Lysine-Directed Site-Selective Bioconjugation for the Creation of Radioimmunoconjugates

Abstract

The synthesis of radioimmunoconjugates via the stochastic attachment of bifunctional chelators to lysines can yield heterogeneous products with suboptimal in vitro and in vivo behavior. In response to this, several site-selective approaches to bioconjugation have been developed, yet each has intrinsic drawbacks, such as the need for expensive reagents or the complexity of incorporating unnatural amino acids into IgGs. Herein, we describe the use of a simple and facile approach to lysine-directed site-selective bioconjugation for the generation of radioimmunoconjugates. This strategy relies upon on the selective modification of single lysine residues within each light chain of the monoclonal antibody (mAb) with a branched azide-bearing perfluorophenyl ester (PFP-bisN₃) followed by the ligation of dibenzocyclooctyne (DBCO)-bearing payloads to these bioorthogonal handles via the strain-promoted azide-alkyne cycloaddition. This methodology was used to create [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, a radioimmunoconjugate of the HER2-targeting mAb pertuzumab labeled with desferrioxamine (DFO) and the positron-emitting radiometal zirconium-89 (⁸⁹Zr). [⁸⁹Zr]Zr-^SSKDFO-pertuzumab was compared to a pair of analogous probes: one synthesized via random lysine modification ([⁸⁹Zr]Zr-DFO-pertuzumab) and another via thiol-maleimide chemistry ([⁸⁹Zr]Zr-^malDFO-pertuzumab). The bioconjugation strategy was assessed using ESI mass spectrometry, SDS-PAGE, and autoradiography. All three immunoconjugates demonstrated comparable binding to HER2 via flow cytometry and surface plasmon resonance (SPR), and ⁸⁹Zr-labeled variants of each were synthesized in >99% radiochemical yield and molar activities of up to ~55.5 GBq/μmol (10 mCi/mg). Subsequently, the in vivo behavior of this trio of ⁸⁹Zr-immunoPET probes was interrogated in athymic nude mice bearing subcutaneous HER2-expressing BT-474 human breast cancer xenografts. [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, [⁸⁹Zr]Zr-^malDFO-pertuzumab, and [⁸⁹Zr]Zr-DFO-pertuzumab produced positron emission tomography (PET) images with high tumoral uptake and high tumor-to-healthy organ activity concentration ratios. A terminal biodistribution study complemented the PET results, revealing tumoral activity concentrations of 126.9 ± 50.3 %ID/g, 86.9 ± 53.2 %ID/g, and 92.5 ± 27.2 %ID/g at 144 h post-injection for [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, [⁸⁹Zr]Zr-^malDFO-pertuzumab, and [⁸⁹Zr]Zr-DFO-pertuzumab, respectively. Taken together, the data clearly illustrate that this highly modular and facile approach to site-selective bioconjugation produces radioimmunoconjugates that are better-defined and more homogeneous than stochastically modified constructs and also exhibit excellent in vitro and in vivo performance. Furthermore, we contend that this lysine-directed strategy holds several key advantages over extant approaches to site-selective bioconjugation, especially in the context of production for the clinic.

Graphical Abstract

INTRODUCTION

Over the last two decades, monoclonal antibodies (mAb) have become increasingly critical tools in oncology for the delivery of a wide array of payloads to malignant tissue.¹ Antibody-drug conjugates (ADCs) are without question the standard-bearers of this phenomenon, but mAbs have also been harnessed to carry fluorophores, photosensitizers, immunomodulatory factors, and (most relevant to the work at hand) radionuclides to tumors as well.^2-4 In nuclear medicine, the remarkable selectivity and affinity of mAbs for cancer antigens have been leveraged for both imaging — via positron emission tomography (PET) and single photon emission computed tomography (SPECT) — and radioimmunotherapy (RIT).^5,6 Practically speaking, the multi-day serum half-life of mAbs means that effective radioimmunoconjugates must contain radionuclides with complementary physical half-lives, most notably zirconium-89 (⁸⁹Zr; t_1/2 ~ 3.3 d) for PET; indium-111 (¹¹¹In; t_1/2 ~ 2.8 d) for SPECT; lutetium-177 (¹⁷⁷Lu; t_1/2 ~ 6.6 d), yttrium-90 (⁹⁰Y; t_1/2 ~ 2.7 d), and iodine-131 (¹³¹I; t_1/2 ~ 8.0 d) for β-particle RIT; and actinium-225 (²²⁵Ac; t_1/2 ~ 9.9 d) for α-particle RIT.⁷ A wide variety of radioimmunoconjugates have been translated to the clinic for immunoPET, immunoSPECT, and RIT, and recent years have played witness to several success stories. The ZEPHIR trial, for example, underscored the value of HER2-targeted immunoPET as a theranostic tool in conjunction with the ADC trastuzumab-emtansine.⁸ Furthermore, clinical studies suggest that PD-L1-targeted immunoPET could have an important role in identifying patients likely to respond to checkpoint blockade immunotherapy.⁹ Shifting gears to therapy, GD2-targeted RIT with [¹³¹I]I-3F8 has produced remarkable results in pediatric patients with CNS tumors, while several ²²⁵Ac-labeled mAbs have shown promise as therapeutics for hematological malignancies.^10,11

The intrinsic biological and structural complexity of mAbs suggest great care should be taken in the construction of radioimmunoconjugates. Yet the vast majority of these sophisticated tools remain synthesized in a remarkably unsophisticated way: the stochastic modification of lysines within the mAb.¹² This approach relies upon the reaction between deprotonated lysine residues and bifunctional chelators bearing amine-reactive moieties such as N-hydroxysuccinimidyl esters or phenyl isothiocyanates, and it is undeniably simple and facile. However, the presence of 40+ lysines distributed throughout the structure of the mAb means that this route produces a complex mixture of regioisomers that is heterogeneous with respect to both the number of payloads appended per mAb and the location of those payloads.¹² The indiscriminate nature of the strategy can also lead to the inadvertent attachment of payloads in the antigen-binding domains of the mAb, thereby impairing the immunoreactivity of the radioimmunoconjugate. Taken together, these challenges can produce radioimmunoconjugates with suboptimal in vitro and in vivo behavior.^12-14

