A novel liposomal drug delivery system for PMMA bone cements

Abstract The population in developed countries is ageing and the number of people experiencing joint‐related conditions, such as osteoarthritis, is expected to increase. Joint replacements are currently the most effective treatment for severe joint conditions and although many of these procedures are successful, infection developing after the procedure is still an issue, requiring complex and expensive revisions. Whilst incorporating a powdered antibiotic within the bone cement can reduce infection rates, the powder frequently agglomerates, resulting in poor antibiotic release characteristics and compromised mechanical performance of the cement. To overcome these issues, a novel delivery system consisting of antibiotic‐loaded nano‐sized liposomes was developed for inclusion into polymethyl methacrylate (PMMA) bone cement. This system was tested in a commercial cement (Palacos R) and consistently delivered a higher percentage (22%) of the incorporated antibiotic when compared to the powdered antibiotic cement (9%), meaning less antibiotic needs to be incorporated than with conventional cement. The novel system resulted in a controlled and gradual release of antibiotic over a longer, 30‐day period and enhanced the toughness, bending strength and Vickers hardness of the cement, without altering its polymerization or molecular structure. This new material has the potential to significantly reduce infections in cemented joint replacements leading to enhanced patient quality of life and reduced healthcare costs. © 2015 Wiley Periodicals, Inc. J Biomed Mater Res Part B: Appl Biomater, 104B: 1510–1524, 2016.


INTRODUCTION
Polymethyl methacrylate (PMMA) bone cements have been widely employed in hip and knee replacements since their introduction by Sir John Charnley in 1953. 1,2 In England and Wales 178,000 hip and knee replacements, of which 102,700 were cemented, were performed in 2012, a 300% increase since 2004. 3 Bone cement has several roles: to secure the implant in place; to distribute stresses between the stiff metallic implant and the bone; and, frequently, to prevent infection by releasing an antibiotic. The latter role was first developed in 1969 when Buchholz et al. 4 observed that the monomer of bone cement was eluted over long periods of time after setting. This led to the concept of local delivery of powdered antimicrobial agents from bone cement. The most successful of the antimicrobial agents tested is gentamicin sulfate, a broad spectrum antibiotic, which is currently employed in several commercial bone cements. Gentamicin is an aminoglycoside antibiotic that inhibits protein synthesis by binding to the 30S ribosome subunit of bacteria. 5 Gentamicin is available as a dry powder and is heat stable, making it ideal for the high setting temperatures of bone cement. 5,6 Localised release of antibiotic is imperative for joint replacements as systemic treatment is ineffective due to the limited circulation in the area adjacent to the implant. Incorporating a powdered antibiotic in bone cement has proved to be successful, resulting in a reduction in primary implant infection rates from 7% to lower than 1%. 7 In a recent survey conducted by the National Joint Registry for England and Wales it was shown that between 2003 and 2010, there has been an increase from 85% to 93% and 92 to 97%, respectively in the percentage of antibiotic-loaded bone cement used for primary hip and knee replacements. 8 This trend not only demonstrates the perceived success of this particular technique but also reflects a growing desire to prevent post-surgical infections in joint replacements. Infected joint replacements are difficult to treat and require the complete removal of the cement and implant. This procedure is costly for both patients and health services due to its complexity, high risk and poor surgical outcome. 9 Correspondence to: W. N. Ayre; e-mail: ayrewn@cardiff.ac.uk Contract grant sponsor: Arthritis Research UK and Cardiff University; contract grant number: 18461 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
Although the use of gentamicin sulfate powder in PMMA bone cement has been successful in reducing the rate of infections in hip and knee replacements, 2,500 revision procedures were still required in the UK in 2012 due to infection 3 and 22,000 revisions of infected hip and knee replacements were performed in the US in 2009. 10 These figures are projected to increase, as are the costs associated with this complex procedure. 3,[10][11][12] Furthermore, there are concerns that the antibiotic powder may have a detrimental effect on the mechanical properties of the cement. Commercial cements employ 0.5-1 g of gentamicin sulfate per 40 g of cement and this can cause significant reductions in mechanical and fatigue properties of the cement. [13][14][15][16] The powdered antibiotic is poorly dispersed and agglomerated in the cement and, as a result, only a small percentage of the incorporated gentamicin sulfate is released, the majority in an uncontrolled burst from surface agglomerations within the first few hours. 14 Subinhibitory concentrations are thought to be released thereafter, which may encourage the establishment of an antibioticresistant bacterial population. 17 To overcome these problems we propose the use of liposomes to disperse therapeutic agents within bone cement and to control their release characteristics. Liposomes are spherical vesicles composed of lipid molecules that are capable of encapsulating a variety and combination of both hydrophilic and hydrophobic therapeutic agents. The structure, size and composition of the liposome can be tailored for specific applications. Liposomes have been extensively used in medicine for the treatments of infections, cancers, pain management, vaccinations and in ophthalmology. [18][19][20] Various therapeutic agents have been incorporated within liposomes, including antimicrobial and antineoplastic therapeutic agents, chelating agents, steroids, vaccines and genetic materials. 21 The majority of these liposomal systems are administered as an aqueous suspension and there is little evidence of their use in polymers or non-aqueous environments; the hydrophilic head groups on the outermost layer of the liposome are generally not compatible with hydrophobic materials, such as PMMA bone cements. Due to the hydrophobicity of methyl methacrylate, phase separation of the aqueous lipid suspension is likely to occur. For this reason, the use of gentamicin solution in the liquid component of bone cements has also proven unsuccessful as significant reductions in mechanical properties were observed. 22,23 This was found to occur as a result of the water content and phase separation increasing the size and number of pores. This problem severely limits the range of antibiotics and other drugs that can be incorporated in the cement. To overcome the issue of phase separation, surfactants may be used. Miller et al. attempted to use surfactants with gentamicin solution but found reductions in mechanical properties nevertheless due to the presence of water. 24 Pelleted liposomes, obtained through ultra-centrifugation in conjunction with amphiphilic polymers (containing a hydrophobic and a hydrophilic portion) may overcome this issue. Non-toxic, neutral surfactants, such as Pluronic block copolymers, which improve the miscibility of solutions, may be employed to overcome phase separation of the liposomes in methyl methacrylate. Pluronics consist of two hydrophilic polyethylene oxide (PEO) chains connected by a hydrophobic polypropylene oxide (PPO) chain (PEO-PPO-PEO).
The aim of this study was to incorporate gentamicin sulfate into 100 nm extruded liposomes, coated with Pluronics, within the liquid component of a commercial bone cement, Palacos R. The dispersion of the liposomes was studied in methyl methacrylate and within the polymerised cement. Release characteristics, antibacterial activity and mechanical and fatigue properties were evaluated.

