HaloTag‐Based Reporters for Fluorescence Imaging and Biosensing

Visualizing the structure and dynamics of biomolecules is critical to understand biological function, and requires methods to fluorescently label targets of interest in their cellular context. Self‐labelling proteins, which combine a genetically encoded tag with a small‐molecule fluorophore, have attracted considerable attention for this purpose, as they can overcome limitations of fluorescent proteins. Among them, the HaloTag protein is the most broadly used, showing fast specific labelling with a small, easy to functionalize and cell‐permeant ligand. Synthetic chemistry and protein engineering have provided a portfolio of powerful imaging tools exploiting HaloTag, along with general methods to optimize and adapt them to specific applications. Here, we provide an overview of fluorescent reporters based on the HaloTag protein for imaging and biosensing, highlighting engineering strategies and general applications.


Introduction
Fluorescence microscopy is the modality of choice for visualizing the structure and dynamics of biological systems, ranging from isolated biomolecules to entire organisms. This is inherently dependent on the availability of robust fluorescent reporters combining labelling specificity with excellent optical properties. Fluorescent proteins (FPs) have revolutionized fluorescence microscopy, offering simple, widely accessible imaging inside living cells and organisms. [1] However, despite important improvements, FPs still suffer from several limitations. They generally display low brightness and photostability especially in the red-shifted region, are intrinsically pH sensitive and require oxygen for chromophore maturation. Additionally, fine-tuning their optical and biochemical properties remains challenging, with fully rational optimization yet to be achieved. To overcome these issues, substantial effort has been dedicated to adapting synthetic fluorophores for labelling proteins in biological systems. A groundbreaking approach involves the use of protein tags, which selectively bind to a synthetic ligand that can be easily appended to small-molecule fluorophores. This strategy combines the highly attractive photophysical properties of synthetic dyes with the genetic specificity of the protein tag. Since the pioneering tetracysteine tag, developed by Roger Tsien, [2] the past two decades have seen a variety of these hybrid, "chemigenetic" [3] labelling systems emerge. [4] Among them, HaloTag combines multiple advantageous features including extremely fast covalent labelling and high specificity with a small cell-permeant ligand. [5] HaloTag was engineered in 2008 from the Rhodococcus dehalogenase DhaA, a 33 kDa bacterial enzyme that catalyzes the conversion of haloalkanes to hydroxyalkanes through nucleophilic substitution followed by base-catalyzed hydrolysis. The reactive site is located at the end of an access tunnel, approximately 15 Å from the protein surface. Mutation of residues around the catalytic site, in particular of the histidine base (H272), prevented release of the substrate and resulted in a covalent ester bond (Figure 1a,b). Several linker structures were investigated, establishing the canonical structure of the HaloTag ligand (HTL) which consists of a chlorohexane and two ethylene glycol motifs. The HTL is most commonly appended with a primary amine head group, providing easy synthetic attachment to CO 2 H-substituted fluorophores by formation of an amide bond. Virtually all fluorophore ligands to date use the original PEGchlorohexane ligand design (with different PEG lengths in some instances). However, recent work shows that re-engineering of the alkane chain can be beneficial for some substrates, and that replacement of the chloride head group can expand the functionalities of the system. [6] HaloTag was originally engineered using fluorescein and tetramethylrhodamine (TMR) as fluorophores, with the ligand appended on the 6-position of the pendant phenyl ring. Several rounds of mutagenesis dramatically increased the kinetics of the labelling reaction with these dye ligands and improved protein expression, stability and solubility. [7] The resulting HaloTag7 (hereafter referred to as HaloTag) is to date the primary variant in use, with an apparent rate constant of 1.9 × 10 7 M À 1 s À 1 with TMR, orders of magnitude faster than other covalent self-labelling tags such as SNAP-Tag (4.3 × 10 5 M À 1 s À 1 ) or CLIP-Tag (1.9 × 10 4 M À 1 s À 1 ). [8] HaloTag has found a wide variety of applications beyond imaging including in vitro analysis (such as biomolecule immobilization and protein assays), and perturbation or manipulation of biological systems, which have been reviewed elsewhere. [9] Nevertheless, fluorescence imaging remains its main area of application. The HaloTag protein can be fused to the terminus of a protein of interest, similarly to fluorescent proteins, and advances in genome editing greatly facilitate its use to visualize endogenous targets. [10] Considerable effort has been dedicated to designing new fluorescent substrates for labelling in live samples, and to engineering advanced functional reporters such as biosensors, in order to visualize highly dynamic processes in complex biological environments. In this review, we detail the different molecular scaffolds and engineering strategies for advanced reporters for fluorescence microscopy using HaloTag, and highlight their general applications.
a HaloTag ligand, assuming an appropriate linker length to prevent the fluorophore from clashing with the protein. The past decade has seen tremendous advances in the design and optimization of fluorophores for use with HaloTag. This has resulted in a plethora of bright and photostable multicolor ligands and rational generalizable methods for tuning their physicochemical properties.

Fluorophore ligands based on traditional dye scaffolds
Representatives of most of the common families of fluorophores have been adapted to HaloTag, providing a palette of dye ligands with optical properties covering the entire spectral range. [11] Several multicolor ligands are commercially available through Promega, the company that originally developed the HaloTag technology. [12] The list includes landmark fluorophores such as the blue 7-amino-4-methylcoumarin (Figure 2a), the green fluorescein, and its fluorinated derivative Oregon Green (Figure 2b), [13] with the HaloTag ligands retaining the photophysical properties of the parent dyes. As fluorescein derivatives are generally cell-impermeant, acetyl ester-masked versions are also available. These non-fluorescent derivatives can cross the cell membrane, before release of the constitutive fluorophore through cleavage by intracellular esterases. [14] Across the visible range, rhodamine derivatives have been by far the most widely used fluorophores, and the next section will discuss in detail the diversity of rhodamines applied as HTLs (section 2.2). In the far-red region of the spectrum, oxazine HaloTag ligands include ATTO655, used in living cells for single molecule localization microscopy (SMLM) of the actin cytoskeleton (Figure 2c), [15] and ATTO700 which shows good performance for STED Alexander Cook is a PhD student in Dr. Claire Deo's group at EMBL Heidelberg. He studied Biology and Chemistry within the Natural Sciences programme at Durham University (UK) during his MSci degree, which he finished in 2020. Since starting his PhD later that year in the group, his research focuses on the development of hybrid probes and sensors for photoacoustic imaging.
