Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive materials for ocular retentive delivery systems

Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive materials for ocular retentive delivery systems

Accepted Manuscript Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive mat...

2MB Sizes 0 Downloads 2 Views

Accepted Manuscript Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive materials for ocular retentive delivery systems

Rohan V. Pai, Jasmin D. Monpara, Pradeep R. Vavia PII: DOI: Reference:

S0168-3659(19)30453-5 https://doi.org/10.1016/j.jconrel.2019.07.037 COREL 9884

To appear in:

Journal of Controlled Release

Received date: Revised date: Accepted date:

5 February 2019 22 July 2019 25 July 2019

Please cite this article as: R.V. Pai, J.D. Monpara and P.R. Vavia, Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive materials for ocular retentive delivery systems, Journal of Controlled Release, https://doi.org/10.1016/j.jconrel.2019.07.037

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. 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.

ACCEPTED MANUSCRIPT

Exploring molecular dynamics simulation to predict binding with ocular mucin: An in silico approach for screening mucoadhesive materials for ocular retentive delivery systems Rohan V Pai 1 , Jasmin D Monpara 1

1

and Pradeep R Vavia

1*

Department of Pharmaceutical Sciences and Technology, Institute of Chemical Technology,

MA

NU

SC

RI

PT

Nathalal Parekh Marg, Matunga East, Mumbai - 400019, Maharashtra, India.

Prof. Pradeep R. Vavia,

ED

*Corresponding Author:

EP T

Centre for Novel Drug Delivery Systems,

Department of Pharmaceutical Sciences and Technology,

AC C

Institute of Chemical Technology,

Nathalal Parekh Marg, Matunga East, Mumbai - 400 019, Maharashtra, India. Tel No.: +91 22 3361 2220 Fax No.: +91 22 2414 5614 Email: [email protected]

Page 1 of 39

ACCEPTED MANUSCRIPT Abstract In recent times, molecular dynamic (MD) simulations have been applied in the area of drug delivery, as an in silico tool to predict the behaviour of nanoparticles with respect to their interaction with larger biological entities like bilayer membranes, DNA and dermal surface. However, the predictions must be systematically evaluated by extensive studies with actual biological entities in order to deem the in silico models accurate. Thus, in the present study, MD simulation was used to screen ligands with respect to ocular mucoadhesion. Mucin-4, a

PT

cell surface-associated mucin was selected as the substrate for the in silico study due to its abundance across the ocular surface. The ligands were then incorporated into a delivery

RI

system like nanostructured lipid carriers (NLC) and assessed for mucoadhesion by relevant

SC

in vitro and in vivo techniques. The in silico study suggested chitosan oligosaccharide (COS) to have an extensive mucoadhesive potential towards ocular mucin followed by stearylamine

NU

(STA) and cetrimonium bromide (CTAB) which showed intermediate and low mucoadhesion respectively. The corresponding in vitro assessment by spectrophotometry and nanoparticle tracking analysis showed a similar outcome wherein COS was found to be extensively

MA

mucoadhesive, followed by both STA and CTAB, which showed mucoadhesion to a nearly equal extent. The findings of in vivo confocal imaging following topical administration to rats

ED

showed that while COS and STA adhered extensively to the ocular surface, CTAB showed negligible adhesion. MD simulation was thus found to accurately predict interactions critical to mucoadhesion and the same could be fairly correlated well by relevant mucoadhesion

AC C

EP T

studies both in vitro and in vivo.

Keywords Mucin, Molecular dynamics simulation, Chitosan oligosaccharide, Nanoparticle tracking analysis, Ocular mucoadhesion, In silico Page 2 of 39

ACCEPTED MANUSCRIPT Introduction In the past decade, much work has been done on understanding the mucosal barriers and designing drug delivery strategies for them. This growing interest comes from the fact that the mucosal membrane covers a large area of the human body ranging from ocular, nasal, oral and lung surface to the whole of the gastrointestinal tract [1]. The drug delivery strategies towards mucosal delivery are categorized into mucopenetrating, mucolytic and mucoadhesive systems [2]. While the mucopenetrating and mucolytic systems work by

PT

disrupting the mucosa either physically or chemically, the mucoadhesive systems take advantage of the physicochemical properties of the mucous and its contents to prolong or

RI

retain delivery systems at the mucosal surfaces. A better understanding of how the mucous

SC

interacts with the drug delivery systems can help in improving therapeutics to the vast number of diseases associated with it.

NU

Mucous is a complex hydrogel composed of over 90% of water, 2-5% of mucins, lipids, salts, non-mucin proteins and enzymes like lysozyme [2]. Thus the major proteinaceous component

MA

of mucus is mucins, the factor responsible for the viscous nature of mucous as they are large glycoproteins having a molecular weight ranging from 0.5-20 MDa [3]. Around 20 different types of mucins have been identified across the mucosal surfaces and based on their

ED

functional properties, mucins are further classified into either secreting mucins, like MUC2, MUC5AC and MUC5B or cell-surface associated mucins like MUC4, MUC6, MUC7 and

EP T

MUC19 [4]. While the secreting or gel-forming mucins are constantly cleared off their sites through the stringent clearance mechanisms like peristalsis, ciliary beating and other motions, the cell-surface associated mucins remain intact since they are anchored to the cell

AC C

membranes [2]. Due to the type of amino acids present, mucins carry a net negative charge at the physiological pH, and thus positively charged ligands are known to have more affinity to the mucins. Moreover, the presence of hydroxyl, carboxylic and sulphate groups on the glycan branches of mucins present a rich surface for hydrogen bonding, particularly relevant for polar hydrophilic molecules like polymers. Also, the presence of branched structures attached to the protein core of membrane-bound proteins can act as an excellent substrate for intercalation with the chains of the polymer [5]. Thus, such mucin-rich surfaces like the eye can be explored for retaining nano drug delivery systems by making use of polymers or charged substrates, taking advantage of the mechanical or the electrostatic mechanisms of mucoadhesion which occur via forces like hydrogen bonding and ionic interactions that can be predicted by techniques like molecular dynamics simulations. Mucin-4 (MUC4) is Page 3 of 39

ACCEPTED MANUSCRIPT abundantly produced throughout the entire surface of the ocular epithelium and stratified epithelium of conjunctiva and cornea, providing a vast surface for adhesion [6,7]. In recent times, nanostructured lipid carriers (NLC) have been widely used for delivery of ocular therapeutics for the treatment of various eye diseases and also to other mucus-rich sites [8– 10]. These systems which are lipophilic in nature have been made mucoadhesive by use of materials which are known to interact with the mucin through interactions like hydrogen bonding and electrostatic interactions which can be reasonably predicted using computational

PT

techniques like molecular dynamics [11].

RI

Molecular dynamics (MD) simulation is a force-field based technique which can predict interactions based on molecular motions like vibration, bond stretching, angle bending and

SC

bond formation [11]. MD simulations have been long used to simulate interactions of drugs, proteins and receptors in the field of computer-aided drug design and discovery [12–14]. MD

NU

simulations have also been applied to polymeric systems to predict their interactions with structures like DNA and hormones [15]. Ru et al. have predicted the disintegration and

behaviour [11].

MA

dissolution of a fairly large dosage form like an ocular tablet and studied its dissolution However, the use of MD simulations in predicting interactions like

mucoadhesion and screening potential entities based on it has not been fully explored yet.

ED

Thus, the present work aims at investigating MD simulation as an approach for rationally screening entities with mucoadhesive potential for use in drug delivery systems meant for sites.

Charged

ligands like chitosan oligosaccharide,

EP T

mucoadhesive

stearylamine

and

cetrimonium bromide were screened with respect to their interactions with MUC4 due to their propensity towards ocular mucoadhesion [16–21]. The ligands were chosen such that all

AC C

of them carried a net cationic charge. A ligand with a net anionic charge was also studied by MD simulation alone in order to evaluate the effect of the negative charge on interaction with mucin. The screening has been performed using an in silico technique of molecular dynamic simulation and emphasis is given on the contacts and interactions between the ligands and an ocular cell surface-associated mucin (MUC4). To test the capabilities of MD simulation towards predicting mucoadhesive interactions, the results obtained through the simulations have then been correlated with relevant in vitro and in vivo retention studies using semisynthetic as well as biological mucin.

Page 4 of 39

ACCEPTED MANUSCRIPT Materials and Methods Materials Chitosan oligosaccharide (~3000 Da) having a glucosamine content of ~80% with a degree of deacetylation greater than 90% was procured from Sisco Research Laboratories Ltd. Type-III mucin (640 kDa) was procured from Sigma-Aldrich. Precirol ATO 5 (Glyceryl distearate) was obtained from Gattefosse India Pvt. Ltd. Capmul MCM C8 (Glyceryl caprylate) was procured from ABITEC Corporation, USA. Synperonic PE/F 87 (7000 Da) was provided by

PT

Croda India Pvt. Ltd. Stearylamine and cetrimonium bromide were procured from S.D Fine

reagents used were of analytical grade.

