Cyanine Masking: A Strategy to Test Functional Group Effects on Antibody Conjugate Targeting

Abstract

Conjugates of small molecules and antibodies are broadly employed diagnostic and therapeutic agents. Appending a small molecule to an antibody often significantly impacts the properties of the resulting conjugate. Here, we detail a systematic study investigating the effect of various functional groups on the properties of antibody–fluorophore conjugates. This was done through the preparation and analysis of a series of masked heptamethine cyanines (CyMasks)-bearing amides with varied functional groups. These were designed to exhibit a broad range of physical properties, and include hydrophobic (−NMe₂), pegylated (NH-PEG-8 or NH-PEG-24), cationic (NH-(CH₂)₂NMe₃⁺), anionic (NH-(CH₂)₂SO₃⁻), and zwitterionic (N-(CH₂)₂NMe₃⁺)-(CH₂)₃SO₃⁻) variants. The CyMask series was appended to monoclonal antibodies (mAbs) and analyzed for the effects on tumor targeting, clearance, and non-specific organ uptake. Among the series, zwitterionic and pegylated dye conjugates had the highest tumor-to-background ratio (TBR) and a low liver-to-background ratio. By contrast, the cationic and zwitterionic probes had high tumor signal and high TBR, although the latter also exhibited an elevated liver-to-background ratio (LBR). Overall, these studies provide a strategy to test the functional group effects and suggest that zwitterionic substituents possess an optimal combination of high tumor signal, TBR, and low LBR. These results suggest an appealing strategy to mask hydrophobic payloads, with the potential to improve the properties of bioconjugates in vivo.

Graphical Abstract

INTRODUCTION

Conjugates of monoclonal antibodies (mAbs) and small molecules are indispensable tools for diagnostic and therapeutic applications. While long used for various cellular imaging applications (i.e., immunolabeling), the past decade has seen significant progress in the translation of mAb conjugates into clinical settings. For example, conjugates of fluorescent molecules and mAbs have emerged as powerful tumor imaging agents for fluorescence-guided surgery.^1,2 In the context of therapeutics, antibody drug conjugates (ADCs) harness the specificity of antibodies to selectively deliver potent payload molecules.³ Despite significant progress validating these approaches, significant challenges remain. In particular, many small molecule payloads interfere with the intrinsic targeting and clearance of the parent antibody.^4–6 Various strategies are being explored to circumvent this issue. These include homogenous labeling strategies, which have been explored extensively for ADCs. While such strategies improve certain in vivo properties, they increase the complexity of the labeling process.^7–9 For optical probes, the use of highly charged, persulfonated fluorophores is a long established strategy designed to address the issue of fluorophore aggregation on the protein surface.¹⁰ This strategy led to the creation of numerous broadly used fluorophores, including the “Cy” and “Alexa Fluor” series. While these probes are optimized for in vitro efforts, more recent results from our group and others have shown that the highly charged, but net neutral (i.e. zwitterionic) fluorophores can improve the targeting of mAbs in vitro and in vivo.^4,16

A critical, still outstanding question is what functional group can best mitigate the impact of small molecule labeling on mAb targeting. Here, we pursue a strategy centered on comparing antibody conjugates of a series of modified near-infrared (NIR) fluorescent dyes. While intrinsically hydrophobic, these molecules can be synthetically modified with polar functional groups. Several prior studies, including those from our group, have examined the impact of modifications on the in vivo properties of mAb–fluorophore conjugates.^11–21 While these prior studies provided key insights, they are not systematic from a “functional group” perspective making it hard to draw larger lessons (see Figure S1 for previous molecules). Here, we address this issue with single-point alterations to substituents distal to the cyanine scaffold. We hypothesize that this approach will provide insights that can broadly inform the design of various bioconjugates.

Below, we detail the synthesis and evaluation of a series of masked heptamethine cyanines (CyMasks) and their antibody conjugates. These probes enable the unbiased assessment of chemical masking strategies on tumor targeting and off-target accumulation of mAb conjugates. In line with prior results, we find that the most hydrophobic probe exhibits the most non-selective in vitro uptake and the lowest in vivo tumor targeting. We also find that anionic and cationic masking groups lead to increased dye aggregation and reduced tumor signals. Notably, pegylated probes, which have been broadly applied in other contexts,^22–27 exhibit excellent in vitro uptake but reduced in vivo tumor targeting. By contrast, zwitterionic probes, bearing both quaternary ammonium and sulfonate functional groups, exhibit excellent in vitro and in vivo properties. These results suggest that the combination of highly charged, but net neutral, substituents may offer significantly improved biophysical properties with implications to a range of bioconjugates.

