Individual radiosensitivity reflected by γ-H2AX and 53BP1 foci predicts outcome in PSMA-targeted radioligand therapy

Purpose γ-H2AX and 53BP1 are fundamental for cellular DNA damage response (DDR) after radiation exposure and are linked to cell repair, arrest, or apoptosis. We aimed to evaluate whether DDR-markers in peripheral blood lymphocytes (PBLs) may have predictive potential for outcome in metastatic castration-resistant prostate cancer (mCRPC) patients receiving [177Lu]Lu-prostate-specific membrane antigen (PSMA) radioligand therapy (RLT). Methods We prospectively enrolled 20 men with advanced mCRPC scheduled for PSMA-targeted RLT. Prior to the first cycle of [177Lu]Lu-PSMA RLT, all patients underwent [18F]F-PSMA-1007 positron emission tomography (PET)/computed tomography (CT) for assessment of tumor PSMA expression (assessing maximum standardized uptake value (SUVmax) of all tumor lesions). Blood samples were collected prior to, + 1 h after, and + 24 h after administration of [177Lu]Lu-PSMA, and DDR-markers γ-H2AX and 53BP1 were determined in PBLs through immunocytofluorescence. We then tested the predictive performance of DDR-markers relative to clinical and PET-based parameters for progressive disease (PSA-PD) after 2 cycles. In addition, the predictive value for progression-free survival (PSA-PFS, provided as median and 95% confidence interval [CI]) was explored. Results Low baseline 53BP1 and γ-H2AX foci (P = 0.17) tended to predict early PSA-PD, whereas low SUVmax was significantly associated with higher risk for PSA-PD (P = 0.04). In Kaplan–Meier analysis, there was a trend towards prolonged PSA-PFS in patients with higher baseline 53BP1 of 6 months (mo; 95%CI, 4–9 mo) compared to 3 mo in patients with low 53BP1 (95% CI, 2–3 mo; P = 0.12). Comparable results were recorded for higher γ-H2AX expression (6 mo [95% CI, 3–9 mo] relative to 3 mo [95% CI, 2–4 mo] in patients with low γ-H2AX; P = 0.12). SUVmax, however, did not demonstrate predictive value (P = 0.29). Consistently, in univariate Cox-regression analysis, baseline 53BP1 foci demonstrated borderline significance for predicting PSA-PFS under [177Lu]Lu-PSMA RLT (P = 0.05). Conclusion In this prospective study investigating mCRPC patients undergoing [177Lu]Lu-PSMA RLT, low baseline DDR-markers in PBLs tended to predict poor outcome. Although the study group was small and results need further confirmation, these preliminary findings lay the foundation for exploring additive radiosensitizing or treatment intensification in future studies with high-risk individuals scheduled for RLT. Supplementary Information The online version contains supplementary material available at 10.1007/s00259-022-05974-8.


Introduction
Prostate-specific membrane antigen (PSMA)-targeted approaches have gained increasing acceptance for the diagnosis and treatment of patients afflicted with metastatic castration-resistant prostate cancer (mCRPC) [1][2][3]. As a type 2 transmembrane protein, PSMA is heavily overexpressed on the surface of prostate cancer cells [4], providing a target for radiolabeled imaging and therapy [1,[5][6][7][8]. For instance, [ 177 Lu]Lu-PSMA-RLT demonstrated high efficacy and safety in a first prospective phase 2 study (LuPSMA) with over half of the investigated patients demonstrating a biochemical response [prostate-specific antigen (PSA) decline of over 50%] [1]. Further building on these promising results, Hofman et al. also recently revealed a favorable outcome for patients treated with [ 177 Lu]Lu-PSMA-RLT when compared to the standard chemotherapeutic regimen using cabazitaxel [9]. Finally, the recently published prospective phase 3 VISION trial reported on longer progression-free (PFS) and overall survival for patients treated with [ 177 Lu] Lu-PSMA-RLT compared to standard care [3,10].
However, although patients had to demonstrate significant tumor PSMA expression to be enrolled into the VISION trial, 54% of the RLT-treated patients did not reach biochemical response [3]. In previous studies, high PSMA expression on pretherapeutic 68 Ga-labeled PSMA-targeted PET/ CT emerged as a strong predictor for treatment response [7,11,12]. Yet, all studies reported patients who did not respond although demonstrating increased PSMA expression [11][12][13]. In light of these results, more reliable predictors are needed. Gafita et al. recently reported on the predictive performance of nomograms containing both PET-based and clinical parameters [14], emphasizing the complexity of predicting treatment response in mCRPC patients receiving [ 177 Lu]Lu-PSMA RLT. Thus, investigating the underlying pathophysiological response caused by RLT-based radiation might help to better understand the biological mechanisms of response and non-response.
Used for PSMA-targeted treatment, 177 Lu as a primarily beta-emitting isotope induces a plurality of DNA damages, including DNA double-strand breaks (DSBs), ultimately leading to subsequent cell death [15,16]. Following DSBs, a cascade of cellular DNA damage response (DDR) mechanisms is activated, involving the histone-variant γ-H2AX and the p53-binding protein 1 (53BP1) [16][17][18]. However, beta-radiation from [ 177 Lu]Lu-PSMA does induce DSBs not only in tumor cells but also in benign cells such as peripheral blood lymphocytes (PBLs) which become exposed during circulation [19,20]. The time course of DSB formation in PBLs and the potential to obtain data on the individual dose-response relationships in vivo have recently been reported in [ 177 Lu]Lu-PSMA RLT [19,20]. Marked interindividual heterogeneity on both amount and kinetics of DNA damage response (DDR) foci during [ 177 Lu]Lu-PSMA RLT was also reported [19,20]. Given the broad range of measured γ-H2AX and 53BP1 levels potentially reflecting individual radiosensitivity, these parameters may also hold potential for risk stratification, serving as "liquid biopsy" to identify patients that most likely benefit from RLT [21,22].
In the present prospective study, we investigated the predictive performance of DSB markers relative to other established parameters, for the prediction of both early progression and progression-free survival in [ 177 Lu]Lu-PSMA RLT.

