Using the Chick Embryo Model to Examine the Effects of Hypoxia Pre-conditioning of Uveal Melanoma Cells on Tumor Growth and Metastasis

Abstract Purpose Highly dynamic oxygen gradients occur within tumors that can result in a hypoxic response, contributing to tumor progression and metastasis. Evidence in uveal melanoma (UM) suggests an upregulated hypoxia response in some poor prognosis UM characterized by HIF1α signaling. We aimed to investigate the effects of exposure to hypoxia on tumor growth and dissemination in the chick embryo chorioallantoic membrane (CAM) model. Methods UM cell lines (MP41, 92.1, MP46, and OMM1) were grown in two-dimensional culture and pre-exposed to hypoxic (1% O2) conditions for 72 h. The effects of this hypoxia pre-conditioning on cell number and clonogenicity as compared with 21% O2 (“normoxia”) were investigated prior to transplantation of the cells onto the CAM. Nodule-forming efficiency (NFE), nodule size, and the presence/absence of tumor cell dissemination were determined macroscopically and histologically. Results Exposure of UM cell lines to hypoxia upregulated HIF1α expression compared to cells cultured in normoxia. A 72-h pre-exposure to hypoxia significantly reduced cell number and clonogenicity in the MP41 and OMM1 cell lines while it had little effect in 92.1 and MP46 cells. When 72-h hypoxia pre-conditioned cells were grown in three-dimensions on the CAM, a reduction in NFE and nodule size was observed when compared with normoxic UM cells. All nodules were composed of proliferating (Ki-67+) Melan-A + cells and displayed chick blood vessel recruitment. Spread of UM cells into the adjacent CAM was observed; however, dissemination to the chick liver was only seen with 92.1 cells grown under normoxia. Conclusions Hypoxia pre-conditioning does not appear to drive a metastatic phenotype in UM; however, further understanding of how oxygen dynamics within the tumor microenvironment regulates HIF1 signaling is needed to determine whether inhibitors of HIF signaling represent a therapeutic option in metastatic UM.


Introduction
Despite effective treatment of primary uveal melanoma (UM), metastatic spread of UM cells from the eye to the liver is the leading cause of death in patients with UM. 1 Metastases arise predominantly in tumors that have particular clinical, histological and genetic characteristics. Indeed, loss of one copy of chromosome 3 (i.e. monosomy 3), gains in 8q, BAP1 mutations/loss of nuclear BAP1 expression, and a class 2 gene signature are all strongly associated with a high metastatic risk. [2][3][4] Several biomarkers have been associated with the class 2 gene expression signature potentially providing further insight into the functional changes regulating this UM cell dissemination. One of these factors is hypoxia inducible factor 1a (HIF1a), a hallmark of tumor hypoxia. 5 It is well-established that hypoxia generates intratumoral oxygen gradients, contributing to tumor plasticity and heterogeneity, and in turn promoting a more aggressive and metastatic phenotype. 6,7 Cycles of hypoxia-re-oxygenation occur in solid tumors, happening rapidly over minutes or more slowly over days. [8][9][10] In neuroblastoma, pre-conditioning for 3 days with hypoxia demonstrated that the cells do not have to be hypoxic anymore to exhibit pro-metastatic capabilities suggesting retention of a "memory" of prior hypoxic exposure. 11 These hypoxic microenvironments can lead to alterations in tumor vasculature and cancer cell differentiation. This is supported in UM, such that HIF1a expression is observed in spindle cells lining structures associated with vasculogenic mimicry (VM). 12 Furthermore, it is associated with a de-differentiated embryonic-like phenotype in class 2 UM, 5 which is consistent with HIF1a signaling and an upregulated hypoxia response noted in a subset of poor prognosis UM by The Cancer Genome Atlas study. 13 The chick embryo chorioallantoic membrane (CAM) model has been utilized previously for the study of angiogenesis 14 and has more recently established itself as a quick and economical xenograft model for cancer research. [15][16][17] We have previously demonstrated and reported the successful utilization of the CAM to grow UM cell lines as threedimensional (3D) tumor nodules over a short period of time. 17 In this study, we expand on our previous work, to further investigate the role of hypoxic pre-conditioning in UM in the CAM model, and to understand how this affects tumor growth and spread.