RESULTS AND DISCUSSION

Synthesis and Characterization of CyMask and mAb Conjugates.

In designing the CyMask series, we sought to examine a range of polar substituents and chose representative zwitterionic, cationic, anionic, pegylated, and hydrophobic functional groups. Specifically, we synthesized six CyMask dyes containing amides substituted with PEG-8, PEG-24, N-ethyl-sulfonate (Sulfo), N-ethyl-trimethyl-ammonium (Quat), N-trimethyl-propyl-ammonium-N-butyl-sulfonate (Zwit), and a hydrophobic derivative, methylamine (Me). As shown in Scheme 1, the CyMask-NHS esters were synthesized from the known precursor (1) using our previously reported Smiles rearrangement method.¹¹ Compound 1 underwent substitution with N-methyl-ethanolamine in MeCN at 80 °C to obtain 2. The subsequent Smiles rearrangement (C4′-N to C4′-O) reaction was carried out using an initial TFA-induced rearrangement followed by electrophilic trapping in THF in the presence of NaHCO₃ at 50 °C to give methyl ester 3. This common precursor 3 was used to couple with primary amines 4a–f to give amides 5a–f, and subsequent saponification with LiOH provided the free carboxylic acids (6a–f). We compared the photophysical properties of all six CyMask carboxylic acids (6a–f) and found similar optical properties in PBS pH 7.4 and 10% FBS (Table 1 and Figures S2 and S3) across this series and relative to other C4′-O-linked heptamethine cyanines.^11–16 These observations confirm the minimal impact of the distal modification on the optical properties of the free molecules.

Scheme 1.

Synthesis of Compounds in CyMask Series and Conjugation with Panitumumab

Table 1.

Key Spectroscopic Properties

CyMask-CO2H	λmax,abs (nm)	λmax,em (nm)	ΦF	εa (M−1 cm−1)	brightness (ε × ΦF)	net charge	molecular weight
CyMask-Me-CO2H (6a)	766	781	0.14	161,000	22,500	0	779.0
CyMask-PEG-8-CO2H (6b)	766	781	0.13	203,000	26,400	0	1483.9
CyMask-PEG-24-COOH (6c)	766	781	0.15	182,300	27,500	0	2893.6
CyMask-Sulfo-CO2H (6d)	764	780	0.11	174,000	19,000	-2	965.4
CyMask-Quat-CO2H (6e)	764	780	0.12	184,000	22,000	+2	923.3
CyMask-Zwit-CO2H (6f)	764	780	0.11	184,000	20,240	0	1165.5

^aMolar absorption coefficients of all dyes were measured in 1:1 MeOH-PBS pH 7.4.

We then examined the impact of the masking functional group on the properties of mAb conjugates. The carboxylic acids 6a–f (CyMask-CO₂H) were converted to CyMask-NHS esters (7a–f) using N,N,N′,N′-tetramethyl-O-(N-succinimidyl) uronium tetrafluoroborate (TSTU) and were used in the bioconjugation step. We choose to first generate conjugates with panitumumab (Pan), an anti-EGFR FDA-approved monoclonal antibody, which has been used extensively for imaging applications.^2,11–16 All six NHS esters were reacted with panitumumab (Pan) in 50 mM PBS at pH 7.4 with molar excesses of 3, 6, and 10 equiv to obtain the lysine-labeled Pan conjugates with degrees of labeling (DOL) of 1, 2, and 3 (±0.3), respectively. Two purification methods were applied to the remove unlabeled probe. The dye conjugates were passed through a spin desalting column and stored overnight at 4 °C to dissociate non-covalently bound free dyes. The resulting solutions were purified a second time by passing through a desalting column eluting with pH 7.4 PBS. All conjugates provided homogenous solutions with the stable absorption spectra over a several-month timeframe.

