Antibacterial Titanium Implants Biofunctionalized by Plasma Electrolytic Oxidation with Silver, Zinc, and Copper: A Systematic Review

Patients receiving orthopedic implants are at risk of implant-associated infections (IAI). A growing number of antibiotic-resistant bacteria threaten to hamper the treatment of IAI. The focus has, therefore, shifted towards the development of implants with intrinsic antibacterial activity to prevent the occurrence of infection. The use of Ag, Cu, and Zn has gained momentum as these elements display strong antibacterial behavior and target a wide spectrum of bacteria. In order to incorporate these elements into the surface of titanium-based bone implants, plasma electrolytic oxidation (PEO) has been widely investigated as a single-step process that can biofunctionalize these (highly porous) implant surfaces. Here, we present a systematic review of the studies published between 2009 until 2020 on the biomaterial properties, antibacterial behavior, and biocompatibility of titanium implants biofunctionalized by PEO using Ag, Cu, and Zn. We observed that 100% of surfaces bearing Ag (Ag-surfaces), 93% of surfaces bearing Cu (Cu-surfaces), 73% of surfaces bearing Zn (Zn-surfaces), and 100% of surfaces combining Ag, Cu, and Zn resulted in a significant (i.e., >50%) reduction of bacterial load, while 13% of Ag-surfaces, 10% of Cu-surfaces, and none of Zn or combined Ag, Cu, and Zn surfaces reported cytotoxicity against osteoblasts, stem cells, and immune cells. A majority of the studies investigated the antibacterial activity against S. aureus. Important areas for future research include the biofunctionalization of additively manufactured porous implants and surfaces combining Ag, Cu, and Zn. Furthermore, the antibacterial activity of such implants should be determined in assays focused on prevention, rather than the treatment of IAIs. These implants should be tested using appropriate in vivo bone infection models capable of assessing whether titanium implants biofunctionalized by PEO with Ag, Cu, and Zn can contribute to protect patients against IAI.


Introduction
Implant-associated infections (IAI) are a devastating complication for patients receiving bone implants in total joint arthroplasty, trauma surgeries, and malignant bone tumor resections [1][2][3]. These infections form a tremendous burden for both patients and society. As the number of implantations continues to grow annually [4][5][6], the need for a cure increases. Given that the treatment of such infections is highly costly from both financial and societal points of view, the focus has shifted towards the prevention of IAI through the development of implants with intrinsic antibacterial activity.
Antibiotics form the primary source of antibacterial agents used to treat bacterial infections. However, a vast number of IAI is caused by Staphylococci and multiple strains have developed high levels of antibiotic resistance [7,8], raising concerns for the future

Data Extraction
Extracted information included the type of the titanium substrate, electrolyte composition, PEO processing parameters, surface topography, XRD phase composition, surface content of the incorporated elements, the release profile of the metallic (i.e., Ag, Cu, and Zn) ions, antibacterial assays, tested pathogens, eukaryotic cell types, and the outcomes (i.e., antibacterial behavior and cytocompatibility). The results were considered significant when p < 0.05.

Search Results
A total of 1261 studies were identified in the two search engines: 1190 from Google Scholar and 71 from Scopus. After screening the titles and abstracts, 1158 studies were excluded. The primary reasons for exclusion were no antibacterial or biocompatibility tests, PEO performed in combination with other surface modification techniques, and document types: reviews, patents, conference abstracts/papers, citations and case reports. As a result, 103 studies were selected for full-text analysis. The analysis led to the exclusion of 59 studies, as they failed to meet the inclusion criteria. Finally, 49 studies were included in this systematic review and were used for a qualitative analysis of their data and for comparison with each other. A flow diagram was created to represent the entire systematic search of the relevant studies (Supplementary Figure S1). The outline of the review is presented in Figure 1.

Summary of Study Characteristics
A summary of the study characteristics is presented in Figure 2. Of the analyzed studies, 43% used Ag, 26% used Cu, and 21% worked with Zn, while 9% investigated a combination of Ag, Cu, and Zn (i.e., using two or more metallic agents. Various types of parameters were reported in the studies (Figure 2A), including the PEO processing parameters (98%), phase composition (87%), surface content of the incorporated elements (80%), and ion release kinetics (48%). Furthermore, 92% of the studies quantified the antibacterial activity, which was reported to be >50% for 100% of the studies using Ag, 93% of the studies using Cu, and 73% of those employing Zn, as well as 100% of the studies combining multiple metallic agents ( Figure 2B). Of those studies, 57% tested the efficacy of the surfaces against S. aureus, 31% of the studies tested their specimens against E. coli, while 12% of the studies chose other bacterial species. Furthermore, the antibacterial activity was determined against adherent bacteria in 42% of the studies, while 35% of the studies assessed the antibacterial activity of their specimens against planktonic bacteria, and 23% assessed both.

