Impact of the molar activity and PSMA expression level on [18F]AlF-PSMA-11 uptake in prostate cancer

This two-part preclinical study aims to evaluate prostate specific membrane antigen (PSMA) as a valuable target for expression-based imaging applications and to determine changes in target binding in function of varying apparent molar activities (MAapp) of [18F]AlF-PSMA-11. For the evaluation of PSMA expression levels, male NOD/SCID mice bearing prostate cancer (PCa) xenografts of C4-2 (PSMA+++), 22Rv1 (PSMA+) and PC-3 (PSMA−) were administered [18F]AlF-PSMA-11 with a medium MAapp (20.24 ± 3.22 MBq/nmol). SUVmean and SUVmax values were respectively 3.22 and 3.17 times higher for the high versus low PSMA expressing tumors (p < 0.0001). To evaluate the effect of varying MAapp, C4-2 and 22Rv1 xenograft bearing mice underwent additional [18F]AlF-PSMA-11 imaging with a high (211.2 ± 38.9 MBq/nmol) and/or low MAapp (1.92 ± 0.27 MBq/nmol). SUV values showed a significantly increasing trend with higher MAapp. Significant changes were found for SUVmean and SUVmax between the high versus low MAapp and medium versus low MAapp (both p < 0.05), but not between the high versus medium MAapp (p = 0.055 and 0.25, respectively). The effect of varying MAapp was more pronounced in low expressing tumors and PSMA expressing tissues (e.g. salivary glands and kidneys). Overall, administration of a high MAapp increases the detection of low expression tumors while also increasing uptake in PSMA expressing tissues, possibly leading to false positive findings. In radioligand therapy, a medium MAapp could reduce radiation exposure to dose-limiting organs with only limited effect on radionuclide accumulation in the tumor.


Small animal PET/CT imaging.
To determine the impact of varying molar activities, each C4-2 xenograft bearing mouse underwent three PET/CT scans within a timeframe of 14 days. On day 1, 7 and 14, mice received either high/low/medium MA app (n = 2) or medium/low/high MA app (n = 4), respectively. The mean adminis- The images were acquired in list mode using a small animal PET scanner (β-cube, Molecubes, Ghent, Belgium) with a spatial resolution of 0.85 mm and an axial field-of-view of 13 cm. All PET scans were reconstructed into a 192 × 192 × 384 matrix by an ordered subsets maximization expectation (OSEM) algorithm using 30 iterations and a voxel size of 400 × 400 × 400 µm 3 .
Immunohistochemical evaluation. After the last scan, two mice/cell line xenograft were sacrificed and tumors were collected for immunohistochemical (IHC) evaluation. Sections were taken from the centre of the tumor sample and stained using Haematoxylin/Eosin, incubated with a primary PSMA antibody (1:400, 2 h, ab133579, Abcam) and counterstained using Haematoxylin (Mayer). Sections were digitally scanned with a virtual scanning microscope (Olympus BX51, Olympus Belgium SA/NV, Berchem, Belgium) at high resolution (20 × magnification).
Western blot analysis. The tumors of the other mice were harvested for western blot (WB) analysis to quantify the PSMA expression levels of the tumor tissues. After thorough rinsing of the tumors, tissue lysates were prepared using RIPA sample buffer. Protein concentrations were determined using the Pierce BCA Protein assay (ThermoFisher Scientific). Samples containing 50 µg of total protein were loaded on 8.5% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) gels. After the transfer, the nitrocellulose membranes were blocked using Odyssey Blocking Buffer (Li-cor Biosciences, Bad Homburg, Germany). Membranes were incubated overnight using a monoclonal PSMA Rabbit anti-Human primary antibody (1:2000, MA533086, Invitrogen, Fischer Scientific, Belgium) and polyclonal Rabbit anti-Human beta actin primary antibody (1:1000, PA1-183, Invitrogen, Fischer Scientific, Belgium), followed by 1 h incubation with IRDye® 800CW Goat anti-Rabbit IgG secondary antibody (1:15,000, Li-cor Biosciences, Bad Homburg, Germany) ( Supplementary Fig. 3). The relative densities of both PSMA and β-actin bands were determined using ImageJ 17 .
Data analysis. Co-registration and analysis of the PET/CT images was performed using PMOD (PMOD Technologies®, Zürich, Switzerland). Images were visually analysed using the PROMISE criteria. This standardizes image interpretation based on activity uptake in reference tissues (blood, liver and salivary gland) and subdivides suspicious lesions into no, low, intermediate or high PSMA expression 18 . Volumes of interest (VOIs) were drawn manually for delineating the tumor, kidneys, bladder, salivary and lacrimal glands, heart (blood), liver, spleen, muscle and bone. Tissue uptake in each VOI was corrected for residual activity in the syringe and radioactive decay and expressed as SUV mean and SUV max . Furthermore, tumor-to-organ ratios of tissues with a high likelihood for metastases including tumor-to-liver (TLR), tumor-to-muscle (TMR) and tumor-to-bone ratio (TBoR) were determined. Statistical analysis. All uptake parameters (SUV mean , SUV max , TLR, TMR and TBoR) were reported as mean ± SD. The statistical analysis was performed in R 19 using the Mann-Whitney U test for comparison of uptake between different PSMA expressing xenografts and the Kruskal Wallis test for comparing radioligand uptake of varying MA app , followed by post-hoc pairwise comparison using the Wilcoxon-signed rank test with holm correction for multiple testing. The significance level was set on p ≤ 0.05.

