- Email: [email protected]
Preclinical models of patient derived xenografts (PDX) for immuno-oncology research
Poster Session – Preclinical Models, Wednesday 29 November 2016 level molecular validation in patients’ tumor cells. Primary tumor cells from ALL patients were amplified in severely immune-compromised NSG mice to generate PDX ALL cells. For genetic engineering, PDX ALL cells were lentivirally transduced in two consecutive rounds using a blue or green fluorochrome (mtagBFP or eGFP) followed by a knockdown construct. The shRNA was embedded into the context of a microRNA 30 background which allowed expressing the shRNA directly linked to a transgenic marker. The two transfections were performed such that blue cells expressed the control knockdown construct, while green cells expressed knockdown of XIAP so that control cells and XIAP knockdown cells could be distinguished by flow cytometry. This setting allowed performing competitive in vivo proliferation assays in which both cell populations could be simultaneously monitored in the same mouse. Results: Our knockdown strategy reduced XIAP expression by well more than 90% in ALL PDX cells which remained stable over several months of passaging in mice. We found that PDX ALL cells with knockdown of XIAP suffered a severe growth disadvantage over control cells. Blue control cells significantly overgrew green cells with XIAP knockdown. The inhibitory effect of XIAP knockdown was dose-dependent as PDX ALL cells with high expression of the knockdown marker showed a more pronounced growth inhibition than cells with low marker expression. In one sample, cells with high knockdown of XIAP completely vanished after 6 weeks. Conclusion: We established a novel technique enabling molecular target validation in patients’ tumor cells in vivo by combining efficient gene silencing in PDX cells with convenient readout systems. This novel approach enables prioritizing putative therapeutic targets on a convenient, molecular and reliable level. XIAP plays an essential role in ALL and ALL depends on XIAP for in vivo growth. XIAP represents a promising target for treatment of ALL and novel inhibitors should target XIAP to treat patients with ALL. No conflict of interest. 164 Poster (Board P135) Development and characterization of HER2+ T-DM1-resistant breast cancer PDX models M. Wick1 , A. Moriarty1 , M. Quinn1 , T. Vaught1 , M. Rundle1 , A. Tolcher1 , D. Rasco1 , A. Patnaik1 , K. Papadopoulos1 . 1 START, Preclinical Research, San Antonio, USA Background: Ado-trastuzumab emtansine (T-DM1) is an antibody–drug conjugate (ADC) approved for treatment of high HER2 expressing (3+), trastuzumab resistant breast cancers. While this agent is initially effective, resistance often develops. To aid in developing new therapies for HER2+ T-DM1-resistant breast cancer and better understand resistance mechanisms, we developed two T-DM1-resistant breast PDX models designated ST1616B/TDR and ST1360B/TDR in athymic nude mice and characterized both for receptor expression, genomic alterations and in vivo drug sensitivity. Materials and Methods: ST1616B/TDR and ST1360B/TDR models were developed by chronic in vivo drug conditioning of ST1616B and ST1360B parent models over several passages; resistance was confirmed with weekly high dose drug injections. Resulting models were characterized by immunohistochemistry, RNA in situ hybridization and NGS and tested in vivo towards several targeted therapies including trastuzumab, pertuzumab and T-DM1 and results compared with data from parent and other HER2+ PDX models. In vivo study endpoints included tumor volume and time from treatment initiation with T/C values and tumor regression reported at study completion. Results: Both TDR models retained high HER2 expression and DNAbased characteristics of their respective parent models. ST1616B/TDR and ST1360B/TDR were found resistant to chronic single agent treatment of trastuzumab, pertuzumab or T-DM1. Interestingly co-administration of pertuzumab and either trastuzumab or T-DM1 restored sensitivity of either agent towards the ST1360B/TDR model. Conclusion: We have established and characterized two HER2+ T-DM1resistant breast PDX models one of which can be resensitized with co-administration of pertuzumab. These models are valuable tools in developing new therapies to T-DM1 resistance. No conflict of interest. 