Evaluation of ¹⁸F-EF5 for detection of hypoxia in localized adenocarcinoma of the prostate

Abstract

Background

A common feature of solid tumours that are resistant to therapy is the presence of regions with low oxygen content (i.e., hypoxia). Oxygen electrode studies suggest that localized prostate adenocarcinoma is commonly hypoxic, although conflicting data have been reported between immunohistochemical detection of hypoxia-induced proteins in biopsy specimens and positron emission tomography (PET) imaging of ¹⁸F-labeled hypoxia reporters. Although the 2-nitroimidazole ¹⁸F-EF5 is well-established to label hypoxic tumour cells in pre-clinical tumour models and clinical trials of multiple primary tumour sites, it has yet to be tested in prostate cancer. The purpose of this study was to evaluate the feasibility of using ¹⁸F-EF5 to detect hypoxia in clinical prostate tumours.

Material and methods

Patients with localized adenocarcinoma of the prostate were recruited for pre-treatment ¹⁸F-EF5 PET scans. Immunohistochemistry was conducted on diagnostic biopsies to assess the expression of glucose transporter 1 (GLUT1), osteopontin (OPN), and carbonic anhydrase IX (CAIX). Immunoreactivity scores of staining intensity and frequency were used to indicate the presence of tumour hypoxia.

Results

We found low tumour-to-muscle ratios of ¹⁸F-EF5 uptake that were not consistent with tumour hypoxia, causing early termination of the study. However, we observed GLUT1 and OPN expression in all prostate tumour biopsies, indicating the presence of hypoxia in all tumours.

Conclusion

Our data do not support the use of ¹⁸F-EF5 PET to detect hypoxia in prostate adenocarcinoma, and suggest the use of immunohistochemistry to quantify expression of the hypoxia-inducible proteins GLUT1 and OPN as indications of prostate tumour hypoxia.

Keywords:

• Hypoxia
• EF5
• positron emission tomography
• glucose transporter-1
• osteopontin
• carbonic anhydrase-9

Introduction

Solid tumours often contain regions with low oxygen content (hypoxia). Preclinical studies have linked hypoxia to resistance to radiation therapy [1] and chemotherapeutics [2,3], and mechanistic observations support that hypoxia promotes metastatic growth [4]. Clinical data mirror these reports, with solid tumour hypoxia associating with worse patient outcome regardless of primary tumour origin [1] or treatment strategy [5,6]. Overall, tumour hypoxia is a validated predictor of therapy resistance and tumour aggression across many tumour types, although the most appropriate method to quantify tumour hypoxia is not always clear. The gold standard measurement of tissue oxygenation is direct quantification via an oxygen microelectrode probe within a needle inserted into the tissue. Multiple studies have found that the majority of cancerous prostate tissue exhibits a median oxygen electrode reading below 5 mmHg [7,8]. Cellular responses to hypoxia commonly occur at oxygenation levels below 2% O₂ (∼15 mmHg), and oxygen electrode readings below 5 mmHg are strongly correlated with tumour radiation resistance [1,9]. Indeed, decreased tumour oxygenation is associated with higher tumour stage [5] and poor outcomes in prostate cancer patients [5,7,8]. Thus, localized prostate adenocarcinomas are commonly hypoxic and hypoxia predicts worse outcome, although quantification of hypoxia is not routinely performed in the clinic. Several less invasive methods to quantify tumour hypoxia have been developed over the past several decades, including assessment of hypoxia-induced protein expression in biopsy samples and drugs that can bind in hypoxic cells for subsequent quantification in biopsy tissue or by non-invasive imaging.