The photophysical properties of these mAb-CyMask conjugates were analyzed in detail. The absorbance spectra were recorded in PBS pH 7.4 (Figure 1). While displaying nearly identical absorbance maxima, significant variation in the H-aggregate peak at 700 nm (Figure 1a insert) was observed across the various conjugates in PBS, but not in denaturing conditions (1:1 PBS:MeOH, Figure S5). This peak, which results from cyanine dimerization on the protein surface, is a well characterized consequence of protein labeling.^28–30 The magnitude of the H-aggregate was found to be dependent on both the probe and the DOL.^31–33 As a convenient means to quantify these effects, we examined the ratio of the absorbance values at 766 (monomer) and 700 (dimer) nm (Figure 1a–d). Across the series, the values decrease along series Sulfo > Me > Zwit > PEG-8 > Quat > PEG-24. Notably, PEG-24 provided almost complete inhibition of H-aggregate formation, as noted elsewhere.^23,34 The observation that sulfonated substituents led to the most substantial H-aggregate formation is somewhat surprising. Prior results clearly indicate that sulfonates appended to the indolenine heterocycles improve the properties of the resulting mAb conjugates.^29,30,35,36 However, these results suggest that distal sulfonates are not sufficient to inhibit H-aggregation, and may even induce formation. Lastly, we evaluated the quantum yield (Φ_F) of the mAb conjugates at DOL 3. Notably, the Φ_F of thee conjugates did not vary dramatically (from 4.8 to 7.5%) in pH 7.4 PBS (Figure 1e). Additionally, we observe little correlation between the Φ_F and either H-aggregation or the in vivo results described below. These observations suggest that the photon output of the fluorophore–mAb conjugates is relatively insensitive to the CyMask functional group, which simplifies the analysis of the in vivo data reported below.

Figure 1.

Characterization of Pan-CyMask conjugates. (a) Normalized absorption of Pan-CyMask conjugates (DOL 3, pH 7.4 PBS). A zoomed-in view is shown in the inset to highlight the monomer and H-aggregated peaks. (b) Normalized absorption of Pan-CyMask conjugates (DOL 2, pH 7.4 PBS). (c) Normalized absorption spectra of Pan-CyMask (DOL 1, pH 7.4 PBS). (d) Aggregation ratio (absorbance ratio at 700/766 nm) of Pan-CyMask conjugates at DOL 1, 2, and 3 (pH 7.4 PBS) calculated from the absorption spectra. (e) Φ_F (measured by a relative method) of Pan-CyMask conjugates (DOL 3, pH 7.2 PBS).

Comparison of Panitumumab-CyMask Conjugates.

We sought to characterize the impact of the CyMask modifications on tumor targeting and biodistribution of these agents. These studies were carried out using athymic nude female mice bearing EGFR+ MDA-MB-468 xenografts implanted subcutaneously in the right flank (5 × 10⁶ cells, n = 5 per group).¹⁶ The mice were injected with 50 μg of the Pan-CyMask conjugates (DOL 3). Fluorescence images were obtained using an IVIS imaging system before injection and at 4, 24, 48, 72, and 168 h p.i. As early as 4 h, all the conjugates showed a significant tumor signal, which increased over the next 7 days (Figure 2a and Figures S6 and S7). After 7 days, the highest radiance fluorescence intensity was observed with the Zwit and Quat conjugates, with values that are ~5.5 and 5.0-fold higher compared to the Me conjugate, respectively. Analysis of the tumor-to-background ratio (TBR) indicates a rank order of Zwit ≈ Quat > PEG-24 > Sulfo ≈ PEG-8 ≈ Me 1 week p.i. (Figure 2b,c). While the fluorescence signal in most healthy organs remain low, significant background uptake was observed in liver, specifically for cationic and anionic dye conjugates, with the strongest signal at the 4 h time point (Figure 2d and Figures S6 and S7). We note that for PanCyMask-PEG-8 and -PEG-24, we observe an apparent bladder signal at 4 h p.i. (Figure S6). This signal may reflect a small amount of non-covalently bound dye that remained following purification or potentially some other degradation pathway. Of note, the in vivo imaging data were confirmed in a subsequent ex vivo study using Pan-CyMask-Zwit and Pan-CyMask-Me conjugates, where mice (n = 3 per group) were analyzed ex vivo at 48 h p.i. (Figure S8). Ex vivo histopathological analysis of tumor slices injected with Pan-CyMask-Zwit and Pan-CyMask-Me showed a heterogenous near-infrared signal throughout the tumor with the highest signal around vasculature (Figure 2e and Figure S9). In the tissue samples, we observed approximately twice the number of clearly labeled cells with Pan-CyMask-Zwit (~20–30%) as compared to Pan-CyMask-Me (~10–20%). Overall, these results indicate that the zwitterionic conjugates outperform the charged conjugates (Sulfo and Quat) with respect to liver uptake. Additionally, the zwitterionic conjugates improved mAb tumor uptake relative to hydrophobic, anionic, and pegylated conjugates.

Figure 2.