Liposome preparation
Phosphatidylcholine (PC) from egg yolk (99%), cholesterol (C, 99), gentamicin sulfate (590 mg gentamicin base per mg) and Pluronics L31 and L61 were purchased from Sigma Aldrich (Sigma-Aldrich Company Ltd., Gillingham, UK). Pluronic L43 was provided by BASF (BASF Corporation, Connecticut, USA); chloroform (HPLC grade) was purchased from Fisher Scientific (Fisher Scientific UK, Loughborough, UK). 175 mg of PC and 25 mg of C (weight ratio of 7:1) were weighed and dissolved with 5 mL of chloroform in a 50 mL round bottom flask. The chloroform was evaporated using a rotary evaporator at 608C (above the phase transition temperature of PC) under vacuum. The lipid bilayer was resuspended in 40 mL of 5 mg mL 21 gentamicin sulfate solution by vortex mixing. The suspension was then held at 608C for 30 min prior to 10 extrusions under nitrogen pressure (8 bars maximum) using a Lipex extruder (Northern Lipids, British Columbia, Canada) vertically through a 400 nm polycarbonate membrane (Whatman, UK). The suspension was subsequently extruded a further 10 times through a 100 nm polycarbonate membrane. Average liposome diameter was determined by photon correlation spectroscopy (Beckman Coulter N4 PLUS). L31 Pluronic (2%w/ w) was added and the suspension centrifuged at 100,000 g (25,000 RPM) for 1 hr at 48C (Beckman Optima LE-80K) to create a pellet. The 200-mg pellet was resuspended in 20 mL of the liquid MMA component of Palacos R) by the process of titruration. The method was repeated for L43 and L61 Pluronics. These Pluronics were employed as they created the most stable liposome suspensions in terms of sedimentation rate and dispersion. 25 Transmission electron microscopy Uranyl acetate (98.0%) and methyl methacrylate (98.5%) with 30 ppm hydroquinone were purchased from Sigma-Aldrich (Sigma-Aldrich Company., Gillingham, UK). Transmission electron microscopy (TEM) was used to image the dispersion of liposomes in MMA and water. 100 mg of PC:C liposomes and 100 mg of PC:C liposomes with L61 Pluronic were prepared as described in the Liposome Preparation section at 5 mg mL 21 concentration, mixed with an equal volume of 4%w/v aqueous uranyl acetate solution and incubated for 60 min. The liposomes were centrifuged into 20 mg pellets and resuspended into 20 mL of water or MMA. Ten mL of the diluted suspension was placed on a Formvar carbon film on a 400 mesh Nickel grid (EM Systems Support, Macclesfield, UK) and allowed to dry in air. MMA alone was also dried on a grid as a control. Grids were visualised using a Philips CM12 TEM (Philips Research, Eindhoven, Netherlands) operating at 80 kV. Images were recorded using an SIS MegaView III digital camera (Olympus Soft Imaging Solutions GmbH, M€ unster, Germany).