Ethics approval and consent to participate. The study was approved by the Ghent University Ethical
Committee on animal experiments (ECD 18/116). All animals were kept and handled according to the European guidelines (Directive 2010/63/EU). The manuscript complies with the ARRIVE guidelines.

Results
Semi-quantitative analysis of PSMA expression levels in prostate cancer cell lines. The PSMA expression levels were determined by both western blot and immunohistochemical assessment (Fig. 1). WB analysis detected the highest presence of PSMA expression in C4-2 tissue, while only moderate to low expression levels could be observed for 22Rv1. Semi-quantification of blots revealed a β-actin normalized density for PSMA expression of 16.36 ± 3.05 for C4-2 and 0.94 ± 0.21 for 22Rv1 tumors. PC-3 tumors did not show PSMA expression. Similar results were demonstrated by IHC staining, where all C4-2 cells were clearly positively stained for PSMA, while 22Rv1 showed weaker and more heterogeneous staining of the tumor tissue.

Relationship between PSMA expression levels and [ 18 F]AlF-PSMA-11 tumor uptake.
Comparative images of [ 18 F]AlF-PSMA-11 uptake (medium MA app ) in different PSMA expressing tumors are presented in Fig. 2. Tumor uptake was positively correlated with higher expression levels. According to the PROMISE criteria, PC-3 tumors were PSMA negative (score 0) as tumor uptake was comparable with activity in the blood (heart). 22Rv1 tumors were visually detectable and the tumor activity was higher than the liver and salivary glands, which corresponds to high PSMA expression (score 3). C4-2 tumors showed good image contrast with tumor uptake above activity in the salivary glands (score 3) (Supplementary Table 1 www.nature.com/scientificreports/  www.nature.com/scientificreports/ Quantitative VOI analysis demonstrated significant differences for SUV mean and SUV max as well as tumor-toorgan ratios including tumor-to-liver, tumor-to-blood and tumor-to-muscle in function of PSMA expression levels ( Table 1). When comparing C4-2 (high PSMA expression) to 22Rv1 (low PSMA expression), SUV mean and SUV max values were approximately 3.22 and 3.17 times higher, respectively (p < 0.0001). Similar ratios of 2.77, 2.23 and 2.96 were found for tumor-to-liver, tumor-to-blood and tumor-to-muscle, respectively (p < 0.0001). SUV values were significantly lower for PC-3 tumors compared to both C4-2 (p < 0.01) and 22Rv1 tumors (p < 0.001).
Impact of the molar activity on tumor uptake. To (Fig. 3). Kidney uptake decreased as well with lower MA app while activity in de bladder Table 1. SUV values and tumor-to-organ ratios (liver, blood and muscle) for different PCa tumor cell lines with varying PSMA expression levels: PC-3 (no PSMA expression), 22Rv1 (low PSMA expression) and C4-2 (high PSMA expression). P values were calculated using the Wilcoxon signed-rank test with Holm correction with C4-2 as a reference, the significance level was set at p = 0.05, **p < 0.01, ****p < 0.0001. a Expression levels were determined by semi-quantitative western blot analysis and are expressed β-actin normalized density. b Activity in blood was determined by delineation of the heart.

