Extremely Bright, Near-IR Emitting Spontaneously Blinking Fluorophores Enable Ratiometric Multicolor Nanoscopy in Live Cells.

New bright, photostable, emission-orthogonal fluorophores that blink without toxic additives are needed to enable multicolor, live-cell, single-molecule localization microscopy (SMLM). Here we report the design, synthesis, and biological evaluation of Yale676sb, a photostable, near-IR-emitting fluorophore that achieves these goals in the context of an exceptional quantum yield (0.59). When used alongside HMSiR, Yale676sb enables simultaneous, live-cell, two-color SMLM of two intracellular organelles (ER + mitochondria) with only a single laser and no chemical additives.

Introduction

Single-molecule localization microscopy (SMLM)[1−4] is a powerful technique for visualizing intracellular architecture at the nanoscale[5] and across large fields of view.[6] The technique is characterized by the detection and localization of fluorescent markers that cycle rapidly between emissive (ON) and non-emissive (OFF) states. For optimal results, the sample and imaging conditions must maintain the majority of fluorescent markers in the OFF state, such that the neighboring molecules in the emissive ON state can be treated as sparse single emitters.[7−11] Organic fluorophores are favored over fluorescent proteins for SMLM because they are generally brighter and more photostable and because their photophysical properties can be fine-tuned using chemistry.[7,12−17,17−21] The challenge is that many SMLM-compatible organic fluorophores require the addition of exogenous nucleophiles, redox modulators, and/or oxygen depletion systems to switch efficiently between ON and OFF states. These additives can be cytotoxic and damage or alter biological samples.[7,12,13,15,17] An additional challenge is that many established SMLM-compatible fluorophores are cell-impermeant[7,12,13,20,22,23] and/or require cytotoxic high-power and/or short-wavelength lasers.[7,12,16,18,22,24,25]

The spontaneously blinking fluorophore (SBF) hydroxymethyl Si-rhodamine (HMSiR)[15] (Figure a) reported by Urano and co-workers overcomes many of these limitations. It is cell-permeant and photostable and is believed to cycle rapidly between ON and OFF states by virtue of a pH-dependent spirocyclization reaction that occurs in the absence of chemical additives[15] (Figure a). For HMSiR, the midpoint of this pH-dependent equilibrium (referred to as pKcycl) occurs at approximately pH 6.0. Thus, at pH 7.4 roughly 98% of the HMSiR molecules in solution occupy the OFF state, which enables facile detection and localization of the sparse subset of molecules that are emissive (ON).[15] HMSiR’s cell permeability, photostability, and ability to blink in the absence of chemical additives has enabled multiple minimally invasive single-color SMLM experiments, including those that visualize organelle membrane dynamics in live cells for extended times,[26] others that resolve the morphology of dopaminergic neurons in an intact Drosophila melanogaster adult brain,[27] and still others that enable turn-on visualization of intracellular protein targets.[28]

Figure 1

(a) Structure and pH-dependent equilibrium of the spontaneously blinking fluorophore HMSiR.[15] (b–d) Structures of previously reported fluorophores considered as potential HMSiR partners for multicolor live-cell SMLM[16,18,21,29] (e) Structure of the spontaneously blinking fluorophore reported herein, Yale676sb.

Despite these advances, there remains a need for new SBFs that effectively partner with HMSiR to enable multicolor live-cell SMLM experiments without the need for chemical additives or photoactivation.[13] Although two green-emitting SBFs whose emission spectra are separable from HMSiR have been reported (Figure b),[18,19] including one (HEtetTFER) that can be paired with HMSiR for two-color SMLM in fixed cells,[18] their use demands high-intensity lasers that excite at 488 and 561 nm, respectively. These light sources can induce substantial cytotoxicity as phototoxicity is especially pronounced in the blue and green spectrum.[14] Two other previously reported SBFs are excitable in the far-red/near-IR (Figure c),[20,21] but they are spectrally indistinguishable from HMSiR and therefore not suitable for two-color experiments. Although both the fluorescent protein mEos3.2 and CP550 (Figure d), a carbopyronin fluorophore that reacts irreversibly with intracellular glutathione,[20] have been paired with HMSiR for two-color, live-cell SMLM,[15,20] these experiments require an additional ∼560 nm laser, which is inferior to red-light excitation for live-cell microscopy.[10,14] Furthermore, sequential multicolor imaging with multiple lasers is slow and the images are prone to sample motion artifacts. Finally, the spontaneously blinking carborhodamine, HMCR550, which was designed using quantum calculations, has an excitation maximum at 560 nm and would likewise require multiple lasers to pair with HMSiR (650 nm excitation) for a two-color live-cell SMLM experiment.