SC

Atomistic molecular dynamics simulation with mucin

RI

Chemical Ltd. Coumarin-6 was provided by Neelicon Food Dyes & Chemical Ltd. All other

Molecular dynamic (MD) simulation was carried out using Maestro 11.1 module of the

NU

Schrödinger® Discovery Suite Release 2016-2. The structure of MUC4 was downloaded as a protein database file from SWISS-MODEL repository maintained by BIOZENTRUM,

MA

University of Basel and the structures of chitosan oligosaccharide (COS), stearylamine (STA), cetrimonium bromide (CTAB) and glyceryl stearyl citrate (GSC) were downloaded from PubChem database having molecular weights and other physicochemical properties as

ED

described in Table 1.

Table 1. Physicochemical properties of the ligands used in MD simulation study with MUC4

Molecule

Mol. Wt.

SASA (A⁰

2

)

MolSA (A⁰

2

)

Net Charge

501.487

~700

~420

+

STA

269.513

~670

~375

+

CTAB

364.456

~650

~380

+

AC C

COS

EP T

type mucin.

SASA= Solvent accessible surface area; MolSA: Molecular surface area as calculated with 1.4 A⁰ probe during simulation The structures for COS, STA and CTAB, were checked for errors or inconsistencies, and the possible tautomers, isomers, ionisation and protonated states were then generated by LigPrep module using Epik, and the geometry of the structures was optimised using the optimised potentials for liquid simulations (OPLS3) force field. Similarly, MUC4 was checked and prepped using the protein preparation wizard module. Further, the energy of the ligands was minimised with a minimisation step in MacroModel module using OPLS3 force field. Prior to Page 5 of 39

ACCEPTED MANUSCRIPT simulation, each structure was separately aligned with MUC4 and the energy of the whole system was minimised using the minimisation step of Desmond module. The system to be simulated was built within an orthorhombic boundary of dimension 10 Å x 10 Å x10 Å using a water model as the solvent system having transferable intermolecular potential with 3 points (TIP3P). Molecular dynamics for each of the combinations were then performed in Desmond for a time period of 10 ns with a recording interval of 10 ps in a constant temperature and constant pressure ensemble at 300 K and 1.01325 bar. The trajectories

PT

generated for each of the models were then analysed for mainly hydrogen bonding, electrostatic bonding and hydrophobic interactions using simulation interaction diagram

RI

generated by the Desmond module.

SC

Preparation of nanostructured lipid carriers

NLC containing the respective individual ligands were prepared by the emulsification-

NU

ultrasonication method by slightly modifying the procedure reported previously in the literature [22]. The composition of the NLC was as shown in Table 2.

MA

Table 2. The composition of NLC prepared with oil, lipid, surfactant and the ligands used in the MD simulation study. MCM C8

ATO 5

PE/F 87

COS

STA

CTAB

(% w/v)

(% w/v)

(% w/v)

(% w/v)

(% w/v)

(% w/v)

0.9

1

-

-

-

0.9

1

0.02

-

-

0.6

0.9

1

-

0.02

-

0.6

0.9

1

-

-

0.02

0.6

COS-NLC

0.6

STA-NLC CTAB-NLC

EP T

NLC

ED

Batch

AC C

MCM C8= Capmul MCM C8; ATO 5= Precirol ATO 5; PE/F 87= Synperonic PE/F 87; COS= Chitosan oligosaccharide; STA= Stearylamine; CTAB= Cetrimonium bromide Briefly, Capmul MCM C8 and Precirol ATO5 were weighed and transferred to the vial along with either STA or CTAB, except in the case of chitosan oligosaccharide, which was added separately as described in the following paragraph. Simultaneously, a 1% w/v solution of surfactant was prepared by dissolving Synperonic PE/F 87 in double-distilled water. Both the components were then heated to 70 °C. The hot surfactant solution was then added wholly to the hot lipid blend, under agitation on a vortex mixer (Remi, India) for 30 seconds to form a coarse emulsion. The coarse emulsion was then subjected to ultrasonication using a probeultrasonic cell disruptor (Dakshin Instruments, India) operated for 2 min at 200 W. Page 6 of 39

ACCEPTED MANUSCRIPT Subsequently; the vial was allowed to cool down to room temperature to form the NLC. The NLC were stored in stoppered vials at 25º C throughout the study. Coumarin-6 (Cou-6) loaded NLC were also prepared similarly by dissolving 0.01% w/v of Cou-6 in the oil phase of the NLC while keeping the rest of the procedure same. The preparation of chitosan oligosaccharide coated NLC was carried out by adding 1 ml of the COS solution, drop-wise to 10 ml of NLC, under continuous stirring at 400 rpm (IKA,

PT

Germany). The mixture was then allowed to stir together at room temperature for a period of 60 min.

RI

Particle size analysis

SC

The particle size of the NLC was evaluated by dynamic light scattering (DLS) using a Malvern Zeta Sizer (Malvern Instruments Ltd, UK) in deionised water. The calculations based on light scattering intensity were performed by software from the correlation functions

NU

using pre-defined algorithms.

MA

Small-angle neutron scattering (SANS)

The study was executed on the SANS diffractometer in Dhruva reactor at Guide Tube Laboratory, Bhabha Atomic Research Centre, Mumbai, India. The samples were poured in a

ED

0.5 cm path-length UV grade quartz sample holder with tight-fitting Teflon stoppers and sealed with parafilm. The sample to detector distance was kept constant at 1.8 m for all the runs, and the samples were bombarded with a monochromatized beam of 5.2 Å with a spread

EP T

of Δλ/λ ~ 15%. The angular distribution of the scattered neutrons was recorded using an indigenously built a 1 m long one-dimensional helium position sensitive detector (PSD). The PSD is filled with helium gas and works on the principle of charge division and allows a

AC C

simultaneous recording of the data over the Q range of 0.017–0.35 Å-1 . The temperature in all the measurements was kept fixed at 30 ºC [23]. Scattering intensities from the sample solutions were corrected for detector background and sensitivity, empty cell scattering, and sample transmission and solvent intensity was subtracted from that of the sample, and the resulting corrected intensities were normalised to absolute cross-section units. The actual and modelled data points obtained from SANS were fitted for appropriate models using SASfit. In vitro mucin binding by zeta potential measurements The zeta potential of the NLC formulations was evaluated in deionised water and in the presence of type-III mucin. Briefly, an accurately weighed quantity of mucin was transferred to phosphate buffer, pH 6.8 to get a 0.5% w/v solution. Then, 1 mL of this solution was Page 7 of 39

ACCEPTED MANUSCRIPT mixed with an equal quantity of the respective NLC and allowed to stir for 1 hour, followed by zeta potential measurements. Zeta potential was determined for the NLC in both deionised water and mucin using a Zeta Sizer Nano ZSP (Malvern, UK). For determination of zeta potential, the samples were suspended in water or mucin and placed in disposable polystyrene electrophoretic cells. The total zeta runs were 100, and the count rate was 250 particles/sec. Zeta potential of plain 0.5% w/v type-III mucin solution was also measured under similar

PT

conditions. In vitro mucin binding efficiency by spectrophotometry

RI

The in vitro mucin binding efficiency of the prepared NLC formulations was evaluated with type-III mucin by modifying a previously reported spectrophotometric method [24]. An

SC

accurately weighed quantity of mucin was dissolved in phosphate buffer, pH 6.8 to obtain a concentration of 0.5 % w/v. Then, 1 ml of respective NLC formulation was added to 0.5 ml

NU

of mucin solution and incubated at 37 ºC for 2 h in an incubator (IKA, Germany). The NLC dispersions were then centrifuged (Thermo Scientific, USA) at 20 ºC at 35,000 RPM for 2 h.

MA

The supernatant was isolated, and the free mucin content in the supernatant was estimated, and the spectra were recorded by UV-Visible spectrophotometer (Jasco, Japan) at 255 nm. The samples were analysed using the respective blanks prepared under similar conditions,

ED

and the mucin binding efficiency was then calculated by using Equation 1 [24].

EP T

Equation 1

( )

AC C

In vitro mucoadhesion by nanoparticle tracking analysis (NTA) Mucoadhesion was assessed with respect to the diffusion coefficient by using the nanoparticle tracking analyser (Nanosight, Malvern), having a C-mount microscope equipped with a tracking camera to record the motion of the particles. For NTA, 0.5 ml of the NLC formulation was suspended in an equivalent amount of 0.5 % w/v mucin solution (1.8 cps viscosity). The samples were then loaded into the prism cell illuminated with an 80-micron wide beam of a red laser. The video of the laser scatter due to the Brownian motion of NLC were captured over multiple frames at 30 fps for a time period of 90 seconds by the attached charged coupled device (CCD) camera. The captured video was then processed using NTA software version 2.3 with respect to diffusion coefficient as the measured parameter at 25º C and a viscosity of 1.8 cps. The procedure was also performed with all the formulations using Page 8 of 39

ACCEPTED MANUSCRIPT de-ionised water (0.993 cps viscosity) in place of mucin to serve as the baseline reading. The relative decrease in the diffusion coefficient of the particles in mucin as compared to that in deionised water served as a measure of the mucoadhesive potential of the particles. In vivo ocular mucoadhesion by confocal imaging Ocular mucoadhesion was evaluated on Sprague Dawley rats by confocal microscopy after administration of Cou-6 loaded NLC formulations. The protocol (BVC/IAEC/1/2018) was

PT

approved by the Institutional Animal Ethics Committee of the Bombay Veterinary College, Mumbai. Briefly, 20 µL of the respective Cou-6 loaded NLC were administered topically

RI

onto the eyes of the rats in the conjunctival sac. After a period of 30 min, the rats were sacrificed, the eyes were surgically removed and gently washed with 5 ml of phosphate-

SC

buffered saline, pH 7.4 with the help of a syringe. The excised eyes were then fixed overnight in 10% neutral buffered formalin solution, followed by sequentially immersing them in 10%,

NU

20% and 30% w/v sucrose solution overnight, till they sank to the bottom. The eyes were then embedded in optimal temperature cutting compound and flash-frozen in liquid nitrogen.

