ABSTRACT
Expanding the palette of fluorescent dyes is vital for pushing the frontier of biological imaging. Although rhodamine dyes remain the premier type of small-molecule fluorophore due to their bioavailability and brightness, variants excited with far-red or near-infrared light suffer from poor performance due to their propensity to adopt a lipophilic, nonfluorescent form. We report a general chemical modification for rhodamines that optimizes long-wavelength variants and enables facile functionalization with different chemical groups.
The development of hybrid small-molecule:protein labeling strategies enable the use of chemical fluorophores in living cells and in vivo1. Optimizing small-molecule dyes for these complex biological environments is important, as synthetic fluorophores are often brighter and more photostable than fluorescent proteins2. We recently developed general methods to improve3 and fine-tune4 rhodamine fluorophores by incorporating four-membered azetidines into the structure, yielding the ‘Janelia Fluor’ dyes. Although our existing tuning strategies allow optimization of short-wavelength rhodamines, we discovered these methods cannot be applied to analogs excited with far-red and near-infrared (NIR) light because such dyes strongly favor formation of a colorless species. Here, we report a new complementary tuning strategy that allows rational optimization of a broader palette of fluorophores. This general method also serves as a basis for facile functionalization, enabling the synthesis of novel cell- and tissue-permeable rhodamine labels for biological imaging experiments.
The bioavailability and behavior of rhodamine dyes is dictated by a key property—the equilibrium between the nonfluorescent lactone and the fluorescent zwitterion (Fig. 1a). Our previous work on rhodamine dyes3–6 allows definition of a general rubric that directly correlates the lactone–zwitterion equilibrium constant (KL–Z) of the free dye to the cellular performance of HaloTag7 ligand derivatives (Fig. 1b). Dyes with high KL–Z exist almost exclusively in the zwitterionic form, making them useful as environmentally insensitive biomolecule labels8. Rhodamines with intermediate KL–Z values exhibit improved cell and tissue permeability due to the modestly higher propensity of the molecule to adopt the lipophilic lactone form and rapidly traverse biological membranes4,6. Dyes exhibiting even smaller KL–Z values (KL–Z = 10−2–10−3) preferentially adopt the closed lactone form, which can be exploited to create ‘fluorogenic’ ligands6,9,10 as binding to the HaloTag protein shifts the equilibrium to the fluorescent form. This property also decreases in vivo utility, however, due to problems with solubility and sequestration in membranes. Finally, dyes with extremely small KL–Z values (< 10−3) exist completely in the nonfluorescent lactone form, rendering them effectively unusable in biological experiments.
We compared a series of Janelia Fluor rhodamine analogs with different fluorophoric systems (1–8, Fig. 1c). Compounds 2, 4, and 5 were described previously and include the azetidine-containing rhodamine (2), which we termed ‘Janelia Fluor 549’ (JF549), and the carbo- and Si-rhodamine analogs 4 and 5 (JF608 and JF646, respectively; Fig. 1c)3. We expanded the wavelength range of the JF dyes by synthesizing new analogs based on known rhodamine structures containing nitrogen11,12 (NCH3; 1), sulfur13,14 (S; 3 and SO2; 8), and phosphorous15,16 (PO2H; 6 and P(O)Ph; 7) starting from bis-arylbromides5 (Fig. 1c, Supplementary Note). We evaluated the spectral and chemical properties of these dyes: absorption maximum (λmax), extinction coefficient at λmax (ε), fluorescence emission maximum λem, fluorescence quantum yield (Φ), and KL–Z (Table 1). Comparing KL–Z and λmax uncovered an inverse correlation (Fig. 1d), with the short wavelength NCH3-containing JF502 exhibiting a high KL–Z = 5.5 and the near-infrared (NIR) dyes containing SO2 and P(O)Ph showing a low KL–Z ≈ 10−4.