165 Poster (Board P136) The Charles River PDX Compendium: A database of wellcharacterized PDX models with molecular and drug sensitivity profiles for preclinical studies B. Zeitouni1 , M. Landesfeind1 , A.L. Peille1 , M. Weidner1 , P. Bronsert2 , T. Giesemann1 , J. Schueler1 , T. Metz1 , V. Vuaroqueaux1 . 1 Charles River Laboratories, Oncotest GmbH, Drug Discovery, Freiburg, Germany; 2 ¨ Universitatsklinikum Freiburg, Pathologie, Freiburg, Germany Patient-derived xenograft tumor (PDX) models are relevant for anti-cancer agent testing because they are useful for predicting drug response in
Poster abstracts
S59
patients. The availability of large collections of such models has become necessary in helping forecast responses in clinical trials. In addition, the molecular characterization of models is crucial for adequate tumor selection prior to drug testing and for predictive biomarker discovery. Here, we present the current status of a database with more than 400 PDX covering major cancer types. These PDX have been extensively characterized for histological features, molecular data, and for tumor growth with sensitivity to relevant standard-of-care compounds and targeted therapies. A total of 442 PDX growing in immunocompromised mice have been successfully established from patient tumors from diverse cancer types. A histology review of stroma content, differentiation and vasculature level was performed on all PDX models and a tissue microarray (TMA) was prepared as a “compendium-associated screening tool” for on-demand target investigation. PDX models were profiled by Affymetrix HGU133 plus 2.0 microarrays or RNA-Seq for gene expression, by Affymetrix SNP 6.0 chips for chromosomal gains/losses, and by whole-exome sequencing for point mutations and insertions-deletions. Raw data were analyzed by PDXspecific bioinformatics pipelines, and processed data were integrated into a growing database for tumor selection, data visualization and analytics. The PDX compendium is enriched in the most common tumor types including 100 lung (23%), 64 colorectal (14%), 48 pancreatic (11%), 35 renal (8%), 29 Asian gastric (7%) and 28 melanoma (6%) cancers. Histological review highlighted the persistence of the overall patient tumor morphology, as well as the diversity of stroma content (mainly related to the cancer types), vascularization grade and tumor differentiation stage. Molecular data analyses showed typical gene expression patterns and copy-number alteration/mutation landscapes with APC (57%) being found highly mutated in colorectal models, KRAS (83%) in pancreatic models, or BRAF (57%) with CDKN2A copy-number losses (39%) in melanoma models. All these data revealed the strengths of the collection particularly in the maintenance of typical histological features and the observation of known biomarkers found in patient tumors, but also its limitations including the difficulty to establish hormone-dependent PDX models. The availability of well-characterized preclinical tumor models with histological and molecular data improves the selection of appropriate models for in vivo/in vitro drug testing and is essential to increase the success rate of preclinical studies and facilitate predictive biomarker discovery. No conflict of interest. 166 Poster (Board P137) Preclinical models of patient derived xenografts (PDX) for immuno-oncology research M. Stecklum1 , A. Wulf-Goldenberg1 , B. Brzezicha2 , I. Fichtner3 , J. Hoffmann3 . 1 EPO GmbH Berlin-Buch, Humanized mice, Berlin, Germany; 2 EPO GmbH Berlin-Buch, PDX models, Berlin, Germany; 3 EPO GmbH Berlin-Buch, Experimental Oncology, Berlin, Germany Objective: The preclinical evaluation of novel immune checkpoint modulators is dependent on tumor models in mice with functional immune cells. In previous experiments, we have demonstrated, that hematopoietic stem cells (HSC) can proliferate and differentiate in vivo to form a functional humanized immune system. Further we have shown engraftment of PDX on these humanized mice. In the next step, we determined PD-L1 expression as a target for immunotherapy on different tumor cell lines and PDX and evaluated the functionality