Increased HIF-1α Expression Correlates With Cell Proliferation and Vascular Markers CD31 and VEGF-A in Uveal Melanoma

Abstract

Purpose.: Overexpression of hypoxia inducible factor-1 α (HIF-1α) has been found in several cancers and is thought to correlate with aggressive disease. The purpose of our study was to investigate the influence of HIF-1α on clinical outcome in uveal melanoma (UM) along with proliferative (MIB-1) and vascular (CD31, VEGF-A) markers.

Methods.: A retrospective analysis was carried out on UM tumors from 88 patients. HIF-1α, MIB-1, CD31, and VEGF-A expression, as well as necrosis, were assessed by immunohistochemistry and hematoxylin/eosin on paraffin-embedded UM tumor sections by using a tissue microarray. The bivariate analysis involving HIF-1α expression and clinicopathologic covariates was performed by using the χ² test. The association of clinicopathologic covariates and HIF-1α expression with patient survival was evaluated by using the Kaplan-Meier approach and Cox proportional-hazards regression analysis.

Results.: Among our study population, 56 patients (63.6%) had high levels of HIF-1α expression. High expression of HIF-1α was associated with high expression of MIB-1 (P = 0.04), CD31 (P = 0.03), and VEGF-A (P < 0.0001), as well as necrosis (P = 0.04). However, high HIF-1α expression was not correlated with cell type, largest macroscopic tumor dimension or thickness, anterior margin, pigmentation, or mitotic figures. Patients with high HIF-1α expression did not show a reduced survival when compared to patients with low HIF-1α expression (P = 0.92). Finally, HIF-1α expression was not increased after irradiation.

Conclusions.: An increase in HIF-1α expression was significantly associated with proliferative (MIB-1) and vascular (CD31 and VEGF-A) markers, as well as necrosis, in UM. However, there was no correlation between high HIF-1α expression and patient survival.

Introduction

Angiogenesis is a process whereby new blood vessels are formed and is considered essential for growth and progression of solid malignant tumors. ¹ If the formation of neovessels is insufficient to provide enough oxygen to proliferating tumor cells, tissue hypoxia results. Therefore, it has been estimated that 50% to 60% of solid tumors contain hypoxic and/or anoxic areas resulting from an imbalance between oxygen supply and consumption. ² In cancer, hypoxia has been associated with malignant progression, metastasis, and resistance to radiotherapy. ³  

The hypoxia-inducible factors 1 and 2α (HIF-1α and HIF-2α) are key transcription factors regulating the expression of a variety of genes involved in glycolysis and angiogenesis. HIF-1 is a basic helix-loop-helix-PAS transcription factor that plays an essential role in oxygen homeostasis. ^{4 –6} It is a heterodimer composed of an α (HIF-1α) and a β (HIF-1β) subunit, and HIF-1α is the unique oxygen-regulated subunit that determines HIF-1 activity. ^7,8 HIF-1α transactivates numerous genes, which either increase oxygen availability or allow metabolic adaptation to oxygen deprivation in presence of hypoxic stress. Among these, some are involved in vascular growth and cellular metabolism, such as vascular endothelial growth factors (VEGFs), while others can stimulate cell proliferation or induce apoptosis. ^2,9,10 A recent immunohistochemistry survey using human malignant and normal tissues has shown that the expression of both HIF-1α and HIF-2α is commonly increased in a variety of tumors, including bladder, breast, colon, glial, hepatocellular, ovarian, pancreatic, prostate, and renal tumors. ¹¹ In addition, high HIF-1α expression correlates with poor patient outcome in multiple cancers. ² Collectively, these findings highlight that HIF-1α expression or activation is a common event in cancer and suggest that this transcription factor may play a central role in tumorigenesis. 