Radiolabelled ¹⁸F-fluoromisonidazole (¹⁸F-FMISO), ¹⁸F-fluoroazomycin-arabinoside (¹⁸F-FAZA), and ¹⁸F-[2-(2-nitro-1H-imidazol-1-yl)-N-(2,2,3,3,3-pentafluoropropyl) acetamide] (¹⁸F-EF5), and the copper-based radiotracer ⁶⁴Cu-diacetyl-bis[N4-methylthiosemicarbazone] (⁶⁴Cu-ATSM), have been used to non-invasively measure tumour hypoxia by positron emission tomography (PET). The arguments for and against each radiotracer have been reviewed previously [10–12]. Despite wide testing in breast, lung, head, and neck, and cervical cancers, limited data have been published on PET imaging of hypoxia radiotracers in prostate cancer. Uptake of ¹⁸F-FMISO in a small study of four patients with localized, locally advanced, or metastatic prostate cancer showed a wide range in ¹⁸F-FMISO demarcated fractional hypoxic volume from 0% to 94%, although anatomical computed tomography (CT) or magnetic resonance imaging (MRI) were not used to identify the prostate gland in this study [13]. More recently, Mainta et al. reported significant uptake of ¹⁸F-FMISO in 7 out of 11 high-grade prostate cancer cases by PET/MRI [14], and Supiot et al. used PET/CT to identify significant ¹⁸F-FMISO uptake in one-third of patients with intermediate-risk prostate cancer [15]. However, minimal uptake of ¹⁸F-FAZA has been reported in prostate cancer tissue compared to benign prostate tissue in 14 patients with untreated localized prostate adenocarcinoma, despite significant hypoxia-inducible factor 1α (HIF1α) staining [16]. The authors concluded that ¹⁸F-FAZA PET imaging is not suitable for detecting hypoxic regions in prostate cancer and suggested the degree of hypoxia present may have been below the limit of radiotracer detection. Thus, despite hypoxia being observed by oxygen electrode measurements in prostate tumours [5,7,8] and significant uptake of ¹⁸F-FMISO found in a proportion of prostate cancer patients [13–15], the suitability of other PET-based radiotracers to detect hypoxia in prostate cancer patients is unclear.

EF5 has been extensively used in pre-clinical studies across many tumour types to label and quantify hypoxic tumour cells and has shown promising results in clinical settings with head and neck cancer, including correlation with patient outcome [17]. PET measurement of ¹⁸F-EF5 uptake has been used to quantify tumour hypoxia in lung cancer [18], head and neck cancer [17,19,20], and ovarian cancer [21]. Yapp et al. previously found ¹⁸F-EF5 PET to effectively detect hypoxia in the Shionogi model of prostate cancer, in both the androgen-dependent and independent phases of the model [22,23], but ¹⁸F-EF5 PET has yet to be tested for detecting hypoxia in clinical prostate cancer.

The purpose of the current study was to evaluate the suitability of ¹⁸F-EF5 for the detection of hypoxia in localized prostate adenocarcinoma by PET. We, therefore, compared ¹⁸F-EF5 uptake with immunohistochemistry (IHC) based assessment of the hypoxia-induced proteins glucose transporter-1 (GLUT1), osteopontin (OPN), and carbonic anhydrase-9 (CAIX).

Methods

Patients

Eight men with newly diagnosed localized adenocarcinoma of the prostate were recruited for this study. Between 8 and 12 core needle biopsies were collected from each patient for pathology assessment at diagnosis, with 42–100% of biopsied tissue containing tumour (Table 1). A subset of 2–3 tumour-positive biopsy specimens was selected from each patient for IHC evaluation of GLUT1, OPN, and CAIX staining. Biopsies were selected for IHC analysis based on the following criteria: greatest proportion of tumour tissue in the biopsy, highest total Gleason score within the biopsy, and preference for selecting biopsies from opposite sides of the prostate to account for spatial heterogeneity. The average time between diagnostic biopsy acquisition and ¹⁸F-EF5 PET scan was approximately 3 months. This study was approved by the BC Cancer Research Ethics Board.

Table 1. Patient characteristics, pathology, and ¹⁸F-EF5 measurements.

Patient #	Age (years)	Clinical stage	iPSA (ng/mL)	Prostate volume (cc)	# Biopsies containing tumour/total biopsies	Maximum Gleason Score	Prostate ^18F-EF5 (SUVmax)	Muscle ^18F-EF5 (SUVmax)	Tumour/muscle ratio (SUVmax/SUVmax)
1	73	T2c	8.9	44	6/10	3 + 4	1.5	1.4	1.07
2	61	T4	24	26	8/8	4 + 5	1.8	1.5	1.20
3	75	T2b	6.3	27.7	4/8	3 + 3	2.1	2.2	0.95
4	71	T2c	10	30	12/12	4 + 3	2.4	1.9	1.26
5	72	T2b	7.26	25	8/8	4 + 4	1.9	1.9	1.00
6	65	T1c	5.4	18	4/8	3 + 4	2.1	2.1	1.00
7	71	T2c	1.3	39	7/12	3 + 4	2.4	2.2	1.09
8	74	T2a	8	23	5/12	4 + 3	2.9	2.2	1.32

