Chemigenetic Far-Red Labels and Ca2+ Indicators Optimized for Photoacoustic Imaging

Photoacoustic imaging is an emerging modality with significant promise for biomedical applications such as neuroimaging, owing to its capability to capture large fields of view deep inside complex scattering tissue. However, widespread adoption of this technique has been hindered by a lack of suitable molecular reporters for this modality. In this work, we introduce chemigenetic labels and calcium sensors specifically tailored for photoacoustic imaging, using a combination of synthetic dyes and HaloTag-based self-labeling proteins. We rationally design and engineer far-red “acoustogenic” dyes, showing high photoacoustic turn-ons upon binding to HaloTag, and develop a suite of tunable calcium indicators based on these scaffolds. These first-generation photoacoustic reporters show excellent performance in tissue-mimicking phantoms, with the best variants outperforming existing sensors in terms of signal intensity, sensitivity, and photostability. We demonstrate the application of these ligands for labeling HaloTag-expressing neurons in mouse brain tissue, producing strong, specifically targeted photoacoustic signal, and provide a first example of in vivo labeling with these chemigenetic photoacoustic probes. Together, this work establishes a new approach for the design of photoacoustic reporters, paving the way toward deep tissue functional imaging.


■ INTRODUCTION
The understanding of neural encoding and information processing, spanning spatial scales from individual neurons to entire brain regions, is a fundamental challenge in neuroscience.A variety of direct and indirect imaging methods can be used to monitor neural activity across these scales, from macroscopic techniques such as MRI hemodynamic contrast, 1 to single cell fluorescence microscopy. 2 Among these, a prevalent approach is the monitoring of fluctuations in neuronal calcium concentration, using indicators that exhibit fluorescence changes upon binding to calcium ions. 3−7 Nevertheless, despite continuous improvements, fluorescence imaging is inherently limited to superficial tissue layers due to light scattering, which restricts imaging depth to a couple of millimeters at best. 8In contrast, photoacoustic imaging (PAI), which relies on light absorption and subsequent ultrasound emission due to the nonradiative deexcitation of chromophores, can achieve centimeter-deep imaging of large fields of view (up to ∼15 × 15 × 10 mm) at ∼50−100 μm resolution, hence standing as an ideal modality for mesoscopic-scale neuroimaging of large volumes such as the entire mouse brain. 9,10However, despite the availability of highly performant fluorescent calcium indicators, there are currently no effective photoacoustic Ca 2+ biosensors, which has hindered the widespread adoption of PAI in the neuroscience field.This can be explained by the stringent requirements for such reporters, including a high extinction coefficient (ε) in the far-red/near-infrared (NIR) to minimize background from endogenous chromophores such as hemoglobin, a low fluorescence quantum yield (Φ) to maximize radiationless relaxation and thus photoacoustic signal, high photostability, and the possibility to genetically target them to specific cells or subcellular features. 11Moreover, essential requirements including high sensitivity, selectivity, physiologically relevant binding affinity, and fast kinetics must be fulfilled by effective biosensors.The design principles of photoacoustic Ca 2+ indicators have mirrored those of their fluorescent analogues, built from either synthetic or genetically encoded chromophores.However, given the fundamentals of the photoacoustic effect, 12 sensors relying on absorption changes inherently possess the potential for markedly higher sensitivity than quantum yield-modulated sensors. 13This renders mechanisms such as photoinduced electron transfer and Forster resonance energy transfer less promising for the development of photoacoustic reporters.To date, only a limited number of calcium sensors for PAI have been reported and suffer from important limitations.Current small-molecule calcium indicators lack genetic targetability, are generally difficult to deliver in complex biological systems, and exhibit modest sensitivity. 14,15−18 Sensitive far-red GECIs have been notoriously difficult to engineer, and the few reported ones have not yet been used in PAI, 19−21 likely due to low sensitivity and/or insufficient photostability, which highlights the need for alternative approaches toward PA calcium sensors.
