Computational Models of the Gastrointestinal Environment. 1. The

Jan 18, 2017 - Estelle J. A. Suys , Dallas B. Warren , Christopher J. H. Porter ... Simone Aleandri , Lida Kalantzi , René Holm , Anita Nair , Christ...
0 downloads 0 Views 2MB Size
Subscriber access provided by University of Colorado Boulder

Article

Computational models of the gastrointestinal environment 1. The effect of digestion on the phase behaviour of intestinal fluids Woldeamanuel A. Birru, Dallas B. Warren, Stephen J. Headey, Hassan Benameur, Christopher J.H. Porter, Colin W. Pouton, and David K Chalmers Mol. Pharmaceutics, Just Accepted Manuscript • DOI: 10.1021/acs.molpharmaceut.6b00888 • Publication Date (Web): 18 Jan 2017 Downloaded from http://pubs.acs.org on January 19, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Molecular Pharmaceutics is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Computational models of the gastrointestinal environment 1. The effect of digestion on the phase behavior of intestinal fluids Woldeamanuel A. Birru,2 Dallas B. Warren,2 Stephen J. Headey,1 Hassan Benameur,4 Christopher J. H. Porter,23 Colin W. Pouton2* and David K. Chalmers1* 1

Medicinal Chemistry; 2Drug Delivery, Disposition and Dynamics; and 3ARC Centre of Excellence in Convergent Bio-Nano Science and Technology, Monash Institute of Pharmaceutical Sciences, Monash University, 381 Royal Pde Parkville, Victoria 3052, Australia. 4 Capsugel Research and Development, Strasbourg, France

ABSTRACT: Improved models of the gastrointestinal environment have great potential to assist the complex process of drug formulation. Molecular dynamics (MD) is a powerful method for investigating phase behaviour at a molecular level. In this study we use multiple MD simulations to calculate phase diagrams for bile before and after digestion. In these computational models, undigested bile is represented by mixtures of palmitoyl-oleoylphosphatidylcholine (POPC), sodium glycodeoxycholate (GDX) and water. Digested bile is modelled using a 1:1 mixture of oleic acid and palmitoylphosphatidylcholine (lyso-phosphatidylcholine, LPC), GDX and water. The computational phase diagrams of undigested and digested bile are compared and we describe the typical intermolecular interactions that occur between phospholipids and bile salts. The diffusion coefficients measured from MD simulation are compared to experimental diffusion data measured by DOSY-NMR, where we observe good qualitative agreement. In an additional set of simulations, the effect of different ionization states of oleic acid on micelle formation is investigated.

KEYWORDS: bile, gastrointestinal tract, molecular dynamics, DOSY-NMR, phase behaviour, phase diagram, phospholipids, digested phospholipids, bile salts, oleic acid

Introduction Oral administration is the preferred method for the delivery of many drugs into systemic circulation. On their path to the bloodstream from the gastrointestinal (GI) tract, pharmaceuticals encounter a complex and dynamically changing environment; having a variable pH profile and containing the surfactant and lipid components of bile, digestive enzymes, food and food digestion products. While there are a variety of in vivo and in vitro methods available to model the environments of the GI tract,1, 2 the complex nature these systems makes them difficult to study. One approach to improving our understanding of the processes occurring during oral drug delivery is to develop model systems that reproduce the key features of the intestinal environment using a relatively small number of pure molecular components. The behaviour of drugs within such in vitro systems can be more easily investigated within the laboratory environment and the study of the simplified system can be further closely coupled to computational models, which can provide a molecular level description of digestive processes. One of the key processes within the GI tract is the digestion of phospholipids, and other compounds containing ester groups, by gastric and pancreatic lipases. Phosphatidylcholines are quickly hydrolysed to lysophosphatidylcholine and free fatty acids by phospholipase A2 (Figure 1).3 We have recently shown, using in vitro model systems, that the digestion process significantly alters the phase behaviour of systems of bile salt/phospholipid and bile salt/digested phospholipid.4 Nota-

bly, the formation of mixed micelles rather than lamellar phases occurs much more easily once the phospholipid is digested. The digestion of the endogenous phospholipids that are secreted in bile is therefore expected to affect the phases present within the GI tract. Since most work in the literature investigates the structure and behaviour of undigested bile this work undertakes a comparison between digested and undigested bile. The formation of the mixed micelles of bile salts and phospholipid and their transition to vesicles/lamellae with increasing concentration of lipids was observed about 40 years ago in an experimental study performed by Small et. al.5 Historically, a number of simple models have been proposed for the nature of bile salt/phospholipid mixed micelles (see for example models by Small6, Mazer et al.7, Ulmius et al.8 and Nichols and Ozarowski9). These models generally imply that bile salt/phospholipid micelles are quite ordered in nature with the molecules of bile salt and phospholipid arranged in clearly distinct locations within the micelle and, that a single description is appropriate over a range of bile salt/phospholipid micelles ratios and concentrations. These models are simplifications, in that they do not describe how the micelles change in size with variation in phospholipid and bile salt concentrations and also do not describe the dynamic nature of micelles. A more complex model for phospholipid/bile salt micelles was proposed by Marrink and Mark,10 who used molecular dynamics (MD) simulations to model single micelles of

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

bile/phospholipid mixtures. In this early MD study, the micelles were of fixed size but nevertheless, the simulations reveal that rather than being well-ordered, the micelles are irregular in structure and dynamic and that the phospholipid and bile salts are not strongly localized within the micelle. MD simulations model the dynamic behaviour of molecules by solving Newton’s equations of motion, most commonly using a classical mechanics approximation (the force field) to describe the inter- and intramolecular interactions within the system. MD simulation has wide application in the study of biomolecular systems; for example in modelling ligand binding to G protein-coupled receptors11, 12 or in investigating the structures formed by nucleic acids13. MD simulations are well suited the modelling of bilayers,14, 15 micelles16-20 and or combinations of colloidal phases to the modelling of colloidal systems21 are increasingly being used to study the interactions between drugs, drug formulations and their biological environments.22-24 In this study, we use an extensive series of MD simulations to investigate the ternary systems of water, bile salt and phospholipid in the micelle and vesicle regions of the

Page 2 of 17

phase diagram, and also investigate the changes that result from hydrolysis of the phospholipid into molecules of lysophosphatidylcholine and fatty acid. The methodology used in this work is based on our previous studies, which show that the use of multiple MD simulations can be used effectively to understand the lipid structures that form in aqueous solution.21, 22, 25 The computational models in this work are designed to complement the experimental model systems that we have developed using pure molecular components4 in order to establish a close relationship between our theoretical and experimental studies. The influence of pH on phase behaviour is also investigated. Additionally, we investigate the molecular diffusion coefficients of the associated particles formed by undigested and digested phospholipids by using NMR. In the following article, the current work is extended through a companion study that uses similar computational and experimental methods to investigate drug distribution within systems that model the behaviour of a simple drug formulation within the intestinal environment.26

Figure 1. Structures of 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphocholine (POPC), 1-palmitoyl-2-hydroxy-sn-glycerol-3-phosphocoline (LPC) and oleic acid (OA), which are used as models for phospholipid and hydrolysed phospholipid in this work. Glycodeoxycholic acid, sodium salt (GDX) which is used as a representative bile salt is also shown.