Due to the inverse correlation of KL–Z and λmax, short-wavelength dyes would typically require tuning KL–Z lower to improve tissue permeability and create fluorogenic HaloTag ligands (Fig. 1d) 4,6,10. In previous work4, we found incorporation of 3,3-difluoroazetidines into the Janelia Fluor structure decreases KL–Z and elicits a hypsochromic shift of ~25 nm. This strategy transformed JF549 (2; KL–Z = 3.5) into JF525 (9; KL–Z = 0.68, Fig. 1e,f, Table 1); the JF525–HaloTag ligand shows improved in vivo bioavailability and is an excellent component of the hybrid voltage indicator Voltron2. This modification also transformed JF608 (4; KL–Z = 0.091) into the highly fluorogenic JF585 (10; KL–Z = 0.001, Fig. 1e,f, Table 1). Given the relatively high KL–Z = 4.33 observed for JF502 (1) we applied this same tuning strategy to yield the fluorinated JF479 (11) that showed the expected decrease in λmax and KL–Z = 2.88 (Fig. 1e,f, Table 1, Supplementary Fig. 1a–c). The JF479–HaloTag ligand (12, Fig. 1g) exhibits similar spectral properties to enhanced green fluorescent protein (GFP) when attached to the HaloTag protein (Fig. 1h) allowing imaging with low bleed-through compared to the similarly cell-permeable—albeit brighter—JF503 ligand (13) when multiplexed with JF525-SNAP-tag ligand (14; Fig. 1i, Supplementary Fig. 1d–i)4,7.
For the far-red and NIR rhodamines (5–8, Fig. 1c) the KL–Z vs. λmax relationship reveals the need for alternative tuning strategy to increase KL–Z (Fig. 1d). This would improve the in vivo performance of Si-rhodamine dyes such as JF646 and rescue the NIR P(O)Ph- and SO2-containing dyes (7–8). We previously showed that halogenation of the pendant phenyl ring system can substantially increase the KL–Z of Si-rhodamine dyes5, presumably by lowering the pKa of the benzoic acid moiety; this substitution also elicits a bathochromic shift17. For example, JF646 (5; λmax/λem = 646 nm/664 nm) exhibits a KL–Z = 0.0014 but the fluorinated analog, JF669 (15; λmax/λem = 669 nm/682 nm), is higher with KL–Z = 0.262 (Fig. 1j,k, Supplementary Fig. 1j,k). This manifests in a higher absorptivity in aqueous solution with 5 exhibiting ε = 5,600 M−1cm−1 but the fluorinated analog 15 showing ε= 112,000 M−1cm−1 (Fig. 1l, Table 1). We surmised this strategy would be general and prepared the fluorinated PO2H-, P(O)Ph-, and SO2-containing rhodamines (16–18, Fig. 1j) by replacing phthalic anhydride with tetrafluorophthalic anhydride in our synthetic scheme (Supplementary Fig. 1j, Supplementary Note). This modification increased KL–Z and ε and elicited a ~23 nm shift in λmax (Fig. 1j–l, Table 1, Supplementary Fig. l–n). In particular, the fluorinated P(O)Ph-derivative 17 strongly absorbs visible light in aqueous solution (ε= 87,000 M−1cm−1; λmax = 722 nm) compared to the parent compound 7 (ε < 200 M−1cm−1; Fig. 1l). This trend was generalizable to oxygen- and sulfur-containing rhodamines based on 2 and 3 where the fluorine substitution on the pendant phenyl ring also increased KL–Z and λmax (19–20, Fig. 1m,n, Table 1, Supplementary Fig. 1o,p).
We then explored conjugatable versions of these new far-red and NIR dyes. In addition to increasing λmax and KL–Z, the halogenated phenyl ring motif can also serve as an electrophile for attack by thiols through a nucleophilic aromatic substitution reaction (SNAr), an established strategy for preparing xanthene dye derivatives8,18. The reactivity of other nucleophiles was unknown, however; we discovered that N3−, CN−, NH3, and NH2OH could react with JF669 (15) to provide derivatives 21–24(Fig. 2a). This reaction type was generalizable to other fluorinated rhodamines and regioselective at the 6-position (Supplementary Note). Although beyond the scope of this report, we briefly investigated some of these derivatives, finding azide 21 was an excellent reactant in strain-promoted ‘click chemistry’ with cyclic alkynes19 25 and 26 to form triazole adducts 27 and 28 (Supplementary Fig. 2a), validating the regiochemistry of the amine addition to form 22 using intermediates 21 and 29 (Supplementary Fig. 2b), and testing the reactivity of amine-containing ion-chelating groups 30 and 31 which generated novel prototype far-red indicators for K+ and Zn2+ (32–33, Supplementary Fig. 2c–e).