In this work, we report the first generation of chemigenetic reporters for photoacoustic imaging.This approach, which amalgamates small-molecule dyes with self-labeling proteins such as HaloTag, 22,23 was recently established as a platform for the design of multicolor fluorescent Ca 2+ biosensors. 24,25ecognizing the potential of this approach for PAI, we rationally engineered a series of far-red/NIR "acoustogenic" ligands, showing high turn-ons in photoacoustic signal upon binding to HaloTag (Figure 1).Adapting these ligands to calcium-sensitive HaloTag-based proteins, we developed a suite of calcium sensors with excellent performance in tissuemimicking phantoms.Importantly, our novel acoustogenic ligands can label HaloTag-expressing neurons in mouse brain tissue, providing high, specific PA contrast.

■ RESULTS AND DISCUSSION
To develop efficient labels and calcium sensors for PAI, we sought to adapt the recently established "chemigenetic" approach, initially designed for fluorescent reporters. 26,27The most common strategy uses the self-labeling protein HaloTag, with rhodamine-based ligands, which exist in equilibrium between a nonabsorbing lactone form and a highly absorbing and fluorescent zwitterionic form.When preferentially adopting the closed form in solution, dyes can exhibit fluorogenic behavior, i.e., a substantial fluorescence turn-on upon binding to the HaloTag protein, due to a shift in the equilibrium toward the open form.−30 Importantly, the HaloTag protein can be modified and appended with genetically encoded sensing motifs, yielding analyte-responsive self-labeling tags.In these tags, the conformational change of the protein upon binding to its target leads to a change in absorption and emission of the fluorogenic dye ligand. 24,25,31,32We reasoned that meticulous engineering of far-red/NIR-absorbing and nonfluorescent dyes, featuring a similar open-closed equilibrium, would lead to "acoustogenic" HaloTag ligands.These dyes, while maintaining low fluorescence, would show large turn-ons in photoacoustic signal upon binding to the HaloTag protein and could in turn function as acoustic reporters in HaloTag-based calcium sensors (Figure 1).
Design, Synthesis and Properties of Acoustogenic Scaffolds.To design highly acoustogenic ligands, we investigated different families of dyes, focusing on scaffolds combining a far-red/NIR absorption maximum with a minimal Φ and bearing the paradigmatic o-carboxylic acid group on the pendant phenyl ring responsible for the open-closed equilibrium in rhodamine derivatives (Figures 2 and S1).Following previous work on the design of chromogenic HaloTag ligands, 28,29,33 we reasoned that dyes presenting a detectable (ε > 200 M −1 cm −1 ) but low absorption in aqueous buffer could result in large absorption turn-ons of the corresponding HaloTag ligands upon binding to the HaloTag protein.As analogous far-red scaffolds are often predominantly closed, we systematically examined both H-and F-substitution on the pendant phenyl ring, as fluorination at these positions leads to a shift of the equilibrium toward the open form. 29The absorption properties of the free dyes were measured in aqueous buffer (Figures 2 and S2), and in MeCN/H 2 O mixtures (Figure S3), to characterize their open-closed equilibrium and therefore assess their potential acoustogenicity.
−36 We also examined recently reported dihydroquinoline-fused Si-rhodamines, which display NIR absorption and low Φ (compounds 6, 7). 36Compounds 1−7 were synthesized using reported methods (Synthetic Schemes in SI). 36,37Among them, compounds 1 and 3 showed no detectable absorption (ε < 200 M −1 cm −1 ), supporting that these are too closed to show exploitable chromogenicity.In contrast, compounds 2, 4, 5, 6 displayed a weak but detectable absorption (200 < ε < 2000 M −1 cm −1 ), which is indicative of an absorption turn-on of the corresponding HaloTag ligands upon binding to the protein. 29ompound 7 showed a higher extinction coefficient, and measurements in MeCN/H 2 O mixtures revealed that compounds 6 and 7 have a high propensity for aggregation in water, with 7 presenting ε = 98,000 M −1 cm −1 in 40% MeCN/ H 2 O, suggesting that the dye is fully open but strongly aggregates in aqueous buffer (Figure S3).