Cell line
Expression level a SUVmean SUVmax Tumor-to-liver ratio Tumor-to-blood ratio b  www.nature.com/scientificreports/ increased. PSMA expressing tissues such as salivary glands, lacrimal glands and spleen were clearly visible on high MA app PET images but were barely visible on medium MA app images and not detectable on low MA app images (Supplementary Table 2). When comparing intra-individual tumor uptake of the high, medium and low MA app , SUV mean (2.14 ± 0.54, 1.48 ± 0.47 and 0.58 ± 0.10, respectively) and SUV max (4.44 ± 1.25, 3.27 ± 1.07 and 1.15 ± 0.19, respectively) showed a significantly decreasing trend (both p < 0.001). Post-hoc pairwise comparison revealed statistically significant changes between the high and low MA app (p < 0.05) as well as between the medium and low MA app (p < 0.05), but not between the high and medium MA app (p = 0.055 and 0.25, respectively) (Fig. 4a). Additionally, 22Rv1 xenograft bearing mice (n = 3) received an additional PET/CT scan with high MA app (246.3 ± 33.1 MBq/µg). Comparison of SUV mean and SUV max values demonstrated a statistically significant difference between the high and medium MA app (both p = 0.031).
Tumor-to-organ ratios were determined for tissues with a high likelihood for metastases including liver (TLR), muscle (lymph nodes) (TMR) and bone (TBoR). In C4-2 xenografts, tumor-to-organ ratios remained constant between the high and medium MA app for TLR (15.71 ± 1.99 vs 17.32 ± 3.55, p = 0.46), TMR (20.69 ± 3.82 vs 19.64 ± 5.37, p = 0.74) and TBoR (3.77 ± 0.78 vs 3.33 ± 1.03, p = 0.31) while a statistically significant difference could be observed for high versus low MA app and medium versus low MA app (Fig. 4b). A similar trend between the high and medium MA app was found for 22Rv1 xenografts, however, the tumor-to-liver ratio remained constant. On the other hand, SUV mean values of PSMA expressing tissues such as salivary and lacrimal glands, spleen, and kidneys significantly decreased between the high and medium MA app (p < 0.01) (Fig. 5). Tumor-to-salivary gland ratios increased from 2.30 ± 0.37 for the high MA app to 5.35 ± 0.59 for the medium MA app . While kidney uptake significantly decreased, uptake in the bladder rose for the medium and low MA app .