We then sought to create carboxy derivatives compatible with different labeling strategies such as the HaloTag. We explored the reactivity of malonates and related carbon nucleophiles, all of which showed reaction with fluorinated rhodamines with regioselectivity at the 6-position (Supplementary Note). In particular, the addition of masked acyl cyanide20 reagent 34, an umpolung-type acyl anion equivalent, to JF669 resulted in intermediate 35, which could be deprotected to yield a reactive acyl cyanide suitable for conjugation with the HaloTag ligand amine (36) to form JF669–HaloTag ligand 37 (Fig. 2a). This late-stage, regioselective introduction of the carboxy handle has distinct advantages over previous rhodamine syntheses that generate isomeric mixtures21. Compound 37 was an excellent label for cell biological experiments (Fig. 2b,c) and was also blood–brain-barrier permeable, labeling HaloTag-expressing neurons throughout the mouse brain after intravenous injection (Fig. 2d, Supplementary Fig. 3a). This chemistry was generalizable across fluorinated rhodamines, allowing facile synthesis of HaloTag ligands 37–42 from dyes 15–20 with λmax ranging from the green to NIR (Supplementary Fig. 3b,c, Supplementary Note). These compounds selectively labeled HaloTag fusions in cells, demonstrating that fluorination on the pendant phenyl ring does not prohibit HaloTag labeling (Fig. 2e).
Finally, we sought to create a bright, fluorogenic NIR HaloTag ligand suitable for biological imaging. The SO2-containing rhodamine JF724 (18) possessed a promising KL–Z = 10−3 for creating fluorogenic compounds; the JF724–HaloTag ligand (42) showed a 15-fold increase upon reaction with HaloTag protein in vitro (Supplementary Fig. 3d). Nevertheless, this dye was plagued with a low Φ = 0.05 (Table 1), making it suboptimal for imaging experiments. In contrast, the P(O)Ph-containing fluorophore JF722 (17) exhibits a larger Φ = 0.11 but also a relatively high KL– Z = 0.026; the JF722–HaloTag ligand (41) was not fluorogenic (Supplementary Fig. 3e). We investigated whether our two tuning strategies could work synergistically, using JF571 (19, Fig. 1m) as a proof-of-concept. We introduced a fluorine substituent on each azetidine ring to create JF559 (43; Supplementary Fig. 3f) and found this dye exhibits KL–Z = 6.22, intermediate between JF549 (2; KL–Z = 3.5) and JF571 (19; KL–Z = 7.93; Table 1, Supplementary Fig. 3g–i), demonstrating the compatibility of these strategies. The JF559–HaloTag ligand (44) could be used in live cell labeling (Supplementary Fig. 3j,k). We then applied this modification to JF722 by synthesizing derivatives with fluorine substituents on the azetidine ring, yielding JF711 (45, Fig. 2f, Table 1, Supplementary Fig. 3l). Compound 45 exhibited a further improvement in Φ = 0.17 (Table 1)and was predicted to yield fluorogenic ligands based on its KL–Z = 10−3 (Fig. 2g). The JF711–HaloTag ligand 46 (Fig. 2h) showed a 5-fold increase upon binding HaloTag (Fig. 2i) with excellent performance in live-cell imaging experiments (Fig. 2j).
