Morphological versus spherical cellular geometry models: impact on dose-response of CA20948 cells to 177Lu- and 161Tb-labeled DOTA-TATE and DOTA-LM3 Authors: Kaat Spoormans, Melissa Crabbé, Lara Struelens, Michel Koole Published: 2025 ABSTRACT AbstractBackgroundAccurate cellular dosimetry is essential to investigate fundamental mechanisms of targeted radionuclide therapy. The aim of this study was to assess how morphological cellular geometry models influence cellular dosimetry estimates, in comparison to simplified spherical models that do not properly represent an adherent cell geometry.MethodsVirtual cell models of the CA20948 cell line were generated by confocal microscopy of SSTR2 and DAPI staining and served as input to derive morphological S-values for 177Lu and 161Tb. Absorbed dose–response relationships were established for [177Lu]Lu-DOTA-TATE, [161Tb]Tb-DOTA-TATE, [177Lu]Lu-DOTA-LM3 and [161Tb]Tb-DOTA-LM3 using S-values from both morphological and spherical cell geometries.ResultsThirty-four cell geometries were modeled and a spherical model with equivalent volume was generated with a radius for the cell and nucleus of 8.6(7) µm and 5.5(6) µm, respectively. Compared to spherical cell models, morphological cell models significantly changed the S-value with an increase of 13% (177Lu) and 22% (161Tb) with the cell membrane as source region and a decrease of 11% (177Lu) and 12% (161Tb) with the cytoplasm as source region. Absorbed dose-response relationships based on morphological cell geometries showed a linear dose-response model for [177Lu]Lu-DOTA-TATE and [161Tb]Tb-DOTA-TATE with α = 0.22[0.18,0.26] Gy-1, and a linear-quadratic dose-response model for [177Lu]Lu-DOTA-LM3 and [161Tb]Tb-DOTA-LM3 with α = 0.000[0.000,0.022] Gy-1 and β = 0.064[0.055,0.072] Gy-2. The assumption of a spherical cell model did not significantly affect the dose-response models, while underestimating the cell dimensions did induce a rescaling of the dose-response models.ConclusionThese findings validate the use of simplified spherical models for CA20948 cells but highlight the importance of a correct estimation of the cell dimensions. Keywords: Cell dosimetry, Morphological cell model, PRRT, SSTR2, Terbium-161, Lutetium-177, DOTA-TATE, DOTA-LM3 Introduction Radiobiological assessments within targeted radionuclide therapy (TRT) rely to a large extent on 2D in vitro evaluations. Clonogenic survival assays, for example, are key experiments to assess the radiosensitivity of different cell types to various radiation treatments. Such radiobiological assessments become even more relevant when evaluating responses in function of absorbed doses with the reliability and translatability toward in vivo or clinical scale largely depending on the accuracy of absorbed dose calculations. For example, the evaluation of the relative biological effectiveness (RBE) within TRT results in large variabilities, partly due to uncertainties in absorbed dose calculations, especially for short-range Auger and alpha particles [1]. Estimations of the absorbed dose to the cell nucleus often assume spherical cells and nuclei, however, this does not properly represent the adherent cell geometry. Tamborino et al. created morphological virtual cell models for the U2OS human osteosarcoma cell line, and thereby showed a major impact of the cell volume (inverse exponential trend) and shape modeling (up to 65%) [2]. On the contrary, it has been stated that the approximation of spherical cells should be relatively accurate for more ellipsoidal geometries [3]. This demands for further investigation. Therefore, this study created morphological cell models for adherent CA20948 cells, a rat pancreatic cell line that overexpresses the somatostatin receptor 2 (SSTR2), and that is frequently used to investigate Peptide Receptor Radionuclide Therapy (PRRT) [4]. This cell type was recently utilized to evaluate the absorbed dose-response relationship of 177Lu- and 161Tb-radiolabelled agonists (DOTA-Tyr3-octreotate (DOTA-TATE)) and antagonists (DOTA-[p-Cl-Phe-cyclo(D-Cys-Tyr-D-4-amino-Phe(carbamoyl)-Lys-Thr-Cys)D- Tyr-NH2] (DOTA-LM3)) [4]. Based on the developed morphological cell model, we calculated S-values for the radioisotopes 177Lu and 161Tb through Monte-Carlo simulations, and compared dose-response curves of 177Lu- and 161Tb-labeled DOTA-TATE and DOTA-LM3 using these morphological S-values with the dose-response curves that were determined with S-values that were based on a simplified spherical cell model. As such, the aim of this study was to determine whether simplified spherical geometries are sufficient for accurate cellular dosimetry or if morphological cell models for adherent