of the humanized mice by the treatment with the checkpoint inhibitors ipilimumab and nivolumab. Methods: For humanization of mice HSC from cord blood were transplanted intravenously into 3 week old irradiated nod scid gamma mice. Tumor cell lines and PDX were screened for PD-L1 expression on RNA level by PCR and protein level by FACS, western blot and immunohistochemistry. PD-L1 positive and negative tumor cell lines and PDX models were subcutaneously co-transplanted into these humanized mice and followed for growth. Response to the CTLA-4 checkpoint inhibitor ipilimumab and PD-1 checkpoint inhibitor nivolumab was evaluated. Tumor and organs (bone marrow, spleen, thymus) were investigated at the end of the experiments by FACS and exemplary by immunhistochemistry for T cells and other immune cells and the expression of PD-1, CTLA-4 and PD-L1. Results: All investigated cell lines (e.g. MDA-MB-231, ES-2, Colo205) and PDX models (e.g. colon, head and neck, melanoma, lung) engrafted successfully on humanized mice. 12 from 14 PDX models showed no difference in tumor growth compared to non-humanized mice. These PDX models seem not to be recognized by the remaining innate immune response of the mice and tolerated by the human immune cells. Treatment with ipilimumab or nivolumab alone or in combination delayed the tumor growth, accompanied by an increase of T-cells in blood and tumor. PDX tumors with a reduced growth on humanized mice (due to innate immune response) showed a stronger response to the checkpoint inhibitors. Conclusion: HSC can be transplanted in immunodeficient mice and establish a “functional” human immune system. Tumor cell lines and PDX
S60
Poster abstracts
Poster Session – Radiation Interactive Agents, Wednesday 29 November 2016
models can be successfully engrafted on humanized mice − generating a fully human preclinical test system for immuno-oncology. No conflict of interest. 167 Poster (Board P138) Radiation-induced lung fibrosis is associated with interstitial M2 macrophages and hybrid alveolar macrophages L. Meziani1 , M. Mondini1 , B. Petit2 , M.C. Vozenin2 , E. Deutsch1 . 1 ´ Gustave Roussy Cancer Campus, Radiotherapie, villejuif, France; 2 ´ CHUV, Oncologie/Radiotherapie, Lausanne, Switzerland Background: Radiation-induced fibrosis (RIF) is a delayed complication of radiotherapy often associated with chronic inflammatory process and macrophage infiltration. Nowadays, macrophages are suggested to be important cellular contributors to fibrogenic process, but their implication in the context of RIF is not well known. Material and Methods: To investigate the role of macrophages in RIF we have used a classical experimental model of lung fibrosis developed in C57Bl/6 mice after 16 Gy thorax-IR. Results: We confirmed the fact that total lung irradiation at 16 Gy (IR) induces an interstitial fibrosis associated with delayed recruitment of pulmonary macrophages. We then profiled both alveolar macrophages (AM) and interstitial macrophages (IM) during the various steps of the fibrogenic process. We found a transient depletion of AM associated with cytokine secretion during the acute post-IR phase (15 days), followed by an active repopulation and an enhanced number of AM during the late post-IR phase (20 weeks). Interestingly, AM were mostly recruited from the bone marrow and exhibit a hybrid polarization (M1/M2) associated with up-regulation of Th1 and Th2 cytokines. The number of M2-polarized IM significantly increased during the late time points after irradiation and a down-regulation of Th1 cytokine was measured in tissue lysate. Conclusions: These results suggest a differential contribution of hybrid AM vs M2-IM to fibrogenesis. Interestingly, in contrast to activated hybrid AM, activated M2-IM were able to induce fibroblast activation in vitro mediated by an enhanced TGF-b1 expression suggesting a profibrotic role of M2-IM. Specific depletion of hybrid AM using intranasal administration of clodrosome increased RIF score and enhanced M2-IM infiltration suggesting a protective role of hybrid AM. These present study shows a dual and opposite contribution of alevolar versus intertitial macrophages in RIF and identify M2-IM as a potential therapeutic target to treat RIF. No conflict of interest.
