Divergent Synthesis of Ultrabright and Dendritic Xanthenes for Enhanced Click‐Chemistry‐Based Bioimaging

Abstract Biorthogonal labelling with fluorescent small molecules is an indispensable tool for diagnostic and biomedical applications. In dye‐based 5‐ethynyl‐2′‐deoxyuridine (EdU) cell proliferation assays, augmentation of the fluorescent signal entails an overall enhancement in the sensitivity and quality of the method. To this end, a rapid, divergent synthetic procedure that provides ready‐to‐click pH‐insensitive rhodamine dyes exhibiting outstanding brightness was established. Compared to the shortest available synthesis of related high quantum‐yielding rhodamines, two fewer synthetic steps are required. In a head‐to‐head imaging comparison involving copper(I)‐catalyzed azide alkyne cycloaddition reactions with in vitro administered EdU, our new 3,3‐difluoroazetidine rhodamine azide outperformed the popular 5‐TAMRA‐azide, making it among the best available choices when it comes to fluorescent imaging of DNA. In a further exploration of the fluorescence properties of these dyes, a set of bis‐MPA dendrons carrying multiple fluorescein or rhodamine units was prepared by branching click chemistry. Fluorescence self‐quenching of fluorescein‐ and rhodamine‐functionalized dendrons limited the suitability of the dyes as labels in EdU‐based experiments but provided new insights into these effects.


Introduction
In the field of bioimaging and diagnostics, fluorescent labelling has emerged as a potent tool for the elucidation of structures, dynamics, interactions and functions of biomolecules such as proteins, [1] nucleic acids, [2] polysaccharides [3] and lipids. [4] Furthermore, efficient fluorescence resonance energy transfer (FRET)-compatible fluorophores and quenchers are increasingly sought after for their use in real-time PCR applications, including as components of molecular beacons, [5] TaqMan probes, [6] and Scorpion primers. [7] In the context of cell proliferation detection, fluorescence labelling of nucleic acids helps to assess genotoxicity of new pharmaceuticals and to evaluate anticancer drugs. [8] In order to obtain an effective image and produce high signal-to-noise ratios, probes must exhibit strong fluorescent signals. A known disadvantage of fluorescent labels over traditional radioactive labels is their moderately lower sensitivity. [9] This difference can lead to undesirable results, especially in the framework of oncology, where slowly proliferating cancerogenic cells have been reported to escape detection. [10] Upon chemical functionalization with multiple dyes to enhance fluorescence signal, proteins and other biomolecules can become inactivated due to their large structural alteration. [11] Fluorescence brightness, that is, the product of a fluorophore's extinction coefficient and fluorescence quantum yield (ɛ · φ), is used to compare the fluorescent properties of different dyes. [12] Therefore, to overcome concerns related to fluorescence intensity and biomolecule alteration, fluorophores with enhanced brightness values are highly desirable.
Click chemistry remains the gold standard labelling strategy used in imaging experiments to attach fluorophores to the biomolecule of interest, [13] with EdU-based assays being particularly useful for cell proliferation detection, and where the clickable thymidine analogue EdU is metabolically incorporated during active DNA synthesis. [14,15] Unlike halogenated thymidine analogues such as 5-bromo-2'-deoxyuridine (BrdU), EdU assays do not require harsh DNA denaturing conditions, thus preserving cellular and tissue integrity. [14,16] In this study, we disclose a facile, three-step synthesis of highly bright ready-to-click rhodamine dyes and systematically explore their photophysical properties. Moreover, we demonstrate the applicability of these dyes in the framework of EdUbased assays. In a further exploration of their synthetic utility and fluorescent properties, we prepare dendrons containing multiple branched fluorophores with the aim of augmenting fluorescence signal without increasing the number of labellingsites within a given alkyne-modified DNA molecule. [17] The use of related, but not identical constructs has been demonstrated to enable a wide variety of applications, particularly in drug and gene delivery, [18] cancer therapy [19] and tissue engineering. [20] Controlled synthesis of large dendritic scaffolds remains a considerable challenge, particularly where structural characterization is concerned, thus prompting the need for new and improved synthetic strategies. In the context of fluorescent scaffolds, self-quenching has also proven problematic. [9,21,22] In certain instances, fluorescence intensity in antibody-and DNAconjugates has nonetheless been successfully enhanced, prompting us to investigate this approach ourselves. [9,23,24] To such an end, we demonstrate the synthesis by means of click chemistry, of a set of bis-MPA dendrons carrying either multiple fluorescein units, or our best-performing rhodamine. After subsequent characterization of the photophysical properties of these fluorogenic compounds, we evaluate their suitability as fluorescent labels for EdU-based experiments as well as other potential applications.