MA

5 µm thick sagittal sections of the corneoscleral region at the centre of the eye, along the limbus, were sliced using a cryostat (Leica Biosystems) and observed under a confocal microscope (LSM510-Meta System, Carl Zeiss) with an attached CCD camera and green

ED

fluorescence filter for Cou-6. The fluorescence intensity of Cou-6 was also evaluated using ImageJ with mean density as the parameter [25].

EP T

Statistical analysis and data interpretation All the data obtained were expressed as the mean ± standard deviation (SD). Data were analysed using two-way ANOVA wherever applicable, followed by post hoc Sidaks’s

AC C

multiple range test using Graph Pads Prism statistical analysis software, version 7.0. P < 0.05 was statistically significant (P < 0.05*; P < 0.01**; P < 0.001***; P < 0.0001****). Results and Discussion

Atomistic molecular dynamics simulation with mucin The main intermolecular interactions between mucins and other molecules are believed to be hydrogen bonding, hydrophobic interactions and electrostatic interactions, all of which can be well predicted between two entities using MD simulations [1,3]. The ligands carry a net positive charge owing to either -NH3 + or -N+ present in their protonated states as depicted in Figure 1. This characteristic makes them good candidates to adhere to the mucins, which have an overall negative charge owing to the ample aspartic acid and glutamic acid residues Page 9 of 39

ACCEPTED MANUSCRIPT [3]. Since the ligands also show hydrogen bond accepting potential, they are expected to undergo hydrogen bond interactions with the different amino acid residues on mucin as well. Supplementary Video S1 shows the trajectories of the 10 ns MD simulation of MUC4 with COS, STA and CTAB while Figure 2 shows the respective frames captured every 2 ns. The red and blue surfaces over the MUC4 depict the negatively and positively charged regions

MA

NU

SC

RI

PT

respectively, corresponding to the charged amino acid residues present.

Figure 1. Detailed structures of (a) Chitosan oligosaccharide (b) Stearylamine and (c)

charge in the protonated state

ED

Cetrimonium bromide, showing nitrogen and oxygen centres with the corresponding net

EP T

From the images, taken at every 2 ns, the intermolecular contact was highest for COS, and it was found to be in contact with MUC4 almost throughout the simulation time of 10 ns, followed by STA and CTAB. The different types of interactions occurring between MUC4

AC C

and the ligands, as predicted by the simulation have been depicted by dashed tubes in the frames for each ligand in Figure 2 and multiple contacts, if any are seen as overlapped dashed tubes. COS was able to form simultaneous multiple contacts which included hydrogen bonding, ionic interactions and hydrophobic interactions while for STA and CTAB, the occurrences of interactions were less and limited to mostly one interaction at any given time.

Page 10 of 39

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 2. Image stack of frames obtained after every 2 ns during the analysis of the 10 ns

ED

trajectory for MD simulation with MUC4, showing contacts if any for MUC4 with chitosan oligosaccharide, stearylamine, cetrimonium bromide and glyceryl stearate citrate throughout the course of the simulation. (See Supplementary Video S1 for the composite of all 1000

EP T

frames and interactions) The red and blue areas on MUC4 correspond to the negatively and positively charged region, respectively. The hydrogen bonding, ionic interactions, hydrophobic interactions or water bridges as depicted by yellow dashed tubes for hydrogen

AC C

bonding, purple dashed tubes for ionic interactions and by sky blue dashed tubes for hydrophobic contact (See Supplementary Figure S2). Water molecules have been deselected for the purpose of visualisation. (For interpretation of the references to colour in this figure legend, the reader is advised to refer to the web version of this article) Figure 3 shows the interaction fractions of different MUC4 residues for the 4 main types of interactions which have been normalised over the course of the simulation. The values suggest the extent to which a particular interaction is maintained at a given residue. The interactions have been expressed as a stacked bar chart wherein each coloured stack indicates the fraction of the total interactions observed at the corresponding protein residue for that ligand. Values over 1 indicate that some protein residues have made multiple contacts of the Page 11 of 39

ACCEPTED MANUSCRIPT same subtype with the ligand. COS showed several occurrences of hydrogen bonding between -NH3 +, -O- and many residues on MUC4. This was evident, given the 3 centres of NH3 + present in the chitosan oligosaccharide molecule, which increases its hydrogen bond donor/acceptor potential [26]. Also, as predicted, the negatively charged aspartate residue interacted heavily with the cationic -NH3 + centres through ionic interactions. Since the simulation was performed in an

PT

aqueous environment of TIP3P water model, occurrences of water bridges on some of the MUC4 residues were also observed. Water bridges are a type of hydrogen bonding

RI

interaction between the protein and ligands, mediated through the water molecules. The hydrophobic interaction was found to have the least occurrence and was only limited to the

AC C

EP T

ED

MA

NU

SC

tyrosine residue, and this low prevalence could be due to the hydrophilic nature of the ligand.

Figure 3. Simulation interaction diagram obtained after analysis of the 10 ns MD simulation trajectories for protein-ligand (PL) contacts at different residues on MUC4 for (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide showing hydrogen bonding, ionic interactions, hydrophobic contacts and water bridges Similarly, for STA, it was found that hydrogen bonding predominated over other interactions, followed by ionic and hydrophobic interactions, which could be attributed to the presence of Page 12 of 39

ACCEPTED MANUSCRIPT NH3 + like in case of COS. Although the interactions were spread over a vast number of residues of MUC4 as compared to COS, the magnitude of the interactions was far less compared to COS. Perhaps this could be due to the extended 18 carbon fatty acid chain, resulting in steric hindrances while interacting with MUC4. On the other hand, CTAB showed virtually no hydrogen bonding, including water bridges throughout the simulation

AC C

EP T

ED

MA

NU

SC

RI

PT

time.

Figure 4. Cumulative values represented as percentages for the different interactions between MUC4 and (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide, irrespective of the residues at which they occurred and the prevalence of the respective interactions, namely (d) hydrogen bonding (e) ionic interaction (f) hydrophobic interaction and (g) water bridges in the corresponding ligands. For CTAB, only ionic and hydrophobic interactions were found to occur throughout the simulation and were restricted to a mere 4 residues of MUC4. Moreover, the fraction of the interactions was far less when compared to COS or STA, except for fraction for hydrophobic contacts, which was found to be higher than that for COS and STA. Although like STA, only one nitrogen in the -N+ state was present in the CTAB and thus the interactions were Page 13 of 39

ACCEPTED MANUSCRIPT expected to be similar in magnitude to that of STA. However, the interactions were found to be negligible with hardly any contacts for most of the simulation time. This negligible interaction of CTAB with MUC4 could be due to the presence of 3 -CH3 groups surrounding the -N+ (Figure 1), thus further sterically hindering its interaction with potential residues on

AC C

EP T

ED

MA

NU

SC

RI

PT

the protein as compared to STA.

Figure 5. Timeline representation of all protein-ligand contacts summarised for (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide. The residues which make more than one specific contact with the ligand, are represented by a darker shade of orange according to the scale to the right of the plot. (For interpretation of the references to colour in this figure legend, the reader is advised to refer to the web version of this article) Page 14 of 39

ACCEPTED MANUSCRIPT In order to get a better understanding of the interaction fraction of the screened ligands compared to one another, the values from Figure 3 for the respective ligands, were broken down into individual interactions and expressed as a doughnut plot in Figure 4a, b and c, irrespective of the residue at which they occurred. The interaction fraction values have been expressed as cumulative values in terms of percentages. COS and STA interacted with the MUC4 through all the types of interactions mentioned earlier, whereas CTAB only interacted via hydrophobic and ionic interactions with MUC4. Among the different interactions in the

PT

individual ligand, hydrogen bonding was found to be more prevalent for both COS and STA while interacting with MUC4, accounting for about 80% of the total interactions while for

RI

CTAB, hydrophobic and ionic interactions were more prevalent. However, with respect to the

SC

individual interactions when compared to their occurrence in different ligands, all the interactions were found to be more prevalent in COS followed by STA as seen from Figure

NU

4d-g, except for hydrophobic interactions which were slightly higher for CTAB than for COS and STA.