In vivo and ex vivo fluorescence imaging of MDA-MB-468 xenograft tumor-bearing mice intravenously injected with Pan-CyMask conjugates (50 μg, DOL 3). (a) In vivo fluorescence images (right dorsum) at pre-injection and 4 h, 2 days, and 7 days post injection (p.i.). (b) Tumor signal (total radiant efficiency normalized to tumor size) after 7 days, (c) the tumor-to-background ratio after 7 days, and (d) the liver-to-background ratio after 4 h. Statistical analysis was performed using the Student’s t test. * p-value ≤ 0.05, *** p-value ≤ 0.001. (e) Representative histopathological images of tumor tissue sections (48 h, post injection) with Pan-CyMask-Me and Pan-CyMask-Zwit. The image is pseudo-colored: pink (near-infrared channel; Cy7), blue (DAPI), and green (autofluorescence).

We then studied the cellular uptake of these fluorophore conjugates in EGFR+ (MDA-MB-468) or EGFR– (MCF-7) cell lines. Epifluorescence microscopy demonstrated the high specific uptake of the CyMask conjugatesin MDA-MB-468 cells compared to the control cell line MCF-7 when incubated with the same concentration of Pan-CyMask conjugates and incubation time (250 μg/mL, 7 h) (Figure 3a). This approach indicates similar cellular uptake of all six Pan-CyMask conjugates. We quantified these differences in more detail by determining an apparent binding constant (EC₅₀) using flow cytometry for three of the conjugates. The Pan-CyMask-PEG-24 conjugate showed slightly lower binding affinity (EC₅₀ = 7.6 nM) compared to the Pan-CyMask-Zwit (EC₅₀ = 4.5 nM) and Pan-CyMask-Me (EC₅₀ = 6.9 nM) (Figure 3b). Overall, these data studies suggest modest differences in in vitro cellular uptake between the CyMask series and suggest that in vivo differences are largely due to altered clearance.

Figure 3.

In vitro studies of Pan-CyMask conjugates in MDA-MB-468 (EGFR+), MCF-7 (EGFR−), and Kupffer cells. (a) Epifluorescence imaging of MDA-MB-468 (top row) cells and MCF-7 (bottom row) after incubation with Pan-CyMask conjugates (250 μg/mL, 7 h, 37 °C). (b) Binding affinity of Pan-CyMask conjugates in MDA-MB-468 cells. Cells were incubated with Pan-CyMask conjugates (0.1–0.3 μM) at an ice bath (4 °C) for 1 h in pH 7.4 PBS (with 1 mg/mL BSA containing 0.01% sodium azide) and fixed using paraformaldehyde. Mean fluorescence intensity (MFI ± SD, n = 3) was measured by flow cytometry. An amount of ~10,000 cells were counted. (c) Non-specific uptake of mAb conjugates in Kupffer cells. Cells were incubated with Pan-CyMask conjugates (0.1 mg/mL) at 37 ° C for 24 h in serum-containing media; MFI ± SD (n = 3) was measured by flow cytometry. An amount of ~10,000 cells were counted.

Next, we investigated a potential hepatic clearance pathway using an in vitro cellular model. Prior work examining the uptake of ADCs and other conjugates has revealed a critical role for stellate macrophages, or Kupffer cells, in mediating the uptake and clearance of bioconjugates labeled with hydrophobic payloads.^37,38 To study the effect of the masking group on dye conjugates with respect to the non-specific uptake, we examined the uptake of these dye conjugates in rat Kupffer cells using flow cytometry.³⁸ After incubation of Pan-CyMask conjugates with rat Kupffer cells for 24 h, the most hydrophobic dye conjugate, Pan-CyMask-Me, exhibited increased uptake compared to other conjugates (Figure 3c). Notably, irrespective of the nature of the masking group, more hydrophilic dye conjugates showed similar non-specific uptake by Kupffer cells. These results are in line with the prior observations that suggest a role for Kupffer cells in the uptake of hydrophobic small-molecule–mAb conjugates but suggest that additional mechanisms may be at play for the high uptake of the positively charged Pan-CyMask-Quat conjugate in vivo.

Comparison of Anti-CD-276 mAb-CyMask Conjugates.

