A semi-mechanistic model to describe preclinical tumour viral dynamics of an oncolytic virus
Carolin Mueller (1), Andreas Ackermann (2), Ulrike Schmid (3), Shaonan Wang (4), Jan-Georg Wojtyniak (3), Eva Germovsek (3)
(1) Ulm University, Germany, (2) Cancer Immunology and Immune Modulation, Boehringer Ingelheim Pharma GmbH Co. KG., Germany, (3) Translational Medicine and Clinical Pharmacology, Boehringer Ingelheim Pharma GmbH Co. KG., Germany, (4) Translational Medicine and Clinical Pharmacology, Boehringer Ingelheim Investment Co., Ltd., China
Introduction/Objectives:
Although cancer treatment is continuously improving, cancer is still one of the leading causes of death worldwide, accounting for almost every sixth death [1,2]. Novel therapeutic options are therefore needed, and oncolytic viruses (OVs) may be a good alternative due to their selective mode of action, i.e. they only productively replicate in tumour cells, causing their lysis. One OV currently investigated is a pseudotyped variant of the Vesicular Stomatitis Virus carrying the envelope GP of the visceral non-neurotropic WE-HPI strain of the lymphocytic choriomeningitis virus (VSV-GP) [3]. To obtain more information on its dynamics in the tumour, VSV-GP was encoded with the gene for enzyme luciferase (VSV-GP-Luc) due to its ability to produce bioluminescence [4]. We aimed to describe the tumour viral dynamics of VSV-GP-Luc in mice.
Methods:
Immune-competent BALB/c mice, implanted with the CT26.CL25-IFNAR1-KO, CT26-RIG-I KO or CT26-IFNAR1-KO tumour models, were used in the preclinical experiments. The following OV dose amounts were tested: 0, 105, 106, 107, 108 and 109 median tissue culture infectious dose (TCID50), with the OV administered intravenously (IV), or a combination of IV and intratumoural (IVIT). We searched the literature for existing viral dynamics models, excluding mechanistic models that also describe e.g. immune response, due to the limited experimental data. The identified literature models were then re-fit using NONMEM 7.4.3. One tumour mm3 was assumed to represent 105 tumour cells [5], and 1 TCID50 equalled 1 infectious virion. The models were compared and evaluated using visual diagnostics together with summary statistics, such as, Akaike-Information Criterion (AIC).
Results: A total of 168 OV-treated, and 96 vehicle mice were included in the preclinical studies, out of which 10 (<4% of total) outlying individual mice were excluded. Available data included 1980 tumour volume measurements, and 1022 bioluminescence observations. Out of the five identified distinct literature models [6-11], the models from Parra-Guillen et al [10] and from Phan and Tian [11], with a combination of additive and proportional residual error for both dependent variables, described our VSV-GP-Luc data best in terms of AIC. However, the model from Phan and Tian [11] was less stable, and included one more parameter, therefore we selected the model from Parra-Guillen et al [10] as our final. Due to the data limitations not all model parameters were estimated in NONMEM, i.e. the tumour growth rate (0.13/day) was obtained by linear regression using vehicle data. Additionally, baseline tumour volume (143 mm3), background bioluminescence signal (1114 ph/s/sr), and the ‘tumour bioavailability’, i.e. the proportion of the IV dose reaching the tumour (0.0002), were also obtained in R 3.5.3. The models were fitted to each administration type separately due to problems with the estimation when all data were combined. Infection rate was estimated as mean (RSE%) 0.006 (12%) mm3/day for the IVIT and 0.004 (56%) mm3/day for the IV administration, viral clearance was 12.3 (13%) /day for the IVIT and 9 (14%) /day for the IV administration, infected cell death rate was 1.3 (1%) /day for the IVIT and 1.1 (17%) for the IV administration, and burst size was fixed to the one from the IV group of 3.0 (33%) TCID50/tumour cell. Additionally, we estimated a so-called conversion factor (7.3x104 for the IVIT and 1.1x105 for the IV) to account for different units of the bioluminescence signal and the dose.
Conclusions:
The final model described the VSV-GP-Luc preclinical tumour volume data relatively well, however, the OV effect on the tumour cells was overpredicted in the lower dose groups of the IV administration, and underpredicted for the highest dose group of the IVIT administration. The model described the general trends in the VSV-GP-Luc bioluminescence data, however, there were still misspecifications. Collecting more informative data in the future might help addressing these and making the model more useful for further oncolytic virus drug development.
References:
[1] World Health Organisation, https://www.who.int/news-room/fact-sheets/detail/cancer, last accessed 17 Feb 2023
[2] International Agency for Research on Cancer, World Health Organisation, https://gco.iarc.fr/today/home, last accessed 17 Feb 2023
[3] Porosnicu et al, Future Oncol. 2022; 18(24):2627-2638
[4] Schreiber et al, Br J Cancer. 2019; 121(8):647-658
[5] Del Monte, Cell Cycle. 2009; 8(3):505-6
[6] Marzban et al, R Soc Open Sci. 2021; 8(11):210787
[7] Baccam et al, J Virol. 2006; 80(15):7590-9
[8] Li et al, Math Biosci Eng. 2020; 17(4):2853-2861
[9] Titze et al, Eur J Pharm Sci. 2017; 97:38-46
[10] Parra-Guillen et al, Front Pharmacol. 2021; 12:705443
[11] Phan and Tian, Comput Math Methods Med. 2017; 2017:6587258