Fluorescent microscopy
To assess the dispersion of the liposomes in a commercial cement, 16 mL of a 5 mg mL 21 liposome suspension in water was prepared as described in the Liposome Preparation section, however fluorescent TopFluor cholesterol (FC; 23-(dipyrrometheneboron difluoride)-24-norcholesterol; INstruchemie BV, Delfzijl, Netherlands) was used to substitute for a portion of the cholesterol component to give a weight ratio of 7:0.9:0.1 of PC:C:FC. The suspension was divided into four aliquots of 4 mL. A 2%w/w of Pluronic L31, L43, or L61 was added to three of the aliquots and the suspension was centrifuged as previously described. Similarly, the remaining 4-mL aliquot of the liposomal suspension alone was also centrifuged. The four 20 mg pellets were individually resuspended in 2 mL of the liquid MMA component of Palacos R (Heraeus Medical, Newbury, UK). This was mixed with 4 g of the Palacos R powder as outlined by the ISO5833 guidelines and the cement manufacturer's instructions. 26 After mixing the cement was compressed between two glass slides to create a thin sample capable of transmitting light. All cement samples were stored in the dark until observed using an Olympus IX50 fluorescent microscope.
Detection of structural changes using FTIR Fourier Transform Infrared Spectroscopy (FTIR) was employed to detect changes in the polymer structure due to the presence of the liposomes. Palacos R, Palacos R1G and the liposomal Palacos R cement with L61 Pluronic were scanned using attenuated total reflectance (ATR) with a Perkin Elmer Spectrum One FT-IR Spectrometer with FT-IR Spectrum software (Perkin Elmer, Massachusetts, USA) between 4000 and 450 cm 21 with a resolution of 4 cm 21 . The surface of the cement samples were scanned at six dif-ferent sections on the sample, taking 30 scans at each section to obtain an average spectrum.
Antibiotic release O-phthaldialdehyde (97%, HPLC), ethanol (99.8%, HPLC), 2-mercaptoethanol (99.0%) and sodium borate (99.0%) were purchased from Sigma Aldrich (Sigma-Aldrich Company, Gillingham, UK). Sodium chloride (99.9%), potassium chloride (99%), calcium chloride (99%), and sodium bicarbonate (99.7%) were purchased from Fisher Scientific (Fisher Scientific UK, Loughborough, UK). Palacos R and Palacos R1G cements (Heraeus Medical, Newbury, UK) with or without liposomes were mixed as outlined in the manufacturer's instructions and the ISO5833 guidelines. 26 Table I shows the compositions of the cements tested. An ultra-high molecular weight PTFE mould was manufactured to produce 10 mm diameter by 2-mm thick cylindrical samples. All samples were abraded with a 250 grit silicon carbide paper to the stated dimensions with a tolerance of 60.2 mm. Each sample weighed 0.40 6 0.01 g and five samples for each test group were examined. Each sample was stored in 5 mL of Ringer's solution (8.6 mg mL 21 NaCl, 0.3 mg mL 21 KCl and 0.33 mg mL 21 CaCl 2 , buffered to a pH of 7.4 with NaHCO 3 27 ) at 378C. After 6 hr, 1, 2, 3, 7, 15, 30, and 60 days the Ringer's solution was removed and stored in the dark at 2208C before assaying; 5 mL of fresh Ringer's was added as replacement. Ringer's solution was replaced at the selected time points in order to approach sink conditions, a condition where the drug concentration in the eluent can be assumed negligible in comparison to the maximum solubility of the drug (50 mg mL 21 for gentamicin sulfate). The solutions were thawed overnight at room temperature in the dark and the concentration of gentamicin released was determined using an o-phthaldialdehyde (PHT) method developed by Sampath et al. 28 and Zhang et al., 29 whereby a PHT reagent reacts with the amino groups of gentamicin sulfate to yield measurable fluorogenic products. The reagent was prepared by adding 2.5 g of ophthaldialdehyde, 62.5 mL of ethanol and 3 mL of 2mercaptoethanol to 560 mL of 0.04M sodium borate solution in distilled water. The PHT reagent was stored in an amber glass bottle in the dark for 24 hr prior to use. 1 mL of the sample eluate was mixed with 1 mL PHT reagent and 1 mL isopropanol and left for 40 min to react and its absorbance compared against a gentamicin calibration curve (ranging from 0 to 100 lg mL 21 ), in order to determine the concentration of gentamicin released by the samples at each time point. Average gentamicin concentrations for each time point were calculated from the five samples and the gentamicin release was determined as a percentage of the theoretical maximum amount of gentamicin sulfate in each sample.

Mathematical modelling of release
Mathematical models were applied to the antibiotic release data for the commercial bone cements in order to better understand the release kinetics. In the equations described, R t is the total percentage gentamicin sulfate released, t is the time period, n is an exponential function and k is a diffusion related constant. Equation (1) was developed by Korsmeyer et al. 30 who stated drug release was exponentially related to time.
Higuchi et al. 31 proposed Eq. (2), whereby the release of drugs is dependent upon Fickian diffusion of water into the material.
Equation (3) is associated with a Noyes-Whitney dissolution process. 32 The Noyes-Whitney dissolution process states that the dissolution rate of a solid is dependent on its diffusion coefficient, the surface area of the solid, the concentration at the boundary layer and the length of the boundary layer.
Equation (4) is an expanded version of Korsmeyer's work developed by Lindner and Lippold 33 which takes into account an additional coefficient (b) to represent the initial burst of antibiotics from the surface of the cement. The additional coefficient was added to the Higuchi and Noyes-Whitney equations [Eqs. (1) and (3)] and applied to the release data.
All models were applied to the data using the curve fitting toolbox in Matlab (MathWorks, Cambridge, MA), which calculates the coefficients by minimizing the sum of squared residuals.
Agar diffusion assay Three 10 mm diameter by 2 mm thick cylindrical samples of Palacos R, Palacos R1G, and the liposomal Palacos R cements (L31, L43, and L61) were prepared as previously described in the Liposome Preparation section. Staphylococcus aureus (S. aureus, NCIMB 9518) was cultured in tryptone soya broth (TSB) for 18-24 hr at 378C. A sterile cotton swab was used to spread the inoculum across the TSA Petri dish. The Petri dish was turned through 608 and the process was repeated to ensure complete surface coverage. A 10 mg gentamicin disc (Oxoid, Hampshire, UK) was placed on the Petri dish as a control and pressure was applied to the top of the disc to ensure full surface contact. The dish was divided into segments and the bone cement samples were individually placed into each segment well separated on the agar, in the same manner. The Petri dish was then incubated at 378C for 24 hr, after which, the zones of inhibition around the samples and gentamicin disc were measured. Images of the zones of inhibition were taken and analyzed using ImageJ software (National Institutes of Health, MA). The zones of inhibition were measured as the radius of the zone minus the radius of the sample. Two measurements for each zone of inhibition were taken, perpendicular to one another and the experiment was repeated in triplicate (n 5 3).