A growing collection of site-selective bioconjugation strategies — the most selective of which have been termed ‘site-specific’ — have recently been developed to circumvent the problems associated with stochastic modification methods. The best-known approach relies upon the reduction of the interchain disulfide linkages within the mAb, thereby exposing eight free cysteine residues that can be modified with payloads bearing thiol-reactive groups such as maleimides.¹⁵ The heavy chain glycans — the biantennary sugar chains attached to the Fc region of the mAb — have also been exploited for site-selective bioconjugation via a chemoenzymatic approach based on the incorporation of azide-modified sugars into the glycans and their subsequent modification with DBCO-bearing payloads via the strain-promoted azide-alkyne cycloaddition (SPAAC) ligation.^16-18 Other approaches to site-selective bioconjugation have taken a more active role in creating orthogonally reactive sites, using genetic engineering to build mAbs containing unnatural amino acids or peptide tags with their own unique reactivities.^19-22 For example, several groups have harnessed sortase A for the enzyme-mediated bioconjugation of vectors bearing terminal LPXTG tags.^23-25 Regardless of the chemistry involved, site-selective (and site-specific) approaches to bioconjugation have repeatedly been shown to create radioimmunoconjugates that are better defined and more homogeneous than their traditionally synthesized cousins.^12,26 Even more importantly, site-selectively modified radioimmunoconjugates have also been shown to boast better in vivo performance than analogues created using stochastic methods.^14,16,27

Despite their clear advantages, each approach to precision bioconjugation comes with its own set of drawbacks. Thiol-mediated strategies, for example, (i) require the reduction of the mAb, (ii) can still produce mixtures of regioisomers based on the presence of as many as 8 free cysteines after reduction, and (iii) traditionally utilize the Michael addition between thiols and maleimides which is reversible in vivo. While the use of 'second-generation’ thiol-reactive synthons — e.g. phenyloxadiazolyl methyl sulfones, 'hydrolyzable’ maleimides, and dibromomaleimides — has lessened concerns surrounding the latter, the former pair of challenges has proven harder to solve.^28,29 Glycans-targeted methodologies avoid these thiol-related complications but require expensive enzymes and can alter the FcγR-binding properties of the radioimmunoconjugate.^16,30,31 Finally, methods that employ genetic engineering are promising but are inherently complex and lack modularity, particularly with respect to creating radioimmunoconjugates from existing mAbs. Clearly, the field is still looking for an approach to bioconjugation that combines the homogeneity of precision methods with the ease of stochastic modification.

Herein, we describe a novel lysine-directed approach to site-selective bioconjugation and its use for the synthesis of a ⁸⁹Zr-labeled radioimmunoconjugate. The ubiquity of lysines within IgG makes them a surprising target for site-selective bioconjugation, but the local environment of a residue’s primary amine can influence its reactivity dramatically and create an opportunity for selective modification. This bioconjugation strategy has two steps. First, the mAb is modified with a branched azide-bearing perfluorophenyl ester (PFP-bisN₃) that selectively reacts with a single lysine residue on the light chain of the mAb. Second, dibenzocyclooctyne (DBCO)-bearing payloads are attached to this azide-bearing immunoconjugate via the SPAAC reaction. As a proof-of-concept, this methodology was used to create [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, a radioimmunoconjugate of the HER2-targeting mAb pertuzumab labeled with desferrioxamine (DFO) and the positron-emitting radiometal zirconium-89 (⁸⁹Zr). [⁸⁹Zr]Zr-^SSKDFO-pertuzumab was compared to a pair of analogous probes — one synthesized via stochastic lysine modification ([⁸⁹Zr]Zr-DFO-pertuzumab) and another via thiol-maleimide chemistry ([⁸⁹Zr]Zr-^malDFO-pertuzumab) — using surface plasmon resonance spectroscopy, mass spectrometry, longitudinal stability assays, and bead-based immunoreactivity experiments. Subsequently, the in vivo behavior of this trio of ⁸⁹Zr-immunoPET probes was interrogated in athymic nude mice bearing subcutaneous HER2-expressing BT-474 human breast cancer xenografts and SKOV-3 human ovarian cancer xenografts.

Before we move on, it is important to note that we are not the first to target lysines for site-selective bioconjugation.³² Cao et al., for example, have developed a site-selective conjugation strategy dubbed ‘pClick’ that is based on the proximity-induced reactivity between a peptide cross-linker — FPheK — and lysine residues.³³ In addition, Hwang et al. have leveraged the enhanced reactivity of a single lysine within a humanized catalytic antibody — h38C2 — to create an approach to site-selective bioconjugation based on the ligation between this primary amine with (normally thiol-reactive) phenyloxadiazolyl methylsulfones.³⁴ Despite the undeniable elegance of both of these methods, the former depends upon the presence of a peptide linker, while the latter requires genetic engineering to create immunoglobulins containing the catalytic h38C2 domain.

More germane to the study at hand, Bhat et al. demonstrated that bifunctional probes bearing amine-reactive fluorophenolic esters preferentially react with the K188 residues of mAbs.³⁵ Critically, these residues lie within the constant (C_L) region of the kappa light chain rather than the variable (V_L) region, meaning that bioconjugation approaches using fluorophenolic esters are compatible with all κ IgG1 mAbs. This approach was further explored by Pham et al., who utilized selective mutagenesis to demonstrate that the selectivity for the solvent-exposed K188 was dependent on the neighboring His189 and Asp151 residues. The authors suggested two explanations for this phenomenon. First, they posited that D151 may activate its spatially-adjacent neighbor H189 to act as an acid-base catalyst in order to accelerate the reaction of fluorophenyl esters with the amine of K188. Second, the authors offered up the possibility that the acylation may initially occur at H189 followed by an acyl transfer to K188. Mechanistic details aside, the authors also illustrated that the selectivity of fluorophenolic esters for K188 could be significantly increased by lowering the reaction temperature or leveraging flow chemistry.³⁶ To wit, the authors determined that the K188-selectivity of a PFP-biotin reagent could be increased from ~70% to >95% by decreasing the reaction temperature from room temperature to 4 °C. Despite this promise, these initial investigations used a single-step approach that proved prone to variability stemming from changes to the payload. Our work builds upon this foundation but offers unparalleled facility and modularity by turning to a two-step procedure in which a single acylating agent is used for bioconjugation and bioorthogonal chemistry is then leveraged for the attachment of payloads. To the best of our knowledge, this work represents the first report of the use of lysine-directed site-selective bioconjugation for the synthesis of radioimmunoconjugates.

