Mixed material wear particle isolation from periprosthetic tissue surrounding total joint replacements

Abstract Submicron‐sized wear particles are generally accepted as a potential cause of aseptic loosening when produced in sufficient volumes. With the accelerating use of increasingly wear‐resistant biomaterials, identifying such particles and evaluating their biological response is becoming more challenging. Highly sensitive wear particle isolation methods have been developed but these methods cannot isolate the complete spectrum of particle types present in individual tissue samples. Two established techniques were modified to create one novel method to isolate both high‐ and low‐density materials from periprosthetic tissue samples. Ten total hip replacement and eight total knee replacement tissue samples were processed. All particle types were characterized using high resolution scanning electron microscopy. UHMWPE and a range of high‐density materials were isolated from all tissue samples, including: polymethylmethacrylate, zirconium dioxide, titanium alloy, cobalt chromium alloy and stainless steel. This feasibility study demonstrates the coexistence of mixed particle types in periprosthetic tissues and provides researchers with high‐resolution images of clinically relevant wear particles that could be used as a reference for future in vitro biological response studies.


| INTRODUCTION
Aseptic loosening is the most common cause of long term total joint replacement (TJR) failure in the United Kingdom 1 [4][5][6] Despite aseptic loosening being the leading cause of TJR failure to date, this complication has declined in prevalence since the introduction of highly wear resistant biomaterials (e.g., highly crosslinked and anti-oxidant polymers). [7][8][9] With approximately 58% and 82% of hip and knee replacements enduring 25 years of use, respectively, 10,11 understanding the characteristics and effects of wear debris remains essential for determining the long term performance of devices.
The volume of wear produced at the primary bearing surface has decreased substantially over time and is suggested to be the main reason for reduced revision rates of hip replacements with up to 13 years of clinical follow-up. 8,9 Despite this, factors other than wear volume, such as wear particle size, shape and chemistry, are known to affect the biological response associated with implants. 12 How these other important factors change between devices, or over the lifetime of an individual device, is not yet understood. 13 The progressive reduction of wear at the primary bearing surface has made isolating lower volumes of wear particles more challenging, requiring the development of more sensitive methods. [14][15][16][17] In addition, the focus of research has somewhat changed to surfaces other than the primary bearing, such as the backside of acetabular liners or cups and corrosion at modular junctions. [18][19][20][21] These changing requirements emphasize the importance of isolating and characterizing wear particles produced across all wear modes. 22 The production of mixed wear particle populations was recognized in early generations of joint replacement 23 and wear particle isolation methods were developed to isolate both high and low density materials concurrently. [24][25][26][27][28] Since the development of these methods, several key principles have been established for isolating specific materials, 29 all of which would need to be considered when multiple materials are being studied. For example, it is generally accepted that acid and base digestion results in a more complete protein digestion. 27, 30 Yet, these aggressive digestion techniques dissolve certain wear particle types like polymethylmethacrylate (PMMA), calcium phosphate and nanoscale metal wear particles, rendering them unmeasurable. 25,26,31,32 Also, these methods did not have the sensitivity nor resolution to isolate and characterize nanoscale wear particles. Maloney, et al. 24 were the first to use enzymatic digestion in a dual material isolation method. The average size of the isolated UHMWPE and titanium wear particles was 0.5 ± 0.3 μm and 0.7 ± 0.3 μm, respectively. These results have since been generally representative of the wear particle sizes isolated from tissue samples for both THR and TKR. 25,26,33,34 However, the size distribution was skewed in favor of the larger and visually defined particle types which were often agglomerates of wear debris rather than individual particles. Ultimately, these limitations led researchers to optimize methods for isolating specific materials, such as UHMWPE 33 or cobalt chromium alloy associated with metal-on-metal devices 31,35 as opposed to broader multi-material approaches.
Highly sensitive wear particle isolation methods have since been developed. [14][15][16][17]36 These modern methods are effective at isolating, characterizing, and imaging wear particles of all size ranges from highly wear-resistant devices. Yet, these methods remain focused on the highest wearing materials, often associated with the primary bearing surface (i.e. UHMWPE in a metal on UHMWPE construct) and do not attempt to capture the full range of high-and low-density particle types present within individual human tissue samples.
In this research, a more exploratory approach to isolating wear particles was taken, whereby a broader spectrum of particle types was targeted rather than one specific material type. The utility of such an approach is for where the composition and morphology of wear affecting the surrounding biology is less obvious, unknown or where primary bearing wear is considered to be of less concern, such as in devices used for smaller joints (e.g., shoulder, elbow, ankle, finger, spine). A modified version of two established state of the art wear particle isolation methods was developed. 16,36 The aim of this study was to determine whether the modified method can reliably isolate and characterize UHMWPE (<1.0 g cm À3 ) and high-density wear particles (>2.0 g cm À3 ) from human tissue samples. Periprosthetic tissue from THR and TKR patients at revision surgery containing generally well understood wear particle types were characterized and compared to the literature.