Mechanical properties
The compression and bending properties of Palacos R, Palacos R1G, and the liposomal Palacos R cement samples were determined using a Zwick Roell ProLine table-top Z050/ Z100 materials testing machine with TestXpert II software (Zwick Testing Machines, Herefordshire, UK) following the ISO5833 standard. 26 Cylindrical compression samples (6 mm diameter and 12 mm length) were tested at a crosshead speed of 20 mm min 21 . Rectangular bending samples (75 mm length, 10 mm width, and 3.3 mm thickness) were tested at a cross-head speed of 5 mm min 21 in four-point bending. Five compression and five bending samples were tested to determine average compression and bending properties.

Fracture toughness
Linear elastic fracture mechanics (LEFM) was used to determine the fracture toughness of cement samples. Fractured bending samples were employed for single-edge notched three-point bending (35 mm length, 10 mm width, and 3 mm thickness). A sharp chevron notch (4.5-5.5 mm in length) was created through the center of the sample using a surgical scalpel blade mounted onto a modified microtome, as described by Evans et al. 34 and the initial crack length and specimen dimensions were measured using a travelling microscope (Pye Scientific, Cambridge, UK). The specimens were loaded at a crosshead speed of 5 mm min 21 in three-point bending, with the span between the rollers set to 40 mm, until failure occurred. The load and displacement were recorded and the critical stress intensity factor (K ic ) of the cement was calculated as described in the ISO13586:2000 standard. 35 Five samples were tested for each group to produce mean fracture toughness values.

Glass transition temperature
The glass transition temperature of Palacos R, Palacos R1G, and the liposomal Palacos R cements were determined by dynamic mechanical analysis using a Rheometric Scientific V500 (TA Instruments, Elstree, UK). Small matchstick-like samples (30 mm length, 3 mm width, and 2 mm height) were prepared and loaded sinusoidally at a frequency of 1 Hz in a dual cantilever configuration whilst heated from 358C to 2008C at a rate of 108C min 21 with the applied stress and strain measured. The phase lag (d) between the stress (r) and strain (e) was measured and the storage modulus (E 0 ) and loss modulus (E 00 ) were calculated using Eqs. (5) and (6) respectively. The complex viscosity (tan d) was calculated as shown in Eq. (7) and plotted against temperature. The temperature at which tan d was at its maximum was considered to be the glass transition temperature (T g ). Five samples were tested to obtain a mean value for T g .
Vickers hardness Palacos R, Palacos R1G and the liposomal Palacos R cement disc shaped samples (10 mm diameter and 2 mm height) were prepared and the surfaces polished with a 4000 grit silicon carbide paper. The microhardness of the samples was tested according to the ISO6507-2 standard. 36 Five indentations, 1 mm apart and 1 mm from the edges of the sample, were placed on both sides of each sample using a Zwick 3212 indenter fitted with a 1368 square-based pyramid diamond indenter (Zwick Testing Machines Ltd., Herefordshire, UK) with a load of 200 g for 10 s. Indentations in the vicinity of large pores (>50 mm) were disregarded and repeated. For each indentation, two diagonal lengths were measured using a travelling microscope and averaged. Equation (8) was employed to calculate the average hardness value (in MPa), where F is the load in kg (0.2) and d is the mean diagonal length of the indentation in mm.

Fatigue
Fatigue crack growth tests were performed using disc compact tension (DCT) specimens, with dimensions conforming to the ASTM E399 standard 37 as described by Evans et al. 34 A modified microtome was used to cut a chevron notch, which ensured symmetrical crack growth. The initial crack length and specimen dimensions were measured using a travelling microscope (Pye Scientific, Cambridge, UK). To apply the load symmetrically across the sample, loose plates with oversize holes were employed, which allowed the loading pins to rotate freely without friction. The crack length was monitored using Krak-gauges and a constant current supply and amplifier designed and built by Evans et al. 34 was used to obtain accurate crack growth measurements at low crack growth rates. Precision components were employed and the current regulator and preamplifier were kept at a constant temperature in an Instron environmental chamber (Instron SFL, Bristol, UK) to achieve very low noise and offset drift. Two fatigue samples for Palacos R, Palacos R1G and the liposomal Palacos R cements were tested. All tests were carried out in an Instron environmental chamber at 378C using a 5 kN Dartec servohydraulic testing machine with an MTS FlexTest GT controller and MTS Multipurpose software (MTS, Eden Prairie, MN).
Samples were cyclically loaded in load control with a sine wave at 5 Hz between 100N and 10N (R-ratio of 0.1), allowing for measurements of crack growth rates through stress intensities of 0.3 to 0.9 MPam 1/2 . Prior to loading, samples were subject to a precracking procedure to ensure an initial steady crack growth rate of 10 29 m/cycle. For precracking, a high cyclic load of 200 N at 5 Hz was applied and the load reduced by 10% after every 0.2 mm of crack growth. The crack length and number of cycles were measured during loading and the crack growth rate (da/dN in m/cycle) was calculated for every 0.2 mm of crack growth. Similarly, the corresponding stress intensity factor range (DK) for each sample and crack length (ranging from 6 to 12 mm, in steps of 0.2 mm) was calculated using Eq. (9), where P max is the maximum load, P min is the minimum load, b is the sample thickness, a is the crack length from the center point of the holes to the tip of the crack, W is the width of the specimen from the center point of the holes to the bottom of the specimen and a is a/W and is >0.2.
Scanning electron microscopy The fracture surfaces of the fatigue testing samples were gold coated using an E65x sputter coater (Emitech, Kent, UK) and imaged using an EBT1 scanning electron microscope (SEM Tech, Southampton, UK) at 15 KeV. Similarly, the surfaces of the antibiotic release samples were imaged to study surface porosity.

