This article is mentioned in:

Introduction

Biochemical failure and local recurrences of prostate cancer after radiotherapy could be reduced by escalation of the radiation dose to the prostate (1,2). As local recurrences frequently originate at the site of the macroscopic tumor, the focus for further dose escalation has to be on focal tumor areas most in need of aggressive treatment (3–5).

Based on recent studies, we know that hypoxia-inducible factor (HIF) may be related to prostate cancer outcome, as higher HIF expression was found to be associated with increased risk of biochemical failure and with reduced survival (6–8). Because of this relationship, tumor areas with high HIF expression may be candidates for focal boost.

HIF is a transcription factor which can be regulated by hypoxia-dependent and hypoxia-independent mechanisms. The hypoxia-dependent mechanism has been the most investigated; levels of HIFα subunits in the cell increase under hypoxic conditions resulting from an imbalance between oxygen demand and oxygen delivery to the tissue. In this pathway, HIF transcription factors act as key regulators of the cellular response to hypoxia and modulate the expression of genes involved in processes such as cell metabolism, proliferation and angiogenesis. The hypoxia-independent pathway of HIF-regulation is the induction of HIFα expression by increased oncogenic signaling in cancer cells (9). Most studies have focused on the presence of HIF-1α. However, HIF-2α may play an additional and important role in prostate cancer (8,9). The role of HIF-2α in prostate cancer is still underexposed and its distribution in whole-mount prostate histology not yet described. This is the first study describing the distribution of HIF-2α expression in whole-mount prostate histology.

For the non-invasive detection of prostate tumor areas, dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is one of the tools currently used in clinical practice (10). The aim of the present study was to investigate the relation of the DCE-MRI parameters Ktrans and kep with the immunohistochemical expression of HIF-1α and HIF-2α in whole-mount prostate histology.

Materials and methods

Patients

The study was approved by the Institutional Review Board of the University Medical Center Utrecht, and informed consent was obtained from all patients. Twenty patients were enrolled in the study. All patients had biopsy-confirmed prostate cancer and were scheduled for a robotic-assisted laparoscopic prostatectomy (RALP). None of the patients was treated with androgen deprivation therapy. Five patients were excluded from analysis, due to the unavailability of pathological material in 2 patients, considerable motion during MR image acquisition in 1 patient, and in 2 patients the prostate specimen was severely deformed which hampered accurate registration. Clinical characteristics of the 15 eligible patients are shown in Table I.

Table I Patient characteristics.

Table I

Patient characteristics.

Characteristics	Value
Patient no.	15
Patient age (years)a	63 (48–74)
PSA level (ng/ml)a	12 (5–29)
Pathological tumor stageb
T2a	2
T2c	5
T3a	8
Gleason scoreb
3+3=6	2
3+4=7	10
3+5=8	3
Tumor no.	34
<0.22 cm3	14
>0.22 cm3	20
Tumor volumea	0.73 (0.01–9.36)
<0.22 cm3	0.17 (0.01–0.21)
>0.22 cm3	1.59 (0.42–9.36)

a Median and (range).

b Patient numbers.

MR imaging

Multi-parametric imaging was performed on a 3T MRI-scanner (Philips Achieva, Best, The Netherlands), including a T2w sequence and DCE-MRI. No endorectal coil was used. T2w images were acquired with a fast spin-echo sequence, TR/TE=8396/120 msec, echo train length 13, acquisition matrix 256×256, field of view (FOV)=20 cm, slice thickness 3 mm, slice gap 1 mm. The DCE-MRI protocol consisted of a 3D spoiled gradient echo sequence (20 transverse slices, 2.5-mm slice thickness, TR/TE=4.0/1.7 msec, flip angle 8°, FOV=40 cm, acquisition matrix 160×128). Scans were repeated 120 times at 2.5-sec intervals. A concentration of 0.1 ml/kg of gadobutrol (1.0 M) (Gadovist®, Schering AG, Berlin, Germany) contrast was injected at 2 ml/sec, followed by a saline flush. The tracer kinetics data were analyzed with the Tofts model, as previously described (11). This yielded 3D absolute Ktrans and kep parameter maps. A generic arterial input function (AIF) was used for all patients, due to measurement errors on the patient-specific input functions. To account for patient-specific differences in AIF, DCE-MRI parameters were normalized to the median value in normal PZ for correlation analyses (12).