cells are required. Materials & methods Antibody staining, imaging and segmentation CA20948 cells were seeded in an 8-well chamber slide (Ibidi, Germany), and after 24 h cells were stained with a SSTR2a primary antibody (rabbit monoclonal, Ab134152, Abcam) and Alexa fluor 568 secondary antibody (goat anti-rabbit, A11011, Thermo Fisher Scientific) to identify the cell membrane. In addition, DAPI counterstaining was used to highlight the cell nucleus, after performing fixation with 2% PFA, permeabilisation with 0.1% PBS-T and blocking with 5% bovine serum albumin. Cells were imaged using a confocal microscope (∆z = 1 μm using a Zeiss LSM 880, Cell and Tissue Imaging Core, KU Leuven). Cell membrane and cell nucleus were segmented based on the SSTRa- and DAPI-stained 3D images of the cells using 3D Slicer software (http://www.slicer.org; Fig. 1). A representative image of (A) SSTR2 (red) and DAPI (blue) staining, (B) segmentation of the cell membrane (yellow), nucleus (blue) and cytoplasm (red) and (C) 3D reconstruction of the segmentations S-value simulations The segmentation of the cell membrane, cytoplasm and cell nucleus of n = 34 randomly selected cells were used as input for Monte Carlo simulations using GATE [6] to compute morphological S-values for the absorbed dose evaluated to the cell nucleus. A minimum of 20 cells were needed to derive an average S-value with a 95% confidence interval of [S ± 10%]. The segmentations were loaded as an ImageNestedParametrisedVolume with the density of water and a dose actor was added to the cell nucleus via the DoseByRegions method. One million electrons were simulated on the cell membrane or within the cytoplasm using a voxelised source, with the energy spectrum according to the emission spectra for the β−, internal conversion (IC) and Auger electrons of 177Lu and 161Tb from ICRP107 [7]. The physics list emstandard_opt4 was used with the cut-value set to 0.01 µm. The simulation was verified by comparing a spherical segmentation, with a conventional spherical volume as geometry and source, as well as MIRDCell S-values. The average volume of the cell and cell nucleus of the morphological cell models was also calculated and used as input to calculate S-values for a spherical model of the cell and nucleus with equivalent volumes. For the spherical model, we considered either a uniform activity and the absorbed dose to the entire cell (later referred to as ’uniform sphere’) or the activity distributed on the cell membrane or in the cytoplasm and only the absorbed dose to the cell nucleus. Radiosensitivity evaluation Absorbed doses to the cell nucleus were calculated with the morphological S-values for CA20948 cells to estimate cell-radiosensitivity parameters from the clonogenic assay, performed on CA20948 cells that were treated with either [177Lu]Lu-DOTA-TATE, [161Tb]Tb- DOTA-TATE, [177Lu]Lu-DOTA-LM3 or [161Tb]Tb-DOTA-LM3, as extensively described in [5]. For the absorbed dose calculations, kinetics of cellular uptake up to 4 h, clearance of the radionuclide up to 7 d, cell binding properties determined from saturation binding assays and internalization kinetics up to 7 d were considered to calculate the contributions from self-dose, cross-dose and dose from the medium. The resulting absorbed dose-response curves were fitted to either a linear. or linear-quadratic dose-response model, depending on which model was selected by the extra-sum-of-square F-test, and a 95% confidence interval was determined for the fitting parameters. The parameter α represents the cell radiosensitivity to the radiation (i.e. single-hit cell killing events) and β accounts for superposition of radiation damage (i.e. multi-hit cell killing events). For dose-response curves with a common best-fit model, a fit to a shared curve was evaluated using the extra-sum-of-square F-test, to determine whether a difference in absorbed dose-response could be confirmed. Statistical significance was considered from p = 0.05. Results were compared with findings from a previous study [5], where spherical instead of morphological S-values were considered, to calculate the absorbed dose to the cell nucleus for generating the absorbed dose-response curves. Dose-response curves were again compared by the extra-sum-of-square F-test. Results The staining and segmentation (Fig. 1) of CA20948 cells (n = 34) resulted in an average cell and nucleus volume of 2629(663) µm3 and 689(211) µm3 with equivalent spherical radii of 8.6(7) µm and 5.5(6) µm, respectively (mean (standard deviation)). The different cell shapes and dimensions resulted in a large variation of S-values, with all S-values significantly different from S-values based on