Radiation Interactive Agents 168 Poster (Board P139) Validation of 68 Ga-HBED-CC PSMA-PET/CT and multiparametric MRI for gross tumor volume delineation in patients with primary prostate cancer based on comparison with histology reference C. Zamboglou1 , V. Drendel2 , C.A. Jilg3 , H.C. Rischke1 , T.I. Beck4 , T. Krauss5 , F. Schiller4 , P.T. Meyer4 , A.L. Grosu1 . 1 Medical Center, University of Freiburg, Department of Radiation Oncology, Freiburg, Germany; 2 Medical Center, University of Freiburg, Department of Pathology, Freiburg, Germany; 3 Medical Center, University of Freiburg, Department of Urology, Freiburg, Germany; 4 Medical Center, University of Freiburg, Department of Nuclear Medicine, Freiburg, Germany; 5 Medical Center, University of Freiburg, Department of Radiology, Freiburg, Germany Background: The accurate detection and display of intraprostatic tumor amount is important for diagnosis and treatment planning in patients with primary prostate cancer (PCa). Most of the ongoing studies use multiparametric magnetic resonance imaging (mpMRI) for detection of PCa and to define the gross tumor volume (GTV) for focal dose escalation with radiation therapy. However, recent studies validated radiolabelled tracers targeting the prostate-specific membrane antigen (PSMA) and reported a good performance. The aim of this study was to compare 68 Ga-HBED-CCPSMA PET/CT with multiparametric MRI for diagnosis and GTV-delineation in patients with primary PCa based on correlation with histology reference. Methodology: Seven patients with biopsy-proven primary PCa underwent 68 Ga-HBED-CC-PSMA PET/CT and MRI followed by radical prostatectomy. Resected prostates were scanned by ex-vivo CT using a special localizer and histopathologically prepared. PCa tissue in histology was processed to obtain GTV-histo and digitalized on ex-vivo CT. Ex-vivo CT including GTV-histo and mpMRI data were matched to in-vivo CT. The overlap between GTV-histo and consensus contours based on MRI (GTV-MRI), PSMA PET (GTV-PET) or the combination of both (GTV-union) was
measured. Sorensen-Dice coefficient (DSC) was calculated for each patient. Furthermore, the prostate in each in-vivo CT slice was separated into four equal segments and sensitivity and specificity was calculated based on histology reference. Results: 225 of 340 segments (66.2%) included GTV-histo. Sensitivity and specificity for GTV-union, GTV-PET and GTV-MRI were 82% and 67%, 75% and 87%, 70% and 82%, respectively. GTV-histo had a significant higher overlap with GTV-union compared to GTV-MRI (p = 0.016) and GTVPET (p = 0.016). The mean DSC for GTV-union, GTV-PET and GTV-MRI was 0.51 (±18) 0.45 (±0.17) and 0.48 (±0.19), respectively. Conclusion: Our study supports the combined usage of mpMRI and 68 GaHBED-CC-PSMA PET/CT for diagnosis and GTV-delineation in primary PCa. No conflict of interest. 169 Poster (Board P140) Dynamics in 18-fluoromisonidazole PET/CT and perfusion-weighted 3-Tesla MRI parameters as biomarkers for predicting treatment outcome in HNSCC H. Bunea1 , A. Bunea1 , N. Wiedenmann1 , C. Stoykow2 , M. Mix2 , L. Majerus1 , U. Ludwig3 , H.C. Rischke1 , M. Langer4 , P.T. Meyer2 , M. Bock3 , A.L. Grosu1 . 1 University Medical Center Freiburg, Radiation Oncology, Freiburg, Germany; 2 University Medical Center Freiburg, Nuclear Medicine, Freiburg, Germany; 3 University Medical Center Freiburg, Medical Physics, Freiburg, Germany; 4 University Medical Center Freiburg, Radiology, Freiburg, Germany Introduction: Tumor hypoxia in squamous cell carcinoma of the head and neck (HNSCC) is associated with poor prognosis. Reoxygenation as a result of treatment leads to increased radiosensitivity. Early adaptive radiotherapy planning may allow for a more individualized treatment. Therefore, in the following study, the dynamics of hypoxia during chemoradiation (RCTx) is detected with FMISO PET/CT and correlated to perfusion MRI parameters. Perfusion-weighted MRI parameters can be correlated with tumor hypoxia and thereby have the potential to serve as predictors of treatment failure. In particular, the volume transfer constant between