MRI-pathology registration

After prostatectomy, the prostate specimens were processed and registered as previously described (12). In short, the prostate was cut into 3-mm thick slices, and whole-mount microscopic slices stained with hematoxylin and eosin (H&E) were created. Tumor areas were delineated by a pathologist and assigned with a Gleason score. The registration process was performed in three steps: i) reconstruction of the prostate from 3-mm macroscopic slices, ii) registration of microscopic slices to the reconstructed prostate, and iii) registration between the reconstructed prostate and the T2w MR images. On average, the registration error between MR images and histopathological data was approximately one voxel (reconstructed voxel size 2.5×2.5×2.5 mm3) (12).

Immunohistochemistry

After deparaffination and rehydration, endogenous peroxidase was blocked and antigen retrieval was performed. For microvessel density analysis, slides were incubated for 1 h at room temperature with a monoclonal mouse anti-human CD31 antibody (1:50) (Leica Microsystems, Wetzlar, Germany). For HIF-1α and HIF-2α, slides were incubated overnight at 4°C using a monoclonal mouse anti-human HIF-1α antibody (1:50) (BD Transduction Laboratories, Lexington, KY, USA) and a polyclonal rabbit anti-human HIF-2α antibody (1:700) (Novus Biologicals, Littleton, CO, USA), respectively. Subsequently, slides were incubated with a secondary antibody (CD31 and HIF-1α: Novolink Polymer Detection System; Leica Microsystems and HIF-2α: BrightVision; Immunologic, Duiven, The Netherlands). All slides were developed with diaminobenzidine followed by hematoxylin counterstaining.

Tissue evaluation

Three whole-mount microscopic slides per patient containing normal tissue, the index tumor and probable secondary tumors, were chosen for further analysis. CD31-stained slides were digitized at microscopic resolution and analyzed in ImageScope v10.0 using the Microvessel Analysis Algorithm v1 (Aperio Technologies, Vista, CA, USA) to assess the microvessel density (MVD). To compensate for variation in staining intensity, the settings of the algorithm were evaluated qualitatively and optimized for every batch by adjusting on test regions.

Scoring of HIF-1α and HIF-2α was performed by consensus of two observers. The percentage of positively stained nuclei in whole tumor regions was estimated per slide, ignoring diffuse cytoplasmic staining.

Regions of interest (ROIs) were defined based on the HIF-expression pattern. Areas within a tumor with a distinct HIF-expression pattern were investigated as separate regions, which could result in more regions of interest within a single tumor. This was carried out separately for HIF-1α and HIF-2α.

Statistical analysis

Median Ktrans, kep and MVD were calculated for each ROI. For analysis of Ktrans, kep and MVD, ROIs were selected within tumors with a volume >0.22 cm3 (diameter >3 voxels = 7.5 mm), as partial volume effects and registration errors have a large influence on smaller tumors. Normalized DCE-MRI parameters were used for statistical analysis to account for interpatient variability. Differences between ROIs were assessed with the Mann-Whitney U and Kruskal-Wallis test, and Spearman’s correlation coefficients were calculated (SPSS v20, SPSS Inc., Chicago, IL, USA). P-values ≤0.05 were considered to indicate statistically significant differences.