| Wear particle isolation method
A novel method was developed to separate high-and low-density materials by density ( Figure 1). The separation step used ultracentrifugation to float particles with a material density of less thañ 1.0 g cm À3 (e.g., UHMWPE; Density = 0.93 g cm À3 ) and to sediment particles with a material density greater than~2.0 g cm À3 (e.g. calcium phosphate; density =~3.1 g cm À3 ). Low-density particles were treated using an adapted version of the method published by Richards et al. 36 (referred to as Process 1), whereas an adapted version of the method by Lal et al. 16 was used to process high density particles (referred to as Process 2).

| Particle separation step
Ultra-pure water was added to fill each ultracentrifugation tube and the tubes were centrifuged at 125,755 g at 20 C for 3 h (Optima  2.5 | Process 2: Isolation of particles with a density >2.0 g cm À3 A density gradient was compiled in a 12.5 ml ultracentrifuge tube with  were centrifuged for 4 h at 202,048 g at 25 C. The sodium polytungstate was discarded and the pellet was sonicated with 2 ml of ultrapure water. The sample was transferred to a clean 12.5 ml ultracentrifuge tube and washed in ultrapure water three times followed by centrifugation at 154,693g at 20 C for 1 h. In between each centrifugation wash, the ultrapure water was exchanged. Finally, the pellet of wear particles, which was free from contaminants, was filtered and characterized.

| Particle characterization
High-and low-density particles were filtered onto separate 0.015 μm filter membranes (GE Whatman, UK), left to dry overnight at room temperature and mounted directly onto aluminium stubs prior to being coated with carbon to a thickness of 5 nm. The filter membranes were imaged using a Hitachi SU8230 high resolution cold-field emission scanning electron microscope (CFE-SEM) at between 50-and 200,000-times magnification. The backscatter detector on the SEM was used to highlight high-density materials. 37 Particle composition was identified using energy-dispersive X-ray (EDX) spectroscopy. Particle morphology was described qualitatively in accordance with ASTM F1877-16. 38 Quantitative metrics such as major diameter (D max ), aspect ratio and roundness measurements were also captured for each particle manually using ImageJ. 39

| Statistics
Non-parametric descriptive statistics (e.g., median size, interquartile range) were reported for wear particle size and parametric characteristics were reported for morphology (e.g., mean aspect ratio with standard deviation). The proportion of wear particles in the submicron (<1 μm), 1-10 μm and >10 μm size ranges were also reported.
T A B L E 2 Number and type of isolated wear particles with a comparison between the isolated particle materials and the known explant materials

| RESULTS
High-and low-density wear particles were isolated and characterized from all THR and TKR tissue samples. At least three different material types were isolated from 6 of 10 THR samples and 5 of 8 TKR samples (Table 2). UHMWPE was identified by its morphology and was assumed, following the exclusion of contaminants, to be the only low-density material isolated in process 1 (low-density particles). Contaminants such as silica and residual proteins were identified at low frequencies and excluded with the aid of EDX analysis. and granules for both joint replacement types ( Figure 2). For both THR and TKR, 73% of the isolated UHMWPE particles were in the submicron size range, which was the smallest particle population identified overall.
However, the largest individual wear particle was also UHMWPE with a major diameter of 465 μm. This particle was isolated from a THR sample and was fibrillar with straight morphology (Figure 2A).
PMMA particles were observed as relatively large (>10 μm), granular, irregular, angular morphologies but sometimes granular,  Titanium alloy particles were the most frequently identified metallic wear particle type. These were typically flake-like shards, in the micron size range with an elongated morphology ( Figure 5A).
However, titanium alloy also presented as flake-like irregular particles ( Figure 5B,C). Stainless steel wear particles were smooth and flake-like, which were generally similar in size and morphology to titanium alloy particles, except generally larger and less elongated ( Figure 5D-F). Cobalt chromium alloy was the rarest particle type with only 12 particles characterized from three samples ( Figure 5G-I). These particles were the only type to manifest as flake-like stacked-sheets ( Figure 5H).
The proportion of particles observed in the three pre-defined size ranges (<1, 1-10, and >10 μm) were similar between THR and TKR samples when all particle types were considered ( Figure 6). However, there were notable differences between size ranges when comparing all low-density particles combined to all high-density particles F I G U R E 6 Wear particle size characteristics by joint type, material type and size range. Left sided graphs are the cumulative frequency of particles in each size range and the right sided graphs show the proportion of particles in each conventional size range. THR, Total hip replacement (blue); TKR, Total knee replacement (gray); UHMWPE, Ultra-high molecular weight polyethylene; D max , major diameter combined. For example, 73% of all UHMWPE particles were in the submicron size range for THRs, whereas 51% of high-density particles were submicron in size. Of all the wear particle types isolated, UHMWPE and ZrO 2 particles were the only types to be predominantly in the submicron size range. All other material types were larger than 1 μm, on average (Table 3). PMMA was the only particle type to be consistently in the >10 μm size range. Relative particle size between material types were generally consistent and comparable between both joint replacement types.

