Hierarchical design of hyaluronic acid-peptide constructs for glioblastoma targeting: Combining insights from NMR and molecular dynamics simulations

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Abstract The main bottleneck of glioblastoma still relies on the existence of the blood brain-blood brain tumor dual barrier, along with the lack of therapy specificity. The present work deals with the question of whether (and how) different targeting hyaluronic acid (HA)-peptide [c(RGDfK) and/or H 7 K(R 2 ) 2 ] moieties hierarchically interact with each other, to ensure a unique entity with specificity to glioblastoma. A dual experimental-computational approach, encompassing nuclear magnetic resonance and molecular dynamics simulations is enclosed. Relevant contact patterns based on the identification of the stabilizing/destabilizing noncovalent interactions within the constructs are detailed. The synthesis pathway requires the HA-c(RGDfK)-H 7 k(R 2 ) 2 association hierarchy, stemming from the size and amino acid residue rearrangement, in the 1:1 molar ratio, to obtain a stable conjugate ultimately able to interact with the tumor cell membrane. To our knowledge, the structural and mechanistic rationale for the formation of hybrid polymer-peptide constructs, including HA-c(RGDfK)-H 7 k(R 2 ) 2, for glioblastoma has not been addressed so far.


Introduction
Brain tumors, as heterogeneous malignant neoplasms in the central nervous system, are considered deleterious diseases responsible for incurring a low survival rate. The global estimated number of new cases of brain and other nervous system cancers in 2018 was 3.5 per 100,000 people, while the number of deaths was 2.8 per 100,000 people, which accounts for an 80% mortality rate. Glioblastoma (GBM), a grade IV astrocytoma, arises as the most common, aggressive and lethal malignant brain tumor in humans 1 . The current therapies rely on the surgical resection, followed by radiotherapy and adjuvant chemotherapy 2 . However, the diffuse nature of the tumor hampers an efficient surgical resection, also associated to the inability of removing brainstem structures. The most challenging issues in GBM therapy include the (i) complexity and heterogeneity molecular biology of the tumor, (ii) high growth rate due to the marked angiogenesis contribution, (iii) presence of the blood brain barrier (BBB), a severely restrictive layer, and (iv) structural complexity of the brain, which limits the amount of drugs that achieve therapeutic concentrations in the invasive regions 3 . The armamentarium of strategies currently under investigation, envisioning glioblastoma treatment, is still poorly understood and far from consensual [4][5][6] . These include, among others, the use of nanosystems, which have been designed aiming at overcoming the limitations and improving the treatment efficacy, ultimately on the basis of a personalized medicine perspective. For that, the functionalization of particle surface to actively targeting the tumor cells and disrupting the tumor microenvironment is mandatory 7 . In this context, and taking advantage of tumor brain related molecular processes, such as transmembrane receptor overexpression (e.g. epidermal growth factor receptors), and the physicochemical tumor features, i.e. lower pH, hypoxia and enzymatic expression, different supramolecular assemblies have been proposed [8][9][10][11][12][13][14][15] . Peptide engineering, involving the residue-and/or site-specific modification of peptides with polymer backbones have been endorsed to design hybrid constructs potentially addressed to fit this purpose.
This work aims at synthesizing a triple targeting polymer-peptide conjugate for the treatment of GBM, mechanistically hypothesized to gather the suitable properties to enhance both tumor targeting and anti-tumor activity. Hyaluronic acid (HA), a component of the extracellular matrix (ECM), and synthesized at the inner face of the plasma membrane as a free linear polymer, is proposed as biopolymer backbone 16 . HA is a biodegradable, biocompatible and a non-immunogenic glycosaminoglycan, composed of repeating disaccharides of J o u r n a l P r e -p r o o f 4 glucuronic acid and N-acetylglucosamine, which has been largely used for tumor targeting 17 .
The strategy is supported by the fact that many types of tumor cells, including GBM ones, overexpress HA receptors (e.g. CD44). However, poor selectivity is evidenced, especially due to the facile saturation of CD44. HA:peptide conjugates, including HA:c(RGDfK) and/or HA-c(RGDfK):H 7 k(R 2 ) 2 combinations, arise as an interesting approach to improve tumorspecific targeting 18 . Cyclo (arginine(R)-glycine(G)-aspartic acid(D)-phenylalanine-lysine), c(RGDf) peptide is a promising ligand due to the respective high-binding selectivity for αvβ3 and αvβ5 integrins, heterodimeric receptors that mediate tumor growth, metastasis and tumor angiogenesis and are overexpressed on the endothelial cells of tumor angiogenic vessels, as well as in GBM cells (e.g. U-87 MG cell line) [19][20][21] . In turn, H 7 K(R 2 ) 2 , an arginine(R)-rich peptide that possesses the pH trigger sequence H 7 , can respond to the acidic pH microenvironment in glioma tissues due to the ionization of polyHis switching from hydrophobic to hydrophilic under acid conditions, thus being considered a tumor-specific pHresponsive peptide 22,23 . On the other hand, this peptide presents cell-penetrating peptide (CPP) characteristics, which endow the ability to cross the BBB and to accumulate in the brain in a seemingly energy independent manner. Figure 1 shows the chemical structures of hyaluronic acid and both peptides, c(RGDfK) and H 7 K(R 2 ) 2 , as well as potential targeting receptor groups identified through the SwissTargetPrediction server 24 . peptides. For convenience, the C atom and peptide residues labelling for each molecule are included. Note that similarly to MD simulations, the dimer of HA was considered for the target prediction. The "Adhesion" target group refers to the possibility of HA to target enzymes whose interaction may potentially inhibit cancer cell adhesion.  The synthesis and characterization of polymer-peptide conjugates raise several questions in what concerns the hierarchical topology of such structures. How can the synthesis pathway be tailored to provide a unique stable system? Novelty consists on explaining polymer-peptide interactions based on a dual experimental-computational framework, encompassing nuclear magnetic resonance (NMR) and molecular dynamics (MD) simulations.

