NMR Structural Elucidation of Mitoxantrone—Gonadotropin Releasing Hormone (GnRH) Conjugates Implicated in Hormone Dependent Cancer

1. Introduction

Cancer is a multifactorial disease characterized by genetic alterations that disrupt the natural balance between cell proliferation and apoptosis [1]. Hormone-dependent cancers comprise a diverse group that affects thousands of individuals and includes two of the most common types of cancer in men (prostate cancer) and women (breast cancer) [2,3]. Steroid and peptide hormones play a pivotal role in regulating cell proliferation through endocrine, paracrine, or autocrine action [4,5,6]. Any disruption in these regulatory aspects leads to carcinogenesis.

Sex steroid hormones, such as estrogens and androgens, have been strongly associated with the development of hormone-dependent malignancies, including breast, ovarian, endometrial, and prostate cancers in both women and men [7]. Similarly, peptide hormones such as Gonadotropin Releasing Hormone (GnRH) have also been implicated in cancer development [8,9]. Receptors for GnRH, which have a significant role in regulating reproductive function and sex steroid production, are overexpressed in certain cancer types. Consequently, their expression on the cell surface marks them as potential targets for therapeutic intervention [10,11,12].

The most common approaches involved the development of altered GnRH peptide analogues acting as agonists (e.g., leuprolide) or antagonists (e.g., degarelix) or non-peptide antagonists (e.g., elagolix) [13,14]. However, the serious adverse effects associated with these therapies prompted the development of novel therapeutic strategies that incorporated cytotoxic agents with the GnRH peptide for targeted drug delivery. The developed conjugates presented enhanced selectivity and efficacy while minimizing systemic toxicity [15]. The first such conjugate, developed and synthesized by A. R. Schally and co-workers, was the AEZS-108 (Zoptrex), in which doxorubicin (DOX) is attached to a GnRH analogue [DLys⁶]GnRH via a glutaryl spacer [16]. Despite the encouraging outcomes from the initial clinical studies, AEZS-108 was discontinued after Phase III trials failed to demonstrate enhanced survival compared with doxorubicin. This outcome was attributed to the early release of doxorubicin into the bloodstream, highlighting the insufficient stability of the conjugate [17].

In this study, we present the structural elucidation of two comparable conjugates based on the structure of AEZS-108. These conjugates consist of a GnRH peptide analogue connected to mitoxantrone through a disulfide bond (Figure 1). The specific design aims to enhance the selectivity and efficacy of the cytotoxic agent while improving the stability of the conjugate. The GnRH peptide analogues were designed to be structurally similar to the commercially available agonist Leuprolide, except at position 6, where D-Leu was replaced by D-Lys or D-Cys [18]. Drug release is facilitated through a thioredoxin-mediated reductive mechanism. Mitoxantrone (MTX) was selected due to its improved toxicity profile compared with doxorubicin [18,19].

Both conjugates were evaluated in vitro and exhibited high affinity for the GnRH receptor. Moreover, they induced dose- and time-dependent anti-proliferative effects in different cancer cell lines, comparable to those of mitoxantrone alone. The findings demonstrate effective GnRH receptor targeting treatment with selectivity and suggest that the therapeutic benefits of these conjugates may arise from improved pharmacokinetics and reduced systemic toxicity [18,20,21]. Despite these promising biological results, the three-dimensional solution structures and the impact of mitoxantrone conjugation on the conformational properties of the GnRH peptide analogue remained unknown. Such structural characterization is essential for understanding receptor recognition and guiding the rational design of improved GnRH peptide-drug conjugates. To address this knowledge gap, the conformational properties of con3 and con7 were investigated by two-dimensional NMR spectroscopy combined with molecular dynamics simulations to obtain solution-state structural models and provide insights into conjugate-GnRH receptor interactions.

2. Results and Discussion

2.1. Structure Identification

Initially, NMR experiments of the synthesized analogues were performed in D₂O (Tables S1; S3; S5; S6), as it mimics the aqueous intracellular environment. However, these spectra proved unsuitable for complete ¹H assignment and structural elucidation because rapidly exchanging protons were not observable due to their exchange with deuterium from the solvent. To overcome this limitation, the chemical shift assignments of the two conjugates, con3 (Table 1), con7 (Table 2), and their respective peptides (Tables S2; S4), were carried out in DMSO-d₆. This solvent provides an amphiphilic environment that suppresses proton–deuterium exchange, allowing the observation of exchangeable protons. The assignments of con3 and con7 in DMSO-d₆ were achieved using a combination of ¹H NMR, ¹H-¹H NOESY, and ¹H-¹⁵N HSQC experiments, together with data from the corresponding peptide spectra in DMSO-d₆ and comparative information obtained from the D₂O NMR experiments.