Franziska Walterspiel studied Chemistry at the University of Heidelberg (Germany) and completed her MSc degree at the Max Planck Institute for Medical Research in 2020. Currently, she is a PhD student in the group of Dr. Claire Deo at EMBL Heidelberg. Her PhD project focuses on the development of hybrid photoswitchable systems for super-resolution microscopy.
Claire Deo is a group leader at EMBL Heidelberg, Germany. She holds a PhD from ENS Paris-Saclay (France) for which she developed photoswitchable catalysts. Her postdoctoral work at Janelia Research Campus (USA) focused on engineering new fluorophores and calcium biosensors. Research in her group focuses on the development of new molecular tools for biological imaging across scales, at the interface of synthetic chemistry and protein engineering. microscopy. [16] Cyanines, with their large extinction coefficients and high photostability, have been the preferred far-red fluorophores, particularly useful for imaging deep in tissue. For example, IR800 with an extended HaloTag ligand could be applied to labelling tumor receptors in mice (Figure 2d). [17] Finally, HaloTag is also compatible with inorganic emissive systems, including organometallic complexes, [18] and nanomaterials such as quantum dots, capable of labelling cell-surface receptors to visualize processes such as endocytosis. [19]

Rhodamine-based ligands
Several key properties make the rhodamine family the fluorophore scaffold of choice for biological imaging. In addition to their high brightness and tunability, they exist in equilibrium between a non-fluorescent, closed form, and an open, highly emissive form (Figure 3a), which can be exploited for the design of fluorogenic dyes (see section 2.3). [20] HaloTag, which was engineered using TMR-HTL, has a clear preference for rhodamines as fluorophore partners, showing up to several orders of magnitude faster binding than with other substrates. [8] A tremendous effort has been dedicated to improving rhodamine dyes, and applying them as HaloTag labels. The structures and properties of the dyes discussed in this section are reported in Figure 3.
Multicolor rhodamines. Along with the original ligand TMR, the simplest member of the rhodamine family, the greenemitting Rhodamine 110 (Rh110) and its cell-impermeant sulfonated counterpart Alexa Fluor 488 (AF488) [21] are also commercially available as HaloTag ligands. A variety of strategies have been reported to improve and diversify the properties of these simple scaffolds, and provide multicolor, bright and photostable ligands with different biochemical properties. Large red-shifts in absorption and emission have been achieved by replacement of the central heteroatom of the xanthene core. Substitution of the oxygen bridge in TMR with CMe 2 or SiMe 2 provided CRh and SiR, with absorption maxima red-shifted by around 60 nm and 90 nm respectively. [22] SiR was a landmark dye, as one of the first cell-permeable, bright far-red fluorophores. [23] Its use with HaloTag enabled multiplexed imaging in living cells, including in STED microscopy, [10a] showing higher cellular brightness and performance than when used with the self-labelling protein SNAP-Tag. [24] Further redshifts can be achieved with bridging SO 2 or P(O)R groups, leading to near-infrared (NIR) HaloTag ligands such as NR666. [25] However, the fluorescence quantum yield of these NIR dyes is substantially lower. Various other central atom substitutions such as NMe, [25a] Ge, [26] S, [27] or cycloketal [28] have been introduced in rhodamine HaloTag ligands displaying green to far-red emission and retaining efficient binding in cells, which further demonstrates the versatility of the rhodamine core.
Improving and fine-tuning photophysical and biochemical properties. Several strategies have been reported to improve the photophysical properties of rhodamines across the entire spectral range. A successful approach involves modification of the N-alkyl substitution. N-monoalkylation of Rh110 with simple alkyl groups results in a small red-shift, providing the bright ligands 520R and 500R. [29] Fusing the N-alkyl groups to the xanthene core increases brightness and simultaneously elicits a further red-shift, providing bright ligands such as Alexa Fluor 594 (AF594) [30] and 580R, the latter containing additional hydroxyl groups to improve water solubility while retaining cell permeability, and has been successfully applied to live cell STED imaging. [26,29] A major advance by the Lavis group was the replacement of the N,N-dimethylamino groups in TMR with azetidine rings, resulting in the Janelia Fluor (JF) dyes. [31] This simple structural modification minimizes quenching originating from twisted internal charge transfer (TICT), resulting in a substantial increase in brightness as exemplified with JF549, 2.8-fold brighter in vitro than the spectrally matched TMR. JF549-HaloTag ligand shows excellent performance in live cells, with fast labelling kinetics and high specificity. It is still to date one of the brightest rhodamine ligands, used in numerous biological experiments ranging from single molecule imaging to labelling in living organisms. [32] The azetidine modification is generalizable and was also implemented on C-and Si-rhodamines to provide the bright JF608 and JF646. [31] Furthermore, introducing substituents on the 3-position of the azetidine groups results in small shifts in wavelength without compromising brightness, yielding fine-tuned dyes spectrally matched to standard laser lines such as the 3,3-difluoroazetidine substituted JF525. [33] Substitution on the azetidine rings additionally offers a straightforward method to adjust biochemical properties for different applications. For example, appending carboxylate or sulfonate groups to the azetidines of JF dyes can generate cellimpermeant ligands, retaining the brightness of the parent rhodamine. [34] Other modifications of the rhodamine scaffold and their use as HaloTag ligands have been recently reported and further diversify their photophysical and biochemical properties. Hell and coworkers replaced the N-alkyl groups with cyano substituents, resulting in negatively charged dyes such as CR1, which, while spectrally comparable to TMR, can show less offtarget labelling inside cells. [35] Alkylation at other positions on rhodamines has been scarcely reported. However, it was recently shown that replacement of the methyl groups on the central Si atom of SiR could result in a small red-shift and increase in brightness while providing an alternative position to append functional groups including the HaloTag ligand. [36] Simple structural modifications can also improve fluorophore photostability, a key requirement for extended and quantitative imaging. Although the photobleaching mechanisms of rhodamines are complex and not fully understood, [37] several dye ligands with improved photostability have been reported. For example, N,N'-di-tert-butylrhodamine (tBuRh) shows less propensity for photoinduced dealkylation than TMR. [38] Recently, the replacement of the hydrogen atoms in rhodamine N-alkyl groups with deuterium led to TMR-d12, 26 % brighter in vitro and slightly more photostable than TMR, as well as the pyrrolidine-d8-substituted JFX554 and JFX650 which also show higher brightness and photostability than their hydrogenated counterparts, establishing an isotopic effect to increase the performance of rhodamines and their corresponding HaloTag ligands. [39]

Fluorogenic ligands
While the majority of the ligands described above are "always on" fluorophores, HaloTag provides the opportunity to generate fluorogenic ligands, which show an increase in fluorescence upon binding to their target. These are highly desirable for imaging, reducing background and circumventing the need for extensive washing steps, and provide a mechanism for fluorescence modulation which can be further exploited for the design of functional reporters (discussed in sections 3 and 4). Different approaches have been reported for the design of highly fluorogenic HaloTag ligands, exploiting the interaction of environment-sensitive fluorophores with the protein surface to alter their emission properties ( Figure 4).
