In this study, the potential for therapy prediction of PDOs in pancreatic cancer research was confirmed and the results were used to determine the most ideal way of comparing in-vitro drug response to patients’ clinical response. In contrast to exposing PDO models to single drugs, a more realistic approach is thought to expose PDOs to all drugs of each therapeutic regimen at the same time, i.e. multi-drug testing. Although this takes into account synergistic effects of each drug [70–72], very few studies actually implemented multi-drug testing [73–76]. In previously published studies, results of multi-drug testing were correlated to the patient’s clinical response of one patient only [76]. In this work, a methodology for performing combinatorial drug-tests for both mFOLFIRINOX and gemcitabine/nab-paclitaxel was introduced. In order to assess the potential of multi-drug testing for in-vivo therapy prediction, two different classification methods for response were implemented (direct clustering vs. pharmacokinetic). The AUC-based direct clustering performed best, reaching an accuracy of 85% for first-line therapies, while the pharmacokinetic classification method had a prediction accuracy of 77%. The lower prediction accuracy of the pharmacokinetic classification is taking into account log[IC50] values, which have previously been reported to vary significantly, also being easily affected by curve-fitting. Therefore, the AUC as the integral area of dose response curves is thought to be a more robust value and led to higher prediction accuracies when applied on multi-drug testing [77].
In the clinic, most PDAC patients are treated with combination therapies, either mFOLFIRINOX or gemcitabine/nab-paclitaxel, whereas gemcitabine monotherapy is only recommended for patients with reduced performance status [78]. For both multi-drug regimes, mechanisms of synergy have been identified in-vitro. Inhibitors of topoisomerase-1, including irinotecan, increase the effect of DNA-damaging agents such as oxaliplatin by decelerating the repair of DNA-crosslinks [26]. Folic acid increases the intracellular amount of folates, which leads to an enhanced inhibition of thymidylate synthase by 5-FU [79]. Interestingly, A Di Paolo, P Orlandi, T Di Desidero, R Danesi and G Bocci [80] demonstrated that the enhanced anti-tumor effect of 5-FU combined with folic acid depends on the time of administration and is only present when administering both drugs simultaneously. Paclitaxel has been shown to increase the concentration of gemcitabine in the tumor tissue. This is due to decreased levels of the enzyme cytidine deaminase, which is mainly responsible for eliminating gemcitabine [27]. Beyond in-vitro studies, the synergistic effects of 5-FU and oxaliplatin as well as 5-FU and folic acid were validated in clinical trials [81, 82]. The combination of 5-FU and oxaliplatin showed a significant benefit in patients with advanced PDAC [82]. In our PDO cohort, mFOLFIRINOX was consistently potent across all models, whereas the effectiveness of gemcitabine/paclitaxel was found to be more variable. Considering the evidence of synergistic drug effects both in-vivo and in-vitro as well as the results of this work, multi-drug tests are thought to be critical in future PDO-based translational trials. This is underlined by the high prediction accuracy of multi-drug testing in the study presented, emphasizing synergic effects in organoid-based drug testing. For multi-drug testing, however, the ratio of the drugs needs to be defined. Most researchers previously used IC25-, IC50- or cmax/plasma to determine drug ratios [73, 74, 76], whereas E Hadj Bachir, C Poiraud, S Paget, N Stoup, S El Moghrabi, B Duchene, N Jouy, A Bongiovanni, M Tardivel, LB Weiswald, et al. [75] estimated tumor tissue concentrations of drugs (see Supplementary Table 3). To the best of the authors knowledge, this study is the first, employing pharmacokinetic modeling to deduce optimal concentrations for translational pharmacotyping.
A major limitation of this study remains the small patient cohort. It should be noted, however, that this patient cohort is in the range to previously published PDO cohorts [20, 22, 24, 66, 83, 84]. The majority of previously reported PDO-studies included ten to fifteen patients. AK Beutel, L Schutte, J Scheible, E Roger, M Muller, L Perkhofer, A Kestler, JM Kraus, HA Kestler, TFE Barth, et al. [20] and H Tiriac, P Belleau, DD Engle, D Plenker, A Deschenes, TDD Somerville, FEM Froeling, RA Burkhart, RE Denroche, GH Jang, et al. [66] used a larger cohort of PDOs to develop the in-vitro based statistical classification methods, but were only able to clinically correlate a small fraction of patients. The clinical outcome was known for 16 of 28 patients and 9 of 66 patients respectively. Furthermore, statistical approaches used in previous studies as well as definition of clinical response were highly heterogeneous. It is also important to note that we included patients with both metastatic and primary resectable disease, which represents a heterogeneous patient cohort.
In conclusion, the high accuracy for matching PDO-drug response to the in-vivo response of patients is in accordance to previously reported translational studies [20, 22, 24, 66, 83, 84]. Moreover, this study emphasized the need to uniformly define the way PDAC PDOs are pharmacotyped (single drug testing vs. multi-drug testing) and are further classified into sensitive and resistant models. PDOs have previously been investigated in several studies and have shown promising results for predicting patients’ treatment response. Up until today, however, only single agent testing was used in PDOs to correlate the in-vivo multi-drug regimens, not taking into account essential synergistic effects of drugs. For the first time, multi-drug testing based on pharmacokinetically calculated drug concentrations was employed and results were compared to the clinical response of patients. Based on the findings of this study, multi-drug testing should be conducted in the future.