Fehr, D. | bioRxiv Science

bioRxiv2026

Simulation-guided non-thermal low-intensity ultrasound reprograms the tumor immune microenvironment and engages systemic antitumor immunity in a syngeneic orthotopic mouse model of breast cancer

IntroductionTherapeutic ultrasound has been extensively studied in ablative and sonodynamic contexts, leaving the intrinsic bioactivity of continuous non-thermal low-intensity ultrasound (LIU) largely uncharacterized. ObjectivesTo characterize the tumor biological and immunomodulatory effects of non-thermal continuous LIU in complementary in vitro and in vivo breast cancer models, underpinned by a standardized exposure platform characterized through finite element simulations and experimental validation. MethodsAcoustic and thermal fields were characterized and optimized using in silico simulations and validated against hydrophone and temperature measurements to ensure homogeneous, non-thermal exposure (1MHz, 1W/cm2, 100% duty cycle). 4T07 murine mammary carcinoma spheroids received 20min LIU treatment, and metabolic activity, apoptosis, and intracellular stress-associated markers were assessed. In a syngeneic orthotopic 4T07 mammary carcinoma model in BALB/c mice, up to six LIU treatment cycles were administered; tumor growth, survival, histopathology, immunohistochemistry, bulk tumor RNA sequencing, spleen volume and plasma cytokine profiles were assessed. ResultsIn vitro and intratumoral temperatures remained within the physiological range ([≤]39{degrees}C) throughout exposure. In spheroids, LIU reduced ATP content by more than 40% and significantly increased apoptotic, Hsp70 and Hsp90 cell fractions. In vivo, cyclic LIU slowed tumor growth, increased intratumoral necrosis, and significantly prolonged time to humane endpoint compared to untreated controls. LIU promoted early intratumoral myeloid cell infiltration and shifted the tumor transcriptome (2,573 differentially expressed genes), with enrichment in gene sets associated with immunogenic cell death, pattern-recognition, inflammatory, and innate and adaptive immune programs and downregulation of pro-tumorigenic pathways. LIU enriched the transcriptional signatures of M1 macrophage polarization and, notably, B-cell compartment engagement, which has not previously been reported for standalone continuous mechanical ultrasound. LIU significantly attenuated tumor-associated splenomegaly and elevated plasma IL-1, TNF-, and IL-10. ConclusionThese results establish a reproducible preclinical platform and provide a hypothesis-generating mechanistic basis for evaluating LIU as an adjunct to immune checkpoint blockade. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=124 SRC="FIGDIR/small/743931v1_ufig1.gif" ALT="Figure 1"> View larger version (49K): org.highwire.dtl.DTLVardef@31d366org.highwire.dtl.DTLVardef@12df6aborg.highwire.dtl.DTLVardef@9d91adorg.highwire.dtl.DTLVardef@c72b8a_HPS_FORMAT_FIGEXP M_FIG C_FIG

cancer biology↗

bioRxiv2019

Classification-based Inference of Dynamical Models of Gene Regulatory Networks

Cell-fate decisions during development are controlled by densely interconnected gene regulatory networks (GRNs) consisting of many genes. Inferring and predictively modeling these GRNs is crucial for understanding development and other physiological processes. Gene circuits, coupled differential equations that represent gene product synthesis with a switch-like function, provide a biologically realistic framework for modeling the time evolution of gene expression. However, their use has been limited to smaller networks due to the computational expense of inferring model parameters from gene expression data using global non-linear optimization. Here we show that the switch-like nature of gene regulation can be exploited to break the gene circuit inference problem into two simpler optimization problems that are amenable to computationally efficient supervised learning techniques. We present FIGR (Fast Inference of Gene Regulation), a novel classification-based inference approach to determining gene circuit parameters. We demonstrate FIGRs effectiveness on synthetic data as well as experimental data from the gap gene system of Drosophila. FIGR is faster than global non-linear optimization by nearly three orders of magnitude and its computational complexity scales much better with GRN size. On a practical level, FIGR can accurately infer the biologically realistic gap gene network in under a minute on desktop-class hardware instead of requiring hours of parallel computing. We anticipate that FIGR would enable the inference of much larger biologically realistic GRNs than was possible before. FIGR Source code is freely available at http://github.com/mlekkha/FIGR.

systems biology↗