¹⁸F-EF5 synthesis & imaging

Preparation of ¹⁸F-EF5 was conducted at the TRIUMF particle accelerator centre (University of British Columbia, Vancouver, Canada) with a molar activity of 2.6 ± 0.2 Ci/mmol (95 ± 9 GBq/mmol) via an electrophilic addition reaction using ¹⁸F-F₂ as previously described [24]. Patients were given an intravenous (iv) infusion of 300–374 MBq of ¹⁸F-EF5 as a slow bolus (over 60 s) followed by a 20 mL normal saline flush. Patients received 20 mg of furosemide iv 30 min prior to imaging to void the bladder. Two or three hours after ¹⁸F-EF5 infusion, patients were positioned supine and centred on the scanner bed, with a CT topogram acquired to define the PET scanning field centred on the prostate gland. Non-contrast CT images of the pelvic region were acquired for attenuation correction and localization of PET emission data. Immediately thereafter, PET data were acquired in 3D mode using a Siemens Biograph 16 HI-REZ PET/CT scanner (Siemens Medical Solutions, Knoxville, TN, USA). PET imaging consisted of two fields of view over the pelvis for 4 min per bed position. The voxel size was 1.02 × 1.02 × 1.5 mm (x,y,z), with an axial resolution of 4.2 mm at the radial centre and 5.4 mm at a 10 cm radius. Regions of interest were centred on the prostate gland and a 30 mm diameter region of the gluteus maximus muscle, and used to generate maximum standardized uptake values (SUVmax). A tumour to muscle SUVmax ratio of 1.5 or greater was considered hypoxic based on previous publications in other tumour sites [17,20].

Histology

GLUT1 IHC was carried out at the BC Cancer Centre for Translational and Applied Genomics (Vancouver, Canada). CAIX and OPN IHC was performed by the BC Cancer Molecular and Cellular Immunology Core Facility (Victoria, Canada). Primary antibodies were anti-GLUT1 (Millipore, 07-1401, 1:200), anti-OPN (Abcam, ab69498, 1:250), and anti-CAIX (BioScience Slovakia, AB1001, 1:200). Immunoreactive sites were visualized with DAB chromogen followed by haematoxylin counterstain. GLUT1, OPN, and CAIX were individually scored for both staining intensity and proportion of positive staining within the sections. The intensity was scored as negative (0), weak (1), moderate (2), or strong (3). The frequency of positive tumour tissue was quantified using a scoring system based on percentage of tissue staining positively as no staining (0), <1% (1), 1–10% (2), 11–33% (3), 34–67% (4), or >68% (5). Immunoreactive scores (IRS) were calculated by adding together the staining intensity scores and the frequency scores for each biopsy. IRS from multiple biopsies from the same patient were averaged to produce an average IRS for each tumour.

Statistics

Statistical analyses were conducted using Prism8 software (GraphPad). A non-parametric Kruskal–Wallis test with Dunn’s multiple comparison analysis was used to determine the significance of immunoreactive scores. Agreement between GLUT1, OPN, and CAIX staining was interpreted using the Bland–Altman method [25].

Results

Patients were aged 61–75 years with stage T1c to T4 prostate adenocarcinoma and initial prostate serum antigen (iPSA) concentrations ranging from 1.3 to 24 ng/mL (Table 1). Volume of the prostate on transrectal ultrasound ranged from 18 to 44 cc. Gleason scores ranged from 3 + 3 to 4 + 5 across all diagnostic biopsy specimens used for clinical pathology and IHC analyses.

PET/CT scans were conducted 2 h post-injection of ¹⁸F-EF5. Sample PET/CT images from patient #4 are shown in Figure 1(A), with three image slices through the pelvic region showing radiotracer clearance through the bladder (white arrow) and the largely absent uptake in the prostate (dashed circle). Prostate ¹⁸F-EF5 SUVmax values ranged from 1.5 to 2.9, although SUVmax values for muscle had a similar range (1.4–2.2), producing tumour-to-muscle (T/M) SUVmax ratios of 0.95 to 1.32 (Table 1). While multiple patients had small focal areas of apparent ¹⁸F-EF5 uptake that may be consistent with the focal nature of prostate adenocarcinoma, we observed similar small focal areas of ¹⁸F-EF5 uptake in skeletal muscle and T/M SUVmax ratios were below the commonly used threshold of 1.5 to identify hypoxia. T/M ratios did not correlate with any clinical features including clinical stage, Gleason score, or extent of disease as indicated by prostate volume or the proportion of tumour-containing biopsies (Table 1).