Statistical analysis
An analysis of variance (ANOVA) was carried out to establish significant differences between groups of samples using the data analysis package in Excel (Microsoft, Reading, UK). Significance between groups was defined as those with a calculated p values of <0.05.

Liposome diameter (with and without Pluronics in water)
Prior to extrusion the average diameter of the liposomes was 588 6 243 nm. After 10 extrusions through a 400 nm polycarbonate membrane and 10 extrusions through a 100 nm polycarbonate membrane, the average diameter of the liposomes was 100 6 19 nm. After ultra-centrifugation and resuspension, the diameter was 111 6 16 nm. The addition of Pluronics L31, L43, and L61 did not alter the average diameter of the liposomes (108 6 28 nm, 106 6 21 nm, and 101 6 30 nm, respectively).
Transmission electron microscopy (TEM) Figure 1 shows TEM images of liposomes with and without the addition of Pluronic L61 resuspended in water or MMA. After centrifugation and resuspension the individual liposomes were intact and approximately 100 nm in diameter, demonstrating that centrifugation at 100,000g for 1 hr did not damage the lipid bilayer and confirming the results obtained by photon correlation spectroscopy. Small agglomerations were observed when liposomes were resuspended in water after centrifugation [ Figure 1(a)], whilst large agglomerations were observed when the liposomes with Pluronic L61 were resuspended in water [ Figure 1(b)]. When resuspending the liposomes in MMA without Pluronic L61, the liposomes were found to coalesce potentially due to the hydrophilic interactions between the polar heads of the phospholipids in the hydrophobic MMA [ Figure 1(c)]. Improved dispersion was achieved however when liposomes with Pluronic L61 were resuspended in MMA [ Figure 1(d)]. Individual liposome diameters were~100 to 200 nm, confirming the results observed by photon correlation spectroscopy.

Fluorescence microscopy
The fluorescent liposome pellets were resuspended in the liquid component of Palacos R bone cement prior to Detection of structural changes using FTIR FTIR was employed to detect potential changes in polymerization or cement structure arising from incorporating liposomes. Figure 3 shows the FTIR spectra for Palacos R, Palacos R1G and liposomal Palacos R with L61 Pluronic using attenuated total reflectance. FTIR peaks representative of PMMA were observed in all samples. These peaks were: between 2850 and 2950 cm 21 for methylene (CH 2 ) and methyl (CH 3 ) stretches; between 1300 and 1500 cm 21 for the CH bend; at 1400 and 1450 cm 21 for methyl groups (CH 3 ); at 1240 cm 21 for the CACAO stretch; at 1140 cm 21 for the OACAC stretch; at 1730 cm 21 for the C@O stretch; and between 1300 and 900 cm 21 for the CAO stretch. Broad bands were observed at~3448 cm 21 and sharp bands at 1636 and 1051 cm 21 for Palacos R1G and Palacos R with liposomes due to the presence of gentamicin sulfate. No other differences were observed between all three cements, demonstrating the liposomes did not interact or alter the polymerization or structure of Palacos R.

Antibiotic release
A strong linear correlation was obtained between gentamicin sulfate concentration and absorbance at 340 nm using the o-phthaldialdehyde method (R value of 0.9987), with acceptable standard deviations up to gentamicin sulfate concentrations of 90 lg mL 21 (absorbance of 1.6). The percentage antibiotic release from Palacos R, Palacos R1G and the liposomal Palacos R cements with L31, L43, and L61 Pluronics after 72 hr and 1440 hr are shown in Figure 4(a,b), respectively. The liposomal Palacos R cements performed more efficiently than the Palacos R1G. At 6 hr, a similar percentage release was experienced by Palacos R1G and the liposomal Palacos R cements ('2.7%), however after this time the liposomal Palacos R cements achieved much higher percentages. At 72 hr, the liposomal Palacos R released 11-12% of the total gentamicin sulfate used, whilst Palacos R1G released~6%. Over the first 72 hr gentamicin release from the liposomal Palacos R cements was more linear than the Palacos R1G, which experienced an initial burst of release followed by lower levels of release thereafter [ Figure 4(a)]. Furthermore, the standard deviations experienced for all liposomal Palacos R cements within the first 72 hr were less than those of the commercial Palacos R1G cement containing powdered gentamicin. After 1440  hr, the liposomal Palacos R cements had released 21-22% of the incorporated gentamicin, more than Palacos R1G [8.9%, Figure 4(b)].