plasma and interstitial space Ktrans is an indirect measure of the capillary permeability and blood flow. A subsequent rise of Ktrans, ve during RCTx is associated with a good response to treatment. Methods: A prospective serial imaging study was conducted in patients undergoing definitive RCTx (70 Gy, concomitant cisplatin) for HNSCC: in weeks 0, 2 and 5 3T-MRI and FMISO PET were acquired. Tumor hypoxia was assessed in FMISO PET 2.5 h p.i. Gross tumor volume in MRI (GTVMRI) was defined as the area of high signal on T2-weighted images using the T1-weighted images for anatomic cross reference. Perfusion parameters Ktrans and ve were calculated from a dynamic T1-weighted study after contrast agent injection. MRI and PET scans were matched using iPlan contouring tool (v. 3.0.0, BrainLAB AG). Hypoxic subvolume (HSV) of GTVMRI was defined after normalization to the FMISO background in the contralateral sternocleidomastoid muscle, thresholded by 1.4. Volumetric parameters between weeks 0, 2 and 5 were compared and related to treatment response in terms of local recurrence (LR) and stable disease (SD). Statistical analysis was done with Spearman correlation. Before t-test analysis, normal sample distribution was confirmed with Shapiro–Wilk test. Results: Between 2014 and 2015 10 male patients, treated for HNSCC with RCTx, were included. All patients received a total dose of 70 Gy. In total, 30 FMISO-PET/CT data sets and 27 MRI data sets were obtained. Mean follow up (FU) was 14.6 months (4−28 months). In weeks 0−5, patients with LR showed a mean Ktrans-decrease of 19%, whereas in weeks 0−2 an increase of SUVmax (57%) was shown. Patients with SD showed Ktrans-increase (36%) and SUVmax-decrease (−61%). HSV diminished in all patients. The correlation analysis was significant between D GTVMRI and D Ktrans in week 0−2 (p = 0.037) and between D SUVmax (week 0−5) and D Ktrans (week 0−2), p = 0.045. Conclusion: As was previously shown we conclude that changes in SUVmax are crucial in week 2. In our limited patient cohort and the short FU, we found that a decrease in Ktrans might indicate a poorer outcome. Finding markers in bioimaging may allow individualization of treatment by dose painting and adaptive radiotherapy. No conflict of interest.
Poster abstracts
S59
patients. The availability of large collections of such models has become necessary in helping forecast responses in clinical trials. In addition, the molecular characterization of models is crucial for adequate tumor selection prior to drug testing and for predictive biomarker discovery. Here, we present the current status of a database with more than 400 PDX covering major cancer types. These PDX have been extensively characterized for histological features, molecular data, and for tumor growth with sensitivity to relevant standard-of-care compounds and targeted therapies. A total of 442 PDX growing in immunocompromised mice have been successfully established from patient tumors from diverse cancer types. A histology review of stroma content, differentiation and vasculature level was performed on all PDX models and a tissue microarray (TMA) was prepared as a “compendium-associated screening tool” for on-demand target investigation. PDX models were profiled by Affymetrix HGU133 plus 2.0 microarrays or RNA-Seq for gene expression, by Affymetrix SNP 6.0 chips for chromosomal gains/losses, and by whole-exome sequencing for point mutations and insertions-deletions. Raw data were analyzed by PDXspecific bioinformatics pipelines, and processed data were integrated into a growing database for tumor selection, data visualization and analytics. The PDX compendium is enriched in the most common tumor types including 100 lung (23%), 64 colorectal (14%), 48 pancreatic (11%), 35 renal (8%), 29 Asian gastric (7%) and 28 melanoma (6%) cancers. Histological review highlighted the persistence of the overall patient tumor morphology, as well as the diversity of stroma content (mainly related to the cancer types), vascularization grade and tumor differentiation stage. Molecular data analyses showed typical gene expression