Figure 1. ¹⁸F-EF5 uptake in prostate cancer. (A) Representative ¹⁸F-EF5 PET (left), CT (centre), and merged PET/CT (right) images of three slices from superior (top) to inferior (bottom) through the pelvic region of patient #4 imaged 2 h after ¹⁸F-EF5 injection. ¹⁸F-EF5 signal in the bladder is indicated by the white arrow; prostate regions are indicated by the dashed circles. (B) Representative ¹⁸F-EF5 PET (top), CT (centre), and merged PET/CT (bottom) images from patient #8 taken 2 and 3 h after ¹⁸F-EF5 injection.

We also performed PET/CT scans 3 h after ¹⁸F-EF5 injection in a subset of 4 patients (Table 2). We found that between 2 and 3 h after ¹⁸F-EF5 injection, the SUVmax T/M ratios decreased in two patients, was unchanged in one patient, and increased in one patient, with T/M ratios ranging from 0.84 to 1.56. Patient #8 had a T/M ratio >1.5 and there may be a potential region of hypoxia indicated by ¹⁸F-EF5 uptake in the prostate (Figure 1(B)). However, several focal spots of ¹⁸F-EF5 uptake were again noted in regions outside of the prostate, making it difficult to conclude whether ¹⁸F-EF5 was identifying tumour hypoxia in this patient. It should also be noted this patient did not exhibit tumour characteristics that have previously been associated with prostate tumour hypoxia in some studies [14,26] including higher clinical stage, higher Gleason scores, or greater extent of disease compared to the other patients (Table 1). This study was terminated after insufficient ¹⁸F-EF5 uptake could be identified in the first eight patients recruited, and we then proceeded to test for the presence of tumour hypoxia in these patients based on IHC staining of hypoxia-inducible proteins.

Table 2. ¹⁸F-EF5 uptake measured 2 and 3 h post-injection of radiotracer.

Patient #	Tumour/muscle ratio 2 h (SUVmax/SUVmax)	Tumour/muscle ratio 3 h (SUVmax/SUVmax)
1	1.07	1.0
2	1.20	n.d.
3	0.95	n.d.
4	1.26	n.d.
5	1.00	n.d.
6	1.00	1.0
7	1.09	0.84
8	1.32	1.56

Patient biopsies were stained for GLUT1, OPN, and CAIX [27–30], and positive staining was restricted to cancerous tissue in the biopsies. The majority of GLUT1 staining in prostate cancer tissue was intracellular, with similar staining in the cervical cancer tissue employed as a positive control for hypoxic tumour tissue (Supplemental Figure 1(A–C)). Cytosolic GLUT1 staining in IHC has been previously reported in prostate tumour cells, as well as other adenocarcinomas [31], while intracellular GLUT1 localized to Golgi vesicles has been previously reported in prostate carcinoma cell lines [32]. We found cytosolic OPN staining regions in prostate tumours and the positive control tissue (Supplemental Figure 1(G–I)), and strong membranous CAIX staining in the positive control cervical cancer tissue (Supplemental Figure 1(D)) consistent with the known function of CAIX as a membrane glycoprotein involved in pH regulation [33]. Interestingly, we did not observe CAIX staining in most of the prostate tumour samples and in the few prostate lesions positive for CAIX, we observed cytoplasmic and nuclear staining of CAIX with little membrane localization (Supplemental Figure 1(E–F)).