Summary of Study Characteristics
A summary of the study characteristics is presented in Figure 2. Of the analyzed studies, 43% used Ag, 26% used Cu, and 21% worked with Zn, while 9% investigated a combination of Ag, Cu, and Zn (i.e., using two or more metallic agents. Various types of parameters were reported in the studies (Figure 2A), including the PEO processing parameters (98%), phase composition (87%), surface content of the incorporated elements (80%), and ion release kinetics (48%). Furthermore, 92% of the studies quantified the antibacterial activity, which was reported to be >50% for 100% of the studies using Ag, 93% of the studies using Cu, and 73% of those employing Zn, as well as 100% of the studies combining multiple metallic agents ( Figure 2B). Of those studies, 57% tested the efficacy of the surfaces against S. aureus, 31% of the studies tested their specimens against E. coli, while 12% of the studies chose other bacterial species. Furthermore, the antibacterial activity was determined against adherent bacteria in 42% of the studies, while 35% of the studies assessed the antibacterial activity of their specimens against planktonic bacteria, and 23% assessed both. Cytocompatibility was tested in 71% of all studies, of which 10% tested against multiple cell types ( Figure 2C). Of the studies assessing the cytocompatibility of their specimens, 78% used a cell line while 22% used cells obtained from a donor. The addition of the metallic antibacterial agent resulted in cytotoxicity for 13% of the Ag studies, 10% of the Cu studies, 0% of the Zn studies, and 0% of the studies combining two or more metals. Meanwhile, improved cell response (i.e., enhanced cell viability and/or osteogenic differentiation) was observed for 7% of the Ag surfaces, 50% of the Cu surfaces, and 33% of the Zn surfaces, as well as for 50% of the surfaces combining Ag, Cu, and Zn.

Synthesis and Characterization of PEO Biofunctionalized Surfaces
PEO is an electrochemical process that converts the outer oxide layer of valve metals into a ceramic surface layer and is applied to enhance corrosion resistance [44], dielectric properties [45], and biocompatibility [46] of the substrates. A PEO setup has two electrodes: the cathode and anode ( Figure 3A). Usually, either a constant current or voltage is applied, leading to the formation of an oxide layer on the anode (i.e., the specimen to be Cytocompatibility was tested in 71% of all studies, of which 10% tested against multiple cell types ( Figure 2C). Of the studies assessing the cytocompatibility of their specimens, 78% used a cell line while 22% used cells obtained from a donor. The addition of the metallic antibacterial agent resulted in cytotoxicity for 13% of the Ag studies, 10% of the Cu studies, 0% of the Zn studies, and 0% of the studies combining two or more metals. Meanwhile, improved cell response (i.e., enhanced cell viability and/or osteogenic differentiation) was observed for 7% of the Ag surfaces, 50% of the Cu surfaces, and 33% of the Zn surfaces, as well as for 50% of the surfaces combining Ag, Cu, and Zn.

Synthesis and Characterization of PEO Biofunctionalized Surfaces
PEO is an electrochemical process that converts the outer oxide layer of valve metals into a ceramic surface layer and is applied to enhance corrosion resistance [44], dielectric properties [45], and biocompatibility [46] of the substrates. A PEO setup has two electrodes: the cathode and anode ( Figure 3A). Usually, either a constant current or voltage is applied, leading to the formation of an oxide layer on the anode (i.e., the specimen to be treated). After dielectric breakdown, the oxide layer is thickened by spark discharges that lead to PEO biofunctionalization results in an altered surface morphology and chemical composition. In order to relate the antibacterial activity to certain surface characteristics, the surface of the biofunctionalized specimens is usually characterized (Tables 1-4). The important surface parameters in this regard are the surface topography, chemical composition, phase composition, and ion release profile. In the following sections, we will discuss the results regarding each of these parameters in more detail. PEO biofunctionalization results in an altered surface morphology and chemical composition. In order to relate the antibacterial activity to certain surface characteristics, the surface of the biofunctionalized specimens is usually characterized (Tables 1-4). The important surface parameters in this regard are the surface topography, chemical composition, phase composition, and ion release profile. In the following sections, we will discuss the results regarding each of these parameters in more detail.       Ref.
Ag and Cu

Titanium Substrate
Of the reviewed studies, most used commercially pure (CP) titanium (62%), followed by Ti6Al4V (23%), Ti6Al7Nb (4%) [15,32], Ti40Nb [52], Ti29Nb13Ta4.6Zr [59], and Ti15Mo [86]. Titanium is used for bone implants because of its mechanical properties, corrosion resistance and chemical biocompatibility [46,89]. Ti6Al4V has a higher strength to weight ratio than CP titanium and is, therefore, the natural choice for load-bearing applications, such as joint replacing implants, while CP titanium is more frequently applied for non-load bearing applications, such as maxillofacial implants [90]. Clinical studies comparing the long-term outcomes of patients treated with either CP-Ti or Ti-alloys are lacking [91,92].
Ti6Al4V implants may release vanadium and aluminum ions that can induce cytotoxicity [93]. Other alloys employing niobium have, therefore, been developed, including Ti6Al7Nb and Ti40Nb, which have similar mechanical properties, but do not induce cytotoxicity [94]. In addition, the cytotoxic effects of Al and/or V can be mitigated by PEO, since it reduces the ion release of those species [89]. PEO is easily scalable and can be applied to human-sized implants [95]. In order to translate the results from in vitro studies, it is, therefore, interesting to investigate the antibacterial behavior of substrates that are designed and produced like an implant, for instance, through additive manufacturing. This also highlights one of the advantages of PEO, namely that it can be applied on highly porous surfaces [31].

PEO Electrolyte
The bioactivity of PEO-biofunctionalized implant surfaces is determined for a large part by the composition of the PEO electrolyte, as the elements in the electrolyte eventually make up the chemical composition of the implant surface. More than 50% of the studies included in this systematic review used electrolytes with Ca and P elements. The presence of Ca and P in the electrolyte can result in the formation of hydroxyapatite, which forms more than 60% of bone tissue and is associated with a Ca/P ratio of 1.67 [96,97]. Calcium acetate and calcium glycerophosphate were the primary source of Ca, while CaCO 3 [52] and C 12 H 22 CaO 14 [67] were also used in some studies. P is usually added in the form of calcium glycerophosphate, β-glycerophosphate, H 3 PO 4 [52,69], K 4 P 2 O 7 [60], NaH 2 PO 4 [48,49,65,71,74,75], NaPO 3 [67], or Na 5 P 3 O 10 [85]. Another element used in about 30% of the included studies is Na in the form of NaOH, NaH 2 PO 4 [48,49,65,71,74,75], NaPO 3 [67], Na 5 P 3 O 10 [85], or Na 2 SiO 3 [73,76,78,98]. The addition of Na roughens the surface and enhances the Ca/P ratio [99], which has been shown to enhance the osteogenic cell response [100,101]. In addition, the implantation of Na through plasma immersion has been found to stimulate the osteogenic differentiation of cells [102]. Moreover, KOH [60,70,88] is used as an alternative base for NaOH given its similar effects on osteogenic differentiation [103].