Here we report the rational design of a new near-IR-emitting SBF that pairs effectively with HMSiR to enable simplified two-color SMLM experiments in live cells (Figure d). Yale676sb emits at 694 nm, the longest wavelength of any reported SBF, and possesses, to our knowledge, a higher quantum yield (0.59) than any previously reported nanoscopy-compatible Si-rhodamine (SiR) fluorophore. Yale676sb and HMSiR can be excited simultaneously with a single 642 nm laser and imaged ratiometrically for simultaneous multicolor SMLM of two distinct intracellular organelles (ER + mitochondria) in live cells.

Results

New Spontaneously Blinking Fluorophores: Design Considerations

Three distinct chemical and photophysical properties are needed to ensure compatibility with HMSiR for ratiometric two-color, live-cell SMLM. The first is an emission maximum > 690 nm to ensure adequate separation from HMSiR (emission maximum = 670 nm) via ratiometric imaging.[5,30,31] The second is a pKcycl value between 5.3 and 6.0 to ensure the sparsity of emissive/ON molecules.[15,21] The third requirement is a high quantum yield; although a quantum yield > 0.2 can yield respectable SMLM images, higher values are always more desirable.[13,22] The challenge is that the quantum yields of rhodamine-based fluorophores typically decrease as the absorption and emission maxima increase (Supporting Information (SI) Figure S1). As a result, molecules that absorb and emit at higher, less cytotoxic wavelengths that are compatible with live cells are relatively dim. This correlation is reflected in the relatively low quantum yield of HMSiR (0.31) when compared to those of the green-light-emitting SBFs HMJF526 (0.87)[16] and HEtetTFER (0.76).[18] We therefore sought a design approach that would yield fluorophores possessing both long-wavelength emission and high quantum yield.

HMSiRindol, HMSiRjulol, and HMSiRTHQ

Previous work has demonstrated that introduction of heterocyclic indoline,[32,33] julolidine,[34] or tetrahydroquinoline[32] moieties into the core of a SiR chromophore can shift the excitation and emission maxima by up to 50 nm relative to SiR itself (Figure a). To evaluate whether these effects would be preserved in the context of a HMSiR core, we synthesized HMSiR, as well as the heterocyclic derivatives HMSiRindol, HMSiRjulol, and HMSiRTHQ (Figure b and Supporting Information Schemes S1–S4) according to a recently reported general method for Si-rhodamine fluorophore synthesis.[35] We then characterized the photophysical properties and aqueous spirocyclization equilibrium (pKcycl) of each new fluorophore (Figure b–d).

Figure 2

(a) Structures and photophysical properties (ƛabs, ƛem, and Φ) of previously reported SiR fluorophores with red-shifted absorption and emission spectra and HMSiR analogs HMSiR, HMSiRindol, HMSiRjulol, and HMSiRTHQ. Normalized (b) absorption and (c) emission spectra of HMSiR, HMSiRindol, HMSiRjulol, and HMSiRTHQ in 0.2 M sodium phosphate (pH = 4.5 for HMSiR, pH = 2.0 for HMSiRindol, HMSiRjulol and HMSiRTHQ). (d) pH-dependent change in absorbance of 2 μM HMSIR (650 nm), HMSiRindol (697 nm), HMSiRjulol (684 nm), and HMSiRTHQ (678 nm) as a function of pH in 0.2 M sodium phosphate buffer at room temperature. The absorbance of each fluorophore was monitored at the wavelength of maximal absorbance in panel b.

Each of the new fluorophores displayed absorption (Figure b) and emission (Figure c) maxima that were red-shifted by at least 23 nm relative to HMSiR, with the emission maxima increasing in the order HMSiR < HMSiRTHQ < HMSiRjulol < HMSiRindol. As expected, the absorption and emission maxima of the HMSiR series were nearly identical to those of the analogous SiR variants reported previously.[32−34] The pKcycl of each new HMSiR analog was determined from a plot of the pH dependence of the absorption of each fluorophore at the absorption maximum of the open/ON form (Figure d and SI Figure S2); the pKcycl is the pH at which the concentration of the open/ON state equals that of the closed/OFF state.[15] The pKcycl values of HMSiRindol and HMSiRTHQ were 6.4 and 6.9, respectively, both significantly higher than the value for HMSiR (6.0). The pKcycl value of HMSiRjulol (pKcycl = 9.0) was shifted even more dramatically, presumably because the additional electron-donating alkyl groups disfavor cyclization. A related previously reported rhodamine analog with julolidine groups also displayed a high pKcycl. value.[15] The absorbance vs pH curves for HMSiRindol, HMSiRjulol, and HMSiRTHQ are sigmoidal, whereas that of HMSiR is bell-shaped due to cyclization of the protonated fluorophore at low pH; this protonation is disfavored when the exocyclic amine is constrained by a five- or six-membered ring.[15,36]

