Background Accurate predictions of tumor dissemination risks and medical treatment outcomes are critical to personalize therapy. Patient-derived xenograft (PDX) models in mice have demonstrated high accuracy in predicting therapeutic outcomes, but methods for predicting tumor invasiveness and early stages of vascular/lymphatic dissemination are still lacking. Here we show that a zebrafish tumor xenograft (ZTX) platform based on implantation of PDX tissue fragments recapitulate both treatment outcome and tumor invasiveness/dissemination in patients, within an assay time of only 3 days. Methods Using a panel of 39 non-small cell lung cancer PDX models, we developed a combined mouse-zebrafish PDX platform based on direct implantation of cryopreserved PDX tissue fragments into zebrafish embryos, without the need for pre-culturing or expansion. Clinical proof-of-principle was established by direct implantation of tumor samples from four patients. Results The resulting ZTX models responded to Erlotinib and Paclitaxel, with similar potency as in mouse-PDX models and the patients themselves, and resistant tumors similarly failed to respond to these drugs in the ZTX system. Drug response was coupled to elevated expression of EGFR, Mdm2, Ptch1 and Tsc1 (Erlotinib), or Nras and Ptch1 (Paclitaxel) and reduced expression of Egfr, Erbb2 and Foxa (Paclitaxel). Importantly, ZTX models retained the invasive phenotypes of the tumors and predicted lymph node involvement of the patients with 91% sensitivity and 62% specificity, which was superior to clinically used tests. The biopsies from all four patient tested implanted successfully, and treatment outcome and dissemination were quantified for all patients in only 3 days. Conclusions We conclude that the ZTX platform provide a fast, accurate, and clinically relevant system for evaluation of treatment outcome and invasion/dissemination of PDX models, providing an attractive platform for combined mouse-zebrafish PDX trials and personalized medicine.

Lung cancer is one of the most deadly diseases accounting for 24% of all cancer deaths worldwide. One reason for this high mortality is the high interindividual heterogeneity but generally poor efficacy of current treatments, leading to an urgent need for new and more effective drugs. Understanding the individual variability in the efficacy of new treatment candidates, delineating whether they should be combined with existing chemotherapeutics and to what extent they affect metastatic dissemination of the tumor cells are key preclinical indicators needed to increase the chance of success in clinical trials. Developing such data, however, requires animal models that recapitulate individual differences of different lung cancer patients, include insights into metastatic activity and allow analysis of a large amounts of treatment combinations for each patient model. As such, an in vivo screening system which has higher throughput than mouse models and at the same time allows analysis of metastatic activity would be very valuable in mimicking human disease. Here we conducted zebrafish patient tumor derived xenograft (PDX)-studies based on cisplatin sensitive and -resistant lung cancer PDX material, to test the efficacy of a a novel antibody, CAN04, under development for this indication. CAN04 targets Interleukin-1 Accessory Protein (IL1RAP) and has shown synergistic effects with cisplatin in murine models of cancer. CAN04 is currently in phase II development in combination with chemotherapy in lung cancer and pancreatic cancer. CAN04 was given either alone or with cisplatin at three different concentrations, and the effects on primary tumor growth and metastasis three days after tumor implantation was evaluated. We show that CAN04 was able to synergize with cisplatin in causing almost complete (85%-98%) tumor regression even of cisplatin-resistant tumors, compared to non-treated controls. The effects were concentration- and model-dependent. Interestingly, in the cisplatin-resistant model, the antibody and cisplatin co-treatment led to robust inhibition of metastatic dissemination, which was not seen in either group alone. This substantiates the beneficial therapeutic efficacy of combining CAN04 to cisplatin treatment in lung cancer. In conclusion, zebrafish-PDX (ZTX) models are powerful tools for evaluating individual differences in drug sensitivity on both primary tumor growth and metastasis and are suitable for screening various drug concentrations and/or combinations in multiple models with results being generated within one or a few weeks. We further conclude that CAN04 is inducing cisplatin sensitivity and synergize with cisplatin to inhibit metastasis, at least in some cisplatin resistant lung cancers. Citation Format: Zaheer Ali, Anna Nilsson, Malin Vildevall, Julia Schueler, David Liberg, Anna Fahlgren, Lasse DE Jensen. Zebrafish patient tumor-derived xenograft models used for pre-clinical evaluation of CAN04 for lung and pancreatic cancer [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6126.