Results and Discussion
Rhodamines, first described in the 1880s, [25] are a family of xanthene dyes which exhibit excellent brightness, exceptional photostability and low pH-sensitivity. [26,27] Traditionally, rhodamine dyes (3) have been synthesized by means of an acidcatalyzed condensation reaction between a phthalic anhydride (1) and an aminophenol (2) (Scheme 1.1). This reaction is, however, characterized by harsh conditions, low yields, mixtures of regioisomers and incompatibilities with several functional groups. [26,[28][29][30] With the aim to omit this step, Lavis and coworkers reported a synthesis based on the use of fluorescein ditriflates (5) as key intermediates and a Pd-catalyzed cross coupling for the formation of CÀ N bonds (Scheme 1.2). [28,31] Following this strategy, milder reaction conditions, higher yields and the use of a wide range of nitrogen nucleophiles were successfully carried out. In this work, the latter methodology has been modified for the synthesis of ready-to-click rhodamine dyes in a facile and straightforward manner for bioconjugation with DNA via click chemistry. This new route only comprises 3 synthetic steps and a late-stage formation of fluorescein azide ditriflate 9, which enables a divergent synthesis of different rhodamine dyes through Buchwald-Hartwig cross-coupling (Scheme 1.3).
The synthetic pathway starts with 6-carboxyfluorescein (7) as an inexpensive, isomerically pure starting material. The 6carboxyl group was converted in situ in an N-hydroxysuccinimide (NHS) ester using a catalytic amount of 4-(Dimethylamino)pyridine (DMAP), triethylamine and N,N'-disuccinimidyl carbonate (DSC) as a coupling reagent. Upon activation as an NHS ester, the bifunctional linker azidopropan-1-amine was added to provide fluorescein azide (8) as a bright orange solid in 70 % yield. The next step involved conversion of the phenolic groups into triflates using trifluoromethanesulfonic anhydride (Tf 2 O) and pyridine to yield fluorescein azide ditriflate 9 in 81 % yield as the key intermediate of the synthesis. Subsequent Buchwald-Hartwig cross-couplings were performed using Pd 2 (dba) 3 /XPhos as the catalytic system and Cs 2 CO 3 as base. In a similar approach as previously reported in the literature, [30,32] catalyst and ligand loadings were increased to suppress triflate hydrolysis, which represented the main side-reaction of this step. Different 3-substituted azetidines were used as nitrogen nucleophiles in the cross-coupling reaction since rhodamine dyes containing azetidine rings were found to exhibit considerable quantum yields. [28,31] In this manner, dyes RD H2 (10), RD m (11), and RD F2 (12) were successfully synthesized as pink-topurple solids in 42 %, 60 % and 74 % yield, respectively. A study of the photophysical properties of these dyes was performed by UV-Vis and fluorescence spectroscopic techniques at different pH values (Table 1). In order to observe the properties of the zwitterionic and cationic forms of the rhodamines, [26] measurements were taken at close-to physiological (pH = 7.3) and acidic (pH = 1.9) pH values, respectively (see Supporting Information, Figures S19-21 for the calculated pk a values). As for the absorption and emission maxima (λ abs , λ em ), the synthesized dyes absorb green and green-to-yellow light (λ abs = 530-556 nm) and emit green-to-yellow and yellow light (λ em = 554-580 nm). A larger hyposochromic 5(blue) shift in both λ abs and λ em was observed when using azetidine derivatives containing more electron-withdrawing groups. Thus, λ abs and λ em of these dyes obey the following tendency: RD H2 > RD m > RD F2 . This set of rhodamine dyes showed high ɛ max values (ɛ max = 53000 to 68000 M À 1 cm À 1 ) with low pH sensitivity. Likewise, the rhodamines showed high to very high quantum yield values (φ = 0.67 to 0.89) with minimal pH-dependent effects, except for RD m (φ = 0.67 at pH 7.3 and φ = 0.86 at pH 1.9). This quenching behavior, as described in the literature, [28,34,35] is indicative of an intramolecular photoinduced electron transfer (PeT) due to the presence of unprotonated morpholino amines. Consequently, the quantum yield value of RD m increased dramatically at pH 1.9.
