The Advent of Biomolecular Ultrasound Imaging

Ultrasound imaging is one of the most widely used modalities in clinical practice, revealing not only human prenatal development but also arterial function in the adult brain. Ultrasound waves travel deep within soft biological tissues and provide information about the motion and mechanical properties of internal organs. A drawback of ultrasound imaging is its limited ability to detect molecular targets due to a lack of cell-type specific acoustic contrast. To date, this limitation has been addressed by targeting synthetic ultrasound contrast agents to molecular targets. This ultrasound molecular imaging approach has proved to be successful but is restricted to the vascular space. Here, we introduce the nascent field of biomolecular ultrasound imaging, a molecular imaging approach that relies on genetically encoded acoustic biomolecules to interface ultrasound waves with cellular processes. We review ultrasound imaging applications bridging wave physics and chemical engineering with potential for deep brain imaging.


INTRODUCTION
Ultrasound imaging is used daily in clinical practice to assess the anatomical and physiological features of organs. Next to diagnostic applications in obstetrics and cardiology, we are witnessing a growing interest for ultrasound imaging in the field of neuroscience (Rabut et al., 2020).
Transcranial ultrasound Doppler has been used for decades to assess the cerebrovascular function of major arteries of the brain (Aaslid et al., 1982). In 2011, functional ultrasound neuroimaging (fUS) (Mace´et al., 2011;Rabut et al., 2019) has been introduced as a breakthrough modality that relies on neurovascular coupling to map neuronal activity with a higher spatiotemporal resolution and portability than fMRI (Deffieux et al., 2018). fUS has been used to track epilepsy crises in human neonates (Demene et al., 2017) or to delineate tumor-brain interfaces in neuro-oncology patients (Imbault et al., 2017;Soloukey et al., 2020). More recently, a technique inspired by optical super-resolution named 3D ultrasound localization microscopy (ULM) has generated vascular maps of the living brain with a 30 microns resolution (Heiles, 2019). Unfortunately, these ultrasound imaging methods are not inherently sensitive to cellular and molecular processes.
To observe biological processes at the cellular scale, ultrasound engineers can now rely on endogenous, synthetic or biomolecular contrast agents ( Fig. 1) as vascular or intracellular reporters (Rabut et al., 2020). Red blood cells (RBCs) are a unique cell type that can be detected with high specificity thanks to their motion. RBC motion differs significantly from collective tissue motion and induces a phase shift in ultrasound signals (Bonnefous and Pesque´, 1986) that can be captured to map blood vessels (Bercoff et al., 2011) (Fig. 1A). Lipidshelled microbubbles (MBs) are the main class of synthetic ultrasound contrast agents (Fig. 1B). MBs scatter ultrasound more efficiently than RBCs thanks to their highly compressible gas core, and are administered intravenously to image blood perfusion (Kaul, 2008). MBs have also been engineered to target specific endothelial biomarkers (Ferrara et al., 2007). The combination of ultrasound imaging with targeted MBs is the most established molecular ultrasound imaging strategy so far. In 2014, genetically encoded acoustic biomolecules called gas vesicles (GVs) have been introduced as ultrasound analogs to the green fluorescent protein (Tsien, 1998;Shapiro et al., 2014) (Fig. 1C). GVs enable acoustic labeling of cells which opens the possibility of tracking cellular processes with ultrasound (Maresca et al., 2018b(Maresca et al., , 2018c Endogenous, synthetic, and biomolecular ultrasound contrast agents enabled various molecular imaging applications such as ultrasound imaging of erythrocyte aggregation (Yu et al., 2011), ultrasound imaging of vascular endothelial growth factor (VEGF) (Wang et al., 2018), or ultrasound imaging of gene expression (Bourdeau et al., 2018;Farhadi et al., 2019). GV engineering has also led to the development of acoustic biosensors that could be used across a wide spectrum of applications (Lakshmanan et al., 2016(Lakshmanan et al., , 2020. This review article presents an overview of recent ultrasound technologies for deep imaging of the living brain. We cover recent ultrasound neuroimaging methods, ultrasound contrast agents used to interface with the brain, molecular imaging applications in the vascular space, and extravascular biomolecular imaging applications. Our focus is on methods to interface a penetrant form of energy -ultrasound waves -with cellular processes occurring in the brain. Light transmission based methods such as photoacoustics (Wang and Yao, 2016) are not discussed here.

