Preferential Light-Chain Labeling of Native Monoclonal Antibodies Improves the Properties of Fluorophore Conjugates

Abstract

Site specific labeling methods have significant potential to enhance the properties of antibody conjugates. While studied extensively in the context of antibody-drug conjugates (ADCs), few studies have examined the impact of homogenous labeling on the properties of antibody-fluorophore conjugates (AFCs). We report the application of pentafluorophenyl (PFP) esters, which had previously been shown to be reasonably selective for K188 of the kappa light chain of human IGG antibodies, toward producing AFCs. We show that simple replacement of N-hydroxy succinimide (NHS) with PFP dramatically increases the light-chain specificity of near-infrared (NIR) AFCs. Comparing the properties of AFCs labeled using NHS and PFP-activated esters reveals that the latter exhibits reduced aggregation and improved brightness, both in vitro and in vivo. Overall, the use of PFP esters provides a remarkably simple approach to provide selectively labeled antibodies with improved properties.

Graphical Abstract

Introduction

Antibody-drug conjugates (ADCs) are an emerging approach for cancer therapy that combine the potency of small molecule anticancer drugs with the ability to target tumor-associated antigens.^1–3 Related antibody-fluorophore conjugates (AFCs) enable the optical detection of target-expressing cells, with clinical applications in the context of fluorescence-guided surgery (FGS).^{4, 5} Numerous AFC constructs are currently being clinically evaluated for FGS, with most consisting of clinically approved therapeutic monoclonal antibodies (mAbs) and the heptamethine cyanine IR-800CW.⁵ By far the most common way to prepare ADCs and AFCs is through amide formation between a lysine residue on the antibody and an active ester (i.e. NHS ester) of the payload.^6–11 Due to the abundance of lysine residues (>50 in a typical IgG1) and the high reactivity of NHS esters, heterogeneous mixtures are obtained. For ADCs, this heterogeneity is, at a minimum, an issue for batch-to-batch reproducibility and, at worst, can reduce the therapeutic index.¹² For AFCs, the impact of homogenous labeling has not been studied extensively, though deleterious effects of over-labeled conjugates have been proposed.¹³

To circumvent these issues, a number of approaches have been developed to produce increasingly homogeneous small molecule-antibody conjugates through site-specific labeling chemistry.^{2, 14–16} The most well-studied is the Thiomab approach developed by Genentech, in which a cysteine residue is installed on the protein and labeled using maleimide or related thiol-reactive chemistry. In early studies, this approach was shown to improve the therapeutic index of an anti-MUC16-MMAE ADC.¹⁷ Depending on the site in which the engineered cysteine was introduced, this approach can improve the in vivo pharmacokinetics and metabolic stability of ADCs.^{18, 19} On the other hand, the effect of labeling chemistry on the properties of AFCs is much less studied, and we are only aware of three prior examples of homogenously-produced protein-NIR fluorophore conjugates.^20–22

Of particular interest are site-selective labeling methods that can be used on native mAbs. Such approaches benefit from employing broadly available fully humanized mAbs, which have been the basis of numerous FDA-approved therapuetics.²³ A unique approach to site-specific labeling of native mAbs is the observation that a single lysine residue (K188) in the kappa (κ) light chain constant domain of human mAbs could be preferentially labeled (w/ ~70% selectivity) by employing pentafluorophenyl (PFP) esters as the active ester species for the conjugation.²⁴ Building on this work in the patent literature, Pham and co-workers studied the mechanism of this selectivity in greater detail. They found that highly homogeneous (>95% selectivity) Fab fragment conjugates can be obtained with preferential K188 labeling by optimization of the reaction conditions (4 °C or flow chemistry) or in combination with specific point mutations in the κ light chain region.²⁵ This preferential light-chain labeling strategy, although demonstrating only ~70% selectivity for native intact mAbs, represented a straightforward way for us to study the effect of labeling chemistry/site on the properties of AFCs.

Herein we report our initial findings applying this labeling strategy to a heptamethine cyanine fluorophore. Labeling the anti-EGFR mAb Panitumumab with an NHS ester leads to AFCs that are preferentially labeled on the heavy chain and exhibit significant fluorescence quenching. On the other hand, using a PFP ester provides conjugates with preferential light chain labeling. These selectively labeled bioconjugates are significantly brighter and exhibit higher tumor signals when administered intravenously to tumor-bearing mice. Overall, this approach, which is notably straightforward, demonstrates that labeling chemistry can significantly impact the properties of AFCs.