Given its remarkably high extinction coefficient, quantum yield and low pH-dependence, RD F2 dye (ɛ max = 60000 M À 1 cm 1 and φ = 0.89 at pH 7.3) stood as the best-performing dye in terms of fluorescence brightness. With this in mind, we assessed the applicability of RD F2 to an EdU cell proliferation assay ( Figure 1a). HEK-293T cells were pulsed for 2 h with EdU, fixed, permeabilized and Cu-catalyzed in situ click reactions were performed in parallel with RD F2 and the well-established 5-TAMRA-PEG3-azide. After image acquisition by fluorescence microscopy (Figure 1b), both dyes provided images with characteristic nuclear patterns of DNA replication and similar fluorescence intensity. The emission filter used in this assay, however, exhibited an acquisition window overlapping more extensively the emission spectrum of TAMRA in detriment of that of RD F2 . This indicates, therefore, that RD F2 performs at least as robustly as the well-established TAMRA dye upon clickmediated in situ EdU labelling.
Encouraged by these results, we investigated the use of dye-containing dendrons as fluorescent labels to further amplify the fluorescence signal without increasing the number of labelling-sites on DNA (Scheme 2). We therefore synthesized and characterized the photophysical properties of a family of fluorescein-based dendritic dyes carrying two (FD2, 12), four (FD4, 13) and eight (FD8, 14) branched fluorescein substituents. In order to synthesize these multivalent fluorescent dyes, we chose a set of 2,2-bis(methylol)propionic acid (bis-MPA) polyester dendrons as functional, biodegradable and low-cytotoxic polymeric scaffolds suitable for biological applications. [17] The present strategy consisted of functionalizing the different dendritic structures with units of fluorescein azide via a copper-catalyzed click reaction to yield FD2, FD4 and FD8 in 61 %, 67 % and 40 % yield, respectively. FD8 was further functionalized to include an azide moiety into its focal point to enable a possible bioconjugation via click chemistry. Upon treatment of FD8 with trifluoroacetic acid, the Boc-protected amine was released in quantitative yields. The subsequent coupling reaction with the NHS ester of an azide-containing linker afforded FD8-N 3 (16) in 67 % yield.
In an analogous approach to that of the synthesized rhodamine dyes, the photophysical properties of these dendrons were evaluated at different pH values (Table 2 and Figure 2). The spectral properties of fluorescein are known to be pH-sensitive, especially due to the equilibrium between the mono-and dianion forms with a pK a of 6.4, [37] the dianion being the most fluorescent species. [37,38] For this reason, measurements were taken at close-to physiological (pH = 7.3) and basic (pH = [c] Maximum extinction coefficients (ɛ max ) were calculated by a linear regression analysis obeying the Beer-Lambert law.
Regarding the emission of the fluorophores, a decrease of the fluorescence quantum yield (φ) was observed along with the increasing number of fluorescein units ( Table 2). More specifically, quantum yields at pH 7.3 for fluorescein azide, FD2, FD4 and FD8 were, in that order, 0.77, 0.13, 0.04 and < 0.01. Analogously, the decay in quantum yield was observed at pH 9.1 although a wide variation as a function of pH was noted, particularly for fluorescein azide and FD2. Thus, the quantum yield values at pH 9.1 were 0.91, 0.26, 0.05 and < 0.01 for fluorescein azide, FD2, FD4 and FD8, respectively.
The fluorescence brightness (Table 2) diminished upon increasing the number of dyes per molecule at both pH values.  [a] Measurements were taken in 10 mM HEPES, pH 7.3 buffer at room temperature.
[b] Measurements were taken in 10 mM sodium borate, pH 9.1 buffer at room temperature.
