Visual Abstract
Abstract
The present study was carried out to investigate whether PET imaging can be used as a potential substitute for immunohistochemical analysis of tumor samples in prostate cancer (PC) patients. Correlation between imaging signals of 2 PET tracers and the corresponding target structures was assessed. The first tracer was [68Ga]Ga-PSMA (prostate-specific membrane antigen)-HBED-CC (N,N′-bis [2-hydroxy-5-(carboxyethyl)benzyl]ethylenediamine-N,N′-diacetic acid) [68Ga]Ga-PSMAHBED-CC ([68Ga]PSMA), which is already implemented in clinical routines. The second tracer was 16β-[18F]fluoro-5α-dihydrotestosterone (16β-[18F]FDHT), which binds to the androgen receptor (AR). The AR is particularly interesting in PC, because AR expression status and its shift during therapy might directly influence patient care. Methods: This prospective, explorative clinical study included 10 newly diagnosed PC patients. Each patient underwent [68Ga]PSMA PET/MRI and [18F]FDHT PET/MRI scans before prostatectomy. Cancer SUVs were determined and related to background SUVs. After prostatectomy, tumor tissue was sampled, and AR and prostate-specific membrane antigen (PSMA) expression was determined. AR and PSMA expression was evaluated quantitatively with the open-source bioimage analysis software QuPath and with a 4-tier rating system. Correlation between imaging signals and marker expression was statistically assessed. Results: For [18F]FDHT, the SUVmax/SUVbackground ratio showed a significant, strong correlation (r = 0.72; P = 0.019) with the AR optical density of the correlating tissue sample. The correlation between PSMA optical density and the [68Ga]PSMA SUVmax/SUVbackground ratio was not significant (P = 0.061), yet a positive correlation trend could be observed (r = 0.61). SUVmax/SUVbackground ratios were higher for [68Ga]PSMA (mean ± SD, 34.9 ± 24.8) than for [18F]FDHT (4.8 ± 1.2). In line with these findings, the tumor detection rates were 90% for the [68Ga]PSMA PET scan but only 40% for the [18F]FDHT PET scan. The 4-tier rating of PSMA staining intensity yielded very homogeneous results, with values of 3+ for most subjects (90%). AR staining was rated as 1+ in 2 patients (20%), 2+ in 4 patients (40%), and 3+ in 4 patients (40%). Conclusion: [18F]FDHT PET may be useful for monitoring AR expression and alterations in AR expression during treatment of PC patients. This approach may facilitate early detection of treatment resistance and allows for adaptation of therapy to prevent cancer progression. [18F]FDHT PET is inferior to [68Ga]PSMA PET for primary PC diagnosis, but the correlation between [68Ga]PSMA SUVs and PSMA expression is weaker than that between [18F]FDHT and the AR.
- [68Ga]PSMA PET
- [18F]FDHT PET
- AR protein expression
- PSMA protein expression
- primary hormone-naïve prostate cancer
Prostate cancer (PC) is causing significant mortality and morbidity worldwide and accounts for approximately 3.8% of deaths caused by cancer in men (1). Although novel diagnostic and therapeutic options led to a decrease in PC related mortality, the incidence of PC is increasing (2).
The androgen receptor (AR) plays a central role in PC development and progression (3,4). Multiple therapies for PC target the hormonal axis connected to the AR (5). Therefore, information on AR expression status and its shift during therapy and along the treatment course would possibly allow prediction of treatment response and imminent resistance to therapy (6).
16β-[18F]fluoro-5α-dihydrotestosterone (16β-[18F]FDHT) binds to the AR and has been discussed as being particularly useful in deciphering the role of the AR in resistant and progressive metastatic PCs (7). Fox et al. performed [18F]FDHT and [18F]FDG PET scans on 133 metastatic castration-resistant PC patients in a prospective clinical study. The authors were able to demonstrate that PET-based assessment of AR expression with [18F]FDHT and glycolytic activity with [18F]FDG can detect tumor heterogeneity affecting survival (7). Data on [18F]FDHT kinetics are limited to castration-resistant PC patients; its use as an imaging agent is still restricted to clinical studies, and it has not been applied in clinical routines (8,9).