The final criterion needed to ensure compatibility with HMSiR for ratiometric two-color, live-cell SMLM is a high quantum yield. The quantum yields measured for HMSiRindol, HMSiRjulol, and HMSiRTHQ also paralleled the values for the analogous SiR variants; the quantum yield of HMSiRindol, like SiR700, was low (0.13), whereas those of HMSiRjulol and HMSiRTHQ (0.43 and 0.38, respectively) were comparable to that of HMSiR (0.31) (SI Figure S3).

These data indicate that neither HMSiRTHQ, HMSiRjulol nor HMSiRindol possess the characteristics necessary to partner with HMSiR for two-color SMS nanoscopy. Although all three fluorophores exhibit emission maxima that are shifted by at least 23 nm from that of HMSiR, and HMSiRjulol and HMSiRTHQ display acceptable quantum yields (0.43 and 0.38), none feature pKcycl values low enough to prevent significant multiemitter artifacts at physiological pH. In each case, chemical modifications are needed to increase the electrophilicity of the xanthene core, favor spirocyclization, and decrease pKcycl. Ideally, these modifications should also increase quantum yield to increase brightness and resolution, but as outlined below, this goal is complicated by the complex interplay between quantum yield, emission maximum, and pKcycl.

Interplay between Quantum Yield, Emission Maximum, and pKcycl

The quantum yields of rhodamine fluorophores are limited by a nonradiative decay process known as twisted intramolecular charge transfer (TICT).[37−39] TICT involves the excited-state transfer of an electron from the exocyclic nitrogen of the fluorophore to the neighboring carbon π system with concomitant twisting of the Caryl–N bond; the charge-separated state subsequently decays to the ground state without emission of a photon. Processes that decrease the propensity for Caryl–N bond rotation increase quantum yield. For example, the quantum yields of rhodamine B and tetramethyl rhodamine (TMR) are higher in viscous solvents[37] and at low temperature where Caryl–N bond rotation is inhibited.[37,40] Indeed, the modestly increased quantum yields of HMSiRjulol (0.43) and HMSiRTHQ (0.38) relative to HMSiR (0.31) can be ascribed to restricted Caryl–N bond rotation,[34] although these effects appear to be less dramatic in the SiR series than with conventional rhodamines: rhodamine 101, the rhodamine analog of HMSiRjulol, displays a near-perfect quantum yield of 0.99.[40]

TICT is also inhibited in fluorophores in which the ionization potential (IP) of the exocyclic nitrogen is increased by electron-withdrawing groups (EWGs).[18,38,41] Addition of EWGs to a fluorophore core also decreases pKcycl by lowering the energy of the fluorophore’s lowest unoccupied molecular orbital (LUMO).[15,18] However, the addition of EWGs typically induces moderate to large decreases in excitation and emission wavelength maxima. For example, an EWG-containing fluorophore reported by Lv et al. possesses an exceptional quantum yield (0.66) but is blue-shifted by ∼20 nm relative to HMSiR (ƛabs/ƛem = 631 nm/654 nm).[41] We reasoned that combining the effects of restricted aryl-N bond rotation with an EWG would simultaneously reduce pKcycl and increase quantum yield by inhibiting TICT. If these changes were introduced into the HMSiRTHQ scaffold, even a moderate decrease in excitation and emission maxima would not jeopardize the emission shift needed to remain orthogonal to HMSiR. HMSiRTHQ was preferred as a starting point because its pKcycl (6.9) and quantum yield (0.38) are both close to those of HMSiR, in contrast to HMSiRindol, whose quantum yield is low (0.13), or HMSiRjulol, whose pKcycl is very high (9.0).

