While MR diffusion imaging can deliver useful information on the main fiber tracts in living brains (Jbabdi et al., 2015), ex vivo techniques are currently necessary to determine neuronal connectivity near or at synaptic resolution.

Light microscopy is limited in terms of imaging depth in scattering biological media. For this reason, serial mechanical sectioning followed by multiphoton imaging has been used in neuronal tracing (Ragan et al., 2012). Ex vivo tissue can also be made optically clear (Chung and Deisseroth, 2013; Renier et al., 2014) and imaged with light-sheet microscopy for non-destructive optical sectioning (Voie and Spelman, 1995; Power and Huisken, 2017; Mano et al., 2018). The resolution in light microscopy can be further improved by tissue expansion (Chen et al., 2015) and stochastic labeling of neurons (Livet et al., 2007).

Neuronal processes can have submicrometer diameters, with dendritic spine necks as small as 0.09 μm in diameter (Bartlett and Banker, 1984; Arellano et al., 2007; Gipson and Olive, 2017) and follow complicated trajectories around each other and relative to the supporting astrocytes, vasculature, and microglia. Therefore, it is challenging to disentangle neuronal processes in densely labeled tissue using, e.g., the brainbow technique (Livet et al., 2007), even at the best-possible optical resolution. Thus, a complete characterization of all synaptic connections will ultimately require imaging the entire brain network via EM. Advances in volumetric EM techniques (Eberle and Zeidler, 2018; Yin et al., 2020) already permitted complete EM brain volumes from Drosophila (Zheng et al., 2018) and partial connectomes of larval zebrafish (Hildebrand et al., 2017) to come online, with substantial advances also made for the mammalian cortex (Motta et al., 2019).

PA imaging enables the structural interrogation of the brain at scales ranging from micro- to macroscopic. Either intrinsic or extrinsic photoabsorbers can generate image contrast. The main intrinsic absorbers in brain tissue are DNA, RNA, cytochrome, lipids, and water. Lipid concentrations vary in white matter vs. gray matter and between brain regions (Chavko et al., 1993). Since their photoacoustic spectra are masked in vivo by hemoglobin, lipid imaging is conducted ex vivo after brain perfusion (Li L. et al., 2016). Mesoscale PAT imaging of entire mouse brains ex vivo has revealed structures such as the neocortex, cerebellum, cerebral aqueduct, corpus callosum, hypothalamus, hippocampus, inferior colliculus, olfactory bulb, and ventricles (Li L. et al., 2016; Olefir et al., 2019). Furthermore, dopaminergic neurons contain neuromelanin, an endogenous pigment that can be visualized ex vivo by photoacoustic tomography (Olefir et al., 2019).

PA microscopy, on the other hand, can resolve DNA/RNA (Yao et al., 2010), lipids (Wang et al., 2011), cytochromes (Zhang et al., 2013), hemoglobin (Yao et al., 2015), and melanin (Zhang et al., 2006) by selecting the appropriate wavelengths of laser illumination. Non-labeled brain tissue blocks measuring a couple of cubic millimeters were serially sectioned, scanned, and volumetrically reconstructed, allowing the mapping of nuclei positions within the blocks (Wong et al., 2017). PA prescans of unstained samples leveraging the intrinsic contrast of the previously mentioned photoabsorbers might provide a useful reference before executing complex fixation or staining protocols, leading to geometric tissue distortion.

Besides imaging the central nervous system, the spinal cord and peripheral nerves can also be accessed by PAI. Ex vivo PAT at an excitation wavelength of 1730 nm was employed to assess white-matter loss after spinal cord injury in adult rats (Wu et al., 2014). Myelin is the principal lipid constituent in nerves and was shown to produce distinctive PA microscopic signatures (Matthews et al., 2014). Nerves could also be visualized with hybrid PA/US systems in ex vivo preparations (Mari et al., 2015) and PAT in in vivo mouse models (Li R. et al., 2016). Imaging myelin in peripheral nerves might improve diagnostic accuracy if implemented in future clinical practice.