In summary, we expanded the palette of Janelia Fluor dyes by replacing the central oxygen in JF549 (2) with nitrogen (NCH3, 1), sulfur (S and SO2, 3,8), and phosphorous (PO2H and P(O)Ph, 6–7; Fig. 1c). As the nitrogen-containing dye JF502 (1) exhibited relatively high KL–Z and long λmax, we applied our established tuning strategy4 to transform this dye into the GFP-like JF479 (11; Fig. 1e–i). The NIR-excited dyes 7 and 8 exhibited the opposite problem with low KL–Z values that rendered the compounds unusable in biological environments (Fig. 1d). We therefore established a complementary general method to increase both λmax and KL–Z by incorporating fluorines on the pendant phenyl ring of rhodamine dyes (Fig. 1j,k) followed by facile, generalizable SNAr chemistry to install groups for bioconjugation (Fig. 2a). This strategy yielded the bioavailable JF669–HaloTag ligand (37, Fig. 2b–e) along with other new fluorophores (38–42, Fig. 2e) and could be combined with our previous tuning method to create the fluorogenic NIR-excited JF711 HaloTag ligand (46, Fig 2f–j). Although we have focused here on HaloTag ligands and mammalian cells, we expect this general rubric relating KL–Z to cellular performance (Fig. 1b) to be applicable to other ligands and biological systems6. We also anticipate this expanded fluorophore palette to enable the rational design of finely tuned labels for biological imaging experiments in cells or animals and the new derivatization chemistry to facilitate the synthesis of novel ligands, labels, stains, and indicators, particularly those excited by far-red or NIR light.
AUTHOR CONTRIBUTIONS
L.D.L. and J.B.G. conceived the project. J.B.G. contributed organic synthesis and 1-photon spectroscopy measurements. A.N.T. and H.C. contributed cultured cell imaging experiments. B.M. contributed in vivo labeling and tissue imaging experiments. N.F. contributed organic synthesis. R.P. contributed 2-photon spectroscopy measurements. J.L-S and T.A.B. directed the project. L.D.L. directed the project and wrote the paper with input from the other authors.
COMPETING FINANCIAL INTERESTS STATEMENT
The authors declare competing interests: J.B.G. and L.D.L. have filed patent applications whose value may be affected by this publication.
ONLINE METHODS
Chemical synthesis
Methods for chemical synthesis, full characterization of all novel compounds, and crystallographic confirmation of regioselective SNAr can be found in the Supplementary Note.
UV–vis and fluorescence spectroscopy
Fluorescent and fluorogenic molecules for spectroscopy were prepared as stock solutions in DMSO and diluted such that the DMSO concentration did not exceed 1% v/v. Spectroscopy was performed using 1-cm path length, 3.5-mL quartz cuvettes or 1-cm path length, 1.0-mL quartz microcuvettes from Starna Cells. All measurements were taken at ambient temperature (22 ± 2 °C). Absorption spectra were recorded on a Cary Model 100 spectrometer (Agilent). Fluorescence spectra were recorded on a Cary Eclipse fluorometer (Varian). Maximum absorption wavelength (λabs), extinction coefficient (ε), and maximum emission wavelength (λem) were taken in 10 mM HEPES, pH 7.3 buffer unless otherwise noted; reported values for ε are averages (n = 3). Normalized spectra are shown for clarity. For prototype ion indicators 32 and 33 (Supplementary Fig. 2c) the compounds were dissolved in 10 mM HEPES, pH 7.3 buffer alone or with either 100 mM KCl or 10 μM ZnCl2; the fluorescence emission spectra of these solutions were recorded using λex = 575 nm and λem = 625–825 nm.
Determination KL–Z and εmax
We calculated KL–Z using the following equation22: KL–Z = (εdw/εmax)/(1 – εdw/εmax). εdw is the extinction coefficient of the dyes in a 1:1 v/v dioxane:water solvent mixture; this dioxane–water mixture was chosen to give the maximum spread of KL–Z values4. εmax refers to the maximal extinction coefficients measured in different solvent mixtures empirically determined depending on dye type: 0.1% v/v TFA in ethanol for the Si-rhodamines (5 and 15); 0.1% v/v trifluoroacetic acid (TFA) in 2,2,2-trifluoroethanol (TFE) for all the other rhodamines. We note that accurate determination of low KL–Z values (≤ 10−3) is complicated by the relatively poor sensitivity of absorbance measurements. We estimated KL–Z = 10−3 when we observed a small but significant absorbance signal in 1:1 v/v dioxane:water solvent mixture over the dye-free control, and KL–Z ≈10−4 when we observed no significant absorbance of the dye solution.
Quantum yield determination
All reported absolute fluorescence quantum yield values (Φ) were measured in our laboratory under identical conditions using a Quantaurus-QY spectrometer (model C11374, Hamamatsu). This instrument uses an integrating sphere to determine photons absorbed and emitted by a sample. Measurements were carried out using dilute samples (A < 0.1) and selfabsorption corrections23 were performed using the instrument software. Reported values are averages (n = 3).