Our studies focused on the model C4’-O-alkyl heptamethine cyanine fluorophore diSulfo-FNIR, which we prepared from C4’-amine-substituted 1 using our previously reported electrophile-integrating Smiles rearrangement chemistry (Scheme 1).²⁶ This chemistry enables the facile preparation of C4’-O-alkyl heptamethine cyanines, which have improved stability to the presence of biological thiols compared to C4’-O-aryl substituted derivatives.^{26, 27} We have used this chemistry over the last several years to study the effect of substituents and charge balance on the properties of heptamethine cyanines in the context of AFCs, which culminated in the development of the efficient mAb label, FNIR-Tag.^28–31 In practice, diSulfo-FNIR could be obtained from 1 and glutaric anhydride, as we previously reported (Scheme 1).³² DiSulfo-FNIR could be converted to the corresponding NHS-ester, diSulfo-FNIR-NHS using TSTU. Alternatively, the corresponding PFP ester, diSulfo-FNIR-PFP, could be accessed using the commercially available bis(pentafluorophenyl) carbonate and N-methylmorpholine.

Scheme 1.

Synthesis of N-hydroxysuccinimdyl (NHS) and pentafluorophenyl (PFP) reactive esters of diSulfo-FNIR.

We then examined a variety of labeling conditions using diSulfo-FNIR-PFP. Our initial screening conditions used Panitumumab at a concentration of 7 μM in different buffers (PBS or HEPES), with a 10-fold molar excess of diSulfo-FNIR-PFP at 4 °C for 18 h (Table 1). We experimented with the addition of an organic cosolvent (up to 20%). Through this screening, we found that using pH 7.0 PBS rather than typically employed pH 8.0 PBS buffer produced conjugates with improved light chain to heavy chain (LC:HC) ratios as determined by reducing SDS-PAGE. We also found PBS with 10% DMF was preferable to HEPES buffer. Although these conditions produced a favorable LC:HC ratio, the degree of labeling (DOL) was lower than expected (~ 1.1:1). Therefore, we examined sequential addition of the active pentafluorophenyl ester (2x) separated by 3 hours, which improved the DOL considerably to a value of 1.7, while retaining the same LC:HC selectivity (1.7:1). This conjugate, PFP-FNIR-Pan, was prepared on up to a 2 mg scale and was purified using centrifugal filtration followed by dialysis. Of note, to date, mass spectral analysis of the conjugates modified with this charged fluorophore has been quite difficult, and we have not been able to analyze them to a finer level of molecular detail. However, we hypothesize that the selective light chain labeling we observe likely reflects at least preferential formation of K188 conjugates, as reported previously.²⁵

Table 1.

Reaction conditions screening for selective light chain labeling.

Entry	R	Buffer/Co-Solvent	Conc.	T	Time	LC:HC
1	PFP	40 mM PBS (pH 8)/20% DMF	7 μM	4 °C	16 h	0.7
2	PFP	40 mM PBS (pH 8)/5% DMF	7 μM	4 °C	16 h	0.8
3	PFP	40 mM HEPES (pH 8)/10% DMF	7 μM	4 °C	16 h	0.7
4	PFP	40 mM HEPES (pH 7)/10% DMF	7 μM	4 °C	16 h	1.1
5	PFP	40 mM PBS (pH 8)/10% DMF	7 μM	4 °C	16 h	0.8
6	PFP	40 mM PBS (pH 7)/10% DMF	7 μM	4 °C	16 h	1.7
7	PFP	40 mM HEPES (pH 8)/1% DMF	7 μM	4 °C	16 h	0.5
8	PFP	40 mM HEPES (pH 7)/1% DMF	7 μM	4 °C	16 h	0.8
9	PFP	40 mM PBS (pH 8)/1% DMF	7 μM	4 °C	16 h	0.6
10	PFP	40 mM PBS (pH 7)/1% DMF	7 μM	4 °C	16 h	0.7
11	NHS	1M PBS (pH 8.5)	500 μM	r.t.	1 h	0.1

In order to evaluate the properties of labeled PFP-FNIR-Pan, we prepared the corresponding conjugate with diSulfo-FNIR-NHS, NHS-FNIR-Pan, with a matching DOL of 1.7. In this case, the LC:HC ratio is 0.1 (Fig. 2) – highlighting the dramatic effect of the PFP labeling strategy. Of note, using the optimized conditions for PFP-FNIR-Pan with diSulfo-FNIR-NHS, resulted in only a modest improvement in the LC:HC selectivity from 0.1 to 0.2 (Fig. S1). This result demonstrates that the PFP ester, not the labeling conditions, are responsible for the enhanced light chain selectivity.

Figure 2.

a) General scheme for labeling conditions screening using the model fluorophore diSulfo-FNIR. b) Gel electrophoresis (Cy7 excitation) of diSulfo-FNIR labeled light and heavy chain fragments after SDS-PAGE (reducing conditions). The light chain: heavy chain ratio (LC:HC) was determined by dividing the quantified signals from the two fragments.