Western blot
UM cell lines were incubated for varying times up to 96 h in normoxic and hypoxic conditions. HIF-1a accumulates under hypoxic conditions and on re-oxygenation is rapidly degraded. Degradation is mediated through hydroxylation of prolines 402 and 564. For this reason, we examined the expression of hydroxylated HIF-1a. Cells were lysed by adding 200 mL RIPA buffer (Merck Life Science UK Limited, Dorset, UK) containing protease inhibitor cocktail (Merck Life Science UK Limited) on ice for 30 min. Lysates were spun at 10 000 g for 5 min at 4 C and the supernatant was retained for western blot. Protein quantification was undertaken using Pierce BCA protein assay kit (Thermo Fisher Scientific) and absorbance was measured using the Spark Multimode Microplate Reader (Tecan, M€ annedorf, Switzerland) at 562 nm. Proteins were separated by SDS-PAGE using Any kD TM Mini-PROTEAN V R TGX pre-cast gels (Bio-Rad, Watford, UK) and transferred using the Trans-Blot Turbo Transfer System (Bio-Rad) onto 0.2 mM PVDF membranes (Bio-Rad). Membranes were blocked using Tris-Buffered Saline, 0.1% Tween (TBS-T) with 5% milk and incubated overnight at 4 C with rabbit anti-Hydroxy-HIF-1a antibody (Cell Signaling Technology, London, UK) diluted 1:1000 with TBS-T, 5% milk. Membranes were washed and incubated with goat anti-rabbit HRP antibody (DAKO, Agilent, Cheshire, UK) for 1 h. Membranes were further washed and incubated with SuperSignal West Pico PLUS (Thermo Fisher Scientific) HRP substrate and bands visualized with the GeneGnome XRQ imaging system (SYNGENE, Cambridge, UK). b-actin 1:5000 (Merck Life Science UK Limited) in TBS-T, 5% milk and a goat anti-mouse HRP secondary antibody (DAKO) was used as loading control. MP41 cells treated with the proteasome inhibitor MG132 were used as a positive control.
Sulforhodamine B assay of cell number UM cell lines were incubated for 72 h in normoxic and hypoxic conditions before harvesting and plating into flat bottom 96-well plates (Nunclon delta-treated; Thermo Fisher Scientific) at densities of either 4000 (MP41, 92.1, OMM1) or 8000 cells (MP46) per 100 mL/well. After plating, cells were incubated for up to 96 h in normoxic conditions. At 24-h time points, media was discarded and cells were fixed by adding 100 mL 10% Trichloroacetic acid (TCA) (Merck Life Science UK Limited) to each well and incubating at 4 C for 1 h. TCA was discarded by applying running tap water and plates were allowed to air dry. 100 mL of 0.4% Sulforhodamine B (SRB) (Merck Life Science UK Limited) in 1% acetic acid was then added to each well and the plates incubated at room temperature for 1 h. SRB solution was discarded and wells underwent several washes with 1% acetic acid to remove excess dye before leaving the plates to dry. Stained cells were solubilized by adding 10 mM Tris buffer and rocking for 10 min. Absorbance was measured at 595 nm using Spark Multimode Microplate Reader (Tecan), and background absorbance at 690 nm was subtracted from these readings.

Clonogenic assay
UM cell lines were incubated for 72 h in normoxic and hypoxic conditions before harvesting and plating into six-well plates (Nunclon delta-treated; Thermo Fisher Scientific) at a cell density of 500 cells/well in 2 mL of media. Plates were incubated in normoxic conditions until colonies of !50 cells were visible (14-16 days). At this time point, both normoxic and hypoxic conditioned UM cells were fixed and stained with 1% Gentian violet (Merck Life Science UK Limited) in 70% ethanol at room temperature for 10 min. Gentian violet was discarded and excess dye removed by submerging plates in water. Colonies of !50 cells were counted using a stereomicroscope.

CAM implantation
CAM implantation was carried out as previously described. 17 In brief, fertilized chicken eggs were incubated at 37 C and then "windowed" at embryonic day 3 (E3). UM cell lines incubated in either normoxic or hypoxic conditions for 72 h were harvested at E7, counted and spun to produce 10 mL aliquots containing 2 Â 10 6 cells in media for implantation onto the CAM. Previously windowed eggs were opened and the surface of the CAM was dried by applying sections of sterilized lens cleaning tissue (Thermo Fisher Scientific). The "wet" cell pellets were gently pipetted onto the surface of the CAM, eggs were resealed and incubated for 7 days until E14. At E14 eggs were opened, examined using a M165 FC dissecting microscope (Leica Microsystems, Wetzlar, Germany) and GFP-expressing tumor nodules present on or beneath the surface of the CAM were photographed and measured. Embryos were then dissected for internal examination of metastasis. Nodules present on the CAM and livers of all embryos with CAM nodules were fixed in 10% Neutral Buffered Formalin (NBF) for subsequent processing, embedding and histological examination.

Statistics
Statistical analyses were performed using SPSS, version 27.0 (SPSS, Chicago, USA) and GraphPad Prism (GraphPad, San Diego, CA). Statistical significance was considered at p < .05.

Results
In our studies, UM cell lines exposed to 1% O2 for 72 h demonstrated expression of hydroxylated HIF1a, which was absent in cells cultured under normoxic conditions (Supplementary Figure 1). All experiments were thus performed with 72-h normoxic and hypoxic pre-incubated cells.