Uveal melanoma (UM) is the most common primary intraocular tumor in adults, accounting for 70% of all eye malignancies. More than 90% of UMs involve the choroid. The overall mean age-adjusted incidence for this eye cancer in the United States is 4 to 6 per million. ^12,13 Ocular treatment of UM is termed “radical” if the eyeball is removed by enucleation, or “conservative” if preservation of vision is attempted. In most centers, the first choice of conservative treatment is brachytherapy, which is delivered by using a saucer-shaped radioactive applicator containing iodine-125 or ruthenium-106. Despite successful eradication of the ocular tumor, approximately 45% of all patients develop metastatic disease after 15 years, which is almost invariably fatal. ^12,14 The most important histopathologic and cytogenetic factors predicting metastatic disease are the following: largest tumor diameter (LTD), tumor thickness, ciliary body involvement, extrascleral extension, epithelioid cell type, high mitotic rate, monosomy of chromosome 3, and chromosome 8q gain. ^15–17 Moreover, similar to most solid tumors, angiogenesis is also essential for the growth and progression of UM. Presence of a tumor angiogenic network composed of at least three back-to-back closed vascular loops is a feature strongly associated with death from metastatic disease in UM. ^18,19  

Although the HIF family of transcription factors is upregulated in skin melanoma, and promotes metastases, ^20,21 the role of HIF-1α in UM progression remains unclear. In this study, we investigated by tissue microarray (TMA) whether overexpression of HIF-1α is a predictive parameter in UM, correlating or not with resistance to radiations and other prognostic factors. 

Materials and Methods

This study followed the tenets of the Declaration of Helsinki and was approved by our institutional human experimentation committees. Written informed consent was obtained from the enucleated subjects. 

Tumor Samples and Clinicopathologic Data

Uveal melanoma tumor specimens (ciliary body and choroidal melanomas) were obtained from patients undergoing enucleation at Centre Hospitalier Universitaire (CHU) de Québec (Quebec City, QC, Canada) or CHU de Caen (Caen, France). Patient clinicopathologic data included sex, age, pre-enucleation irradiation, metastasis, and overall survival. The specimens were stained with hematoxylin/eosin (H&E) for histologic evaluation using light microscopy. Cell type was classified by using a modified Callender's classification (spindle, mixed, or epithelioid). ²² When only one large, unequivocal epithelioid cell was seen in approximately five fields at a magnification of ×400, the tumor was considered to be mixed cell type. ²³ The LTD and thickness were both measured by ultrasonography before the enucleation procedure. Anterior margin (i.e., ora serrata and/or ciliary body involvement) was determined by reviewing the slides. The degree of pigmentation was noted as previously described: absent (no pigmentation), minimal (cytologic details evident), moderate (cytologic details partially obscured), or heavy (cytologic details totally obscured). ²² Mitotic figures were counted in 15 high-power fields with a total magnification of ×400 by using an eyepiece grid, and the number of mitoses was averaged. Necrosis was graded as absent (necrosis < 10% of the tumor) or significant (necrosis > 10% of the tumor) ^24,25 (see Supplementary Fig. S1). 

Tissue Microarray Construction

The TMA was built from archived Bouin- or formalin-fixed, paraffin-embedded UM tumors collected from 1988 to 2010. Hematoxylin/eosin-stained tumor sections were used to mark regions with highest nuclear grade, avoiding areas of necrosis. Three tissue cores per tumor (1 mm in diameter) were placed into an empty recipient paraffin block, using a semi-automated tissue arrayer Minicore 3 (Alphelys, Plaisir, France). Cores were spaced at intervals of 1 mm in the x- and y-axes. One section of the TMA block was stained with H&E for morphologic assessment, and additional serial sections were cut for immunohistochemistry. Five-μm-thick sections were mounted on positively charged glass slides, left to dry at 60°C overnight, and stored in the dark at 4°C until staining. 

Immunohistochemistry

Tumor samples were processed for immunohistologic analysis in a Dako Autostainer Plus Link (Dako, Mississauga, ON, Canada) according to the manufacturer's protocol using Envision peroxidase procedure (ChemMate TM Envision HRP detection kit; DakoCytomation, Glostrup, Denmark). Demasking procedure was done by using Dako's TRS buffer at pH 6, except for the MIB-1 antibody (at high pH 7). The antibodies used were HIF-1α (1:1000; Abcam, Toronto, ON, Canada), MIB-1 (prediluted; Dako, Burlington, ON, Canada), CD31 (1:30; Dako), and VEGF-A (1:100; Santa Cruz Biotechnology, Dallas, TX). The Envision Flex with High pH kit (Dako) was used for MIB-1 staining, whereas the LSAB2 System HRP kit (Dako) was used for HIF-1α, CD31, and VEGF-A antibodies. Secondary antibody without primary antibody was used as negative control. Positive controls were performed with renal, splenic, or tonsil lymphoid tissues. 