RESULTS

Model System

The proof-of-concept system for this investigation is centered on pertuzumab, a humanized mAb that targets the HER2 antigen frequently over-expressed by breast and gastric cancers among others.^37,38 We paired the antibody with ⁸⁹Zr — a positron-emitting radiometal whose half-life (t_1/2 ~ 3.3 d) dovetails well with the pharmacokinetic profile of full-length mAb — and its gold-standard chelator, desferrioxamine (DFO).^{39 89}Zr-DFO-labeled radioimmunoconjugates of pertuzumab are well characterized both preclinically and in the clinic, which allows us to concentrate on evaluating the bioconjugation strategy rather than the imaging agent itself.^16,40-42 Finally, two HER2-expressing human cells lines were used for the in vitro and in vivo characterization of the [⁸⁹Zr]Zr-DFO-pertuzumab radioimmunoconjugates: BT-474 human breast ductal carcinoma cells and SKOV-3 human ovarian epithelial adenocarcinoma cells.

Bioconjugation

The lysine-directed site-selective bioconjugation strategy requires a straightforward two-step procedure. First, pertuzumab was incubated with a branched, azide-bearing perfluorophenyl ester (PFP-bisN₃) to produce ^SSKN₃-pertuzumab (Figure 1; SSK = site-selective lysine). This azide-bearing intermediate was then modified with DBCO-DFO via the strain-promoted azide-alkyne cycloaddition click ligation to yield ^SSKDFO-pertuzumab. The entire two-step procedure was completed in 40 h (including two overnight incubations) and afforded the completed immunoconjugate in a cumulative yield of >75% after purification via ultrafiltration following each of the two reaction steps. Two additional immunoconjugates — one randomly modified, one site-selectively — were created for comparison. In the first, the lysines of the mAb were randomly modified with NHS-PEG₈-N₃ to produce an azide-bearing immunoconjugate that was subsequently reacted with DBCO-DFO to yield DFO-pertuzumab. For the second, the ‘gold-standard’ for thiol-mediated bioconjugations — the flawed but ubiquitous maleimide — was used to create another site-selectively modified immunoconjugate. To this end, the interchain disulfides of the mAb were reduced with tris(2-carboxyethyl)phosphine (TCEP) to expose free cysteine residues that were subsequently reacted with a maleimide-bearing variant of DFO (^malDFO) to produce ^malDFO-pertuzumab. After size exclusion chromatography, each of these immunoconjugates were afforded in >90% yield.

Figure 1.

(A) The lysine-directed site-selective bioconjugation strategy for ^SSKDFO-pertuzumab using PFP-bisN₃, a branched bifunctional synthon bearing two azides and an amine-reactive perfluorophenyl ester; (B) the structure of PFP-bisN₃; (C) schematics of the bioconjugation routes for DFO-pertuzumab and ^malDFO-pertuzumab.

Structural and Biological Characterization

The next step in the investigation was the structural characterization of the trio of immunoconjugates. The selectivity of the lysine-directed bioconjugation strategy for the light chain was interrogated via ESI-MS (Figure 2A). These data showed a single bioconjugation event on the light chain of ^SSKN₃-pertuzumab with nearly no detectable modification to the heavy chain, consistent with the selective modification of K188. The mass spectra of the analogous randomly labeled immunoconjugate (N₃-pertuzumab) clearly show modifications to both the heavy and light chains of the mAb with a significant preference for conjugation to the heavy chain. SDS-PAGE of the DFO-modified immunoconjugates reinforces this finding (Figure 2B). A pronounced shift can be observed in the molecular weight of the light chain of ^SSKDFO-pertuzumab relative to the parent antibody, again with minimal change to the heavy chain. In contrast, both DFO-pertuzumab and ^malDFO-pertuzumab display subtle shifts to the molecular weight of both the heavy and light chains. Next, MALDI-ToF mass spectrometry was employed to determine the degree-of-labeling (DOL) of each of the immunoconjugates, ultimately revealing similar DOLs for the three constructs: 3.9 ± 0.02 DFO/mAb for ^SSKDFO-pertuzumab, 3.9 ± 0.05 DFO/mAb for DFO-pertuzumab, and 4.3 ± 0.14 DFO/mAb for ^malDFO-pertuzumab (Table 1). As the number of cargoes in an immunoconjugate can influence its hydrophobicity and thus pharmacokinetic profile, the DOL of each of these immunoconjugates was kept the same.⁴³ Finally, the monomeric content and aggregation of DFO-pertuzumab, ^malDFO-pertuzumab, and ^SSKDFO-pertuzumab were interrogated using size exclusion HPLC shortly after purification as well as upon a short period of storage at 4 °C in PBS pH 7.4. The constructs demonstrated relatively low amounts of aggregation over a period of 14 days under these conditions (Figure S1). Yet only ^SSKDFO-pertuzumab reproducibly remained nearly completely in the monomeric form (>99%) upon synthesis and over this period of storage; both DFO-pertuzumab and ^malDFO-pertuzumab exhibited detectable aggregation yielding 93.5% and 96.0% of monomeric immunoconjugate, respectively, after 14 days.

Figure 2. The structural characterization of the immunoconjugates.

(A) ESI-MS of pertuzumab, ^SSKN₃-pertuzumab, and N₃-pertuzumab; (B) reducing SDS-PAGE of the immunoconjugates; HC = heavy chain and LC = light chain.

Table 1.

Conjugation and degree-of-labeling results for each of the pertuzumab immunoconjugates (n = 3).

