Multiplexed Ultrasound Imaging of Gene Expression

Introduction: Ultrasound offers a unique combination of advantages for in vivo imaging of deep tissues (up to ~10 cm) of living subjects with concomitant high spatial resolution (down to ~50 µm), alongside low cost, compactness, and high temporal resolution, with acquisition rates of 1-100 ms per frame. Although ultrasound is a relatively old technology, acoustic reporter genes (ARGs) that encode air-filled, protein nanostructures called gas vesicles (GVs), were established more recently in 2018, and enabled ultrasound imaging of gene expression for the first time. [Ref1] Since this first milestone, advancements have been made to develop ARGs for imaging mammalian gene expression, [Ref2] and for ultrasound imaging with nondestructive pulse sequences thereby permitting continuous, real-time imaging through detection of so-called nonlinear signals, [Ref3]. However, ARGs have so far been limited to a “single color” readout i.e. the signal is either there or it’s not, signifying the presence or absence of only a single variable at a time. Our goal here was to address this limitation by engineering ARGs for "multi-colored" or multiplexed ultrasound imaging of gene expression.

Materials and

Methods: The GV is structurally comprised of a base shell in addition to a scaffolding protein called gvpC that contributes to its rigidity. We first deleted gvpC from the cassette, hypothesizing that this construct would generate increased nonlinear signal yield for more sensitive nondestructive imaging (Fig 1a). We leveraged the light-scattering property of assembled GVs (Fig 1b) to scan for circuit parameters that resulted in successful assembly of this engineered construct (Fig 1c). Following acoustic characterization of this unscaffolded construct (Fig 1d-f), we then embarked on a directed evolution campaign to find a second construct for multiplexing (Fig 2a). More specifically, we generated a gvpC library with error-prone PCR, cloned the library back into the GV cassette, and transformed it into expression bacteria. After plating, we arrayed individual colonies, and induced expression in 96-well format before loading samples for acoustic measurements on a custom-built motorized ultrasound system (Fig 2a). We were looking for a second structure with (1) a comparable nonlinear signal yield, (2) a distinct nonlinear signal threshold, and (3) a compatible collapse threshold. Finally, we engineered tumor-homing bacteria, S. typhimurium and E. coli Nissle 1917, [Ref4] with these next-generation ARG constructs for signal deconvolution on ultrasound images. All images were acquired on a Verasonics Vantage system using an X-wave amplitude modulation (xAM) pulse sequence for nonlinear imaging with an L22-14v transducer (f=15.6 MHz) and analyzed using custom Matlab scripts.

Results, Conclusions, and Discussions: After identifying synthetic circuit conditions for in cellulo assembly of the unscaffolded nanostructure, we called it ARG560 based on its nonlinear threshold of 560 kPa (Fig 1a-c). This construct exhibited significantly increased nonlinear yield relative to the original wild-type cassette for improved imaging sensitivity (19.5 dB gain, [cell]min = ~104 cells/ml, Fig 1d-f). From our directed evolution campaign (Figure 2a-b), we identified a second construct, which we call ARG710, with ideal properties: (1) comparable nonlinear signal yield to ARG560, (2) a distinct nonlinear signal threshold (Po = 710 kPa), and (3) a compatible collapse threshold to ARG560 (Fig 2c). For pressure-domain multiplexing, we generated an unmixing matrix (M) for signal deconvolution via linear unmixing, based on signals from pure populations (Fig 2d). Qualitatively, signal deconvolution successfully delineated signals from pure and mixed populations of ARG560- and ARG710-expressing bacteria (Fig 2d-e).

The ability to track multiple variables simultaneously within an imaging field would enable a deeper, more dimensional understanding of the biological system under study, and has signified a critical advance in other imaging modalities. [Refs5-6] ARG560 and ARG710 enable this for ultrasound imaging for the first time. Future work will immediately focus on multiplexing gene expression in vivo and translating this system for multiplexed ultrasound of mammalian gene expression.

References
1. Bourdeau et al. Nature 553, 86–90 (2018)
2. Farhadi et al. Science 365, 1469–1475 (2019)
3. Hurt, Buss, Duan et al. Nature Biotechnology 41, 919–931 (2023)
4. Gurbatri et al. Science 378, 858–864 (2022)
5. Matz et al. Nature Biotechnology 17, 969–973 (1999)
6. Allouche-Arnon et al. Nature Biotechnology 40, 1143–1149 (2022)

Acknowledgements (Optional):