpressed all the studied protein when compared to COC, OM, and OOC in either basal (A, except for NOTCH1 and HBEGF in OOC) and parabasal (B) layers. Immunoexpression between basal and parabasal layers within the same lesions was also evaluated in OKC (C), COC (D), OM (E), and OOC (F). Differences were only seen for NOTCH1 and HIF-1α in OKC and ADAM-12 in OM. In OKC, NOTCH1, and HIF-1α were more expressed in parabasal layers, while ADAM-12 in OM showed a higher expression in the basal layer. * p < 0.05; ** p < 0.01; *** p < 0.001.

Discussion
OKC has been investigated by several studies trying to elucidate the mechanisms associated with its intriguing biological and invasive behavior. Recently, the invadopodia-associated proteins (cortactin, MT1-MMP, Tks4, and Tks5) were described in this lesion [7], but what causes the invasive mechanism in this lesion is still unclear. In this study, the role of hypoxia as a microenvironmental factor influencing OKC biology was investigated as a possible cause.
Our results provide evidence of increased expression of NOTCH1, HIF-1α, ADAM-12, and HBEGF in OKC when compared to COC, OM, and OOC. Interestingly, when comparing the basal and parabasal layers of OKC, we observed predominant staining of NOTCH1 and HIF-1α in the epithelial layers near to the cyst lumen. It suggests an increased hypoxic environment when the cells are distant from the oxygen supply provided by the surrounding connective tissue. It links tissue hypoxia, characterized here by higher expression of HIF-1α in parabasal layers than in the basal layer, to biological activity, the overexpression of NOTCH1. These findings support the evidence of tissue hypoxia in the upper suprabasal cells in OKC, wherein the expression pattern resembles the pattern also seen in NOTCH1. Although a cause-effect relationship cannot be determined by the methods used here (i.e., immunohistochemistry), HIF-1α overexpression is known to initiate the NOTCH1 signaling pathway [15], suggesting that a possible biological mechanism is triggered by tissue hypoxia.
OKC and OOC are histologically similar, showing a main difference in the type of keratin layer. In OKC, the keratin layer shows nucleated keratinocytes, while in the OOC, the keratin component is nuclei-free [1]. Prior to OOC being recognized as a different entity, OKC was classified as orthokeratinized or parakeratinized, and the last considered the most aggressive form. Here, the proteins studied and associated with more aggressive behavior were more expressed in OKC than in OOC, except for NOTCH1 and HBEGF in the basal layer that showed no differences.
Activation of NOTCH1 results in its release from the plasma membrane and translocation to the nucleus [24], resulting in the activation of the transcription of several genes. These genes have been previously associated with tumor growth, by either increasing cell proliferation or reducing cell death [25]. Here, we observed nuclear NOTCH1 immunostaining in specific regions of the basal layer of OKC cells. This nuclear staining suggests NOTCH1 nuclear translocation and gene activation. It may result in the transcription of genes related to cell proliferation and invasion [15], possibly by inducing the activation of invadopodia-related genes [12]. A previous study has shown the overexpression of invadopodia-related proteins in the OKC basal layer [7]. Thus, these findings suggest that hypoxia occurs in OKC, which overexpresses NOTCH1. NOTCH1 is found in the nucleus of basal cells, suggesting its activation as a possible biological event influencing OKC invasion. NOTCH1 signaling is known to promote the survival and proliferation of cancer cells, which is potentiated by hypoxia through stabilization of the active form of NOTCH1 by HIF-1α [12]. In this investigation, a higher expression of HIF-1α was identified in OKC when compared to COC, OM, and OOC.
Studies have shown that the nuclear localization of HIF-1α is associated with the formation of invadopodia under hypoxic conditions, and subsequently, more aggressive behavior in some types of tumors [12,13]. It can be hypothesized that a hypoxic environment within the upper layers of OKC results in the activation of HIF-1α and NOTCH1. Together, they would induce the transcription of genes that will further lead to invadopodia formation.
We also observed a higher expression of NOTCH1 and HIF-1α in the parabasal layers in relation to the basal layer in OKC. In parabasal layers, both proteins showed intense nuclear and cytoplasmic immunostaining [24]. HIF-1α is classically known to be transient in the cytoplasm, which is quickly degraded by the proteasome. In this case, the cytoplasmic labelling of HIF-1α was probably a consequence of the proteins responsible for its degradation in cytoplasm having reached their saturation point, resulting in higher cytoplasmic levels [25].
Additionally, increased expression of NOTCH1 and HIF-1α in the parabasal layers may be connected to cellular mechanisms involved in cystogenesis, such as hypoxia and apoptosis [17].
Under hypoxic conditions, the p53 tumor suppressor gene is activated via HIF-1α signaling [26]. It is known that p53 is directly associated with apoptosis and cell proliferation [27], events that are fundamental for cystic formation. Furthermore, the regulation of NOTCH1 has been shown to occur via p53 [27].
Here, the increased expression of NOTCH1 and HIF-1α observed in OKC can trigger the transcription of a myriad of host genes, including the metalloprotease ADAM-12 [12]. The identification of high cytoplasmic levels of ADAM-12 in all epithelial layers in OKC supports this hypothesis. Inhibition of ADAM-12 has previously been shown to reduce HBEGF cleavage and cell migration, while its overexpression is linked to ADAM-12-mediated HBEGF cleavage and EGFR dysregulation [28]. The expression pattern for HBEGF was similar to that of ADAM-12. ADAM-12 probably induces the release of HBEGF in OKC. This could potentially lead to increased proliferation and proteolytic activity of OKC cells.
In addition, ADAM-12 has also been directly implicated in the formation of invadopodia through its cytoplasmic tail and activation of Src proteins kinases. Src proteins participate in the formation and activation of the invadopodia [29], which has the potential to contribute to the locally aggressive behavior of OKC.
In summary, our results suggest that hypoxia occurs more intensively in OKC than in COC, OM, and OOC. In OKC, hypoxia appearred to be more intense in parabasal layers near to the cystic lumen than in the basal layer, as observed by higher HIF-1α expression in these regions of the lesion. A similar pattern of expression within OKC epithelial layers was observed for NOTCH1, suggesting the activation of NOTCH1 by HIF-1α. ADAM-12 and HBEGF were found to be overexpressed in OKC when compared to COC, OM, and OOC, a possible result of downstream activation of the NOTCH1 and HIF-1α signaling pathways and transcription. Together, these results support a biological role for microenvironmental hypoxia in OKC that may contribute to its intriguing biological behavior. Although the HIF-1α expression pattern seen in OKC can be suggestive of tissue hypoxia, the method used in this study cannot determine that the altered expression of the associated proteins studied here were regulated by it. Further mechanistic studies may provide supportive evidence of the role of hypoxia in OKC clinical behavior.