A Nonaggregating Heptamethine Cyanine for Building Brighter Labeled Biomolecules.

Heptamethine cyanines are broadly used for a range of near-infrared imaging applications. As with many fluorophores, these molecules are prone to forming nonemissive aggregates upon biomolecule conjugation. Prior work has focused on persulfonation strategies, which only partially address these issues. Here, we report a new set of peripheral substituents, short polyethylene glycol chains on the indolenine nitrogens and a substituted alkyl ether at the C4' position, that provide exceptionally aggregation-resistant fluorophores. These symmetrical molecules are net-neutral, can be prepared in a concise sequence, and exhibit no evidence of H-aggregation even at high labeling density when appended to monoclonal antibodies or virus-like particles. The resulting fluorophore-biomolecule conjugates exhibit exceptionally bright in vitro and in vivo signals when compared to a conventional persulfonated heptamethine cyanine. Overall, these efforts provide a new class of heptamethine cyanines with significant utility for complex labeling applications.

Near-infrared (NIR) wavelengths can facilitate the visualization of biological processes in complex organismal settings. The use of fluorophore-labeled biomacromolecules is an enduring strategy employed across the spectrum of fundamental to applied biomedical science.[1,2] However, fluorophore conjugation often alters the properties of both the fluorophore and the molecule to which it is attached.[3] Specifically, important parameters such as brightness, target binding, in vivo stability, and pharmacokinetics (PK) are often impacted.[4−7] An important component of these issues is the formation of dye aggregates. In particular, H-aggregates, which are characterized by the appearance of a hypsochromic (blue-shifted) absorption band and a reduction in fluorescence intensity, are a common consequence of fluorophore bioconjugation.[3,8]

The most common strategy to avoid aggregation, pioneered with the “Cy” dyes, is the introduction of multiple anionic sulfonate groups.[9−11] This broadly adopted strategy has led to extensively used commercial dyes, including Licor’s IR-Dye800CW (IR-800CW, Figure A).[11] Notably, IR-800CW conjugates of the anti-EGFR antibody, panitumumab, and the small molecule, folate, are advancing through various clinical trials for fluorescence-guided surgical excision of solid tumors.[12−14] However, while dye sulfonation is a highly successful strategy in the visible range, persulfonated NIR fluorophores are still prone to the formation of aggregates at even moderate labeling density. Because NIR fluorophores are intrinsically less emissive than their counterparts in the visible range, strategies to circumvent the formation of nonemissive aggregates are a pressing need.

Figure 1

Prior strategies to improve heptamethine cyanine aggregation (A), efforts to improve their in vivo properties (B), and studies reported here (C).

Our approach to develop cyanine fluorophores for emerging imaging applications has been to identify novel chemical strategies that directly modify the core chromophore unit.[15−17] Relevant to these studies, we developed a general chemical strategy to modify the C4′ position of the heptamethine scaffold with an O-alkyl group.[18] These molecules are formed through an electrophile-integrating N- to O-rearrangement reaction from the corresponding readily available C4′ amino precursors. Significantly, the extensively studied C4′-O-aryl (e.g., IR-800CW) and related compounds form C4′-thio adducts upon exposure to thiol-containing biomolecules, as revealed recently in studies from our lab and others.[19−21] By contrast, these C4′-O-alkyl cyanines are resistant to this chemical pathway, circumventing the potential for these adducts to have deleterious effects. This chemistry has been employed to examine the role of substituent patterns on the properties of these molecules and their antibody conjugates.[22,23] For example, we examined the use of tetralkyl ammonium functional groups appended to indolenine nitrogens, a strategy examined productively for small-molecule-fluorophore conjugates.[24−28] However, in the case of antibody conjugates, H-aggregation remained a persistent issue, and the resulting compounds showed improved properties only at modest labeling density.[29]

To improve the properties of heptamethine cyanines for biomolecule labeling applications, we drew on recent studies that led to the identification of UL-766 (Figure B).[20] Fluorophores that undergo exclusive renal clearance could enable ready visualization of the ureter, a sensitive duct often injured during abdominal surgery. The compound that emerged for this application following significant screening efforts is the highly water-soluble but net-neutral UL-766. This net-neutral compound contains short ethylene glycol chains at the indolenine nitrogen positions and a trimethylalkylammonium salt at the C4′ position. In the design of these studies, we hypothesized that modification of this scaffold for bioconjugation through the tetraalkylammonium linker might deliver unique properties for biomolecule labeling.