The absorbance and emission spectra of the two conjugates were obtained in pH 7.4 PBS (50 mM) at a 500 nM mAb concentration (Fig. 3A). As expected, with NHS-FNIR-Pan the absorbance spectra contained a blue-shifted band arising from the nonemissive H-aggregate dimer species.^{33, 34} By contrast, the absorbance spectrum of PFP-FNIR-Pan more closely resembles that of the free dye (Fig S2). The dramatic difference between the optical properties of the conjugates is also readily observed by the difference in emission intensity. When solutions with equal mAb concentration (500 nM) were excited at 730 nm, PFP-FNIR-Pan was ~3x brighter than the NHS-FNIR-Pan conjugate. In line with previous observations, dilution of the same conjugates into MeOH:PBS (protein denaturing conditions) restores the emission intensity of the non-selectively labeled control, albeit not completely (Fig. S3). We then examined the two conjugates by flow cytometry to determine the impact of labeling method on cellular uptake. When incubating PFP-FNIR-Pan or NHS-FNIR-Pan with MDA-MB-468 cells (EGFR+), we observed the PFP-labeled mAbs was significantly brighter as measured by the mean fluorescence intensity of the labeled cells (Fig. 3B). By contrast, no significant uptake of either of the conjugates was observed in MCF-7 cells (EGFR–).

Figure 3.

a) Absorbance and emission spectra of NHS-FNIR-Pan, black trace and PFP-FNIR-Pan, red trace with a DOL of 1.7 and a mAb concentration of 500 nM in PBS (50 mM. pH 7.4). b) Flow cytometric analysis of DOL 1.7 PFP-FNIR-Pan and PFP-FNIR-Pan in MCF-7 (EGFR –) and MDA-MB-468 (EGFR+) cells.

Finally, we examined the difference between these two conjugates in the context of in vivo tumor imaging. Prior studies with AFCs have demonstrated that even moderate changes to the fluorophore label density has an impact on the clearance of AFCs.²⁹ We have also found that AFC aggregation (due to high label density) and poor pharmacokinetics are related, often resulting in increased liver uptake.^28,33 Equal doses of NHS-FNIR-Pan and PFP-FNIR-Pan (10 ug, DOL 1.7) were administered intravenously in MDA-MB-468 xenograft (n = 5 per group) bearing mice. The PFP-FNIR-Pan exhibited significantly improved tumor uptake relative to NHS-FNIR-Pan at all timepoints and significantly higher tumor to background ratios (TBR) at 48 h and 1 week post injection (Fig. 4).

Figure 4.

a) Representative side and ventral view images of MDA-MB-468 xenograft tumor-bearing mice (n = 5) that received either NHS-FNIR-Pan or PFP-FNIR-Pan, mice were dosed by tail vein injection (10 ug Ab, DOL 1.7, 1.2 nmol effective dye dose). b) Quantification of Tumor Signal and Tumor to background ratios (TBR) of the two conjugates at the timepoints indicated. The tumor signal values, as determined by a region of interest (ROI) were normalized to tumor size for each mouse. Data points are displayed as mean ± SD, and the p-values were evaluated by student t-test (*** p-value ≤ 0.001).

In this study, we have employed a selective mAb labeling technique to produce AFCs with enriched labeling in the LC constant domain of a human mAb (Panitumumab). The selectivity of the LC chain labeling observed with a charged fluorophore was similar to early reports that mainly employed hydrophobic, uncharged payloads.²⁵ Notably, the selectively labeled conjugates showed no signs of aggregation in their optical spectra, retained fluorescence emission similar to that of the unconjugated dye, and showed improved uptake in MDA-MB-468 cells relative to the random lysine-labeled conjugate using diSulfo-FNIR-NHS. Additionally, PFP-FNIR-Pan exhibited significantly higher tumor signal than NHS-FNIR-Pan in an MDA-MB-468 xenograft mouse model.

We hope these studies highlight the potential for this simple PFP-based method to provide antibody conjugates of improved homogeneity. Additionally, this work provides further insight into the relationship between fluorophore labeling site and AFC properties. These studies indicate more homogeneous conjugates reduce the propensity for fluorophore H-aggregates to form. This property, in turn, leads to improved optical properties and tumor signal. In combination with prior work, we and others have now shown that the properties of AFCs can be improved by either homogenous labeling or by the use of highly polar, but net-neutral (i.e. zwitterionic), fluorophores.^28,33 Notably, however, the latter benefits from achieving higher DOL. Going forward, we anticipate that, much as the properties of ADCs have been improved through the use homogenous protein labeling methods, the same will be true for AFCs.

Figure 1.

a) Structure of model heptamethine cyanine fluorophore diSulfo-FNIR and the activated esters used in this study. b) General schematic for Panitumumab labeling with diSulfo-FNIR activated esters to produce either random lysine labeled (NHS) or selective light chain labeling (PFP).

Highlights.

Pentafluorophenyl esters of cyanine fluorophores can be readily prepared

These esters selectively label the light chain of native monoclonal antibodies

The resulting conjugates are brighter in vitro and in vivo