Materials and Methods NMR Methods The phospholipids 1-palmitoyl-2-oleoyl-sn-glycerol-3phosphocholine (POPC) and 1-palmitoyl-2-hydroxy-snglycerol-3-phosphocoline (LPC) were obtained from Avanti Polar Lipids, Inc. in powder form. Glycodeoxycholic acid, sodium salt (GDX) was obtained from Calbiochem, oleic acid (OA) (> 99% pure) was obtained from Sigma-Aldrich. Deuterium oxide (D, 99%) was obtained from Cambridge Isotope

Laboratories Inc. Sodium hydroxide, sodium phosphate monohydrate and sodium chloride were analytical grade. The blank fasted state simulated intestinal fluid27 buffer (FaSSIF) was based on the composition of FaSSIF buffer, minus the phospholipid and bile salt and is composed of: 0.00348 g of NaOH and 0.03954 g of NaH2PO4.H2O and 0.06186 g of NaCl in 10 ml of purified water. The pH was adjusted to 6.5 ± 0.02 using 1 M NaOH and 1 M HCl . Phospholipid solutions were prepared using the evaporated film method; 0.0375 g of lipid (POPC or LPC+OA) was dissolved in 5 ml of methanol in a round bottom flask. The meth-

ACS Paragon Plus Environment

Page 3 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

anol was removed using a rotary evaporator and placed under high vacuum for 1 hour to ensure that all the methanol was evaporated. The resulting lipid film was dispersed in 1.5 g of the deuterium oxide aqueous phase blank FaSSIF buffer, generating a 2.5% w/w solution. The aqueous solutions of 2.5% w/w GDX were prepared by dissolution of 0.0875 g of GDX in 3.5 g of blank FaSSIF buffer made using D2O. Seven different proportions of undigested and digested phospholipids with bile salt were prepared as shown in Table 1. These representative mixtures are taken from the phase transition curve identified by using turbidity and particle size measurement experiments conducted by Birru et al.4 Table 1. Mass fractions (W) of GDX, POPC and LPC+OA (total lipid content 2.5% w/w) used in NMR diffusion experiments Sample

WGDX

1

1

0

2

0.866

0.133

9

0.867

0.133

3

0.799

0.200

10

0.667

0.332

4

0.733

0.266

11

0.601

0.398

5

0.667

0.333

12

0.468

0.531

6

0.599

0.400

13

0.335

0.665

7

0.460

0.533

14

0.167

0.832

WPOPC

Sample

WGDX

WLPC+OA

8

1

0

Diffusion ordered spectroscopy (DOSY) spectra were acquired on a Bruker AVANCE 600 MHz NMR spectrometer using a pseudo 2D version of a stimulated echo sequence with a 100 ms longitudinal echo gradient delay and bipolar gradient pulses of 1.5 ms.28 Each pseudo 2D spectrum consisted of 12 1D spectra where the gradient strength was varied linearly from 5 to 95%. Spectra were processed using Topspin 1.3.

Molecular Dynamics Simulations MD simulations were performed using GROMACS version 4.5.429 Calculations were performed using the Victorian Life Sciences Computation Initiative (VLSCI) Linux cluster comprised of 1088 Intel Nehalem compute cores running at 2.66GHz connected using InfiniBand. The GROMOS 53a6 united atom force field30 was used to represent POPC, GDX, LPC, OA and OLAT. This force field is parameterized to reproduce free energies of solvation in water and cyclohexane and has been used extensively to model proteins, micelles and membranes. The cis double bond in oleic acid was modelled using dihedral parameters developed by Barchar et al.31, 32 Heavy hydrogen atoms (4 amu) were used to enable an increased time step the additional mass of the hydrogen was offset by reducing the mass of the attached heavy atom.33 Water was modelled using rigid SPC water and constrained using SETTLE.34 The remaining solute bonds were constrained by the LINCS algorithm.35 Periodic boundary conditions were employed using a 15 nm cubic periodic cell which, for these systems, provides an appropriate trade-off between periodic cell artefacts and CPU time. A cut-off distance of 0.9 nm was used for short range electrostatic interactions and van der Waals interactions and the particle-mesh Ewald (PME) method36 was used for long range electrostatic interactions. A reference temperature of 310 K was used in all simulations with a coupling time constant of 0.1 ps. Water and non-water atoms were coupled to independent heat baths. All constant-pressure simulations used a reference pressure of 1 bar with a coupling

time constant of 2.0 ps and a compressibility of 4.5 x 10-5 bar1 . Starting model structures were built using the Silico37 script random_box. The required numbers of solute and water molecules were randomly positioned in the simulation cell, producing systems ranging from 150,000 to 300,000 atoms with approximate dimensions 15 × 15 × 15 nm. MD simulations were established using: (1) A steepest descents minimization of 500 steps. (2) A constant volume simulation of 5,000 steps with a time step of 2 fs. (3) A constant pressure simulation of 10,000 steps with a time step of 2 fs using Berendsen isotropic pressure coupling. (4) A simulation of 50,000 steps using a 2 fs time step, Parrinello-Rahman pressure coupling38 and the vrescale thermostat.39 The production simulations were run for 200 ns using a time step of 5 fs, the Parrinello-Rahman barostat and v-rescale temperature coupling. Molecular aggregation was analysed using the Silico script find_aggregate, which combines molecules into aggregates by comparing distances between carbon atoms. Two molecules are considered to be part of the same aggregate if they have carbon atoms separated by a distance of less than 0.4 nm. Visualization of the simulation trajectories was performed using VMD40 and images for publication were produced using PyMol.41 Radial distribution functions (RDFs) were calculated with the GROMACS program g_rdf with a bin width of 0.004 nm run using the last 10 ns of the production run. Spatial distributions functions (SDF), which describe the three dimensional probability of finding a particle around a given reference, were calculated with g_sdf distributed with GROMACS 4.0.7 using a bin width of 0.09 nm and a cell size 8×8×8 nm.