Discussion
Although the diagnostic applications of PSMA as a target for imaging have been the subject of many clinical trials, preclinical studies are essential for a comprehensive investigation of several important aspects of PSMA PET imaging. In this paper, [ 18 F]AlF-PSMA-11 was used for imaging experiments to further understand PSMA The relationship between target expression levels and [ 18 F]AlF-PSMA-11 uptake was determined by PET/ CT imaging of mice bearing PCa tumors with varying PSMA expression levels. The use of a medium MA app for the expression level experiments was selected based on several practical considerations. A larger group of mice could be imaged on the same day, while the use of the high MA app would significantly limit the size of the study population. Also, using the high MA app would introduce a larger variation in the MA app between animals, as the amount of PSMA-11 precursor is so small, that minor differences in amount of precursor added to the stock solution would lead to large differences in MA app . Semi-quantitative western blot analysis showed that PSMA expression levels of C4-2 tumors were approximately 16 times higher compared to 22Rv1 tumors, while SUV values were only 3 times higher. A positive association between PSMA expression and SUV values could be found. Although 22Rv1 tumors had only low PSMA expression levels on western blot, activity uptake was above the liver, scoring positively (score 3) according to the PROMISE criteria. This suggests the ability of [ 18 F] AlF-PSMA-11 to detect PCa lesions, even with low PSMA receptor abundancy. Increasing the MA app of [ 18 F] AlF-PSMA-11 increased absolute SUV mean and SUV max values (Fig. 4), but due to higher salivary gland uptake, the PROMISE score did not increase. However, TMR and TBoR did increase significantly, which could be beneficial for the detection of lymph nodes and bone metastases with low PSMA expression levels, although this should be confirmed in a murine model of bone metastases. Lückerath et al. determined the in vivo correlation between [ 68 Ga]PSMA-11 PET/CT and PSMA expression using four murine PCa cell lines. [ 68 Ga]PSMA-11 PET was able to detect changes in PSMA expression at low levels, but this sensitivity was lost at higher PSMA levels, which is in agreement with the reported non-linear correlation 20 .
PSMA expression was shown to be an important prognosticator for overall survival of metastatic castrate resistant prostate cancer (mCRPC) patients under [ 177 Lu]PSMA treatment. The presence of low average-PSMA expressing tumors was found to be a negative prognostic factor in these patients 21 . The threshold for determining the PSMA expression level was based on the criterium set by Hofman et al. who defined high PSMA expression as any metastatic lesion with a SUV max of 1.5 times the liver SUV. Here, [ 177 Lu]PSMA therapy resulted in 50% decline in PSA for 17/30 mCRPC patients presenting with high PSMA expression metastases 22 . The liver seems to be a solid choice as reference organ as our results show that changes in the MA app did not significantly alter tumorto-liver ratios (between the high and medium MA app ). It should however be noted that the difference in chemical structure between [ 177 Lu]PSMA-617 and [ 18 F]AlF-PSMA-11 could also have an influence on biodistribution and clearance pathways. Additional research should determine to which extent PSMA expression levels could be used as an exclusion criterium for [ 177 Lu]PSMA therapy. In our study, the difference between the medium and high MA app was more pronounced in low expression 22Rv1 tumors. SUV values, TMR and TBoR increased significantly when administering the high MA app . This can be attributed to the lower amount of receptors available to bind radiolabelled PSMA molecules and more rapid saturation in the presence of a larger peptide mass.
Overall, the highest mean absolute SUV mean and SUV max values were obtained after administration of [ 18 F] AlF-PSMA-11 with high MA app , but the difference between the medium and high MA app was not statistically significant, both for SUV values as for tumor-to-organ ratios (TLR, TMR and TBoR). There was however a substantial decrease of activity uptake in PSMA expressing tissues (salivary and lacrimal glands, spleen and kidneys) between the high and medium MA app . The use of medium MA app [ 18 F]AlF-PSMA-11 could therefore reduce activity uptake in PSMA positive healthy tissues while maintaining image contrast of tumor metastases compared to the surrounding tissue. This would be beneficial for both diagnostic purposes and radioligand therapy. Prostate cancer lesions could be more reliably detected due to the constant tumor-to-organ ratio while www.nature.com/scientificreports/ less non-specific uptake could facilitate scan interpretation by limiting the potential pitfalls of PSMA PET caused by aspecific physiological uptake 23 . In radioligand therapy, healthy tissue would be less exposed to the radiation while absolute tumor uptake would remain approximately constant. During PSMA radioligand therapy, radiation exposure caused by activity uptake in PSMA-expressing nontumor tissues can cause dose-limiting toxic side effects. Radioligand uptake in the salivary glands causes xerostomia and can be a reason for treatment discontinuation 24,25 . Lowering the MA app for RLT ligands could be a useful tool for reducing radiation exposure and should be further investigated as an alternative for botulinum toxin injections 26 or sialendoscopy and dilatation combination with saline