Pre-conditioning of cells in hypoxic conditions reduces cell number and colony formation
Three UM cell lines, MP41, MP46, and OMM1, demonstrated a significant reduction (p < .05) in cell number following a 72-h pre-exposure to 1% O 2 and then growth in normoxia for up to 96 h as compared with cells cultured in normoxia alone. This was greatest for OMM1 and MP41 cells with 60% and 38% reduction at 96 h respectively. While for 92.1 cells, pre-exposure to hypoxia had no significant effect on cell number as compared with normoxia alone (Figure 1).
Hypoxia has previously been reported to re-programme cancer cells toward a cancer stem-like cell (CSC) phenotype. 21 To investigate whether pre-exposure to hypoxia may select for increased CSC populations in UM, we investigated clonogenicity of the UM cell lines following a 72-h pre-exposure to 1% O 2 and then growth in normoxia, as compared with cells cultured in normoxia alone. Hypoxia significantly reduced clonogenicity in OMM1 (p < .0001) and MP41 (p < .0162) cells only (Figure 2).
Pre-conditioning of cells in hypoxic conditions reduces nodule-forming efficiency and nodule size when UM cells are implanted on the CAM Previous studies of neuroblastoma cells grown on the chick embryo CAM demonstrated that the metastatic phenotype can be controlled by neuroblastoma cell hypoxic pre-conditioning (3 days at 1% O 2 ). 11 Given the effects of hypoxic pre-conditioning on colony forming efficiency and cell number , we investigated the metastatic phenotype of the UM cell lines when pre-exposed to hypoxic or normoxic conditions before transplantation onto the CAM. UM cell lines demonstrated variable tumor nodule-forming efficiencies (NFE) when transplanted onto the CAM as has been reported for cell lines of other cancer types (reviewed in Ref. 22 ) Under normoxic conditions, NFE was highest for the 92.1 cell line (55%) and lowest for the MP46 cell line (28%) ( Table 1). This was affected by a 72-h pre-exposure of the cells to 1% O 2 resulting in a reduction of NFE by around 50% in 92.1, OMM1 and MP41 UM cell lines. No nodules formed with MP46 cells following 72-h pre-exposure of the cells to 1% O 2 . However, the reduction in NFE by hypoxia failed to reach statistical significance (p < .05) in all UM cell lines examined (Table 1). With regard to tumor nodule size formed by normoxic UM cells on the CAM over the 7-day period, MP41 cells formed the largest nodules and OMM1 cells formed the smallest nodules (Table 1). For 92.1 and MP41 cells, when compared with normoxia, nodule size was reduced following 72-h pre-exposure of the cells to 1% O 2 , although this was not statistically significant likely due to the small number of nodules formed overall. OMM1 cells showed little change in nodule size following 72-h preexposure of the cells to 1% O 2 as compared with normoxia.

Macroscopic and histological features of UM nodules growing on the CAM
All UM nodules formed on the CAM stained positively with Melan-A and Ki67 demonstrating actively growing tumor cells (Figure 3). aSMA positivity indicated recruitment of chick blood vessels into the tumor nodule, although metastatic spread to the chick liver was only observed in a single xenograft of 92.1 cells grown under normoxic conditions and transplanted on the chick embryo CAM (Figure 4). Although tumor NFE and nodule size were reduced following pre-conditioning of UM cells in hypoxia, tumor nodules were similarly positive for Melan-A, Ki67 and retained recruitment of aSMA positive chick blood vessels. In a number of tumor nodules either grown in normoxic or hypoxic conditions, spread of the UM cells into the adjacent CAM tissues, a short distance from the xenograft site, could be seen (Figure 3(B,C)).