Immunoreactive scores (IRS) for individual biopsies are summarized in Table 3. We found that 21 biopsies across 8 patients stained positively for GLUT1, with 1/8 lesions displaying weak staining intensity, 6/8 lesions displaying moderate staining intensity, and 1/8 lesions displaying strong staining intensity in at least one biopsy. Analysis of the frequency of GLUT1 staining revealed that more than 67% of the prostate tumour stained positively for GLUT1 in at least one biopsy from all patients. Similar to GLUT1, all eight prostate lesions stained positively for OPN with one lesion displaying weak staining and seven lesions showing moderate OPN staining intensity. OPN positivity ranged from 1 to 10% in one lesion, 34–67% in two lesions, and more than 67% positive staining in five prostate lesions. Conversely, only three out of eight prostate tumours stained positively for CAIX, with two lesions displaying weak CAIX staining intensity and one lesion showing moderate CAIX staining intensity. CAIX positive tissue was 1–10% in two patients and more than 67% in only one patient. CAIX did not co-localize with GLUT1 or OPN staining (Figure 2(A–C)), and combining individual biopsy IRSs to generate a global tumour IRS for each marker indicates that GLUT1 and OPN staining is significantly higher than CAIX across patients (Figure 2(D)). Matched biopsy specimens sectioned and stained for GLUT1, OPN, and CAIX show consistency between GLUT1/OPN staining and mismatch with CAIX expression in seven out of eight prostate lesions in our cohort (Supplemental Figure 2(A–C)). We used the Bland-Altman method to quantitatively assess potential agreement between GLUT1, OPN, and CAIX IRS within individual biopsies (Figure 3). Briefly, data points indicate the difference between IRS for the indicated markers within each biopsy (2–3 per patient), which is plotted against the average of the two IRSs [25]. A ‘mean difference’ (dashed line) close to zero and a narrow 95% confidence interval (dotted lines) indicates strong agreement between the two markers. The mean difference between GLUT1 and OPN IRS was 0.59 (95% CI limits: −1.8 to 3.0), while the mean IRS difference between GLUT1 and CAIX was 5.1 (95% CI limits: 0.25–9.93), and the mean IRS difference between OPN and CAIX was 4.2 (95% CI limits: −0.54 to 9.01). These data indicate substantially greater agreement between GLUT1 and OPN than either marker with CAIX. Together, these data indicate that CAIX does not agree with either OPN or GLUT1 expression and that there is strong agreement between OPN and GLUT1 staining in prostate cancer biopsies consistent with previous work [30].

Figure 2. Immunohistochemical detection of GLUT1, OPN, and CAIX expression in prostate cancer tissue. (A) Representative images of moderate GLUT1 staining, (B) moderate OPN staining, and (C) negative CAIX staining in serial sections of biopsy ‘a’ from patient #5. Boxed regions in the left images are enlarged on the right. Scale bars = 200 µm (left) or 50 µm (right). (D) Immunoreactive scores (IRS) were calculated for each patient by averaging IRS values for GLUT1, OPN, and CAIX IHC staining from 2 to 3 biopsies per patient. Data are mean ± SD (n = 8, *p < 0.05, ***p < 0.001, Kruskal–Wallis test with Dunn’s post-hoc analysis).

Figure 3. Bland–Altman plots comparing similarities between GLUT1, OPN, and CAIX immunoreactive scores. (A) The difference between GLUT1 IRS values and OPN IRS values within individual biopsies are plotted on the y-axis (closer to zero is more similar) with the average GLUT1/OPN IRS value for the biopsy on the x-axis (higher IRS values indicate more intense and frequent staining). Data points represent individual biopsies (n = 17–22). Dashed line indicates the mean difference between the two markers with 95% CI indicated by the dotted lines; a mean difference close to zero indicates strong agreement between the two markers. (B) GLUT1 IRS compared to CAIX IRS. (C) OPN IRS compared to CAIX IRS.

Table 3. GLUT1, OPN, and CAIX IHC scores, Gleason scores, and ¹⁸F-EF5 measurements.