2.2.Synthesis of peptide-polymer conjugates
Hyaluronic acid (HA) modification was achieved by amine coupling. First, carboxylic groups of HA were activated by EDC/sulfo-NHS method as previously reported 25 . HA (5x10 -4 mmol) was dissolved in ultrapurified water followed by the addition of EDC (10x10 -4 mmol) and sulfo-NHS (10x10 -4 mmol). The reaction mixture was stirred gently for 2 h followed by the addition of c(RGDfK) (5x10 -4 mmol) and/or H 7 K(R 2 ) 2 (5x10 -4 mmol) and stirring for 3 h. The reaction mixture was dialyzed against ultrapurified water using a 14 kDa cutoff membrane for 24 h to remove unreacted reagents.

2.3.Structural analysis based on NMR
The 1 H and 13 C NMR spectra were obtained on a Bruker Avance III HD 500 MHz NMR spectrometer. The 13 C spectra were recorded using proton decoupling techniques taking advantage of the nuclear Overhauser effect. The methyl signal of tert-butyl alcohol was used as  solvated with approximately 20200 water molecules.

Simulation details
MD simulations were performed using the GROMACS package (version 4.6.5) 31 and the all atom amber99sb force field 32 , under periodic boundary conditions, and with a NPT ensemble, see 33 . TIP3P waters were employed for modelling aqueous solvation. A constant temperature and pressure of 300 K and 1 bar, respectively, were imposed in all simulations, by the coupling constants of 0.5 ps and 1 ps, respectively. An equilibration run of 40 ns was performed prior to each production run, maintaining the pressure at 1 bar. No pressure coupling was imposed during the production runs, allowing to keep the size of the simulation box constant. Lennard-Jones interactions and electrostatic interactions were assessed using a cut-off of 0.9 nm and the particle mesh Ewald (PME) method, respectively 34 . The constraints in the binding partners were imposed by the LINCS algorithm 35 . The occurrence and strength of noncovalent interactions are inferred, respectively, when δg inter > 0, and by the magnitude of the descriptor at a point in space.

Analysis of noncovalent interactions
∇ρ IGM,inter is obtained from the sum of the N atoms in the different components, referred as A and B, in the x-direction, IGMPlot employs the pre-computed atomic charge densities for estimating a pro-molecular density that produces a minimal effect on the noncovalent interactions. δg inter allows identifying NCI regions, and ∇ 2 ρ, a second derivative (Laplacian) of the density, is used for discriminating favorable/unfavorable NCIs. The decomposition of the Laplacian term into J o u r n a l P r e -p r o o f

Results and discussion
The rationale behind polymer-peptide conjugate synthesis is assuming increasingly importance, in particular for tumour targeting purposes. Betting on supramolecular structures that entail a certain level of complexity, associated to e.g. the size, nature and type of amino acid residues, and also to variations in conformational behavior must be supported by a deep understanding of the mechanistic aspects involved in the binding process and stability of polymer-peptide conjugates. In this context, several questions are raised: How can polymer-peptide supramolecular constructs be formed? Is the peptide order addition relevant?
What governs these hierarchical constructs at the molecular level? Which are the main interaction forces established? These challenging aspects are explored in what follows, combining insights from NMR and MD simulations.

Synthetic route proposed for polymer:peptide contructs
The hierarchical design of hyaluronic acid-peptide constructs for glioblastoma targeting is hypothesized in

Hierarchical construction of peptide conjugates
The type and extent of the interactions between HA, c(RGDfK) and H 7 K(R 2 ) 2 are critical for establishing the binding sites, exploring conformational changes, and providing an in-depth knowledge of the dynamics of polymer-peptide conjugation. Such information is only experimentally available when fingerprint techniques, including NMR spectroscopy, are utilized.