Results MD Simulations of Model Undigested and Digested Bile To investigate the molecular factors controlling the formation of micelles or lamellar phases in intact and digested bile, we modelled two systems using MD; undigested bile was modelled using a ternary system of phospholipid (POPC), bile salt (GDX) and water and digested bile was modelled using 1:1 molar ratio of palmitoylphosphatidylcholine and oleic acid (as expected from digestion), bile salt (GDX) and water. GDX was chosen as a representative bile acid based on our previous experimental studies which show that pure GDX behaves in a similar manner to a physiological mixture of bile salts in the solubilisation of phospholipids.4 Oleic acid was modelled as being fully protonated (denoted OA). The effect of changing the ionization state of the oleic acid is considered further below. We note that both the digested and undigested models represent the fasted state, and that additional lipids derived from food would change the nature of the colloidal species within the system. All MD simulations were run for a total of 200 ns starting from a random arrangement of lipid and water molecules and, as we have previously observed for other molecular species,21, 22, 25 the lipids rapidly and spontaneously form colloidal aggregates. From the initial random configuration, aggregates start to form within 10 ns. After 50 ns the aggregates become quite stable and are in equilibrium throughout the remaining 150 ns of simulation. Table 2 records the molecular composition of the simulations performed and the nature of the colloidal aggregates

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

present at the end of the simulation. In both the digested and undigested model systems, the total number of aggregates and mean size of the aggregates formed in the simulations was calculated using find_aggregate. This program identifies hydrophobic aggregates of molecules (e.g. micelles or vesicles/lamellae) in a given periodic system. At the completion of 200 ns, the colloidal structures produced in each simulation were categorised as being micellar or phase separated; having separate aqueous and oil phase regions. The endpoints were characterised as being micellar if more than two discrete aggregates were present. The micellar systems are further subdivided into ‘micelles’, small approximately spherical aggregates, ‘prolate micelles’, which are larger and slightly elongated, and ‘wormy micelles’. In many simulations, only a single aggregate (or occasionally two aggregates) formed during the simulation which we have characterised as ‘phase separated’. In many cases, the aggregate clearly formed leaflets with the POPC or LPC+OA molecules tails aligned which we have classed as ‘lamellar’, but in a number of cases, large aggregates form that do not have a clearly defined lamellar structure. At low total lipid content, these structures resemble large micelles and at high lipid content form disordered rods or planes spanning the simulation cell. We have characterised as ‘phase separated’, but we expect that larger (and longer) simulations would produce a more lamellar arrangement. It is important to note that, in many of the simulations, the sizes of the molecular aggregates formed is limited by the number of molecules present in the simulated system and the simulation results should be interpreted with care when only a small number of aggregates are present at the end of the simulation. For example, simulations 6 and 7 both produce a single micelle-like aggregate that, on closer inspection, has a lamellar structure, that is, the ‘micelle’ resembles a small section of bilayer with the hydrocarbon tails of the phospholipids aligned in separate leaflets. In these simulations, globular aggregates have formed because there are insufficient molecules present to form a complete bilayer. This is confirmed by inspection of simulations 16 and 17 that respectively contain the same proportions of bile salt and phospholipid as simulations 7 and 8 but at higher concentrations. Simulation 16 forms a disordered lamellar system and 17 forms a well-ordered bilayer, indicating that these phases would also form in the more dilute systems if a significantly larger system were to be studied. In Table 2 we have annotated the cases where we expect the apparent phase produced by the simulation does not reflect the thermodynamically stable phase. These annotations are made where a micelle-like structure is observed, but only a small number of aggregates are present (often only a single aggregate), or the observed aggregates have a large range of sizes. Figure 2 shows a selection of final structures obtained from the GDX/POPC/GDX/water simulations. The simulations were performed in four series (A-D) which traverse the phase diagram from high GDX mass fraction to high POPC mass fractions and contain successively greater lipid content. The simulations in set C maintain a constant amount total lipid (POPC+GDX) of 15% by mass. Simulations 1-4 are the start-

Page 4 of 17

ing points for each series and consist only of GDX in water. These simulations clearly display the non-classical surfactant behaviour of bile-salt micelles; the micelles are generally irregular in shape and not strongly organized, changing dynamically through the course of the simulation as observed previously.25 As the GDX concentration is increased in simulations 1 to 4, the micelles progress from being roughly spherical with free GDX present (simulation 1) through elongated or prolate micelles (2), to extended wormy micelles (4). Simulations 5, 9, 13 and 19 contain GDX and POPC in the ratio of 2:1 by mass and form mixed micelles structured with the hydrophobic POPC tails in the interior and the GDX located on the surface. Simulations 5 and 9, having lower total lipid content, form discrete, roughly spherical micelles while simulation 13 with more lipid present, forms a mix of wormy micelles and spherical micelles. Simulations 6, 10 and 14 contain equal masses of POPC and GDX. These simulations also form spherical micelles at lower total lipid concentrations and wormy micelles as the lipid content is increased. Simulations 7, 11 and 15 contain more POPC than GDX and, rather than forming discrete micelles, form single aggregates with evidence of lamellar structure. This suggests that mixtures of POPC and GDX in these ratios would be expected to form large lamellar structures such as vesicles or phase separated systems. The final group of simulations, 8, 12 and 17, are POPC/water systems. When the lipid content is low, these simulations produce single apparent micelles with a lamellar structure. When the lipid content is higher, a bilayer forms that spans the simulation cell. Again, these simulations indicate the formation of lamellar phases. Figure 3 shows the structures produced by simulations of bile salt and digested triglyceride composed of equimolar amount of LPC and OA. The equimolar ratio is required as a single POPC molecular upon digestion produces a single LPC and a single OA molecule. Again, the simulations were performed in four series (D-G) with series F maintaining constant total lipid content. The structures produced by simulations with high ratios of GDX to LPC+OA and a total lipid content below about 12.5 % (simulations 20, 21, 24, 25 and 28) produce discrete mixed micelles with the OA in the core surrounded by LPC and with GDX on the exterior. These micelles become progressively more oblate as the proportion of LPC+OA increases. As the total lipid content increases (simulations 31, 32, 36 and 37) the micelles become elongated and form extended, wormy micelles. At higher ratios of LPC+OA to GDX (simulations 22, 26, 29, 33 and 34) only one or two large oblate aggregates form, often with a lamellar structure. This suggests that these ratios would be expected to form lamellar structures or a phase separated system. The simulations containing only LPC+OA and water (simulations 23, 27, 30 and 35) produce clearly lamellar systems with simulation 30 producing a vesicle-like structure which is expected to be an artefact resulting from the limited size of the simulation. Simulation 35, which contains a larger amount of LPC+OA produces a simple bilayer sheet.

ACS Paragon Plus Environment

Page 5 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Table 2. Simulations modelling the ternary POPC/GDX/water (undigested) and LPC+OA/GDX/water (digested) phase systems. No. of Molecules

Composition (% w/w)