2.2. Structure Elucidation

Cross-peaks in the NOESY spectra of all analogues corresponding to spatial proximity between non-overlapping proton resonances were analyzed. The analysis started on the spectral regions between 0.8–2.0 ppm (aliphatic protons) and 6.0–8.0 ppm (aromatic protons), which were examined in detail to identify well-resolved, non-overlapping cross-peaks suitable for distance estimation. The analysis was then extended to the remaining spectral regions. The complete lists of observed cross-peaks are provided in Table 3 for con3 and con7, and Table S7 for the corresponding peptides.

2.3. MD in water and DMSO

Clustering analysis in water identified two dominant clusters for each conjugate, representing 50% (Figure 2. A) and 19% of structures in the simulation time for con3, and 45% (Figure 2. B) and 25% for con7, respectively. The dominant clusters overall exhibited highly similar conformations, with only minor differences in the orientation of the mitoxantrone tricyclic moiety. In all clusters, the peptide backbone retained the U-shaped conformation, indicating that the conformation of conjugates depends mainly on the peptide backbone rather than the mitoxantrone moiety. In contrast, analysis in DMSO revealed a single dominant cluster for each conjugate, present for approximately 85% of the simulation time (Figure 2).

In con3 (Figure 2. A), the peptide backbone adopts a closed U-shaped conformation instead of the helical structure observed for the free peptide (Figure S3. A). On the other hand, the peptide backbone of con7 (Figure 2. B) adopts a more compact U-shaped conformation than its corresponding free peptide (Figure S3. B). The mitoxantrone moiety remains in close proximity to the peptide backbone in both conjugates during the water simulations but is positioned farther away in DMSO (Figure 2). In both solvents, the disulfide bond is exposed to the solvent, facilitating its proposed reduction mechanism by the thioredoxin system. Previous structural studies of leuprolide and GnRH peptide analogues have identified His², Trp³ and Arg⁸ as key residues involved in receptor recognition [22]. In the conformations observed here, the side chains of these amino acids are exposed to the solvent (Figure S3 and Table S8), potentially enhancing receptor interactions and facilitating GnRHR activation.

1.4. MD with NOE constraints

The NMR refinement, performed using the NMR experimental constraints (Table 3 and Table S7), yielded the predicted structures for conjugates in Figure 3 and for the respective peptides in Figure S3. The dominant conformations from the unrestrained MD simulations were used as the starting structures for the refinement protocol. In both the peptides (Figure S3) and the conjugates (Figure 3), the peptide backbone retains a bent conformation. In the peptide component of con7 (Figure S3. B), the backbone adopts an almost helical conformation, while in all other cases the conformations resemble mostly a U-shape (Figure S4. B). Another notable feature observed in the two conjugates is the compact conformation of the mitoxantrone moiety (Figure S4). In the NMR-refined structures of both conjugates, the drug is positioned in close proximity to the peptide backbone, while the spacer in con7 (Figure S4. B) is folded onto itself. Furthermore, the spacer conformation exposes the disulfide bond to the solvent, potentially facilitating its reduction as aforementioned. Also, the side chains of His², Trp³, and Arg⁸ exhibit a solvent-exposed orientation in both conjugates (Figure S3), which is consistent with the high GnRHR binding affinity observed in our previously reported study [18]. The main limitation of NMR refinement modeling is the incomplete definition of all the NOE constraints. Extensive signal overlapping and possible compact conformation of the peptide in solution hindered the assignment of NMR signals for every single proton; consequently, undetermined NOE constraints could not be considered for the refinement.

The comparison between the unconstrained MD simulations in water and DMSO and the NOE-restrained simulations revealed a high degree of structural agreement. In all structures, the peptide backbone adopted a bent, U-shaped conformation, while the mitoxantrone moiety remained away from the peptide backbone only in the DMSO MD simulations. The NMR-refined structures had slightly more open conformations than those obtained from the corresponding MD simulations, due to the limited number of NOE restraints available for refinement. Nevertheless, the major structural features were consistently reproduced by both the computational and experimental approaches. Importantly, in all structures the disulfide bond remained solvent-exposed, indicating that it is accessible to the thioredoxin system for an effective release of the mitoxantrone moiety. This structural observation is compatible with the biological activity of the conjugates and further supports the proposed thioredoxin-mediated drug release mechanism [18,20,21].