Microscopic brain imaging in vivo is dominated by multiphoton detection of fluorescent sensors. Besides imaging neurons, multiphoton imaging has also been used for the quantification of vascular architecture. In particular, measuring the microvasculature diameter during sensory stimulation can estimate the cerebral blood volume (CBV) and clarify its relationship to neuronal activation (Kong et al., 2015; Kim et al., 2019). Label-free PA microscopes can image brain microvasculature by detecting absorption signals from individual red blood cells (Kong et al., 2015; Kim et al., 2019). An integrated microscope combining reflection-mode photoacoustic, multiphoton, and second harmonic detection was used to image the complex 3D neurovascular system of mice in vivo (Song et al., 2016). Although oxygen metabolism was not quantified in this study, the ability to image neuronal activity at a subcellular level and simultaneously record hemodynamic parameters would prove valuable in elucidating the underlying mechanisms of neurovascular coupling (Song et al., 2016).

Photoacoustic tomography in vivo is dominated by vascular contrast due to high hemoglobin concentrations that mask other photoabsorbers. Brain structures such as the third ventricle, superior sagittal sinus, and cerebral bridging veins were visualized in vivo (Razansky et al., 2011; Burton et al., 2013; Olefir et al., 2016; Chen et al., 2017). The carotid arteries and Willis Circle could be imaged with PA by illuminating through the oral cavity (Lin et al., 2015), delivering data similar to what, e.g., time-of-flight MRI techniques provide.

Anatomical imaging with a hybrid PA/US system enabled the correction of skull-induced artifacts in mice, facilitating microvasculature visualization with a field-of-view covering the entire cortex (Estrada et al., 2018). Imaging brain vasculature in vivo can also be achieved optically with short-wave infrared imaging (Bruns et al., 2017; Carr et al., 2018) and US (Errico et al., 2015). Both of these techniques feature several centimeters in penetration depth (Macé et al., 2011; Bruns et al., 2017) and excellent performance in capturing flow dynamics. In this configuration, photoacoustics could contribute valuable data on oxygen metabolism.

Concerning pathological brain conditions, PA is naturally suited for studying brain ischemia due to its sensitivity to blood oxygenation. PAT systems lend themselves to label-free detection of stroke by recording an asymmetry in blood oxygenation. Furthermore, areas of compromised perfusion in which brain tissue may still be viable, the so-called penumbra, can be identified (Kneipp et al., 2014). Both OR-PAM and PAT were able to identify ischemia 3-5 min after onset. Hemorrhagic transformation could also be detected (Lv et al., 2020).

OR-PAM systems can be easily modified to enable localized photocoagulation by increasing the laser beams’ energy, such that small vessels can be occluded with high precision, leading to local ischemia (Hu et al., 2011a; Liao et al., 2013; Lv et al., 2020). After photocoagulation, ischemia caused a breakdown of the BBB, leading to extravasation of Evans Blue bound to albumin to visualize the extent of ischemia (Lv et al., 2020).

The ability to visualize anatomical structures based on contrast derived from exogenous and endogenous photoabsorbers makes PA imaging a valuable tool for non-invasive neoplastic disease monitoring, including brain tumors. Volumetric imaging of glioblastoma in nude mice was demonstrated without contrast agents by exploiting differential contrast under illumination with 800 and 850 nm. The authors also quantified tumor O₂ saturation, which increased 15 days after inoculation, whereas the tissue became hypoxic 42 days after inoculation (Balasundaram et al., 2018). In another study, integrin expressing tumor cells were tagged with an IRDye800-based contrast agent while also assessing tumor SO₂, which was 13% lower than in healthy brain tissue (Li M. et al., 2008).

Melanoma cells offer strong intrinsic contrast for PAI. Secondary melanoma brain tumors were monitored non-invasively with a reflection-mode PAM in small rodents (Staley et al., 2010). Circulating melanoma cells could be imaged in the brain of mice perfused with artificial cerebrospinal fluid, which was possible because melanoma cells naturally contain the highly absorbing pigment melanin. Still, high hemoglobin absorption prevented a transfer of the technique to in vivo detection (Deán-Ben et al., 2020). In addition to cell visualization, it is possible to successfully manipulate circulating B16F10 mouse melanoma cells with acoustic and photoacoustic pressure waves. Circulating cells in lymphatic vessels were trapped, focused, and redirected, showcasing how this method might be used for the non-invasive diagnosis of neoplastic disease (Galanzha et al., 2016). Identification of circulating melanoma cells in the superficial venous system of human subjects was also achieved with a PAT system. However, more extensive studies are needed to determine the value of including this imaging method in the staging of this neoplastic disease (Hai et al., 2020).

