Neuron-specific enolase has potential value as a biomarker for [18F]FDG/[68Ga]Ga-PSMA-11 PET mismatch findings in advanced mCRPC patients

Background PSMA-targeted radioligand therapy (PSMA-RLT) yielded impressive results in the metastasized castration-resistant prostate carcinoma (mCRPC) setting. High expression of PSMA is essential for successful PSMA-RLT. However, some patients develop [18F]FDG-avid lesions with low or no PSMA expression ([18F]FDG/[68Ga]Ga-PSMA-11 mismatch findings on PET/CT) in the course of treatment. Those lesions are not affected by PSMA-RLT and a change in therapy management is needed. To enable early mismatch detection, possible blood parameters as indicators for the occurrence of [18F]FDG/[68Ga]Ga-PSMA-11 mismatch findings on PET/CT were evaluated. Methods Retrospective study of N = 66 advanced mCRPC patients with dual [68Ga]Ga-PSMA-11 and [18F]FDG PET/CT imaging within 4 weeks, who were referred for or received [177Lu]Lu-PSMA-617 radioligand therapy. Prostate-specific antigen (PSA), neuron-specific enolase (NSE), gamma-glutamyltransferase (GGT), and alkaline phosphatase (ALP) were tested as indicators for the occurrence of [18F]FDG/[68Ga]Ga-PSMA-11 mismatch findings. Additional to absolute values, relative changes (ΔPSA, ΔNSE, ΔGGT, ΔALP) over a period of 4 ± 1 weeks prior to [18F]FDG PET/CT were analyzed. Results In total, 41/66 (62%) patients revealed at least one [18F]FDG/[68Ga]Ga-PSMA-11 mismatch finding on PET/CT. These mismatch findings were detected in 13/41 (32%) patients by screening for and in 28/41 (68%) patients during PSMA-RLT. NSE serum level (55.4 ± 44.6 μg/l vs. 18.5 ± 8 μg/l, p < 0.001) and ΔNSE (93.8 ± 124.5% vs. 2.9 ± 39.5%, p < 0.001) were significantly higher in the mismatch group than in the non-mismatch group. No significant differences were found for serum PSA (p = 0.424), ΔPSA (p = 0.417), serum ALP (p = 0.937), ΔALP (p = 0.611), serum GGT (p = 0.773), and ΔGGT (p = 0.971). For NSE and ΔNSE, the maximum value of the Youden index in ROC analysis was at a cut-off level of 26.8 μg/l (sensitivity 78%, specificity 96%) and at + 13.9% (sensitivity 84%, specificity 75%), respectively. An introduced scoring system of both parameters achieved a sensitivity of 90% and a specificity of 88% for the occurrence of [18F]FDG/[68Ga]Ga-PSMA-11 mismatch. Conclusion We observed a significantly higher absolute serum concentration and a higher relative increase of NSE in advanced mCRPC patients with [18F]FDG-avid and insufficient PSMA expressing metastases ([18F]FDG/[68Ga]Ga-PSMA-11 mismatch findings on PET/CT) in our cohort. NSE might be used as a potential laboratory indicator for [18F]FDG/[68Ga]Ga-PSMA-11 mismatch findings, if this observation is confirmed in future, ideally prospective, studies in larger patient cohorts.


Introduction
With over 1,000,000 new cases and approximately 300,000 deaths worldwide in 2012, prostate carcinoma is one of the most frequent malignant diseases in men [1]. A significant fraction of prostate carcinoma patients' progresses to the lethal metastasized castration-resistant prostate carcinoma (mCRPC) setting [2,3]. During the last decade, however, the development of new treatment options of men with mCRPC has led to an improved survival time [4]. In addition to chemotherapy with docetaxel or cabazitaxel [5,6] and next-generation androgen receptor signaling inhibition with abiraterone or enzalutamide [7,8], radioligand therapy targeting the prostate-specific membrane antigen (PSMA) is a potential option in palliative settings [9,10]. PSMA-targeted radioligand therapy (PSMA-RLT) with [ 177 Lu]Lu-PSMA-617 yielded impressive results in palliative settings while causing only moderate side effects [11,12]. PSMA-targeted based positron emission tomography (PET)/computer tomography (CT) with radiolabeled PSMA ligands, such as [ 68 Ga]Ga-PSMA-11, is frequently used for imaging of prostate cancer in clinical routine [13]. PSMAtargeted PET/CT is not only used for the staging of prostate cancer, but it is also a useful tool for therapy monitoring [14,15] and indispensable for verifying PSMA expression prior to PSMA-RLT [16]. High expression of PSMA is essential for successful PSMA-RLT in patients with mCRPC. However, some patients develop lesions with low or no PSMA expression under ongoing treatment [17]. Those lesions are not affected by PSMA-RLT and a change in therapy management is needed [9,18]. [ 18 F]FDG PET/ CT using 18 F-labeled fluorodeoxyglucose ([ 18 F]FDG) in addition to a PSMA-targeted PET/CT may be a suitable method for detection of those lesions. Mostly, prostate carcinoma cells have a low glucose metabolism due to energy gain by lipids and other energetic molecules but in advanced late-stage disease the glucose metabolism is highly increased by the Warburg effect and shifting to aerobic glycolysis after numerous mutation events [19]. Therefore, a combination of [ 18 [20]. Early detection of these mismatch findings is thus needed to provide patients with alternative therapy options, additionally to or instead of PSMA-RLT [18]. However, frequently performing [ 18 F]FDG PET/CT in addition to PSMA-targeted PET/CT is very cost intensive and associated with additional radiation exposure for the patient, which should be avoided even in a palliative setting.