Lung cancer accounts for the 2nd most common cancer among men and women, representing 24% of cancer deaths worldwide. Standard-of-care treatments vary considerably depending on the tumor type and staging. Identifying which patients will benefit from treatment with a certain drug remains one of the major challenges in the clinic. Genetic analyses are widely used but have low applicability as only ~10% of the patients have mutations coupled to available targeted therapies and relatively low sensitivity as therapeutic effects are absent in ~50% of the predicted responders. Mouse-PDX models can accurately determine drug response rates for 50-60% of the patients, but are not well suited for evaluating metastatic risk. As metastasis is a major cause of disease-associated mortality and no drugs that target metastasis exist today, there is considerable need to develop new drugs able to impair metastatic dissemination in lung (and other) cancers. To meet this need, zebrafish-PDX (ZTX) models are ideally positioned as a synergistic complement to mouse-PDX models allowing evaluation of drug responses, in a non-rodent in vivo system with the turnaround time and scalability of an in vitro platform Here we generated zebrafish- and mouse-PDX models based on 20 patient NSCLC samples and compared the efficacy of standard-of-care treatment (erlotinib and paclitaxel) on primary tumor growth/regression as well as metastatic dissemination in the zebrafish-PDX models. The ZTX models exhibited variable sensitivity to the drugs tested with 11 of 20 and 16 of 20 models being sensitive to erlotinib and paclitaxel respectively. The efficacy of erlotinib and/or paclitaxel was identical in 9 of 11 mouse- and zebrafish-PDX models where these drugs were compared head-to-head. The models metastasized within three days of tumor implantation in the zebrafish larvae, seeding an average of 2 - 8 metastatic lesions per model in the caudal hematopoietic plexus. Paclitaxel and erlotinib inhibited metastasis in 7 of 20 and 6 of 20 models respectively. The anti-metastatic activity did not correlate with the activity against the primary tumor. Investigations as to what extent this correlates with invasive phenotypes observed in histological preparations of the mouse-PDX models, and clinical data on metastasis in the patients, are currently ongoing. In conclusion we provide evidence of the accuracy of the ZTX models in predicting anti-tumor responses to commonly used drugs in NSCLC compared to mouse-PDX models and demonstrate that ZTX models provide a sensitive method for determining metastatic risk and the anti-metastatic efficacy of NSCLC relevant drugs. Citation Format: Zaheer Ali, Malin Vildevall, Anna Nilsson, Julia Schueler, Anna Fahlgren, Lasse DE Jensen. Zebrafish patient tumor-derived xenograft models synergize with mouse-PDX models for understanding variation in anti-cancer drug responses [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6125.