Patient population
Prior to their first treatment cycle, 20 mCRPC patients scheduled for [ 177 Lu]Lu-PSMA-I&T RLT were included between August 2020 and Mai 2021 in this prospective monocentric study for assessing the value of DDR-markers in the context of therapy (Table 1). All patients were selected for [ 177 Lu] Lu-PSMA-I&T RLT compliant to the German Medicinal Products Act, AMG §13.2b after progressing under previous treatments, including radiation therapy, androgen deprivation, and second-line antihormonal treatment, and had progressed under or were unfit for chemotherapy. In addition, all patients demonstrated PSMA-expression on pretherapeutic [ 18 F]F-PSMA-1007 PET/CT. The institutional review board

Imaging procedure and interpretation
All PET/CTs were acquired on a dedicated PET/CT system (Biograph mCT 128 Flow; Siemens Healthineers; Knoxville, TN, USA), with an extended field-of-view for the PET and a 128-slice spiral CT. Per patient, 306 ± 23 MBq of [ 18 F] F-PSMA-1007 was administered. Prior to PET acquisition, a low-dose non-enhanced helical CT (120 kV, mA modulated, pitch of 1.2, reconstructed axial slice thickness of 5.0 mm) for attenuation correction was performed. 100 min p.i. and after voiding of the bladder, whole-body PET was acquired using continuous bed motion at a speed of 0.9 mm/s for the chest and abdomen and 2.1 mm/s for the legs. All studies were reconstructed using Ultra HD (iterative algorithm with time-of-flight and point-spread function information including 2 iterations, 21 subsets; matrix, 200; zoom, 1.0; Gaussian filter, 5.0). No intravenous contrast material was injected.
For image analysis, a commercial software package enabling simultaneous and fused review of PET and CT data was used (syngo.via; V10B; Siemens Healthcare). Malignant lesions on all scans were segmented manually by one reader (L.W.) and verified by an experienced boardcertificated specialist (R.A.W.), if necessary. To this end, an isocontour volume of interest was placed on individual lesions, including all voxels above 45% of the maximum using a 3-dimensional segmentation method allowing for a semi-automatic volumetric assessment [23]. This enabled for calculation of the whole-body PSMA-TV, averaged SUV max (mean SUV max of all tumor lesions in one patient), and averaged SUV mean as described in [12].