RECENT ADVANCES IN ULTRASOUND NEUROIMAGING
Recent ultrasound neuroimaging methods rely exclusively on intravascular acoustic contrast (Fig. 2). Neurovascular signals are separated from the global brain ultrasound backscatter thanks to their motion or their specific frequency content. In the mammalian brain, blood flow velocities range from tens of centimeters per second in major arteries (Aaslid et al., 1982) to less than1 mm=s in capillaries supplying neurons with oxygen (Ivanov et al., 1981). At 15 MHz -a frequency often used for pre-clinical ultrasound imaging -the ultrasound voxel size is typically 10 6 lm 3 , which contains about 100 cortical neurons (Keller et al., 2018) or 10 5 RBCs in mice (McGarry et al., 2010). Ultrasound neuroimaging is therefore mapping the dynamics of cell populations rather than individual cells (Table 1).

Mapping cerebrovascular function with ultrafast ultrasound Doppler imaging
With the introduction of high framerate plane wave ultrasound imaging, referred to as ultrafast ultrasound , ultrasound imaging can now capture thousands of images per second. This ultrasound imaging approach was originally proposed by Bruneel et al. (1977) but only recently made possible thanks to modern multi-core computing architectures . Instead of forming ultrasound images line by line, ultrafast ultrasound imaging relies on tilted plane wave transmissions to insonify brain tissues at kilohertz framerates (Montaldo et al., 2009). Ultrafast ultrasound Doppler imaging (UDI) exploits these high framerates to achieve dense spatiotemporal sampling of RBC motion in tissue regions of interest. The UDI signal is a combination of blood scattering, tissue scattering, and noise. A critical problem in UDI is therefore to distinguish signal contributions arising from blood from signal contributions arising from other tissue types and noise. Recent approaches rely on spatiotemporal filtering of ultrafast ultrasound datasets (Demene et al., 2015).
Experimental comparisons have shown that UDI is 30-50 times more sensitive than conventional Doppler imaging (Mace´et al., 2011; Mace et al., 2013). This increase in sensitivity is key to capture blood flow variations in smaller vessels. UDI was successfully applied in a range of medical disciplines such as rheumatology (Maresca et al., 2014), cardiology (Ekroll et al., 2015;Maresca et al., 2018aMaresca et al., , 2018bMaresca et al., , 2018c, stroke imaging (Hingot et al., 2020), or intensive care . The capacity of UDI to map flow variations in small brain vessels led to the fUS imaging breakthrough. fUS relies on neurovascular coupling that serves as an indirect readout of neural activity (like in fMRI) (Mace´et al., 2011). Since 2011, fUS imaging ( Fig. 2A) was reported in many animal models (Rabut et al., 2020), using experimental paradigms to study brain-wide circuits during sleep, active behavior, motor planning or recently optogenetic stimulation (Brunner et al., 2020). In a clinical setting, fUS imaging has been used to monitor neonates (Demene´et al., 2014) and during tumor resection surgery (Imbault et al., 2017;Soloukey et al., 2020).
A critical parameter in ULM is the ability to detect individual MBs. Initially, a frame to frame subtraction or a rolling background technique was used to subtract static echoes (Siepmann et al., 2011;Christensen-Jeffries et al., 2015). This technique is hardly applicable  x-AM imaging of GV (Maresca et al., 2018c) in vivo because tissue motion can be important and outof-plane, leading to inaccurate MB localization . Spatiotemporal filters have been successfully used in this context as well to separated MB signals from tissue clutter (Desailly et al., 2017). Pixel oriented filtering methods such as the non-local means technique are also providing good results (Song et al., 2018). Current ULM methods rely on MB motion. If MBs are targeted to endothelial biomarkers or stopped by a vascular occlusion, their ULM detection is no longer possible. This problem can be solved by relying on the unique nonlinear frequency content of MB echoes rather than on MB motion. Nonlinear ultrafast ultrasound imaging of MBs has been demonstrated (Couture et al., 2012;Tremblay-Darveau et al., 2016;Muleki-Seya et al., 2020) and lends itself well to ULM processing (Zhao et al., 2020). In this context, special attention must be given to MB destruction, especially when mapping targeted MBs for molecular imaging applications.