In the report below, we detail the synthesis, characterization, and testing of a novel NIR fluorophore that circumvents the issue of H-aggregation and provides exceptionally bright fluorophore–antibody conjugates at high labeling density. This optimized dye, FNIR-Tag (Figure C), can be synthesized in a short, scalable (up to 0.5 g) synthetic sequence and is used here as the NHS ester and alkyne functionalized variants, though others are certainly feasible. Similarly labeled monoclonal antibody (mAb) conjugates of FNIR-Tag show superior tumor localization and brightness when compared with IR-800CW conjugates in an in vivo imaging study of mice bearing EGFR+ tumors. FNIR-Tag-mAb conjugates are also less subject to liver uptake than conjugates of IR-800CW. We also evaluated FNIR-Tag conjugates of virus-like particles (VLPs) derived from the bacteriophage Qβ. These conjugates also proved significantly brighter in both cellular and in vivo studies. Overall, these studies provide a NIR cyanine with excellent properties for complex, high-density labeling applications.

Results and Discussion

Synthesis and Photophysical Properties of a Net-Neutral, Bioconjugatable Fluorophore

The synthesis of the FNIR-Tag derivatives from previously described 1 is outlined in Scheme .[20] The addition of amine 2 (available in 3 steps from commercial material as described in the Supporting Information) to 4′-chloroheptamethine cyanine 1 in the presence of DIPEA at 120 °C for 25 min provided the deep blue C4′-N-linked heptamethine cyanine 3 in 47% yield after reversed-phase purification. In a one-pot process, carboxylic acid 4 was prepared by stirring 3 in neat TFA for 5 min at 60 °C, during which time LC-MS analysis of the reaction indicated both the expected removal of the t-butyl ester group and C4′ N- to O-transposition.[20] After removing the TFA, the mixture was exhaustively alkylated by treatment with excess NaHCO3 and MeI in DMF at 60 °C, resulting in the formation of both the desired carboxylic acid 4 (33%) and the methyl ester 5 in yields of 20–30%. The latter could be saponified by treatment with 1 M NaOH in MeOH:H2O to provide 4 (see Supporting Information). The carboxylic acid 4, FNIR-Tag, can be converted to FNIR-Tag-NHS using N,N,N′,N′-tetramethyl-O-(N-succinimidyl)uronium tetrafluoroborate (TSTU) in 88% yield. The conversion of FNIR-Tag-NHS to FNIR-Tag-alkyne was carried out using propargyl amine in 50 mM PBS (pH 7.4) in 71% yield.

Scheme 1

Synthesis of FNIR-Tag-NHS and FNIR-Tag-alkyne

Properties of unconjugated FNIR-Tag and IR-800CW carboxylic acids are shown in Table . As expected, the C4′ O-alkyl cyanine exhibits an absorption/emission maxima (765/788 nm) well in the NIR region that is similar to IR-800CW at 774/795 nm. Both dyes have similar absorption coefficients (ε), absolute quantum yields (ΦF), and brightness (ε × ΦF) in PBS solution with free IR-800CW being slightly brighter (Table ). Also as expected, both of the free dyes have similar photostability (Figure S1).

Table 1

Properties of the Free Dyes Used in This Study in PBS

	λabs (nm)	λem (nm)	ε (M–1 cm–1)	ΦF (PBS)	ε × ΦF	rel. brightness (PBS)
IR-800CWCO2H	774	795a	240 000	0.087	23 490	1.2
FNIR-Tag-CO2H	765	788b	200 000	0.099	19 800	1.0

λexcitation = 740 nm.

λexcitation = 730 nm.

Conjugation of FNIR-Tag with the anti-EGFR mAb panitumumab was carried out in 50 mM PBS (pH 7.4) with molar excesses of 2.2, 4.4, and 8.1 to provide the desired lysine-labeled panitumumab conjugates with degree of labeling (DOL) of 1, 2, and 4 (± 0.2), respectively (Figure ). We found that NHS-ester labeling proceeds with greater efficiency at pH 7.4 then at pH 8.5, even though the more basic pH is more commonly employed for NHS-ester labeling studies. We attribute this observation to the enhanced chemical reactivity of FNIR-Tag NHS-ester, which perhaps arises from an inductive effect of the electron withdrawing quaternary amine. Similarly labeled samples with IR-800CW (DOL 1, 2, and 4) were obtained by carrying out the labeling using 1 M PBS (pH 8.5). To confirm that the dyes were covalently attached, all samples were purified, incubated at 25 °C for 18 h, and then repurified by size exclusion chromatography (which led to a modest decrease in dye signal, Figure S2). Samples were analyzed by SDS-PAGE to confirm conjugate purity (Figure S3).