spherical geometries (Table 1; Fig. 2). The use of morphological cell models resulted in a significantly increased S-value for the cell membrane (13% for 177Lu and 22% for 161Tb) and a significantly decreased S-value for the cytoplasm (11% for 177Lu and 12% for 161Tb), compared to the use of a spherical cell model where the uniform model largely overestimated the S-values, especially for the internal conversion (IC) and Auger electrons. Table 1The average morphological self-dose S-values for CA20948 cells in µGy Bq−1 s−1 cell−1 ± standard deviation (n = 34 cells)UniformCell MembraneCytoplasmIsotope ParticleMorphologicalSpherep-valueMorphologicalSpherep-value161TbBeta199104(20)920.002133(26)1510.0004161TbIC707348(81)2820.00004463(106)5360.0004161TbAuger49510(8)40.0000340(13)330.003161TbAll1401463(105)3780.00005637(144)7200.002177LuBeta227116(23)1020.0011150(30)1710.0003177LuIC3218(3)170.01222(4)250.00015177LuAuger58.51.9(13)0.5< 0.000016(2)5.20.005177LuAll317136(26)1200.0008179(36)2010.0009 The average morphological self-dose S-values for CA20948 cells in µGy Bq−1 s−1 cell−1 ± standard deviation (n = 34 cells) The self-dose S-values for a uniform spherical (uniform activity distribution and absorbed dose evaluation) and a spherical or morphological CA20948 cell model with activity distributed either across the cell membrane (CM) or in the cytoplasm (Cyto), and the absorbed dose evaluated in the cell nucleus The corresponding absorbed dose-response curves for morphological cell geometries are shown in Fig. 3. The dose-response curves for [177Lu]Lu-DOTA-TATE and [161Tb]Tb- DOTA-TATE were best fitted with a shared (p = 0.29) linear (p = 0.9 and p = 0.4, respectively) absorbed dose-response curve with α = 0.22[0.18,0.26] Gy−1. [177Lu]Lu-DOTA-LM3 and [161Tb]Tb-DOTA-LM3 were better fitted to a shared (p = 0.7) linear-quadratic (p = 0.04 and p < 0.0001, respectively) survival curve with α = 0.000[0.000,0.022] Gy−1 and β = 0.064[0.055,0.072] Gy−2. When comparing these absorbed dose-response curve fits with those when assuming an equivalent spherical model (Fig. 3 solid vs. dashed lines), no significant difference was observed for either radiolabeled DOTA-TATE (p = 0.3) or DOTA-LM3 (p = 0.7). The clonogenic survival probability of CA20948 cells in function of the absorbed dose to the cell nucleus using morphologic cell geometries to model cell nucleus, cytoplasm and cell membrane. The corresponding fits of the linear and linear-quadratic model are shown with a solid grey line. The dashed line shows the dose-response curves when assuming a spherical cell model with an equivalent volume. The dotted line shows the original dose-response curves from previous literature when assuming a spherical cell model with underestimated dimensions [3] Discussion We created morphological virtual models for the adherent CA20948 cell geometry, and thereby obtained significantly different S-values compared to equivalent spherical cell models, with an increase of 13% (177Lu) and 22% (161Tb) for the cell membrane and a decrease of 11% (177Lu) and 12% (161Tb) for the cytoplasm. However, the impact was lower for the CA20948 cells than for the U2OS cells, for which a decrease of 37% for the cell membrane and 60% for the cytoplasm was reported for 177Lu [2]. This can be attributed to the more epithelial-like geometry of the U2OS cells, which deviates more from a spherical geometry. Furthermore, assuming a uniform activity distribution and evaluating the absorbed dose to the entire cell had an even larger impact than the cell geometry, and largely overestimated the absorbed dose, especially for the short-range Auger and IC electrons of 161Tb. The assumption of a uniform activity distribution and absorbed dose evaluation is often made at tissue level, and could induce a large overestimation of the absorbed dose, especially when working with short-range radiation types, such as pure or combined Auger emitters like 161Tb, 125I or 111In [8, 9]. We established absorbed dose-response curves using the morphological S-values in combination with cell survival and uptake data from previously published dose-response curves [5]. The latter one assumed spherical cell models with dimensions estimated using an automated cell counter (Moxi Z, Avantor, Pennsylvania, USA), which overestimated the cellular S-values compared to morphological S-values by 79% (177Lu) and 98% (161Tb) for the cell membrane and by 127% (177Lu) and 159% (161Tb) for the cytoplasm. As a result, the absorbed dose-response curves were significantly rescaled (p < 0.0001) when applying S-values from morphological cellular models, thereby increasing the radiosensitivity parameters that are estimated by fitting