Fluorogenic ligands based on a change in electronic conjugation. Rhodamine derivatives bearing a 3-carboxylic acid group exist in equilibrium between a closed, non-fluorescent lactone form, and an open, fluorescent zwitterionic form ( Figure 3a). This equilibrium is environment-sensitive, which can lead to a fluorescence turn-on upon binding to a biomolecular target. Although this has also been observed with other tags such as SNAP-Tag, the fluorescence turn-ons are generally higher with HaloTag, [40] which is another highly attractive feature of this tag. Structural insights support that this behavior arises from the tight interaction of the dye with protein surface residues, shifting the equilibrium towards the open form ( Figure 4a). [8] As a result, the fluorescence increase observed with rhodamines is primarily due to an increase in absorption, although the protein environment also affects quantum yield to a smaller extent. Classic O-rhodamines preferentially exist as the open form in solution and are hence generally non-fluorogenic, although minor fluorescence increases upon binding can be observed with certain dyes due to a small increase in quantum yield. In contrast, the equilibrium of Si-rhodamines is largely shifted towards the closed form in solution. SiR was the first reported fluorogenic rhodamine HTL, exhibiting an 8-fold increase in absorption upon binding to HaloTag. [23] Following this initial observation, substantial efforts have been made to understand the basis of this phenomenon. Quantification of the open-closed equilibrium and determination of structure-properties relationships enabled the design of ligands with improved fluorogenic behavior and the expansion of this approach to multicolor derivatives (Figure 4b). In particular, the introduction of electron-withdrawing substituents on the N-alkyl groups was found to shift the equilibrium towards the closed form. This was exemplified by the replacement of azetidines in JF646-HTL with 3-fluoroazetidines to lead to JF635-HTL, increasing absorption turn-on from 22 to 113-fold upon binding to HaloTag. [31,33] This strategy also led to the highly fluorogenic C-rhodamine JF585 bearing 3,3'-difluoroazetidine groups, with a 79-fold absorption turn-on. [33] These derivatives showed good bioavailability for in vivo applications. For example, JF635-HTL labelled HaloTag-expressing neurons in Drosophila brain tissue and JF585-HTL stained neurons in the brain of living mice when delivered intravenously, and could additionally be imaged using two-photon microscopy. Designing a highly fluorogenic Orhodamine with this approach is more challenging, as these are substantially more shifted towards the open form. This required combining azetidine fluorination with direct fluorination of the xanthene scaffold, providing JF526-HTL with a 16-fold absorption increase. [41] Johnsson, Wang et al. recently provided an alternative general method to design multicolor fluorogenic rhodamines, by substitution of the 3-carboxylic acid responsible for the formation of the lactone with electron-deficient sulfonamides. [40,42] The resulting multicolor dyes MaP555, MaP618 and MaP700, show 35, 1000 and 650-fold fluorescence turn-ons respectively, among the highest reported with Hal-oTag. This strategy was further expanded to the design of ratiometric dyes, incorporating an N,N-dimethylsulfonamide moiety into a hybrid coumarin-benzopyrylium dye, showing a ratiometric > 2000-fold change in fluorescence upon binding due to the modification of the electronic conjugation, switching the fluorescence from green to red (Figure 4c). [43] As these methods result in a shift of the equilibrium towards the closed form, they are ideally suited to O-, C-and Sirhodamines. However, far-red scaffolds such as sulfo-and phospho-rhodamines are already substantially shifted towards the closed form, and require alternative strategies to turn them into fluorogenic dyes. Consequently, the Lavis group incorporated fluorines on the pendant phenyl ring, which gave access to fluorogenic NIR HaloTag ligands including JF711 and JF724. [25a] These dyes however do not open completely, and show limited brightness due to their relatively low fluorescence quantum yields, which is a general limitation of NIR dyes. When applied to Si-rhodamine, this fluorination strategy led to the non-fluorogenic and bioavailable JF669 which showed excellent labelling in vivo.
Fluorogenic ligands based on molecular rotors and solvatochromic fluorophores. Fluorogenic behaviors have also been reported based on other mechanisms, with the most common approach exploiting the conformational restriction imposed by HaloTag on the bound dye, which can inhibit quenching due to TICT and rotational flexibility (Figure 4d). In contrast with the previous approach, the observed fluorescence turn-on stems here from an increase in quantum yield. Recently, this was shown for rhodamine derivatives bearing an aniline quenching group (Figure 4e). [44] Halo-SiR5 showed a 140-fold increase in fluorescence quantum yield upon binding, and was used to label neurons in cleared brain tissue. Various other molecular rotors show an emission turn-on upon binding to HaloTag, including easily accessible styrylpyridinium dyes showing up to 27-fold fluorescence increase (Figure 4f), [45] and rhodaninederived Red-Halo2 and NIR-Halo1 which show > 100-fold turnons, although their brightness is substantially lower than that of rhodamines ( Figure 4g). [46] Turn-on was also observed with benzothiadiazoles bearing a shortened HTL, due to their interaction with a tryptophan residue on the HaloTag surface, enabling wash-free imaging of HaloTagged targets in live cells ( Figure 4h). [47]

Functional fluorophore ligands
Beyond static labelling with fluorescent and fluorogenic ligands, HaloTag also offers the possibility to target functional reporters to specific subcellular locations or proteins of interest. A variety of stimuli-responsive fluorophore HTLs have been reported, including light-responsive fluorophores, which are key components for SRM, and biosensors able to dynamically report on the concentration of small molecules or ions. Generally, the HaloTag protein acts as a conventional targeting method, where the bound ligand retains the behavior of the free dye. In addition, the fluorogenic behavior of certain dyes can be transferred to functional ligands and result in synergistic behavior where HaloTag-binding activates or enhances function. Importantly, careful design of the linker between the dye and the HTL is critical to maintain responsiveness.