Mathematical modelling of release
The results from the mathematical modelling of the gentamicin sulfate release from Palacos R1G and the liposomal Palacos R cement are shown Table II. Coefficient values (k, n, and b) that most accurately model the experimental antibiotic release data are shown. The table also shows the goodness of fit (R value) between the experimental data and the models. For all cements, the n-coefficient for the Korsmeyer-Peppas model was found to be <0.5, demonstrating that the release of gentamicin sulfate from the cements was not governed by Fickian diffusion alone. The Higuchi model for the  liposomal Palacos R cements however had higher R-values than Palacos R1G demonstrating release from the liposomal cements is more related to diffusion-based mechanisms. The Korsmeyer-Peppas model was found to best model the Palacos R1G release profile, whilst the Noyes-Whitney dissolution model fitted the experimental release data for the liposomal Palacos R cements the best. The R values obtained from the Noyes-Whitney and the expanded Noyes-Whitney models were higher for the liposomal Palacos R cements than the Palacos R1G cements. This demonstrates a more dissolution-based process taking place in the liposomal Palacos R cements. The additional coefficient (b) was found to improve the accuracy of the mathematical models for all the cements tested confirming there was a burst of antibiotic from the surface of the cement. Interestingly, the burst coefficient improved the models for Palacos R1G to a greater extent than the liposomal Palacos R cements.
Agar diffusion assay Figure 5 shows the zones of inhibition obtained from the agar diffusion assay and the average radii of the zones. No zones of inhibition were obtained for Palacos R cement, demonstrating no antimicrobial activity against S. aureus [ Figure 5(b)]. Palacos R1G was found to produce the largest zone of inhibition in one sample, however the variation between samples was large, potentially due to variable dispersion of the gentamicin sulfate in the cement and/or poor release properties [ Figure 5(c)]. The liposomal Palacos R cements (L31, L43, and L61) however produced consistent inhibition zones demonstrating reliable antimicrobial activity against S. aureus [ Figure 5(d-f)]. Furthermore, these results highlight the availability of the encapsulated gentamicin sulfate to inhibit S. aureus growth. Fatigue Figure 6 shows the fatigue crack propagation results for Palacos R, Palacos R1G and the liposomal Palacos R cements with L31, L43, and L61 Pluronics and Table IV shows the fatigue test constants obtained when the Paris law was applied to the fatigue data. All three liposomal Palacos R cements performed similarly in terms of crack growth rates. Palacos R1G had a higher rate of crack growth when compared to Palacos R, demonstrating the addition of gentamicin sulfate powder to bone cement reduces the crack resistance of the cement. Interestingly, the addition of welldispersed 100 nm liposomes to the cement increased the fatigue crack growth resistance of the cement. Higher stress intensities were required to propagate crack growth in the liposomal cement samples and lower rates of crack growth (da/dN) were observed. When applying the Paris law to the fatigue data, the variables 'A' and "m" are obtained, where "A" is the crack growth rate at a stress intensity of 1 MPam 1/2 and "m" is the slope of the plot. The rate of crack growth at 1 MPam 1/2 was increased as a result of using gentamicin sulfate powder in the cement, however the use of sub-micron liposomes was found to reduce the crack growth rates of the cement.
Scanning electron microscopy   to air entrapped during mixing and cleaving of PMMA beads during fracture. The liposomal Palacos R cements were found to have smaller sized pores (20-50 lm), whilst Palacos R1G had larger pores (50-100 lm). Closer inspection of the liposomal Palacos R cements found small welldispersed pores (<10 lm) throughout the fracture surface. Rough fracture surfaces were also observed for the liposomal Palacos R cements, indicating non-brittle fracture. SEM images of the surface of the samples used at the end of the antibiotic release study (Figure 8