Immunoconjugate	DFO/mAb	Conjugation Type	Conjugation Target
DFO-pertuzumab	3.9 ± 0.05	Random	Lysine residues
SSKDFO-pertuzumab	3.9 ± 0.02	Site-selective	K188
malDFO-pertuzumab	4.3 ± 0.14	Site-selective	Free cysteine residues

The biological evaluation of the immunoconjugates followed their structural characterization. Flow cytometry experiments illustrated that the trio of constructs bound HER2-expressing BT-474 cells in a manner seemingly identical to their parent antibody (Figure 3A). Flow cytometry was also used to determine their EC₅₀ values with BT-474 cells, yielding a tight range of values between 1.5-2.9 μg/mL (Figure S2 and Table S1). Surface plasmon resonance data reinforced these observations, yielding similar K_D values for pertuzumab (1.7 × 10⁻⁹ M), ^SSKDFO-pertuzumab (0.8 × 10⁻⁹ M), DFO-pertuzumab (1.7 × 10⁻⁹ M), and ^malDFO-pertuzumab (2.3 × 10⁻⁹ M) as well as similar kinetic parameters (Table 2 and Figure S3). Finally, ELISA analysis revealed that the three immunoconjugates display unperturbed FcγRI engagement relative to unmodified pertuzumab (Figure S4 and Table S2).

Figure 3. The biological characterization of the immunoconjugates.

(A) Fluorescence-associated cell sorting (FACS) of pertuzumab, DFO-pertuzumab, ^SSKDFO-pertuzumab, ^malDFO-pertuzumab, and non-specific hIgG₁ using HER2-expressing BT-474 cells and an AlexaFluor488-labeled secondary antibody (n = 3); (B) bead-based immunoreactivity of [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, and [⁸⁹Zr]Zr-^malDFO-pertuzumab using HER2-coated magnetic beads (n = 3). Statistical significance was determined via a two-tailed t test with a Welch’s correction using GraphPad Prism software. * = p-value < 0.05; ** = p-value < 0.01.

Table 2.

Kinetic and thermodynamic binding parameters for pertuzumab, DFO-pertuzumab, ^SSKDFO-pertuzumab, and ^malDFO-pertuzumab for HER2 measured via surface plasmon resonance.

Antibody	ka (M−1 s−1)	kd (s−1)	KD (M)
pertuzumab	1.46 ×105 (± 6.69 × 100)	2.51 ×10−4 (± 5.77 × 10−7)	1.72 ×10−9 (± 4.03 × 10−12)
DFO-pertuzumab	1.06 ×105 (± 1.31 × 101)	1.75 ×10−4 (± 6.31 × 10−7)	1.65 ×10−9 (± 6.15 × 10−12)
SSKDFO-pertuzumab	1.00 ×105 (± 2.6 × 100)	7.98 ×10−5 (± 4.87 × 10−6)	0.80 ×10−9 (± 4.85 × 10−11)
malDFO-pertuzumab	1.38 ×105 (± 1.77 × 101)	3.16 ×10−4 (± 5.17 × 10−7)	2.29 ×10−9 (± 4.03 × 10−12)

Radiopharmaceutical Chemistry and In Vivo Evaluation

The radiosyntheses of [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, and [⁸⁹Zr]Zr-^malDFO-pertuzumab proved to be a straightforward endeavor. The DFO-bearing constructs labeled with [⁸⁹Zr]Zr⁴⁺ were radiolabeled and purified using standard protocols, ultimately producing radioimmunoconjugates in high yield, high purity, and specific activities of 5-10 mCi/mg (Figure S5).⁴⁴ [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, and [⁸⁹Zr]Zr-^malDFO-pertuzumab were incubated in human serum at 37 °C for 5 days to investigate their serum stability. Longitudinal radio-thin layer chromatography and radio-size exclusion chromatography measurements revealed that the radioimmunoconjugates remained >95% stable to demetallation and aggregation over that time-period (Figure S6, Figure S7, and Table S3). The trio of radioimmunoconjuages was subsequently analyzed via radioactive SDS-PAGE and autoradiography (Figure S8). As expected, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab was radiolabeled selectively on the light chain with very minimal activity observed associated with the heavy chain. Interestingly, both [⁸⁹Zr]Zr-DFO-pertuzumab and [⁸⁹Zr]Zr-^malDFO-pertuzumab were radiolabeled almost exclusively on the heavy chain, even though the ESI-MS and the stained SDS-PAGE clearly showed bioconjugation on both the light and heavy chains (Figure 2). The most likely explanation for this phenomenon lies in the accessibility of the chelators within the immunoconjugates, with those appended to the heavy chains more accessible to [⁸⁹Zr]Zr⁴⁺ than those on the light chains. Next, a bead-based assay was performed to assess the immunoreactivity of the radioimmunoconjugates. To this end, each of the ⁸⁹Zr-labeled mAbs were incubated with magnetic beads coated with a recombinant His-tagged HER2 antigen. All three radioimmunoconjugates demonstrated immunoreactive fractions of >70% (Figure 3B), though both [⁸⁹Zr]Zr-^SSKDFO-pertuzumab (89 ± 1.5%) and [⁸⁹Zr]Zr-DFO-pertuzumab (87 ± 2.7%) boasted higher values than their maleimide-based cousin (72 ± 2.6%).

The final step in the development of these radioimmunoconjugates was their in vivo evaluation in murine models of HER2-expressing cancer. To this end, female nude athymic mice bearing subcutaneous BT-474 human breast cancer xenografts in the right shoulder were injected intravenously with ~100 μCi (5 mCi/mg, 20 μg) of either [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, or [⁸⁹Zr]Zr-^malDFO-pertuzumab, and scans were acquired every 24 hours for 6 days (Figure 4A). Each radioimmunoconjugate clearly delineated the HER2-expressing tumor tissue as early as 24 h post-injection. Moreover, the activity concentrations in the xenograft increased throughout the course of the experiment, with each cohort displaying >80 %ID/g in the tumor at 144 h p.i. Relatively little uptake was observed in the healthy tissues of the mice (i.e. blood, liver, kidneys, bone, etc.) during the experiment, ultimately producing images with excellent tumor-to-background contrast. A biodistribution conducted at the end of the experiment confirmed the imaging data, revealing activity concentrations of >80 %ID/g for each candidate at 144 h post-injection: 92.5 ± 27.2 %ID/g for [⁸⁹Zr]Zr-DFO-pertuzumab, 126.9 ± 50.3 %ID/g for [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, and 86.9 ± 53.2 %ID/g [⁸⁹Zr]Zr-^malDFO-pertuzumab (Figure 4B and Table S4). A longitudinal biodistribution study in which mice bearing BT-474 xenografts were administered a smaller mass dose of each radioimmunoconjugate (32-36 μCi; 6-7 μg) reinforced the steady accretion of the radioimmunoconjugates in the tumor tissue and the high tumor-to-healthy organ activity concentrations produced by the tracers (Figure S9 and Table S5). To wit, in this experiment, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab produced tumoral activity concentrations of 29.6 ± 21.5 and 79.9 ± 27.7 %ID/g at 72 and 144 h post-injection alongside a tumor-to-blood activity concentration ratio of >10 at the later time point.