MA

The overall outcome of the 10 ns simulation with respect to protein-ligand contacts has been summarised in Figure 5, which gives a timeline representation of the number of contacts between MUC4 and the respective ligands. The top panel in blue shows the total number of

ED

specific contacts that MUC4 makes with respect to time while the bottom panel in orange shows the specific residues that interact with the ligand in each trajectory frame. The plot

EP T

indicates extensive contact between MUC4 and COS throughout the 10 ns simulation including simultaneous multiple contacts at the same time for many frames, followed by STA which showed significant time between consecutive contacts and CTAB which showed only

AC C

intermittent contact and were limited to only one contact at any given time. The timeline in Figure 5 can be well understood if viewed simultaneously with Supplementary Video S1, which is a composite of all 1000 frames captured during the 10 ns simulation. The video also reiterates results obtained from simulation interaction diagram that COS was in contact with MUC4 starting from 1.2 ns till the end of the simulation via multiple interactions or contacts at the same time point. While STA also showed significant contact time with MUC4, the number of interactions were only limited to 1-3 in some frames with occasional detachments in between. On the other hand, CTAB remained in contact with MUC4 only intermittently. To further demonstrate the ability of MD simulation in being responsive to interactions critical to mucoadhesion, particularly the ionic interactions, a simulation was run with a negatively charged ligand like glyceryl stearyl citrate. Bansil et al have reported that the Page 15 of 39

ACCEPTED MANUSCRIPT mucins can have a tendency of being anti-adhesive towards negatively charged species [3]. Thus, the simulation between MUC4 and GSC was expected to proceed with minimal contact. As seen from Figure 2, Figure 5 and Supplementary Video S1, virtually no contact was observed between GSC and MUC4 throughout the simulation time of 10 ns for any of the said interactions. Moreover, since GSC also contains -O- like in the case of COS, it was expected that hydrogen bonding interactions would be predominant. However, none were recorded, which could suggest that the electrostatic repulsive forces could be acting more

PT

strongly as compared to the weak hydrogen bonding, thereby hindering any contacts through other forces. Also, since no contacts or interactions were recorded during the trajectory

RI

analysis, the interaction fraction and contact plots could not be generated for this run.

SC

The overall outcome of the MD simulation study suggests the influence of the functional group on the interactions, mainly the hydrogen bonding and electrostatic. Wang et al. have

NU

reported an extensive study with chitosan oligosaccharide wherein they studied the interaction of polynucleotides with chitosan oligosaccharide having a different substitution at

form

stronger

electrostatic

MA

the nitrogen, namely -NH3 +, -NH2 + and -N+ [26]. The authors observed that -NH3 + was able to attractions

with

the

negatively

charged

-PO 4 -

on

the

polynucleotide as compared to -NH2 + and -N+. Although the ligands used in our study are

ED

different, our findings are on similar lines with respect to the functional groups, wherein we also observed appreciable electrostatic interactions for COS and STA having -NH3 + than

EP T

CTAB which had a -N+. Further, Wang et al. have also postulated that the increased electrostatic interactions might also be responsible for the hydrogen bonding arising out of the enhanced contact between the substrate and the ligand owing to the -NH3 +--PO4 - contacts,

AC C

which again decreased from ligands having -NH3 + to ligands having -N + [26]. Our MD simulation also yielded a similar result wherein COS, having 3 centres of -NH3 + was able to form a greater extent of hydrogen bonds with MUC4 than CTAB which had the -N + and remain in contact with MUC4 for a longer. The in silico study was able to predict interactions between the mucin and the different ligands used. Moreover, based on the differences in the ligands, the simulation was found to respond differently, giving interactions that differed significantly and could be related to the physicochemical properties of the ligands used. in silico studies must be validated with relevant biological studies in vitro and in vivo. However, direct assessment of the ligands as such in their molecular state was beyond the scope of the techniques used to asses mucoadhesion due to their instrumental limitations pertaining to the size of the ligands at the Page 16 of 39

ACCEPTED MANUSCRIPT molecular level. Hence, for this reason, the ligands had to be embedded in a carrier system to act as a substrate so that the interactions of the ligands could be then easily quantified indirectly at the particulate level. Thus, for this purpose, the ligands used in the MD simulation were incorporated into NLC and evaluated by relevant studies to evaluate mucoadhesion and ocular retention to support the in silico findings. Preparation of NLC and particle size measurement was prepared by the emulsification-ultrasonication method, which involves the

PT

NLC

formation of a primary emulsion using a suitable surfactant system followed by a size

RI

reduction of the droplets by application of ultrasonic energy [22]. The NLC prepared by the method appeared stable visually, and as seen from Table 3, all the NLC were within the sub-

SC

200 nm particle size range.

Table 3. The particle size of the NLC incorporated with the different ligands used in the MD

NU

simulation study and their corresponding fluorescence intensity obtained from images of confocal micrographs after administration topically over the eyes. Fluorescence Intensity

(nm)

(x 105 RFU)

181.7 ± 22.8

1.15

ED

NLC COS-NLC

EP T

STA-NLC CTAB-NLC

Particle Size

MA

Batch

169.2 ± 8.6

37.86

153.9 ± 18.5

24.80

139.7 ± 13.9

2.91

RFU= Relative Fluorescence Units

AC C

The differences in the particle size could be due to the incorporation of either COS, STA and CTAB, which may act as stabilisers through electrostatic mechanisms to varying extents [27]. Small-angle neutron scattering (SANS) NLC was subjected to small-angle neutron scattering in order to understand its particulate nature and surface characteristics. It provides accurate information on the structure, shape and size of the scattering particles in the size scale of 10-1000 Å [28,29]. The SANS data of the surfactant-stabilised NLC systems were analysed using a model of spherical particles while the COS coated NLC was analysed by a spherical core-shell model to account for the COS coating. The data plots obtained were analysed by comparing the scattering from the models to the obtained experimental data. Appropriate corrections were Page 17 of 39

ACCEPTED MANUSCRIPT made for instrumental smearing, where the calculated scattering profiles were smeared by the appropriate resolution function to compare with the measured data. The fitted parameters for the analysis were then optimised using a nonlinear least-square fitting program to the model scattering. A typical SANS spectrum shows certain distinct regions like the Guinier region towards the lower Q values and the Porod region towards the higher Q values. While the Guinier region

PT

gives information regarding the radius of gyration of the particles which can be used to make approximations of its particle size, the Porod region gives considerable information regarding

RI

the shape, interparticle distances and conformations of the particles [30]. Scattering curves obtained by the analysis of actual data points of these formulations are shown in Figure 6.

SC

All the formulations showed a scattering curve exhibited by a typical particulate system [28], which confirms the nearly spherical nature of the formed NLC, as observed from the slope of

AC C

EP T

ED

MA

NU

the scatter in the Guinier region of the plot at lower Q values.

Figure 6. The scattering intensities of experimental and fitted data points obtained after small-angle neutron scattering (SANS) for (a) COS NLC (b) STA NLC and (c) CTAB NLC, along with the graphical representation of the formed NLC in the inset. (d) An overlay of the SANS data for COS, STA and CTAB NLC, depicting the differences in the scattering pattern. Page 18 of 39

ACCEPTED MANUSCRIPT Moreover, for COS-NLC, a slight deflection was observed towards in the Porod region of the scatter, indicating a core-shell type of structure [29,31]. This could be due to the COS fibres cluttering around the NLC, forming a core and shell form of a structure. Overlay of the SANS spectra in Figure 6d for STA, CTAB and COS NLC clearly shows the differences in the scattering pattern arising due to the presence of the COS coat around the NLC for COS-NLC, whereas, for STA and CTAB NLC, the spectra appear identical and correspond to a

PT

particulate model. In vitro mucin binding by zeta potential measurements

RI

The charge on the surface of a nanocarrier is an important parameter which affects the adhesion to mucin as electrostatic forces are one of the key mechanisms of mucoadhesion and

SC

the same was observed in the simulation studies from the simulation with GSC [32,33]. Many authors have previously reported the requirement of a cationic charge on the particles for

NU

enhanced mucoadhesion as compared to negligible adhesion for neutral or negatively charged particles [34,35]. Thus, a positive charge could be a prerequisite for good mucoadhesive

AC C

EP T

ED

MA

properties of nanocarriers.