Mapping blood perfusion with nonlinear ultrasound pulse sequences
Several ultrasound pulse sequences dedicated to MB detection have been reported over the years. In neuroscience, van Raaij et al. (2011Raaij et al. ( , 2012 showed that high-frequency nonlinear ultrasound imaging of MBs could capture rat brain activity evoked by forepaw stimulation (Fig. 2C).
A first class of pulse sequences relies on amplitude modulation (AM) of ultrasound pulses transmitted in tissues. Low amplitude ultrasound pulses elicit a linear response from MB and tissues, whereas high amplitude ultrasound pulses elicit a nonlinear response from MBs but not from tissues. The AM response of MBs is unique and differs from that of tissue. Ultrafast AM imaging has been used for characterization of kidney and tumor perfusion (Tremblay-Darveau et al., 2016).
A second class of ultrasound pulse sequences relies on the pulse inversion (PI) of ultrasound imaging pulses. In PI imaging, pairs of phase inverted pulses are transmitted into tissues and the echoes recorded are subsequently summed. This process retains nonlinear signatures that are unique to MBs circulating in tissues. Ultrafast PI imaging has been used to study blood perfusion (Leow et al., 2015).
A drawback of nonlinear ultrasound pulse sequences is their susceptibility to nonlinear propagation artifacts that misclassify tissues as contrast agents (ten Kate et al., 2012;Maresca et al., 2017). This occurs when ultrasound waves travel through large inclusions of nonlinear contrast agents. Imaging methods based on crosspropagating plane waves are currently investigated and have shown to significantly reduce nonlinear propagation artifacts (Maresca et al., 2018c).

ENDOGENOUS, SYNTHETIC AND BIOMOLECULAR ULTRASOUND CONTRAST AGENTS
Ultrasound waves are backscattered by microscale structures such as cells or ultrasound contrast agents.
Individual particles with dimensions below a tenth of the wavelength (k US =10) are referred to as Rayleigh scatterers (Helfield, 2019) and their reflective power is characterized by the scattering cross section r Rs , with f US the ultrasound wave frequency, V Rs the volume of the Rayleigh scatterer, j Rs and j 0 the compressibility of the scatterer and of the surrounding medium respectively, q Rs and q 0 the mass density of the scatterer and of the surrounding medium respectively. Eq.
(1) states that microscale particles exhibiting a density and/or compressibility contrast with surrounding tissue scatter ultrasound waves (Fig. 1A). In addition, r Rs scales with ultrasound frequency to the power 4. An example is RBC contrast which is hardly visible at 1 MHz but easily detectable at 10 MHz and above (Szabo, 2004).
Last, it is worth noting that Rayleigh scattering can also arise from structural inhomogeneities within cells such as genetically encoded GVs (Sehgal and Greenleaf, 1984;Bourdeau et al., 2018;Farhadi et al., 2019).

RBCs as indirect reporters of neuronal activity
While the acoustic contrast arising from RBCs does not inform directly on cellular or molecular events, it can measure functional hyperemia induced by neurovascular coupling (Boido et al., 2019). In a recent study, Aydin et al. (Aydin et al., 2020) reported that functional hyperemia measured with fUS imaging is a robust reporter of underlying neuronal calcic activity. The hemodynamic response function (HRF) of the brain typically lags $2 s behind the electric response (Masamoto and Kanno, 2012). In fUS imaging, the HRF has been modeled as a gamma-distribution function with no post-stimulus undershoot (Aydin et al., 2020).
It is therefore possible to exploit ultrasound backscattering induced by RBC motion to capture neural activity with ultrasound. However, that measure is indirect.