Next, we investigated the behavior of pyrrole-functionalized O-xanthenes, a recently reported class of NIR dyes, 38 and synthesized compounds 8−11 by Suzuki coupling from the fluorescein ditriflates.These dyes presented absorption maxima between 637 and 679 nm (a 90−110 nm red-shift compared to their rhodamine relatives) due to an extended electronic conjugation.8, 10, and 11 were strongly shifted toward the closed form, but tetrafluorinated 9 showed a higher ε = 5100 M −1 cm −1 .This value is far lower than the expected maximum for these scaffolds (∼10 5 M −1 cm −1 ), characteristic of possible chromogenic behavior.
Finally, we examined triarylmethane lactones (compounds 12-19), closely resembling Malachite Green, which absorbs in the far-red and is virtually nonfluorescent in solution due to rotational flexibility. 39,40The parent nonfluorinated (12) and tetrafluorinated ( 13) Malachite Green lactones were synthesized in one step by Friedel−Crafts acylation of N,Ndimethylaniline with the corresponding phthalic anhydride.While 12 was too closed, compound 13 with ε = 13,900 M −1 cm −1 , was structurally reminiscent of the landmark fluorogenic dye SiR, 41 and had substantial room for further functionalization.We hypothesized that strategies to shift the dye toward the closed form which have been developed for rhodamines could be applied to this scaffold, in order to decrease the absorption of the free dye, and in turn lead to higher chromogenicity and acoustogenicity.Following a well-established approach, we set out to replace the N,N-dimethyl substituents with various substituted azetidines (14, 15, 16). 28he reactivity of the azetidines prevented synthesis of these target compounds by Friedel−Crafts acylation or Pd-catalyzed cross coupling, and they were synthesized using a recently reported synthetic method involving lithiation of 2,3,4,5tetrafluorobenzoic acid before addition to the corresponding benzophenones. 36The trend observed was very similar as for SiR, with electron withdrawing substituents shifting the equilibrium toward the closed form.Among these azetidinesubstituted derivatives, compounds 14 and 16 displayed extinction coefficients in the desired range (4700 and 2100 M −1 cm −1 , respectively).As a second approach to tune the equilibrium, we examined the effect of substituents on the aromatic rings, introducing fluorines at the 2′,7' positions (17), previously shown to shift the equilibrium toward the closed form. 42Unfortunately the fluorination had too strong an effect, completely closing the dye.We also introduced methyl groups on the 1′,8' positions (18), a modification shown to elicit small wavelength shifts in arylmethane dyes, 43 but so far unexplored in the context of the open-closed equilibrium or rhodaminelike dyes.Interestingly, this resulted in a small shift toward the closed form, with a 2-fold lower extinction coefficient for 18 compared to 13, along with a favorable ∼30 nm red-shift in absorption.Finally, we also investigated the impact of partial defluorination of the bottom ring with compound 19, serendipitously isolated as a byproduct during the synthesis of the HaloTag ligand.As expected, this previously unexplored modification provided a desired small shift of the equilibrium toward the closed form (ε = 10,700 M −1 cm −1 ).