Photoactivatable, photoswitchable and blinking fluorophore ligands
With high brightness and photostability, synthetic dyes are ideal candidates for SRM, to visualize molecular targets below the diffraction limit. HaloTag is substantially smaller than antibody labels and is live-cell compatible, which make it an attractive labelling method for SRM in both fixed and live cells. Consequently, several classes of fluorophores tailored for SRM have been applied with HaloTag, relying either on spontaneously blinking or photo-responsive dyes.
Spontaneously blinking fluorophores. Spontaneously blinking dyes, that transiently undergo a reversible fluorescence turn-on without requiring light or exogenous chemicals, are particularly attractive for SRM in living cells. Urano et al. first reported the spontaneous blinking of a hydroxymethyl-substituted SiR (HMSiR), due to the open-closed equilibrium of the dye, analogous to the mechanism responsible for fluorogenicity ( Figure 5a). [48] HMSiR, with an on-time of 245 ms, displays suitable blinking rates for SMLM, exemplified on HaloTaglabelled tubulin in living cells. Although substantial effort has been made to computationally guide the design of novel blinking dyes, [49] one difficulty in their application is that HaloTag binding can directly affect the blinking properties, shifting the spirocyclisation equilibrium. Indeed, Urano et al. showed that the self-blinking carborhodamine HMCR550 was unsuitable for SMLM when attached to HaloTag on the traditional 6-position, because of a significant opening of the dye upon binding. However, attaching the ligand to the 5-position minimized this protein interaction and enabled SMLM of tubulin in live cells (Figure 5a). [50] Tuning the equilibrium into the blinking range can be also be achieved by substitution of the 3carboxylic acid with an ortho-trifluoroethylamide, which led to HaloTag-compatible self-blinking analogs of 500R and TMR, [42] or with substituted lactams resulting in dye ligands compatible with MINFLUX imaging. [51] Intermolecular reactions can also lead to blinking behaviors, as was observed for pyronines undergoing reversible nucleophilic addition of glutathione. The corresponding HaloTag ligands were applicable in living cells, using only endogenous glutathione. [52] An alternative approach to blinking-based SRM is PAINT, which exploits the transient binding of a fluorescently labelled DNA strand to a complementary strand on a target of interest. Expanding this technique for imaging protein targets was achieved using HaloTag labelled with an oligonucleotide ligand, [53] and this approach enabled resolution of individual HaloTagged-Nup96 protein within the nuclear pore complex. [54] Recently, the Johnsson group established a different strategy for HaloTag-based PAINT microscopy. [6b] Replacing the chloride leaving group on the conventional HaloTag ligand with a sulfonamide resulted in a non-covalent ligand, with affinity to HaloTag in the nanomolar range (Figure 5b). The reversible binding events led to blinking suitable for SRM, taking advantage of the brightness of conventional rhodamine fluorophores and showing high photostability due to fluorophore renewal from free ligands in solution. The applicability of these exchangeable HaloTag ligands was demonstrated for PAINT and MINFLUX modalities on various protein targets.
Photoactivatable fluorophores. Fluorophores that can be irreversibly turned on with light are another class of candidates for SRM, with the advantage of controllable activation. Photoactivatable fluorophores traditionally have their fluorescence suppressed by caging groups, which can be removed upon illumination, typically with large fluorescence turn-on. A classic example exploits o-nitrobenzyl photocages to lock fluorescein and rhodamines in the closed lactone form, which have been implemented as HaloTag ligands. [55] Recently, the Hell group designed highly water-soluble versions of these photocages, leading to a series of rhodamine HTLs which undergo rapid photolysis upon far UV illumination and are suitable for MINSTED imaging (Figure 5c). [56] An alternative photocaging method for rhodamines uses a diazoketone to lock the dye in the closed form, where illumination induces a rearrangement leading to the formation of open, fluorescent products. [57] Applied to the JF dyes, this approach provided photoactivatable versions of JF549 and JF646, which show high fluorescence turn-ons and are useful for single-particle tracking and SMLM (Figure 5d). [58] While the use of a small caging group is advantageous, a main limitation of the diazoketone strategy is the complex photoreaction, which additionally forms unwanted dark photoproducts.
Alternative photochemical transformations involving structurally minimal caging groups have been reported including photoactivatable pyronines displaying light-induced protonation or cyclisation. PA-SiR was serendipitously found to undergo a light-induced protonation upon illumination at 405 nm, reconstituting the conjugated pyronine core with a high fluorescence turn-on, and applied to SMLM in both fixed and live cells with HaloTag (Figure 5e). [59] However, the resulting fluorophore reacts rapidly with water or thiols, forming nonfluorescent adducts. In contrast, the xanthone-based PaX dyes undergo an intramolecular cyclisation to form the fluorescent pyronine in an efficient photoconversion reaction, which avoids dark photoproduct formation (Figure 5f). [60] A similar transformation applied to imine-xanthones led to photoactivatable dyes with large Stokes shifts and comparable performance in SRM as the PaX dyes. [61] Other specific light-triggered reactions involving non-rhodamine ligands further expand the repertoire of fluorophore photoactivation, such as the photolysis of a quenched azido-dicyanodihydrofuran to form amino fluorescent products, [62] and, more recently, photodeoximation of an oxime-caged acridine upon illumination with blue light (Figure 5g). [63]  Reversible photoswitchable fluorophores. Finally, some fluorophores present reversible photoactivation, and can therefore undergo multiple on-off cycles before photobleaching. In the presence of phosphine or thiol-containing buffers, certain standard rhodamine and cyanine derivatives, including Alexa-Fluor488 and AlexaFluor647, display such photoswitching activity, making them appropriate for SMLM in fixed samples. Intrinsically photoswitchable dyes, which do not require exogenous chemicals, are more widely applicable. For example, rhodamine lactams such as rhodamine-phthalimide undergo a transient opening under illumination, and can be used for SMLM. [64,65] This behavior was subsequently expanded to other lactams such as Rh-Gly (Figure 5h). [66] Recently, Rivera-Fuentes et al. Incorporated a hydrazone photoswitch within the lactam ring, further exploiting the open-closed equilibrium of rhodamines as a turn-on mechanism (Figure 5i). [67] Upon illumination, the hydrazone undergoes isomerization from E to Z, with the latter isomer displaying blinking behavior due to spontaneous transient opening. This approach offers controllable activation while requiring only low light intensity, which facilitates long-term imaging. Combining this scaffold with HaloTag required careful molecular engineering, as here again the interaction with the protein directly affects blinking behavior. The Si-rhodamine derivative, appended with an extended ligand at the 5-position to minimize turn-on upon binding, enabled SMLM of the endoplasmic reticulum (ER) in living cells.