Results

Of the 15 patients, most showed HIF-1α and HIF-2α-positive tumor areas (Table II). The distribution of HIF expression was highly heterogeneous: i) negative and positive tumors were found within patients, ii) negative and positive areas were found within tumors, and iii) within a single tumor the range in HIF-positivity could be large with 5–35% for HIF-1α and 2–90% for HIF-2α. In Fig. 1, this heterogeneous distribution is illustrated for HIF-2α. Tumor areas showing HIF expression, could either be very small or comprise the entire tumor. There was a marked difference in the distribution of HIF-1α and HIF-2α. HIF-1α expression was often present focally in small regions within tumors, while HIF-2α frequently was expressed in whole tumor regions. Furthermore, the maximum HIF-1α expression was 50%, whereas HIF-2α ranged to 90%. HIF expression was sometimes found in normal tissue, but was not taken into account.

Figure 1 Example of (A) strong HIF-1α staining and (B) strong HIF-2α staining in the same patient. Second patient example showing the heterogeneity of HIF-2α expression within one tumor. (C) Overview of HIF-2α-positive areas with HIF-2α expression of (D) 5%, (E) 10% and (F) 90%. Scale bar, 50 μm.

Table II Distribution of HIF-1α and HIF-2α in patients and tumors.

Table II

Distribution of HIF-1α and HIF-2α in patients and tumors.

		HIF-1α [n/total (%)]	HIF-2α [n/total (%)]
Patients	HIF-positive	12/15 (80)	14/15 (93)
Tumors
Total	HIF-positive	20/34 (59)	28/34 (82)
<0.22 cm3	HIF-positive	7/14 (50)	9/14 (64)
	Maximum HIF-expressiona	20%	35%
>0.22 cm3	HIF-positive	13/20 (65)	19/20 (95)
	Maximum HIF-expressiona	50%	90%

{ label (or @symbol) needed for fn[@id='tfn3-or-29-06-2249'] } HIF, hypoxia-inducible factor.

a Percentage of tumor cells with positive staining.

For all HIF-1α ROIs, median values (range) were: Ktrans 0.241 min−1 (0.108–0.599), kep 0.356 min−1 (0.049–0.581) and MVD 135/mm2(50–230). For HIF-2α these values were: Ktrans 0.256 min−1 (0.109–0.609), kep 0.362 (0.051–0.888) and MVD 115/mm2(55–225). No differences in median Ktrans, kep and MVD were observed between different HIF-expression levels (Fig. 2). In addition, classification into groups according to different HIF-expression levels (e.g. HIF-positive vs. HIF-negative and 0–1% vs. >1%) did not show any significant differences. As shown in Fig. 3, the perfusion parameters and MVD in HIF-negative tumor regions (with 0% HIF expression) overlapped with those in all other HIF-positive ROIs.

Figure 2 Patient example showing the absence of correlations between (A) percentages of HIF-2α expression and (B) Ktrans, (C) kep and (D) MVD. HIF-1α expression in this patient was 0% in all tumor regions.

Figure 3 Graphs indicating the large heterogeneity of Ktrans, kep and MVD for the different HIF-expression levels. Bar, median.

Furthermore, ROIs were classified into groups with high/low Ktrans and high/low kep, using median normalized Ktrans and kep values as threshold. No differences in HIF expression were present between the groups (P>0.16).

Percentages of HIF-1α and HIF-2α expression were not correlated with median Ktrans, kep and MVD (P>0.2). HIF-2α expression was correlated with the Ktrans 5th percentile (r=0.30, P=0.02) and with the MVD 95th percentile (r=−0.39, P=0.002). HIF-1α and HIF-2α were not correlated with each other (P=0.49) and both did not correlate with Gleason score (P>0.77).

Larger tumors (>0.22 cm3) were more often HIF-2α-positive (P=0.023) than smaller tumors and showed higher levels of HIF-2α expression (P=0.041). This association of HIF expression with tumor volume was less clear for HIF-1α (Fig. 4). In addition, larger tumors showed more frequent presence of Gleason grade 4 (P=0.005), but for Gleason grade 5 no difference between smaller and larger tumors was observed (P=0.955).