Synthetic MBs and nanobubbles
MBs are the first and most widespread ultrasound contrast agents. Early developments aimed at controlling their scattering power as well as their stability (Ophir and Parker, 1988). Modern synthetic MBs are stable enough to pass through the heart and lungs, and their in vivo lifetime allows for multiple recirculations which lengthens the diagnostic window (Frinking et al., 2020).
To stabilize MBs, a majority of them are now coated with a phospholipid layer, and their gas core is made of perfluorinated gases such as sulfur hexafluoride (Schneider et al., 1995), perfluoropropane (Unger et al., 1994), or perfluorobutane (Schneider et al., 1997(Schneider et al., , 2011. More than a dozen agents have now been commercialized with various technologies. Commercial MB diameters range from 0.5 to 10 lm (Frinking et al., 2020) which lets them circulate in most vessels after intravenous injection (Fig. 1B). The scattering power of MBs is directly linked to their size and ability to resonate at medical ultrasound frequencies (Helfield, 2019). Recently, monodisperse MB size distributions have triggered interest as these can boost MB echogenicity further (Segers et al., 2018).
Increasing the sensitivity of molecular ultrasound imaging with monodisperse MBs would be particularly valuable as just a fraction of targeted MBs end up binding to endothelial markers (Talu et al., 2007;Shih et al., 2013). Finally, Nakatsuka et al. (2012) demonstrated that it was possible to engineer MBs that become reflective only when levels of thrombin are significant but remain dormant in normal physiological conditions.

Acoustic biomolecules
The adoption of molecular ultrasound imaging based on MBs remains limited because of MB size, intravascular confinement, and short MB half-life in circulation (Kaufmann and Lindner, 2007;Wang et al., 2018).
Recently, GVs were introduced as a new class of genetically encoded ultrasound contrast agents that have the potential to become the ''GFP for ultrasound" (Rabut et al., 2020). GVs are ancient hollow protein nanostructures evolved as motility devices by aquatic microorganisms to regulate their buoyancy (Fig. 3A). In 2014, Shapiro et al. (2014) realized that the gas content of GVs would make them bright in ultrasound images. The GV nanostructure consists of two proteins. GV protein A (GvpA) forms the spindle-shaped backbone structure and GV protein C (GvpC) binds externally to GvpA and stiffens the assembled GV (Walsby, 1994).
Both linear and nonlinear ultrasound imaging of GVs was reported in literature (Shapiro et al., 2014;Cherin et al., 2017;Yang et al., 2017). Nonlinear GV scattering is thought to arise from large GV shell deformations referred to as buckling (Maresca et al., 2017). At medical ultrasound frequencies, this buckling behavior occurs at pressures above a few hundreds of kPa and makes GVs particularily suited for AM ultrasound imaging (Maresca et al., 2017).
Thanks to their biogenic origin, GVs lend themselves to molecular and genetic engineering (Fig. 3). The acoustic properties of GVs have been successfully manipulated by altering GvpC binding to GvpA. Complete GvpC removal was shown to weaken GVs and increase their nonlinear scattering (Lakshmanan et al., 2016;Maresca et al., 2017). GvpC removal is also lowering GV collapse pressue. This characteristic was leveraged to engineer acoustic reporters with distinct acoustic collapse pressures and used for multiplexed imaging (Lakshmanan et al., 2016). Specific biomarkers can also be bound to to the outer wall of GVs, following the example of targeted MBs. In summary, GvpC provides a convenient handle for GV shell functionalization.