PEO Processing Parameters
The electrical parameters of the PEO process affect the surface morphology [42], including the porosity [104], pore size [105], pore shape [106], and pore density [107], as well as the surface chemistry [83,84]. Of the included studies, 54% controlled the voltage, 31% controlled the current density, and 13% controlled both, while 1 study did not report the PEO processing parameters. The oxidation times ranged between 0 and 180 min, with 21% between 0-4 min, 50% between 5-9 min, 19% between 10-14 min, 6% between 15-19 min, and 4% ≥ 20 min. As the current density, voltage, or oxidation time increases, the spark discharge energy amplifies, affecting the mass of the oxide layer formed by a single pulse and resulting in enhanced growth of the oxide layer [40,108]. Furthermore, as temperature of the local discharge area increases, the plasma effect is enhanced, resulting in larger pore sizes and the transformation of amorphous TiO 2 to anatase and rutile phases. Meanwhile, the intensity of the spark discharge enhances with time, meaning that prolonged oxidation times result in the formation of hydroxyapatite on the implant surface [109,110]. As such, PEO processing parameters largely affect the chemical and phase composition as well as the surface topography of the implant surface.

Surface Morphology
As PEO greatly affects the surface topography of titanium surfaces, all studies investigated the surface topography by scanning electron microscopy (SEM) and most studies reported a porous surface topography with rounded pores ( Figure 4A). PEO transforms the native titanium oxide layer into a highly porous surface with interconnected porous networks, which is frequently described as a volcanic landscape with micropores that are <10 µm in diameter. In addition, flake-like morphologies [35,55,63] and needle-like structures [58] are often observed. Furthermore, the thickness and porosity of the oxide layer were shown to depend on the composition of the PEO electrolyte and PEO processing parameters [54,111]. The specifications of the surface morphology in turn were shown to greatly influence the antibacterial behavior [112] and osteogenic properties [113,114] of the implant surfaces. such, PEO processing parameters largely affect the chemical and phase composition as well as the surface topography of the implant surface.

Surface Morphology
As PEO greatly affects the surface topography of titanium surfaces, all studies investigated the surface topography by scanning electron microscopy (SEM) and most studies reported a porous surface topography with rounded pores ( Figure 4A). PEO transforms the native titanium oxide layer into a highly porous surface with interconnected porous networks, which is frequently described as a volcanic landscape with micropores that are <10 µm in diameter. In addition, flake-like morphologies [35,55,63] and needle-like structures [58] are often observed. Furthermore, the thickness and porosity of the oxide layer were shown to depend on the composition of the PEO electrolyte and PEO processing parameters [54,111]. The specifications of the surface morphology in turn were shown to greatly influence the antibacterial behavior [112] and osteogenic properties [113,114] of the implant surfaces.
These phases were shown to affect the biological response. For instance, TiO2 is transformed from an amorphous phase into crystalline anatase and rutile phases that were shown to produce reactive oxygen species (ROS) [117], which in turn contribute to the desired antibacterial behavior [118].

Content of the Antibacterial Elements Incorporated in the PEO Layers
The antibacterial activity of Ag, Cu, and Zn may be present on the implant surface depending on the dose [119][120][121]. Therefore, it is important to quantify the content of these
These phases were shown to affect the biological response. For instance, TiO 2 is transformed from an amorphous phase into crystalline anatase and rutile phases that were shown to produce reactive oxygen species (ROS) [117], which in turn contribute to the desired antibacterial behavior [118].

Content of the Antibacterial Elements Incorporated in the PEO Layers
The antibacterial activity of Ag, Cu, and Zn may be present on the implant surface depending on the dose [119][120][121]. Therefore, it is important to quantify the content of these elements on the implant surface after PEO biofunctionalization. This analysis is usually done either by energy-dispersive X-ray spectroscopy (EDS; Figure 4B) or X-ray photoelectron spectroscopy (XPS). Among the included studies, 80% reported the elemental composition of the surface, while 20% did not. The studies generally reported the elemental composition either in terms of atomic% or weight% and found them to correlate with the amount of Ag, Cu, and Zn dispersed in the PEO electrolyte. The amount of Ag incorporated in the implant surfaces tended to be lower (1.35 ± 1.82 wt%) than Cu (7.70 ± 10.17 wt%) and Zn (18.79 ± 12.06 wt%), reflecting the lower minimal inhibitory concentration (MIC) of Ag (0.03-8 µg/mL) as compared to Cu (256-448 µg/mL) and Zn (765 µg/mL) [122]. However, EDS does not exclusively measure the elemental composition of the surface but may penetrate deeper into the oxide layer. This is an important point, because it is not clear to what extent the species present deeper inside the oxide layer, which can be up to 10 µm in thickness, and contribute to the antibacterial properties of biofunctionalized implants [15]. The amount of active agents present on the implant surface may not be directly related to the antibacterial activity, since the form in which the element is present on the surface (i.e., ionic species, nanocrystals, or nanoparticles) affects the antibacterial properties as well [123,124].