[c] Maximum molar extinction coefficients (ɛ max ) were calculated by a linear regression analysis obeying the Beer-Lambert law.
Therefore, even though the absorption increased upon increasing the number of fluorescein moieties, the strong decay in quantum yield rendered a decrease in the fluorescence intensity. This behavior is confirmed by the fluorescence spectra shown in Figure 2. This decrease in fluorescence intensity was attributed to self-quenching, presumably resulting from the relatively small Stokes shift of fluorescein and the proximity of the dyes within the dendritic structures. [9,38,39] It is noteworthy to mention the bimodal shape of the absorbance spectrum of FD8 at pH 7.3. In comparison to the other analogous dendrons, FD8 exhibited a hypsochromic shift in which a second absorption band arose at 462 nm. This was suggestive of the formation of H-aggregates caused by π-π interactions between the different dye units. [23,[40][41][42] . These interactions were less prevalent at pH 9.1 with the more polar and water-soluble dianion form of fluorescein. [43,44] Hence, the absorption spectrum recovered the original pattern. After evaluating the fluorescein-based family of dendritic fluorophores, we concluded that the macromolecules are illsuited for use in nucleic acid detection compared with their individual monomeric substituents. The strong interactions between the different dye units led to self-quenching, which was evidently undesirable for the synthesis of dendron-based fluorescence amplifiers.
Raddaoui, Stazzoni et al. [24] developed a TAMRA-functionalized dendritic fluorophore which was bioconjugated to EdUlabelled DNA via click chemistry and achieved fluorescence amplification. Inspired by this work, we focused our efforts on the preparation of a rhodamine-based dendritic fluorophore using our model dye RD F2 . In the event, after functionalizing an alkyne-presenting bis-MPA dendron via copper-catalyzed click chemistry, RD F2 D4 was afforded as a bright, pink powder in 85 % yield (Scheme 2). Once the photophysical properties were measured (Table 2 and Figure 3), RD F2 D4 was found to exhibit absorption and emission wavelengths very similar to RD F2 at both pH 7.3 and pH 1.9. As for the extinction coefficient, lower values than those of RD F2 were noted as well as an important pH-dependence (ɛ max = 33000 M À 1 cm À 1 at pH 7.3 and ɛ max = 57000 M À 1 cm À 1 at pH 1.9). In terms of quantum yield, an important decay of the fluorescence emission was observed with a negligible effect of pH (φ = 0.13 at pH 7.3 and φ = 0.14 at pH 1.9). Therefore, as shown in Figure 3, the fluorescence intensity of RD F2 D4 did not surpass that of RD F2 . The most plausible explanation of this behavior is, once again, the selfquenching caused by the relatively small Stokes shift of RD F2 as well as the relative proximity of the dye units within the dendron. A tendency of RD F2 D4 to form H-aggregates in aqueous solution is also observed by the presence, at both pH values, of an additional shoulder at 505 nm (pH = 7.3) or 510 nm (pH = 1.9). The suitability of RD F2 D4 as a fluorescent amplifier is, thus, limited by its inherent self-quenching.
Previous work has demonstrated aggregates tend to diminish upon temperature elevation due to changes in the dynamic equilibrium, and thus resulting in an increase of fluorescence intensity. [45À -47] Conversely, non-radiative rate constants and collisional quenching also tend to be amplified upon heating, leading to a decrease of fluorescence. [48] In order to evaluate which of the two effects predominate in our dendritic fluorophores, the fluorescence of these molecules was measured at varying temperature values between 25°C and 65°C, as shown in Figure 4. In the event, the fluorescein-based dendrons FD2 and FD4 exhibited comparable fluorescence decay behavior. FD8, meanwhile, showed a diminished change in fluorescence intensity upon heating compared with FD2 and FD4. This behavior could be indicative of a reconversion of aggregates to monomers in solution. For RD F2 D4, a less pronounced decay of fluorescence is shown in comparison to its fluorescein analogue FD4, indicating that aggregation might occur at a lesser extent for the former compound at higher temperatures. Overall, it is apparent that non-radiative processes and collisional quenching prevail for all fluorophores upon increase of temperature, leading to a general decay in their fluorescence intensity.