MOLECULAR ULTRASOUND IMAGING USING VASCULAR AGENTS Targeted MBs as acoustic reporters of endothelial function
Many vascular inflammation diseases and types of cancer are characterized by differential expression of endothelial biomolecules. Targeted MBs have been successfully used to bind to multiple membrane proteins such as Pand E-selectin, VCAM-1 and ICAM-1, locally enhancing imaging contrast in diseased areas (Hamilton et al., 2004;Kaufmann and Lindner, 2007;Chadderdon et al., 2014). In practice, ultrasound imaging with targeted MBs requires a significant difference in concentration between bound and unbound MBs. A 5-10 min window after intravenous injection is enough for the reticuloendothelial system to clear non-functionalized MBs from the bloodstream. MB clearance is also in part due to acoustic collapse or gas diffusion through the MB phospholipid shell (Kosareva et al., 2020). A vast majority of methods for discerning bound and freely circulating MBs rely on destruction-replenishment strategies. In destruction-replenishment methods, MBs are first destroyed using a high-intensity ultrasound pulse. The subsequent signal is assumed to be arising from freelycirculating MBs and can be subtracted from the initial signal that accounts for both bound and unbound MBs (Abou-Elkacem et al., 2015).
Molecular ultrasound imaging using targeted MBs has been recently combined with ULM to co-localize vascular anatomy and vascular markers of angiogenesis (Zhao et al., 2020) (Fig. 4A). This promising multimodal approach could be used to study longitudinal cancer development and vascular remodeling in living organisms.

GV-enhanced hemodynamic fUS imaging
GVs were recently investigated as hemodynamic enhancers for sensitive transcranial fUS imaging in mice   (Fig. 4B). This study constitutes a first step towards biomolecular fUS imaging of neural activity. Purified GVs were used as linearly scattering nanoparticles circulating alongside RBCs, increasing the overall blood backscatter and raising the sensitivity of Doppler methods.
A first finding was that thanks to their nanoscale, inherent stability, and large numbers in circulation, GVsenhanced UDI could measure slow flow velocities more accurately than conventional UDI and MB-enhanced UDI . One possible explanation is that unlike MBs, GVs are dominated by Stokes drag over other forces such as buoyancy. The capacity of GVs to enhance Doppler signals down to 50 mm/s flow velocities is critical to map the neurovascular coupling in the smallest vascular compartments (Boido et al., 2019).
A second finding was that intravenous bolus injection of GVs did enhance ultrafast Doppler signals in the mouse brain. While the peak enhancement provided by GVs was inferior to that of MBs (+34% compared to baseline level for GVs and +149% for MBs), GVs yielded a significantly smoother Doppler signal enhancement (the mean variance of the fast time fluctuation of the GV-enhanced Doppler signal was 0.8% compared to 4.5% for the MBenhanced Doppler signal).
A third finding was that bolus injections of GVs provided a pseudo-steady transcranial enhancement of fUS signals over the course of 4 minutes in mice, whereas bolus injections of MBs deteriorated the quality of brain activity maps. In the future, a comparison of different modes of administration (bolus versus infusion) and a broader range of doses would help optimize this imaging approach.
It is worth noting that in contrast-free fUS imaging, the percent change in fUS signals at 15 MHz is lower in mice ($5%) (Brunner et al., 2020) than in rats ($15%) (Rabut et al., 2019) for a given brain stimulation protocol (e.g. whisker stimulation). This might be due to differences in RBC volume -45.5 mm 3 in laboratory mice (de Jong et al., 2001) versus 60.7 mm 3 in laboratory rats (Garcia et al., 2012). This justifies the need for methods capable  Fig. 3. GV engineering . (A) GV hollow protein nanostructure, gene cluster encoding GVs and surface arrangement of the GV shell proteins GvpA and GvpC. Adapted with permission from (Maresca et al., 2018b). (B) Heterologous expression of gas vesicles in E. coli. Adapted with permission from (Bourdeau et al., 2018). (C) GvpC bioengineering pipeline, adapted with permission from (Maresca et al., 2018b). (D) Design of an acoustic biosensor of Tobacco Etch Virus (TEV) protease activity. Adapted with permission from (Lakshmanan et al., 2020). (E) Nonlinear ultrasound imaging of acoustic biosensors of TEV activity. Scale bar = 1 mm. Adapted with permission from (Lakshmanan et al., 2020).
of enhancing fUS sensitivity in mice. The benefit of GVs at lower ultrasound frequencies and in larger brains remains to be demonstrated. In any case, efforts to engineer brighter GVs will be beneficial across medical ultrasound frequencies (Hurt et al., 2021). Together these results demonstrated that GVs could become a preferred contrast agent for fUS imaging thanks to their noise-free enhancement of fUS signals across a wide range of cerebral blood flow velocities.