Acoustogenic HaloTag Ligands as Photoacoustic Labels.Based on the absorption properties of the free dyes, we synthesized and characterized the HaloTag ligands of selected compounds in each family (presenting 200 < ε < 15,000 M −1 cm −1 , Figure S4).These were synthesized by standard amide coupling from the 6-CO 2 H precursor for nonfluorinated ligands, and directly from the free dyes for the fluorinated compounds, using an adapted version of the recently published masked acyl cyanide chemistry. 29We measured their photophysical properties in the absence or presence of HaloTag protein (Figures 3, S5, and Table S1).All ligands showed an absorption increase upon binding to HaloTag7, albeit to different extents, with dyes with lower extinction coefficients generally leading to higher turn-ons upon binding (Figure S6).Si-rhodamine 4-HTL showed a high turn-on, with ΔA/A 0 = 62.This compound however showed very slow binding (>24 h in solution), which is prohibitive for biological applications (Figure S7).We hypothesized that this was primarily due to unfavorable interactions of the phenyl rings on the protein surface, so attempting to minimize this deleterious effect, we synthesized the unsymmetrical compound 20-HTL, by replacing one aniline with an azetidine (Figure S4).This compound maintained high absorption turnon upon binding (ΔA/A 0 = 42) and low Φ < 0.01, and showed substantially faster binding to the protein than the symmetrical 4-HTL, although still slower than conventional HaloTag ligands (Figure S7).Piperazine-SiR 5-HTL was too closed, with low absorption when bound to HaloTag, and additionally presented a relatively high protein-bound Φ = 0.11, so it was therefore deemed unsuitable for PA.Compound 6-HTL showed fast binding kinetics and a high ΔA/A 0 = 94, but similarly a high Φ = 0.13 with HaloTag.In the pyrrolexanthene family, fluorinated 9-HTL showed a good ΔA/A 0 = 9.2.Finally, we investigated the behavior of the Malachite-Green-derived HaloTag ligands.To further expand the series of fine-tuned derivatives, we additionally synthesized the hydroxy-substituted 21-HTL, hypothesized to present improved water solubility.We also synthesized the difluoro derivative 22-HTL, by hydrodefluorination of the ester intermediate.We note that this approach had not been previously explored to fine-tune dyes, and the regioselectivity of the reaction was confirmed by X-ray crystallography, showing removal of the fluorine at the 4-position in the major isolated product (see Crystallography in SI).The Malachite Green derivatives showed rapid binding kinetics, with good ΔA/A 0 up to 21 and generally followed the expected trend, with dyes more strongly shifted toward the closed form showing higher ΔA/A 0 , but lower protein-bound absorption.Importantly, these compounds remained virtually nonfluorescent after binding (Φ < 0.01), demonstrating that HaloTag does not elicit the conformational restriction observed with fluorogen activating proteins for Malachite Green which results in a fluorescence turn-on undesirable for PAI. 44In an attempt to reduce the trade-off in HaloTag-bound absorption, we also evaluated selected ligands with HaloTag9, a recently published mutant which shows higher absorption turn-on with certain rhodamine derivatives (Table S1 and Figure S5). 45Although we observed similar absorption as with HaloTag7 for the HTLs of 14, 16, and 22, the absorption of the bound dye to HaloTag9 was up to 70% higher for compounds 9-HTL, 13-HTL and 18-HTL, showing that our ligands have potential for even greater turn-ons with tailored protein engineering.Overall, six of the ligands 9-HTL, 13-HTL, 14-HTL, 16-HTL, 18-HTL, 22-HTL showed fast labeling kinetics, large absorption turn-on, high far-red/NIR absorption and low Φ upon binding to HaloTag, suggesting excellent performance as photoacoustic labels.
We therefore measured their photoacoustic properties, using a custom-built spectrometer (Figure S8).Specifically, the spectral features and photoacoustic turn-ons upon HaloTag binding (ΔPA/PA 0 ) at λ PA = λ max , were in excellent agreement with absorption measurements, confirming that absorption can serve as a robust proxy for the design of such photoacoustic reporters in which fluorescence is minimal (Φ < 0.01) (Figures S9 and S10).To support the proof-of-principle of our approach, we compared the photoacoustic signal intensity of 13-HTL with the spectrally matched highly fluorescent JF635-HTL 28 (Figure S11).While both HaloTag conjugates have comparable extinction coefficients, the photoacoustic signal was 17-fold larger for 13-HaloTag7, explained by the fact that its fluorescence quantum yield (Φ ∼ 0.001) is about 750-fold smaller than that for JF635-HaloTag7 (Φ = 0.75).This shows that strong suppression of fluorescence is a paramount requirement for high photoacoustic signal in this type of chromophore.