Discussion
In this study we have shown that pre-conditioning of UM cells with hypoxia for 72 h reduces cell number and clonogenicity in 2D culture, and that this is consistent with a reduced capacity of these cells to form tumor nodules on the CAM. There was no evidence in this study that hypoxia drives a metastatic phenotype. We provide further evidence to support the use of the chick embryo CAM model for the growth of 3D UM nodules.
Tumors experience a highly dynamic range of oxygen gradients. 23 As tumors grow in size, diffusion distances increase and there is localized outstripping of the blood supply, resulting in a mixed population of cells at differing oxygen tensions at any one time. 23 Hypoxia therefore, has been broadly separated into three types namely acute, chronic and cycling. In chronic hypoxia, cells experience low oxygen tensions for prolonged periods of time as tumors grow and outstrip the local blood supply while acute (short periods of time) and cycling (intermittent hypoxia) hypoxia occur as the tumor adapts to ongoing changes in the microenvironment, resulting in cellular responses, such as accumulation of reactive oxygen species, changes in metabolism, radioresistance and spontaneous metastasis. 10,24 Transcriptional responses to hypoxia are primarily mediated by HIF with expression of the HIF1a subunit increasing dramatically when cells are exposed to low oxygen levels. 25 Previous studies have demonstrated that hypoxic pre-conditioning of neuroblastoma cells for 3 days drives a change to a more aggressive phenotype that is mediated by HIF1a and continues following cellular re-oxygenation. 11 High levels of HIFa have been observed in UM tissue 26,27 and, indeed, increased HIF-1a expression was shown to be significantly associated with proliferative and vascular markers; although there was no correlation between HIF-1a expression levels and patient survival. 26,27 Importantly in these studies, HIF1a expression was observed throughout the whole tumor and not necessarily restricted to hypoxic areas. In the study by Jha et al. HIF1a expression was significantly associated with highly pigmented UM cases. 26 It has been reported that melanin serves as a scavenger of oxygen and creates hypoxic conditions during its synthesis. 28 This has led to the suggestion by several authors that an oxygen-independent mechanism also regulates HIF1a expression, which is consistent with reports of HIF1a expression in UM cell lines grown under both normoxic and hypoxic conditions. 27,29 In our study, however, HIFa expression was highest in UM cell lines cultured at 1% O 2 as compared with cells grown at 21% O 2 , consistent with data from Hu et al. who also reported increased HIF1a when 92.1 and OMM1 cells were exposed to hypoxia for 24 h as compared with cells in normoxia. 30 We further demonstrated that OMM1 and MP41 cells showed the highest sensitivity in culture to hypoxic pre-exposure for 72 h with significant decreases in cell number and clonogenicity compared with cells grown in normoxia. El Filali et al. similarly demonstrated variable effects of hypoxia on proliferation in six UM cell lines exposed for up to 11 days. 31 Of note in these experiments was a significant reduction in the proliferation of 92.1 cells grown in hypoxia as compared with no effect in our studies using this cell line. This is likely due to chronic exposure to hypoxia in the studies by El Filali 31 compared with acute hypoxic pre-conditioning in our experiments again highlighting the importance of understanding the contributions of acute, chronic and cycling hypoxia to tumor growth and metastasis.  When the UM cell lines were transplanted onto the chick embryo CAM, there was no significant effect of hypoxia on tumor nodule formation and size compared with normoxia for three of the cell lines; the MP46 cell line failed to form tumor nodules following hypoxia pre-conditioning. Importantly, a single metastatic lesion in the chick liver following the incubation of 92.1 cells in normoxia was identified throughout the experiments demonstrating spontaneous metastasis of these cells. We hypothesize that additional spontaneous metastatic lesions could be identified if longer incubation times were possible in this model, as we have previously reported metastatic foci in the chick liver by 92.1 and OMM1 cells when injected directly into the chick circulation. 17 Furthermore, although local invasion of UM cells into the CAM tissue was observed in some of the nodules formed for all UM cell lines, this was not dependent on oxygen concentrations. This is consistent with data to support the functional importance of HIF-1a signaling in promoting the invasive capacity of UM cells in both hypoxia and normoxia. 29 Interestingly, targeting hypoxia-induced gene expression with KCN1, a hypoxia inducible factor-1 (HIF-1) pathway inhibitor, has been shown to reduce tumor size and prolong survival in an orthotopic mouse model of UM using the 92.1 and Mel290 cell lines. 32 At the molecular level, KCN1 antagonized hypoxia-induced expression of two genes, P4HA1 and P4HA2, which regulate collagen maturation and deposition resulting in a disordered collagen VI structure, and reduced UM cell invasion. Similarly, Voropaev et al. demonstrated that shRNA knockdown of CREB or HIF-1 in Mel270 UM subcutaneous xenografts in SCID mice, dramatically decreased UM progression. 33  The use of the chick embryo CAM model is a strength of this study as it allows the UM cells to grow in 3D, with blood vessel recruitment and hence variations in oxygen tensions across the nodule, as compared with cells in 2D culture. A weakness of this study, however, is that it examined UM cell lines only. We have recently demonstrated the importance of the immune tumor microenvironment in metastatic UM, [34][35][36] and interestingly Bronkhorst et al. reported that conditioned media from cultured UM cell lines induced chemotaxis of monocytes, which was independent of the normoxic or hypoxic state. 37 We are currently working with 3D co-cultures of UM cells that incorporate immune infiltrates and other features of the tumor microenvironment of metastatic UM. The ability to xenograft such nodules onto the chick embryo CAM will enable more representative pre-clinical models for drug testing.
In conclusion, the chick embryo model continues to represent a valuable short-term animal model for pre-clinical cancer research. Further understanding of the role of differential hypoxia dynamics and HIF1 signaling in UM is needed to determine whether inhibitors of HIF signaling represent truly viable therapeutic options in this devastating disease.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was funded by a UK-India Education and Research Initiative Grant.