Fluorescent biosensor ligands
Biosensors, which show a fluorescence turn-on upon binding to a small molecule or ion, or in response to environmental changes, are crucial tools to investigate cellular signaling networks. HaloTag enables their specific localization to desired subcellular compartments of interest. In contrast with FP-based biosensors, HaloTag-targeted synthetic biosensors benefit from the superior photophysical properties of the fluorophores, while showcasing binding motifs that have not yet been reported for protein scaffolds.
Biosensors for ions and small molecules. A variety of synthetic biosensors for ions and other biologically relevant molecules have been combined with HaloTag to visualize fluctuations in their concentration at a specific subcellular location. As for other functional ligands, the challenge lies in the appropriate choice of linkage to the HaloTag ligand, to preserve sensor activity without compromising other critical features such as cell permeability and labelling specificity. The vast majority of examples are targetable ion sensors. pH can be monitored using fluorescein and rhodol derivatives, due to the intrinsic pKa of the phenolate. SNARF-1 displays a ratiometric response to pH changes in the physiological range, and was targeted to different organelles with HaloTag (Figure 6a). [68] Careful engineering of the HaloTag linker was critical to maintain sensitivity, and only extended linkers were suitable. Another approach consists of a piperazine-substituted rhodamine, which shows a fluorescence turn-on upon protonation, due to suppression of photoinduced electron transfer (PeT) quenching, and was applied to monitor exocytosis when targeted to synaptic vesicles. [69] For metallic cations, virtually all synthetic indicators rely on PeT from an ion chelating motif as the mechanism for fluorescence modulation. Numerous selective chelating moieties have been developed and combined with synthetic fluorophores to generate biosensors for a variety of physiologically relevant ions. One of the most studied is the calcium ion, with virtually all calcium indicators exploiting the BAPTA binding motif developed by Roger Tsien. [70] The TMR-based RhoCa-Halo, was among the first targetable synthetic calcium sensor outside of the blue region of the visible spectrum, enabling multiplexed calcium imaging. [71] In this design, the HTL was attached to the BAPTA motif, resulting in a low-affinity sensor, displaying relatively low brightness. Appending the BAPTA on alternative positions of the rhodamine scaffold later enabled optimal positioning of the HTL on the pendant phenyl ring. Lavis et. al attached the BAPTA group on the azetidine ring of JF dyes, providing bright sensors detecting single action potentials in cultured neurons, which outperform spectrally matched FP-based sensors. [72] In addition, adapting this design to JF646 resulted in a modestly fluorogenic biosensor showing a 2.4-fold fluorescence turn-on upon binding, and provided the first example of subcellular calcium visualization in the far-red (Figure 6b). Recently, the Johnsson group developed a suite of targetable calcium sensors based on the MaP dyes, introducing the BAPTA on the sulfonamide (Figure 6c). [73] Among them, the far-red MaPCa-656 variants showed excellent fluorogenicity (> 100-fold turn-on) and good calcium sensitivity. Different affinity variants are well suited to visualize calcium flux in the cytosol (high affinity), or in the endoplasmic reticulum (low affinity) in cultured neurons. In a different design, Dumat et al. reported the sensor Ca-DIP, based on a BAPTA-functionalized fluorogenic molecular rotor, which also exploits fluorogenicity to gate sensor function (Figure 6d). [74] While the calcium sensitivity of this dye is similar to that of the JF and MaP-based sensors, the brightness is lower, primarily due to the fluorophore scaffold.
The engineering efforts for other cations have not been as extensive as for calcium. Nevertheless, for zinc, two fluoresceinbased targetable sensors have been reported, using the Zn 2 + binding motif dipicolylamine attached to the fluorophore scaffold (Figure 6e,f). ZP1 can be applied to detect intracellular zinc fluctuations in response to exogeneous stimulation, and was used as a ratiometric probe by fusing HaloTag to mCherry. [75] In contrast, ZIMIR is cell impermeant and has substantially lower affinity for zinc. Its HaloTag-mediated membrane targeting results in the dynamic detection of local ion concentration, exemplified by monitoring Zn 2 + / insulin release in stimulated beta cells. [76] HaloTag-targetable biosensors for other ions have also been reported for K + and Na + , using crown ether chelating motifs. [77] Another fluorogenic design is the Mg 2 + sensor Mag-SÀ Tz. This APTRA-based ratiometric sensor is functionalized with a quenching tetrazine, which reacts through inverse electron demand Diels Alder (IEDDA) with a strained alkyne HaloTag ligand (Figure 6g). [78] The biorthogonal reaction results in covalent targeting of the sensor, and abolishes quenching.
PeT-based targeted biosensors also exist for small molecules, including a fluorescein-based ligand that undergoes an irreversible cyclisation upon reaction with nitric oxide. [79] One limitation of such reactive probes is that they can undergo reaction before specific labelling of the HaloTag, which complexifies experiments. To circumvent this limitation, the Rivera-Fuentes group developed a HaloTag-gated glutathione (GSH) biosensor (Figure 6h). [80] This SiR-based compound is closed and unreactive towards GSH in solution. Upon binding to HaloTag, the shift in equilibrium towards the open form enables nucleophilic addition of GSH with concomitant fluorescence decrease. Fusion with a redox-insensitive fluorescent protein enabled a ratiometric readout and quantitative determination of GSH concentration.