Assessment of DDR-markers
As part of the study procedure, dedicated blood samples for the evaluation of DDR-markers were collected before administration of [ 177 Lu]Lu-PSMA-I&T, + 1 h after administration and + 24 h after administration. Blood samples were prepared as previously described in [22]. In brief, after isolating PBLs through Ficoll (GE Healthcare, Chicago, IL, USA) density-gradient from EDTA blood, PBLs from mCRPC patients were frozen in an RPMI1640 medium with 15% fetal calf serum 500 U/ml penicillin, 0.5 mg/ml streptomycin, and 2 mM l-glutamine, supplemented with 10% glycerol at − 80 °C until further use. Subsequent PBLs were centrifuged on cover glasses (Cytospin ROTANTA 460/460R centrifuge, Hettich), fixed (3% (w/v) PFA, 2% (w/v) sucrose in PBS for 10 min and at RT) and permeabilized (0.2% (v/v) Triton X-100 in PBS for 3 min). Afterwards, cells were incubated simultaneously with antibodies against Phospho (S139)-Histone H2AX (Millipore, Burlington, MA, USA) in a ratio of 1:200 and against 53BP1 (Bethyl Laboratories, Montgomery, TX, USA) in a ratio of 1:400 for 1 h followed by at least three PBS washing steps and simultaneously incubation with Alexa Fluor anti-mouse IgG 488 and Alexa Fluor anti-rabbit IgG 546 (Invitrogen, Carlsbad, CA, USA; both at a ratio of 1:250) for another 45 min. Finally, DAPI (Invitrogen) was used to counterstain DNA and cells were mounted with ProLong® Gold (Invitrogen). Further details can be found in [22].
γ-H2AX and 53BP1 foci were counted under a Leica DMI6000B microscope (Wetzlar, Germany) using a 63 × objective lens at 1.6 × magnification with a manual focus through the whole nucleus enabling 3-dimensional foci detection. DDR-markers were assessed independently in several different areas of the slide leading to a median of 64 analyzed cells per sample. Cells demonstrating apoptotic morphology or an intensely stained nucleus were excluded from the analysis. In addition, monocytes and granulocytes were excluded from the analysis according to morphological criteria as described in [22]. DDR-markers related to intact PBLs were summed up and divided by the total number of PBLs to access the number of foci per cell.

Assessment of clinical outcome
Before and 6-8 weeks after 2 cycles of RLT, PSA levels were determined and %-change to baseline was assessed. In patients discontinuing RLT, the last available PSA was used for response assessment. According to prostate cancer clinical trials working group 3 (PCWG3) criteria [26], treatment response after 2 cycles of RLT was determined. Early progressive disease (PSA-PD) was defined as a PSA increase of ≥ 25% from baseline, stable disease (PSA-SD) was defined as PSA-change between + 25% and − 50% and partial response (PSA-PR) as a PSA decrease over 50%. Patients without PSA-PD continued RLT with PSA measurements at and 4-6 weeks after each cycle. For the determination of progression-free survival (PSA-PFS), we calculated the time between the day of the first treatment cycle and the date of PSA-based PD as defined above. Patients were followed up until progression, termination of RLT due to other medical reasons, or the end of the follow-up period in December 2021.

Statistical analysis
For statistical analyses, we used GraphPad Prism 9 (Graph-Pad Software, San Diego, CA, USA) and SPSS Statistics 27 Inc. (IBM; Chicago, IL, USA). We performed simple linear regression to demonstrate the association between the two DDR-markers γ-H2AX and 53BP1. Changes in DDR-marker levels at different time points were assessed through repeated-measures one-way ANOVA. Stratifying patients into high versus low values by median univariate logistic regression analysis was performed for DDR-marker levels and PET-derived and clinical parameters to explore the relationship of parameters with PSA-PD. The predictive performance for progression-free survival (PSA-PFS) of these parameters was then further assessed in Kaplan-Meier analysis and log-rank test. Finally, univariate Cox-regression for PSA-PFS was performed using continuous variables. A P value < 0.05 was considered significant.
Representative cases are provided in Figs. 3 and 4.