| DISCUSSION
A modified wear particle isolation method was developed and implemented with the aim of isolating mixed wear particle populations of commonly used materials in TJRs from the same periprosthetic tissue sample. UHMWPE and a range of high-density wear particle types were isolated from all tissue samples in this study, which confirmed the feasibility of the method. The isolation of several particle types from all the tissue samples processed in this research confirmed the presence of mixed particle populations, the majority of which were in the submicron size range. Approximately half of the tissue samples analyzed in this study were from devices revised for aseptic loosening after more than 10 years of implantation. How these particles influenced the biological response cannot be determined from this study, but recent research has linked such particle types to osteocytic osteolysis, 6  commonly identified. 33 The broad spectrum of UHMWPE particle sizes and morphologies reflects the different wear processes, such as adhesion, abrasion or fatigue, occurring at the interfacing surfaces. 44 Topolovec, et al. 37 more recently noted that 51% of their isolated UHMWPE particles were smaller than 0.1 μm and that only 7% of particles were larger than 0.5 μm. The median UHMWPE particle size identified in this current study were slightly smaller than the mean values reported in the literature, which was expected as wear particle distributions are non-parametric. 45 The interquartile ranges for both joint types were consistent with the typical submicron UHMWPE size distribution reported in the published literature.
High density wear particles were isolated from all THR and TKR tissue samples in this study. The most common high-density wear particle type identified was ZrO 2 , which presented as both mulberry-like particles and granules for both joint types. In several samples, vast numbers of these particles were found embedded within larger PMMA particles isolated from the same tissue samples. Zirconium dioxide particles are an additive used in bone cement (PMMA) to act as a radiopacifier in a similar way to how barium sulphate is used.
Manufacturers tend to add one or the other to bone cement to improve its visibility on X-ray imaging. Bone cement particles have been isolated from tissue samples previously. [46][47][48]et al. 49 was the first and only study to isolate zirconia particles from tissue samples and found the average size to be 0.28 ± 0.08 μm. This was the same average ZrO 2 granule size identified in the current study.
Lerouge et al. 49  The majority of previous research into metal wear debris is focused on cobalt chromium alloy metal-on-metal hip replacements, 31,45,50,51 which tend to produce nanometer-sized (~30-100 nm) metallo-organic composite spheroids. 44 However, metal wear particles are also commonly found within periprosthetic tissue surrounding non-metal-onmetal THR and TKR, 19,37,52,53 which is more relevant to the devices in this study. The composition, size and shape of such metal particles depends on the source, location and the mechanism of wear. For example, trunnion wear against the inside of a ceramic THR head has been shown to produce titanium alloy wear particles sized between 0.02 and 0.05 μm, 54 whereas titanium alloy wear debris produced at the stem, acetabular cup or supporting screws of ceramic-on-ceramic THRs were substantially larger at 0.61 ± 0.31 μm. 49 The titanium alloy wear debris isolated in this current study was predominantly micron-sized and elongated. The origins of the titanium alloy wear particles in the THR samples are the cup or stem components. One explanted TKR had a worn-through tibial tray which resulted in a large number of titanium alloy particles being isolated from the associated tissue sample. For the other TKR samples, less obvious wear of the tibial tray was the likely origin of the particles.
Only a very small number of cobalt chromium alloy and stainless steel wear particles were characterized in this current study. The tissue samples were predominantly from surrounding cemented devices and the layer of bone cement separating the tissue from the metallic components could have reduced the probability of exposure to such particles. Cobalt chromium alloy particles were found predominantly in one TKR tissue, which featured cobalt chromium alloy femoral and tibial components. The flake-like stacked sheet morphology of the particles was indicative of abrasive wear, which is expected of TKRs. 19 The origin of the stainless steel particles was unclear for several samples but the possible sources include monoblock hip stem components, which was stainless steel for three samples; accessory components such as a Posterior Lip Augmentation Device (PLAD), 55 which was noted for one of THR systems included in this study; and/or the surgical instrumentation used at implantation/retrieval. Stainless steel scissors and scalpels were also used to dissect the tissue samples into small pieces in preparation for digestion and may have introduced contamination despite the significant care taken not to do so.
The method developed for this study implemented several techniques to capture the full particle size range, including ultra-high resolution imaging, nano-scale filtration pore sizes and manual characterization.