Amyloid plaque deposition is a hallmark of Alzheimer’s disease. Inflammatory processes and abnormal oxygen metabolism also play an important role in neurodegenerative processes and may be accessed by molecular imaging.

Positron emission tomography is considered the gold standard in Alzheimer’s plaque and inflammation imaging. Despite recent developments in simultaneous PET/MR acquisition, molecular information is limited to low spatial resolution in the 0.5–1 millimeter range (Zhang X. Y. et al., 2017). Optical imaging in preclinical settings has been facilitated by the development of NIR contrast agents targeting amyloid plaques (Tong et al., 2015). The plaques were visualized with an OR-PAM system after i.v. congo red injection, which is a classical photoabsorbing stain for amyloid. Plaque position, morphology, and relation to vasculature were imaged both in vitro and in vivo through a glass window. Again for this application, multispectral acquisition allowed blood vessels to be simultaneously imaged without contrast agent injection (Hu et al., 2009).

Inflammatory processes can be imaged with PA by using macrophage targeting probes such as CDnir7 (Kang et al., 2014). Alzheimer’s disease’s inflammatory component was quantified non-invasively in vivo using PAT detection after CDnir7 administration (Park et al., 2019).

Cerebral metabolic dysfunction has been shown to accompany plaque deposition in AD (de la Monte and Tong, 2014). PAT investigation of arcAβ mice showed a decrease in the cerebral metabolic rate of oxygen (CMRO₂), while the oxygen extraction fraction (OEF) remained unchanged in control animals (Ni et al., 2018).

Hemodynamic imaging also lends itself to the study of pathologies that induce massive neuronal activation followed by a hemodynamic response such as epileptic seizures. Proof of principle work in this pathology has been performed with spherical arrays in rats (Wang et al., 2012) while seizures in awake animals were imaged using portable sensor arrays (Tang et al., 2015). PAT helped to characterize the spatial spread of pharmacologically induced seizure activity in the mouse (Zhang et al., 2018) and could also discern patterns related to seizure activity in thalamic regions beyond the reach of optical imaging (Gottschalk et al., 2017). Tissue perfusion changes showing increased blood flow in seizure foci measured with PA are also in agreement with optical imaging spectroscopy studies (Harris et al., 2014).

As accessible as the neurovascular coupling is for many imaging methods, it only provides a rough estimation of the metabolic activity in a given brain region. Thus, it is probably biased toward energy-intense presynaptic neurotransmitter recycling (Logothetis, 2008; Gauthier and Fan, 2019), while it can not differentiate between excitatory or inhibitory neurotransmission. This contrast mechanism can also not resolve individual cells or the timing of action potentials, which can be accessible with molecular contrast agents.

For PA imaging, the building blocks for contrast agents ideally fulfill the following set of photophysical properties: a large molar extinction coefficient at wavelengths at which tissue absorbance is relatively low such as in the near-infrared window, a distinct absorption peak to facilitate spectral unmixing, a low quantum yield for maximum light conversion to thermal energy via non-radiative decay, and a high photostability (Weber et al., 2016).

Concerning the interaction with the biological tissues of interest, a molecular contrast agent should also possess high stability in biological media such as blood and cerebrospinal fluid, low toxicity, and immunogenicity. For neuroimaging, the contrast agent’s size and charge should also enable permeability through the blood-brain-barrier (BBB) (Milej et al., 2017), such that techniques to open the BBB do not have to be employed.

There is a wide variety of contrast mechanisms applicable in preclinical neuroimaging. It is thus useful to differentiate between the following molecular contrast agent classes with respect to which molecular state or process they indicate: (i) targetable labels that can map the biodistribution of a target structure, (ii) turn-on probes, whose signal can be irreversibly activated by a molecular interaction, and (iii) reversible sensors, which can dynamically adopt different signaling states in response to an analyte of interest (Figure 2, columns).

Based on their composition, we can also differentiate synthetic from semi-genetic or fully genetic contrast agents (Figure 2, rows).

Synthetic agents are produced through chemical and nanotechnological methods. Semi-genetic contrast agents have a genetically encoded component that interacts with at least one synthetic component. In distinction, genetic contrast agents are generated via genetic programming of a biosynthetic pathway within a cellular (or cell-free) biomolecular machinery, which may require endogenous co-factors.