In this study, selected serum parameters were therefore tested as indicators for the occurrence of [ 18 F]FDG/ [ 68 Ga]Ga-PSMA-11 mismatch findings on PET/CT in this study, including the prostate-specific antigen (PSA) as the routine control and response parameter [21], alkaline phosphatase (ALP) as known to be elevated by bone metastases [22], gamma-glutamyltransferase (GGT) as a parameter of liver function affected by liver metastases [23], and neuron-specific enolase (NSE) as a possible parameter for transdifferentiating to a neuroendocrine type of prostate carcinoma [24].  [26]. Before data acquisition, all patients were advised to empty their bladder. No diuretics were applied. All PET/CT scans were performed using a Biograph 40 mCT PET/CT scanner (Siemens Medical Solutions, Knoxville, TN, USA) with EANM Research Ltd. accreditation. The PET acquisition was performed from vertex to mid-femur with 3-min acquisition time per bed position. A bed position covers 21.4-cm extended field-of-view (TrueV). The PET datasets were reconstructed using an iterative 3-dimensional OSEM (ordered-subset expectation maximization) algorithm (3 iterations; 24 subsets) with Gaussian filtering and a slice thickness of 5 mm. Random correction, decay correction, scatter attenuation, and attenuation correction were applied. The CT was performed in low-dose technique using an X-ray tube voltage of 120 keV and a modulation of the tube current by applying CARE Dose4D with a maximal tube currenttime product of 30 mAs. All PET/CT data sets were visually analyzed using certified analysis software (Sectra PACS-Sectra Medical Systems GmbH, Cologne, Germany). [ 18 F]FDG and [ 68 Ga]Ga-PSMA-11 PET/CT were read simultaneously by three experienced physicians (at least 5 years of experience in PET reading) searching for mismatch findings. A mismatch finding was defined as metastasis with remarkable [ 18 F]FDG uptake and no or considerably less concordant [ 68 Ga]Ga-PSMA-11 uptake based on visual analysis. The decision of a mismatch finding was taken in consensus of all PET readers.

Statistical analysis and scoring system
For statistical analysis, Mann-Whitney U test was applied using Prism 8 (GraphPad Software, San Diego, USA) to determine significant differences between the groups. A p value < 0.05 was regarded as statistically significant. For power calculation of Mann-Whitney U test, we calculated the effect size; values higher than 0.5 were considered as a strong effect size. For each parameter that turned out to be statistically significant, a receiver operating characteristic (ROC) analysis was performed and a cut-off point was determined using the maximal value of Youden index. In addition, a scoring system combining all significant parameters was introduced. This scoring system was constructed for ease of clinical use and on the basis of sensitivity-specificity analysis of the strongest parameter, assisted by the other significant parameters, enhancing the power and accuracy of the score.