Design and Properties of Photoacoustic Calcium Sensors.Given the high performance of our acoustogenic HaloTag ligands in vitro, we then set out to use them for the design of Ca 2+ sensors.For this purpose, we focused on the Ca 2+ -sensitive self-labeling proteins HaloCaMP and rHCaMP. 24,25In these sensors, the conformational change of the protein upon binding calcium ions alters the local environment of a HaloTag-ligand fluorophore, hereby shifting its equilibrium toward the open form with a concomitant absorption increase.We reasoned that these calcium-binding self-labeling proteins would lead to sensitive photoacoustic sensors when used with our acoustogenic ligands.Indeed, all novel ligands tested showed a calcium-dependent change in absorption and photoacoustic signal when used with these platforms (Table S2, Figures S12, and S13).Generally, the two HaloCaMP variants, 1a and 1b, afforded positive-going sensors (i.e., with PA signal turn-on upon binding calcium), while rHCaMP led to an inverse response with the same ligands.Importantly, absorption and photoacoustic properties were again in excellent agreement for the calcium sensors (Figures 4a and S14).Compound 9-HTL and Malachite Green derivatives 13-HTL, 14-HTL and 18-HTL showed comparable sensitivity with HaloCaMP1a (ΔA/A 0 between 0.7 and 1.4,  ΔPA/PA 0 between 1.1 and 2.0), and larger sensitivity with HaloCaMP1b, with ΔPA/PA 0 up to 7.5.Following the trend observed with HaloTag, compounds 16-HTL and 22-HTL showed low absorption and photoacoustic signal in the Ca 2+bound state and were therefore excluded.Overall, the results align with the observations made with the Janelia Fluor ligands which HaloCaMP was originally evolved against, with variant 1a generally resulting in higher absorption, and variant 1b, leading to a larger ΔA/A 0 with the same dye. 24With ligand 13-HTL, both HaloCaMP variants showed reasonable pH stability in the physiological range of the neuronal cytosol (pH ∼ 7), similar to the original HaloCaMPs with JF635 (Figure S15), and high selectivity for Ca 2+ over Mg 2+ (Figure S16).We then performed calcium titrations to determine the calcium affinity of these sensors (Figure 4b,c and Table S2).The four ligands tested provide high affinity sensors with HaloCaMP1b with a dissociation constant (K d ) ranging from 10 to 33 nM, which is in the same range as the original HaloCaMP1b sensors, 24 and GECIs such as jGCaMP7s, 46 thus well-suited for in vivo imaging of cytosolic neuronal calcium.Surprisingly, using the same dyes with the HaloCaMP1a scaffold led to low affinity sensors, with K d orders of magnitude larger than with the original JF derivatives (∼60−900 nM). 2413-HaloCaMP1a and 18-HaloCaMP1a clearly showed two binding phases (K d 1 = 300 and 980 μM, K d 2 = 18 and 10 mM, respectively), while 9-HaloCaMP1a and 14-HaloCaMP1a presented an apparent single binding event, with a K d of 6.9 and 4.8 mM, respectively.The mechanism by which the dye ligand determines the apparent affinity of the sensor is unclear.−50 Overall, 13-HaloCaMP1a provided the largest photoacoustic signal in the Ca 2+ -bound state (Figure 4d), and 18-HaloCaMP1b showed the highest sensitivity, with ΔPA/ PA 0 = 7.5 (Figure 4e).Our probes are also far more photostable than competing protein sensors like the far-red, biliverdin-dependent, fluorescent GECI NIR-GECO1 19 (Figure S17).After illumination for 1 h at λ max , the photoacoustic signal of NIR-GECO1 decreased to 65%, while 13-HaloCaMP1a and 13-HaloCaMP1b retained 90 to 95% of their original signal.The parent chromophores show a similar trend, with 13-HaloTag7 substantially more photostable than mIFP (Figure S17a,c), and the relative photostabilities between mIFP and NIR-GECO1 are consistent with reported values. 19,51Overall, ligands 9, 13, 14 and 18 used in combination with HaloCaMP variants provide a series of eight photoacoustic Ca 2+ -indicators with a wide range of affinities, signal intensities, and sensitivities.Importantly, these indicators clearly outperform existing sensors in the far-red/ NIR wavelength range such as NIR-GECO1 and published synthetic PA calcium sensors in sensitivity, 14,19 with the benefit of showing a photoacoustic signal increase upon binding calcium (Figure 4e and Table S2).