Conclusion
The improvement and development of new dyes is of utmost importance for fluorescence-based diagnostic advancements. In cell proliferation assays, high signal-to-noise ratios allow reliable detection of smaller number of proliferating cells. To achieve these requirements, fluorophores with enhanced fluorescent properties must be identified, synthesized, and introduced into biomolecules without jeopardizing their biological activity.
In this work, the successful development of a direct, divergent pathway for the synthesis of ready-to-click rhodamine  dyes afforded three fluorophores, namely, RD H2 , RD m and RD F2 , showing remarkable photophysical properties, low pH-dependence, and high potential as fluorescent probes in EdU cell proliferation assays. The development of this new synthetic pathway, therefore, proves as a promising strategy for the rapid development of new click chemistry-conjugable fluorophores.
To further increase the fluorescence signal of our fluorophores, a family of three bis-MPA dendritic scaffolds containing 2, 4 and 8 alkyne functionalities were decorated via click chemistry. In this case, fluorescein azide was used as a standard fluorophore to afford FD2, FD4, and FD8, respectively. As for the photophysical properties, which were found to be highly pH-dependent, an increase of the maximum extinction coefficient along with the number of dye units was observed. In contrast, the fluorophores showed a decrease of quantum yield together with the increasing number of dye moieties which, in turn, caused a decay of the fluorescence intensity. We decided then to use our model rhodamine dye RD F2 to prepare RD F2 D4. This latter molecule showed lower values of both extinction coefficient and quantum yield in comparison to RD F2 and, thus, a decay of the fluorescence intensity. Ultimately, the effect of temperature on fluorescence intensity of the dendritic fluorophores was evaluated, leading to a general decay of fluorescence. This behavior suggests that collisional quenching prevails over aggregate dissociation upon temperature increase.
Albeit the observed self-quenching of fluorescence was not favorable in this study, comprehensive insights of the behavior of these dye-functionalized dendrons were acquired. It is apparent that the development of these dendritic structures remains challenging [9,21,22] and further optimization must be carried out. Relatively high lipophilicity of the employed fluorophores appears to be the main reason for the observed self-quenching. Introduction of water-soluble functionalities into the dye moieties would reduce closeness between them, elude aggregation and possibly enable amplification of the fluorescent signal. Nevertheless, the strongly absorbing and nonfluorescent properties of these dendrons could serve in other applications as quenchers for Förster resonance energy transfer (FRET) experiments. [49] Experimental Section Materials and methods for chemical synthesis and full characterization of all new compounds are found in the Supporting Information.
UV-vis and fluorescence spectroscopy: Samples for spectroscopy were prepared as stock solutions in DMSO and diluted such that the DMSO concentration did not exceed 1 % (v/v). Absorption spectra were recorded on an Agilent Cary 50 spectrometer. Fluorescence spectra were recorded on a Shimadzu RF-5301PC fluorescence spectrometer. All measurements were performed at ambient temperature using 1 cm path length, 3.5 mL quartz cuvettes from Hellma Analytics. Absorbance/extinction coefficient and emission scans are averages (n = 3).
Maximum extinction coefficient determination: All reported maximum extinction coefficients (ɛ max ) were calculated by a linear regression analysis obeying the Beer-Lambert law. Measurements were carried out using moderately concentrated samples (A < 0.8) to ensure linearity between absorbance and concentration. All absorbance values are averages (n = 3). Plots are found in the Supporting Information.
Quantum yield determination: All reported fluorescence quantum yield values (φ) were determined using the comparative method. [50] Fluorescein (φ = 0.91 in 0.1 M aqueous NaOH) [36] and rhodamine 6G (φ = 0.95 in EtOH) [28] were used as references for fluorescein-and rhodamine-based fluorophores, respectively. Reported refractive index values were taken for the aqueous solutions and buffers [51,52] (n = 1.33) and EtOH [53] (n = 1.36). Measurements were carried out using dilute samples (A < 0.1) to minimize inner filter effects. [54] All absorbance and integrated fluorescence intensity (fluorescence area) values are averages (n = 3). The slopes of the plots of fluorescence area vs. absorbance were used in the comparative method by means of the following equation: . Identical illumination intensity and exposure time were used for imaging samples treated with the two dye-azides. Images were processed using the ImageJ software.