patterns and copy-number alteration/mutation landscapes with APC (57%) being found highly mutated in colorectal models, KRAS (83%) in pancreatic models, or BRAF (57%) with CDKN2A copy-number losses (39%) in melanoma models. All these data revealed the strengths of the collection particularly in the maintenance of typical histological features and the observation of known biomarkers found in patient tumors, but also its limitations including the difficulty to establish hormone-dependent PDX models. The availability of well-characterized preclinical tumor models with histological and molecular data improves the selection of appropriate models for in vivo/in vitro drug testing and is essential to increase the success rate of preclinical studies and facilitate predictive biomarker discovery. No conflict of interest. 166 Poster (Board P137) Preclinical models of patient derived xenografts (PDX) for immuno-oncology research M. Stecklum1 , A. Wulf-Goldenberg1 , B. Brzezicha2 , I. Fichtner3 , J. Hoffmann3 . 1 EPO GmbH Berlin-Buch, Humanized mice, Berlin, Germany; 2 EPO GmbH Berlin-Buch, PDX models, Berlin, Germany; 3 EPO GmbH Berlin-Buch, Experimental Oncology, Berlin, Germany Objective: The preclinical evaluation of novel immune checkpoint modulators is dependent on tumor models in mice with functional immune cells. In previous experiments, we have demonstrated, that hematopoietic stem cells (HSC) can proliferate and differentiate in vivo to form a functional humanized immune system. Further we have shown engraftment of PDX on these humanized mice. In the next step, we determined PD-L1 expression as a target for immunotherapy on different tumor cell lines and PDX and evaluated the functionality of the humanized mice by the treatment with the checkpoint inhibitors ipilimumab and nivolumab. Methods: For humanization of mice HSC from cord blood were transplanted intravenously into 3 week old irradiated nod scid gamma mice. Tumor cell lines and PDX were screened for PD-L1 expression on RNA level by PCR and protein level by FACS, western blot and immunohistochemistry. PD-L1 positive and negative tumor cell lines and PDX models were subcutaneously co-transplanted into these humanized mice and followed for growth. Response to the CTLA-4 checkpoint inhibitor ipilimumab and PD-1 checkpoint inhibitor nivolumab was evaluated. Tumor and organs (bone marrow, spleen, thymus) were investigated at the end of the experiments by FACS and exemplary by immunhistochemistry for T cells and other immune cells and the expression of PD-1, CTLA-4 and PD-L1. Results: All investigated cell lines (e.g. MDA-MB-231, ES-2, Colo205) and PDX models (e.g. colon, head and neck, melanoma, lung) engrafted successfully on humanized mice. 12 from 14 PDX models showed no difference in tumor growth compared to non-humanized mice. These PDX models seem not to be recognized by the remaining innate immune response of the mice and tolerated by the human immune cells. Treatment with ipilimumab or nivolumab alone or in combination delayed the tumor growth, accompanied by an increase of T-cells in blood and tumor. PDX tumors with a reduced growth on humanized mice (due to innate immune response) showed a stronger response to the checkpoint inhibitors. Conclusion: HSC can be transplanted in immunodeficient mice and establish a “functional” human immune system. Tumor cell lines and PDX
S60
Poster abstracts
Poster Session – Radiation Interactive Agents, Wednesday 29 November 2016
models can be successfully engrafted on humanized mice − generating a fully human preclinical test system for immuno-oncology. No conflict of interest. 167 Poster (Board P138) Radiation-induced lung fibrosis is associated with interstitial M2 macrophages and hybrid alveolar macrophages L. Meziani1 , M. Mondini1 , B. Petit2 , M.C. Vozenin2 , E. Deutsch1 . 