1.5. Docking

The molecular docking simulations were performed using the conformations obtained after the NMR refinement. The results are summarized in Table 4 and in Figure 4. Due to their size, the peptide conjugates dock in a vertical fashion and do not move deeply into the receptor binding site like Leuprolide (Figure 4. C). This binding mode potentially resembles the dynamic process of GnRH recognition, in which the peptide approaches the receptor from the extracellular side. Based on the analysis of the docking conformations, there are no hydrogen bonds formed between the peptides and the protein, suggesting that binding is predominantly stabilized by nonpolar interactions such as van der Waals interactions. Notably, the docking results are consistent with the solution conformations obtained from MD simulations in water and NMR-guided MD simulations, indicating that these conformations are compatible with GnRHR recognition while allowing distinct orientations of the mitoxantrone moiety upon receptor binding.

Importantly, docking scores of these analogues compared with Leuprolide suggest that substitution of the D-Leu⁶ residue with either D-Cys or D-Lys is well tolerated within the GnRH receptor binding pocket and is compatible with the previously reported receptor binding affinity study and efficacy evaluation [18,20,21]. Con7 exhibited a docking score of −13.201 kcal/mol, which was more favorable than that of the reference agonist Leuprolide (−12.335 kcal/mol) and agrees with our earlier work [18]. In contrast, the docking score of con3 (−9.869 kcal/mol) was lower than that of Leuprolide (-12.335 kcal/mol), but its experimental binding affinity was higher [18]. The docking score of con7 was comparable to that of its corresponding free peptide (−13.292 kcal/mol), suggesting that the presence of the spacer and the bulky mitoxantrone moiety at position 6 does not significantly interfere with peptide–receptor recognition. Con3 exhibited a lower docking score than its corresponding free peptide (−13.494 kcal/mol), most likely due to the absence of the spacer and the shorter D-Cys side chain, which reduces the conformational flexibility. As shown in Figure 4. C, Leuprolide binds deeper within the receptor pocket than either conjugate. In both conjugates, the mitoxantrone moiety and the disulfide bond (Figure 4. A and B) remain solvent-exposed and easily accessible by the thioredoxin system for reduction, which is consistent with our preceding findings [18]. The disulfide bond of con3 is positioned slightly deeper in the binding pocket because of the absence of the spacer. The flexibility of con7, due to the presence of the spacer, leads to better exposure of the disulfide bond to the solvent and likely promotes its reduction by the thioredoxin system, as previously reported [18,20,21].

3. Materials and Methods

3.1. Synthesis of con3 and con7

The synthesis of conjugates con3 and con7 has been described in our previous work [18]. Briefly, the peptides were synthesized using the Fmoc/tBu solid-phase methodology, and conjugation to the mitoxantrone analogue was achieved by disulfide bond formation. The final products were purified by reverse-phase high-performance liquid chromatography (RP-HPLC) (purity >95%) and characterized by electrospray ionization mass spectrometry (ESI-MS).

3.2. NMR Spectroscopy

DMSO-d₆ (99.8 atom% D) was purchased from Deutero GmbH (Kastellaun, Germany). High-precision 5 mm NMR tubes (Deutero GmbH) were used throughout the study. High-resolution NMR spectra were recorded at 298 K on a Bruker Avance Neo 800 MHz spectrometer equipped with a cryogenically cooled 5 mm ¹H/¹³C/¹⁵N/D z-gradient probe. All experiments were acquired in phase-sensitive mode using standard pulse sequences and phase-cycling routines provided in the Bruker pulse program library. Detailed descriptions of the NMR experiments are provided in the Supplementary Materials (Section 2). Raw NMR data were processed and analyzed using MestReNova software (version 14.3.3-33362) [23].

3.2.1. Structure Identification

Structure identification of the conjugates con3 (Table 1), con7 (Table 2), and their respective peptides (Supplementary Materials; Section 3) through NMR experiments was performed in DMSO-d₆ under the conditions described in Supplementary Materials; Section 2. Chemical shift assignments were obtained from ¹H NMR, ¹³C NMR, 2D ¹H-¹H NOESY, and 2D ¹H-¹⁵N HSQC spectra. Additionally, data from the assignment of all analogues in D₂O (Supplementary Materials; Section 1 and Section 3) were used to support the assignment in DMSO-d₆.