Precision medicine approaches based on DNA or RNA analyses, are often not identifying actionable mutations in the patients, and have not been established for predicting presonse to commonly used drugs such as platins and taxols. In epithelial ovarian cancer (EOC), where paclitaxel and carboplatin is the most commonly used first-line treatment, however, local practice and a host of clinical studies have/are evaluating combinations with other drugs, implying that first line treatment regimens may vary in different countries and regions. As EOC patients that respond to first-line treatment have much better prognosis that those who don't, there is an urgent need for methods to predict the outcome to various available first line therapies, prior to treatment onset, so that the optimal treatment can be identified and offered to every patient. Zebrafish tumor xenograft (ZTX) models have recently been used to predict treatment outcome to first line therapy in multiple myeloma, neuroendocrine cancer, colorectal cancer and gastric cancer, but these studies have been small case studies including only a few patients. Furthermore, EOC, and the main drugs used to treat them have, however, not been studied in ZTX models in the past. Here we present results from the first 81 patients in an ongoing clinical study to delineate the sensitivity and specificity of ZTX models to predict treatment outcome to paclitaxel and carboplatin. Of these 81 patients, 21 presented with a primary tumor only, whereas 60 had metastatic disease. To date, 25 patient-derived tumor xenograft models have been developed by random selection of samples from this patient cohort, of which 12 were from paired primary and metastatic lesions from 6 patients. We show that tumors generated from metastatic omental lesions exhibited similar growth and metastatic dissemination as those generated from samples taken from ovarian lesions. Among the 25 patient models generated, 15 were analyzed for efficacy of paclitaxel and carboplatin in the ZTX models. We found that 11 of 15 responded to paclitaxel, 12 of 15 to carboplatin and 10 of 15 to both drugs. In conclusion, ZTX EOC models established from surgical samples allow prediction of treatment outcomes to 1st line treatment and thereby may aid treatment planning in the future. Citation Format: Karthik Selvaraj, Malin Vildevall, Lina Wirestam, Zaheer Ali, Anna Erkstam, Annelie Abrahamsson, Åsa Rydmark Kersley, Preben Kjölhede, Stig Linder, Charlotta Dabrosin, Anna Fahlgren, Lasse Jensen. Zebrafish tumor-derived xenograft-models for improved diagnosis and treatment planning in ovarian cancer patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3000.

Precision medicine in oncology aims to identify the most effective treatment for any given patient based on individualized analyses of patient material. Currently, precision medicine relies on sequencing of DNA or RNA to identify patient tumor-specific mutational profiles that may be coupled to drug response. These techniques, however, fail to reveal actionable mutations in approximately 85% of the cancer patients, and have not been established at all for many commonly used drugs including cisplatin-based treatments used in urinary bladder cancer. While mouse-PDX models can determine drug response rates with high accuracy in most patients and for most drugs, such techniques are too slow and expensive to be relevant for first line treatment planning. Urinary bladder cancer patients are often treated with cisplatin-containing combination therapy, with the hope of down-staging tumors before surgery. 60%, however, do not respond or even progress on this treatment, and these patients would benefit from immediate surgery upon diagnosis. To help identify non-responding patients, we show here that patient-derived tumor xenograft models can be established in zebrafish larvae (ZTX models) and that the resulting tumors exhibit differential responses to the two main cisplatin-containing treatments GC and MVAC. Preliminary results from the first 19 patients are presented. Two tumor biopsies were destroyed during transport and two did not allow isolation of sufficient viable cells for implantation. From the remaining 15 samples an average of 2,6 million cells with average viability of 53% were isolated and used to implant at least 60 2-days old larvae. All 15 samples implanted in the larvae and survived and/or grew exhibiting varying degrees of metastatic dissemination (average between 2 and 13 metastasized cells per embryo and model) within only three days from implantation. Four ZTX models exhibited different responses to GC and MVAC demonstrating that these treatments are not equally effective in all patients. Non-response in ZTX models was associated with tumors having re-appeared in the bladder upon radical cystectomy in all patients undergoing surgery prior to Dec. 5th 2019 (n=3). GC inhibited metastasis in all models (average 69% inhibition), whereas MVAC inhibited metastasis in 40% of the models (average 36% inhibition). In conclusion: The ZTX urinary bladder cancer platform presented here overcome limitations associated with long assay time and high cost of other functional models within precision medicine as well as the low hit-rate of actionable mutations associated with genomic techniques. ZTX models will therefore likely become a powerful method for functional precision medicine within oncology, in the near future. Citation Format: Zaheer Ali, Anna Nilsson, Malin Vildevall, Larissa Rizzo, Ylva Huge, Amir Sherif, Anna Fahlgren, Lasse DE Jensen. Translation of zebrafish tumor-derived xenograft-models for improved diagnosis and treatment planning in urinary bladder cancer patients [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6124.