1 ´ Gustave Roussy Cancer Campus, Radiotherapie, villejuif, France; 2 ´ CHUV, Oncologie/Radiotherapie, Lausanne, Switzerland Background: Radiation-induced fibrosis (RIF) is a delayed complication of radiotherapy often associated with chronic inflammatory process and macrophage infiltration. Nowadays, macrophages are suggested to be important cellular contributors to fibrogenic process, but their implication in the context of RIF is not well known. Material and Methods: To investigate the role of macrophages in RIF we have used a classical experimental model of lung fibrosis developed in C57Bl/6 mice after 16 Gy thorax-IR. Results: We confirmed the fact that total lung irradiation at 16 Gy (IR) induces an interstitial fibrosis associated with delayed recruitment of pulmonary macrophages. We then profiled both alveolar macrophages (AM) and interstitial macrophages (IM) during the various steps of the fibrogenic process. We found a transient depletion of AM associated with cytokine secretion during the acute post-IR phase (15 days), followed by an active repopulation and an enhanced number of AM during the late post-IR phase (20 weeks). Interestingly, AM were mostly recruited from the bone marrow and exhibit a hybrid polarization (M1/M2) associated with up-regulation of Th1 and Th2 cytokines. The number of M2-polarized IM significantly increased during the late time points after irradiation and a down-regulation of Th1 cytokine was measured in tissue lysate. Conclusions: These results suggest a differential contribution of hybrid AM vs M2-IM to fibrogenesis. Interestingly, in contrast to activated hybrid AM, activated M2-IM were able to induce fibroblast activation in vitro mediated by an enhanced TGF-b1 expression suggesting a profibrotic role of M2-IM. Specific depletion of hybrid AM using intranasal administration of clodrosome increased RIF score and enhanced M2-IM infiltration suggesting a protective role of hybrid AM. These present study shows a dual and opposite contribution of alevolar versus intertitial macrophages in RIF and identify M2-IM as a potential therapeutic target to treat RIF. No conflict of interest.
Radiation Interactive Agents 168 Poster (Board P139) Validation of 68 Ga-HBED-CC PSMA-PET/CT and multiparametric MRI for gross tumor volume delineation in patients with primary prostate cancer based on comparison with histology reference C. Zamboglou1 , V. Drendel2 , C.A. Jilg3 , H.C. Rischke1 , T.I. Beck4 , T. Krauss5 , F. Schiller4 , P.T. Meyer4 , A.L. Grosu1 . 1 Medical Center, University of Freiburg, Department of Radiation Oncology, Freiburg, Germany; 2 Medical Center, University of Freiburg, Department of Pathology, Freiburg, Germany; 3 Medical Center, University of Freiburg, Department of Urology, Freiburg, Germany; 4 Medical Center, University of Freiburg, Department of Nuclear Medicine, Freiburg, Germany; 5 Medical Center, University of Freiburg, Department of Radiology, Freiburg, Germany Background: The accurate detection and display of intraprostatic tumor amount is important for diagnosis and treatment planning in patients with primary prostate cancer (PCa). Most of the ongoing studies use multiparametric magnetic resonance imaging (mpMRI) for detection of PCa and to define the gross tumor volume (GTV) for focal dose escalation with radiation therapy. However, recent studies validated radiolabelled tracers targeting the prostate-specific membrane antigen (PSMA) and reported a good performance. The aim of this study was to compare 68 Ga-HBED-CCPSMA PET/CT with multiparametric MRI for diagnosis and GTV-delineation in patients with primary PCa based on correlation with histology reference. Methodology: Seven patients with biopsy-proven primary PCa underwent 68 Ga-HBED-CC-PSMA PET/CT and MRI followed by radical prostatectomy. Resected prostates were scanned by ex-vivo CT using a special localizer and histopathologically prepared. PCa tissue in histology was processed to obtain GTV-histo and digitalized on ex-vivo CT. Ex-vivo CT including GTV-histo and mpMRI data were matched to in-vivo CT. The overlap between GTV-histo and consensus contours based on MRI (GTV-MRI), PSMA PET (GTV-PET) or the combination of both (GTV-union) was