To test the generality of these observations, we applied CyMask with a fully human anti-CD276 (B7 homolog H3, B7H3) antibody that binds both mouse CD276 and human CD276 with similar affinity.^39,40 This target is overexpressed in the cancer cells and tumor neovasculature of multiple solid tumor types, and this antibody has been applied to create potent ADCs.^39–44 For the in vivo studies, we chose to compare the effects of positive charge (Quat), no charge (Me), and net neutral charge (Zwit) CyMask on tumor targeting and clearance. The conjugates were injected intravenously into mice bearing 200–250 mm³ JIMT-1 breast tumors grown orthotopically in the mammary fat pad. The tumor fluorescence intensity and TBR was highest for m276-SL-CyMask-Zwit (10.2) after 7 days (Figure 4a,b and Figures S10 and S11). These studies provide additional evidence that zwitterionic masking can improve the properties of mAb conjugates.

Figure 4.

(a) Quantification of tumor uptake of m276-SL-CyMask-Me, Quat, and Zwit conjugates (50 μg; DOL 3) in JIMT-1 tumors after 168 h. (b) Tumor-to-Background ratio at 168 h post injection. (c) IVIS images at 4 h, 2 days, and 7 days post injection. Tumors are highlighted in red dotted circles. Statistical analysis was performed between groups at 168 h. Data points are displayed as mean ± SD, and the p-values were evaluated by the Student’s t test. * p-value ≤ 0.05, ** p-value ≤ 0.01, *** p-value ≤ 0.001.

CONCLUSIONS

In the studies outlined above, we detail an unbiased comparison of charge masking functional groups on the in vivo targeting properties of mAb–fluorophore conjugates. These studies reveal that zwitterionic and cationic shielding outperforms pegylated, anionic, and hydrophobic conjugates with respect to tumor uptake. Furthermore, cationic shielding leads to the highest off-target liver uptake at early imaging time points. Notably, significant in vitro non-specific uptake by Kupffer cells occurs with the hydrophobic derivative but not the charged or pegylated conjugates. In total, these results suggest that the zwitterionic shielding groups may offer the optimal combination of excellent targeting and minimal off-target uptake. These results align with prior studies from our group and others, which found that zwitterionic heptamethine cyanines exhibit improved tumor uptake when compared to conventional persulfonated heptamethine cyanines.^16,17,45

These observations complement prior work to optimize the in vivo properties of therapeutic protein and nano-material modalities. In particular, extensive efforts have investigated the impact of mAb sequence modifications on clearance and tumor targeting. The general theme to emerge from these efforts is that cationization and anionization, the introduction of positively and negatively charged amino acids, respectively, both increase the plasma clearance rate, albeit through different mechanisms.⁴⁶ Cationization generally increases tissue uptake, which can be advantageous for solid tumor uptake, but also leads to increased uptake in off-target organs.⁴⁷ By contrast, anionization decreases tissue uptake due to increased whole-body (including blood) clearance through enhanced hepatic clearance.⁴⁶ In general, these studies have settled on the notion that the isoelectric point (i.e., the pH at which the protein is net neutral) of native antibodies, which is typically between 7 and 8, is preferred. While efforts to chemically introduce charged functional groups, as we have done here, are rare, extensive efforts have examined the impact of pegylation strategies.^{23,34,47–51} While offering significant benefits with respect to plasma half-life, this approach increases the size of protein significantly with consequent impacts on binding interactions.^52–55 Additionally, particularly when appended to long half-life proteins, such as mAbs, there is a potential for antibody-mediated immune responses.^56–58

Complementing efforts to augment the properties of proteins, the nanomaterials community has extensively studied the impact of charge and polarity as passivation strategies to improve the properties of otherwise hydrophobic species. In this context, there is significant evidence that zwitterionic functional groups reduce non-specific adsorption at the particle/liquid interface.^37,59 The resulting decrease in non-specific interactions leads to reduced off-target uptake in various contexts. While, the exact mechanism of this effect is not entirely clear, previous studies have suggested a combination of charge shielding, minimization of membrane penetration, and protein stabilization.⁶⁰

Here we introduce a series of optical probes that enable the quantitative assessment of functional group effects on mAb targeting in vivo. These results suggest that zwitterionic functional groups, which are highly charged, but net neutral modifications, may represent an important addition to the mAb-bioconjugate landscape. This strategy, which maintains the original charge profile of the parent protein, while simultaneously introducing a highly charged functional group, may represent a promising approach to mask the undesirable hydrophobicity of drug conjugates in the circulation. Future studies will investigate the potential impact of these modifications in homogeneous mAb conjugates, where we hypothesize that charge will also play a significant role.^61,62 Additionally, we suggest that the incorporation of zwitterionic functional groups into otherwise hydrophobic therapeutic payloads may have significant promise. Studies toward these goals are underway in our group and will be reported in due course.