Design of the Bright, Near-IR-Emitting SBF, Yale676sb

To test this hypothesis, we synthesized Yale676sb, a variant of HMSiRTHQ in which two N-methyl groups were replaced symmetrically by monofluorinated N-ethyl groups (Figure and SI Scheme S5). As predicted, Yale676sb was characterized by a 10-fold more favorable spirocyclization equilibrium than HMSiRTHQ (pKcycl = 5.9 vs 6.9) and a greatly improved quantum yield (0.59 vs 0.38) (SI Figure S3). Interestingly, Yale676sb exhibited absorption and emission ƛmax that are both virtually identical to those of HMSiRTHQ. Addition of a stronger difluorinated N-ethyl group to generate Cal664sb resulted in a further increase in quantum yield to 0.74 (SI Figure S3) but, in this case, led to an emission ƛmax that was too close to that of HMSiR (667 nm vs 677 nm) to support two-color ratiometric imaging. The photophysical properties associated with Yale676sb suggest that it should be an ideal partner for HMSiR: an emission maximum > 690 nm, a pKcycl value between 5.3 and 6.0, and a high quantum yield. The quantum yield of Yale676sb (0.59) is, to our knowledge, higher than any Si-rhodamine derivative prepared and utilized for fluorescence nanoscopy.

Figure 3

Structures and photophysical properties (ƛabs, ƛem, and Φ) of (a) HMSiR and HMSiR2FlEt, and (b) HMSiRTHQ, Yale676sb, and Cal664sb. Normalized (c) absorption and (d) emission spectra of HMSiR, HMSiRindol, HMSiRjulol, and HMSiRTHQ in 0.2 M sodium phosphate (pH = 4.5 for HMSiR; pH = 2.0 for HMSiRindol, HMSiRjulol, and HMSiRTHQ). (e) pH-dependent spirocyclization equilibria. Normalized absorption of open form of 2 μM HMSiR, HMSiR2-FlEt, HMSiRTHQ, Yale676sb, and Cal664sb as a function of pH in 0.2 M sodium phosphate buffer at room temperature.

To deconvolute the effects of aryl-N bond rotation and the monofluoro electron-withdrawing group, we also prepared HMSiR2-FlEt, which carries the same monofluorinated N-ethyl groups but allows aryl-N bond rotation (SI Scheme S7). HMSiR2-FlEt was characterized by a minimal change in absorption and emission ƛmax relative to HMSiR; however, it displayed a 10-fold more favorable spirocyclization equilibrium than HMSiR (pKcycl = 5.0 vs 6.0), a value too low for efficient blinking at physiological pH of 7.4.[21] Its improvement in quantum yield was more modest relative to Yale676sb (0.51 vs 0.59). These comparisons emphasize the benefits of combining restricted aryl-N bond rotation with an EWG.

Evaluation of the Single-Molecule Properties of Yale676sb

To ensure that the bulk photophysical parameters of Yale676sb would translate into efficient single-molecule parameters, we evaluated its properties under SMLM imaging conditions. We quantified the “ON time” to determine the dye’s compatibility with HMSiR by imaging single dye molecules immobilized on glass coverslips (SI Figure S4). By monitoring individual molecules, we were able to determine the ON time to evaluate the compatibility of Yale676sb and HMSiR. Because both dyes are imaged on the same camera using ratiometric imaging, similar ON times allow a single camera integration time to be effective for acquiring data from both fluorophores.[42] From these data, we determined that Yale676sb has an ON time of 4.5 ms at pH 7.4, which is close to the ∼10 ms ON time reported for HMSiR, and in theory should allow even faster imaging.[15] This short ON time, combined with the high quantum yield also makes the Yale676sb/HMSiR combination suitable for high-speed imaging, with camera frame rates as high as 400 frames per second (fps). With an OFF time of 3.8 s, we expect an ON fraction or duty cycle of 0.0012.

Single-Color Live-Cell SMLM with Yale676sb

We next tested whether Yale676sb would support single-color, live-cell SMLM imaging. U2-OS cells that were engineered to overexpress the endoplasmic reticulum (ER)-localized protein Halo-Sec61β[43] were treated with 300 nM Yale676sb-CA (SI Scheme S8) for 30 min, washed, and immersed in a standard live-cell imaging solution using a custom-built SMLM instrument (see the SI discussion of methods). Figure a shows a representative super-resolution image (out of n = 16 images) acquired over 5 s. These images revealed multiple tubules in the cell periphery that were ∼99 ± 15 nm (mean ± s.d.) wide, a value comparable to ER morphology metrics acquired using both STED and 4Pi-SMS.[44] A time series illustrates changes in ER morphology that occur over the course of 10 s (Figure B). On average, we detected ∼800 photons per blink, corresponding to a localization precision distribution with a peak at ∼20 nm (Figure c,d).