In contrast, the clinical utility of [68Ga]Ga-PSMA (prostate-specific membrane antigen)-HBED-CC (N,N′-bis [2-hydroxy-5-(carboxyethyl)benzyl]ethylenediamine-N,N′-diacetic acid) [68Ga]Ga-PSMAHBED-CC ([68Ga]PSMA) PET has been widely accepted, and it is regularly used in routine clinical practice for primary detection of PC and in recurrent disease (10,11).
Few clinical studies are available linking histopathologic patterns in tumor tissue to imaging signals (7,12,13). To date, there are no studies published that have quantitatively analyzed whether PET-based assessment of AR and prostate-specific membrane antigen (PSMA) expression in PC correlates with histopathologic expression of these markers. However, this knowledge is essential to reliably assess tumor heterogeneity and monitor alterations of AR and PSMA during therapy using noninvasive PET scans as a potential substitute for histopathologic analysis via repeated biopsies.
The aim of the present study was to investigate whether PET imaging can be used as a substitute for immunohistochemical (IHC) analysis of tumor samples in patients with newly diagnosed PC. For this purpose, [68Ga]PSMA and [18F]FDHT PET images were correlated with AR and PSMA IHC expression in PC tissue.
MATERIALS AND METHODS
Ethics
This study was conducted at the Medical University of Vienna (Vienna, Austria) and the Ludwig Boltzmann Institute Applied Diagnostics (Vienna, Austria) in accordance with the Declaration of Helsinki and the Good Clinical Practice Guidelines of the International Conference on Harmonization. The study was approved by the Ethics Committee of the Medical University of Vienna. Before inclusion, all study subjects gave oral and written informed consent to study participation.
Trial Design and Study Population
The present study was designed as a prospective, explorative clinical study. A total of 10 patients with newly diagnosed PC were included. Main inclusion criteria were age of greater than or equal to 18 y, histologically or cytologically confirmed prostate adenocarcinoma, and planned radical prostatectomy. Main exclusion criteria were any contraindication for performing a PET/MRI scan and a patient’s ineligibility for the size of the PET/MRI gantry.
Each patient underwent PET/MRI scans with [68Ga]Ga-PSMAHBED-CC and [18F]FDHT before surgery. Scans were scheduled on 2 separate study days, allowing a scan-free interval of at least 24 h between the 2 scans. A blood sample was taken before the [18F]FDHT PET scan to determine serum testosterone and prostate-specific antigen levels.
Up to 6 wk after the first scan, subjects were admitted to the clinical ward at the Department of Urology and a radical prostatectomy was performed by a urologist in accordance with standardized procedures. The timing and indication of surgery were not influenced by study participation. Tumor tissue obtained during surgery was used for IHC analysis of AR and PSMA expression in addition to the routine pathologic workup at the Department of Pathology (Medical University of Vienna).
PET/MRI
Radiosyntheses
All radiotracers for the study were produced in-house at the Radiopharmacy Unit, Vienna General Hospital, Vienna, Austria, by applying standard procedures in accordance with the state of the art for radiopharmaceutical preparations. Quality control was performed according to the European Pharmacopoeia. For details of the radiosynthesis, see the supplemental material (supplemental materials are available at http://jnm.snmjournals.org).
Imaging Protocols
All PET/MRI examination were conducted on a Biograph mMR (Siemens), consisting of a PET detector integrated with a 3.0-T whole-body MRI scanner. Two different imaging protocols were used. At our hospital, every newly diagnosed PC patient receives a diagnostic [68Ga]PSMA PET/MRI examination using a multiparametric MRI protocol with contrast enhancement to accurately evaluate the primary tumor and the prostate region (12). However, because [18F]FDHT PET is not yet established for routine clinical use at our institution and there are no previous studies using [18F]FDHT and PET/MRI scanners, the protocols described in previous studies regarding this tracer in metastatic castration-resistant PC patients were followed. For the clinical evaluation, the most relevant MRI sequences of the pelvic region were acquired (7,14,15). The PET images were reviewed by 2 trained nuclear medicine physicians.