Biosensors for biophysical properties. HaloTag-targeting was also applied to biosensors reporting on membrane properties, providing labelling of specific cells, or even of specific membranes of interest inside a cell. This was demonstrated with synthetic voltage indicators, which can directly report on electrical activity with submillisecond temporal resolution. The Miller group developed PeT-based voltage biosensors based on rhodamine fluorophores, with a lipophilic molecular wire attached to the pendant phenyl ring. Attachment of the HTL through the 3-carboxylic acid of the dyes resulted in targetable sensors enabling specific labelling and recording of neurons in brain slices, and Drosophila. [81] An extremely long linker to the HTL was necessary to provide sufficient flexibility for the targeted sensor to properly intercalate within the membrane, and this strategy was extended to the red-shifted SiR-based isoBeRST (Figure 6i). [82] The Flipper membrane tension probes developed by the Matile group, based on dithienothiophene push-pull systems, can also be combined with HaloTag to visualize changes in membrane properties in different organelles via changes in fluorescence lifetime. [83] HaloFlippers can target internal membranes where HaloTag is expressed as a fusion of a membrane protein, and report on the physical properties of the local membrane environment (Figure 6j). Precise engineering of the linker was crucial to retain function without impairing cellular uptake, and the targeted sensor was applied to various internal membranes including the endoplasmic reticulum, mitochondria, peroxisomes, lysosomes, and the Golgi apparatus. In subse- quent work, the same group designed Supraflippers, which allow release of the membrane probe inside the membrane of interest for untethered sensing, and indirectly exploit HaloTag for probe delivery. [84] Fluorophore ligands with biomolecule binders. Finally, fluorophore HaloTag ligands have also been attached to proteinbinding molecules, to monitor specific interactions, or to capture a tagged protein of interest. Bruchez and coworkers designed a functional ligand to visualize cell-cell interactions. [85] This method, termed FAP-DAPA, relies on the cell surface expression of a low affinity fluorogen-activating protein (FAP) on a target cell, and HaloTag on the surface of a distinct cell. The HaloTag is labelled with a tethered Malachite Green ligand, which is cell-impermeant and non-fluorescent in solution (Figure 6k). Upon interaction with a neighboring FAP-expressing cell, the Malachite Green shows far-red fluorescence turn-on, providing a readout of intercellular contact. Attachment of a protein binding group to a fluorophore ligand can also be exploited to isolate and purify the protein of interest, following its imaging inside cells. For this purpose, the Lavis group developed biotinylated rhodamine HTLs. [86] The far-red JF646 derivative efficiently labelled intracellular proteins, which could then be isolated from cell lysate by capturing with streptavidin beads.

Engineering HaloTag platforms for improved optical properties and novel functionalities
In addition to the targeting of a large diversity of ligands with various optical and functional properties, HaloTag-based reporters can also be optimized and functionalized through protein engineering. This allows further expansion of the capabilities of HaloTag, with applications including alternative imaging modalities, multiplexing and biosensing, while taking advantage of the photophysical properties of structurally minimal synthetic fluorophores. Several modifications of the HaloTag protein have been reported, inspired by the engineering strategies applied to fluorescent proteins, and resulted in a variety of mutants and topological variants (circularly permuted or split) which can further be used for insertion or fusion of sensing domains to afford biosensors (Figure 7a,b). The concurrent engineering of the protein and synthetic components to optimize their interactions can improve the photophysical properties of the resultant hybrid probe, and achieve functional behaviors that synthetic scaffolds alone cannot display.

HaloTag mutants and topological variants
Mutations in the HaloTag protein have been reported for three different purposes: (i) to modify the biochemical properties of the protein, (ii) to alter the dye-protein interactions, and (iii) to modify the ligand binding mechanism.
The current HaloTag7 is the product of several rounds of mutagenesis, which have dramatically improved properties including solubility and stability. Further improvements are however possible, and Kritzer et al. recently showed that HaloTag7 can be prone to oxidation by formation of an intramolecular disulfide bond, which reduces activity. [87] Mutations of the two cysteine residues C61S and C262S yielded HaloTag8, which can be a more robust alternative for use in oxidizing environments.
Mutations in the HaloTag sequence can also be used to adapt the tag to specific fluorophores, by altering dye-protein interactions. HaloTag presents a negatively charged surface, which leads to slow binding with negatively charged substrates. Introduction of positively charged lysine residues on the surface modifies its electrostatic potential, resulting in HOB, which shows higher binding rates with oligonucleotides and with the negatively charged dye AlexaFluor488. [8,88] As demonstrated by fluorogenicity, the protein scaffold can strongly affect the optical properties of bound fluorophores. Consequently, a minimal number of mutations in the vicinity of the fluorophore binding site can further alter fluorescence. For example, three point mutations (M175Y + V245A + L271D) led to a 4-fold higher fluorescence turn-on and 42-fold increase in binding rate with the styrylpyridinium ligand 1d (Figure 4f) compared to HaloTag. [89] Similarly, the Johnsson group engineered three novel variants of HaloTag tailored for rhodamine dyes to extend their application for FLIM. [90] Site-saturation mutagenesis led to the identification of new variants HaloTag9 (Q165H + P174R), HaloTag10 (Q165W) and HaloTag11 (M175W). Compared to the parent protein, HaloTag9 increases quantum yields and extinction coefficients of rhodamine ligands due to a change in electrostatic surface potential and an orientational shift of the bound dye. The other two variants reduce quantum yield, and hence fluorescence lifetime, due to PeT quenching from a proximate tryptophan. Together, the different variants enable multiplexed FLIM of up to three targets with a single fluorophore ligand, and up to eight targets when combined with other fluorophore labels. [91] The multiplexing capability was further exploited to design single wavelength cell-cycle Fucci biosensors, using HaloTag7 and HaloTag9 attached to human Cdt1 and Geminin respectively. [90] Advantageously, further multiplexing is possible, as demonstrated by simultaneous imaging of the HaloTag-Fucci labelled with MaP618 and a GFP-based RhoA GTPase activity biosensor.