3.2.2. Structure Elucidation

The structure elucidation of conjugates con3 and con7 was achieved using two-dimensional ¹H-¹H NOESY experiments. Cross-peaks corresponding to spatial proximity between non-overlapping proton resonances were analyzed. The resulting distance restraints were incorporated into molecular dynamics (MD) simulations to refine the structural model. The cross-peaks observed from conjugates con3 and con7 NOESY spectra are shown in Table 3 and for the respective peptides in Table S7.

3.3. Molecular Modeling

Molecular dynamics (MD) simulations were initially performed in water to evaluate the conformational aspects of the conjugates under physiological conditions. However, since NMR structural assignments could only be achieved in DMSO-d₆, MD simulations were also carried out in DMSO to enable the direct comparison between the computational models and the NMR experiments. MD production runs were conducted in water and DMSO solvents with and without constraints. The simulation run was carried out for 150 ns using AMBER22 software [24]. Structure preparation: The 2D structures of the analogues were designed with UCSF Chimera 1.16 [25]. Geometry optimization was performed with the GAMESS R1 [26,27] software using the density-functional theory (DFT) [28,29] and the Hartree–Fock (HF) approximation. The B3LYP/6-311G atomic basis set was applied, while the convergence criterion was set at 0.0001 [29,30]. Detailed information on the construction of the parameters and the MD protocol followed is provided in Supplementary Materials (Section 6).

3.4. Docking

The structures obtained from the NMR refinement process (Supplementary Materials Section 5) were employed for molecular docking simulations using Autodock Vina [3.1] [31,32]. The conjugates were treated as a single ligand and prepared with the prepare_ligand tool in the software. The GnRH receptor 1 AlphaFold canonical structure (AF-P30968-F1, Uniprot ID: P30968) was employed for the docking due to the absence of any missing residues compared to the crystal structure (Figure S5) and was prepared using the prepare_receptor tool provided by Autodock Vina. The structure of leuprolide was obtained from the Protein Data Bank Chemical Component Dictionary (PDB ID: 1YY2) and used as the reference ligand for comparison. The binding site of the receptor was defined based on the center-of-mass of elagolix (Figure S5) with coordinates (x, y, z) = (-26.906, -13.2388, 2.45162) and a box size of 35 x 40 x 39 Å. For the docking protocol, an exhaustiveness value of 32 was used.

4. Conclusions

The conformations of the two synthesized mitoxantrone–GnRH conjugates, con3 and con7, were evaluated by high-resolution NMR spectroscopy combined with molecular dynamics simulations in DMSO and water. ¹H and ¹³C assignments were achieved in both solvents, and NOESY-derived distance restraints provided valuable structural information that was incorporated into the MD refinement protocols. Computational and experimental data indicate that both conjugates adopt compact bent conformations characterized by a U-shaped arrangement of the peptide backbone. Despite the sequence difference at position 6, both conjugates exhibit similar conformations, suggesting that this substitution has only a limited influence on the global structure. The mitoxantrone moiety does not significantly alter the peptide backbone conformation, while the key residues His², Trp³, and Arg⁸ remain exposed to the solvent, thereby preserving their ability to interact with the GnRH receptor. Moreover, the orientation of the mitoxantrone moiety exposes the disulfide bond to the solvent, which is consistent with the proposed mechanism of intracellular controlled drug release mediated by the thioredoxin system. The conformational characteristics observed in the unconstrained simulations were further confirmed by NMR-restrained molecular modeling, which generated refined structural models consistent with experimental NOE data. A limitation of the present NMR study was the inability to assign all observed NOE cross-peaks due to signal overlap; therefore, only confidently assigned NOE-derived distance restraints were included in the refinement protocols. Nevertheless, the resulting structural models provide a reliable representation of the predominant solution conformations of both conjugates. The docking studies showed that the structural models of the experimentally determined conformations are compatible with GnRH receptor recognition. Combined with the NMR spectroscopy and molecular dynamics results, these findings provide a framework that explains the previously reported biological evaluation of these conjugates [18,20,21]. In particular, the preservation of the receptor-recognition residues and the solvent-exposed disulfide bond may account for the high GnRH receptor affinity relative to Leuprolide and further validate the proposed thioredoxin-mediated drug-release mechanism. Moreover, the docking study of con7 highlights the importance of the flexible spacer for receptor-binding affinity while maintaining accessibility to the disulfide bond. Overall, the structural information obtained in this study enhances our understanding of GnRH receptor recognition and supports the rational design and optimization of next-generation GnRH peptide–drug conjugates and their continued development as selective drug-delivery platforms for GnRH receptor–expressing malignancies.