[18F]FDHT PET/MRI Protocols
The [18F]FDHT PET/MRI examinations were performed 60 min after intravenous injection of [18F]FDHT at 3 MBq/kg of body weight. A static 10-min sinogram mode for the pelvis and 16-min partial-body PET (skull base to knees) were performed with 4 bed positions, each with a 4-min sinogram mode. For details of the sequence parameters, see the supplemental material.
Reconstruction parameters for PET were 3 iterations/21 subsets and summation of the 10-min pelvic acquisition for visual and semiquantitative analysis. MRI-based attenuation correction was applied using DIXON-VIBE sequences comprising in- and opposed-phase as well as fat- and water-saturated images.
[68Ga]PSMA PET/MRI Protocols
For [68Ga]PSMA PET/MRI studies, a 45-min dynamic list mode PET acquisition of the pelvis started immediately after the intravenous injection of [68Ga]Ga-PSMAHBED-CC at 2 MBq/kg of body weight. This scan was followed by partial-body PET (skull base to midthigh) performed with 4 bed positions, 4-min sinogram mode each. Reconstruction parameters for PET were 3 iterations/21 subsets and summation of the last 10-min pelvic acquisition for visual and semiquantitative analysis. MRI-based attenuation correction was applied using DIXON-VIBE sequences comprising in- and opposed-phase as well as fat- and water-saturated images. For details of the sequence parameters, see the supplemental material.
To improve image quality, especially of pelvic and abdominal images, forced diuresis with 20 mg of furosemide and 20 mg of hyoscine butylbromide (Buscopan; Boehringer Ingelheim) applied intravenously was done before the [68Ga]PSMA application, and all patients received a bladder catheter.
PET Data Analysis
PET data were analyzed using HybridViewer 3D (Hermes Medical Solutions) software. Anatomically exact regions of interest based on MRI data were defined. Semiautomated threshold-based volumes of interest were generated in areas with focally increased [68Ga]PSMA or [18F]FDHT uptake and evaluated with respect to the following semiquantitative data: SUVmax, SUVmean, and SUVpeak. A threshold of 90% of the SUVmax corrected for local background was applied for semiautomated PET imaging of both [68Ga]PSMA and [18F]FDHT. Background SUVmean was measured in the gluteus muscle for each subject separately (SUVbackground). Ratios of tumor SUV to SUVbackground were calculated for SUVmax, SUVmean, and SUVpeak.
IHC Analysis and Handling of Samples
IHC analysis was performed on tumor tissue at the Department of Pathology (Medical University of Vienna) using the automatic staining system VENTANA BenchMark ULTRA (Roche Tissue Diagnostics). For details of IHC analysis, see the supplemental material. Interpretation of marker expression was performed by the same qualified uropathologist. Membranous PSMA and nuclear AR quantification for each sample was semiquantitatively determined by a qualified uropathologist masked with regard to clinical data using a 4-tier system (0, 1+, 2+, and 3+). All regions present on the histologic slide were evaluated. On the basis of the overall expression of said markers, the rating “0” indicates no expression and the rating “3+” indicates the strongest expression. If heterogeneous AR or PSMA expression was present, then the regions were separated into 2 regions with low versus high protein expression. Membranous and cytoplasmic PSMA expression and nuclear expression for AR were also quantitatively determined using the open-source bioimage analysis software QuPath (v.0.3.0). The area chosen for analysis was spatially matched with the area of SUVmax detection in [18F]FDHT PET imaging. For AR, the “positive cell detection” function was used to automatically detect positive stained cells in an area of 4 mm2 on the basis of the average 3,3′-diaminobenzidine staining intensity within the nucleus and given as diaminobenzidine optical density (OD) mean, ranging from 0 to 1, with a cutoff value of 0.05 to detect positive cells. To also account for the cell density in the marked area, the sum of all diaminobenzidine OD mean values (OD mean sum) was chosen for further analysis. For PSMA, a pixel classifier was trained after annotation of sample positive and negative areas. This pixel classifier was used to detect the positive staining area within the mentioned 4-mm2 area. In the next step, the average diaminobenzidine staining intensity in this positive staining area was calculated and multiplied with the surface area to account for cell density.