Immunohistologic Analyses

The immunohistochemical staining results were evaluated in blinded fashion by two experienced investigators (MM and FM) without knowledge of clinicopathologic data on each individual case. Five high-power fields (magnification ×200) were selected randomly and 200 cells were counted per field. As reported by others, HIF-1α was scored according to the presence of nuclear and cytoplasmic staining. ^26,27 Percentage of positive cells was calculated and mean value of the three tissues cores was used. Cells were graded as negative (0), weakly positive (1%–15%), moderately positive (15%–50%), or strongly positive (>50%). Because the level of HIF-1α required to initiate transcription is currently unknown, as well as the limited number of patients, HIF-1α and VEGF-A immunoreactivity was classified in two grades: low (negative, weak, or moderate; <50%) or high (strong; >50%). ²⁸ MIB-1 and CD31 immunoreactivity was also divided in two grades: low (negative or weak; <15%) or high (moderate or strong; >15%) (for examples of grading see Supplementary Fig. S2). 

Statistical Analysis

Statistical analysis was performed with SAS software (version 9.4; SAS Institute, Cary, NC). Spearman test was used to assess the significance of correlation with clinical and histopathologic parameters. The bivariate analysis involving HIF-1α expression and clinicopathologic covariables was performed by using the χ² test. Association of clinicopathologic variables with patient survival was evaluated by using Kaplan-Meier approach and Cox proportional-hazards regression analysis. Association of HIF-1α staining with patient survival was evaluated by using Kaplan-Meier curves. Differences between groups were tested by the log-rank test. Statistical differences with P value less than 0.05 were considered significant. Cox proportional-hazards regression analysis was used to assess the impact of HIF-1α levels on UM mortality. Patients who died of causes unrelated to UM were censored at the time of death. 

Results

Clinical and Histopathologic Characteristics

Eighty-eight patients were included in this study (Table 1), 50 men and 38 women, whose ages ranged from 14 to 91 years, with a mean of 63.2 ± 13.7 years. Mean LTD and tumor thickness measured by ultrasonography were 15.3 ± 3.7 mm and 9.5 ± 3.8 mm, respectively. The mean follow-up was 66.5 ± 52.6 months, with 54 (61.3%) and 38 (43.2%) patients followed up for more than 3 and 5 years, respectively. Liver metastasis was observed in 41 patients (46.6%). Cell type classification determined 51 epithelioid tumors (58.0%), 17 spindle tumors (19.3%), and 20 mixed tumors (22.7%). Anterior margin was observed in 39 cases (44.3%). Mean number of mitoses was measured as 3 ± 2.6. Pigmentation was graded as absent in 8 cases (9.1%), minimal in 47 cases (53.4%), moderate in 22 cases (25.0%), and heavy in 11 cases (12.5%). Necrosis was graded as absent in 73 cases (83.0%) and minimal or significant in 15 cases (17.0%). 

Table 1

 

Association of Clinicopathologic Covariates and Immunohistochemistry of CD31, MIB-1, and VEGF-A With HIF-1α Expression

	Low HIF-1α Expression, n (%)	High HIF-1α Expression, n (%)	P Value
Age, y			0.51
≤55	8 (25)	17 (30)
>55–69	13 (41)	16 (29)
>69	11 (34)	23 (41)
LTD, mm			0.8
≤10	2 (6)	5 (9)
>10–15	13 (41)	26 (45)
>15	17 (53)	26 (46)
Thickness, mm			0.46
≤3	3 (9)	2 (4)
>3–8	8 (25)	18 (32)
>8	21 (66)	36 (64)
Margin			0.09
Posterior to the equator	14 (44)	35 (63)
Anterior	18 (56)	21 (37)
Cell type			0.76
Spindle	5 (16)	12 (22)
Epithelioid	20 (62)	31 (55)
Mixed	7 (22)	13 (23)
Pigmentation			0.07
Absent or minimal	16 (50)	39 (70)
Moderate and heavy	16 (50)	17 (30)
Mitotic activity			0.56
0–1	10 (31)	18 (32)
2–3	9 (28)	21 (38)
≥4	13 (41)	17 (30)
Necrosis			0.04
Absent	23 (72)	50 (89)
Significant	9 (28)	6 (11)
CD31 expression			0.03
Low	27 (84)	35 (62)
High	5 (16)	21 (38)
MIB-1 expression			0.04
Low	30 (94)	43 (77)
High	2 (6)	13 (23)
VEGF-A expression			<0.0001
Low	24 (75)	15 (27)
High	8 (25)	41 (73)