Figure 4. In vivo evaluation of the radioimmunoconjugates.

(A) Representative PET scans collected 24, 48, 72, 96, 120, and 144 h after the intravenous administration of [⁸⁹Z]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, or [⁸⁹Zr]Zr-^malDFO-pertuzumab [3.7 −3.9 MBq (100-105 μCi), 20-21 μg, in 100 μL of PBS] to athymic nude mice bearing subcutaneous HER2-expressing BT-474 human breast cancer xenografts (n = 5). The images on the left are coronal slices, and those on the right are maximum intensity projections (MIPs); (B) biodistribution data from the mice used for PET imaging collected after the terminal imaging timepoint at 144 h (n = 5).

Complementary PET imaging and terminal biodistribution experiments in athymic nude mice bearing HER2-expressing SKOV-3 ovarian cancer xenografts provided another opportunity to assess the in vivo performance of [⁸⁹Zr]Zr-^SSKDFO-pertuzumab and [⁸⁹Zr]Zr-DFO-pertuzumab (Figure S10, Figure S11, and Table S6). In these experiments, both ⁸⁹Zr-DFO-labeled radioimmunoconjugates clearly visualized the tumor tissue even at early time points and ultimately produced high tumoral uptake and low retention in healthy tissues. Taken together, the preclinical in vivo data plainly show that all three radioimmunoconjugates are excellent imaging agents. However, from a bioconjugation standpoint, the most important message from these data is straightforward: across several experiments, the in vivo performance of [⁸⁹Zr]Zr-^SSKDFO-pertuzumab demonstrated the high tumoral accretion and low healthy tissue uptake that are expected from a highly homogeneous, site-selectively modified radioimmunoconjugate.

DISCUSSION

In the context of radioimmunoconjugates (and all immunoconjugates for that matter), the benefits of site-selective bioconjugation are clear. A more precise approach to synthesis yields immunoconjugates with greater homogeneity, more batch-to-batch reproducibility, and often improved in vitro and in vivo behavior. In this investigation, this lysine-directed site-selective bioconjugation strategy clearly provided a radioimmunoconjugate that was better defined and more homogeneous than analogues synthesized using random lysine labeling or a site-selective maleimide-based method. In in vitro assays, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab exhibited a higher immunoreactive fraction than [⁸⁹Zr]Zr-^malDFO-pertuzumab but not the stochastically-labeled variant. Interestingly, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab was the only construct to reproducibly retain >99% monomeric content. In vivo, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, [⁸⁹Zr]Zr-DFO-pertuzumab, and [⁸⁹Zr]Zr-^malDFO-pertuzumab all demonstrated impressive tumoral uptake, although the biodistribution data uncovered no statistically significant difference between the three radioimmunoconjugates. This almost certainly stems from the highly optimized nature of pertuzumab; more significant differences in in vivo performance are more likely with a less robust mAb and/or payloads that cause greater perturbations to the biophysical properties of the mAb (e.g. hydrophobic toxins).

This report represents — to the best of our knowledge — the first time that a PFP ester-based approach to lysine-directed site-selective bioconjugation has been applied to the synthesis of radioimmunoconjugates. This chemistry was first described in a 2011 patent from Bhat, et al., but the earliest literature report arose in 2018, when Pham et al. compared the site-selective reactivity of N-hydroxysuccinimidyl and fluorophenyl esters of biotin³⁶. More recently, Luciano, et al. used perfluorophenyl esters of the near-infrared fluorophore diSulfo-FNIR to create site-selectively modified antibody-fluorophore conjugates that displayed improved biological photophysical properties compared to randomly labeled analogues⁴⁵. Here, and in these previous studies, the key to this bioconjugation approach lies in the selective reactivity of PFP esters with the primary amine of K188. Mechanistic experiments eschew explanations related to the pKa of K188 or the binding of the PFP moiety to nearby residues. Instead, the selectivity likely has a kinetic origin related to the stabilization of the transition state of the PFP-ester/K188 ligation by the nearby H189 and D151 residues³⁶. In this work, we combine this K188-selectivity with bioorthogonal click chemistry to create a two-step procedure that offers two key advantages over previous methodologies. The branched prosthetic groups allow access to DOLs as high as 4, and the use of a single PFP-bearing bioconjugation reagent (i.e. PFP-bisN₃) provides robust and reliable results by ensuring that the cargo does not influence the selectivity of the bioconjugation reaction.

Finally, it is important to note that this lysine-directed approach to bioconjugation holds several key advantages over extant site-selective methodologies. Unlike methods based on unnatural amino acids, it requires no genetic engineering. Unlike strategies that employ thiol-reactive probes (even ones that offer more stability than maleimides), it eschews the reduction of the mAb and offers greater homogeneity by leveraging only a pair of reactive sites. Unlike methodologies that rely on chemoenzymatic transformations, it does not require enzymes. And unlike approaches that manipulate or remove the heavy chain glycans, it allows for control over FcγRI engagement. Most importantly, we contend that the purely chemical nature of this approach makes it more “translatable” than many existing methods, as the use of bacterially derived enzymes and specialized expression systems can be significant obstacles during the large-scale production of radioimmunoconjugates under Current Good Manufacturing Practices (cGMP) conditions.