Trial Endpoints and Statistical Analysis
The main outcome parameter was the correlation between tracer radiation dose normalized to injected dose, expressed as SUV and PSMA and AR protein expression levels in tissue samples assessed by IHC. Quantitative protein expression levels and the 4-tier ratings were correlated with SUVmax, SUVmean, and SUVpeak and the respective SUV to SUVbackground ratios. In case of heterogeneous protein expression, the areas with stronger staining intensity spatially matched the areas of SUV detection and were therefore chosen for the correlation analysis. Linear correlation was investigated using the Pearson correlation coefficient and reported with the 95% CI.
Statistical analysis was performed using a commercially available computer program (GraphPad Prism 9.3.1 for macOS; GraphPad Software). All data collected were expressed as mean ± SD or median with interquartile range.
RESULTS
Demographics and Clinical Characteristics of Subjects
Between February 2020 and March 2021, 10 patients with newly diagnosed PC were included in the study and completed all study procedures (Table 1). None of the patients had received any hormonal cancer therapy. None of the patients had a testosterone level below the castration threshold of 0.5 ng/mL. Pathologic workup of the tumor tissue revealed a median Gleason score of 8 (7–8).
Patient Characteristics and Pathologic Tumor Characteristics
PET Scans
The tumor detection rate of the [68Ga]PSMA PET scan was 90% and 40% for the [18F]FDHT PET scan. Representative [68Ga]PSMA and [18F]FDHT PET scans are shown in Figure 1.
PET images and IHC stains for 1 study patient. IHC images of PSMA staining and AR staining are magnified 5-fold and 40-fold, respectively. SUVmax/SUVbackground ratios were 14.3 for [68Ga]PSMA PET scan and 5.0 for [18F]FDHT PET scan. Staining of tissue samples showed strong PSMA expression but weak AR expression. H&E = hematoxylin and eosin stain.
Measured SUVs in PC lesions for [68Ga]PSMA and [18F]FDHT are shown in Figure 2. The SUVmax of [68Ga]PSMA PET was 17.0 ± 15.0 (mean ± SD), and that of [18F]FDHT PET was 3.4 ± 0.5. Mean background SUVmean measured in the gluteus muscle was 0.5 ± 0.1 and 0.7 ± 0.1 for [68Ga]PSMA PET and [18F]FDHT PET, respectively. Ratios of tumor SUV to background SUV are shown in Table 2. These were considerably higher for [68Ga]PSMA than for [18F]FDHT. The highest ratios were achieved for SUVmax with values of 34.9 ± 24.8 and 4.8 ± 1.2 for [68Ga]PSMA and [18F]FDHT, respectively.
SUVmax, SUVpeak, and SUVmean of PET scans in PC tissue for [68Ga]PSMA and [18F]FDHT.
Ratios of Tumor SUV to Background SUV* for [68Ga]PSMA and [18F]FDHT PET Scans
Immunohistochemistry
The quantification results of the IHC staining of PSMA and AR in the tumor tissue are shown in Table 3 and sample pictures of the staining are shown in Figure 1. The semiquantitative evaluation of the membranous staining intensity of PSMA yielded very homogeneous results with values of 3+ for most subjects (90%). In contrast to this, the AR staining was rated with 1+ in 2 patients (20%), with 2+ in 4 patients (40%) and with 3+ in 4 patients (40%). Three patients (1, 5 and 6) demonstrated highly heterogeneous AR expression with areas showing negative and positive nuclear AR expression. For subject 5, areas with high AR expression showed considerably lower PSMA expression (Supplemental Fig. 1). In this subject the missing AR staining was found especially in low differentiated, cribriform glands. However, this could not be observed in the other subjects. In contrast, in subject 6 AR staining was only visible in low differentiated, cribriform glands. The Gleason patterns within the different areas of the subjects did not show significant differences [subject 1: 7 (3 + 4); subject 5: 7 (3 + 4), subject 6: 8 (4 + 4)].
Intensity of IHC Staining of PSMA and AR Determined in Two Ways*
The PSMA diaminobenzidine OD measured with QuPath was 2,257,912 ± 1,297,251 (mean ± SD). The nuclear AR diaminobenzidine OD measured with QuPath was 3,857 ± 2,991 (mean ± SD).