Using Spearman analysis, we found that mitotic activity positively correlated with LTD (r = 0.27; P = 0.01), tumor thickness (r = 0.43; P < 0.001), and metastasis (r = 0.35; P = 0.0009). We determined that anterior margin positively correlated with LTD (r = 0.48; P < 0.0001), tumor thickness (r = 0.50; P < 0.0001), and metastasis (r = 0.27; P = 0.01). Using the Kaplan-Meier approach, we found that LTD, tumor thickness, anterior margin, and mitotic activity were the most significant prognostic factors of survival (data not shown). 

The degree of necrosis was positively correlated with tumor thickness (r = 0.28; P = 0.01), mitotic activity (r = 0.33; P = 0.001), anterior margin (r = 0.39; P = 0.0002), and metastasis (r = 0.37; P = 0.0005). Using the Kaplan-Meier approach, we determined that necrosis was a significant predictive factor of survival (P < 0.0001 [log-rank test]): tumors with necrosis had worse prognosis than tumors without necrosis (hazard ratio [HR] = 4.56; P < 0.0001). 

HIF-1α Immunoreactivity and Time-Independent Outcome Analysis

Among our study population, 56 patients (63.6%) showed high expression of HIF-1α. When present in tumor tissues, HIF-1α staining pattern was mixed nuclear/cytoplasmic (Fig. 1A). However, we did not observe specific expression of HIF-1α in close proximity to blood vessels, but rather in the rim of necrotic areas. High HIF-1α expression was associated with moderate and strong immunoreactivity of VEGF-A (P < 0.0001; Fig. 1B), MIB-1 (P = 0.04; Fig. 1C), and CD31 (P = 0.03; Fig. 1D; Table 1). For the 56 patients with high HIF-1α expression, six tumors (11%) had minimal or significant necrosis, compared to nine tumors (28%) among the 32 patients with low HIF-1α expression (P = 0.04) (Table 1). 

Figure 1

 

Immunostaining for HIF-1α, VEGF-A, MIB-1, and CD31 in primary uveal melanomas. (A) Epithelioid uveal melanoma showing strong mixed nuclear and cytoplasmic HIF-1α staining. (B) Epithelioid uveal melanoma with necrosis showing strong cytoplasmic VEGF-A staining. (C) Mixed uveal melanoma showing moderate nuclear MIB-1 (Ki-67) staining. (D) Mixed uveal melanoma showing strong cytoplasmic and paranuclear CD31 staining. Negative controls are in each figure in the top right. Dark brown indicates positive staining. Scale bar: 200 μm.

HIF-1α Time-Dependent Survival Analysis

Similar proportions of patients developed metastasis among those showing high HIF-1α expression (48.2%, 27/56) and low HIF-1α expression (43.8%, 14/32). Survival of patients was compared for high and low HIF-1α expression (Fig. 2). Fifteen patients among our study population were lost to follow-up or died of other causes (17%, 72.3 ± 65.3 months): 11/56 with high HIF-1α expression (19.3%, 68.3 ± 48.9 months) and 4/32 with low HIF-1α expression (12.5%, 83.3 ± 108.1 months). Patients with high HIF-1α expression did not show a significantly reduced survival when compared to patients with low HIF-1α expression (HR = 0.97; P = 0.92 [log-rank test]). 

Figure 2

 

Kaplan-Meier survival curves for primary uveal melanoma tumors with low or high expression of HIF-1α.