Figure 4 (A) HIF-1α and (B) HIF-2α expression vs. tumor volume.

Discussion

This is the first study describing the distribution of HIF-2α expression in whole-mount prostate histology. We described the distribution of nuclear HIF-1α and HIF-2α expression in prostate cancer and investigated the relationship with DCE-MRI parameters and MVD. HIF distribution was highly heterogeneous, with pronounced expression of HIF-2α, while HIF-1α was more limited.

To date, the presence of HIF-2α in primary prostate cancer has been described in regards to data of tissue microarrays only (8,13). Boddy et al(13) suggested that HIF-2α may be the most important isoform in prostate cancer, as they showed that HIF-2α expression was negatively correlated with the presence of prolyl hydroxylase enzyme PHD2. Under hypoxic conditions, this enzyme becomes inactive, with subsequent higher levels of HIF-2α. In the present study, we underscore this suggestion, as we found that the expression of HIF-2α was more pronounced and reached higher levels compared with HIF-1α.

In addition, we observed higher HIF-2α expression levels in larger tumors. In light of the prognostic relevance of HIF expression, this suggests that larger tumors are most clinically relevant. These larger tumors have a higher chance to be detected with MRI compared with smaller tumors (14). However, we found that high HIF expression was present as well in small tumor regions. Unfortunately, these small tumors may not be detected with MRI.

Furthermore, we observed a weak positive correlation between HIF-2α and the Ktrans 5th percentile, which indicates that the minimum Ktrans is higher in tumors with higher HIF-2α expression. This may facilitate tumor detection, as higher Ktrans increases the conspicuity of tumor voxels (12).

Despite these encouraging results, we did not observe any differences and correlations between HIF expression and median DCE-MRI parameters or MVD. The question is whether HIF expression in prostate cancer is regulated in an hypoxia-dependent or hypoxia-independent manner. If hypoxia-dependent, HIF induces angiogenesis by upregulation of vascular endothelial growth factor. Therefore, we would expect high MVD and high Ktrans in regions with high HIF expression. However, HIF expression may be induced by several non-hypoxia-related factors (15). In a recent study, Garcia-Parra et al(16) put a critical note on the association between HIF-1α expression and hypoxia. While they found variable HIF-1α staining, they were not able to show hypoxia in prostate cancer with 18F-FAZA PET/CT or with immunohistochemical staining of CAIX which is a downstream target of HIF-1α. Based on these findings, they suggest that the expression of HIF-1α may be independent of tissue hypoxia.

Hypoxia-independence of HIF expression may clarify why we did not observe a correlation between HIF expression and DCE-MRI parameters and MVD.

For both HIF-1α and HIF-2α, a number of studies have shown an inverse relationship between HIF expression and biochemical recurrence (6–8). This prognostic value of HIF expression appears to be independent of the Gleason score. Similar to other reports, we did not find a correlation between expression of HIF-1α or HIF-2α and Gleason score (6,13,16). Therefore, HIF expression in prostate tumor biopsies may be used as input for individual boost dose prescription, to achieve personalized radiotherapy (17).

This study has some limitations. First, the number of patients was relatively small. Definitive conclusions on the relationship between HIF expression and DCE-MRI must be verified in larger studies. Nevertheless, as in this study the P-values found with correlation analyses were far from significance, clinical relevant correlations might be absent as well in larger cohorts. Second, although we used a robust registration method, a registration error existed between DCE-MRI data and pathology.

In conclusion, in the present study we demonstrated that expression of HIF-1α and HIF-2α is not correlated with DCE-MRI parameters. Larger tumors showed frequently high HIF-2α expression, which suggests that larger tumors are most clinically relevant tumors. Given the pronounced expression of HIF-2α and independence of Gleason score, HIF expression may function as a biomarker to guide boost dose prescription.

References