1-Photon spectroscopy of HaloTag conjugates
HaloTag protein was used as a 100 μM solution in 75 mM NaCl, 50 mM TRIS·HCl, pH 7.4 with 50% v/v glycerol (TBS–glycerol). Absorbance measurements were performed in 1-mL quartz cuvettes. HaloTag ligands 12, 41–42, 46(5 μM) were dissolved in 10 mM HEPES, pH 7.3 containing 0.1 mg·mL−1 CHAPS. An aliquot of HaloTag protein (1.5 equiv) was added and the resulting mixture was incubated until consistent absorbance signal was observed (60–120 min). To measure the fold-increase of absorbance upon HaloTag binding (41, 42, 46) an equivalent volume of TBS–glycerol blank was added in place of enzyme to record the ‘-HT’ absorbance. Absorbance scans are averages (n = 2).
Multiphoton spectroscopy
For the environmentally insensitive compounds 1, 11, and a fluorescein control (Supplementary Fig. 1c) we prepared 5 μM solutions of the free dyes in 10 mM HEPES buffer, pH 7.3. For other rhodamines (Supplementary Fig. 3c), we measured spectra of the HaloTag conjugates: compounds 37–42, 44, and 46(5 μM) were incubated with excess purified HaloTag protein (1.5 equiv) in 10 mM HEPES, pH 7.3 containing 0.1 mg·mL−1 CHAPS as above for 24 h at 4 °C. These solutions were then diluted to 1 μM in 10 mM HEPES buffer, pH 7.3 and the two-photon excitation spectra were measured as previously described48,49. Briefly, measurements were taken on an inverted microscope (IX81, Olympus) equipped with a 60×, 1.2NA water objective (Olympus). Dye–protein samples were excited with pulses from an 80 MHz Ti-Sapphire laser (Chameleon Ultra II, Coherent) for 710-1080 nm and with an OPO (Chameleon Compact OPO, Coherent) for 1000-1300 nm. Fluorescence collected by the objective was passed through a dichroic filter (675DCSXR, Omega) and a short pass filter (720SP, Semrock) and detected by a fiber-coupled Avalanche Photodiode (SPCM_AQRH-14, Perkin Elmer). For reference, a two-photon excitation spectrum was also obtained for the red fluorescent protein mCherry (1 μM), in the same HEPES buffer. All excitation spectra are corrected for the wavelength-dependent transmission of the dichroic and band-pass filters, and quantum efficiency of the detector.
General cell culture and fluorescence microscopy
U2OS cells (ATCC) were cultured in Dulbecco’s modified Eagle medium (DMEM, phenol red-free; Life Technologies) supplemented with 10% (v/v) fetal bovine serum (Life Technologies), 1 mM GlutaMAX (Life Technologies) and maintained at 37 °C in a humidified 5% (v/v) CO2 environment. For confocal imaging of cell nuclei (Fig. 1i, Fig. 2e, Supplementary Fig. 1e,g, Supplementary Fig. 3k), we used U2OS cells with an integrated a histone H2B–HaloTag expressing plasmid via the piggyback transposase. For confocal imaging of the cell surface (Supplementary Fig. 1f,h), we used U2OS cells transiently transfected with a plasmid expressing a C-terminal transmembrane anchoring domain from platelet-derived growth factor receptor (PDGFR) fused to the HaloTag protein (PDGFR– HaloTag); nucleofection (Lonza) was used for transfection. For the Airyscan imaging experiments (Fig. 2b,c) we used U2OS cells transiently transfected with Sec61β–HaloTag expressing plasmid or TOMM20–HaloTag expressing plasmid using FuGENE HD (Promega). Sec61β is an endoplasmic reticulum membrane protein translocator protein and TOMM20 is an outer mitochondrial membrane protein as part of a protein translocase complex. For confocal imaging of mitochondria (Fig. 2j), we used U2OS cells with an integrated a TOMM20–HaloTag expressing plasmid via the piggyback transposase. These cell lines were kept under the selection of 500 μg/mL Geneticin (Life Technologies). Cell lines undergo regular mycoplasma testing by the Janelia Cell Culture Facility. Unless otherwise noted, cells were imaged on a Leica SP8 Falcon confocal microscope with an HC PL-APO 86×/1.20 water objective, a Zeiss LSM 800 confocal microscope with a Plan APO 20×/0.8 air M27 objective or Plan APO 63×/1.4 oil DIC M27 objective, a Zeiss LSM 880 with a C-APO 40×/1.2 W Corr FCS M27, or a Zeiss LSM 880 with Airyscan and a plan- apochromatic 63× oil objective (NA=1.4). The airyscan images were processed using the Zen software from Zeiss, the Leica and Zeiss LSM 800 confocal images were processed using FIJI.