Patient #	Biopsy	GLUT1			OPN			CAIX			Gleason Score	Maximum Gleason Score	Tumour to muscle ^18F-EF5 ratio
		Intensity	Frequency	Immuno-reactive Score	Intensity	Frequency	Immuno-reactive Score	Intensity	Frequency	Immuno-reactive Score
1	a	2	5	7				0	0	0	3 + 4	3 + 4	1.07
	b	1	5	6	1	2	3	0	0	0	3 + 4
2	a	1	5	6	2	4	6	2	1	3	4 + 4	4 + 5	1.2
	b	2	5	7				0	0	0	4 + 4
	c	1	5	6				1	1	2	4 + 5
3	a	2	5	7	1	4	5	0	0	0	3 + 3	3 + 3	0.95
	b	2	5	7	2	5	7	0	0	0	3 + 3
4	a	1	5	6	2	5	7	1	5	6	3 + 4	4 + 3	1.26
	b	1	5	6	1	5	6	1	5	6	3 + 4
	c	1	5	6	1	5	6	1	5	6	3 + 4
5	a	2	3	5	2	4	6	0	0	0	4 + 3	4 + 4	1
	b	2	5	7				0	0	0	4 + 3
	c	1	5	6	2	2	4	0	0	0	3 + 4
6	a	2	3	5	1	3	4	1	1	2	3 + 4	3 + 4	1
	b	3	3	6	1	2	3	0	0	0	3 + 4
	c	2	5	7	2	5	7	0	0	0	3 + 3
7	a	1	5	6	1	5	6	0	0	0	3 + 3	3 + 4	1.09
	b	2	4	6	2	4	6	0	0	0	3 + 4
	c	1	5	6	1	5	6	0	0	0	3 + 3
8	a	2	5	7	1	5	6	0	0	0	4 + 3	4 + 3	1.32
	b	2	5	7	2	5	7	0	0	0	3 + 4

We then compared the GLUT1 and OPN IRS with the T/M ratios for ¹⁸F-EF5 uptake in each patient (Table 3). We found expression of GLUT1 and OPN in biopsies from all prostate tumours imaged by ¹⁸F-EF5 PET, and our dataset does not suggest a relationship between ¹⁸F-EF5 uptake and hypoxia-associated IHC scores. While patient #8 had the highest T/M ratio 2 and 3 h after radiotracer injection, GLUT1 and OPN staining in this tumour was comparable to other patients. Taken together, these data indicate that ¹⁸F-EF5 PET does not adequately detect hypoxia in clinical prostate tumours.

Discussion

Hypoxic regions are a common characteristic of solid tumours. Tumour hypoxia is associated with metastatic dissemination, resistance to therapy, and poor outcome of several tumour types, including prostate cancer [1,3,4]. Early detection of hypoxia in localized prostate adenocarcinoma by non-invasive PET imaging could identify high-risk patients requiring more aggressive treatment approaches, and published evidence suggests ¹⁸F-FMISO can be used to identify hypoxia in prostate tumours by PET [13–15]. The objective of this study was to evaluate the suitability of ¹⁸F-EF5 for the detection of hypoxia in prostate cancer patients. We found minimal tumour ¹⁸F-EF5 uptake despite all prostate tumours scoring positively for two proteins that have been used in the past to indicate hypoxia in prostate tumours, GLUT1 [30,34] and OPN [30,35], suggesting ¹⁸F-EF5 PET is not suitable to identify hypoxia in prostate cancer.

The biodistribution of ¹⁸F-EF5 in patients has been studied previously, showing rapid and uniform distribution throughout the body within less than 15 min of injection [36]. A range of half-lives has been cited for ¹⁸F-EF5 in patients (on the order of 7.5–10.5 h), with some suggestion that shorter half-lives are associated with higher molar activities of ¹⁸F-EF5 (i.e., lower concentrations of ¹⁸F-EF5 for a given 300 MBq injection) [36]. The specific activity of ¹⁸F-EF5 in our study is ∼24-fold higher than these earlier studies [36], but ∼84-fold lower than more recent work showing robust ¹⁸F-EF5 uptake by hypoxic head and neck tumours [19]. Our inability to detect significant ¹⁸F-EF5 uptake in prostate tumours was therefore not limited by the injected dose of ¹⁸F-EF5.

We imaged all tumours 2 or 3 h after ¹⁸F-EF5 injection, and it is possible that the sensitivity of ¹⁸F-EF5 to identify regions of hypoxia may improve with longer time periods allowed for ¹⁸F-EF5 uptake and binding in the tumour. We conducted PET/CT imaging at both 2 and 3 h after ¹⁸F-EF5 injection in four patients but found that the T/M ratio improved with time in only one individual while all four patients exhibited IHC-based evidence of hypoxia with average GLUT1 or OPN IRSs of at least 6. The kinetics of ¹⁸F-EF5 uptake have been previously examined in head and neck cancer patients where it was observed that the ¹⁸F-EF5 T/M ratio can increase during the first 3 h post-injection, but decreases between 3 and 4 h post-injection [20]. Based on these data and the low T/M ratios observed in our study, it is unlikely that ¹⁸F-EF5 PET T/M ratios would significantly increase with longer durations between radiotracer injection and PET imaging.