HA:c(RGDfK)
The 1 H and 13 C NMR spectra were first acquired for the individual components, HA and the c(RGDfK) peptide, and for the synthesized conjugate between HA and the peptide in the   The assignment of 1 H and 13 C spectra was completed based on homonuclear and heteronuclear correlations obtained in the bi-dimensional spectra. 1 H spin systems were identified based on TOCSY data, starting from NH signals (Figure S1), and the 1 H NMR signals fully assigned by combining TOCSY with COSY ( Figure S2) and ROESY spectra. 13 C assignment was based on HSQC and HMBC spectra (Figures S3 and S4, respectively). The  (Figure 1). The same effect is observed for the aromatic region of the phenylalanine, also suggesting an increase of the rigidity of the aromatic ring in the HA-c(RGDfK) conjugate. A pronounced broadening is also seen in the other regions of the spectra, which is most pronounced for the signals of the lysine lateral chain, in particular, the signal of CH 2 -ε of lysine, also indicating that its terminal NH 2 may be bound to the carboxylic group of HA ( Figure 5). This observation is in complete agreement with the loss of the coupling constants (Table S3) and the intensity changes observed for the 13 C signals of the lysine lateral chain due to the signal broadening ( Figure 6).
Notwithstanding, this interaction is not observed when both single molecules are in solution, due to the rigidity of the HA backbone associated to intramolecular hydrogen bonds between the carboxyl group in C6 and the hydroxyl group in C6'. Likewise, no carboxyl groups remain available to establish interaction with terminal NH 2 of lysine moiety in c(RGDfK). Moreover, the 13 C NMR spectra of HA in the presence of the peptide show two signals for each of the carbon atoms C-6 and C-7', one of them shifted to high frequencies and the other one Journal Pre-proof presenting very similar resonance to the respective nuclei in free HA; in addition, the signals corresponding to nuclei C-5 and C-4' appear broadened and duplicated, respectively ( Figure   6). These observations, in particular, the intensities of the bound and unbound signals, suggest that spatially alternating moieties of HA are involved in interactions with the peptide by the carboxylic acid, which is bound to the terminal NH 2 group of lysine. This alternating state allows the formation of other NH … O hydrogen bonds between other moieties of c(RGDfK), such as C6=O … NH (D-Phe), see Table S1.
Additional interactions of the methyl group of the N-acetyl glucosamine moiety with the aromatic group of phenylalanine cannot be ruled out, based on the changes in the corresponding signals, as has previously been described for peptides and HA 39 .

HA-c(RGDfK):H 7 K(R 2 ) 2
The HA-c(RGDfK):H 7 K(R 2 ) 2 conjugate was subsequently prepared by the addition of H 7 K(R 2 ) 2 to the preformed HA-c(RGDfK) in the 1:1 molar ratio. The corresponding 1 H and 13 C spectra are represented in Figures 7 d) and 8 d). Two 1 H signals, (1.97 and 2.84 ppm) for the methyl group CH 3 -8', corresponding to free and conjugated HA, also show the 2:1 intensities ratio as is observed for the HA conjugate obtained with the peptide H 7 K(R 2 ) 2 alone.
In agreement with this, 13    moiety and the c(RGDfK)-phenylalanine phenyl CH groups. Stabilization of the HA-c(RGDfK):H 7 K(R 2 ) 2 construct is also assured by weak hydrogen bonds between the H 7 K(R 2 ) 2 -arginine (R 1 and R 2 ) carbonyl groups, and the c(RGDfK)-lysine C ε H 2 , C β H 2 , C γ H 2 and c(RGDfK)-aspartate C β H 2 groups.
The architecture of HA-c(RGDfK):H 7 K(R 2 ) 2 also exhibits propensity to establish hydrophobic C-H⋯C-H interactions and C-H⋯π dispersion interactions (Figure 9, panel C). These support the decrease of flexibility previously observed due to the formation of a "pocket-like" conformation.
For the former, the C-H bonds belong for e.g. to the H 7 K(R 2 ) 2 -arginine (R 2 ) 1 and c(RGDfK)-lysine (C γ H 2 ). C-H⋯π interactions are established between the H 7 K(R 2 ) 2 -lysine(K) moiety, which acts as C-H donor, and the electron-rich phenyl ring of c(RGDfK)-phenylalanine moiety, which corresponds to the π system (panel C).
These interactions can also be identified in the more intense peaks on the left, at sign(λ 2 )ρ ≈ -0.016. Such weak attractive forces have been recognized as important driving forces in the association process of similar systems involving carbohydrates and aromatic moieties, and also protein-drug complexes 33  interactions, which are discriminated in the isosurfaces represented in Figure 9. Each atom is colored using a gray (no contribution)-to-red (significant relative contribution) gradient reflecting the relative score (%).

Conclusions
Keeping active targeting and also nanoparticle surface modulation purposes in consideration, a hybrid polymer-peptide construct based on HA as biopolymer, and c(RGDfK) and H 7 K(R 2 ) 2 as peptides was successfully designed. The establishment of such molecular assemblies was comprehensively characterized based on NMR and MD simulations. While NMR provided evidenced regarding structured architecture, MD simulation shed light on the relevant contacts between the interaction pairs. These methodologies enabled to set forth a hierarchy in terms of synthesis pathway, governed by the size and amino acid residue rearrangement, and also the appropriate molar ratio