CONCLUSION

In this investigation, we have combined a lysine-directed site-selective bioconjugation strategy and bioorthogonal chemistry to create [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, a HER2-targeted immunoPET probe with excellent in vitro and in vivo behavior. This modular and versatile approach to antibody modification produces better defined and more homogenous immunoconjugates than traditional stochastic methods. Ultimately, we believe that this approach offers several critical advantages over established site-selective bioconjugation strategies, particularly with respect to GMP production for the clinic. Efforts to develop clinical imaging agents using this strategy are currently underway.

METHODS AND MATERIALS

General

All reagents were purchased from Fisher Scientific (Thermo Fisher Scientific, Waltham, MA, USA) unless otherwise noted. Pertuzumab was purchased from Memorial Sloan Kettering Cancer Center. Protein concentrations were determined via UV-Vis spectroscopy using an extinction coefficient (ε₂₈₀) of 2.1 × 10⁵ M⁻¹ cm⁻¹ and a molecular weight of 1.5 × 10⁵ Da. All water used was ultrapure (>18.2 MΩcm⁻¹ at 25 °C). DBCO-DFO and ^malDFO were purchased from Macrocyclics, Inc. (Plano, TX, USA). MALDI mass spectrometry was performed by the Alberta Proteomics and Mass Spectrometry Facility (University of Alberta, Edmonton, AB, Canada). ⁸⁹Zr was provided by 3D Imaging (Little Rock, AR. USA).

Instrumentation

All instruments were calibrated and maintained according to standard quality control practices and procedures. UV-Vis measurements were taken on a Shimadzu BioSpecNano Micro-volume UV-Vis Spectrophotometer (Shimadzu Scientific Instruments; Kyoto, Japan). Radioactivity measurements were taken using a CRC-15R Dose Calibrator (Capintec, Inc; Ramsey, NJ, USA) and Automatic Wizard² gamma counter (PerkinElmer; Waltham, MA, USA). Flow cytometry experiments were performed using FACSCalibur instrumentation (BD Biosciences; Franklin Lakes, NJ, USA). Surface plasmon resonance was performed using a Nicoya OpenSPR-XT instrument (Nicoya Lifesciences, Kitchener, ON, Canada).

Stochastic Lysine Bioconjugation

Synthesis of N₃-PEG₈-pertuzumab

Pertuzumab (5.0 mg, 33.3 nmol, 1.0 eq.) in Chelex 100-treated (Bio-Rad Laboratories; Hercules, CA, USA) phosphate-buffered saline (Chelex PBS, pH 7.4) was diluted to a final concentration of 34.5 μM (5.0 mg/mL). The pH of the solution was adjusted to 8.8-8.9 with 0.1 M Na₂CO₃. NHS-PEG₈-N₃ (30.11 μL, 266.7 nmol, 8 eq., 5 mg/mL in DMSO) was slowly added to the solution and mixed thoroughly. The mixture was allowed to react for 1 h on a thermomixer at 37 °C and 500 rpm. Following the incubation, the antibody mixture was purified via size exclusion chromatography (PD-10 column; GE Healthcare; Chicago, IL) and concentrated with a 2 mL Amicon Ultra centrifugal filter with a 50 kDa molecular weight cut-off (MilliporeSigma). Typical yields following purification and concentration were >90%.

Synthesis of DFO-pertuzumab

N₃-PEG₈-pertuzumab (5.0 mg, 33.3 nmol, 1.0 eq.) in Chelex PBS, pH 7.4 was diluted to a final concentration of 34.5 μM (5.0 mg/mL). DBCO-DFO (56.53 μL, 666.7 nmol, 20 eq., 10 mg/mL in DMSO) was slowly added to the solution and mixed thoroughly. The mixture was allowed to react overnight on a thermomixer at room temperature and 400 rpm. Following the incubation, the reaction mixture was purified via size exclusion chromatography (PD-10 column) and concentrated with a 2 mL Amicon Ultra centrifugal filter with a 50 kDa molecular weight cut-off. Typical yields following purification and concentration were >90%.

Site-Selective Bioconjugations

Lysine-directed site-selective bioconjugation: ^SSKDFO-pertuzumab

Pertuzumab (8.0 mg, ~55 nmol, 1.0 eq.) in ice-cold phosphate-buffered saline (PBS, pH 7.4; MilliporeSigma; Saint Louis, MO, USA) was diluted to a final concentration of 10.2 μM (1.5 mg/mL) and kept on ice. BisN₃-PFP (110 μL, 660 nmol, 12 eq., 6 mM in DMF) was immediately mixed with the antibody solution and the mixture was allowed to react for approximately 19 h on ice. Following the incubation, the pH of the solution was adjusted to between 5.6-5.8 via the addition of 0.1 M HCl. The modified antibody was separated from all the low molecular weight reagents by buffer exchange into Chelex PBS, pH 7.4 using a 4 mL Amicon Ultra centrifugal filter with a 30 kDa molecular weight cut-off until the final buffer exchange ratio of at least 1 × 10⁷ was reached. The resulting ^SSKN₃-pertuzumab in Chelex PBS, pH 7.4, was adjusted to a concentration of 2.06 μM (0.3 mg/mL) and quickly mixed with DBCO-DFO (1 mM in DMF) at a 97:3 v:v ratio of protein to DBCO-DFO and allowed to react overnight at room temperature. The final conjugate was separated from unreacted DBCO-DFO by buffer exchange into Chelex PBS, pH 7.4 using a 15 mL Amicon Ultra centrifugal filter with a 30 kDa molecular weight cut-off (MilliporeSigma) until the final buffer exchange ratio of at least 1 × 10⁸ was reached. The absence of any unconjugated DBCO-DFO in the final preparations of the conjugate was confirmed by size-exclusion HPLC on Zenix-C SEC-300 column (Sepax Technologies, Newark, DE, USA). Typical total yield following both purification steps was >75%.