Figure 2

Absorbance and emission spectra of IR-800CW (A) and FNIR-Tag (B) to panitumumab (500 nM) of DOL 1, 2, and 4 conjugates in 50 mM pH 7.4 PBS. (C) Absolute quantum yields of fluorescence (ΦF) of the panitumumab conjugates (250 nM effective dye concentration) in 50 mM pH 7.4 PBS. (D) Phantom IVIS imaging of DOL4-panitumumab conjugates in 50 mM pH 7.4 PBS. Scale bar represents epi-fluorescent in total radiant efficiency.

The optical properties of these conjugates were analyzed by several methods. The absorption/emission curves are shown in Figure A, B (500 nM in protein concentration in 50 mM pH 7.4 PBS). The formation of a significant H-aggregate peak in the absorption spectrum (∼705 nm) is apparent with the IR-800CW conjugates even at DOL 1, which worsens at higher DOL of 2 and 4. By contrast, FNIR-Tag does not form a significant H-aggregate band over this range of labeling density. Of note, this trend persists in the presence of serum proteins (Figure S4). The impact of H-aggregation on the fluorescent emission of these conjugates is dramatic. With IR-800CW, the fluorescent emission of the antibody conjugates is nearly constant across the range of labeling density, meaning added fluorophores do not increase brightness of the individual antibody conjugates. By contrast, the emission of FNIR-Tag-panitumumab conjugates increases with higher labeling density. Demonstrating that this effect is the result of H-aggregation, diluting the conjugates in 50:50 MeOH:PBS to denature the protein removes the presence of H-aggregate peak and increases the emission of the conjugates (Figures S5 and S6). The impact of aggregation on emissive properties of these conjugates can also be demonstrated by measuring the absolute quantum yields of fluorescence (ΦF) of the differentially labeled conjugates. In the case of IR-800CW-antibody conjugates, quantum yield decreases as a function of increased labeling from ∼9 to ∼2.5%. By contrast, the quantum yield of FNIR-Tag conjugates maintains nearly the same value of the free dye (9–10%) with increasing labeling density. To examine the relative brightness of these conjugates on an in vivo imaging system, we carried out phantom imaging on the PerkinElmer IVIS system. In this context, DOL-4 antibody conjugates of FNIR-Tag could be readily visualized at a lower concentration (5 μg/mL) than that required for similarly labeled conjugates of IR-800CW (10–50 μg/mL) (Figure D).

Comparison of FNIR-Tag and IR-800CW mAb Conjugates In Vivo

The data above suggested that FNIR-Tag-mAb conjugates might have favorable fluorescence properties relative to those of IR-800CW in vivo. We also speculated that reduced dye aggregation might impact the pharmacokinetics of the mAb conjugates. To compare the two mAb conjugates, in vivo imaging was carried out using athymic nude mice bearing EGFR+ tumors implanted in their right flank (n = 5 per group). The mice were injected with 100 μg of either IR-800CW or FNIR-Tag conjugates of panitumumab of DOL 2 or 4. Fluorescence images were recorded before injection and at 10 min, 24 h, 48 h, and 1 week postinjection (Figure A and Figure S7). At the peak accumulation of both fluorophores in the tumor (48 h), the radiance output of FNIR-Tag is 2.5X higher at DOL 2 and 7.1X higher at DOL 4. We also noted that at an initial 10 min time point there was significantly higher liver uptake of the IR-800CW than with FNIR-Tag conjugates (Figure S8). However, at this 100 μg dose, the tumor to background (TBR, background taken in the neck region) ratio was indistinguishable during this time course (Figure S9). The latter observation is perhaps not surprising, as this dose was previously optimized for IR-800CW conjugates and not for conjugates of this new brighter dye.

Figure 3

(A) Images of MDA-MB-468 tumor-bearing mice injected with 100 μg of IR-800CW and FNIR-Tag at DOL 2 and 4 conjugates with panitumumab at preinjection and 10 min, 1 day, 2 day, and 7 day time points postinjection. Scale bar represents epi-fluorescent in total radiant efficiency. Tumor signal (total radiant efficiency, all values × 1010) normalized to tumor size for DOL 2 (B) and DOL 4 (C) (all paired columns are statistically significant differences using Student’s t-test (p ≤ 0.001)).