DISCUSSION
Commercially available antibiotic-loaded bone cements, such as Palacos R1G, employ large amounts of gentamicin sulfate powder to achieve therapeutic levels of release (from 0.5 to  1 g per 40 g of cement). Although these cements have reduced infection rates in cemented joint replacements, there are several limitations associated with this delivery system, as shown in this study. A large dose of the antibiotic is released in an uncontrolled manner from the surface of the cement within the first 6 hr, after which, lower, possibly sub-inhibitory levels, are released over prolonged periods of time. This may promote the evolution of resistant bacteria. 17 Although large amounts of gentamicin sulfate are incorporated in commercial cements, only a small percentage (9% over 60 days) is released, leaving the majority of the antibiotic unused in the bulk of the cement. Large amounts of powdered gentamicin in bone cements have been shown to affect mechanical and fatigue properties in previous studies, [13][14][15][16] with no beneficial influence against the formation of biofilms. 38,39 Furthermore, the poor dispersion of gentamicin sulfate in bone cement can result in poor antimicrobial activity. With these limitations in mind, there is a clear requirement to improve antibiotic-loaded cements.
This study demonstrates that liposomes with an encapsulated antibiotic can be effectively dispersed in PMMA bone cement using specific Pluronics (L31, L43, and L61) as surfactants. It is hypothesized that the hydrophilic PEO chains of the Pluronics attach to the outer hydrophilic polar head groups of the liposome leaving the hydrophobic PPO chain to interact with a hydrophobic environment, as shown in Figure 9. This in turn allows the hydrophilic liposomes to be suspended in a non-aqueous environment such as methyl methacrylate. TEM and fluorescent microscopy showed that in the absence of Pluronics, the liposomes aggregated in the cement, likely due to the hydrophilic outer surfaces of the liposomes attracting one another in a hydrophobic environment. The presence of Pluronics on the surface of the liposomes is thought to render the liposome hydrophobic, enhancing dispersion in non-aqueous environments. When testing various Pluronic-liposome combinations, it was observed that hydrophobic Pluronics (with a long PPO chain) and a low molecular weight (such as Pluronics L31, L43, and L61) produced the most stable suspensions. 25 This resulted in an even dispersion of liposomes throughout the cement. All three Pluronics gave similar results in terms of antibiotic release, antimicrobial efficacy and mechanical, fatigue and material properties as a result of good dispersion properties.
For the liposomal Palacos R cements, there was a small initial burst of gentamicin sulfate released from the cement as surface liposomes came into contact with water. The initial burst of gentamicin sulfate was lower than for the commercial cements as significantly less gentamicin sulfate was employed and fewer agglomerations were present on the surface. This resulted in a more linear, gradual and prolonged release of antibiotic over time from the liposomal systems. The SEM images of the release samples confirm this as the commercial cements had randomly located large pores caused by the release of surface agglomerations of gentamicin sulfate, whilst the liposomal Palacos R cements had much smaller well-dispersed pores. The presence of small well-dispersed pores would promote controlled diffusion of water through the outer layers of the cement. This in turn would allow water to penetrate uniformly to the antibiotic below the surface of the cement, resulting in a more gradual and controlled release; a previous study has demonstrated PMMA bone cement to absorb moisture over 60 days. 40 Large random pores caused by agglomerations of gentamicin sulfate powder would result in less controlled release as water penetration into the cement would be less uniform. The standard deviations for the antibiotic release from liposomal Palacos R cements were also reduced, demonstrating consistency between the samples tested over the early time points. Commercial cements on the other hand demonstrated large standard deviations presumably due to variability in the levels of antibiotic being released from each sample.
After 360 hr, the release from the liposomal Palacos R cements was equal to that of Palacos R1G, despite there being 60% less gentamicin sulfate in the liposomal Palacos R cements. This resulted in a much higher percentage release (11-12% at 72 hr and 21-22% at 1440 hr); more than double that of Palacos R1G. A recent study using liposomal amphotericin B in Simplex P bone cement obtained similar improvements over cements with powdered amphotericin B deoxycholate. 41 Static release tests were carried out to compare the release from liposomal Palacos R cements with the release from Palacos R1G. Although dynamic elution tests would more accurately replicate in vivo conditions, the static release test used provides valuable comparative information on the release of gentamicin sulfate as a result of water penetration. Results obtained using dynamic elution techniques would subject the cement to a continuous circulation of medium, which may encourage degradation of the cement, allowing for increased water penetration and antibiotic release. Although Ringer's solution at 378C replicated similar temperatures, pH and salt concentrations found in vivo, the viscosity of the solution differs from blood, which is increased due to the presence of plasma and particles, such as red blood cells. In vivo therefore, the penetration of blood into the cement may differ from that of Ringer's solution, resulting in different release rates. Centrifugation of liposomes may not harvest all the lipid material for inclusion in the MMA, but calculations on gentamicin content have assumed that all the liposomes were incorporated in cement. Furthermore, a high encapsulation efficiency (nearing 100%) of gentamicin sulfate was also assumed, however studies have shown the encapsulation efficiency of gentamicin sulfate in liposomes to range from 5 to 40%. 42,43 Therefore, the efficiency of gentamicin sulfate release from the liposomal Palacos R cement may be significantly greater than the calculated results.
The high efficiency of delivery from the liposomal system may be attributed to the release mechanisms taking place. Equations for antibiotic release were used in an attempt to establish this mechanism. Although simplified equations were employed, it was possible to establish the dominant release mechanism in the cements tested. It is likely; however, that a combination of mechanisms are occurring during the release of gentamicin sulfate and therefore, combining release equations may provide more accurate models, as demonstrated by Kuhn and Wilson. 44 The results from the mathematical models have highlighted differences between the liposomal Palacos R cements and Palacos R1G. The poor fit for the Higuchi model highlights that antibiotic release from bone cement is governed by more than simple diffusion mechanisms. The release from liposomal cements are thought to be controlled through a more diffusion and dissolution based process when compared to Palacos R1G, most likely because the welldispersed liposomes create a more controlled pathway for water to penetrate into the cement. All cements experienced a burst of release from surface-held antibiotic as demonstrated by the improved fit in the expanded models. This improvement was more dramatic for Palacos R1G, demonstrating Palacos R1G has a greater reliance on surfacedriven release mechanism.
The agar diffusion test demonstrated good bioavailability of the encapsulated gentamicin sulfate in the liposomal Palacos R cements as reproducible zones of inhibition against S. aureus were obtained, consistent in both size and shape between all the liposomal Palacos R samples. Good dispersion of the liposomes on the surface layers of cement would result in an even diffusion of gentamicin sulfate through the agar, creating consistent zones of inhibition. It can be concluded, therefore, that the liposomal system would most likely inhibit bacteria in all regions of the cement, unlike powdered antibiotic based products which were found to have inconsistent zones of inhibition due to an uneven distribution of gentamicin sulfate. The efficacy of the liposomal system was only tested against one strain of bacteria as a bioindicator. A study by Mugabe et al. 45 has shown liposomal delivery of therapeutics to be more effective against bacteria than free drugs. The mechanism of action of the liposomes was shown to be through fusion to the wall of the bacterium. This not only enhanced the antimicrobial activity of the encapsulated therapeutic but also increased drug penetration and delivery, making it an exciting method to combat infections caused by antibiotic-resistant bacteria.
Although the liposomal Palacos R cements were found to have good release and antimicrobial properties, a reduction in compressive strength was experienced although this still remained within the range specified by international standards. It is thought that having a dispersion of submicron liposomes in the cement promotes shear yielding, which occurs near the surface of the material where it is in plane stress. This reduces the compressive strength but increases the toughness of the material as observed. A study using lyophilized liposomes in bone cement also found a similar reduction in compressive strength. 41 The bending modulus of the liposomal Palacos R cement was not significantly different from Palacos R1G. Improvements in bending strength were observed for the liposomal Palacos R cement when compared to Palacos R1G. This is due to the even dispersion of the liposomes which contrasts with the weakening effect of agglomerations of powdered gentamicin sulfate. No improvements in bending strength however were observed when compared to Palacos R cement.
The compression and bending tests were carried out according to the ISO5833 standard, which all cements must undergo prior to approval for medical application. The liposomal Palacos R cements were found to exceed the ISO5833 minimum requirements for compression and bending properties, demonstrating the novel delivery system retains the suitability of Palacos cement as an acrylic cement for orthopaedics.
After the release of gentamicin sulfate from the liposomal Palacos R cements, small well-dispersed pores were observed in the SEM images. It has been hypothesised that small dispersed pores may be beneficial to mechanical and fatigue properties by promoting shear yielding ahead of the crack tip, as well as blunting the crack tip. 46,47 Small pores are also thought to reduce polymerization shrinkage, reducing the likelihood of residual stress-induced cracks. 48 It is postulated that the sub-micron pores and liposomes in the liposomal Palacos R cements behave similar to small rubber particles, inducing toughening mechanisms. Rubber particles have been shown to improve the fracture toughness of bone cement through shear yielding, crack deviation and craze formation, 49-51 but reduce compressive strength as yielding occurs in compression. 52,53 Furthermore, rubber particles have also been shown to reduce the modulus of materials, as experienced with the liposomal Palacos R cements. 54 Toughening through shear yielding mechanisms in the liposomal Palacos R cements was observed due to an increase in fracture toughness and fatigue crack propagation. The fracture surfaces were also found to be rougher in the liposomal Palacos R cements, demonstrating non-brittle fracture.
Tests on the structure of liposomal Palacos R cements demonstrated that liposomes do not affect the molecular structure of the cement; however, the hardness of the cement was significantly increased by the addition of liposomes. As hardness value is linked to the yield stress and toughness of a material, it is expected that the yield stress of the liposomal Palacos R cements would also be significantly higher than Palacos R/R1G. The glass transition temperature was also affected, however at temperatures well below the T g of the cements (for example, in vivo temperatures of 378C), it is expected that the slight change will not be detrimental to the mechanical properties as fatigue testing at 378C was not detrimental to crack growth rates. The ISO5833 standard used for mechanical testing does not specify required testing temperatures. Nevertheless, testing at this temperature would ensure that the changes in T g experienced do not significantly affect the compression and bending properties of the liposomal Palacos R cements. Furthermore, creep testing would provide useful information on the rate of deformation that the novel cement is likely to undergo in vivo.