Discussion
In the present prospective study, we evaluated the predictive performance of γ-H2AX and 53BP1 foci in PSMAtargeted RLT. There was a consistent trend emphasizing the predictive value of reduced DDR-markers at baseline for subsequent poor outcome under [ 177 Lu]Lu-PSMA RLT, but likely due to the small patient cohort significance could not be reached. In this regard, low baseline DDR-marker levels prior to treatment on-set probably reflected low individual radiosensitivity and thus, were linked to early PSA-PD and  shorter PSA-PFS. Thus, we herein provide for the first time data on the role of individual radiosensitivity in the context of PSMA-targeted RLT along with predictive capability, which requires further studies with larger patient populations to provide definite evidence. Nonetheless, the herein presented consistent results may also enable future studies investigating the potential of pretherapeutic administration of radiosensitizing agents, e.g., in patients having low baseline levels of γ-H2AX and 53BP1 foci. Schumann et al. were the first to investigate changes in DDR-marker levels during [ 177 Lu]Lu-PSMA RLT [27]. Enrolling 16 mCRPC patients, a distinctive time-dependent kinetic pattern of both γ-H2AX and 53BP1 foci was observed, with an early increase of both DDR-marker levels within + 4 h after administration and a subsequent decrease thereafter [27]. Ritt [20]. Of note, in both studies, DDR-marker remained elevated compared to baseline even + 24 h after initiation of RLT [20,27]. This is in line with our results also showing a time-dependent kinetic pattern of DDR-marker levels in a prospective cohort of 20 mCRPC patients. Of note, a recent study reported on the predictive value of DDR-markers for therapy response under 177 Lu-DOTATATE [22], while the present investigation is the first to evaluate the predictive performance of γ-H2AX and 53BP1 foci for men treated with [ 177 Lu]Lu-PSMA RLT. Consistent with findings of that study investigating  DDR-markers in patients with gastroenteropancreatic neuroendocrine neoplasms treated with somatostatin receptortargeted peptide receptor radionuclide therapy (PRRT), low levels of 53BP1 foci were linked to shorter PFS [22], supporting the notion of an incremental value of those DDRmarkers for patients scheduled for endoradiotherapy. Despite favorable results of PSMA-targeted RLT compared to other therapies in advanced disease [3,9], a substantial portion of patients will experience early PD, and thus, reliable predictors for subsequent treatment response would be crucial for optimizing patient care [13]. In this regard, the averaged SUV max derived from tumor lesions on pretherapeutic [ 68 Ga]Ga-PSMA PET/CTs has shown promise for predicting early response to treatment and overall survival [11,12]. However, due to various advantages such as improved image quality and longer half-life allowing for a higher throughput in a busy PET practice, recent years have witnessed an increased use of 18 F-labeled PSMA radiotracers, e.g., [ 18 F]F-PSMA-1007 [6]. Given this shift from 68 Ga to 18 F radiotracers in clinical routine, the predictive potential of 18 F-PSMA PET in patients scheduled for RLT needs further exploration. We demonstrated that a low SUV max derived from a pretherapeutic [ 18 F]F-PSMA-1007 PET/CT was significantly associated with early PSA-PD after two cycles of RLT (P = 0.04). Nonetheless, SUV max did not reach significance for the prediction of PSA-PFS in Kaplan-Meier analysis and univariate Cox-regression which might be clinically even more important than early PD for the long-term outcome of mCRPC patients. Although the study population was rather small, this may also be a hint at restricted predictive capability of 18 F-PSMA PET-based quantification. While a strong relation between uptake of PSMA-ligands and histologic PSMA-expression has already been proven for [ 68 Ga]Ga-PSMA-11 in a dedicated ex vivo animal model [28] and immunohistochemical validation of clinical primary prostate cancer samples [29], such evidence has not been provided for [ 18 F]F-PSMA-1007 to date. However, there is an increasing body of evidence of potential pitfalls on scan interpretation for this 18 F-labeled PSMA radiotracer [30]. Along with the high rate of indeterminate  (MIP, a) of a patient scheduled for RLT. Multiple PSMApositive lesions in the skeleton can be seen, e.g., in the ribs or in the left acetabulum. Averaged SUV max was 14.6 (below median, indicative of progressive disease). Pretherapeutic PSA was 523 µg/l. Fluorescence microscopy images (63 × objective lens at 1.6 × magnification), with one merged and three separate channels for γ-H2AX (red), 53BP1 (green), and DAPI for nuclei stain (blue) of DDR-markers in PBLs prior to treatment (b). This patient demonstrated a rather high number of γ-H2AX (n = 0.44) and 53BP1 (n = 0.5) foci per cell, presumably reflecting high radiosensitivity. This patient achieved partial response after 2 cycles of [. 177 Lu]Lu-PSMA RLT (c; PET: 66 days after study inclusion; PSA: 84 days after study inclusion, − 63% compared to baseline) findings on 18 F-PSMA especially in the skeleton [31] and moderate interobserver agreement rates [32], one may question the reliability of 18 F-PSMA PET-based predictors. One way to maximize the predictive performance of PETbased parameters could be a dual tracer approach for patient selection targeting both PSMA and glucose metabolism [9]. Excluding patients with PSMA-negative and simultaneously FDG-positive metastases, Hofman et al. were able to achieve biochemical response in 65% of the RLT-treated patients compared to 46% in the VISION trial, in which only PSMA PET/CTs were performed for patient selection [3,9].