POPC LPC+OA GDX Water

POPC LPC+OA GDX

Sim. No

Num. Aggregatesa

Median Aggregate Size

Structure in final state

12

20

Micelles

1

-

-

236 105570

-

-

5.2

2

-

-

466 100238

-

-

9.7

15

29

Prolate micelles

3

-

-

575

88118

-

-

15.0

10

56

Prolate micelles

4

-

-

982

88118

-

-

22.5

3

47

Wormy micelles

5

64

-

118 105806

1.6

-

3.0

8

19

Micelles

6

88

-

81

105806

2.2

-

2.0

4

45

Micelles

7

118

-

24

105806

3.0

-

0.6

1

142

Phase separated lamellar

8

135

-

-

105806

3.4

-

-

1

135

Phase separated lamellar

9

125

-

233 100518

3.0

-

6.0

8

40

Micelles

10

172

-

165 100073

4.0

-

4.0

5

63

Micelles

11

236

-

47

100191

6.0

-

1.0

1

283

Phase separated lamellar

12

263

-

-

100238

7.0

-

-

1

263

Phase separated lamellar

13

175

-

400

89937

5.0

-

10.0

6

95

Wormy micelles

14

285

-

290

89211

7.5

-

7.5

2

286

Phase separated

15

375

-

233

89937

9.0

-

6.0

1

607

Phase separated lamellar

16

476

-

95

89843

12.4

-

2.4

1

571

Phase separated lamellar

17

567

-

-

89100

15.0

-

-

1

567

Phase separated lamellar

18

111

-

699

89937

3.0

-

17.0

5

44

Wormy micelles

19

260

-

466

89211

6.7

-

11.5

3

190

Wormy micelles

20

-

31

118 105806

-

1.0

3.0

8

21

Micelles

21

-

43

81

105806

-

1.7

2.0

5

26

Micelles

22

-

58

24

105806

-

3.0

0.6

2

70

Micelles

23

-

65

-

105806

-

3.0

-

1

130

Phase separated

24

-

61

233 100518

-

2.4

6.0

8

50

Spherical micelles

25

-

84

165 100073

-

3.6

4.0

6

56

Micelles

26

-

115

47

100191

-

5.0

1.0

2

139

Phase separated

27

-

128

-

100238

-

5.0

-

1

256

Phase separated

28

-

183

233

89937

-

7.6

6.0

6

86

Micelles

29

-

232

95

89843

-

9.9

2.5

2

279

Phase separated

30

-

250

-

89100

-

11.0

-

1

500

Phase separated lamellar

31

-

175

400 100500

-

6.0

9.0

6

122

Wormy micelles

32

-

285

290 110900

-

9.0

6.0

2

429

Phase separated

33

-

375

233 130000

-

11.0

4.0

1

982

Phase separated

34

-

476

95

130000

-

13.0

2.0

1

1047

Phase separated lamellar

35

-

567

-

140500

-

15.0

-

1

1134

Phase separated lamellar

36

-

127

699

89937

-

5.0

17.0

4

282

Wormy micelles

37

-

169

466

89211

-

7.0

12.0

7

121

Wormy micelles

a

Aggregate numbers and median values exclude any isolated molecules.

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 17

Figure 2. Selected final structures from simulations of model undigested bile (POPC/GDX/water) showing the progression from dynamic GDX micelles (e.g. simulation 1), through mixed micelles (e.g. 6) to large aggregates (e.g. 15) and finally to distinct lamellar phase (17). The simulation numbering is described in Table 2. Atom colouring; GDX is grey, POPC is orange and oxygen atoms are red. Boxes indicate the periodic cell. Scale bars are 3.0 nm. Water molecules are not shown.

ACS Paragon Plus Environment

Page 7 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Figure 3. Selected final structures of the simulations of the model digested bile (LPC+OA/GDX/water) showing progression from discrete micelles at high ratios of bile salt/lipid ratios to single, large aggregates as the bile salt concentration is reduced. The simulation numbering is described in Table 2. Atom colouring; GDX is grey, LPC is orange, OA is pink and oxygen atoms are red. The box indicates the periodic cell. Scale bars are 3.0 nm. Water molecules are not shown.

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 17

Figure 4. Ternary phase diagrams showing the phases present at the completion of simulations of (a) model undigested bile (POPC/GDX/water) and (b) model digested bile (LPC+OA/GDX/water). Extrapolated experimental phase boundaries between micelle and vesicle phases are shown as dotted lines.4

Figure 4 shows the MD-derived phase diagrams for the model undigested and digested bile systems; POPC/GDX/water and LPC+OA/GDX/water. Both diagrams have the same general features; a micellar region is present at ratios of bile salt to phospholipid above approximately 3:2 w/w, while phase separation occurs with higher phospholipid content. Roughly spherical micelles form in more dilute micellar solutions where the total lipid concentration is below about 12.5 % w/w. Above this concentration, the micelles are more extended, forming elongated and wormy micelles. In the phase separated region, simulations with low bile-salt content generally produce clear lamellar structures, with well-defined leaflets, while those with greater bile salt content result in more disordered lipid regions. Figure 4 also shows (estimated) experimental phase boundaries for the transition between micellar and phase-separated regions. These boundaries were determined by taking the phase boundaries measured by nephelometry and dynamic light scattering4, which were measured at lower total lipid concentration, and extrapolating these to higher concentration. Possible causes of the discrepancy between experimental and theoretical boundaries in the undigested case are discussed further below.

Figure 5. Spatial distribution functions for the interaction of POPC and LPC with GDX. (a) POPC/GDX/water: water-OW (light blue), POPC: P (tan), POPC: N (dark blue) and POPC carbon atoms C24 and C42 (orange) around GDX in the POPC/GDX/water mixture. (b) LPC+OA/GDX/water: water-OW (light blue), LPC: P (tan), LPC: N (dark blue) and LPC carbon atoms C16 and OA C18 (orange) around GDX in the POPC/GDX/water mixture.

ACS Paragon Plus Environment

Page 9 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Interactions of Bile Salts with Digested and Undigested Lipids To investigate the interactions occurring between the bile salt and phospholipids, we calculated spatial distributions functions (SDFs) for the interactions of GDX within the undigested and digested models. SDFs are the three dimensional equivalent of radial distribution functions. Figure 5 shows the distribution of water atoms (OW) and lipid terminal carbon atoms (POPC: C24 and C42; LPC: C16 and OA: C18) around GDX, revealing that the interactions made by GDX are quite similar in both systems. In each case, the isosurfaces for the water and lipid carbon atoms clearly highlight the hydrophobic and hydrophilic faces of the steroid skeleton and a strong water interaction is evident for the GDX carboxylate group. The hydrophilic face of the steroid is also associated with the phospholipid quaternary nitrogen and phosphate groups. Figure 6 further exemplifies the interactions of GDX with the lipid environment; showing four common relative arrangements of GDX and POPC molecules. In each example, hydrophilic (OH or glycine tail) within GDX interact with the polar phospholipid head group while the hydrophobic steroid backbone interacts with the hydrocarbon chain of POPC. Comparison of MD simulations across the phase diagram reveal different types of interactions predominate in micelles and in oily or lamellar phases. In the micelles, where GDX is in molar excess, the GDX molecules tend to coat the surface the surface of the micelle by interacting with the hydrocarbon chain of POPC exposing the hydroxy groups to the surrounding water molecules and allowing the glycine moiety to interact with the hydrophilic head of POPC, as shown in Figure 6a. This is also observed in other similar MD studies.25, 42, 43 In the lamellar region where bilayer like structures are formed, the GDX molecules tend to arrange themselves parallel to lipids as shown in Figures 6b and 6c.

Figure 6. Four typical relative arrangements of POPC and GDX molecules taken from simulations of the POPC/GDX/H2O system. Atom colouring: carbon, cyan; oxygen, red; polar hydrogen, white; nitrogen, blue and phosphorous, tan.