We then carried out a dose lowering study with the goal of defining an optimal dose for FNIR-Tag-mAb conjugates. We chose to examine three doses, 10, 5, and 1 μg, which span the lower end of the doses used in prior studies. We observed clearly discernible signal using FNIR-Tag-mAb conjugates throughout this range, while IR-800CW-mAb tumor signal could not be visualized below 5 μg (Figure S10). In quantifying these images, significantly higher fluorescence intensity was observed in nearly all studies (Figure A). Significantly improved TBR was observed for the 10, 5, and 1 μg doses of FNIR-Tag conjugates relative to conjugates IR-800CW starting on day 2 with values reaching 5.78 at day 7 with a 5 μg dose (Figure B). Finally, dramatically higher liver to background ratios were observed for the 10 and 5 μg doses of IR-800CW conjugates at 4 h and 1 day time points (Figure C). The observation that FNIR-Tag antibody conjugates are less subject to hepatobiliary uptake is likely to be significant utility, particularly in instances that seek to visualize biological processes in this region.

Figure 4

Tumor signal (total radiant efficiency, all values × 109) normalized to tumor size (A), tumor to background ratio (B), and liver to background ratio (C) of MDA-MB-468 tumor-bearing mice injected with 10, 5, and 1 μg of IR-800CW and FNIR-Tag DOL 4 conjugates of panitumumab (statistical analysis was performed between groups at the same time point using Student’s t-test; *p ≤ 0.5, **p ≤ 0.01, ***p ≤ 0.001).

Synthesis and Evaluation of Labeled VLPs

We extended our evaluation of FNIR-Tag to conjugates of VLPs derived from the bacteriophage Qβ. The Qβ VLP has been explored extensively for applications in vaccinology and drug delivery,[30−33] making it an ideal nanoparticle system for preliminary in vivo imaging studies. The multivalent display of lysine residues on the VLP surface[34] allows labeling to be compared at a higher density than possible using mAbs. Similar to the panitumumab conjugates, VLPs labeled with FNIR-Tag show superior brightness and stability when compared with IR-800CW conjugates.

Short azido linkers were installed on the VLPs using standard NHS coupling to make the particles amenable to modification by copper-catalyzed azide–alkyne cycloaddition (CuAAC) chemistry (see Methods in the Supporting Information for details). Conjugation of FNIR-Tag-alkyne with the Qβ-N3 VLPs, which were prepared as previously described,[34] was carried out via the CuAAC reaction in 100 mM potassium phosphate buffer (pH 7.4) at 50 °C for 3 h with molar equivalents of 0.5, 1, and 2 (relative to azide) to provide conjugates with a staggered degree of labeling (Figure ; see Methods in the Supporting Information for detailed conjugation conditions). As higher degrees of modification might interfere with particle surface properties, we aimed for fewer than 100 modifications on each particle. Conjugates were prepared with IR-800CW (0.5, 1, and 2 mol eq) under similar conditions (Figure ). All samples were subsequently purified by centrifugal filtration. Conjugate purity and the absence of free dye were confirmed by absorbance measurements of the supernatant and retentate after each passage (data not shown). Interestingly, higher loadings were achieved with FNIR-Tag (23, 40, and 70 attached dyes) compared to IR-800CW (9, 20, and 33 attached dyes) under all conditions screened. Notably, reaction with 2.0 mol equiv of IR-800CW yielded significant aggregation of the final product, which was not observed under similar conditions with FNIR-Tag (Figure B). As a result, loadings higher than 33 ± 3 IR-800CW molecules could not be achieved under these conditions (Figure A). Labeling was quantified by relating absorbance measurements (at λabs) for either FNIR-Tag or IR-800CW with the total protein concentration.

Figure 5

Stability and photophysical properties of VLP-fluorophore conjugates. (A) Quantification of the number of attached fluorophores as a function of molar equivalents added in the CuAAC reaction for FNIR-Tag (blue) and IR-800CW (red). An asterisk (*) indicates conditions where significant particle aggregation was observed. (B) Representative image of the final products from the reactions using 2 mol equiv of fluorophore, demonstrating visible aggregates for particles labeled with IR-800CW (right) but none for those labeled with FNIR-Tag under similar conditions (left). (C and D) Absorbance spectra of VLP conjugates with FNIR-Tag (C) and IR-800CW (D) with increasing degrees of fluorophore modification. All samples were at 20 μg/mL total protein in 100 mM potassium phosphate buffer (pH 7.4).

Absorbance curves for the purified conjugates were obtained following dilution in 100 mM potassium phosphate buffer to yield solutions that were 20 μg/mL in protein. Similar to the antibody conjugates, a significant H-aggregate peak was observed in the absorbance spectrum for IR-800CW conjugates with as few as 20 attached dyes and was significantly worse at higher levels of modification (Figure D). No significant H-aggregate band was detected for any of the FNIR-Tag conjugates under these conditions (Figure C).