Page 19 of 39

ED

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 7. Comparative bar plot representing zeta potential values of the respective NLC

EP T

formulations in deionised water and 0.5 % w/v type – III mucin solution. (p > 0.05) Plain mucin was found to have a net negative charge over it owing to the abundance of

AC C

aspartate and glutamate residues in its residue composition, which exist as anionic residues at physiological conditions [36]. All the NLC gave a positively charged NLC in deionised water, owing to the protonated -NH3 + and N + present as indicated by the zeta potential values of 19.8 ± 1.4 mV, 21.5 ± 2.6 mV and 28.6 ± 3.6 respectively, as seen from Figure 7. On the other hand, blank NLC exhibited a slight negative charge (-10.2 ± 3.5 mV), which could be attributed to the lipid present (Glyceryl distearate). Zeta potential was also measured in the presence of a semi-synthetic mucin in order to evaluate the effect of the interaction on the overall charge of the systems, and the changes have been shown as a comparative bar plot. All the systems still gave a positive charge, however, the magnitude of zeta potential was reduced significantly for COS and STA NLC followed by a slight decrease for CTAB NLC indicating the corresponding extent of binding Page 20 of 39

ACCEPTED MANUSCRIPT with the mucin. Perhaps, the differences in binding could also be since mucoadhesion was a function of not only electrostatic forces but also other factors hydrogen bonding as observed in the trajectories. No significant differences were found in the zeta potential of plain NLC in the presence of deionised water or artificial mucin, suggesting no significant interaction between the two. It could be noted that although CTAB NLC showed the highest zeta potential value, however

PT

the corresponding reduction in zeta potential value after its interaction with the mucin was not of the same magnitude. This could be related to the orientation of the -CH3 groups

RI

surrounding the -N+ thereby sterically hindering the interaction with mucin as was observed

SC

in the trajectories obtained from the MD simulation in Supplementary Video S1. In vitro mucin binding efficiency by spectrophotometry

Many well-established methods are available for evaluating mucoadhesion of dosage forms

NU

like tablets, gels and films. These methods include the use of excised mucous membranes mounted on mucoadhesion apparatus and texture analyser [37–39]. Although precise and

MA

robust, these methods may not be suitable for evaluating nanoparticles, microparticles or other nanocarriers in the nano-size range [40]. Thus, the focus has been shifting to other techniques which are based on inherent characteristics of the mucous, like spectral properties, surface characteristics, diffusion coefficient etc. in order to estimate their

ED

viscosity,

interaction with nanoparticles [41,42]. One of the simplest methods that have been used for

shows

an

EP T

evaluating the interaction of nanoparticles with mucin is spectroscopic studies [43]. Mucin absorption

maximum

at

255

nm

AC C

spectrophotometrically detected and quantitated as well.

Page 21 of 39

(Figure

8a)

and

thus

can

be

AC C

EP T

ED

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 8. (a) Ultraviolet spectra of unbound mucin in the supernatant after incubation with the NLC and the corresponding (b) Mucin binding efficiency calculated in percentage, indicating the extent of binding of the respective NLC to mucin. (p > 0.05)

Page 22 of 39

ACCEPTED MANUSCRIPT In the study, the NLC of COS, STA and CTAB were incubated with type-III mucin to allow for interactions between them. Depending upon the different interactions possible, the mucin would bind to the NLC to a different extent and can be quantified spectrophotometrically at 255 nm by estimating the amount of free mucin in the supernatant obtained after centrifugation [24].

The bound

mucin adheres to the NLC and sediments during

centrifugation therefore not interfering in the spectrophotometric measurements.

PT

Figure 8a depicts the overlay of plots obtained after spectrophotometrically analysing the supernatants at 255 nm. The corresponding absorbance values were then used in Equation 1

RI

to obtain mucin binding efficiency in percentage, which is compared to blank NLC in Figure 8b. The results indicate that among the NLC evaluated, plain NLC showed the least binding

SC

efficiency of 4.23 ± 1.2% when compared with the other NLC. Among COS, STA and CTAB NLC, COS NLC showed the highest binding efficiency of 79.14 ± 6.7%, followed by STA

NU

(67.83 ± 5.6%) and CTAB NLC (55.27 ± 7.1%). Although all the NLC had nearly similar extent and nature of the cationic charge, their binding efficiencies were varying. This could

MA

be since the binding efficiency is also a function of other interactions, as discussed earlier. In vitro mucoadhesion by nanoparticle tracking analysis (NTA) Nanoparticle tracking analysis was originally introduced as a technique for measurement of

ED

particle size and its distribution. However, over the period of time, several applications have emerged for NTA, which includes protein aggregation studies, refractive index and

EP T

microrheology [44–46]. In recent years, NTA has also been proposed for application of mucoadhesion measurements or mucin-particle interactions based on its diffusion coefficient of the nanoparticles across mucosal barriers and have been reported to be used for colonic

AC C

and pulmonary adhesion predictions [40,42,43,47]. Martens et al. have used the method for studying the mobility of particles in the vitreous fluid; however, the use of NTA with ocular mucins to predict ocular mucoadhesion could be explored [48]. Estimation of binding efficiency by NTA is based on measurement of diffusion coefficients of rapidly moving particles (Brownian motion) within a mucosal barrier. The rapid Brownian motion of the nanoparticles is analysed using a CCD camera capable of operating at high frame capture speeds. The diffusion coefficient is measured by algorithms in NTA 2.3 software with the help of Stokes-Einstein’s equation, as described in Equation 2 [49]. Equation 2

Page 23 of 39

ACCEPTED MANUSCRIPT

Where, Dt is diffusion coefficient, Kb is Boltzmans constant,

PT

T is temperature,

RI

η is solvent viscosity, r is hydrodynamic radius.

SC

Thus, from the Equation 2, it can be said that, for a given particle of fixed radius, the diffusion coefficient measured at a constant temperature in different media of same

NU

viscosities, would be nearly similar, provided that there are no other interactions. However, when the nanoparticles are analysed using viscous barriers like mucin, their motion would

MA

also depend upon the extent of different types of interactions possible, i.e. hydrogen bonding, electrostatic interactions, hydrophobic interactions and physical entanglements [48]. This would result in differences in the values of diffusion coefficients for different particles based

ED

on the extent of their interaction. Thus, the diffusion coefficient of the NLC was first measured in deionised water to serve as the baseline reading as the diffusion would then only

EP T

be a function of the size of the NLC since the other parameters were constant. The same set of experiment was then run with mucin in place of deionised water. This time the diffusion of the particles was not only a function of their size but also their interaction with mucin

AC C

depending on their surface characteristics. The significance of the differences in values of the diffusion coefficient of the respective NLC in deionised water and mucin can be correlated with the extent of interaction of the individual NLC. A particle with negligible or no interaction with mucin was expected to result in similar diffusion coefficient, whereas for a particle with greater extent of interaction with mucin, the reduction in values of diffusion coefficient was expected to be significant as compared to those obtained in deionised water. Martens et al. in their study of the mobility of surface-modified nanoparticles in vitreous have observed that in a viscous media, the effect of surface characteristics on the mobility of the individual particles was far greater than that of the particle size.

Page 24 of 39

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

ED

Figure 9. Comparative bar plot of diffusion coefficients obtained from nanoparticle tracking analysis of the NLC in deionised water and mucin. The significance of the differences

EP T

between the two values for the corresponding NLC indicates the extent of mucoadhesion of the with the mucin. (p > 0.05)

Figure 9 depicts the diffusion coefficient of plain NLC and COS, STA and CTAB NLC in

AC C

deionised water and mucin solution. Although recorded in the same media (Deionised water), the diffusion coefficient of the NLC was significantly different from each other. Plain NLC showed a low diffusion coefficient value of 449.5 x 104 ± 70.04 nm2 /s owing to the larger particle size as compared with other NLC, which ranged from 600-700 x 104 nm2 /s. The results might suggest that there was an absence of any interaction between the media and the NLC, irrespective of their surface characteristics. However, when the NLC were analysed in mucin solution, significant differences were found between their values as compared to diffusion in deionised water and from each other. The diffusion coefficient for all the NLC in mucin was lower to that of deionised water, which was expected due to the higher viscosity of the mucin (1.8 cps) as compared to deionised water (0.933 cps). However, among individual comparison, it was found that there was a significant reduction in the diffusion Page 25 of 39

ACCEPTED MANUSCRIPT coefficient of COS NLC in mucin (~1.5 fold) as compared to deionised water. On the other hand, both STA and CTAB NLC showed only a marginal decrease in the diffusion coefficient in mucin as compared to that of deionised water. Chuah et al. have previously reported evaluating chitosan oligosaccharide nanoparticles using NTA [47]. Although the authors have focussed more on the aggregation phenomenon of nanoparticles in the presence of mucin, they have also concluded that chitosan oligosaccharide nanoparticles showed

PT

significantly higher mucin binding than the plain ones. In vivo ocular mucoadhesion by confocal imaging

RI

The ocular surface is lined with mucus-secreting cells, extending up to the conjunctiva. Moreover, the tear film secreted over the eye surface is abundant with mucins and therefore,

SC

an excellent substrate for materials having the potential for mucoadhesion like those evaluated in the present study [6,50]. The in vivo ocular mucoadhesion test was performed on

NU

Sprague-Dawley rats by instillation of Cou-6 loaded NLC of COS, STA and CTAB. The experiment was performed by slightly modifying the protocol reported previously and also

MA

compared to Cou-6 NLC [51]. The confocal images were recorded on the sagittal section of the rat eyes, parallel to the corneal epithelium. There have been several reports wherein ocular permeation has been reported to occur after a period of 15-30 min [52–54]. Since the

ED

aim of the study was to evaluate only retention or mucoadhesion at the ocular surface, rats were sacrificed at the end of 30 min to minimise any permeation within the eye tissues and

EP T

give a clear indication of surface adhesion alone [55]. Figure 10 shows the fluorescence micrographs obtained for plain NLC and COS, STA and CTAB NLC through optical and fluorescence filters along with their corresponding overlay.