Figure 4

(a) Super-resolution image of the ER in U2-OS cells using Yale676sb. Average reconstructed signal as a function of position along the seven line profiles indicated by yellow lines is shown. Insets: (b) dynamic ER network remodeling; histograms illustrating range in no. of photons (c) and localization precision (d) associated with single molecules in panel a. (e) Two-color super-resolution image of the ER and mitochondria in U2-OS cells using Yale676sb (magenta) in conjunction with HMSiR (green). Insets: (f, g) super-resolved mitochondrial and ER networks in close proximity. Scale bars: 5 μm for panels a and e; 1 μm for panels b and f. All reconstructions using 5 s of acquired frames.

Ratiometric Two-Color Live-Cell SMLM with Yale676sb and HMSiR

Next we sought to evaluate whether Yale676sb would support live-cell multicolor imaging in combination with HMSiR. U2-OS cells were transiently transfected with Halo-Sec61β (to reveal the ER) and SNAP-OMP25 (to reveal the outer mitochondrial membrane), treated with Yale676sb-CA and HMSiR-BG, and imaged using the identical SMLM setup. As predicted from the absorption and emission spectra of Yale676sb and HMSiR, both dyes could be excited with the same 642 nm laser and ratiometrically separated from two simultaneously acquired images detecting the emission wavelength ranges of 650–680 and 680–750 nm, respectively (SI Figure S5). Figure e shows a two-color super-resolution image, accumulated over 5 s, revealing the intertwined mitochondrial and ER networks of the cell. We detected comparable average photon numbers per frame for the two dyes (∼500 and 590 photons for Yale676sb and HMSiR, respectively), especially given that the filters and excitation wavelength were optimized for HMSiR.

Conclusions

In summary, here we report a new spontaneously blinking Si-rhodamine, Yale676sb, that can be used alongside HMSiR to enable two-color ratiometric SMLM in living cells in physiological media. This new experiment was facilitated by three unique photophysical metrics associated with Yale676sb: (1) an exceptionally high quantum yield for a silicon rhodamine derivative (0.59); (2) an unusually long emission maximum (694 nm); and (3) a pKcycl value (5.9) that is nearly identical to that of HMSiR (6.0).

The unique photophysical metrics associated with Yale676sb result from the simultaneous introduction of both heterocyclic rings as well as electron-withdrawing dialkyl amino groups (DAGs) into the silicon rhodamine core. When either of these structural features is introduced in isolation, at least one of the three critical photophysical metrics required for two-color SMLM becomes nonoptimal. Silicon rhodamine dyes with only heterocycle-containing dialkyl amino groups (such as HMSiRindol, HMSiRTHQ, and HMSiRjulol) display long-wavelength emission (689–716 nm) but resist spirocyclization. As a result, their pKcycl values (6.4–9.0) are too high to ensure adequate distribution of single-molecule emitters (Figure ). By contrast, silicon rhodamine dyes with only electron-withdrawing substituents, such as HMSiR2-FlEt, display a high quantum yield, but their spirocyclization equilibrium is too favorable, and their pKcycl values are too low (Figure ). By combining these two substitution patterns in Yale676sb, the competing effects on pKcycl are balanced, while the red shift from the heterocycle-containing DAG is maintained (Figure ). Moreover, because both the rotational restriction from the heterocycle-containing DAGs and the electron-withdrawing capacity of the 2-fluoroethyl substituent inhibit twisted intramolecular charge transfer, the quantum yield increase from the latter is not only maintained, but enhanced (0.51 vs 0.59).

As expected, switching from a 2-fluoroethyl to a more electron-withdrawing 2,2-difluoroethyl substituent at the nitrogen in Cal664sb further increases the quantum yield, though at the expense of both pKcycl and emission wavelength (Figure ). This pattern would likely continue with increasingly electron-withdrawing substituents. Despite these blue shifts, Cal664sb displays a comparable quantum yield to a previously reported and exceptionally bright Si-rhodamine fluorophore (compound 9 in ref (41)), but with a >30 nm longer emission maximum.[41]

Finally, we note that while the quantum yield increase relative to HMSiR observed with HMSiR2-FlEt is not as dramatic as that observed with Yale676sb, it is comparable to that observed from more commonly used azetidinyl substituents.[16,35,38,46,47] Being that the former requires only one position at each aniline nitrogen to be substituted, whereas the latter requires two, use of the 2-fluoroethyl substituent may serve as an alternative method for increasing quantum yield of rhodamine derivatives, especially those with more complex DAGs. This approach and others described herein may serve as versatile methods for the preparation of even more greatly enhanced fluorescent labels.