Imaging with engineered RBC contrast
An interesting strategy to couple molecular events to hemodynamic fUS imaging could be to artificially trigger vasodilatory responses in the brain using vasodilating peptides (Desai et al., 2016;Ohlendorf et al., 2020). Using fMRI, Desai et al. showed that nanomolar concentrations of the cgrp peptide induced larger cerebral flood flow variations than neurovascular coupling. Since fUS imaging measures local blood flow variations in the brain, molecular ultrasound imaging combining fUS and vasoactive peptides could be envisioned.
Another interesting chemical engineering approach consists of using RBC membranes to ''hide" ultrasound contrast agents from the immune system (Dhanaliwala and Hossack, 2012) and increase the contrast agent circulation time (Hu et al., 2011). RBCs can circulate in the bloodstream for several months whereas the lifetime of MB circulation is of the order of 10 min and prevents lon-gitudinal experiments. Contrast-enhanced transcranial fUS imaging methods would greatly benefit from long circulating contrast agents (Errico et al., 2016;Maresca et al., 2020).

BIOMOLECULAR ULTRASOUND IMAGING APPLICATIONS Ultrasound imaging of gene expression
GVs are the first genetically encoded acoustic biomolecules and could become the ''GFP for Ultrasound" (Rabut et al., 2020). GVs are encoded by highly conserved genetic clusters of 8-12 genes in microorganisms such as Anabaena flos-aquae (Maresca et al., 2018b). Two of these genes encode for the structural proteins GvpA and GvpC that are the main constituents of GV nanostructures. Fully assembled GVs present a 2 nm-thick shell made of periodic repeats of the single protein GvpA which is further reinforced with the external scaffold protein GvpC. echogenic bacterial phenotypes (Bourdeau et al., 2018;Hurt et al., 2021). In 2019, Farhadi et al. (2019) reported genetic expression of GVs in mammalian cells. Their study showed that acoustic contrast arising from GVs and occupying less than 0.1% of the cytoplasm could be detected using a destructive template-based ultrasound imaging sequence named BURST. Using this approach, they reported depth-resolved ultrasound imaging of GV expression in a subcutaneous tumor model (Fig. 4C). In the future, ultrasound imaging of mammalian acoustic reporter genes (mARGs) using nondestructive methods such as cross AM (Maresca et al., 2018c) should emerge (Fig. 4D).
The development of mARGs is still in its infancy and research efforts are needed to facilitate targeted GV expression in various cell-types using viral vectors. In neuroscience, a potential application could be to drive GV-expression with neural activity-dependent promoters such as Arc, fos or E-SARE (DeNardo and Luo, 2017) to observe time integrated neural responses in the brain. A second exciting route is to engineer GVs into acoustic intracellular sensors to study dynamic cellular processes deep into brain tissues.

Acoustic biosensors for deep molecular ultrasound imaging
Following the discovery of GFP, fluorescent proteins were further engineered into genetically-encoded indicators to investigate biological processes at the cellular scale. In neuroscience, genetically encoded calcium and voltage indicators are now established tools to study cognition in animal models. Lakshmanan et al. (2016) showed that the acoustic response of GVs could be shifted from linear to nonlinear by removing GvpC (Fig. 3C), a first critical step towards the development of an acoustic biosensor. In 2020, they reported the first genetically encoded acoustic biosensor with tunable nonlinear constrast in response to protease activity (Lakshmanan et al., 2020) (Fig. 3D). To do so, they inserted a short recognition sequence for the model Tobacco Etch Virus (TEV) endopeptidase into one of the GvpC repeats. They observed that GVs covered with TEV-sensitive GvpC were weakened in the presence of TEV, which led to increased nonlinear ultrasound scattering (Fig. 3E). Similar results were obtained for two other acoustic biosensor designs, one based on the calciumactivated calpain protease, and the other on a degradation tag appended to the C-terminus of GvpC (model proteosome ClpXp). These acoustic biosensors were successfully expressed in probiotic bacteria and imaged in the mouse GI tract.
In neuroscience, proteases are involved in various aspects of neuropathophysiology. For example, intracellular proteases such as calpain and caspases get involved in neuronal dysfunction in Alzheimer's disease. A current limitation is that acoustic biosensors based on GvpC degradation are not reversible and cannot report on dynamic intracellular processes. The feasibility of fast and reversible acoustic biosensors conceptually similar to GCamP (Lin and Schnitzer, 2016) remains an open research question.