An ¹⁸F-EF5 T/M ratio >1.5 is typically used to indicate tumour hypoxia [17,20], and patient #8 imaged 3 h after radiotracer injection had a T/M ratio of 1.56. While the focal ¹⁸F-EF5 signal in this patient may represent tumour hypoxia, the presence of similar ¹⁸F-EF5 signal foci in the surrounding tissue and unremarkable IHC-based hypoxia scoring from this patient reinforce our conclusion that ¹⁸F-EF5 PET should not be relied upon to indicate prostate tumour hypoxia. While biopsies from patient #8 had a high GLUT1 IRS of 7.0, this GLUT1 IRS value was also observed in patient #3 who had the lowest T/M ratio of 0.95. Biopsies from patient #8 and patient #4 had high OPN IRS values of 6–7 and the two highest T/M ratios, although there was a range of T/M ratios for patients with average OPN IRS values greater than 6. Overall, these data suggest that ¹⁸F-EF5 PET exhibits limited sensitivity to detect hypoxia in localized prostate adenocarcinoma lesions that stain strongly for two immunohistochemical markers normally associated with tumour hypoxia.

GLUT1 is commonly over-expressed in hypoxic tumour cells to optimize cell metabolism in anaerobic conditions. GLUT1 mRNA expression was previously reported to distinguish prostate cancer from benign prostate hyperplasia, and high GLUT1 mRNA correlated with Gleason score [29]. In localized prostate adenocarcinoma, GLUT1 expression has been observed in core biopsies in conjunction with HIF1α expression [37] and correlated with poor patient outcome in a tissue microarray-based study [34]. However, while some groups have found a high frequency of GLUT1 staining in prostate tumours [31,37], GLUT1 does not always correlate with Gleason score, tumour stage, or PSA levels, and some groups have found only weak GLUT1 staining in prostate tumours [15,38,39]. Supiot et al. also found that weak GLUT1 staining by IHC was not associated with ¹⁸F-FMISO uptake in a cohort of prostate cancer patients [15]. Since there is a disparity in the literature regarding the use of GLUT1 to indicate hypoxia in prostate tumours, we stained prostate tumour samples with GLUT1 and OPN, which Rupp et al. used in combination to indicate prostate tumour hypoxia (as defined by pimonidazole staining) [30]. Importantly, pimonidazole was previously shown to correlate strongly with hypoxia quantified by magnetic resonance imaging in over 40 prostate cancer patients [26]. Vergis et al. showed increased OPN expression detected via IHC predicted for worse outcomes in prostate cancer patients treated with surgery [35]. Others have shown that increased OPN levels in prostate cancer tissue positively correlated with Gleason scores [40]. We found frequent moderate to high-intensity staining of both GLUT1 and OPN in our patient cohort, consistent with tumour hypoxia. While GLUT1 and OPN staining levels did not associate with Gleason scores, our dataset is of limited size, and biopsies submitted for IHC were selected in part to assess samples with greater Gleason scores rather than a range of Gleason scores. An IHC-based analysis of radical prostatectomy specimens, rather than the diagnostic needle biopsies analyzed in our study, would provide a more complete depiction of the overall levels and distribution of hypoxia within the prostate tumours. Nevertheless, the agreement between GLUT1 and OPN staining supports further use of these markers in combination [30] for detecting hypoxia in prostate cancer.

A likely explanation for the difficulty in using ¹⁸F-EF5 PET to quantify hypoxia in clinical prostate tumours is the diffuse nature of prostate cancer within the prostate gland. Rather than a single tumour mass as is typical of solid tumour types that exhibit high ¹⁸F-EF5 T/M ratios that indicate tumour hypoxia [17–21], prostate tumours tend to be multi-focal, with cancerous cells interspersed throughout the prostate. Since hypoxic tumour cells make up only a fraction of tumour cells in a given lesion, quantifying ¹⁸F-EF5 uptake in a fraction of prostate tumour cells may be limited by the detection sensitivity of PET. Notably however, previous success with ¹⁸F-FMISO PET to detect hypoxia in prostate tumours [13–15] suggests the imaging modality is not likely a significant limitation for using ¹⁸F-EF5 in prostate cancer.