The molecular contrast agent type has, for instance, implications for the delivery method that can be used. Options for delivering contrast agents to the brain include intravenous, intrathecal, intracerebroventricular, or intracerebral injections.

For intravenous delivery, a prolonged circulation time may be desired for imaging blood flow and contrast agent accumulation in the target tissue.

To overcome the intact BBB, high lipid solubility and a particle size under 1 nm are advantageous (Abbott and Romero, 1996; Pardridge, 2005). In neoplastic disease, BBB disruption allows the passage of larger particles up to 100 nm (Houston et al., 2020). Transport processes across the BBB can be exploited by incorporating suitable modifications that enable receptor-mediated transcytosis (Pulgar, 2018).

The BBB can also be temporarily opened using focused ultrasound, thus increasing the spectrum of compounds that can be efficiently delivered to the brain (Chen et al., 2019).

An intracranial injection is an alternative option for the direct delivery of dyes, therapeutics, or viruses to the brain. This method’s limitations are that only a small volume can be administered (usually around 1–2 μL) at low injection rates, usually not exceeding 0.1 μL/min to avoid damaging the brain (Cetin et al., 2006). Alternatively, surgically implanted osmotic pumps can dispense defined volumes over longer periods (Sepúlveda et al., 2012).

Whereas synthetic agents have to be delivered to the tissue of interest and then targeted to extra- or intracellular structures, genetically encoded contrast agents have the advantage that they can be produced directly in genetically defined cells conditioned on specific cellular states. Transferring the necessary genetic information can be achieved by germline transgenesis or via the administration of viral (Sung and Kim, 2019) or non-viral (O’Brien and Lummis, 2006) DNA or RNA delivery vehicles.

Gene expression via viral transduction is preferred in vivo and can be achieved through interstitial injection of neurotropic viruses or systemic vector administration (Foust et al., 2009; Manfredsson et al., 2009; Challis et al., 2019). Popular vectors include Lentiviruses, Herpes Simplex, or Adeno-associated viruses (Cetin et al., 2006; Challis et al., 2019; Ingusci et al., 2019). The selective use of gene promoters allows for an increasing number of neuronal or glial cell types to be specifically targeted. After infection, the contrast agent DNA sequence can be integrated into the host genome (Vargas et al., 2016), or it persists as an episome in the nucleus (Haggerty et al., 2020). Regardless of DNA integration, because neurons do not divide, long-term contrast agent expression can be achieved.

Besides, the knowledge that different neurotropic viruses have a preferred direction of transport within infected neurons can be exploited to identify or target the pre- or post-synaptic connections of a neuronal population of interest (Saleeba et al., 2019).

The specific biodistribution of contrast agents in predefined compartments of the body can reveal local anatomical and functional aspects. Examples include “blood pool agents” that are constrained to the vascular system, nanomaterials that accumulate in tumors or post-stroke tissue due to enhanced permeability and retention effects (PRE) (Nel et al., 2017), or leakiness of the BBB.

At a finer scale, information can be gained from targetable labels that can be directed to specific cells via surface affinities or selective cellular uptake (see next section).

Many synthetic photoabsorbers can serve as blood-pool agents and as the building blocks for targetable agents in PAI: organic contrast agents (e.g., organic dyes), conjugated polymer nanoparticles, and other inorganic nanostructures (e.g., carbon nanoshells).

Small molecular dyes can visualize blood vessels and enable perfusion measurements in healthy brains and brain tumors (Burton et al., 2013). Strong photoabsorbers such as black paramagnetic polyethylene microspheres (Luís Dean-Ben and Razansky, 2018) or oil-dissolved IR-780 dye droplets (Zhang P. et al., 2019) were used for localization microscopy with photoacoustics to go beyond the diffraction limit. Applying these techniques to PA angiography enables high-resolution maps of brain vasculature to be acquired in vivo (Zhang P. et al., 2019).