Comparison of FNIR-Tag and IR-800CW VLP Conjugates in Vitro and in Vivo

We then evaluated VLP-fluorophore conjugate performance both in vitro and in vivo. To compare the brightness of the conjugates, C166 mouse endothelial cells were incubated with 20 nM VLP at 37 °C for 1 h. Samples were extensively washed to remove VLPs not bound to the cellular surface and subsequently fixed to preserve any interactions prior to analysis by flow cytometry. FNIR-Tag conjugates were significantly brighter than IR-800CW conjugates at similar labeling densities (Figure A and B), assuming that cell binding and uptake of the particles is not influenced by the nature of the attached dye. This result is consistent with the data obtained for mAb conjugates (described above). Interestingly, increasing the labeling density of IR-800CW slightly diminished the fluorescence (Figure S11A), whereas the emission of FNIR-Tag conjugates increased with the subsequent addition of fluorophores (Figure S11B). These data further suggest that H-aggregation significantly influences the fluorescence emission of IR-800CW conjugates.

Figure 6

Comparison of VLP conjugate brightness by flow cytometry and whole animal imaging. (A and B) VLP uptake in mouse C166 endothelial cells as measured with conjugates of both FNIR-Tag (red) and IR-800CW (black) at low (A) and high (B) fluorophore loadings. (C) Representative images demonstrating detection of VLP conjugates in the liver at 2 h (left panel) and 24 h (right panel) postinjection. FNIR-Tag = 25 μg dose of Qβ-(FNIR-Tag)70; IR-800CW = 50 μg dose of Qβ-(IR-800CW)20. Scale bar represents epi-fluorescent in total radiant efficiency. (D) Quantitative analysis of the fluorescence intensity detected at 2 and 24 h postinjection. Data are mean value ± standard error. Statistical analysis was performed between groups at the same time point using Student’s t-test. *p ≤ 0.05; **p ≤ 0.01.

Finally, we conducted an initial comparative imaging study in mice to evaluate the performance of the VLP-fluorophore conjugates in vivo. Given the superior brightness of the FNIR-Tag panitumumab conjugates in vivo and the stability of Qβ-FNIR-Tag conjugates at high loadings, we speculated that it would be possible to administer a lower dose of VLPs and still achieve sensitive detection by virtue of the greater number of attached fluorophores. Qβ VLPs and other nanoparticle systems are well-known to be cleared from circulation through the reticuloendothelial system and begin to accumulate in the liver within the first hour of administration;[35] therefore, we sought to image accumulation of both FNIR-Tag and IR-800CW conjugates in the livers of naïve CD-1 mice (n = 3 per group) following intravenous administration. Mice were injected with either 50 μg of Qβ-(IR800CW)20 or 25 μg of Qβ-(FNIR-Tag)70, and fluorescence images were recorded at 2 and 24 h postinjection (Figure C). The attachment of 20 IR-800CW dyes to the Qβ particle was the maximum labeling density that could be achieved without significant sign of insoluble aggregate formation. The detectable fluorescence for the FNIR-Tag conjugates was significantly higher than that of the IR-800CW conjugates at both 2 and 24 h (Figure D), and the signal for the IR-800CW conjugates was not detectable at 24 h even when a high threshold was used. Although preliminary, these data suggest that FNIR-Tag allows for greater labeling densities, which subsequently enables the sensitive imaging of biotherapeutic molecules at much lower doses than can be achieved using IR-800CW.

Conclusion

We developed a novel NIR fluorophore, FNIR-Tag, that has excellent properties for biomolecule labeling and in vivo imaging applications. The net-neutral zwitterionic dye appears to circumvent the issue of H-aggregation upon biomolecule conjugation within a reasonable range of labeling density. Reduced aggregation dramatically increases NIR emission of the corresponding mAb conjugates relative to the extensively used persulfonated fluorophore IR-800CW. Antibody conjugates of FNIR-Tag exhibit superior tumor uptake, reduced liver uptake, and enhanced brightness when compared to a similarly labeled IR-800CW conjugate in an in vivo imaging study in mice bearing EGFR+ tumors. We also observed that modification with IR-800CW induced irreversible aggregation above a moderate degree of labeling. The advantages of the FNIR-Tag conjugates were also apparent from preliminary in vitro and in vivo studies, as they proved significantly brighter in cellular association studies, and their stability at high degrees of modification allowed for superior detection in mice at a lower overall dose compared to VLPs modified with IR-800CW. In total, these studies introduce a new solution to the long-standing problem of NIR fluorophore aggregation. These efforts also illustrate that the charge and placement of peripheral substituents can play a critical role in fluorophore function and provide promising fluorophores for a range of applications where high-density labeling is desirable.