Mutations in the HaloTag protein can also alter fluorophore binding. Replacing the reactive site D106 with an alanine provided dHaloTag7, showing high affinity to modified ligands containing primary alcohol head groups (xHTLs). [6b] This results in non-covalent binding, in a strategy orthogonal to the one shown in Fig 5b. The xHTLs can be appended to diverse rhodamines, with the resulting fluorophore ligands generally retaining similar brightness and fluorogenicity as the covalent analogs, with the advantage of higher photostability. Combining HaloTag7 and dHaloTag7 enabled multiplexing for high photon-demanding modalities such as STED and PAINT. Concurrently, Piehler, Holtmannspötter et al. used an alternative approach for the design of non-covalent HaloTags, by rescuing the catalytic behavior of the original dehalogenase with the mutation N272H. [92] The resulting reHaloTagS and reHaloTagF show longer lifetimes and higher affinities than the xHTLs and were also applied to dual color labelling and SRM. An apparent increase in photostability could be observed here also due to fluorophore renewal, although in this design the fluorophore ligand is consumed by the enzyme.
In addition to point mutations to alter photophysical or binding properties, HaloTag is amenable to topological rearrangements. HaloTag can be split at different locations, resulting in bipartite systems useful for monitoring protein-protein interactions (PPIs). Two splitHaloTag pairs have been engineered by opening the protein at position 58-59 or 155-156. [93] When the two fragments are fused to interacting proteins, reconstitution of the functional HaloTag results in irreversible capture of a fluorescent ligand, hence enabling the visualization of PPIs or the localization of protein complexes at high resolution. For example, the interaction of molybdenum insertase Cnx1 with F-actin could be imaged in plant cells using splitHaloTag. [94] This approach was also applied to image mRNAs with a single aptamer copy, avoiding perturbation of mRNA dynamics. [93b] However, the aforementioned split Hal-oTags cannot be universally applied to endogenous targets because of the large size of both fragments. Developed from the original splitHaloTag, tag-assisted split HaloTag (TA-splitHalo) consists of N-and C-terminal splitHaloTag fragments connected to two orthogonal peptide tags. Advantageously, this tool enables fluorescence labelling via tagging only with short peptides, minimizing perturbation. [95] Finally, the HaloTag can also be circularly permuted (cp), to introduce new termini near the fluorophore binding site. [96] These cpHaloTags are particularly useful for the design of biosensors which will be further discussed in the next section.

HaloTag-based biosensors
The compatibility of HaloTag with insertions, fusions and circular permutations enables the design of HaloTag-based biosensors, taking advantage of the diversity of existing protein sensing domains. In contrast with the targeted biosensors described in section 3.2, these sensors exploit genetically encoded sensing motifs, which simplifies the synthetic compound to be delivered, and are highly tunable simply by varying the fluorophore ligand. The design of these biosensors was directly inspired by the design of FP-based genetically encoded indicators, which function either via FRET, or by exploiting the environmental sensitivity of the chromophore.
Biosensors based on resonance energy transfer (RET) partners. The spectral flexibility afforded by the fluorophore selection makes HaloTag an ideal FRET partner for fluorescent proteins and other chromophores. Hiblot and coworkers systematically investigated the properties of dye-HaloTag conjugates as FRET acceptors for fluorescent proteins. [97] Fusing HaloTag to the Cterminus of EGFP and mutagenesis of the interface led to near quantitative FRET with a variety of rhodamine ligands including red-shifted dyes such as SiR and JF669, despite their minimal spectral overlap with GFP. The spectral tunability of the pair could be further expanded by replacing EGFP with fluorescent proteins ranging from cyan to red. The authors engineered a series of FRET biosensors using the EGFP-HaloTag pair, mimicking the design of ratiometric FP-based biosensors, by introducing a sensing motif between the two reporters ( Figure 8a). Upon binding to the small molecule or ion target, the sensor undergoes conformational change bringing the reporters in close proximity and facilitating FRET. This design led to spectrally tunable biosensors for calcium, ATP and NAD + which show high ratio changes (e. g. ΔR/R 0 up to 37 for Ca 2 + ) and are applicable in living cells.
One of the first examples of a chemigenetic FRET biosensor was Voltron, a voltage indicator based on the microbial rhodopsin voltage sensing domain Ace2 fused to HaloTag, and labelled with JF rhodamine dyes (Figure 8b). [98] Upon membrane depolarization, the absorption of the opsin retinal cofactor increases, leading to more efficient FRET with the dye HaloTag ligand, and resulting in a fluorescence decrease. Due to the brightness and photostability of the dye ligand, Voltron shows 10-fold larger photon yield than existing FP-based voltage sensors, enabling long-term recording of neuronal signals in living mice, zebrafish and Drosophila. A positive-going version, Positron, was subsequently developed via targeted point mutations. [99] An alternative design for FRET-based fluorescent biosensors is the SNIFIT approach (Figure 8c). [100] This design can lead to efficient biosensors for which protein sensing motifs do not undergo large conformational changes, and involves a synthetic compound which is tethered to the sensor and competes with the target analyte. While originally engineered using SNAP-Tag and a FP as the FRET pair, this approach was later expanded to HaloTag. NADP-SNIFIT reports on free NADPH/NADP + ratios by combining an NADP-binding protein (SPR), with SNAP-Tag and HaloTag self-labelling proteins. These are labelled with a sulfamethoxazole (SMX) functionalized TMR-SNAP ligand, and SiR-HTL respectively. [101] The sulfamethoxazole ligand binds SPR in an NADP + -dependent manner, bringing the two fluorophores in close proximity and resulting in high FRET efficiency. Upon binding to NADPH, the SMX ligand is displaced, which separates the two fluorophores and abolishes FRET, leading to a ratiometric fluorescence response. The SNIFIT design is generalizable, as shown with the recent development of a Coenzyme A (CoA) biosensor based on this strategy, the first reported quantitative sensor for this cofactor. [102] Protein engineering provided variants with different affinities for CoA, and their use in FLIM allows absolute measurements of CoA concentration in different subcellular compartments.