Advances in material science have led to the creation of functionalized synthetic nanoprobes, which can be tuned for high absorption in a defined spectral range, exhibit resistance to photobleaching, and be functionalized with the addition of dyes or drugs. Most synthetic nanoparticles developed for photoacoustic applications have been employed for studying brain tumors: noble metal nanoparticles (Shang et al., 2017), copper nanoparticles (Wu et al., 2018; Zhang H. et al., 2019), carbon nanorods (Qian et al., 2018), iron-based nanoparticles (Zhou et al., 2018), MoS2 nanosheets (Chen et al., 2016). Nanoparticles that enable multimodal imaging have been a focus of recent research: PA/MRI/Raman (Kircher et al., 2012), PA/MRI/PET (Fan et al., 2014), PA/MRI (Zhu et al., 2015; Qiao et al., 2018). Circulating dyes and nanoparticles have been shown to accumulate in brain tumors due to a disruption of the BBB and PRE (Kircher et al., 2012; Burton et al., 2013; Fan et al., 2015; Neuschmelting et al., 2018). However, these contrast agents are rapidly cleared from the bloodstream by the reticuloendothelial system and the liver.

An example of fully synthetic turn-on PA contrast agents are BBB-permeable aniline compounds that strongly increase absorbance and thus PA signal around 750 nm upon chelation with Cu²⁺ and may thus be useful for imaging copper accumulation in brains that may be correlated with amyloid plaque formation (Wang et al., 2019). Furthermore, gold nanoparticles conjugated to antibodies have been shown to change their PA spectra upon target binding due to plasmon resonance coupling (Luke et al., 2014).

One of the first semi-genetic turn-on mechanisms adapted for PA imaging is the well-known reporter enzyme β-galactosidase, which converts the soluble synthetic substrate X-gal into a blue precipitate. Delivery of the X-gal substrate to β-galactosidase-expressing cells in culture was shown to generate PA contrast from the resulting precipitate upon 650 nm illumination (Li et al., 2007). Simultaneous detection of the microvasculature and lacZ labeled tumor cells via AR-PAM imaging was useful in identifying tumor feeding vessels (Li L. et al., 2008). Individual LacZ labeled cells could be imaged with an OR-PAM setup showing intracellular contrast agent distribution, and PAT methods were able to detect LacZ labeled cells up to a depth of 5 cm in biological tissue (Cai et al., 2012).

Förster resonance energy transfer (FRET) entails energy transfer from a donor fluorophore to an acceptor chromophore, which can also alter the corresponding PA signals (Wang and Wang, 2012; Wang et al., 2013) (Figure 2, bottom of middle column). An activatable PA reporter for protease activity was explored in vitro using genetically expressed chromoproteins with low QY as a dark FRET acceptor. The authors expressed ultramarine variants linked to EGFP as a FRET donor via a caspase-3 cleavage site (Li Y. et al., 2016). When caspase-3 activity was induced in mammalian cell culture, an increase in EGFP fluorescence was observed due to the loss of FRET upon proteolytic separation of the FRET pair. When FRET-pairs were purified and measured by PA without and with proteolytic cleavage of the linker, a decrease in the PA signal was observed consistent with an increase of donor fluorescence. Modulation of FRET from a dark acceptor could be an interesting PA reporter mechanism, especially if genetically controlled FRET donors with high QY in the near-infrared could be developed.

As discussed above, volumetric PA imaging is a scanless technique that can pick up signals with higher temporal resolution and from deeper tissue layers than multiphoton techniques. Thus, it is an attractive method for mapping dynamic and distributed brain processes. As laid out above, PAI can map neurovascular coupling dynamics at adequate temporal sampling rates and over large portions of the brain.

However, it is desirable to obtain more specific information about the cellular and molecular identity and temporal profile of neuronal activity.

Patch-clamp measurements of neurons naturally provide the best spatial and temporal resolution for characterizing neuronal electrical activity (Petersen, 2017), but it is often valuable to capture the propagation of charge fluxes over many cells of specific identities.

Imaging electrical events in neurons was first achieved with electrochromic sensors, in which membrane potentials cause changes in absorption (Tasaki et al., 1968; Ross et al., 1977; Grinvald et al., 1981; Grinvald and Hildesheim, 2004). Despite the low signal changes of usually less than 1%, impressive imaging results allowed millisecond activation maps resolving individual columns in the visual cortex, a highly influential finding for the field of visual perception (Sharon and Grinvald, 2002).