Ion Release
An important antibacterial mechanism is through the release of metallic ions from the implant surface. These released ions do not only play a role in contact-killing, but also target planktonic bacteria in the implant surrounding, as this area could form a niche for bacteria [125]. Ion release was studied in 48% of the included studies and was measured from 12 h up to 56 days. Overall, the release of Ag, Cu, and Zn ions was found to be higher for the implant surfaces with a higher elemental content and a higher concentration of the active agents in the PEO electrolyte. The combination of Ag with Cu or Zn NPs on the implant surface resulted in enhanced Cu or Zn release while the Ag release was reduced in the first 24 h [19,20]. Similarly, higher concentrations of zinc acetate added to copper acetate resulted in enhanced Zn ion release while Cu ion release was reduced with higher concentrations of zinc acetate [77]. This may stem from galvanic coupling favoring the oxidation and release of one element over the other [126,127]. When studied in detail, this may allow for controlled release profiles and accompanying antibacterial effects.
Ion release results depend on the liquid in which these measurements are performed. Frequently used liquids are phosphate-buffered saline (PBS) and simulated body fluid (SBF) [128]. Ion release does not only depend on the surface content, but also on the form in which the antibacterial agent is present on the surface (i.e., as ionic species, nanoparticles, or other forms) [124]. Ideally, one could control the release of ions to not only prevent infection immediately after surgery, but also ward off late implant-associated infections [129]. However, comparing the reported ions release kinetics is difficult due to the different units, specimen designs, and measurement setups being used. In addition to the previously mentioned parameters, the surface area plays an important role in determining the concentration of the released ions, as a larger area allows for more agents to be incorporated on the surface, in turn leading to a higher release rate [31]. The reported concentrations of release ions should, therefore, be normalized with respect to the surface area of the specimens to enable direct comparison between different studies. The information regarding the surface area is generally not reported in the studies, rendering a direct comparison impossible.

Antibacterial Properties
Surface biofunctionalization by PEO with Ag, Cu, and Zn results in antibacterial surfaces. In the following section, we will first compare the antibacterial activity of PEO biofunctionalized titanium implants bearing Ag, Cu, and Zn found by in vitro and ex vivo studies (Table 5). Then, we will discuss the factors that determine the antibacterial activity. First of all, the types of the bacterial species and strains were shown to affect the susceptibility and resistance of bacteria to antibacterial agents [130], their ability to infect host cells [131], and their pathogenicity [132]. Moreover, the type of assay, the inoculation dose, and the culture time used in the studies may affect the observed antibacterial activity. Finally, the activity against adherent and/or planktonic bacteria is discussed, as the adherence of bacteria may initiate biofilm formation, while planktonic bacteria form a source for reinfection and host cell invasion [133].

Comparing Antibacterial Activities of Ag, Cu, and Zn
All the included studies reported antibacterial activity. Guidelines designate a material as antibacterial when it induces a >99.9% (i.e., 3-log) reduction in the number of viable bacteria [134]. However, this is a guideline for treatment, while the required reduction in the bacterial load for the prevention of IAI is not known. In fact, 48% of the studies using Ag, 14% of the studies with Cu, 10% of the studies with Zn, and 80% of the studies that combined these metallic agents reduced the bacterial load by >99.9%. This indicates that surfaces biofunctionalized with Ag demonstrate the highest degree of antibacterial activity, while Cu and Zn were less effective, which is not surprising given the much lower MIC for Ag as compared to Cu and Zn [122]. Interestingly, combining Ag, Cu, and Zn resulted in much higher levels of antibacterial activity, while the doses of single elements can be reduced [19,20,87,88].
Studies that focused on the antibacterial mechanisms of Ag, Cu, and Zn NPs suggest that two antibacterial mechanisms play a role: ion release killing [135] and the generation of reactive oxygen species (ROS) [136]. Ions released from the implant diffuse across the bacterial cell wall and penetrate into bacteria where vital bacterial structures are targeted. Meanwhile, ROS are highly reactive and cause lysis of the bacterial cell wall. It was found that Cu showed the best antibacterial activity as a result of contact killing [137], while Ag exhibited most of its antibacterial activity through both ion release and contact killing [138]. Furthermore, the synergistic antibacterial properties of AgNPs and Zn ions were observed to stem from long-range Zn ion release and contact-killing effects from Ag through microgalvanic coupling [29,139].
We plotted a 3D graph showing the correlation between antibacterial activity, cytocompatibility, and surface content of the antibacterial agent for the titanium substrates biofunctionalized by PEO with Ag, Cu, or Zn ( Figure 5). Very few studies reported all of these 3 parameters. This analysis shows that Ag indeed resulted in the highest levels of antibacterial activity at lower doses compared to Cu and Zn, yet also induced cytotoxicity more frequently. However, a direct comparison between the included studies, and thereby of Ag, Cu, and Zn bearing surfaces, was hampered by a large number of variables that differ in the various studies and are addressed in the next paragraphs of this section.