Photoacoustic Tomography in Tissue-Mimicking Phantoms.We then set out to characterize and evaluate our new acoustogenic probes in tissue-mimicking phantoms, using a custom-built all-optical Fabry-Peŕot-based photoacoustic tomography setup, which is ideally suited for deeptissue imaging in mice. 52,53−56 The tubes were filled with solutions of purified dye-protein conjugates and positioned at the desired depth from the Fabry−Peŕot detector.We evaluated the performance of our HaloTagbased photoacoustic labels and compared it against mIFP, the chromophore component of NIR-GECO1. 19Ligand 13 bound to HaloTag7 gave an ∼6-fold higher PA signal than mIFP (Figure S18), with detectable signal down to 1.2 cm depth in the 60% milk phantom, much deeper than the mouse brain (∼6 mm) (Figure S19).In contrast, mIFP gave a much lower signal and was not detectable below 8 mm in these scattering conditions.At a depth of 5 mm, equivalent to close to the bottom of the mouse brain, a detection limit of ∼1.5 μM was measured for 13-HaloTag7 in those brain scattering representative conditions, which is well within the protein concentration range achievable in vivo with adeno-associated viruses (Figures 5b and S20). 57,58Notably, the HaloCaMPbased calcium sensors led to similar turn-ons in the phantoms as in spectroscopy, outperforming NIR-GECO1 which showed only a small decrease in signal (Figures 5c and S21).In the Ca 2+ -bound state, 13-HaloCaMP1a gave twice the photoacoustic signal of NIR-GECO1 and 13-HaloCaMP1b, in a scattering medium at ∼5 mm depth (Figure S22).
Labeling and Photoacoustic Tomography of Mouse Brain Tissue.Consequently, we set out to demonstrate the use of newly developed reporters in biological systems.We first assessed the cell permeability and labeling efficiency of the ligands in U2OS cells expressing HaloTag7, fused to the enhanced green fluorescent protein (EGFP) to quantify protein expression (Figure S23).Very faint fluorescence signals from the dyes could be observed in the far-red channel.The bright competitor JF549-HTL was subsequently added to these prelabeled cells, 28 and the minimal fluorescence signal in this channel confirmed that the ligands tested (9-HTL, 13-HTL and 14-HTL) efficiently bind to HaloTag in living cells, with about 95% of available protein labeled (estimated by the bound fraction of the competitor, Figure S23b).Moving to more complex tissues, we then performed labeling of mouse brain slices for photoacoustic imaging.For this purpose, mice were injected with an adeno-associated virus serotype 1 (AAV1) HaloTag7-EGFP under the synapsin promoter into the hippocampus in the right-hand hemisphere of the brain.Following expression, brains were excised, sliced coronally and labeled with 13-HTL by bath loading.Successful specific labeling of HaloTag-expressing neurons in the hippocampus was observed via widefield fluorescence microscopy, as shown by the correlation of the EGFP fluorescence and the weak farred fluorescence from 13-HTL, which was absent prior to labeling (Figures 6 and S24).Photoacoustic tomography showed a strong, specific photoacoustic signal from the labeled hippocampal slices.In contrast, brain slices lacking HaloTag expression showed no detectable fluorescence or photoacoustic signal from 13-HTL after incubation with the dye, demonstrating the specificity of the labeling (Figure 6b).Similar results were obtained by bath labeling with 14-HTL, showing a strong and specifically localized PA signal (Figure S25).
Finally, we set out to investigate the labeling of mouse brains in vivo for photoacoustic imaging.Following stereotactic AAV1 HaloTag7-EGFP injection into the righthand hippocampus and thalamus, dye labeling was performed in vivo via intracerebroventricular injection of 13-HTL in the left lateral ventricle.Photoacoustic tomography of the entire excised brain showed no signal in the control mice (no dye was injected, Figure 7a).In the dye-injected mice, we observed variable results, which support the potential of this dye for in vivo applications, while evidencing limitations in bioavailability and/or the dye delivery method.In the most efficiently labeled mouse, strong photoacoustic signal at λ PA = 646 nm was observed in close proximity to the viral injection site in the hippocampus (Figure 7b).To confirm that the signal specifically arises from the HaloTag-bound dye, the brain was fixed, sliced and fluorescence imaging was performed.Based on the mouse brain atlas, 59 the slices were mapped to their corresponding coronal section in the photoacoustic tomography (PAT) volume.The PA signal unambiguously correlated with the far-red fluorescence signal from the dye in the hippocampus, where HaloTag is strongly expressed, as evidenced by EGFP fluorescence (Figures 7c−f and S26).