Figure 7. Per-residue solvent accessible surface areas of GDX and the digested and undigested phospholipid plotted as a fraction of total phospholipid in the system. a) Simulations 1 and 5-8 (~10% total lipid content) and b) simulations 1 and 24-27 (total lipid content ranges from ~5 to ~10%).

The per-molecule solvent accessible surface area (SASA) of the system components is plotted in Figure 7 as a function of phospholipid content. In both the digested and undigested systems, the SASA of GDX has minimum value of about 3 nm2 when no phospholipid is present. The GDX molecules become increasingly solvent exposed surface as the fraction of either digested or undigested phospholipid is increased. The greater exposure of the GDX is driven by the stronger hydrophobic association of POPC and of LPC+OA, and the SASA of the POPC and LPC+OA correspondingly decreases as the phospholipid content increases. The POPC SASA drops to about 4 nm2 as the phospholipid becomes lamellar and the LPC and OA molecules reach minimum values of about 5 and 3 nm2 respectively.

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Comparison of Diffusion Measured by MD and NMR The simplified model systems, representing undigested and digested bile, are ideally suited to experimental study using spectroscopic methods. Here, we have used NMR-based diffusion measurements to compare the mobility of bile salts in the experimental and MD models. Using the published 1H NMR chemical shift assignments of GDX44, 45 we selected three protons from different sites of GDX (H5, H15 and H25, refer to Figure 1) and measured their diffusion coefficients by diffusion ordered spectroscopy (DOSY). Plots of the experimental diffusion coefficient of the bile salt (GDX) as a function of the lipid mass fraction in undigested and digested systems are shown in Figures 8a and 8b. The 2D DOSY spectra are provided in the supplementary material (Figures S1 and S2). Diffusion coefficients for POPC/GDX solutions could be measured up to a POPC mass fraction of 0.5. Above this ratio, the sample became appreciably cloudy and the lipid precipitated from solution. Similarly, the LPC+OA sample precipitated above a mass fraction of 0.7. Comparison between the GDX/POPC and GDX/LPC+OA systems shows that the diffusion of the undigested micelles slows more rapidly with increased phospholipid content than the digested case. Figures 8c and 8d show the diffusion coefficients calculated from the MD simulation. While there is not direct quantitative agreement between the MD and experimental values, there is good qualitative agreement between the two modes; the overall MD diffusion rates are slightly higher than measured experimentally but show very similar changes with increasing phospholipid concentration. The qualitative agreement between MD simulation and experiment provides experimental support for the MD models. MD Simulations Investigating the Impact of Fatty Acid Ionisation on Mixed Micelle Formation One of the digestion products of the phospholipid is oleic acid. In the aqueous environment the physical properties of oleic acid, or fatty acids in general, are influenced by the ionization state of the carboxylic group which, in isolation, has a pKa around 4.8.46 Because medium and long-chain fatty acid fatty acids form self-assembled structures in the aqueous environment, their apparent pKa (pKaapp) is raised due to the high negative surface charge of the aggregates which lowers the H+ activity and increases the effective pH of the local environ-

Page 10 of 17

ment surrounding the micelle.47 A large range of pKaapp values for oleic acid are reported in the literature (6-8.5).47-49 It is believed that fatty acids in the fully protonated state form oil droplets, in the partially ionized state form crystalline and liquid-crystalline phases and, when fully ionized, form micelles.46, 50, 51 Standard MD simulations do not model proton transfer of acids and bases, but generally treat titratable functional groups, such as amines or carboxylic acids, as having a single state which is selected at the start of the simulation based on the group pKa value and the pH at which the simulation is being performed. While methods for constant pH MD simulations in explicit solvent are under active development (see for example Morrow et al52), high-performance constant pH MD methods are not yet available. So, to gain an understanding of the structural changes that can be expected from changes in the environmental pH, in addition to the simulations with protonated oleic acid (OA) (modeling low pH, simulations 20-37) we performed additional ‘digested’ simulations with oleic acid present as 50:50 mixture of protonated oleic acid (OA) and deprotonated oleate (OLAT) (approximately neutral, simulations 38-41) and in the completely deprotonated form (modeling high pH, simulations 42-45). The details of these simulations are provided in Table 3 and the colloidal structures present at the end of the simulations are shown in Figure 9. Table 3 shows that increasing ionisation of the oleic acid reduces the size of the micelles formed. The effects of ionisation on the interactions of oleic acid, with itself and with water, are made clear by the radial distribution functions (RDFs) shown in Figure 10. When the oleic acid is protonated (OA), there is a large peak in the O1:O1 RDF at 0.3 nm, indicating a strong interaction between the head groups due to hydrogen bonding (Figure 10a). In contrast, when the oleic acid is deprotonated (OLAT), electrostatic repulsion between the head groups prevents close association of the head groups. Figure 10b shows that the O1:water interaction is much weaker for protonated oleic acid than for oleate. The close overlap of the oleic acid and oleate curves for the simulations containing 100% OA, 50:50 OA and OLAT and 100% OLAT show that the oleic acid/oleate-water interactions dominate and the behaviour of the OA/OLAT is little-affected by the surrounding lipid environment.

ACS Paragon Plus Environment

Page 11 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Figure 8. Diffusion coefficients of (a) GDX/POPC/water and (b) GDX/LPC+OA/water mixtures determined from DOSY NMR experiment and (c, d) by MD simulation. The measurement error in the diffusion coefficients (estimated using the peak width at half height of the DOSY spectrum) is < 0.01 log units.

Table 3. Simulations of bile salt with digested phospholipid where the oleic acid component is partially or completely deprotonated. Sim. No

Composition (% w/w)

No of Molecules LPC OA OLAT GDX Water

Num. Aggregatesa

LPC/ OA+OLAT

GDX

Median Aggregate Size

Structure in Final State

38

62

31

31

233 100518

2.5

5.6

10

30

Micelles

39

84

42

42

165 100073

3.4

4.0

7

51

Micelles

40

116

58

58

47

100191

4.8

1.2

4

63b

Micelles

41

128

64

64

-

100238

5.3

0.0

2

128b

Phase separated

42

61

-

61

233

89937

2.7

6.2

13

22

Micelles

43

84

-

84

165

89211

3.8

4.5

7

36

Micelles

44

115

-

115

47

89937

5.3

1.3

6

19 b

Micelles

5

b

Micelles

45 a

128

-

128

-

89843

5.9

0.0

46

b

Aggregate numbers and median values exclude any isolated molecules. A small number of aggregates are observed. The median value will not be a reliable estimate of true aggregate size. cThe observed structure is expected to be affected by the number of molecules present in the modelled system.

11

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 17

Figure 9. Final frames of simulations of (left) LPC+OA/GDX/water, (centre) LPC+50%OA+50%OLAT/GDX/water and (right) LPC+OLAT/GDX/water. Increasing ionisation of the oleic acid results in formation of smaller aggregates. Simulation numbering is described in Table 2. Atom colouring; GDX is grey, LPC is orange, oleic acid (OA) is pink, oleate (OLAT) is yellow and oxygen atoms are red. The box indicates the periodic boundary. Scale bars are 3.0 nm. Water is not shown.