irrigation and steroid injections 27 . A preclinical study by Fendler et al. also showed that administration of [ 177 Lu]PSMA-617 with a molar activity of 62 MBq/nmol achieved the highest tumor-to-salivary gland ratio compared to lower molar activities (31 and 15 MBq/nmol) 28 . When correcting for the administered dose of 60 MBq, the highest molar activity corresponds to 0.97 nmol PSMA-617, which is in the same range as our medium MA app group (0.49 nmol PSMA-11) that also showed the highest tumor-to-salivary gland ratio. However, it was reported that the low to moderate PSMA staining on IHC staining does not correlate with the high PSMA radioligand accumulation. It was hypothesized that tracer accumulation in the salivary glands is therefore a combination of both specific and non-specific uptake 29 . The latter mechanism is not yet elucidated and our results show a large impact of a higher amount of peptide on activity uptake in salivary glands, suggesting that this as a promising direction for further research. The difference in PSMA expression between murine and human salivary glands must also be taken into consideration. As the binding affinity and PSMA concentration in murine salivary glands is lower compared to human salivary glands, it would be expected that the influence of molar activities will be more pronounced in humans 30 . Another organat-risk are the kidneys. Renal clearance can lead to accumulation of radiolabelled peptides in the tubular cells where it irradiates the kidneys, potentially leading to renal dysfunction, particularly in patients with a known history of nephropathy.  33 .
These results indicate potential benefits of the use of cold precursor or PSMA inhibitory small molecules such as 2-PMPA for reducing radiation exposure to the kidneys. Additionally, a high tumor load in patients can also lead to reduced activity uptake in dose-limiting organs 34 . [ 18 F]AlF-PSMA-11 might therefore be applied as a diagnostic tool for individual optimization of RLT protocols considering PET-based tumor load and radioligand MA app .
These results show the importance of the amount of carrier on tumor targeting strategies. The optimal amount should be low to avoid competitive binding of unlabelled molecules at the target site but not too low as the tracer molecules could be trapped in non-specific binding places. For diagnostic purposed, using a high MA app will increase the detection of low expression tumors while also increasing specific uptake in PSMA expressing tissues, possibly leading to false positive findings. For radioligand therapy, further research should focus on finding a balance between administration of a high MA app for high absolute uptake in the tumor tissue while minimizing activity uptake in healthy organs in order to reduce dose-limiting toxicity. Also, although the considerations mentioned before are valid for high PSMA expressing PCa lesions, decreasing the MA app to limit non-specific uptake could also affect the detectability of low PSMA expressing lesions, which could lead to an underestimation of the tumor burden.
A major limitation of this study is the translation from mouse to human. Although the differences in the administered MA app in mice gave well defined results, the activity dose and amount of peptide that is administered compared to total body weight for PET/CT is different for mice and humans, which makes translation into clinical applications more challenging. However, as the molar activity of a PSMA PET tracer decreases over time, it would be beneficial to evaluate the effect on both tumor uptake and non-target uptake in humans. In a retrospective analysis of patients who underwent [ 18 F]rhPSMA-7.3 PET/CT, only minor effects on biodistribution were found for a tenfold decrease of MA app , except for the salivary glands and spleen, which showed significantly lower uptake for the lower molar activity 35 . Overall, these results encourage a prospective clinical trial using [ 18 F] AlF-PSMA-11 comparing a high to medium MA app .
Due to the intra-individual comparison of MA app levels, most mice were imaged multiple times and tumor sizes differed slightly between scans. However, the tumor size was many times higher than the PET resolution and mice were randomized for MA app sequence, therefore minimizing this limitation as much as possible.

Conclusion
This paper evaluated [ 18 F]AlF-PSMA-11 PET for expression-based imaging of prostate cancer and the effect of the amount of carrier administered on tumor uptake and tumor-to-organ ratios. A positive association was found between PSMA expression and tumor uptake. The highest absolute tumor uptake was obtained by the high MA app . And although activity uptake in PSMA expressing organs decreased significantly with lower MA app , there was no statistically significant difference in tumor SUV between the high and medium MA app . These results suggest that administration of a high MA app increases the detection of low expression tumors while also increasing specific uptake in PSMA expressing tissues, possibly leading to false positive findings. In radioligand therapy, a medium MA app could reduce radiation exposure to dose-limiting organs with only limited effect on radionuclide accumulation in the tumor. Overall, it is of utmost importance to validate the association between both PSMA expression and molar activity relative to radioligand activity uptake. www.nature.com/scientificreports/

Data availability
The datasets and images used and/or analysed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/