a linear or linear-quadratic model to the surviving fraction data. However, this difference is mainly attributed to an underestimation of the dimensions of the cell and nucleus (RC = 6 μm vs. 8.6 μm and RN = 4 μm vs. 5.5 µm3), since the assumption of a spherical cell model with equivalent volume did not significantly affect the absorbed dose-response curves. Nonetheless, the general conclusions that were drawn from dose-response evaluations remain consistent for both the small spherical and morphological cell geometry models since for both cell models no significant difference in absorbed dose-response is observed between [177Lu]Lu-DOTA-TATE and [161Tb]Tb-DOTA-TATE, and an additional quadratic response for [161Tb]Tb-DOTA-LM3 is observed compared to [161Tb]Tb-DOTA-TATE. The linear component of the dose-response curves can be interpreted as a measure for the radiosensitivity, which was found to be similar for [177Lu]Lu-DOTA-TATE and [161Tb]Tb-DOTA-TATE (α = 0.22[0.18,0.26] Gy−1) as compared to external 137Cs irradiation (α = 0.21(7) Gy−1 [10]) for CA20948 cells. This suggests that electron irradiation is equally biologically effective compared to gamma irradiation. Both treatment modalities mainly differ in their quadratic component, with β = 0.05(2) Gy−2 for external 137Cs irradiation [10] while this vanishes for [177Lu]Lu-DOTA-TATE and [161Tb]Tb-DOTA-TATE. This could be a result of the reduced dose-rate in TRT, such that sublethal damage gains more time for repair, and consequently the superposition of sublethal damages is reduced. However, these factors should be interpreted with caution since the underlying radiobiological models are taken from external beam radiotherapy, while TRT demands for more sophisticated models that are able to consider cell proliferation, dose heterogeneity, protracted low dose rate, etc. [1]. A limitation of the applied methodology is potential shrinkage of cells after fixation. This might have resulted in a slight overestimation of the calculated S-values. Furthermore, since the S-values are largely dependent on the cell size, the investigated cell type should ideally match the cell dimensions of clinically relevant cells. An even more refined cellular dosimetry model, through the application of microdosimetry, would require a more detailed subcellular localization, such as accumulation of DOTA-TATE in the Golgi apparatus, for example. Moreover, absorbed dose calculations to other radiosensitive targets, such as the cell membrane and mitochondria, will further strengthen our understanding of dose-response relations in TRT. Nonetheless, we found that the basic spherical cell model is valid for evaluating absorbed dose-response relationships, but that determining the cell dimensions are crucial for accurate cellular dosimetry. Conclusion We developed morphological geometry models for the CA20948 cell type and calculated corresponding S-values for the radioisotopes 177Lu and 161Tb. These S-values showed substantial intracellular variability and differed significantly from those derived using conventional spherical cell models. However, the use of these morphological S-values did not alter the absorbed dose-response curves when compared to the spherical models. In contrast, when the cell dimensions were underestimated, absorbed dose-response curves were rescaled, resulting in decreased estimates for the radiobiological parameters α and β that were obtained by fitting a linear quadratic model to the dose-response curves. These findings indicate the importance of accurately determining cell dimensions, but support the use of spherical cellular geometries [8, 11] as a valid simplified model for PRRT intercomparisons using the same cell line. Acknowledgement We thank the Cell and Tissue Imaging Core at KU Leuven for their support during the confocal imaging, and Amelie Coolkens for her assistance during the staining procedures. Abbreviations Targeted Radionuclide Therapy Somatostatin Receptor 2 Peptide Receptor Radionuclide Therapy Internal Conversion Author contributions KS performed the staining, confocal imaging, segmentation, S-value simulations and data analysis with support of MC. The manuscript was prepared by KS and reviewed by MC, LS and MK. Funding This research was in scope of the PhD research of KS that was funded by the Belgian nuclear research center SCK CEN. Data availability The virtual cell models are available from the corresponding author on request. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing Interests The authors declare that they have no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The virtual cell models are available from the corresponding author on request.