Dye loading kinetics
For the dye loading comparison (Supplementary Fig. 1i), U2OS cells stably expressing histone H2B–HaloTag were stained over a time course of 0–4 h with 200 nM of either JF479–HaloTag ligand 12 or JF503–HaloTag ligand 13. Cells were washed 2× with dye-free media and imaged live using widefield microscopy (Nikon Eclipse Ti, Plan APO l 20×/0.75; 470nm Ex/FITC (515/30) Em. Fluorescence was quantified from the average of the summed intensity of nuclear signals in single plane widefield analyzed using Nikon NIS-Elements AR software (Supplementary Fig. 1e,g).
Multiplexed imaging comparison JF503 and JF479
U2OS cells stably expressing histone H2B– HaloTag fusion protein were transiently transfected with TOMM20–pSNAPf plasmids using nucleofection (Lonza). Live cells were simultaneously incubated with JF479–HaloTag ligand (12, 500 nM) or JF503–HaloTag ligand (13, 500 nM) for 3 h, adding JF525–SNAP-tag ligand (14, 1 μM, Supplementary Fig. 2d) for 60 min. These cells were then washed and imaged (Fig. 1i) using tunable white light laser excitation at 488 nm and 532 nm on a Leica SP8 Falcon confocal microscope with an HC PL-APO 86×/1.20 water objective. The images are displayed as maximum intensity projections of confocal image stacks using FIJI.
Mouse in vivo labeling experiments
Adult C57/BL6 male mice were used to express a GFP HaloTag fusion protein throughout the brain by systemic injection using the viral vector: PHP-eB- Syn-HaloTag-GFP (~5 × 1011 infectious units per ml, 100 μl). The virus was injected using a 0.5ml 27G syringe to the retro-orbital sinus. JF669–HaloTag ligand (37) was administered to mice 3–4 weeks after the viral injection. Dye solution was prepared by first dissolving 100 nmol (76 μg) of 37 in 20 μL DMSO. After vortexing, 20 μL of a Pluronic F-127 solution (20% w/w in DMSO) was added and this stock solution was diluted into 200 μL sterile saline for IV (retro-orbital) injection. All experimental protocols were conducted according to the National Institutes of Health guidelines for animal research and were approved by the Institutional Animal Care and Use Committee at the Janelia Research Campus, HHMI.
Statistics
For spectroscopy measurements (Fig. 1l, Table 1, Supplementary Fig. 1a,b,k–p, and Supplementary Fig. 1i,l) reported n values for absorption spectra, extinction coefficient (ε) and quantum yield (Φ) represent measurements of different samples prepared from the same dye DMSO stock solution. For the cell loading experiment (Supplementary Fig. 1i) the following reported n values represent the number of intensity values measured from three fields of view for the time points at 30 s, 1 min, 2 min, 3 min, and 4 minutes, respectively: JF479–HaloTag ligand 12: n= 112, 120, 130, 114, 128; JF503– HaloTag ligand 13: n = 135, 129, 135, 158, 161.
Data availability
The data that support the findings of this study are provided in the Source Data files or available from the corresponding author upon request.
ACKNOWLEDGEMENTS
We thank: S. Sternson (Janelia) for initial discussions on umpolung reagents; K. Svoboda (Janelia) for advice on in vivo labeling experiments; C. Deo and E. Schreiter (Janelia) for purified HaloTag protein, contributive discussions, and a critical reading of the manuscript; the Janelia Cell and Tissue Culture, Anatomy and Histology, and Vivarium teams for assistance with biological experiments. This work was supported by the Howard Hughes Medical Institute.