12

ACS Paragon Plus Environment

Page 13 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

Figure 10. Radial distribution functions, g(r), for mixtures of GDX with oleic acid in the protonated form (OA) and deprotonated form (OLAT). (a) Self interaction of the carboxylic acid oxygen atom O1 in OA and OLAT. (b) Interaction of oxygen O1 with the water oxygen atom (OW) in simulations containing 100% OA molecules, 50% OA and OLAT and 100% OLAT.

Discussion The final states of the 37 simulations of the undigested bile (GDX/POPC/water) and digested bile (GDX/LPC+OA/water) model systems (Table 2) were used to derive computational phase diagrams (Figure 4). Both phase diagrams have two distinct phase regions; 1) a mixed micelle region, which occurs where the proportion of bile salt is sufficiently high to solubilise the phospholipid present and 2) a phase separated region, where the low aqueous solubility of the oily phospholipid dominates and the lipid components separate out to form an oily or lamellar aggregate. Within the mixed micelle region, a transition from small discrete micelles to extended wormy aggregates is observed as the total lipid content of the system is increased. The computationally derived phase diagrams can be compared to our recent experimental study,4 which used the same molecular components as the current computational study. The experimental studies used dynamic light scattering and nephelometry to detect the transition from micelles to larger aggregates (e.g. vesicles) and were performed with a total lipid content between 0.3 and 2.5 % w/w, which is below the minimum concentration of 5 % used in the MD models. To allow comparison between the experimental measurements and theoretical studies, we have assumed that the position of the phase boundary does not change with increased lipid concentration and the boundary can be extrapolated from the measurements at lower concentration. The extrapolated experimental boundaries are shown as dotted lines in Figure 4. For the digested system, the modeled boundary occurs at ~15% w/w LPC+OA content which is in very good agreement with experimental boundary, which occurs at 13% w/w LPC+OA content. In the undigested case, there is a significant difference in the location of the boundary in the undigested system (the experimental and theoretical boundaries differ by >5% w/w POPC content). Several reasons can be postulated for the difference in the experimental and predicted phase boundaries; 1) there may be deficiencies in the computational model for POPC and/or GDX which cause the phase boundary position to be incorrect, although this seems unlikely to cause such a large discrepancy given that POPC have been used widely, 2) the limited size of the simulation will prevent the formation of large colloidal structures such as vesicles and this may strongly influence the

location of the phase boundary, or 3) the limited duration of the MD simulations may not allow sufficient time for phase separation to occur or there and/or the criteria we have used for classification of the phase region (based on the number of aggregates formed) may not correctly discriminate the different phase regions. Of these possibilities, limitations (the last option), seems the most likely, but further investigation of this discrepancy is warranted. Importantly, the MD simulations provide detailed information about the colloidal structures present in the different regions of the phase diagram. In the mixed micelle regions, the micelle structures are dominated by exposed bile salt molecules (GDX) with phospholipid (POPC or LPC+OA) almost completely buried within the micelle interior. The observed structures are consistent with previous studies.10, 25 In the phase separated region there is insufficient bile salt present to solubilise the more hydrophobic lipid molecules resulting in all of the lipid components forming large aggregates (usually only a single aggregate). The aggregates formed in the phase separated region of the POPC/GDX (undigested) system are more bilayer-like than those formed in the LPC+OA/GDX (digested) system. SDFs calculated around GDX (Figure 5) show that the more hydrophobic β-face of the steroid skeleton interacts strongly with the lipid fatty acid tails while the αhydroxy groups and carboxylate group interact strongly with water and phospholipid head groups. The SDFs for the GDX/POPC and GDX/LPC+OA systems are very nearly identical, showing that the interactions are very similar in both cases. Typical relative arrangements of GDX and POPC are shown in Figure 6. The final section of this study investigates the influence of fatty acid ionisation on the phase behaviour of the digested bile using a simple pH model that neglects the equilibration between ionised and unionised states of oleic acid, treating the oleic acid as being; ionised, 50:50 ionised:unionised or completely unionised. GDX is modelled as 100% ionised in all simulations. While these approximations do not allow us to know the precise pH of each system – the models encompass the extremes of pH, from well below the oleic acid pKaapp to well above the pKaapp. The simulations over the pH range show that the ionisation state of the oleic acid makes a significant contribution to the micellisation of the digested bile, par-

13

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ticularly where the concentration of bile salt is lower. Changes in micelle size with ionisation are evident in the simulation series 26, 40, 44 and 27, 41, 45. Improved MD models that provide a more accurate treatment of ionisation will require the development of computationally efficient, large-scale constant pH MD simulations.

Conclusions In this work, we describe the development of atomistic computational models that allow us to investigate the structure of bile colloids and further, investigate the differences in the colloidal structures of undigested and digested bile. Most work to date has focussed on undigested bile. A key feature of our models is that they are composed of pure molecular components that can be reproducibly prepared and investigated in the laboratory4 and are also suitable for MD simulation. Such models are important new tools that can assist in understanding the fundamental physical behaviour of the colloidal environment within the GI tract which strongly influences the absorption of oral drugs. The current study shows that the MD simulations of pure chemical species reproduce the key features of colloidal behaviour that are observed experimentally. Specifically, the formation of bile salt/lipid mixed micelles and the formation of phase separated species, although the limited size of the models means that large structures, such as vesicles, cannot be observed directly in the MD simulations. Furthermore, multiple MD simulations can be used to generate computational phase diagrams that reproduce experimental data from nephelometry and dynamic light scattering.4 The prediction of phase boundaries for complex mixtures is difficult and, in this context, the agreement between the theoretical and experimental phase diagrams is very good for the digested bile model system (LPC+OA/GDX/water); micelle and phase separated regions are correctly predicted and the phase boundary lies close to the experimentally determined position, but in the undigested system (POPC/GDX/water) there is a discrepancy in the predicted and experimental positions of the phase boundary (>5% w/w POPC content). The reason for this is currently unclear and requires further investigation. There is a qualitative agreement between our theoretical models and NMR measurements of diffusion in both digested and undigested systems. This work also provides insight into the ionisation of fatty acids in digested bile, showing that increased ionisation can enhances micelle formation and also that that simple pH models can provide useful information about the influence of pH on GI colloid formation. Improved understanding of the colloidal species within the GI tract is particularly important in understanding the dissolution of pharmaceutical active agents and the development of drug formulations with improved ability to solubilise poorly water soluble compounds in the GI tract and thereby provide better drug bioavailability and consequently more effective and safer medicines.

ASSOCIATED CONTENT Supporting Information 2D DOSY-NMR spectra of GDX in undigested and digested lipid, tables of values for the data presented in Figure 8 and the final structures of all simulations are provided in PDB format. This

Page 14 of 17

material is available free of charge via the Internet at http://pubs.acs.org.