Correlations
An overview of the different correlations is given in Table 4. Correlation of imaging signals with the 4-tier rating was not investigated for PSMA since protein expression was high (rated with “3+”) in all patients. For all investigated correlations, the SUVmax/SUVbackground ratio consistently yielded the strongest correlation with staining intensity.
Pearson Correlation Coefficient and P Value for Correlation of OD and 4-Tier Rating with Different Imaging Parameters*
A strong significant correlation between the AR expression and the [18F]FDHT SUVmax/SUVbackground ratio with correlation coefficients of r = 0.72 (95% CI, 0.17 to 0.93) for the OD (Fig. 3) and r = 0.80 (95% CI, 0.34 to 0.95) for the 4-tier rating could be demonstrated.
Correlation between SUVmax/SUVbackground ratio (SUV ratio) and PSMA diaminobenzidine OD and AR diaminobenzidine OD. Straight line represents linear regression line. PSMA OD was calculated as PSMA diaminobenzidine OD mean × surface area, and AR OD was calculate as AR diaminobenzidine OD mean sum.
In contrast, correlation between the PSMA OD and the Ga-PSMA SUVmax/SUVbackground ratio was not significant (P = 0.061) with a correlation coefficient of r = 0.61 (95% CI, −0.03 to 0.90) (Fig. 3).
DISCUSSION
A significant positive correlation between [18F]FDHT uptake in the PET scans and AR expression in cancer tissue could be demonstrated. For PSMA the PET/IHC correlation showed a positive trend but was not significant.
To our knowledge the present study is the first study that aimed to quantitatively assess whether [18F]FDHT and [68Ga]PSMA uptake in PET scans correlates with histopathologic AR and PSMA expression.
The [18F]FDHT SUVmax/SUVbackground ratio showed a strong significant correlation (r = 0.72; P = 0.019) with AR OD of the correlating tissue sample. Compared with previous studies in humans, lower [18F]FDHT SUVs were observed in the present study (8,9). Larson et al. performed [18F]FDHT PET scans in 7 patients with PC and observed an average SUVmax of 5.28 ± 2.57 (8). An even higher SUVmax of 7.46 ± 3.37 was reported by Vargas et al. in 27 patients with PC (9). These values are 1.6- to 2.2-fold higher than the average SUVmax (3.4 ± 0.5) observed in the present study. Presumably, the underlying reason for this is that for the present study patients were enrolled who did not receive any hormonal pretherapy and therefore had physiologic testosterone blood levels. It can be assumed that [18F]FDHT binding to ARs was competitively inhibited by endogenous dihydrotestosterone (DHT) leading to comparatively low [18F]FDHT uptake. In contrast, previous studies with [18F]FDHT only included patients with testosterone concentrations below the castration threshold (<50 ng/dL) (7,8,14). A finding that supports our theory was described by Larson et al. (8). The authors performed a [18F]FDHT PET scan in castrated PC patients and rescanned 2 of the study subjects after administration of exogenous testosterone. In 1 of the 2 patients, the plasma DHT concentration was considerably higher before the second [18F]FDHT scan and tracer uptake also decreased substantially. In the other patient, tracer uptake was unchanged compared with the initial scan, probably because plasma DHT initially increased, but then decreased again before the [18F]FDHT scan.
Currently, there are 2 studies published, which investigated the correlation between [68Ga]PSMA uptake in PET and IHC PSMA expression. In the prospective study by Rüschoff et al., IHC staining intensity was only determined semiquantitatively, according to the 4-tier rating system. A positive trend was described for membranous and cytoplasmic PSMA expression, which did not reach significance (15). Similar to the study by Rüschoff et al., in the retrospective study by Woythal et al. a 4-tier rating system was used for PSMA staining intensity (16). Woythal et al. did not discriminate between membranous and cytoplasmic PSMA expression and they were able to demonstrate a significant correlation between SUVmax and the immunoreactivity score which incorporates staining intensity and percentage of positive cells (P < 0.001). Unfortunately, the authors do not report the correlation between SUVmax and staining intensity. The patient characteristics of the study by Woythal et al. were comparable to the present study with a mean Gleason score of 7.9, but the sample size was larger (31 primary PC patients). In the present study SUVmax and staining intensity were positively correlated, but without statistical significance (r = 0.6; P = 0.068). Compared with the studies mentioned earlier, we used a more refined, granular and objective method for staining intensity assessment. Unfortunately, membranous and cytoplasmic PSMA expression could also not be discriminated in our study due to artifacts generated by the high cytoplasmic background PSMA staining in the automatic cell detection using QuPath. Therefore, the software was not able to reliably differentiate between membranous and cytoplasmic expression. Interestingly, the correlation between PSMA OD and the [68Ga]PSMA SUVmax/SUVbackground ratio was not significant, yet a positive correlation trend could be observed (Fig. 3). The small sample size may account for this lack of significance. However, the present study was designed as an exploratory pilot study and therefore only 10 patients were included.