AC C

The images show that negligible fluorescence was observed for plain NLC.

Page 26 of 39

AC C

EP T

ED

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

Figure 10. Confocal laser scanning micrographs of the rat eyes showing mucoadhesion at the corneal-scleral region, 30 min after topical instillation of coumarin-6 incorporated (a) Blank NLC (b) COS NLC (c) STA NLC and (d) CTAB NLC. The images were captured in two channels, fluorescence (Left), optical (Centre) and the merged results of the two channels are depicted on the right. The red arrows indicate the upper side or the corneal epithelium and the blue arrows indicate the lower side or the corneal endothelium of the eyes. Page 27 of 39

ACCEPTED MANUSCRIPT On the other hand, ocular sections after administration of COS and STA NLC showed the high intensity of fluorescence at the ocular surface. Although both the images appear of nearly equal intensities visually, analysis of the corresponding fluorescence intensity as given in Table 3 revealed that COS NLC showed 1.5-fold higher fluorescence than STA NLC, thus indicating greater mucoadhesion among the two. The images obtained after CTAB NLC administration showed a very weak fluorescence as compared to COS and STA NLC and the same was observed in the corresponding intensity values obtained, which were significantly

PT

lower than those obtained for COS and STA NLC.

RI

Although the results of the in vivo mucoadhesion study obtained by confocal imaging was not entirely in line with that of the in vitro mucoadhesion by spectrophotometry which showed a

SC

marginal interaction of mucin with CTAB, the findings however aligned well with that of in silico studies which had predicted a similar fate for CTAB i.e. low interaction with the mucin

NU

while showing enhanced contact time with MUC4 for COS and STA. The results of the in vivo mucoadhesion study iterates that the interactions at the surface of the cornea between

MA

delivery systems and the mucin can be effectively predicted by in silico techniques like MD simulation with respect to the types of the interactions involved.

ED

Conclusion

Molecular dynamics simulation was sufficiently accurate in predicting the extent of adhesion of different materials to a biological entity like mucin-based on the different types of

EP T

interactions. The extent of interaction of the particular ligand could be well predicted by the in silico simulation study as noted in the results of in vitro and in vivo studies which were in the same order. Also, the MD simulation of a negatively charged ligand (GSC) with the

AC C

MUC4 did not show any type of contacts within the studied simulation time unlike the cationic ligands, highlighting the relevance of a cationic nature and perhaps suggesting that a cationic charge could be the pre-requisite for mucoadhesive interactions. However, the present study did not account for the physiological factors encountered during in vivo administration like tear turnover and clearance, which could significantly affect the retention of non-adhesive particles, and thus needs to be studied. Other extrinsic factors that may also influence ocular mucoadhesion like blinking and eye movements may also have to be factored in before a robust model could be developed to entirely predict the outcome in silico. Nevertheless, molecular dynamics can be used to partly screen and compare materials for mucoadhesion. Although the studies have been specific for the mucins secreted in the eye (Mucin-4), it could also be similarly applied to other mucoadhesive sites as well. Page 28 of 39

ACCEPTED MANUSCRIPT Declaration of interest The authors declare no conflict of interests for the above work. Acknowledgement The authors would sincerely like to thank Dr. Smita Mahale and Dr. Nafisa Balasinor at National Institute for Research in Reproductive Health (NIRRH), Mumbai for the use of cryostat and confocal microscope facilities. The authors would also like to thank Dr. Vijay

PT

Mendhulkar at the Institute of Science, Mumbai for the use of Nanosight and Dr. V. K. Aswal at the Bhabha Atomic Research Centre (BARC), Mumbai for small angle neutron scattering

RI

study.

[1]

SC

References

L. Wu, W. Shan, Z. Zhang, Y. Huang, Engineering nanomaterials to overcome the

NU

mucosal barrier by modulating surface properties, Adv. Drug Deliv. Rev. (2017). doi:10.1016/j.addr.2017.10.001.

X. Murgia, B. Loretz, O. Hartwig, M. Hittinger, C.-M. Lehr, The role of mucus on

MA

[2]

drug transport and its potential to affect therapeutic outcomes., Adv. Drug Deliv. Rev.

[3]

ED

(2017). doi:10.1016/j.addr.2017.10.009.

R. Bansil, B.S. Turner, Mucin structure, aggregation, physiological functions and biomedical applications, Curr. Opin. Colloid Interface Sci. 11 (2006) 164–170.

[4]

EP T

doi:10.1016/J.COCIS.2005.11.001.

R.R. Hodges, D.A. Dartt, Tear film mucins: front line defenders of the ocular surface;

AC C

comparison with airway and gastrointestinal tract mucins., Exp. Eye Res. 117 (2013) 62–78. doi:10.1016/j.exer.2013.07.027. [5]

X. Murgia, B. Loretz, O. Hartwig, M. Hittinger, C.-M. Lehr, The role of mucus on drug transport and its potential to affect therapeutic outcomes, Adv. Drug Deliv. Rev. (2017). doi:10.1016/j.addr.2017.10.009.

[6]

S.C. Pflugfelder, Z. Liu, D. Monroy, D.Q. Li, M.E. Carvajal, S.A. Price-Schiavi, N. Idris, A. Solomon, A. Perez, K.L. Carraway, Detection of sialomucin complex (MUC4) in human ocular surface epithelium and tear fluid., Invest. Ophthalmol. Vis. Sci. 41 (2000) 1316–26. http://www.ncbi.nlm.nih.gov/pubmed/10798646 (accessed December 21, 2017).

Page 29 of 39

ACCEPTED MANUSCRIPT [7]

F. Taherali, F. Varum, A.W. Basit, A slippery slope: On the origin, role and physiology

of

mucus,

Adv.

Drug

Deliv.

Rev.

124

(2018)

16–33.

doi:10.1016/J.ADDR.2017.10.014. [8]

I.D. Rupenthal, C.R. Green, R.G. Alany, Comparison of ion-activated in situ gelling systems for ocular drug delivery. Part 1: Physicochemical characterisation and in vitro release, Int. J. Pharm. 411 (2011) 69–77. doi:10.1016/J.IJPHARM.2011.03.042. Q. Luo, J. Zhao, X. Zhang, W. Pan, Nanostructured lipid carrier (NLC) coated with

PT

[9]

Chitosan Oligosaccharides and its potential use in ocular drug delivery system, Int. J.

cationic

nanostructured

pharmacodynamics,

lipid

J.

carriers:

Pharm.

doi:10.1016/j.xphs.2017.08.024.

Enhanced

Sci.

106

brain

distribution

(2017)

and

3613–3622.

Q. Ru, H.M. Fadda, C. Li, D. Paul, P.T. Khaw, S. Brocchini, M. Zloh, Molecular

MA

[11]

SC

P. Wavikar, R. Pai, P. Vavia, Nose to brain delivery of rivastigmine by in situ gelling

NU

[10]

RI

Pharm. 403 (2011) 185–191. doi:10.1016/J.IJPHARM.2010.10.013.

dynamic simulations of ocular tablet dissolution, J. Chem. Inf. Model. 53 (2013)

[12]

ED

3000–3008. doi:10.1021/ci4002348.

K.S. Vinay Kumar, G.S. Lingaraju, Y.K. Bommegowda, A.C. Vinayaka, P. Bhat, C.S. Pradeepa Kumara, K.S. Rangappa, D.C. Gowda, M.P. Sadashiva, Synthesis,

EP T

antimalarial activity, and target binding of dibenzazepine-tethered isoxazolines, RSC Adv. 5 (2015) 90408–90421. doi:10.1039/C5RA17926B. S. Chinnasamy, S. Nagamani, K. Muthusamy, Zn 2+ ion of the snake venom

AC C

[13]

metalloproteinase (SVMP) plays a critical role in ligand binding: a molecular dynamics

simulation

study,

RSC

Adv.

5

(2015)

70566–70576.

doi:10.1039/C5RA14693C. [14]

C.A. Rocha, P.R.S. Sanches, R. Marchetto, A. Zottis, The octahydroindene carboxyl substructure from dihydrobetulinic acid is essential to inhibit topoisomerase IB from leishmania donovani, J. Braz. Chem. Soc. 27 (2015) 591–598. doi:10.5935/01035053.20150295.

[15]

M. Yahyaei, F. Mehrnejad, H. Naderi-manesh, A.H. Rezayan, Follicle-stimulating hormone

encapsulation

in

the

cholesterol-modified

Page 30 of 39

chitosan

nanoparticles

via

ACCEPTED MANUSCRIPT molecular dynamics simulations and binding free energy calculations, Eur. J. Pharm. Sci. 107 (2017) 126–137. doi:10.1016/j.ejps.2017.07.007. [16]

J. Alvarez-Trabado, Y. Diebold, A. Sanchez, Designing lipid nanoparticles for topical ocular

drug

delivery,

Int.

J.

Pharm.