Interestingly, HaloTag can also be converted into a bioluminescent tag, which can in turn be used to convert fluorescent biosensors into bioluminescent reporters. Fusion of the luciferase NanoLuc to the C-terminus of HaloTag led to good bioluminescence resonance energy transfer (BRET) efficiency with JF525. [103] As BRET is distance-dependent, more

ChemBioChem
Review doi.org/10.1002/cbic.202300022 efficient transfer could be achieved by inserting a circularly permuted luciferase into the HaloTag near the fluorophore binding site, resulting in H-Luc. [104] The emission of NanoLuc can be shifted from 460 nm for its native substrate up to 680 nm via BRET with H-Luc-bound rhodamine fluorophores. This HaloTag-NanoLuc chimera can be incorporated into bioluminescent sensors named LUCIDs, in a similar design as the SNIFITS. A LUCID biosensor for the drug methotrexate (MTX) was built from H-Luc, which was fused to a circularly permuted dihydrofolate reductase (cpDHFR) as the receptor, and to SNAP-Tag as the second fluorescent partner (Figure 8d). [104] SNAP-Tag was labelled with a Cy5 ligand functionalized with trimethoprim. In the off state, trimethoprim binds to cpDHFR, bringing the Cy5 close to H-Luc labelled with CRh-HTL, facilitating BRET-FRET. Upon binding to the receptor, MTX displaces trimethoprim, preventing FRET, and generating a maximum ratio change of 280 %. Bioluminescent HaloTags have also been used for the design of voltage indicators [105] or to target synthetic BAPTAbased calcium indicators transforming them into bioluminescent biosensors. [73] Biosensors based on environment-sensitive fluorophores. When attached to HaloTag, fluorogenic fluorophores can further act as environment sensitive probes. Fluorogenic HTLs were used to monitor proteostasis. A single point mutation (K73T) in HaloTag resulted in the destabilized protein AgHalo, highly prone to aggregation in response to stress conditions (Figure 8e). [106] Multicolor solvatochromic dyes and molecular rotors with an extended linker were developed to remain nonfluorescent upon binding to HaloTag, and thus to the properly folded AgHalo. [106][107] Upon aggregation, the change in protein conformation results in up to a 57-fold fluorescence turn-on in vitro, leveraging these fluorogenic ligands as reporters for proteostatic changes in living cells. The AgHalo strategy was also employed for monitoring ER-specific proteostasis. [108] While the metastable mutant reports on the global stress response, this approach can additionally be used with the native HaloTag protein fused to proteins of interest. [109] Protein sensing domains that undergo a large conformational change can be fused to HaloTag, to provide intensiometric biosensors. The design of these platforms was directly inspired by FP-based single wavelength sensors such as the Ca 2 + indicators GCaMP, which exploit the environment sensitivity of the GFP chromophore as the mechanism for fluorescence modulation. HaloTag was circularly permuted at position 143, creating new termini near the fluorophore binding site, which were fused to calmodulin (CaM) and a calmodulin binding peptide. This resulted in the HaloCaMP protein, which, when used with fluorogenic ligands such as JF635, shows a farred fluorescence increase in response to calcium due to a shift in the open-closed equilibrium (Figure 8f). [96] In cultured neurons, HaloCaMP shows superior brightness, kinetics, and sensitivity to the FP-based sensor NIR-GECO1. This platform is highly tunable, and the wavelength, brightness and sensitivity can be adapted to the desired application by only varying the fluorogenic ligand. A similar design based on the insertion of the calcium binding domain directly into HaloTag led to a ratiometric sensor when used with a coumarin-benzopyrilium ligand (Figure 4c). [43] This general strategy can also be adapted to other sensing moieties, as exemplified with the voltage sensors HASAP and HArcLight which exploit voltage-sensitive domains, and are able to report on single action potentials in cultured neurons (Figure 8g). [96] In these designs, the sensing component is genetically encoded and only the small-molecule fluorophore ligand is exogenous. While this presents the advantage of having to deliver a simpler chemical, synthetic binding motifs can provide properties that have not yet been replicated by protein domains. The Campbell group designed ion sensors based on a cpGFP-HaloTag protein labelled with ion-binding ligands (Figure 8h). [110] Using a BAPTA-HTL led to a green calcium sensor, in which ion binding shifts the protonation equilibrium of the GFP chromophore towards the fluorescent phenolate form. Generalizing this strategy, a sodium indicator was developed using HTL derivatives of a Na + -chelating crown ether. This alternative approach further highlights the modularity of the HaloTag, and the breadth of possible designs exploiting it as a building block in advanced reporters.

Conclusion and perspectives
Despite being a relatively recent technology, the HaloTag has rapidly become a staple labelling method for fluorescence microscopy, overcoming limitations of fluorescent proteins. Synthetic chemistry has provided a palette of bright and photostable ligands which can be fine-tuned for the desired applications, and display high fluorogenic behavior. General strategies for functional synthetic fluorophores have also been extended to HaloTag, resulting in targeted stimuli-responsive fluorophores to investigate complex processes at subcellular locations. Finally, the possibility to engineer the protein scaffold offers an additional dimension to the design of reporters, expanding the strategies to develop and optimize fluorescent molecular tools. Together, these different strategies have led to robust imaging reporters which have found applications in fluorescence imaging modalities covering widefield to superresolution, multi-photon and lifetime microscopies, and have been used in biological systems ranging from isolated cells to complex organisms.
Looking forward, we expect that the portfolio of HaloTagbased reporters will keep growing, and become even more prevalent in the imaging field. Among foreseeable advances, the continued improvements of functional platforms based on HaloTag will be critical to enable their use in complex biological environments, particularly in vivo, and their generalization to other biochemical targets. In addition, despite the large library of dyes already available, novel fluorophore HTLs are still needed. Indeed, bright and highly fluorogenic NIR ligands are still lacking, as well as fluorescent ligands also compatible with other modalities for correlative imaging (e. g. CLEM, photoacoustic imaging). Ultimately, the broad adoption of HaloTagbased reporters by biologists will also depend on the availability of the synthetic ligands, and new initiatives aiming at rapid dissemination of novel dyes will be key to achieve this goal. [111] From a broader perspective, the development of other self-labelling tags is still a requirement to further expand the capabilities of chemigenetic imaging tools. In particular, the fairly large size of the HaloTag protein can be a concern for certain protein targets, with the risk of perturbating protein localization and function. An important future endeavor will be the development of self-labelling tags combining the fast covalent labelling, dye ligand diversity, and fluorogenic behavior of HaloTag in a smaller protein scaffold. We can expect that this will be realized through improvements of smaller tag proteins (e. g. SNAP-Tag), or through the identification and optimization of novel tags either from existing enzymes or by de novo design. Their implementation into functional reporters, similarly to HaloTag, will enable these tools to be multiplexed, giving access to dynamic information on multiple targets simultaneously. Together, the development of orthogonal tagligand pairs, the continued engineering of functional platforms, and their extension to other imaging modalities will provide robust multiplexable reporters, greatly facilitating biological imaging across scales.