Comparing Antibacterial Activities of Ag, Cu, and Zn
All the included studies reported antibacterial activity. Guidelines designate a material as antibacterial when it induces a >99.9% (i.e., 3-log) reduction in the number of viable bacteria [134]. However, this is a guideline for treatment, while the required reduction in the bacterial load for the prevention of IAI is not known. In fact, 48% of the studies using Ag, 14% of the studies with Cu, 10% of the studies with Zn, and 80% of the studies that combined these metallic agents reduced the bacterial load by >99.9%. This indicates that surfaces biofunctionalized with Ag demonstrate the highest degree of antibacterial activity, while Cu and Zn were less effective, which is not surprising given the much lower MIC for Ag as compared to Cu and Zn [122]. Interestingly, combining Ag, Cu, and Zn resulted in much higher levels of antibacterial activity, while the doses of single elements can be reduced [19,20,87,88].
Studies that focused on the antibacterial mechanisms of Ag, Cu, and Zn NPs suggest that two antibacterial mechanisms play a role: ion release killing [135] and the generation of reactive oxygen species (ROS) [136]. Ions released from the implant diffuse across the bacterial cell wall and penetrate into bacteria where vital bacterial structures are targeted. Meanwhile, ROS are highly reactive and cause lysis of the bacterial cell wall. It was found that Cu showed the best antibacterial activity as a result of contact killing [137], while Ag exhibited most of its antibacterial activity through both ion release and contact killing [138]. Furthermore, the synergistic antibacterial properties of AgNPs and Zn ions were observed to stem from long-range Zn ion release and contact-killing effects from Ag through microgalvanic coupling [29,139].
We plotted a 3D graph showing the correlation between antibacterial activity, cytocompatibility, and surface content of the antibacterial agent for the titanium substrates biofunctionalized by PEO with Ag, Cu, or Zn ( Figure 5). Very few studies reported all of these 3 parameters. This analysis shows that Ag indeed resulted in the highest levels of antibacterial activity at lower doses compared to Cu and Zn, yet also induced cytotoxicity more frequently. However, a direct comparison between the included studies, and thereby of Ag, Cu, and Zn bearing surfaces, was hampered by a large number of variables that differ in the various studies and are addressed in the next paragraphs of this section. Figure 5. The relation between the antibacterial activity, cytocompatibility, and surface content for titanium surfaces biofunctionalized by PEO with Ag, Cu, or Zn. The reported antibacterial activity as a function of surface content and cytocompatibility is depicted by the blue dots. The green, red, and yellow projections enable a comparison between the parameters. Cytocompatibility is depicted as cytotoxicity (−), no effect (0), or enhanced cytocompatibility (+).

Bacterial Species and Strains
Antibacterial results are affected by the tested bacterial species. Of the reviewed studies, 57% used S. aureus, 31% E. coli, and 12% other bacterial species, including S. epidermidis Figure 5. The relation between the antibacterial activity, cytocompatibility, and surface content for titanium surfaces biofunctionalized by PEO with Ag, Cu, or Zn. The reported antibacterial activity as a function of surface content and cytocompatibility is depicted by the blue dots. The green, red, and yellow projections enable a comparison between the parameters. Cytocompatibility is depicted as cytotoxicity (−), no effect (0), or enhanced cytocompatibility (+).

Bacterial Species and Strains
Antibacterial results are affected by the tested bacterial species. Of the reviewed studies, 57% used S. aureus, 31% E. coli, and 12% other bacterial species, including S. epidermidis [63,86], S. sanguinis [61], S. mutans [57,85], P. aeruginosa [82], and P. gingivalis [64]. Given that Ag, Cu, and Zn form an alternative to antibiotics, it is important to analyze the results on antibiotic resistant bacteria, such as MRSA, which are involved in up to 32% of fracture-related infections [140,141]. MRSA was investigated in 9 studies and found to be strongly inhibited by Ag [15,31,32,50,65], Ag and Cu [19], Ag and Zn [20], Cu and Zn [88] bearing surfaces, while one study that included Zn surfaces did not observe any inhibition [86]. Thukkaram et al. observed that the antibacterial effect of Ag containing surfaces against MRSA was lower compared to S. aureus and E. coli, although with increasing doses of Ag, all bacterial species were targeted equally [65]. Furthermore, testing on multiple species was performed in 19% of the included studies. No studies tested multiple species in a single experiment (i.e., co-culture of multiple species), which would be of interest given that 10-20% of IAI are induced by polymicrobial infections [142,143].
We can, thus, conclude that most studies investigated antibacterial behavior against S. aureus. This bacterial species causes 20-46% of IAI [144][145][146]. Other gram-positive species, such as Streptococci caused up to 10% and Enterococci 3-7% of cases [147]. Enterococci have not been tested in studies with PEO-treated surfaces bearing Ag, Cu, or Zn. Gramnegative bacteria, such as Pseudomonas aeruginosa and Enterobacteriaceae induce 6-17% of IAI [143,148]. Given the relatively low rate of IAI induced by Enterobacteriaceae, it is surprising that 31% of the studies investigated the effects of the implant surfaces on E. coli. While some studies that analyzed both S. aureus and E. coli reported a stronger antibacterial effect against E. coli as compared to S. aureus [49,51,55,65,83,98], others reported a similar antibacterial effect for both species [35,37,70,74,84,88]. Interestingly, up to 42% of IAI in patients were caused by culture-negative (i.e., undefined) bacteria [149,150] and therefore warrant an antibacterial agent effective against a wide antimicrobial spectrum.
Among bacterial species, different levels of sensitivity to antibacterial agents have been reported [151], including against Ag and Cu [152]. To what extent the differences between strains plays a role depends on the bacterial species. The differences between strains in terms of their MIC/MBC values was found to be negligible for S. aureus, but were quite large in the case of E. coli strains [153]. It is, therefore, important that the bacterial strain is properly reported, which was done only in 79% of the included studies. Only one study, conducted by Leśniak-Ziółkowska et al., compared different strains within a bacterial species, namely S. aureus (ATCC 25,923 and clinical MRSA 1030) and S. epidermidis (ATCC 700,296 and clinical 15560) [86]. No strain-dependent differences were observed after 4 h using a bacterial adhesion test.