measured. Sorensen-Dice coefficient (DSC) was calculated for each patient. Furthermore, the prostate in each in-vivo CT slice was separated into four equal segments and sensitivity and specificity was calculated based on histology reference. Results: 225 of 340 segments (66.2%) included GTV-histo. Sensitivity and specificity for GTV-union, GTV-PET and GTV-MRI were 82% and 67%, 75% and 87%, 70% and 82%, respectively. GTV-histo had a significant higher overlap with GTV-union compared to GTV-MRI (p = 0.016) and GTVPET (p = 0.016). The mean DSC for GTV-union, GTV-PET and GTV-MRI was 0.51 (±18) 0.45 (±0.17) and 0.48 (±0.19), respectively. Conclusion: Our study supports the combined usage of mpMRI and 68 GaHBED-CC-PSMA PET/CT for diagnosis and GTV-delineation in primary PCa. No conflict of interest. 169 Poster (Board P140) Dynamics in 18-fluoromisonidazole PET/CT and perfusion-weighted 3-Tesla MRI parameters as biomarkers for predicting treatment outcome in HNSCC H. Bunea1 , A. Bunea1 , N. Wiedenmann1 , C. Stoykow2 , M. Mix2 , L. Majerus1 , U. Ludwig3 , H.C. Rischke1 , M. Langer4 , P.T. Meyer2 , M. Bock3 , A.L. Grosu1 . 1 University Medical Center Freiburg, Radiation Oncology, Freiburg, Germany; 2 University Medical Center Freiburg, Nuclear Medicine, Freiburg, Germany; 3 University Medical Center Freiburg, Medical Physics, Freiburg, Germany; 4 University Medical Center Freiburg, Radiology, Freiburg, Germany Introduction: Tumor hypoxia in squamous cell carcinoma of the head and neck (HNSCC) is associated with poor prognosis. Reoxygenation as a result of treatment leads to increased radiosensitivity. Early adaptive radiotherapy planning may allow for a more individualized treatment. Therefore, in the following study, the dynamics of hypoxia during chemoradiation (RCTx) is detected with FMISO PET/CT and correlated to perfusion MRI parameters. Perfusion-weighted MRI parameters can be correlated with tumor hypoxia and thereby have the potential to serve as predictors of treatment failure. In particular, the volume transfer constant between plasma and interstitial space Ktrans is an indirect measure of the capillary permeability and blood flow. A subsequent rise of Ktrans, ve during RCTx is associated with a good response to treatment. Methods: A prospective serial imaging study was conducted in patients undergoing definitive RCTx (70 Gy, concomitant cisplatin) for HNSCC: in weeks 0, 2 and 5 3T-MRI and FMISO PET were acquired. Tumor hypoxia was assessed in FMISO PET 2.5 h p.i. Gross tumor volume in MRI (GTVMRI) was defined as the area of high signal on T2-weighted images using the T1-weighted images for anatomic cross reference. Perfusion parameters Ktrans and ve were calculated from a dynamic T1-weighted study after contrast agent injection. MRI and PET scans were matched using iPlan contouring tool (v. 3.0.0, BrainLAB AG). Hypoxic subvolume (HSV) of GTVMRI was defined after normalization to the FMISO background in the contralateral sternocleidomastoid muscle, thresholded by 1.4. Volumetric parameters between weeks 0, 2 and 5 were compared and related to treatment response in terms of local recurrence (LR) and stable disease (SD). Statistical analysis was done with Spearman correlation. Before t-test analysis, normal sample distribution was confirmed with Shapiro–Wilk test. Results: Between 2014 and 2015 10 male patients, treated for HNSCC with RCTx, were included. All patients received a total dose of 70 Gy. In total, 30 FMISO-PET/CT data sets and 27 MRI data sets were obtained. Mean follow up (FU) was 14.6 months (4−28 months). In weeks 0−5, patients with LR showed a mean Ktrans-decrease of 19%, whereas in weeks 0−2 an increase of SUVmax (57%) was shown. Patients with SD showed Ktrans-increase (36%) and SUVmax-decrease (−61%). HSV diminished in all patients. The correlation analysis was significant between D GTVMRI and D Ktrans in week 0−2 (p = 0.037) and between D SUVmax (week 0−5) and D Ktrans (week 0−2), p = 0.045. Conclusion: As was previously shown we conclude that changes in SUVmax are crucial in week 2. In our limited patient cohort and the short FU, we found that a decrease in Ktrans might indicate a poorer outcome. Finding markers in bioimaging may allow individualization of treatment by dose painting and adaptive radiotherapy. No conflict of interest.