While EGFP was also clearly expressed in the thalamus, a negligible fluorescence signal from the dye ligand was visible in  that region (Figures 7d−f and S27c).In addition, substantially less fluorescence signal (∼6-fold) was observed in the hippocampi of the other mice, highlighting the variability of the labeling (Figure S27c).We also evaluated compound 14-HTL in vivo, which also led to labeling of the hippocampus, evidenced by the specific marking of individual neurons visualized by fluorescence (Figure S28).Together, these results demonstrate that our acoustogenic ligands can specifically label neurons in mouse brain tissues while highlighting that further improvements will be needed for reliable in vivo delivery, ideally systemically.

■ CONCLUSION
While a variety of photoacoustic probes have been reported, such as contrast agents for imaging vasculature, tumors and anatomical structures, and reporters showing irreversible activation by biomolecules or analytes of interest, 60−62 the field is still lacking specific, dynamic reporters, that can be easily targeted to cells and subcellular features of interest.To resolve these issues, we have repurposed the chemigenetic strategy for the design of genetically targeted photoacoustic probes and calcium sensors, based on the HaloTag self-labeling protein and acoustogenic dye ligands.We synthesized and systematically investigated the photophysical properties of a library of 20 dyes.Through synthetic modifications, we rationally optimized the scaffolds for strong photoacoustic signal in the far-red/NIR region and high calcium sensitivity when used in combination with the calcium-sensitive HaloCaMP protein.In the process, we establish the centuryold dyestuff Malachite Green as a robust photoacoustic probe and additionally introduce novel synthetic methods to modify its properties, specifically 1′,8′-methylation and monohydrodefluorination.These modifications, previously unexplored, are likely to be broadly applicable to rhodamine derivatives, thereby expanding the toolbox of general methods to tune dyes.Overall, this work resulted in a series of acoustogenic dye-ligands with up to a 12-fold photoacoustic turn-on, and ultimately 8 calcium indicators with ΔPA/PA 0 up to 7.5 and affinities ranging from nM to mM.These first-generation chemigenetic probes for PAI constitute, to the best of our knowledge, the first example of far-red, positive-going calcium sensors optimized for this modality.Our probes exhibit superior in vitro performance, demonstrated through both spectroscopic and tomographic photoacoustic measurements, outperforming existing PA sensors.Notably, we achieved specific labeling of HaloTag-expressing neurons in mouse brain tissue, leading to a strong photoacoustic signal that could be readily visualized via PAT, showing high potential for in vivo neuroimaging applications.The variable in vivo labeling observed suggests that the bioavailability of the dye ligand is an important limiting factor, affected by solubility, slow diffusion, and tissue permeation.Current efforts in our group focus on improving the bioavailability of these acoustogenic dye ligands, a critical property to achieve robust whole brain labeling.
A key feature of the chemigenetic approach is the high tunability of the system afforded by the synthetic component, precluding the need for re-engineering the protein scaffold.Here, by developing photoacoustic ligands that exhibit high performance with established HaloTag and HaloCaMP proteins, we extend the "plug-and-play" versatility of these systems to imaging across various modalities, simply by varying the small-molecule dye ligand.This feature stands as a powerful asset for imaging across spatial scales using different contrast mechanisms.Nevertheless, further refinements of these first-generation sensors are conceivable through protein engineering, which will require innovative high-throughput screening methods and instrumentation tailored for this modality.We envision that continued improvements in photoacoustic hardware, along with further molecular engineering to refine sensor properties and improve dye bioavailability, will pave the way toward noninvasive calcium imaging across the entire mouse brain, thus positioning photoacoustic imaging to reach its full potential for functional neuroimaging.

■ ASSOCIATED CONTENT
* sı Supporting Information The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c07080.Supplementary figures, general methods, synthetic procedures, and characterization for all new compounds (pdf)

Figure 1 .