Source of Antibacterial Agent
Antibacterial behavior depends not exclusively on the antibacterial agent, but also on the form in which Ag, Cu, and Zn are added to the PEO electrolyte and are subsequently incorporated onto the titanium implant surface [124]. Ag, Cu, and Zn elements are either completely dissolved in the electrolyte or are added in the form of NPs that form a suspension. The former will end-up in the form of chemical compounds present all over the surface, while the latter (NPs) are spread over the surface. NPs may form a reservoir from which ions are released, thereby ensuring prolonged antibacterial activity [154]. In addition, the shape of the NPs determines the antibacterial activity as the surface-to-volume ratio affects the ion release and, thus, the efficacy of the surface biofunctionalization process [155]. Ionic forms only induce antibacterial activity through the action of ions, while NPs also produce reactive oxygen species and induce contact-killing [156]. Among the included studies, 33% used NPs, 64% employed ionic species, and only a study by Zhang et al., combined ions and NPs [87]. This study combined Ag NPs with Zn acetate, which resulted in much higher release of Zn ions compared to Ag ions. Furthermore, the antibacterial activity was assessed against both adherent and planktonic S. aureus after 24 h. The developed surface demonstrated significant antibacterial behavior with increasing concentrations of Ag and Zn leading to further reduction of viable bacteria. The authors reasoned that the antibacterial activity stems from ROS generation by both Ag and Zn as well as Ag + release. Moreover, both Ag and Zn ion concentrations remained below cytotoxicity levels and thus stressed the utility of combining these elements. Studies that investigate the differences in the antibacterial properties induced by NP and ionic forms are lacking.

Analysis Method
Antibacterial properties can be investigated by different assays. Properties often investigated are the antibacterial leaching activity, the killing of adherent bacteria, and the prevention of biofilm formation. Although most of the included studies used only one antibacterial assay (53%), the use of several assays is required for the assessment of the various types of antibacterial properties [157]. Therefore, 32, 8, and 8% of the included studies used 2, 3, and 4 assays, respectively. To determine the leaching effects of the antibacterial ions released from the PEO surfaces, a zone of inhibition assay or a Kirby-Bauer assay is often used. The number of bacteria can be quantified either through a direct CFU count, by spread plate analysis, or by staining the live cells using a fluorescent dye. A few studies referred to ISO [51,59] and ASTM [37,76,78] standards. With SEM, adherent bacteria and/or biofilm formation can be visualized in a non-quantitative manner. A wide variety in the type of assays used in the studies was found, with spread plate analysis (33%), SEM (24%), and viability fluorescence imaging (12%) being the most frequently applied assays.
In addition to in vitro assays, ex vivo models were explored, in which infected implants biofunctionalized with Ag and Cu, Zn, or Sr are inserted into a murine femur [19,20,31,50]. Subsequently, the number of CFU present are quantified (e.g., after 24 h). Although this ex vivo model does not allow to assess the effects of the implants on the immune system or bony ingrowth, some of the other in vivo effects such as those of the extracellular matrix and bone tissue [158] can be captured to some extent. Indeed, the gene expression profile of osteocytes was found to be similar between an ex vivo bone infection model and tissue samples from IAI patients [159]. Thus far, no study has tested the antibacterial activity of titanium implants biofunctionalized by PEO with Ag, Cu, and Zn in vivo.

Duration and Inoculum of Antibacterial Assay
Over two thirds of IAIs are initiated during surgery [160]. A rapid antibacterial response to prevent the adherence of the bacteria that enter the human body peri-operatively is, therefore, desired. Almost all of the included studies (94%) tested the antibacterial properties within 24 h and 10% even within 2 h. However, IAI can also be initiated long after surgery, stemming from hematogenous origins. Prolonged antibacterial activity is, thus, desirable too [72,85,161]. Zhang et al., reported on the antibacterial activity of Cucontaining surfaces for longer periods of time [72]. It was observed that the number of viable adherent bacteria was significantly reduced on surfaces containing 0.67-1.98 wt% Cu up to 96 h. However, this was one of the few studies aiming to assess long-term antibacterial behavior, since prolonged in vitro culture of bacteria is challenging. Research into late IAI is, therefore, primarily performed in vivo [162,163].
The inoculum used in the antibacterial assays is another factor determining the antibacterial behavior of PEO-biofunctionalized implants. The exact number of bacteria required for IAI is unknown, but it was shown that the presence of a foreign body can reduce the infection dose by 6 orders of magnitude [164] due to a hampered immune response [165]. The inoculum used in the included studies varied widely between 250 [53] and 10 9 CFU/mL [58,85], and was not reported in two studies. Currently, most inocula are presented per volume or as a measure of optical density. However, the surface area of the implant is also of importance, as more area with more incorporated antibacterial agent is likely to have a greater antibacterial effect. Therefore, presenting the inoculum per volume per surface area would support comparative analyses of different studies.

Planktonic vs. Adherent Bacteria
As both planktonic and adherent bacteria play an important role in IAIs, antibacterial implants should target both types of bacteria. Planktonic bacteria are present in the fluid and tissue surrounding the implant and have shown to be a reservoir for late-stage reinfections [125]. Once the bacteria adhere to the implant, bacteria should be targeted in order to prevent biofilm formation as this would induce bacterial resistance to antibiotic treatment [166]. In this respect, 42% of the included studies investigated antibacterial activity against adherent, 35% against planktonic, and 23% against both planktonic and adherent bacteria. Targeting both planktonic and adherent bacteria should, therefore, be emphasized more in future studies.

Biocompatibility
In addition to antibacterial properties, PEO-biofunctionalized implant surfaces should not induce cytotoxicity, and ideally even enhance cell response and bony ingrowth. The compatibility of the implants with mammalian cells is, therefore, an important topic that needs to be thoroughly investigated for any such implant. Several of the included studies report the results of such in vitro cytocompatibility experiments, which are affected by the type of the assay, cell type, and cell source (Supplementary Table S1).