Irradiated Versus Nonirradiated Tumors

Enucleation was performed for neovascular glaucoma in 7 cases and for recurrence in 10 cases. We compared necrosis and mitotic activity between nonirradiated and irradiated tumors. We found less mitotic activity in irradiated tumors than nonirradiated tumors (P < 0.05, Table 2). Because previous studies have indicated that irradiation results in a re-oxygenation–dependent increase in HIF-1α activity, we analyzed HIF-1α in irradiated tumors. However, HIF-1α expression was not increased after radiotherapy (Table 2). 

Table 2

 

Analysis of Necrosis, Mitotic Activity, and HIF-1α Immunostaining in Irradiated and Nonirradiated Tumors

	Irradiated, n = 17	Nonirradiated, n = 71	P Value*
Necrosis, absent/significant	15/2	58/13	0.73
Mitotic activity, 0/≥1	9/8	19/52	0.04
HIF-1 α immunostaining, low/high	7/10	25/46	0.78

*  Fisher's exact test.

Discussion

Few reports have studied HIFs expression in UM. Using immunohistochemistry on 50 UM samples, Xu et al. ²⁹ have previously shown that HIF-1α expression is related to tumor size and pathologic types. However, this study did not investigate the link between HIF-1α expression and patient survival because of a lack of follow-up data. Another immunohistochemistry survey on 65 UM tumors has demonstrated a correlation between HIF-1α staining and the class 2 gene expression signature of tumors with high metastatic risk. ³⁰ The other articles studying HIF-1α in UM have focused on cell lines. ^31–34 Our study assessed HIF-1α expression in UM primary tumors and explored the association between its presence and clinicopathologic parameters, including patient survival, as well as expression of proliferative (MIB-1) and vascular (CD31 and VEGF-A) markers. One limitation of our study was the inclusion in the tumor cohort of some samples fixed in Bouin preservative. Unfortunately, we were not able to assess the link between HIF-1α and cytogenetic alterations such as monosomy 3: loss of expression or mutations in the Von Hippel-Lindau gene (VHL) (short arm of chromosome 3) result in constitutive HIF stabilization and activation of HIF-controlled transcriptional programs irrespective of oxygen levels. ³⁵ Another limitation was the use of 1-mm biopsies to build the TMA because the entire tumor could not be analyzed using this technique. We also recognize that our cohort included many large tumors, which are more likely to result in enucleation than small ones. However, using the Kaplan-Meier approach, we found that LTD, tumor thickness, anterior margin, and mitotic activity were the most significant prognostic factors to predict patient survival, suggesting that our cohort was valid. 

A tumor cannot grow larger than 2 mm³ in absence of angiogenesis because the lack of oxygen in the center of the tumor will result in cell apoptosis or necrosis. ²⁶ However, a dense network of newly formed capillaries within a tumor does not necessarily imply that these vessels are fully functional, so the tumor can remain hypoxic. In our study, HIF-1α staining was detected at some distance from tumor blood vessels, reflecting the decrease in oxygen concentration with increasing distance from the capillaries. ³⁶ Contrary to our data, Chang et al. ³⁰ have observed HIF-1α expression confined almost exclusively in spindle-shaped cells lining extracellular matrix patterns and small vascular structures in class 2 tumors. HIF-1α immunoreactivity was also found diffusely throughout the entire tissue in our tumors, in contrast to glioblastoma or endometrial carcinoma where HIF1Aα expression is present around necrotic areas. ^37,38 These latter findings may reflect the existence of alternative oxygen-independent regulatory modes of HIF-1α expression, ³⁶ such as insulin, cytokines, RACK1, and HSP90. ^39–42  

The ability of tumor cells to induce angiogenesis occurs through a multistep process, designated “angiogenic switch,” which ultimately tips the balance toward pro-angiogenic factors. Reduced level of oxygen leads to the stabilization and activation of the transcription factor HIF-1α, which in turn induces the expression of a number of pro-angiogenic factors such as VEGFs. VEGF-A is particularly noteworthy since it is expressed in various cancers including UM. ^31,43,44 Accordingly, we found a significant association between high HIF-1α and VEGF-A expression. CD31 is specific to endothelial cells and its expression underlies the vascular network within a tumor. We showed that high HIF-1α expression was correlated with strong CD31 staining. MIB-1 (also known as Ki-67) is expressed in proliferating cells undergoing DNA synthesis. ⁴⁵ We found a moderate to intense expression of MIB-1 in tumors with high HIF-1α expression, suggesting that hypoxia promotes UM growth. However, the converse could also apply, namely, that increased proliferation rate could induce tumor hypoxia. ³ Thus, a vicious circle is established. 