AUTHOR INFORMATION Corresponding Authors * E-mail: [email protected], [email protected]

ACKNOWLEDGMENT WAB acknowledges the support of a PhD scholarship funded by ARC Linkage Grant LP120100600 awarded to Monash University in collaboration with Capsugel. We would also like to acknowledge the CPU time and technical support provided by the Victorian Life Sciences Computation Initiative (VLSCI) through grant VR0004, MASSIVE, and the Australian National Computational Infrastructure (NCI) through grant y96.

Abbreviations DOSY, diffusion ordered spectroscopy; FaSSIF, fasted state simulated intestinal fluid; GDX, glycodeoxycholic acid, sodium salt; GI, gastrointestinal; LPC, 1-palmitoyl-2-hydroxy-sn-glycerol-3phosphocholine; MD, molecular dynamics; OA, oleic acid; POPC, 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphocholine.

REFERENCES 1. Porter, C. J.; Charman, W. N. In vitro assessment of oral lipid based formulations. Adv Drug Deliv Rev 2001, 50 Suppl 1, S127-47. 2. Crum, M. F.; Trevaskis, N. L.; Williams, H. D.; Pouton, C. W.; Porter, C. J. A new in vitro lipid digestion - in vivo absorption model to evaluate the mechanisms of drug absorption from lipidbased formulations. Pharm Res 2016, 33, (4), 970-82. 3. Whitcomb, D. C.; Lowe, M. E. Human Pancreatic Digestive Enzymes. Dig Dis Sci 2007, 52, 1–17. 4. Birru, W. A. W., D. B.; Williams, H. D.; Benameur, H.; Porter, C. J. H.; Chalmers, D. K.; Pouton, C. W. Digestion of phospholipids after secretion of bile into the duodenum changes the phase behaviour of bile components. Molecular Pharmaceutics 2014, 11, (8), 2825-2834. 5. Small, D. M.; Bourges, M. C.; Dervichian, D. G. The biophysics of lipidic associations. I. The ternary systems: lecithin-bile salt-water. Biochim Biophys Acta 1966, 125, (3), 563-80. 6. Small, D. M. Gastroenterology 1967, 52, (3), 607-610. 7. Mazer, N. A.; Benedek, G. B.; Carey, M. C. Quasielastic light-scattering studies of aqueous biliary lipid systems. Mixed micelle formation in bile salt-lecithin solutions. Biochemistry 1980. 8. Ulmius, J.; Lindblom, G.; Wennerstrom, H.; Johansson, L. B.; Fontell, K.; Soderman, O.; Arvidson, G. Molecular organization in the liquid-crystalline phases of lecithin-sodium cholate-water systems studied by nuclear magnetic resonance. Biochemistry 1982, 21, (7), 1553-60. 9. Nichols, J. W.; Ozarowski, J. Sizing of lecithin-bile salt mixed micelles by size-exclusion high-performance liquid chromatography. Biochemistry 1990, 29, (19), 4600-6. 10. Marrink, S. J.; Mark, A. E. Molecular Dynamics Simulations of Mixed Micelles Modeling Human Bile. Biochemistry 2002, 41, 5375–5382. 11. Thomas, T.; Fang, Y.; Yuriev, E.; Chalmers, D. K. Ligand Binding Pathways of Clozapine and Haloperidol in the Dopamine D-2 and D-3 Receptors. J Chem Inf Model 2016, 56, (2), 308-321. 12. Latorraca, N. R.; Venkatakrishnan, A. J.; Dror, R. O. GPCR Dynamics: Structures in Motion. Chem Rev 2016.

14

ACS Paragon Plus Environment

Page 15 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

13. Sponer, J.; Banas, P.; Jurecka, P.; Zgarbova, M.; Kuhrova, P.; Havrila, M.; Krepl, M.; Stadlbauer, P.; Otyepka, M. Molecular Dynamics Simulations of Nucleic Acids. From Tetranucleotides to the Ribosome. J Phys Chem Lett 2014, 5, (10), 1771-1782. 14. Hong, C.; Tieleman, D. P.; Wang, Y. Microsecond Molecular Dynamics Simulations of Lipid Mixing. Langmuir 2014, 30, (40), 11993-12001. 15. Bennett, W. F. D.; Tieleman, D. P. The Importance of Membrane Defects-Lessons from Simulations. Accounts Chem Res 2014, 47, (8), 2244-2251. 16. Bruce, C. D.; Senapati, S.; Berkowitz, M. L.; Perera, L.; Forbes, M. D. E. Molecular dynamics simulations of sodium dodecyl sulfate micelle in water: The behavior of water. J Phys Chem B 2002, 106, (42), 10902-10907. 17. Bruce, C. D.; Berkowitz, M. L.; Perera, L.; Forbes, M. D. E. Molecular dynamics simulation of sodium dodecyl sulfate micelle in water: Micellar structural characteristics and counterion distribution. J Phys Chem B 2002, 106, (15), 3788-3793. 18. Amani, A.; York, P.; de Waard, H.; Anwar, J. Molecular dynamics simulation of a polysorbate 80 micelle in water. Soft Matter 2011, 7, (6), 2900-2908. 19. Santos, D. P.; Longo, R. L. Molecular Dynamics Simulations of Specific Anion Adsorption on Sulfobetaine (SB3-14) Micelles. J Phys Chem B 2016, 120, (10), 2771-2780. 20. Tang, X. M.; Koenig, P. H.; Larson, R. G. Molecular Dynamics Simulations of Sodium Dodecyl Sulfate Micelles in WaterThe Effect of the Force Field. J Phys Chem B 2014, 118, (14), 38643880. 21. King, D. T.; Warren, D. B.; Pouton, C. W.; Chalmers, D. K. Using Molecular Dynamics to Study Liquid Phase Behavior: Simulations of the Ternary Sodium Laurate/Sodium Oleate/Water System. Langmuir 2011, 27, (18), 11381-11393. 22. Warren, D. B.; Chalmers, D. K.; Pouton, C. W. Structure and dynamics of glyceride lipid formulations, with propylene glycol and water. Mol Pharm 2009, 6, (2), 604-14. 23. Benson, S. P.; Pleiss, J. Molecular Dynamics Simulations of Self-Emulsifying Drug-Delivery Systems (SEDDS): Influence of Excipients on Droplet Nanostructure and Drug Localization. Langmuir 2014, 30, (28), 8471-8480. 24. Shityakov, S.; Salmas, R. E.; Durdagi, S.; Salvador, E.; Papai, K.; Yanez-Gascon, M. J.; Perez-Sanchez, H.; Puskas, I.; Roewer, N.; Forster, C.; Broscheit, J. A. Characterization, in Vivo Evaluation, and Molecular Modeling of Different PropofolCyclodextrin Complexes To Assess Their Drug Delivery Potential at the Blood-Brain Barrier Level. J Chem Inf Model 2016, 56, (10), 1914-1922. 25. Warren, D. B.; Chalmers, D. K.; Hutchison, K.; Dang, W. B.; Pouton, C. W. Molecular dynamics simulations of spontaneous bile salt aggregation. Colloid Surface A 2006, 280, (1-3), 182-193. 26. Birru, W. A.; Dallas B. Warren; Han, S.; Benameur, H.; Porter, C. J. H.; Pouton, C. W.; Chalmers, D. K. Computational and experimental models of the gastroin-testinal environment 2. Phase behavior and drug solubilisation capacity of a Type I lipid-based drug formulation after digestion. Submitted. 27. Marques, M. R. C.; Loebenberg, R.; Almukainzi, M. Simulated Biological Fluids with Possible Application in Dissolution Testing. Dissolut Technol 2011, 18, (3), 15-28. 28. Wu, D. H.; Chen, A. D.; Johnson, C. S. Advances in diffusion ordered NMR spectroscopy. Bull Magn Reson 1995, 17, (14), 21-26. 29. Pronk, S.; Pall, S.; Schulz, R.; Larsson, P.; Bjelkmar, P.; Apostolov, R.; Shirts, M. R.; Smith, J. C.; Kasson, P. M.; van der Spoel, D.; Hess, B.; Lindahl, E. GROMACS 4.5: a high-throughput and highly parallel open source molecular simulation toolkit. Bioinformatics 2013, 29, (7), 845-54. 30. Oostenbrink, C.; Villa, A.; Mark, A. E.; Van Gunsteren, W. F. A biomolecular force field based on the free enthalpy of hydration