Thiol-directed site-selective bioconjugation: ^malDFO-pertuzumab

Pertuzumab (5.0 mg, 33.3 nmol, 1.0 eq.) in Chelex PBS, pH 7.4 was diluted to a final concentration of 34.5 μM (5.0 mg/mL). Next, Tris(2-carboxyethyl)phosphine hydrochloride (TCEP HCl, 33.3 μL, 333.3 nmol, 10 eq., 10 mM in H₂O) was added and the solution was mixed thoroughly. Maleimide-DFO (23.73 μL, 333.3 nmol, 10 eq., 10 mg/mL in DMSO) was slowly added to the solution and mixed thoroughly. The mixture was allowed to react for 2 h on a thermomixer at 37 °C and 500 rpm. Following the incubation, the reaction mixture was purified via size exclusion chromatography (PD-10 column) and concentrated with a 2 mL Amicon Ultra centrifugal filter with a 50 kDa molecular weight cut-off. Typical yields following purification and concentration were >90%.

Radiosynthesis

DFO-pertuzumab, ^SSKDFO-pertuzumab, and ^malDFO-pertuzumab (0.5 mg) were diluted in Chelex PBS (pH 7.4) to a final concentration of 1 mg/mL. [⁸⁹Zr]Zr⁴⁺ [92.5 MBq – 185.0 MBq (2.5 mCi – 5 mCi)] in 1.0 M oxalic acid was diluted with Chelex PBS and the solution pH was adjusted to 7.0-7.5 with 1.0 M Na₂CO₃ (final volume: 100 μL). After CO₂ bubbling ceased, the ⁸⁹Zr solution was added to the antibody solution, mixed thoroughly, and reacted on a thermomixer (500 rpm, 37 °C, 15 min). The reaction progress was assayed using glass-fiber, silica-impregnated instant thin-layer chromatography (iTLC) paper (Pall Corp.; East Hills, NY), eluted with 50 mM EDTA (pH 5.0), and analyzed on an AR-2000 radio-iTLC plate reader using Winscan Radio-TLC software (Bioscan, Inc.; Washington, DC). Following the completion of the reaction, any free [⁸⁹Zr]Zr⁴⁺ was removed from the radioimmunoconjugate using size exclusion chromatography (PD-10 column). The radiochemical purity of the final radiolabeled construct was assayed using radio-iTLC with EDTA as the eluent (50 mM, pH 5.0). All radiolabeling studies were performed in triplicate.

Immunoreactivity Assays

To determine the immunoreactivity of the antibody constructs for their HER2 target, a bead-based immunoreactivity assay was used as described by Sharma, et al.⁴⁶ Briefly, 20 μL of HisPur^™ Ni-NTA magnetic beads were washed twice with PBS + 0.05% Tween-20 (PBS-T). After each wash, the tubes were placed on a magnetic rack, and the supernatant was discarded. Next, 200 μL PBS-T + 10 μL HER2 antigen (0.1 mg/mL in Chelex PBS + 1% BSA, pH 7.4) were added to the beads, and the solutions were mixed thoroughly. The tubes were incubated for 15 minutes at room temperature on a rotating platform. Following incubation, the beads were washed twice with PBS-T. One cohort of beads per antibody did not receive any antigen and was used as a negative control. Next, 1 ng of [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, and [⁸⁹Zr]Zr-^malDFO-pertuzumab (1 μg/mL in PBS + 1% BSA) were added to their respective tubes. Each antibody had a second cohort in which the HER2 antigens were blocked with 5 μL of cold antibody (1 mg/mL in PBS + 1% BSA) immediately prior to the addition of the radiolabeled antibody to demonstrate the specificity of the constructs. The samples were allowed to react for 30 min on a rotating platform at room temperature. Following incubation, the supernatants were collected. The tubes were then washed twice with PBS-T, and the supernatants of each wash were collected. The samples were then measured on an ⁸⁹Zr-calibrated gamma counter, with the activities (counts/minute) background- and decay-corrected to the start of the run. The immunoreactivity was expressed as a percentage by comparing the activity remaining in the beads to the total activity (beads + supernatant + washes) (n = 3).

Cell Culture

The human breast cancer cell line (BT-474) and ovarian cancer cell line (SKOV-3) were purchased from the American Type Culture Collection (ATCC; Manassas, VA) and maintained under sterile conditions in either DMEM:F-12 (1:1) HG medium supplemented with 10% fetal bovine serum, 2 mM l-glutamine, non-essential amino acids, 100 units/mL penicillin, and 100 units/mL streptomycin (BT-474) or McCoys 5a medium supplemented with 10% fetal bovine serum, 100 units/mL penicillin, and 100 units/mL streptomycin (SKOV-3). All cells were stored in an incubator set to 37 °C and 5% CO₂. The cells were passaged upon reaching 80% confluency.

Flow Cytometry

2 × 10⁶ BT-474 cells were added to a microcentrifuge tube and washed 3 times with ice-cold PBS. 50 μL of either pertuzumab, DFO-pertuzumab, ^SSKDFO-pertuzumab, ^malDFO-pertuzumab, or a non-specific hIgG1 isotype control (final concentration: 6 μg/mL) were added to their respective tubes (n = 3). The cells were then incubated on ice for 30 min. Following incubation, the samples were washed 3 times with ice-cold PBS. After the final wash, 50 μL of goat anti-human IgG Alexa 488 antibody was added to the tubes (final concentration: 6 μg/mL), and the samples were incubated for 30 min on ice. The samples were then washed 3 times with ice-cold PBS. The cell pellets were re-suspended in FACS buffer (PBS + 0.05% FBS + 2 mM EDTA), and the samples were measured using FACS Caliber instrumentation. The data was analyzed using FlowJo software. All samples were performed in triplicate. Cells that did not receive any primary or secondary antibody were used as a non-stained control.

To determine the EC₅₀ values of the immunoconjugates, the same protocol as above was followed with the following modifications: the primary antibodies were incubated with the cells with final concentrations of 24, 12, 6, 3, 1.5, 0.75, 0.375, and 0.1875 μg/mL. The secondary antibody, goat anti-human IgG Alexa 488, was added to the tubes with a final concentration of 6 μg/mL, as described above (n = 3).