CONCLUSION
Liposomes have been successfully incorporated into a commercial PMMA bone cement (Palacos R) and, with the exception of compressive strength, the commercial cement properties were not adversely affected. Furthermore, all cement properties were above the ISO5833 minimum requirements for acrylic resin cements and the presence of well-dispersed liposomes throughout the cement improved the bending, fracture toughness, fatigue and hardness properties of the commercial cement. It is speculated that the improvement in fracture toughness and fatigue crack growth resistance is caused by the dispersed liposomes inducing rubber toughening mechanisms, such as shear yielding. The novel drug delivery system demonstrated enhanced dispersion of antibiotics when compared to the current powdered antibiotic system. This resulted in a more controlled, gradual and prolonged release of gentamicin sulfate; more consistent antimicrobial properties and higher efficiency, requiring less gentamicin sulfate to achieve therapeutic levels of release. It is speculated that the liposome dispersion allows water to penetrate and diffuse below the surface of cement in a controlled manner.
Although the initial results are promising, it will now be essential to study the long-term stability of liposomes in MMA and determine how liposome modified bone cements perform in vivo. The system must also be optimised in terms of the antibiotics used; in this regard the ability to employ antibiotics in aqueous solution may broaden the range of agents finding application. Further optimization will include: the size and composition of the liposomes; the encapsulation efficiency and the use of other preparation methods, such as lyophilization, to ensure optimum release, antimicrobial, mechanical and fatigue properties in vivo.
In conclusion, the release of liposomal therapeutic agents from PMMA bone cement may play a vital role in reducing infection rates and potentially improving osseointegration in cemented joint replacements and therefore enhance patient quality of life, whilst reducing costs for healthcare providers.