Necrosis is a common histopathologic feature in many solid tumors and has been proposed as a marker of poor prognosis in renal, breast, lung, and colorectal cancers. ⁴⁶ Necrosis is present at a frequency between 1.1% and 55.8% in UMs depending of the tumor size. ²² In our study, the degree of necrosis was positively correlated with tumor thickness, mitotic activity, anterior margin, and metastasis. Moreover, we found that necrosis was a significant prognostic factor of patient survival, with necrotic tumors associated with worse prognosis than tumors without necrosis. The origin of necrosis is unknown, but immunity may play a role in this process. ^47–49 It is likely that necrosis reflects the presence of hypoxic areas with increased HIF-1α activity. Therefore, HIF-1α expression has been found to be a prognostic biomarker in patients with necrotic astrocytic and invasive bladder tumors. ^50,51 In glioblastoma multiforme showing VEGF upregulation, HIF-1α has been detected in tumor cells closest to necrotic areas, and farthest from blood vessels. ^52,53 The probability of invasion, metastasis, and death are positively correlated with the degree of intratumoral hypoxia in various cancers. ^3,54 However, we did not observe a reduced survival in patients with high HIF-1α expression when compared to patients with low HIF-1α in our cohort. Similar results have been described with a cohort of 89 patients with skin melanoma. ⁵⁵ Our negative result can be explained by the limited number of patients with a follow-up longer than 5 years (43%). Moreover, it is becoming increasingly apparent that HIFs are one common link between hypoxia, chronic inflammation, and tumor progression through their functions in macrophages during cancer development. ^56,57 Because there is a strong implication for inflammation in UM, high HIF expression in our study may be also induced by inflammatory mechanisms. ⁴⁷  

Radiotherapy is a major treatment modality for UM and requires free radicals from oxygen to destroy target cells. It has been previously suggested that hypoxia can compromise the beneficial effects of ionizing radiation, ^58,59 and that cells found in hypoxic areas are resistant to radiation-induced cell death. ^3,60 Previous studies have indicated that irradiation results in a re-oxygenation–dependent increase in HIF-1α activity. ² In addition, HIF-1α stabilization promotes tumor vasculature radioresistance through the release of pro-angiogenic cytokines such as VEGF. ⁶¹ In contrast, HIF-1α can also induce tumor radiosensitivity through the induction of apoptosis. ⁶¹ However, no increase of HIF-1α expression was observed after irradiation in our tumors. 

In conclusion, we observed a strong HIF-1α immunoreactivity in 64% of UM tumors, suggesting that hypoxia may play a role in the progression of this cancer. However, we did not find a significant correlation between high HIF-1α expression and reduced patient survival. Perhaps, the colonization of metastatic sites involved the activation of other hypoxia-related factors such as HIF-2α in UM cells. In certain situations, HIF-1α drives the initial response to hypoxia, while HIF-2α plays a major role during chronic hypoxic exposure. ⁶² For example, in kidney cancer both isoforms have opposing properties on tumor growth. ⁶³ Further experimental work is warranted to explore the roles of HIF transcription factors in UM progression and their potential as targets in novel therapies. 

Supplementary Materials

pdf S1 

Acknowledgments

The authors thank Marcelle Giasson and Marthe Mercier for clinical data collection and management. The authors thank Hélène Gingras and Michèle Orain for technical assistance for immunohistochemistry. 

Supported by the Réseau de recherche en santé de la vision from the Fonds de recherche du Québec – Santé (FRQS) (financial support to The Québec Uveal Melanoma Infrastructure). 

Disclosure: F. Mouriaux, None; F. Sanschagrin, None; C. Diorio, None; S. Landreville, None; F. Comoz, None; E. Petit, None; M. Bernaudin, None; A.P. Rousseau, None; D. Bergeron, None; M. Morcos, None