Results
In  Figure 2 shows an example of two mismatch lesions, one intensely PSMA-positive and one not identifiable on the initial [ 68 Ga]Ga-PSMA-11 PET/CT. NSE serum level (55.4 ± 44.6 μg/l vs. 18.5 ± 8 μg/l, p < 0.001) and ΔNSE (93.8 ± 124.5% vs. 2.9 ± 39.5%, p < 0.001) were significantly higher in the mismatch group than in the non-mismatch group. No significant difference was found for serum PSA (p = 0.424), ΔPSA (p = 0.417), serum ALP (p = 0.937), and ΔALP (p = 0.611) between the mismatch and the non-mismatch group, respectively. Twenty-nine of 66 (44.0%) patients had liver metastases, and 20/29 had at least one mismatch finding in the liver. In patients with liver metastases, GGT (p = 0.773) and ΔGGT (p = 0.971) also revealed no significant difference between the mismatch and the nonmismatch group. For NSE and ΔNSE, the effect size was 0.70 and 0.59, which were considered as strong. Summarized statistics of the data are presented in Table 2. All these statistical comparisons are illustrated by box-plot diagram format for each tested parameter (Fig. 3). A subgroup analysis comparing the patients diagnosed with a mismatch by screening for PSMA-RLT with the patients diagnosed with a mismatch after several cycles PSMA-RLT revealed no significant difference of serum NSE value (67.5 ± 51.7 μg/l vs. 49.9 ± 40.8 μg/l, p = 0.12). The predictive impact of NSE and ΔNSE regarding mismatch detection is discernible from the waterfall plots with highlighting of mismatch and non-mismatch individuals (Fig. 4). ROC analyses of NSE and ΔNSE Only one patient had mismatch findings with serum NSE < 15 μg/l. On the other hand, we observed several mismatch (13/41) and non-mismatch (16/25) patients with NSE in the range between 15 and 30 μg/l (Fig. 4a). To dichotomize these patients and to improve the sensitivity without marked loss of specificity we introduced a scoring system (Combined NSE Score) based on both parameters, which is presented in Fig. 6a. The main part of the scoring system was assigned to the absolute value of NSE being the strongest parameter in ROC analysis. Zero points were given for NSE in the physiological range of < 15 μg/l. For the ease of clinical use, NSE was classified in the critical range (15-30 μg/l) in steps of 5 μg/l, starting with 1 point for 15-20 μg/l up to 3 points for 25-30 μg/l covering the Youden's index (26.8 ng/ml) of the sensitivity-specificity analysis. Greater interval steps were chosen for NSE above 30 μg/l (4 points 30-50 μg/l, 5 points 50-100 μg/l and 6 points > 100 μg/l). ΔNSE was additionally included for enhancing the power and accuracy of the score. Additional points for ΔNSE was set to − 1 point when NSE was decreasing (more than − 20 %), to 0 points with stable (− 20 to + 20  Abbreviations: NSE neuron-specific enolase, PSA prostate-specific antigen, GGT gamma-glutamyltransferase, ALP alkaline phosphatase %) or unknown NSE, and to 1 (20-50%) or to 2 (> 50%) points for increasing NSE. The ROC analysis of the developed scoring system (Fig. 6b)

Discussion
We observed a significantly higher absolute serum concentration and a higher relative increase of NSE in advanced mCRPC patients with [ 18 F]FDG-avid and insufficient PSMA expressing metastases ([ 18 F]FDG/ [ 68 Ga]Ga-PSMA-11 mismatch findings) in our cohort. NSE is a highly specific marker for neurons and peripheral neuroendocrine cells, used as a biomarker for aggressive forms of neuroendocrine tumors and small cell carcinoma [27,28]. Elevated NSE serum levels have also been observed in neuroendocrine subtypes of prostate carcinoma [29,30]. While only a small proportion of prostate cancer patients represent neuroendocrine features at the beginning of disease [31], more patients develop these under therapy [24]. Serum NSE might be increased in patients with mismatch findings as a sign of dedifferentiation or transformation to a neuroendocrine subtype. This could be a mechanism of resistance induced by selective treatment pressure and was particularly observed in patients undergoing anti-androgen therapy [32][33][34]. Neuroendocrine differentiation in prostate carcinoma is a phenotypic change by which prostate cancer cells trans-differentiate into neuroendocrine-like cells. These neuroendocrine-like cells are lacking expression of androgen receptor and prostate-specific antigen. Neuroendocrine-like cells are known to produce peptide hormones and growth factors to promote tumor progression and are apoptosis-resistant contributing to treatment failure [35][36][37]. Some case reports described that PSMA-targeted PET/CT may be not able to detect neuroendocrine prostate carcinoma [38][39][40]. Supporting these observations, Bakht et al. (2018) demonstrated in vitro on patient-derived xenograft (PDX) an inverse correlation between PSMA gene (FOLH1) and neuroendocrine biomarker gene expression [41]. Particularly, suppression of the PSMA gene was observed in 65% of cases which overexpressed the NSE gene (ENO2). They also demonstrated that the most progressions' pathway to neuroendocrine trans-differentiation under ongoing  treatment was associated subsequently with loss of the PSMA expression. Thus, alternative molecular imaging methods are required for neuroendocrine phenotype. The glucose transporter GLUT1 has been found to be overexpressed on cells of the advanced late-stage prostate cancer and its expression was related to tumor aggressiveness [42]. Meziou et al. (2020) assumed that GLUT1 expression may be increased in neuroendocrine prostate carcinoma suggesting that [ 18 F]FDG PET/CT might be an important imaging tool to examine neuroendocrine prostate carcinoma [43]. Few studies demonstrated the clinical utility of [ 18 F]FDG PET/CT in neuroendocrine prostate carcinoma [44][45][46]. Spratt et al. demonstrated in 23 patients with neuroendocrine prostate cancer that [ 18 F]FDG PET/CT has clinical benefit and reported a high detection rate of metastatic disease, especially of lymph node and visceral metastases [46]. This hypothesis of transformation to prostate cancer with neuroendocrine features in our mismatch cohort was confirmed histopathologically in one patient only. More evidence is required to confirm this hypothesis. Thus, consistent histopathological examination of mismatch metastases should be performed in future studies.