The tumor detection rate was 40% in the [18F]FDHT PET scans and 90% in the [68Ga]PSMA PET scans. In line with this finding, the SUVmax for [68Ga]PSMA was about 7-fold higher than that for [18F]FDHT (34.9 ± 24.8 vs. 4.8 ± 1.2, respectively). Despite the low detection rate, ratios of [18F]FDHT cancer SUV to background SUV were always above 1, indicating increased binding of [18F]FDHT in cancer tissue (Table 2). However, the SUV ratios were not high enough to be identified as tumor-positive areas. This explains that a significant correlation between [18F]FDHT SUVmax/SUVbackground ratio and AR OD could be observed, despite the low tumor detection rate.
In future studies the correlation between [18F]FDHT SUVs and AR expression in patients with low testosterone levels should be examined to investigate the hypothesis that endogenous androgens antagonize [18F]FDHT. In these patients the correlation of [18F]FDHT with AR expression might be more pronounced and is of particular clinical importance, since PC is usually progressed in these patients.
Another interesting finding of our study was the highly heterogeneous AR expression in 3 subjects. Magi-Galluzzi et al. performed a retrospective analysis of 40 PC samples and observed decreasing AR staining with increasing Gleason grade (17). To investigate whether the different regions in our subjects indicate different foci of aggressiveness, we determined the Gleason patterns for the different regions of AR expression. However, no significant differences could be observed. Others have observed that AR positive cells were also PSMA positive (18). We did not observe this trend. On the contrary, in 1 subject areas with higher AR expression showed lower PSMA expression. In consequence, no correlation between morphology and expression pattern can be derived from these findings and the influence of fixation artifacts in preanalytics cannot be excluded.
As mentioned earlier, a limitation of the present study was the small sample size. For a more robust statistical analysis a higher sample size would have been preferable, but due to the exploratory character of the study only 10 patients were included. In addition, the effects of hormonal therapy on [18F]FDHT uptake and the correlation of [18F]FDHT uptake with therapeutic response remain to be investigated in future longitudinal studies.
CONCLUSION
The findings of our study suggest that [18F]FDHT PET scans may be useful for monitoring AR expression and alterations of AR expression during PC treatment. Knowledge of changes in AR expression during disease progression could help clinicians recognize imminent resistance to therapy and improve patient outcomes.
DISCLOSURE
No potential conflict of interest relevant to this article was reported.
KEY POINTS
QUESTION: Can PET imaging be used as a substitute for IHC analysis of tumor samples in patients with newly diagnosed PC?
PERTINENT FINDINGS: In this prospective, explorative clinical study in 10 patients with newly diagnosed PC, we investigated the correlation between imaging signals of [68Ga]PSMA and [18F]FDHT and the protein expression of their corresponding target structures (PSMA and AR). The results suggested that [18F]FDHT PET scans may be useful for monitoring AR expression and alterations in AR expression during PC treatment.
IMPLICATIONS FOR PATIENT CARE: Monitoring AR expression during disease progression could help clinicians recognize imminent resistance to therapy and thereby improve patient outcomes.
ACKNOWLEDGMENT
We thank Neydher Berroterán-Infante for his help in establishing the [18F]FDHT synthesis.
Footnotes
Published online Jan. 19, 2023.
- © 2023 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication October 10, 2022.
- Revision received January 12, 2023.