While synthetic voltage dyes also tend to stain glia, genetically encoded voltage-sensitive indicators (GEVIs) can be targeted to selective neuronal types. Several types of GEVIs have been realized based on fusions to voltage-sensing domains or rhodopsins, which can be read out by FRET or QY changes. They have become brighter and increasingly red-shifted, enabling the study of subcellular features of the action potential waveform (Xu et al., 2017; Bando et al., 2019; Panzera and Hoppa, 2019). Challenges persist in applying this technology in vivo as heavily convoluted neuronal cells make it difficult to accurately record subcellular signals with sufficient speed and spatial resolution to avoid partial volume effects. Despite these technical challenges, all-optical electrophysiological (optophysiological) studies have begun to provide insight into the circuit activity underpinning behavior (Adam et al., 2019).

There have been initial attempts to pick up voltage changes by PA via fluorescence quenching of voltage-dependent dyes, which should increase PA intensity (Zhang H. K. et al., 2017). The fluorescent voltage sensor dipicrylamine was imaged with OR-PAM in HEK cells in vitro and PAT in vivo after administration of chemoconvulsants (Rao et al., 2017). Similarly, the voltage sensor VSD IR780 perchlorate was imaged in vivo after seizure induction (Kang et al., 2019). While VSD fluorescence changes were used as a positive control in these studies, in vivo ground-truth electrophysiology during more physiological stimulation paradigms would be desirable to assess these contrast agents’ potential.

Given the small signal change and the limited concentration and slow replacement of photobleached voltage sensors in the cell membrane, PA detection of voltage changes will remain challenging but very desirable to advance fast volumetric voltage imaging.

Calcium transients are too slow to identify the precise timing of action potentials, such as synchronicity between AP trains (Kwan and Dan, 2012; Hoang et al., 2020). However, calcium dynamics lie within the temporal resolution of many current imaging methods, and estimating action potentials by deconvolution is possible under certain conditions (Pachitariu et al., 2018; Hoang et al., 2020). Furthermore, intracellular calcium concentrations can change by two orders of magnitude in response to neuronal spiking (Schiller et al., 1995; Ali and Kwan, 2019) and can be measured in the cytosol and nucleus of the cell, where higher concentrations of calcium sensors can be achieved than possible for membrane-bound voltage sensors.

Volumetric PAI at current frame rates up to 100 fps is better suited for recording Ca²⁺ transients than action potentials due to the slower kinetics of calcium sensors (Chen et al., 2013).

Canonical calcium sensors include synthetic dyes such as Oregon Green 488 Bapta-1 AM-ester, which can reach the cytosol, where endogenous esterases cleave the AM-ester to activate and trap the sensor inside cells (Tsien, 1980, 1981; Stosiek et al., 2003). As for genetically encoded calcium indicators (GECIs), a design based on GFP fused to calcium-binding protein CaM, and its binding peptide (GCaMP) has been substantially optimized over two decades to yield robust tools for routine measurements (Miyawaki et al., 1997; Chen et al., 2013; Yang et al., 2018). Furthermore, ratiometric calcium sensors have been built by crosslinking FRET pairs via a calcium-triggered contractile element from muscle fibers (Mank et al., 2008; Trigo-Mourino et al., 2019).

As a step to enable intracellular calcium imaging with photoacoustics, a synthetic calcium sensor for photoacoustics (CaSPA) was recently designed to selectively respond with a robust blue-shift of the spectrum (metallochromism) upon calcium binding while minimizing quantum yield and ensuring high photobleaching resistance (Roberts et al., 2018) (Figure 2, upper right segment). Due to its esterified BAPTA moiety (Tsien, 1981), the compound could be delivered into cells, organoids, and into zebrafish and exhibited high selectivity. CaSPA is based on a semi-cyanine chromophore that can be modified to red-shift the absorbance spectrum, which is an obvious next goal to separate it from the brain activity-dependent hemoglobin absorbance in the visible range.

It could also be recently shown that GCaMP can be detected by PAI, which was chosen because it exhibits a strong change in absorbance instead of calcium sensors that work by a change in quantum yield (Deán-Ben et al., 2016) (Figure 2, lower right segment). In proof-of-principle experiments, simultaneous fluorescence, and photoacoustic recordings were conducted in GCaMP-expressing zebrafish larvae and ex vivo adult zebrafish brains using chemical calcium triggers. These results have been corroborated by an ex vivo murine brain preparation showing signal changes from GCaMP6f in response to an activating chemoconvulsant (Gottschalk et al., 2019a).