532

(2017)

204–217.

doi:10.1016/J.IJPHARM.2017.09.017. [17]

J. Alvarez-Trabado, A. López-García, M. Martín-Pastor, Y. Diebold, A. Sanchez,

PT

Sorbitan ester nanoparticles (SENS) as a novel topical ocular drug delivery system: Design, optimization, and in vitro/ex vivo evaluation, Int. J. Pharm. 546 (2018) 20–30.

N. Saini, R.K. Sodhi, L. Bajaj, R.S. Pandey, U.K. Jain, O.P. Katare, J. Madan,

SC

[18]

RI

doi:10.1016/J.IJPHARM.2018.05.015.

Intravaginal administration of metformin hydrochloride loaded cationic niosomes

NU

amalgamated with thermosensitive gel for the treatment of polycystic ovary syndrome: In vitro and in vivo studies, Colloids Surfaces B Biointerfaces. 144 (2016) 161–169.

[19]

MA

doi:10.1016/J.COLSURFB.2016.04.016.

J.F. Fangueiro, T. Andreani, M.A. Egea, M.L. Garcia, S.B. Souto, A.M. Silva, E.B. Souto, Design of cationic lipid nanoparticles for ocular delivery: Development, and

cytotoxicity,

ED

characterization

Int.

J.

Pharm.

461

(2014)

64–73.

doi:10.1016/J.IJPHARM.2013.11.025. P. Niamprem, S.P. Srinivas, W. Tiyaboonchai, Penetration of Nile red-loaded

EP T

[20]

nanostructured lipid carriers (NLCs) across the porcine cornea, Colloids Surfaces B

[21]

AC C

Biointerfaces. 176 (2019) 371–378. doi:10.1016/J.COLSURFB.2019.01.018. J.F. Fangueiro, A.C. Calpena, B. Clares, T. Andreani, M.A. Egea, F.J. Veiga, M.L. Garcia, A.M. Silva, E.B. Souto, Biopharmaceutical evaluation of epigallocatechin gallate-loaded cationic lipid nanoparticles (EGCG-LNs): In vivo, in vitro and ex vivo studies, Int. J. Pharm. 502 (2016) 161–169. doi:10.1016/J.IJPHARM.2016.02.039. [22]

M. Zheng, M. Falkeborg, Y. Zheng, T. Yang, X. Xu, Formulation and characterization of nanostructured lipid carriers containing a mixed lipids core, Colloids Surfaces A Physicochem. Eng. Asp. 430 (2013) 76–84. doi:10.1016/J.COLSURFA.2013.03.070.

[23]

V.K. Aswal, P.S. Goyal, Small-angle neutron scattering diffractometer at Dhruva reactor, Curr. Sci. 79 (2000) 947–953.

Page 31 of 39

ACCEPTED MANUSCRIPT [24]

S. Md, R.A. Khan, G. Mustafa, K. Chuttani, S. Baboota, J.K. Sahni, J. Ali, Bromocriptine loaded chitosan nanoparticles intended for direct nose to brain delivery: Pharmacodynamic, Pharmacokinetic and Scintigraphy study in mice model, Eur. J. Pharm. Sci. 48 (2013) 393–405. doi:10.1016/j.ejps.2012.12.007.

[25]

M. Fitzpatrick, Measuring cell fluorescence using ImageJ — The open lab book v1.0, Open

Lab

B.

(2014).

https://theolb.readthedocs.io/en/latest/imaging/measuring-cell-

[26]

PT

fluorescence-using- imagej.html (accessed July 7, 2018). J.-W. Shen, J. Li, Z. Zhao, L. Zhang, G. Peng, L. Liang, Molecular dynamics study on

RI

the mechanism of polynucleotide encapsulation by chitosan, Sci. Rep. 7 (2017) 5050.

[27]

SC

doi:10.1038/s41598-017-05197-0.

S. Zhao, X. Yang, V.M. Garamus, U.A. Handge, L. Bérengère, L. Zhao, G. Salamon,

NU

R. Willumeit, A. Zou, S. Fan, Mixture of nonionic/ionic surfactants for the formulation of nanostructured lipid carriers: Effects on physical properties, Langmuir. 30 (2014)

[28]

MA

6920–6928. doi:10.1021/la501141m.

B. Chu, T. Liu, Characterization of nanoparticles by scattering techniques, J.

[29]

ED

Nanoparticle Res. 2 (2000) 29–41. doi:10.1023/A:1010001822699. H. B., Probing nanoscale structures-the sans toolbox., Natl. Inst. Stand. Technol. Cent.

[30]

A.

Alford,

EP T

Neutron Res. Gaithersburg, MD. (2016). V.

Kozlovskaya,

E.

Kharlampieva,

Small

angle

scattering

for

pharmaceutical applications: From drugs to drug delivery systems, in: Springer,

[31]

AC C

Singapore, 2017: pp. 239–262. doi:10.1007/978-981-10-6038-0_15. S. Napso, D.M. Rein, Z. Fu, A. Radulescu, Y. Cohen, Structural analysis of cellulosecoated oil-in-water emulsions fabricated from molecular solution, Langmuir. 34 (2018) 8857–8865. doi:10.1021/acs.langmuir.8b01325. [32]

M. García-Díaz, D. Birch, F. Wan, H.M. Nielsen, The role of mucus as an invisible cloak to transepithelial drug delivery by nanoparticles, Adv. Drug Deliv. Rev. 124 (2018) 107–124. doi:10.1016/j.addr.2017.11.002.

[33]

G.E. Yakubov, S. Singleton, A.-M. Williamson, Methods for assessing mucoadhesion: The experience of an integrative approach, in: Mucoadhesive Mater. Drug Deliv. Syst., John Wiley & Sons, Ltd, Chichester, United Kingdom, 2014: pp. 197–232. Page 32 of 39

ACCEPTED MANUSCRIPT doi:10.1002/9781118794203.ch09. [34]

P. He, S.S. Davis, L. Illum, In vitro evaluation of the mucoadhesive properties of chitosan microspheres, Int. J. Pharm. 166 (1998) 75–88. doi:10.1016/S03785173(98)00027-1.

[35]

J. Shen, Y. Wang, Q. Ping, Y. Xiao, X. Huang, Mucoadhesive effect of thiolated PEG stearate and its modified NLC for ocular drug delivery, J. Control. Release. 137 (2009)

J. Leal, H.D.C. Smyth, D. Ghosh, Physicochemical properties of mucus and their

RI

[36]

PT

217–223. doi:10.1016/J.JCONREL.2009.04.021.

impact on transmucosal drug delivery., Int. J. Pharm. 532 (2017) 555–572.

[37]

SC

doi:10.1016/j.ijpharm.2017.09.018.

S.S. Darandale, P.R. Vavia, Design of a gastroretentive mucoadhesive dosage form of

NU

furosemide for controlled release, Acta Pharm. Sin. B. 2 (2012) 509–517. doi:10.1016/J.APSB.2012.05.004.

M.N. Wasnik, R.D. Godse, H.A. Nair, Development and evaluation of buccoadhesive

MA

[38]

tablet for selegiline hydrochloride based on thiolated polycarbophil, Drug Dev. Ind.

[39]

ED

Pharm. 40 (2014) 632–638. doi:10.3109/03639045.2014.884124. E. Cevher, M.A.M. Taha, M. Orlu, A. Araman, Evaluation of mechanical and

and

their

EP T

mucoadhesive properties of clomiphene citrate gel formulations containing carbomers thiolated

derivatives,

Drug

Deliv.

15

(2008)

57–67.

doi:10.1080/10717540701829234. A.-C. Groo, F. Lagarce, Mucus models to evaluate nanomedicines for diffusion, Drug

AC C

[40]

Discov. Today. 19 (2014) 1097–1108. doi:10.1016/J.DRUDIS.2014.01.011. [41]

J.Y. Lock, T.L. Carlson, R.L. Carrier, Mucus models to evaluate the diffusion of drugs and

particles,

Adv.

Drug

Deliv.

Rev.

124

(2018)

34–49.

doi:10.1016/J.ADDR.2017.11.001. [42]

J. Grießinger, S. Dünnhaupt, B. Cattoz, P. Griffiths, S. Oh, S. Borrós i Gómez, M. Wilcox, J. Pearson, M. Gumbleton, M. Abdulkarim, I. Pereira de Sousa, A. BernkopSchnürch, Methods to determine the interactions of micro- and nanoparticles with mucus., Eur. J. Pharm. Biopharm. 96 (2015) 464–76. doi:10.1016/j.ejpb.2015.01.005.