Figure 1.General approach for the design of chemigenetic photoacoustic labels and calcium sensors based on the open-closed equilibrium of acoustogenic dyes.

Figure 2 .
Figure 2. General structure, open-closed equilibrium and absorption properties of the dyes studied in this work; (a) Si-rhodamine and pyrrolexanthene derivatives and (b) malachite green lactone derivatives; All measurements were made in 10 mM HEPES, pH = 7.4 except for a in 2,2,2trifluoroethanol containing 0.1% trifluoroacetic acid.b Apparent value due to aggregation.Values are the mean of 3 replicates.

Figure 3 .
Figure 3. (a) General structure of HaloTag ligands and formation of the dye-HaloTag conjugate; (b) extinction coefficient of the HaloTagbound dye vs absorption turn-on upon binding for the HaloTag ligands."HTL" was omitted from the labels for clarity.Yellow: Sirhodamines, blue: pyrrole-xanthenes, red: Malachite Green derivatives.The dashed box highlights compounds presenting both large absorption turn-on and high extinction coefficient bound to HaloTag; Values are the mean of 3 replicates; (c) normalized absorption spectra of selected HaloTag ligands in the absence or presence of HaloTag protein.

Figure 5 .
Figure 5. (a) Schematic representation of the tissue-mimicking phantom setup showing the relative position of the polyethylene (PE) tubes and the Fabry−Peŕot interferometer; (b) PAT maximum intensity projections and line profile (250 μm thick) quantification of 13-HaloTag at 5, 10, 25, or 50 μM, at ∼5 mm depth in 60% milk/ H 2 O scattering medium; (c) PAT maximum intensity projections and quantification of 13-HaloCaMP1a, 13-HaloCaMP1b or NIR-GECO1 in the calcium-bound and the calcium-free states at 50 μM, at ∼3.5 mm depth in 60% milk/H 2 O scattering medium; PAT was performed at λ PA = λ max for each probe.Representative images of 3 replicates.

Figure 6 .
Figure 6.Fluorescence microscopy and photoacoustic tomography of mouse coronal brain slices labeled ex vivo with 13-HTL; (a) Slice with high expression of HaloTag-EGFP in the hippocampus and (b) slice with no detectable expression of HaloTag-EGFP.From left to right: fluorescence images of EGFP channel; fluorescence images of 13-HTL channel; overlay of EGFP and 13-HTL channels; photoacoustic tomography maximum intensity projections (∼2 mm thickness) at λ PA = 646 nm.Magenta arrows highlight signals from the bound dye ligand.(a) Representative images of 16 slices and (b) representative images of 4 control slices, obtained from 3 mice.Scale bars: 1 mm.

Figure 7 .
Figure 7. Photoacoustic tomography of a whole ex vivo mouse brain expressing HaloTag7-EGFP in neurons, labeled with 13-HTL delivered by intracerebroventricular injection in vivo, and fluorescence images of coronal slices; (a, b) PAT maximum intensity axial projection (λ PA = 646 nm) of the entire mouse brain between 2.5 and 5.75 mm from the surface of Fabry−Peŕot interferometer to exclude strong endogenous signal from the olfactory bulbs: (a) control, unlabeled mouse and (b) mouse labeled with 13-HTL in vivo via intracerebroventricular injection.The magenta arrow indicates signal from 13-HaloTag7, the white vertical line indicates the position of the coronal slice; (c) Selected coronal slice image from the PAT volume and corresponding widefield fluorescence images of (d) EGFP channel with white arrows indicating the hippocampus (H) and thalamus (T) regions, (e) far-red channel, and (f) the merge of the two channels; Images are from one of three mice injected with 13-HTL.Scale bars: 1 mm.
CCDC 2336812 contains the supplementary crystallographic data for this paper.These data can be obtained free of charge via www.ccdc.cam.ac.uk/data_request/cif, or by emailing data_request@ccdc.cam.ac.uk, or by contacting The Cambridge Crystallographic Data Centre, 12 Union Road, Cambridge CB2 1EZ, UK; fax: +44 1223 336033.