As with synthetic calcium sensors, one obvious path for improvement will be to develop red-shifted calcium-dependent chromoproteins, e.g., BphPs, as detailed above. For example, NIR-geco1 exhibits absorbance and emission maxima at 678 and 704 nm, respectively (Qian et al., 2019). However, the protein is only moderately metallochromic, and in vivo paw stimulation experiments resulted in only a 0.3% fluorescence signal decrease in the somatosensory cortex of mice (Qian et al., 2019), while no photoacoustic signal changes were shown. Another recently developed genetically encoded Ca²⁺ sensor, iGECI, is a NIR Förster resonance energy transfer (FRET)-based calcium indicator with a fixed maximum absorbance at 670 nm. In vivo imaging measurements yielded a negative 3% change in fluorescence after stimulation. The authors report simultaneous iGECI fluorescence and PA imaging of hemodynamics but no direct PA detection because PA signals are likely too small to exceed the noise (Shemetov et al., 2020). Further, NIR Ca²⁺ sensors like GAF-CaMP2 have also been reported to have modest absorbance change and have only been validated in cell culture (Subach et al., 2019).

None of the constructs listed above has the necessary properties for successful application in PA imaging of calcium transients in the NIR. One way to mitigate this problem might be to develop a reversibly switchable calcium sensor, applying some of the strategies illustrated in the section on reversibly switchable photochromic proteins above, and use more complex signal acquisition and data processing schemes to derive calcium dynamics. Research in this area is intense as, apart from the application in PA imaging, NIR Ca²⁺ sensors would be valuable tools for multiplexing with other visible range Ca²⁺ reporters and optogenetic actuators, enabling more complex neurophysiological experiments. Ca²⁺ imaging with PA could also benefit from progress made in image acquisition hardware and new computational strategies for spectral unmixing and image reconstruction.

A helpful step toward developing such a sensor would include the standardization of PA imaging setups and protocols for high-throughput screening of biosensor candidates (Hofmann et al., 2019).

A central signal transduction pathway in neurons, for which molecular sensors are desired, are GPCRs. GPCRs are a large family of cell surface receptors that interact with a wide variety of external signals, which regulate a myriad of cellular functions (Gurevich and Gurevich, 2019). This function makes them one of the most attractive targets for pharmaceutical intervention. Binding of an extracellular ligand to a GPCR results in G-protein activation, followed by signaling through the cAMP or phosphatidylinositol pathway, which releases intracellular calcium.

This signal transduction chain was used to generate GPCR-reporter cells that express a specific GPCR together with a calcium indicator, such that a calcium-induced fluorescence signal change answered the presence of a specific ligand. By implanting these whole-cell sensors into the brain of mice, dopamine and norepinephrine release could be monitored via fluorescence (Muller et al., 2014).

Direct fluorescence sensors for GPCR activity were designed in analogy to GCaMP by fusing circularly permuted fluorescent proteins to GPCR, which translate conformational changes in the GPCR to the fluorescent protein, altering the fluorescent signals (Patriarchi et al., 2018; Sun et al., 2018). Although these sensor mechanisms may, in principle, also be accessible by photoacoustics, the current wavelength range is far from optimal, as discussed above.

Recently, a different PA contrast mechanism was introduced based on the reversible relocalization of biosynthetic pigments (Lauri et al., 2019). To turn the static pigments into dynamic contrast changes, the authors took inspiration from cuttlefish, which can change their skin’s photoabsorbing patterns by rapidly contracting or dispersing pigment-filled compartments within specialized chromatophore cells via GPCR-induced calcium influxes. As a biomimetic analogon, cells containing melanin-filled melanosomes from Xenopus laevis were transplanted into the midbrain of juvenile zebrafish. The strong PA contrast enabled RSOM (an AR-PAM implementation) to resolve the contraction of individual melanophore cells in response to ligand-induced GPCR activity. The pigments’ contraction and dispersion could also be read out by a change in photoacoustic signal frequency as an additional observable parameter for molecular photoacoustics. This molecular sensor mechanism could be generalized by genetically reprogramming other cell types to relocalize pigments via various mechanisms (Figure 2, lower right segment).