Surface Plasmon Resonance

The affinity of the antibodies for their target, HER2, was measured using surface plasmon resonance. Briefly, protein A was immobilized onto an activated carboxyl sensor using a Nicoya OpenSPR kit as per the manufacturer’s instructions. The mAb — pertuzumab, pertuzumab-^SSKDFO, pertuzumab-^malDFO, or pertuzumab-DFO diluted in running buffer (HBS + 0.05% P-20 + 0.1% BSA) — was captured onto the protein A sensor (25 μg/mL over 300 s) in channel 2. The multicycle kinetics experiment was performed by flowing 1.23, 3.4, 11, 33, and 100 nM HER2 antigen solutions (prepared in running buffer) over the sensor for 300 s in channel 1 + 2. Glycine HCl (10 mM, pH 1.5) was used as a regeneration solution between each injection of antigen. After each injection of antigen, the mAb had to be recaptured on the protein A sensor prior to the next injection of antigen, as the regeneration injection strips the mAb from the protein A. Blank buffer runs were subtracted from the results, and the kinetics were determined using TraceDrawer software.

Gel Electrophoresis

Pertuzumab, DFO-pertuzumab, ^SSKDFO-pertuzumab, and ^malDFO-pertuzumab samples were prepared for SDS-PAGE according to the manufacturer’s instructions (NuPAGE^™, ThermoFisher). Briefly, 2 μg of each antibody was reduced using 10× reducing agent and denatured using 4× LDS buffer. Samples were then diluted to 20 μL using DI water. The reduced antibody samples were then placed on a thermomixer at 85 °C, 300 rpm for 15 minutes. Next, 20 μL of each sample was added to the well of a Novex 4-12% SDS Page Gel, and the gel box was filled with NuPAGE^™ MOPS running buffer. A Novex^™ sharp pre-stained protein ladder was also added on each side of the samples. The gel was allowed to run at 70 V for 2.5 hours, or until the ladder bands reached the bottom of the gel. The gel was then washed 3× with DI water. Enough SimplyBlue^™ SafeStain was added to cover the gel, and the gel was allowed to stain for 90 minutes on a shaker. Following staining, the gel was washed 3× with DI water. Finally, the gel was imaged using Odyssey CLX instrument and analyzed using Image Studio software.

Animals

Five to seven-week-old female athymic nude mice (#00785) were obtained from Jackson Laboratory (Bar Harbor, ME) and allowed to acclimatize approximately 1 week prior to tumor inoculation. Animals were housed in ventilated cages and given food and water ad libitum. All animal work was approved by IACUCs of Hunter College, Weill Cornell Medical College, and Memorial Sloan Kettering Cancer Center.

Subcutaneous Xenografts

Mice were given water containing β-estradiol (8 μg/mL, Sigma Aldrich) one week prior to the implantation of the BT-474 xenografts and continued to receive β-estradiol for the remainder of the study. For the implantation, mice were anesthetized by inhalation of 2% isoflurane (Baxter Healthcare; Deerfield, IL)/oxygen gas mixture. Next, the injection site was sanitized with an ethanol wipe, and 1 × 10⁷ BT-474 cells (100-150 μL) in mixture of 1:1 media:Matrigel (Corning Life Sciences; Corning, NY) were implanted subcutaneously in the right shoulder. To ensure homogenous tumors, the cell suspension was mixed thoroughly prior to each inoculation. The BT-474 tumors reached the ideal size for imaging and biodistribution studies (~100 mm³) after 3 weeks. The procedures used to generate the subcutaneous SKOV-3 xenografts are described in the Supporting Information.

PET Imaging

All imaging of mice was performed under 2% isoflurane/oxygen gas mixture. The PET images were obtained using a microPET Focus 120 small-animal scanner (Siemens Medical Solutions; Malvern, PA). Briefly, mice underwent static scans between 24 and 144 h after the intravenous tail vein administration of [⁸⁹Zr]Zr-DFO-pertuzumab, [⁸⁹Zr]Zr-^SSKDFO-pertuzumab, or [⁸⁹Zr]Zr-DFO-^malDFO-pertuzumab [3.7 −3.9 MBq (100-105 μCi), 20-21 μg, in 100 μL of PBS] for a total scan time of 10-30 minutes. The counting rates in the reconstructed images were converted to activity concentrations (percentage injected dose per gram of tissue [%ID/g]) using a system calibration factor derived from the imaging of a mouse-sized water-equivalent phantom containing ⁸⁹Zr. Maximum intensity projection (MIP) images were generated from 3-dimensional ordered subset expectation maximization reconstruction (3D-OSEM). The resulting images were analyzed using ASIPro VM^™ (Concorde Microsystems; Knoxville, TN).

Biodistribution Study

Following the terminal PET imaging timepoint (144 h p.i.), the mice were euthanized via CO₂ asphyxiation followed by cervical dislocation. The 15 most relevant organs were collected, rinsed in water, dried, weighed, and quantified using an Automatic Wizard² γ-counter calibrated for ⁸⁹Zr (PerkinElmer). The counts/minute in each tissue was background- and decay-corrected to the start of the activity measurement. The %ID/g for each sample was calculated by normalization to the total injected activity.

Statistical Analysis

All data presented are expressed as the mean ± standard deviation of at least three independent experiments. Statistical differences were analyzed with GraphPad Prism software (8.0 GraphPad Software Inc., San Diego, CA, USA) via an unpaired, two-tailed Student’s t test (with a Welch’s correction). * = p < 0.05; ** = p < 0.01.

ABBREVIATIONS

PET

positron emission tomography

DFO

desferrioxamine

HER2

human epidermal growth factor receptor 2

Mal

maleimide

SSK

site-selective lysine

TCEP

tris-(2-carboxyethyl) phosphine

DOL

degree-of-labeling

PFP

perfluorophenyl ester

DBCO

dibenzocyclooctyne

ADC

antibody-drug conjugate

PD-L1

programmed death-ligand 1

GD2

ganglioside G2

CNS

central nervous system

mAb

monoclonal antibody

SPAAC

strain promoted alkyne-azide cycloaddition

IgG

immunoglobulin G

ESI-MS

electrospray ionization mass spectrometry

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

MALDI-ToF

matrix-assisted laser desorption/ionization-time of flight mass spectrometry

EC₅₀

half maximal effective concentration

ELISA

enzyme-linked immunoassay

%ID/g

% of injected dose per gram

PBS

phosphate-buffered saline

DMSO

dimethylsulfoxide

RPM

revolutions per minute

EDTA

ethylenediaminetetraacetic acid

iTLC

instant thin-layer chromatography