Our ¹⁸F-EF5 data agree with a previous report testing ¹⁸F-FAZA to image hypoxia in prostate adenocarcinoma by PET, in which no significant radiotracer uptake or CAIX IHC staining was observed across 14 patients despite heterogeneous HIF1α staining [16]. While CAIX, a transmembrane protein involved in pH regulation, has been used as an endogenous marker of hypoxia in a range of tumour types including breast cancer [41], non-small cell lung cancer [42], and bladder cancer [43], the dependence of CAIX expression on hypoxia in prostate tumours is controversial. While Ambrosio et al. reported that CAIX significantly correlated with HIF1α staining and Gleason scores in 110 prostate cancer biopsies [44], multiple studies of localized prostate adenocarcinoma report that CAIX is absent in tumours that are positive for HIF1α [45] and in tumours positive for both HIF1α and GLUT1 [37]. Our data agree with these latter studies in that we did not observe CAIX staining in most of our patient cohort, and the CAIX staining that was present did not co-localize with GLUT1 or OPN. It is possible that pH regulation in prostate cancer cells is dependent on alternative mechanisms that do not involve CAIX, or perhaps prostate tumours are less acidic and do not require CAIX-mediated pH regulation, although further studies are required to determine the precise role of pH and CAIX in prostate cancer.

The metabolic redox balance specific to malignant prostate cancer cells is a potential explanation for the low uptake of ¹⁸F-EF5 (and ¹⁸F-FAZA) [16] in clinical prostate lesions, although pimonidazole has been shown to bind in hypoxic cells in several clinical prostate tumour studies. One study reports pimonidazole staining in 70% of the tumour-positive regions analyzed across 24 patients [46], another reports positive pimonidazole staining in 34 out of 37 patients [47], and a third study found pimonidazole staining to be associated with increased Gleason score and advanced tumour stage [26]. The ability of pimonidazole to identify a high proportion of prostate adenocarcinomas as hypoxic argues that the limitations of ¹⁸F-FAZA and ¹⁸F-EF5 are unlikely to be explained by issues of prostate tumour cell biochemistry. It is worth noting that anoxic prostate tumour cells in vitro exhibit decreased Cu-ATSM labelling compared to mammary carcinoma cells [48], and this decreased Cu-ATSM labelling was dependent on the relatively high expression of fatty acid synthase found in prostate tumour cells. The metabolic phenotype associated with high fatty acid synthase expression can hijack cellular reducing potential even in hypoxic conditions, resulting in a limited ability to reduce and therefore trap Cu-ATSM inside of the cell [48,49]. To our knowledge, no studies have specifically investigated the effect of fatty acid synthase expression on the binding efficiency for any of the major 2-nitroimidazole based hypoxia reporters (e.g., pimonidazole, EF5, FAZA, misonidazole), although the binding of these hypoxia markers also relies on intracellular reducing potential. Ultimately, with several published studies using pimonidazole to demarcate hypoxic cells in clinical prostate tumours [26,30,46,47], there is a need to evaluate immunohistochemical staining of non-radiolabeled EF5 to further determine whether EF5 is capable of detecting hypoxia in clinical prostate tumours.

Collectively, our data do not support the use of ¹⁸F-EF5-based PET to identify hypoxia in clinical prostate tumours. The use of other methods to quantify hypoxia, such as ¹⁸F-FMISO PET, pimonidazole, and IHC-based analyses of GLUT1 and OPN staining in diagnostic biopsies, is recommended to evaluate hypoxia in prostate cancer.

Disclosure statement

The authors report no potential conflict of interest.

Additional information

Funding

This work was financially supported by a Canadian Association of Radiation Oncology Foundation Sanofi-Aventis Research Innovation Award (CASARIA to MM) and an operating grant from the Canadian Institutes of Health Research (CIHR MOP – [126138] to KLB). BJW was funded by a CIHR Doctoral Research Award. BJW and LRD were supported by funding from the Strategic Training in Transdisciplinary Radiation Science for the twenty-first Century (STARS21) Program. KLB was supported by funding from the Michael Smith Foundation for Health Research Biomedical Research Scholar program.