[43]

A.R. Mackie, F.M. Goycoolea, B. Menchicchi, C.M. Caramella, F. Saporito, S. Lee, K. Page 33 of 39

ACCEPTED MANUSCRIPT Stephansen, I.S. Chronakis, M. Hiorth, M. Adamczak, M. Waldner, H.M. Nielsen, L. Marcelloni,

Innovative

methods

and

applications

in

mucoadhesion

research,

Macromol. Biosci. 17 (2017) 1600534. doi:10.1002/mabi.201600534. [44]

X. Tian, M.R. Nejadnik, D. Baunsgaard, A. Henriksen, C. Rischel, W. Jiskoot, A comprehensive

evaluation

of

nanoparticle

tracking

analysis

(NanoSight)

for

characterization of proteinaceous submicron particles, J. Pharm. Sci. 105 (2016) 3366–

[45]

PT

3375. doi:10.1016/J.XPHS.2016.08.009. E. van der Pol, F.A.W. Coumans, A. Sturk, R. Nieuwland, T.G. van Leeuwen,

RI

Refractive index determination of nanoparticles in suspension using nanoparticle

[46]

SC

tracking analysis, Nano Lett. 14 (2014) 6195–6201. doi:10.1021/nl503371p. V. Filipe, A. Hawe, W. Jiskoot, Critical evaluation of nanoparticle tracking analysis

NU

(NTA) by NanoSight for the measurement of nanoparticles and protein aggregates, Pharm. Res. 27 (2010) 796–810. doi:10.1007/s11095-010-0073-2. L.H. Chuah, C.J. Roberts, N. Billa, S. Abdullah, R. Rosli, S. Manickam, Using

MA

[47]

nanoparticle tracking analysis (NTA) to

decipher mucoadhesion propensity of

curcumin-containing chitosan nanoparticles and curcumin release, J. Dispers. Sci.

[48]

ED

Technol. 35 (2014) 1201–1207. doi:10.1080/01932691.2013.800458. T.F. Martens, D. Vercauteren, K. Forier, H. Deschout, K. Remaut, R. Paesen, M.

intravitreal

EP T

Ameloot, J.F. Engbersen, J. Demeester, S.C. De Smedt, K. Braeckmans, Measuring the mobility

of nanomedicines

with

single-particle

tracking

microscopy,

[49]

AC C

Nanomedicine. 8 (2013) 1955–1968. doi:10.2217/nnm.12.202. NanoSight

NTA

2.3

analytical

software

operating

https://sydney.edu.au/medicine/bosch/facilities/molecular-biology/P554I

manual,

(n.d.).

NTA

2_3

Manual.pdf (accessed June 29, 2018). [50]

I.K. Gipson, The ocular surface: the challenge to enable and protect vision: the Friedenwald lecture., Invest. Ophthalmol. Vis. Sci. 48 (2007) 4390; 4391–8. doi:10.1167/iovs.07-0770.

[51]

D. Liu, J. Li, B. Cheng, Q. Wu, H. Pan, Ex vivo and in vivo evaluation of the effect of coating a coumarin-6-labeled

nanostructured

lipid

carrier with chitosan- N -

acetylcysteine on rabbit ocular distribution, Mol. Pharm. 14 (2017) 2639–2648.

Page 34 of 39

ACCEPTED MANUSCRIPT doi:10.1021/acs.molpharmaceut.7b00069. [52]

Y.-H. Weng, X.-W. Ma, J. Che, C. Li, J. Liu, S.-Z. Chen, Y.-Q. Wang, Y.-L. Gan, H. Chen, Z.-B. Hu, K.-H. Nan, X.-J. Liang, Nanomicelle-assisted targeted ocular delivery with enhanced antiinflammatory efficacy in vivo, Adv. Sci. 5 (2018) 1700455. doi:10.1002/advs.201700455.

[53]

K. Hironaka, Y. Inokuchi, Y. Tozuka, M. Shimazawa, H. Hara, H. Takeuchi, Design

PT

and evaluation of a liposomal delivery system targeting the posterior segment of the

[54]

RI

eye, J. Control. Release. 136 (2009) 247–253. doi:10.1016/J.JCONREL.2009.02.020. W. Zeng, Q. Li, T. Wan, C. Liu, W. Pan, Z. Wu, G. Zhang, J. Pan, M. Qin, Y. Lin, C.

Mucoadhesion,

precorneal

retention,

SC

Wu, Y. Xu, Hyaluronic acid-coated niosomes facilitate tacrolimus ocular delivery: aqueous

humor

pharmacokinetics,

and

doi:10.1016/j.colsurfb.2016.01.014.

MA

E.A. Mun, P.W.J. Morrison, A.C. Williams, V. V. Khutoryanskiy, On the barrier properties of the cornea: A microscopy study of the penetration of fluorescently labeled nanoparticles, polymers, and sodium fluorescein, Mol. Pharm. 11 (2014)

EP T

ED

3556–3564. doi:10.1021/mp500332m.

AC C

[55]

NU

transcorneal permeability., Colloids Surf. B. Biointerfaces. 141 (2016) 28–35.

Page 35 of 39

ACCEPTED MANUSCRIPT Figure captions Figure 1. Detailed structures of (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide, showing nitrogen and oxygen centres with the corresponding net charge in the protonated state Figure 2. Image stack of frames obtained after every 2 ns during the analysis of the 10 ns trajectory for MD simulation with MUC4, showing contacts if any for MUC4 with chitosan

PT

oligosaccharide, stearylamine, cetrimonium bromide and glyceryl stearate citrate throughout the course of the simulation. (See Supplementary Video S1 for the composite of all 1000

positively

charged

region,

respectively.

The

RI

frames and interactions) The red and blue areas on MUC4 correspond to the negatively and hydrogen

bonding,

ionic

interactions,

SC

hydrophobic interactions or water bridges as depicted by yellow dashed tubes for hydrogen bonding, purple dashed tubes for ionic interactions and by sky blue dashed tubes for

NU

hydrophobic contact (See Supplementary Figure S2). Water molecules have been deselected for the purpose of visualisation. (For interpretation of the references to colour in this figure

MA

legend, the reader is advised to refer to the web version of this article) Figure 3. Simulation interaction diagram obtained after analysis of the 10 ns MD simulation

ED

trajectories for protein-ligand (PL) contacts at different residues on MUC4 for (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide showing hydrogen bonding,

EP T

ionic interactions, hydrophobic contacts and water bridges Figure 4. Cumulative values represented as percentages for the different interactions between MUC4 and (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide,

AC C

irrespective of the residues at which they occurred and the prevalence of the respective interactions, namely (d) hydrogen bonding (e) ionic interaction (f) hydrophobic interaction and (g) water bridges in the corresponding ligands. Figure 5. Timeline representation of all protein-ligand contacts summarised for (a) Chitosan oligosaccharide (b) Stearylamine and (c) Cetrimonium bromide. The residues which make more than one specific contact with the ligand, are represented by a darker shade of orange according to the scale to the right of the plot. (For interpretation of the references to colour in this figure legend, the reader is advised to refer to the web version of this article) Figure 6. The scattering intensities of experimental and fitted data points obtained after small-angle neutron scattering (SANS) for (a) COS NLC (b) STA NLC and (c) CTAB NLC, Page 36 of 39

ACCEPTED MANUSCRIPT along with the graphical representation of the formed NLC in the inset. (d) An overlay of the SANS data for COS, STA and CTAB NLC, depicting the differences in the scattering pattern. Figure 7. Comparative bar plot representing zeta potential values of the respective NLC formulations in deionised water and 0.5 % w/v type – III mucin solution. (p > 0.05) Figure 8. (a) Ultraviolet spectra of unbound mucin in the supernatant after incubation with

PT

the NLC and the corresponding (b) Mucin binding efficiency calculated in percentage,

RI

indicating the extent of binding of the respective NLC to mucin. (p > 0.05)

Figure 9. Comparative bar plot of diffusion coefficients obtained from nanoparticle tracking

SC

analysis of the NLC in deionised water and mucin. The significance of the differences between the two values for the corresponding NLC indicates the extent of mucoadhesion of

NU

the with the mucin. (p > 0.05)

Figure 10. Confocal laser scanning micrographs of the rat eyes showing mucoadhesion at the

MA

corneal-scleral region, 30 min after topical instillation of coumarin-6 incorporated (a) Blank NLC (b) COS NLC (c) STA NLC and (d) CTAB NLC. The images were captured in two

ED

channels, fluorescence (Left), optical (Centre) and the merged results of the two channels are depicted on the right. The red arrows indicate the upper side or the corneal epithelium and the

AC C

EP T

blue arrows indicate the lower side or the corneal endothelium of the eyes.

Page 37 of 39

ACCEPTED MANUSCRIPT Highlights 

MD simulation as an in-silico screening model to predict mucoadhesion in silico



In silico simulation accurately predicts the extent of mucoadhesion to ocular mucin



Hydrogen bonds and

ionic interaction play an important role in enhancing

PT

mucoadhesion Cationic charge essential for interaction with mucin than neutral or anionic charge



In silico approach could also apply to other mucins at different mucosal sites

AC C

EP T

ED

MA

NU

SC

RI



Page 38 of 39

ACCEPTED MANUSCRIPT Credit Author Statement Rohan V Pai: Conceptualization, Methodology, Investigation, Writing – Original Draft, Visualization Jasmin D Monpara: Formal Analysis, Resources, Writing – Review & Editing

AC C

EP T

ED

MA

NU

SC

RI

PT

Pradeep R Vavia: Supervision, Project Administration, Validation

Page 39 of 39