Cytocompatibility of Ag, Cu, and Zn Surfaces
Cytocompatibility was investigated in 71% of studies. In those studies, Ag induced cytotoxicity in 13% of the studies, while 10% of the studies investigating Cu and 0% of those employing Zn reported cytotoxic effects. None of the studies combining Ag, Cu, and Zn reported cytotoxicity. Cell response of the implants was improved in 7% of the studies using Ag, 50% of the studies focused on Cu, and 33% of the studies with Zn, as well as for 50% of the studies in which two or more antibacterial agents were combined. The control group often consists of PEO biofunctionalized surfaces without antibacterial elements. Cytotoxicity is, therefore, not considered a major concern by the vast majority of the included studies. Indeed, Cu and to somewhat lesser extent Zn were shown to improve the cytocompatibility of PEO-treated implants.

Cell Type
The cellular response was shown to differ in in vitro experiments between different cell types [176,177]. In the reviewed studies, pre-osteoblasts (32%), osteosarcoma cells (22%), fibroblasts (20%), MSCs (17%) and SV-HFO, macrophages, adipose stem cells, and endothelial cells (each in 1 study) were used. Pre-osteoblasts and MSCs are the main cells responsible for bone formation [178,179]. Osteosarcoma and SV-HFO cells [180] are immortalized cells stemming from the osteogenic lineage. However, osteosarcoma was shown to stem from defective differentiation [181]. Since these titanium implants will be used in bone tissue, it was surprising that 29% of the studies did not analyze the effects of the implants on bone-forming cells. Other cell types may support bone formation through indirect pathways. Endothelial cells play a role in angiogenesis, which plays a major role in bone regeneration as blood vessels carry nutrients and oxygen and facilitate the transport of immune cells to the regenerating bone tissue [182]. Meanwhile, macrophages form an important part of the immune response against IAI. Any potential toxicity of the synthesized implants against this cell type is of concern, as it may hamper the clearance of infections [16,183]. Finally, fibroblasts were shown to regulate osteoblast activity through tight junction interactions [184].

Cell Source
About 22% of the included studies used primary cells, whereas 78% utilized cell lines. Primary cells are more representative of the clinical situation, as they have been isolated from donors. However, their variability is high. Cells from multiple donors, therefore, need to be tested [185]. Cell lines, on the other hand, are homogenous and stable, while exhibiting little variability. However, their immortalized nature makes them differ from the clinical situation [186]. Furthermore, the source of animal species from which the cells were derived differed greatly between the included studies, with 56% using murine cells, 34% human cells, and 10% rat cells. The osteogenic differentiation capacity of stem cells is known to differ between human, mice, and rat MSCs [187,188]. These differences in animal species make it difficult (if not impossible) to directly compare the cytocompatibility results reported in the different studies.

Discussion
In order to prevent IAI, the biofunctionalization of titanium implants by PEO using Ag, Cu, and Zn as the active agents has gained significant momentum in the last decade. Therefore, we systematically reviewed the progress made on those implants and summarized the various types of properties measured for such types of PEO-biofunctionalized implants.

Antibacterial Results
From the results of this study, it can be concluded that Ag is the most potent antibacterial agent followed by Cu and Zn. It is important to stress that different studies utilized different experimental protocols to determine the antibacterial properties of PEObiofunctionalized implants. It was shown that titanium surfaces bearing Ag, Cu, and Zn can kill bacteria through antibacterial leaching activity, contact killing, and the formation of ROS [156,189]. These properties cannot be assessed in a single assay. The use of multiple assays is, therefore, warranted to support the claim of antibacterial activity [157]. Finally, it is important to make sure that the assays assess infection prevention rather than infection treatment.
Furthermore, the bacterial species and strains used were found to affect the level of antibacterial activity. For instance, surfaces demonstrating antibacterial activity against E. coli may not do the same against S. aureus [51,55,98]. Most studies investigated the antibacterial activity of the implants against S. aureus or E. coli. While a large proportion of IAI was induced by S. aureus, only a small proportion of infections were caused by E. coli [144,147]. The rationale for choosing E. coli was, thus, primarily methodological convenience rather than clinical prevalence. Meanwhile, S. epidermidis or polymicrobial infections were rarely studied, even though they cause a significant proportion of IAI [142][143][144]. Moreover, the antibacterial behavior of PEO-biofunctionalized implants should be assessed in environments co-habited by multiple bacterial species, as this was shown to influence the resistance profiles of bacteria [130].
The antibacterial experiments aimed to mimic the clinical situation as closely as possible. In this respect, both adherent and planktonic bacteria should be warded off, as adherent bacteria can form biofilms [133], while planktonic bacteria may infect the peri-implant tissue and form a reservoir for late-stage reinfection [125]. Furthermore, an antibacterial implant should prevent infections that occur immediately after surgery, as that is the point where most IAI occur [160], as well as late-stage infections from hematogenous origins [161]. At the moment, the focus primarily lies on preventing early-stage infections. Ultimately, Ag, Cu, and Zn may form an alternative to antibiotics, as bacteria are developing ever-growing degrees of antibiotics resistance [144,190]. As such, the development of resistance against Ag, Cu, and Zn and combination thereof is worthwhile to investigate given that resistance against Ag, Cu, and Zn was reported in vitro [191][192][193] and in patients [194].
The observed antibacterial activity depends on a wide variety of factors described in this review, including the titanium substrate, composition of the PEO electrolyte, and PEO processing parameters that in turn affect the surface morphology, phase composition, surface content of the incorporated antibacterial agent, and ion release profile. These parameters determine the antibacterial properties and biocompatibility of the implants. The measured antibacterial properties are highly dependent on the bacterial species and strains used, experimental techniques, the duration of the assays, bacterial inoculum, and the type of bacteria against which the implant performance is measured (i.e., planktonic and/or adherent). As for biocompatibility, the type of the assays, cell type, and cell source could all influence the final read-outs. These factors varied between the studies included in this review and make a one-to-one comparison between the different studies challenging.
The antibacterial activity is dependent on the dose of Ag, Cu, and Zn present on the surface of the titanium implants [119][120][121]. It is, therefore, essential to determine the amount of these elements present on the surface. In addition, the Ag, Cu, and Zn ions released from the implant surface are responsible for a significant part of the antibacterial activity, which is why it is important to measure the concentration of the ions released from the implant surface. From the results, it is clear that the surfaces bearing Ag had much lower elemental content and ion release as compared to those bearing Cu and Zn, which was expected due to the lower MIC of Ag as compared to Cu and Zn [122]. Both the surface content and ion release were also dependent on the surface area, as a larger surface area allows for the incorporation of a greater amount of elements and, thus, increased ion release [31]. Therefore, describing these properties relative to the surface area may aid in a comparison between the results of different studies.