Ultrasound imaging of phagolysosomal function
Next to genetic expression of mARGs, purified GVs can be used as an injectable to report on cellular activity. In a proof-of-concept study, Ling et al. (2020) used GVs to image liver macrophage phagolysosomal function with ultrasound. After intravenous injection, purified GVs get cleared from the bloodstream by liver macrophages that recognize them as foreign particles. Next, phagocytosed GVs undergo lysosomal degradation. This process was monitored by Ling et al. using two different ultrasound imaging modes. UDI was used to track GV circulation in brain vessels over time. A sharp increase in UDI intensity was observed 100 s after injection before decreasing over time due to GV retention in the liver. The GV circulation half-life was equal to 230 s. In parallel to UDI, the accumulation of GVs in liver macrophages was monitored using AM ultrasound imaging. AM signals peaked 10 minutes after injection, before decreasing due to lysosomal degradation of GVs. Together, these results showed that ultrasound imaging of phagocytosis and lysosomal degradation rates could be used to diagnose healthly versus diseased liver states in humans.
In neuroscience, purified GVs could be used as an injectable to report on cerebrovascular afflictions such as ischemia-reperfusion injuries, atherosclerosis, or brain tumors.

OUTLOOK AND OPPORTUNITIES IN NEUROSCIENCE
The field of biomolecular ultrasound imaging was born in 2014 with the introduction of GVs as genetically encoded ultrasound contrast agents. GVs have now been used as acoustic reporters genes in bacteria and mammalian cells, engineered into acoustic biosensors of protease activity, and vascular reporters for transcranial functional ultrasound neuroimaging or imaging of macrophage phagolysosomal function.
Next to the introduction of GVs, the field of ultrasound neuroimaging has witnessed major developments. In 2011 (Mace´et al., 2011), functional ultrasound neuroimaging was established as a breakthrough modality that can map brain activity where fMRI fails: in freely moving animals, human neonates, and during intraoperative brain surgery. In 2015 (Errico et al., 2015), ULM was reported as a super-resolution method capable of mapping the whole brain-wide vasculature of living mammals with a 10 mm resolution. In 2018 (Maresca et al., 2018c), a new nonlinear ultrasound pulse sequence enabled depth-resolved, nondestructive tracking of GVs in the GI tract (Fig. 4D). The combination of ultrasound neuroimaging with endongenous, synthetic and biomolecular ultrasound contrast agents will undoubtedly create new opportunities for deep molecular imaging of the living mammalian brain.
Clinical use of biomolecular ultrasound technologies is far ahead, but two translational paths can be envisioned. The first is cell-based. Genetically modified cell agents such as CAR T cells are already sent in the human body to fight cancer (Brentjens et al., 2013). These cells could augmented with the capacity to produce their own acoustic contrast and enable ultrasound monitoring of immune cell therapy. From a regulatory point of view, potential immune responses to GVs -bacteria-derived proteinswill need to be investigated. The second path consists in using GVs as purified nanoparticles administered intravenously. A potential application could be to sense cancer microenvironments with acoustic biosensors. Here again, future studies must be conducted to investigate potential immune reactions.
Bridging the fields of chemical engineering and acoustics, biomolecular ultrasound imaging will certainly lead to a wave of discoveries in neuroscience.