Run for your life: an integrated virtual tissue platform for incorporating exercise oncology into immunotherapy

Abstract

The purpose of this paper is to introduce a novel in silico platform for simulating early-stage solid tumor growth and anti-tumor immune response. We present the model, test the sensitivity and robustness of its parameters, and calibrate it with clinical data from exercise oncology experiments which offer a natural biological backdrop for modulation of anti-tumor immune response. We then perform two virtual experiments with the model that demonstrate its usefulness in guiding pre-clinical and clinical studies of immunotherapy. The first virtual experiment describes the intricate dynamics in the tumor microenvironment between the tumor and the infiltrating immune cells. Such dynamics is difficult to probe during a pre-clinical study as it requires significant redundancy in lab animals and is prohibitively time-consuming and labor-intensive. The result is a series of spatiotemporal snapshots of the tumor and its microenvironment that can serve as a platform to test mechanistic hypotheses on the role and dynamics of different immune cells in anti-tumor immune response. The second virtual experiment shows how dosage and/or frequency of immunotherapy drugs can be optimized based on the aerobic fitness of the patient, so that possible adverse side effects of the treatment can be minimized.

Keywords: Exercise oncology; Glycolysis; Hypoxia; In silico modeling; Personalized medicine; Treg recruitment.

Fig. 1

Model conceptualization. The model simulates the early stage of 2D solid tumor progression from which a growth rate (in terms of tumor area) can be calculated. Once initialized, tumor cells grow in the TME, and with time become more glycolytic, in a rate that depends on the host’s aerobic fitness and tolerance to hypoxia. Tumor cells die through necrosis or apoptosis (lack of oxygen or death by immune response, respectively). Tumor suppressors (“CTLs”) and tumor promoters (“Tregs”) react to cytokine and chemoattractant fields secreted by tumor cells. Tumor cells grow until they saturate the grid

Fig. 2

Aerobic fitness modulates hypoxia-tolerance in the TME. While oxygen levels are identical in the two examples above, the two representative TME react to them differently: the more aerobically fit is the host (FIT), the more tolerant to hypoxia its TME is, and as a result, tumor cells are less glycolytic relative to sedentary hosts (SED)

Fig. 3

Aerobic fitness modulates anti-tumor immune response. The more aerobically fit is the host, the less glycolytic its tumor cells are relative to a sedentary host. Consequently, recruitment of tumor promoters that can block tumor suppressors is down regulated relative to a sedentary host, and tumor growth will be relatively suppressed. Tumor promoters move towards the tumor along the chemo-attractant gradient that glycolytic tumor cells secrete. Tumor suppressors move towards the tumor along a cytokine gradient (“IFNγ”) that necrotic tumor cells secrete. Once infiltrated into the TME, tumor promoters can inhibit the ability of nearby tumor promoters to kill tumor cells

Fig. 4

Effect of aerobic fitness on tumor progression rate. The model was run on 200 virtual subjects, divided into 10 distinct aerobic fitness levels, each with 20 subjects. Each fitness level generated an average growth rate (a). These average growth rates were plotted against the fitness levels on a logarithmic scale (b). The model behaves qualitatively in accordance with a similar plot of tumor doubling times vs. fitness levels from a pilot study in 14 recently diagnosed T1 invasive ductal carcinoma post-menopausal patients (c, “Data”) [13]. The comparison between the two correlations (the observed and the mechanistically generated by the model) can be used to further constrain further calibrations of model parameters

Fig. 5

Distribution of tumor doubling time in the population. We ran the model with 200 subjects with random fitness levels. The statistical distribution of growth rates (a, b) was statistically indistinguishable in a KS test (p > 0.29) from a log normal distribution, such as the one observed of growth rates of invasive ductal carcinoma (c, [26])

Fig. 6

Prevalence of clinical tumors in athletes vs. non-athletes. Epidemiological data show prevalence of solid tumors in non-athletes to be around twice the prevalence in athletes (a) [3, 4]. We used this data point to extract a spatiotemporal calibration of the model by running 40 subjects, aerobically fit and sedentary, and identifying the tumor size (in terms of tumor area) and the time after initiation of 200 cells (in model time steps MCS) in which such a prevalence ratio is achieved (b). The prevalence ratio allows us to impose a spatiotemporal scale on our model (in this case, a scale of 3:2 between model to reality)

Fig. 7

Time series of TME sections in early stage progression of a solid tumor. To probe the intricate dynamics of anti-tumor immune response in the early stages of a solid tumor progression, the model can yield an observation window into the TME in different stages of growth (a–d), and can be used to test competing hypotheses on tumor immune cells population dynamics by comparing these snapshots to real life immunohistochemistry end points (e, [24]), where cross sections from exercised (“FIT”) and sedentary (“SED”) mice show different intratumoral CD8⁺/CD4⁺FOXP3⁺ ratios. Additional plots in Supplemental Material (Fig. S3) show the potential of the model to generate quantitative analysis for TME markers which can be compared to desired pre-clinical end points

Fig. 8

Precision immunotherapy. Aerobically fit patients may require smaller dosage of ICI than sedentary patients, which may lead to personalization of treatment and reduction of adverse effects. Without a mitigated dosage, aerobically fit subjects are more prone to ICI adverse effects than their sedentary counterparts (a, b). Lowering the dosage of ICI for aerobically fit patients relative to their sedentary counterparts can achieve the same reduction in tumor growth (c) but with a lower added toxicity hence lower probability for adverse effects (d). As a result of the ICI, the two tumors in e (sedentary and fit hosts), treated with high and low dosage, respectively, are of the same size, regardless of their initial immunogenicity. IHC of fast and slow growing invasive ductal carcinomas in human females from the study reported in [13] show, respectively, lower and higher ratios of CD4⁺FOXP3⁺ to CD8⁺ T cells (8F)