Biocompatibility
Most of the included studied found cytotoxicity to be a minor concern, with Ag inducing cytotoxicity in 13% of the studies. It was striking that 29% of the included studies did not investigate the effects of the implants on bone-forming cells, even though the implants are intended for bone tissue. In addition, cytotoxicity against other cell types, such as endothelial cells and immune cells is of interest, as these cells contribute to bone regeneration as well [195,196]. Furthermore, the use of cell lines vs. donor cells and different mammalian species complicates the comparisons between different studies [197]. Moreover, biocompatibility needs to be investigated both in vitro and in vivo, as the results of in vitro and in vivo experiments are known to differ, for instance, in the case of Agbearing surfaces [16].
Another way to enhance the cytocompatibility of PEO-biofunctionalized implants is by combining two or more antibacterial metals (i.e., Ag, Cu, and Zn), as synergic effects between various such agents are reported to exist [19,20] and could be used to reduce the concentration of Ag [126,198]. In addition, combining these elements with other osteogenic elements, such as Sr [50] may enhance their antibacterial and biocompatible properties. Finally, the combination of multiple antibacterial elements significantly reduces the risk of the development of bacterial resistance, thereby ensuring that the prolonged use of these elements will remain possible [199].
PEO is frequently applied in combination with other surface treatments, such as hydrothermal treatment [200] and physical vapor deposition [201] to alter the chemical and phase composition of the surface. This may result in improved antibacterial behavior [98]. Furthermore, hydrothermal treatment has resulted in the enhanced formation of hydroxyapatite crystals, yet may reduce corrosion resistance too [202]. A major disadvantage of these additional surface treatments is that they make the entire process lengthier and more complex, thus making it more difficult to upscale the production of clinically sized implants.

Towards Clinically Relevant Implants
A decade of PEO biofunctionalization of titanium implants with Ag, Cu, and Zn confirmed the great potential of this method as an effective, fast, and scalable process. At the moment, however, the research on antibacterial PEO-biofunctionalized titanium implants is still far away from clinical application, as the research was primarily conducted in vitro with few studies also exploring ex vivo models [20,50]. Furthermore, PEO was shown to enhance the osteogenic capacity of titanium implants in vivo [38,39,203], including surfaces bearing Zn [204]. However, these studies did not analyze the antibacterial properties of such implants, which should be evaluated using bone infection models [205]. In this respect, a major limitation of the state-of-the-art techniques is their limited relevance for the assessment of the preventive potential of antibacterial implants (as opposed to their treatment potential). However, studying prevention requires a much larger sample size, as it is associated with lower bacterial loads, meaning that infections are less likely to occur. This lower risk of infection has major ethical and financial implications. In addition, future implants will most likely be fabricated by AM and as such be highly porous. Not only is the risk of infection of such volume-porous implants higher, their IAI treatment is also highly challenging due to their usually high degree of bony ingrowth that may cause significant bone loss during their removal. The development of antibacterial surface treatments for such types of implants is, thus, highly relevant. In fact, the additional surface area of such implants may be exploited to enhance the bioactivity of PEO biofunctionalized implants [31].

Conclusions
In order to combat IAI, the biofunctionalization of titanium implants by Ag, Cu, and Zn has gained significant momentum in recent years and resulted in the synthesis of potent antibacterial and biocompatible surfaces. Implant biofunctionalized with Ag, Cu, and Zn demonstrated significant antibacterial behavior against a wide bacterial spectrum, including antibiotic-resistant bacterial strains. However, the antibacterial properties of these implants were primarily investigated in vitro and occasionally ex vivo. Furthermore, many studies do not reach sufficiently high antibacterial levels, as indicated by international guidelines. Moreover, the biofunctionalization of volume-porous AM implants has not been investigated extensively. Finally, combining Ag, Cu, and Zn on the surface of titanium implants was shown to result in potent antibacterial surfaces with reduced cytotoxicity. In order to take the PEO biofunctionalization of titanium implants by Ag, Cu, and Zn to clinical settings, in vivo studies should be conducted using relevant infection models for both solid and volume-porous bone implants.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/ijms22073800/s1, Figure S1: Flow diagram of the systematic literature search., Table S1: The biocompatibility of PEO-modified Ti-based surfaces bearing single or multiple elements.

Funding:
The research for this paper was financially supported by the Prosperos project, funded by the Interreg VA Flanders-The Netherlands program, CCI grant no. 2014TC16RFCB046.

Conflicts of Interest:
The authors declare no conflict of interest.