and solvation: The GROMOS force-field parameter sets 53A5 and 53A6. J Comput Chem 2004, 25, (13), 1656-1676. 31. Bachar, M.; Brunelle, P.; Tieleman, D. P.; Rauk, A. Molecular dynamics simulation of a polyunsaturated lipid bilayer susceptible to lipid peroxidation. J Phys Chem B 2004, 108, (22), 7170-7179. 32. Martinez-Seara, H.; Rog, T.; Karttunen, M.; Reigada, R.; Vattulainen, I. Influence of cis double-bond parametrization on lipid membrane properties: how seemingly insignificant details in forcefield change even qualitative trends. J Chem Phys 2008, 129, (10), 105103. 33. Feenstra, K. A.; Hess, B.; Berendsen, H. J. C. Improving efficiency of large time-scale molecular dynamics simulations of hydrogen-rich systems. J Comput Chem 1999, 20, (8), 786-798. 34. Miyamoto, S.; Kollman, P. A. Settle - an Analytical Version of the Shake and Rattle Algorithm for Rigid Water Models. J Comput Chem 1992, 13, (8), 952-962. 35. Hess, B.; Bekker, H.; Berendsen, H. J. C.; Fraaije, J. G. E. M. LINCS: A linear constraint solver for molecular simulations. J Comput Chem 1997, 18, (12), 1463-1472. 36. Darden, T.; York, D.; Pedersen, L. Particle Mesh Ewald an N.Log(N) Method for Ewald Sums in Large Systems. J Chem Phys 1993, 98, (12), 10089-10092. 37. Chalmers, D. K.; Roberts, B. P. Silico: A Perl Molecular Toolkit. http://silico.sourceforge.net. 38. Berendsen, H. J. C. P., J. P. M.; Vangunsteren, W. F.; Dinola, A.; Haak, J. R. Molecular Dynamics with Coupling to an External Bath. J Chem Phys 1984, 81, (8), 3684-3690. 39. Bussi, G.; Donadio, D.; Parrinello, M. Canonical sampling through velocity rescaling. J Chem Phys 2007, 126, (1). 40. Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual molecular dynamics. J Mol Graphics 1996, 14, (1), 33-38. 41. The PyMOL Molecular Graphics System, Version 1.5.0.4 Schrödinger, LLC. 42. Sayyed-Ahmad, A.; Lichtenberger, L. M.; Gorfe, A. A. Structure and Dynamics of Cholic Acid and DodecylphosphocholineCholic Acid Aggregates. Langmuir 2010, 26, (16), 13407-13414. 43. Turner, D. C.; Yin, F. C.; Kindt, J. T.; Zhang, H. L. Molecular Dynamics Simulations of Glycocholate-Oleic Acid Mixed Micelle Assembly. Langmuir 2010, 26, (7), 4687-4692. 44. Ijare, O. B.; Somashekar, B. S.; Gowda, G. A.; Sharma, A.; Kapoor, V. K.; Khetrapal, C. L. Quantification of glycine and taurine conjugated bile acids in human bile using 1H NMR spectroscopy. Magn Reson Med 2005, 53, (6), 1441-6. 45. Ijare, O. B.; Somashekar, B. S.; Jadegoud, Y.; Nagana Gowda, G. A. 1H and 13C NMR characterization and stereochemical assignments of bile acids in aqueous media. Lipids 2005, 40, (10), 1031-41. 46. Cistola, D. P.; Atkinson, D.; Hamilton, J. A.; Small, D. M. Phase-Behavior and Bilayer Properties of Fatty-Acids - Hydrated 1-1 Acid Soaps. Biochemistry 1986, 25, (10), 2804-2812. 47. Cistola, D. P.; Hamilton, J. A.; Jackson, D.; Small, D. M. Ionization and Phase-Behavior of Fatty-Acids in Water - Application of the Gibbs Phase Rule. Biochemistry 1988, 27, (6), 1881-1888. 48. Salentinig, S.; Sagalowicz, L.; Glatter, O. Self-Assembled Structures and pK(a) Value of Oleic Acid in Systems of Biological Relevance. Langmuir 2010, 26, (14), 11670-11679. 49. Kanicky, J. R.; Shah, D. O. Effect of Premicellar Aggregation on the pKa of Fatty Acid Soap Solutions. Langmuir 2003, 19, (6), 2034-2038. 50. Small, D. M.; Cabral, D. J.; Cistola, D. P.; Parks, J. S.; Hamilton, J. A. The ionization behavior of fatty acids and bile acids in micelles and membranes. Hepatology 1984, 4, (5 Suppl), 77S-79S. 51. Cabral, D. J.; Hamilton, J. A.; Small, D. M. The Ionization Behavior of Bile-Acids in Different Aqueous Environments. J Lipid Res 1986, 27, (3), 334-343. 52. Morrow, B. H.; Koenig, P. H.; Shen, J. K. Self-Assembly and Bilayer-Micelle Transition of Fatty Acids Studied by Replica-

15

ACS Paragon Plus Environment

Molecular Pharmaceutics

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 16 of 17

Exchange Constant pH Molecular Dynamics. . Langmuir 2013, 29, 14823–14830.

16

ACS Paragon Plus Environment

Page 17 of 17

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Molecular Pharmaceutics

TOC Graphic N+

N+

O O P O O

O O P O O OH

O

O

O O

HO O

O

O

enzymatic hydrolysis O OH

N H

OO Na +

HO

POPC

OA LPC

GDX

Digestion of bile phospholipids

Phase behavior models

17

ACS Paragon Plus Environment