In addition, blood-based markers such as AP have also shown promise for the prediction of treatment response [33,34] and we also confirmed the value of AP levels in the present investigation (P = 0.04), but inconsistent results question the suitability of this parameter to identify patients prone to early therapeutic failure [13]. For instance, Gafita et al. reported on no predictive value of AP for overall survival under [ 177 Lu]Lu-PSMA RLT [35]. As such, liquid biopsies assessing DNA damage due to ionizing radiation, either emitted through external beam radiation or internal RLT, may provide a more causal-system approach to identify patients prone to treatment failure [17].
Following RLT-induced DSBs, a cascade of biological response mechanisms is activated including the phosphorylation of the histone H2AX to γ-H2AX and the assembling of 53BP1, subsequently leading to either repair, arrest, or apoptosis of the referring cell [15][16][17]. Thus, one might speculate that low levels of the DDR-markers 53BP1 and γ-H2AX are associated with reduced responsiveness to radiation subsequently causing ineffective cell repair and less apoptosis [36,37]. In this regard Xiao et al. were able to demonstrate, that low levels of 53BP1 were associated with reduced expression of proteins fundamental for p53-dependent apoptosis such as the ataxia-telangiectasia mutated kinase and the checkpoint kinase-2 in a dedicated colorectal cancer cell line after radiation [38]. Decreased expression of 53BP1 led to radiotolerance with reduced apoptosis and elevated proliferation rates after treatment [38]. However, beta-irradiation due to 177 Lu not only induces DDR mechanisms in malignant but also in benign cells, e.g., PBLs [20,22]. Of note, the evolution of γ-H2AX and 53BP1 foci in PBLs have previously been investigated under RLT [20,27].
In the present study, we expanded the concept of bloodbased assessment of DNA damage and demonstrated clinical value of these parameters beyond dosimetry. We were able to demonstrate a clear and consistent trend emphasizing the predictive value of reduced DDR-markers assessed prior to treatment onset for subsequent poor outcome under [ 177 Lu] Lu-PSMA RLT. Future studies, however, may also measure DDR-foci in circulating tumor cells (CTCs) possibly enabling an even more accurate determination of tumor radiosensitivity [21]. However, relative to PBLs, previous singlecell studies assessing mutations in CTCs of breast cancer patients revealed remarkable heterogeneity within investigated cells from the same patient [39], and thus, DDRmarker in CTCs derived from men affected with prostate cancer may also be prone to such intra-patient heterogeneity, limiting their usefulness for prediction.
The herein presented results also support the hypothesis that low DDR-marker levels reflect low individual radiosensitivity explaining their predictive value for treatment outcome. Although other clinical parameters demonstrated similar or even superior predictive performance, e.g., AP, this is of importance as these parameters can hardly be influenced. By contrast, simultaneous chemoradiotherapy to boost radiosensitivity upon external beam radiation is widely practiced in numerous other tumor entities e.g. non-small-cell lung cancer [40]. Thus, our study provides a rationale for future studies exploring the use of radiosensitizing agents under [ 177 Lu]Lu-PSMA RLT, especially in high-risk individuals exhibiting low baseline DDR-marker levels. Among others, a combination of [ 177 Lu]Lu-PSMA RLT and docetaxel has already demonstrated high response rates with only limited side effects [41]. An even more sophisticated approach was presented by Guo et al. [42]. Conjugating docetaxel-loaded nano-particles with a PSMA-ligand achieved high tumor doses while the whole-body dose was reduced [42]. Given the herein reported findings, patients with low γ-H2AX and 53BP1 foci in PBLs at baseline may benefit from such combination approaches, e.g., by conducting a docetaxel-based radiosensitizing prior to RLT. This study has several limitations. First, although we were able to demonstrate consistent and robust results, the baseline DDR-marker closely failed to reach significance in predicting early PSA-PD and PSA-PFS due to the small number of enrolled patients in this prospective study. Therefore, future studies need to verify these results in a substantially larger cohort with a longer follow-up interval, which would then also allow to investigate overall survival. In addition, factors influencing the individual expression of DDR-markers in PBLs before RLT remain unclear and should also be investigated. Moreover, additional characterization of PBLs, e.g., CD4 + or CD8 + , might also help in better understanding radiosensitivity of specific immune cell types and their influence on treatment outcome after radiation therapy [43]. Future studies may also investigate the additional predictive role of DDR markers when combined with established outcome predictors, e.g., recently suggested nomograms [14].

Conclusion
In this prospective study investigating mCRPC patients undergoing [ 177 Lu]Lu-PSMA RLT, low baseline DDRmarkers in PBLs demonstrated a trend towards association with poorer outcome. Although the study group was small and results need further confirmation to provide definite evidence, these preliminary findings lay the foundation for exploring additive radiosensitizing or treatment intensification in future studies with high-risk individuals scheduled for RLT. included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.