Patients with mCRPC, presenting discordant [ 18 F]FDGavid lesions with low or no PSMA expression, have a poor prognosis with very short survival time [47]. This mismatch phenomenon indicates a more aggressive type of mCRPC in clinical routine and precludes patients from PSMAdirected radioligand therapy [9]. Thus, early diagnosis of mismatch findings is very important before and during PSMA-RLT to provide these patients with alternative therapy options in addition to or instead of PSMA-RLT, including chemotherapy, immunotherapy, bone-seeking radiopharmaceuticals, PARP inhibition, and other novel targeted treatments [18,47].
Our results obtained from 66 mCRPC patients suggest that serum NSE is a potential biomarker for the existence of the described [ 18 F]FDG/[ 68 Ga]Ga-PSMA-11 mismatch. Statistical analysis of our data proposed an NSE cut-off value for a mismatch occurrence of 26.8 μg/l (sensitivity 78%, specificity 96%) and of + 13.9 % increase for ΔNSE (sensitivity 84%, specificity 75%). A combination of both parameters may improve the predictive power. We achieved a sensitivity and specificity of almost 90% with our introduced Combined NSE Score (cut-off: total points ≥ 3, Fig. 6). The simplicity of the score allows easy clinical application as a guide for further diagnostic procedures in the mCRPC setting.
In contrast to NSE, other serum parameters analyzed in this study as PSA, ALP, and GGT were not able to indicate mismatch findings in our cohort. These biomarkers may only reflect total tumor burden giving no specific hint towards [ 18 F]FDG-avid lesions with low or missing PSMA expression.
We suggest including NSE assessments into routine blood tests of mCRPC patients before and during PSMA-RLT. However, these results should be noticed with caution considering a potential bias due to the retrospective Fig. 6 Combined NSE Score (a) and ROC curve for mismatch prediction by the score (b) with maximum value of the Youden index (J) study design and not-representative patient cohort. Quite often, [ 18 F]FDG PET/CT was performed when viable PSMA-negative metastases were suspected before or during PSMA-RLT due to a worsening course of disease or by hint in other imaging methods (CT, MRI). Consequently, more mismatch patients (62% in our cohort) than non-mismatch patients were included in our study, resulting in a preselected cohort of mCRPC patients not representing the normal incidence of mismatch in candidates referred for PSMA-RLT. In a prospective phase-II trial for PSMA-RLT, Hofman [20]. This was the first prospective study which excluded patients with mismatch lesions. A prospective setting including larger non-biased patient cohorts is necessary to confirm our findings. Further limitations are the missing blood samples in 11/66 (16.7%) patients prior to screening. Those patients were included in the scoring system as "unknown" ΔNSE, which reflects a frequent clinical situation in mCRPC patients referred for evaluation of PSMA-RLT. As another point of criticism, no quantitative SUV threshold defining low PSMA expression on [ 68 Ga]Ga-PSMA-11 PET/CT was set, which is also a global problem in literature suffering from defined cut-offs for adequate PSMA expression. Furthermore, additional serum parameters, which were recently reported as prognostic markers in the mCRPC setting for PSMA-RLT, such as the neuroendocrine marker chromogranin A or the lactate dehydrogenase (LDH) [48][49][50], were not available for testing in this study. Serum chromogranin A values might also be higher in patients with mismatch findings similar to NSE. LDH, which is a marker with strong prognostic value for response prediction of PSMA-RLT [48], might also be a potential marker for mismatch. Both parameters should thus also be tested in future studies. If those studies would indicate a mismatch predictive value, these parameters could be included to the scoring system to increase its sensitivity and specificity. Lastly, histopathological confirmation of a transformation process in mismatch lesions towards a neuroendocrine phenotype was performed only in one case. Histopathological examination of mismatch metastases should be included in future prospective studies.

Conclusions
We observed a significantly higher absolute serum concentration and a higher relative increase of NSE in ad-