Imaging magnification has been shown to be a critical factor for accurately characterizing particle size distributions. 56 The earliest wear particle isolation studies were unable to confidently analyze nanometer wear particles due to the resolution of their analysis methods. 24,26 For example, the mean size of UHMWPE wear particles surrounding THRs were reported to be 0.53 ± 0.3 μm, 25 and between 0.58 and 0.79 μm, 26 both of which were determined using automated particle sizers. However, the maximum resolution of these automated methods was 0.4 24,57 and 0.58 μm, 26 both of which are too low to precisely characterize the majority of particles produced in the nanoscale and sub-micron size ranges.
The resolution of these automated methods are approximately 20 times larger than the smallest particle identified in the current study. Tipper, et al. 58 used manual characterization to observe nanometer wear particles using magnifications up to 65,000 times. Magnifications up to 200,000 times were used for wear particle characterization in the current study, which in part explains why the average particle size was smaller than the previous literature and is consistent with the prediction that improvements in technology result in smaller wear particle size distributions. 56 The filter pore size used in this study was 0.015 μm which is smaller than other published methods that typically use 0.1 μm or larger. 59 Using pore sizes too large for the expected particle types means nanometer-sized particles can flow freely through the filter, resulting in their omission from the analysis. Nanometer wear particles have previously been isolated but as they commonly present in larger agglomerates, manual characterization may be more effective than automated methods at accurately identifying them. 36 It is known that manual characterization is labour intensive, subject to observer selection bias and that fewer particles can be analyzed as a proportion of the total wear particle population. 36 However, manual characterization was deemed necessary for the current study because the types of particles present in the tissue samples were unknown and required careful assessment to be accurately identified, imaged, and characterized.
This study tested the feasibility of a novel isolation method, which performed acceptably because UHMWPE and many of the high-density wear particles associated with the patient's devices were isolated from all 18 tissue samples with minimal contamination. The processed tissue samples were however, from largely unknown failed THR and TKR devices and the type of UHMWPE used in these devices was also unknown. Differences between the material types (e.g., GUR 1020, GUR 1050, crosslinked etc.) would likely have added to the high variability of particle characteristics found between samples. Also, the anatomical location of where the sample was extracted from was not known and may also have affected the observed wear particle size distribution. 24 EDX analysis is not sufficient by itself to determine the presence of UHMWPE wear particles because its carbon-oxygen composition is indistinguishable to the filter membrane and carbon coating used during characterization. However, in-house research expertise, personal experience and EDX analysis were combined to visually identify particles with morphology similar to those identified in previous research. Contaminants such as atmospheric silica particles could be excluded based on their elemental composition. The characterization and verification by EDX analysis of individual particles could be viewed as a strength of using manual particle characterization and give confidence to the identification of the particles. However, given billions of particles are produced by each joint replacement, the relatively small sample of characterized particles may be subject to selection bias during manual characterization. A different wear particle size distribution may have resulted if automated analyses were used instead, however both approaches have their strengths and weaknesses.
The full breadth of the available magnification provided by the SEM was utilized in this study to identify and characterize individual particles.
Because of this, more particles at the extremes of the size range would be expected to have been captured which, appears to have been reflected in the results. This differs from automated particle sizers which have a fixed size range and cannot discriminate between target particles and contaminants. Manual characterization also affected the number of particles characterized. Only approximately 100 to 400 UHMWPE wear particles were characterized per sample, the distribution of which may not be representative of the billions of wear particles produced by a joint replacement. 36 Future research should focus on simplifying the method and implementing it in tissues samples from known joint replacement device combinations. A method with this capability may also be useful to further understand the heterogeneity of wear particle populations between sample selection sites. 53

| CONCLUSION
It has been 25 years since the first wear particle isolation methods were developed, over which time the simplicity and effectiveness of such methods have proved to be inadequate for processing of tissues surrounding modern bearing couples. This feasibility study demonstrates the ability to capture high resolution images of a range of different material types of isolated wear particles that are representative of the total wear particle burden within an individual tissue sample. This study identified the prevalence of ceramic wear particles within the tissues of patients with cemented total joint replacements and also the relative scarcity of cobalt chromium alloy wear particles in the same devices. The images and particle characteristics included in this study could provide a reference for researchers attempting to emulate clinically relevant wear particles for use in in vitro biological response studies.