Influence of Zinc and Magnesium Substitution on Ion Release from Bioglass 45S5 at Physiological and Acidic Ph

Biomed. Glasses 2015; 1:93–107

Research Article Open Access

Max Blochberger, Leena Hupa, and Delia S. Brauer* Influence of zinc and magnesium substitution on ion release from Bioglass 45S5 at physiological and acidic pH

DOI 10.1515/bglass-2015-0009 Received May 04, 2015; accepted Aug 14, 2015 1 Introduction

Abstract: Ion release of Mg- and Zn-substituted Bioglass Glass ionomer cements (GIC), or glass polyalkenoate ce- 45S5 (46.1 SiO2-2.6 P2O5-26.9 CaO-24.3 Na2O; mol%; with 0, ments, have been used successfully in dentistry since 25, 50, 75 or 100% of calcium replaced by magnesium/zinc) the 1980’s. They can form chemical bonds to hydroxyap- was investigated at pH 7.4 (Tris buffer) and pH 4 (acetic atite [1], which is the mineral phase of teeth and bone, and acid/sodium acetate buffer) in static and dynamic dissolu- medical grade metal alloys. In addition, they have been tion experiments. Despite Mg2+ and Zn2+ having the same shown to release therapeutic ions [2, 3]. These two proper- charge and comparable ionic radii, they influenced the ties make them of interest for use as bone adhesives to re- dissolution behaviour in very different ways. In Tris, Mg- pair fractures or as injectable bone fillers, e.g. in vertebro- substituted glasses showed similar ion release as 45S5, plasty or kyphoplasty. However, dental GIC contain and re- while Zn-substituted glasses showed negligible ion release large amounts of aluminium ions, which are neuro- lease. At low pH, however, release behaviour was similar, toxic and negatively affect bone mineralisation [4]. with all glasses releasing large percentages of ions within Finding a non-toxic substitute for aluminium- a few minutes. Precipitation of crystalline phases also var- containing GIC has been challenging, as aluminium plays ied, as Mg- and Zn-substitution inhibited apatite forma- an important role in both glass degradation (acid hydrol- tion, and Zn-substitution resulted in formation of zinc ysis of Si—O—Al bonds) and in GIC stability (cross-linking phosphate phases at low pH. These results are relevant of polyalkenoate chains by Al3+). Most studies have fo- for glasses used in aluminium-free glass ionomer bone ce- cussed on glasses of low network connectivity, which ments, as they show that Zn/Mg-substituted glasses re- contain relatively large amounts of either zinc [5] or mag- lease ions similarly fast as glasses containing no Zn/Mg, nesium [3, 6]. Despite zinc and magnesium ions having suggesting that these ions are no prerequisite for ionomer similar ionic radii and field strengths [7], the properties glasses. Zn-substituted glasses may potentially be used as of GIC have been shown to differ greatly depending on controlled-release materials, which release antibacterial whether they were prepared using zinc- or magnesium- zinc ions when needed only, i.e. at low pH conditions (e.g. containing bioactive glass [6]. These differences may be bacterial infection), but not at normal physiological pH caused by magnesium and zinc-containing glasses having conditions. different solubility, particularly at low pH values relevant for GIC setting reactions. Keywords: intermediate oxide; bioactive glass; phospho- Zinc- [8] and magnesium-containing [9] bioactive silicate glass; dissolution; glass ionomer cement; continu- glasses have also attracted interest for other reasons. Zinc ous flow plays a variety of roles in mammalian systems; it is known for its antibacterial effects but is also essential for growth and normal development [10]. It acts as a co-factor for *Corresponding Author: Delia S. Brauer: Otto Schott Institute of over 400 enzymes in the body, including alkaline phos- Materials Research, Friedrich Schiller University Jena, Fraunhofer- phatase, an enzyme essential for normal bone minerali- str. 6, 07743 Jena, Germany; Email: [email protected]; Tel.: +49-3641-948510; Fax: +49-3641-948502 sation and formation. Zinc deficiency, which is common Leena Hupa: Johan Gadolin Process Chemistry Centre, Åbo in the elderly, has been associated with abnormal bone Akademi University, Piispankatu 8, 20500 Turku, Finland growth [11, 12], poor rates of wound repair [13] and a range Max Blochberger: Otto Schott Institute of Materials Research, of dermal, psychiatric and gastrointestinal disorders [14]. Friedrich Schiller University Jena, Fraunhoferstr. 6, 07743 Jena, Zinc has also been shown to have a stimulatory action on Germany

Table 1: Nominal glass composition (mol%) and network connectivity calculated assuming Zn/Mg to be acting as a modifier (NC) or entering the silicate network (NC*).

* Glass SiO2 P2O5 CaO Na2O MgO ZnO NC NC Mg100 46.1 2.6 - 24.4 26.9 - 2.12 4.28 Mg75 46.1 2.6 6.7 24.4 20.2 - 2.12 3.91 Mg50 46.1 2.6 13.4 24.4 13.5 - 2.12 3.44 Mg25 46.1 2.6 20.2 24.4 6.7 - 2.12 2.87 45S5 46.1 2.6 26.9 24.4 - - 2.12 2.12 (Mg0/Zn0) Zn25 46.1 2.6 20.2 24.4 - 6.7 2.12 2.87 Zn50 46.1 2.6 13.4 24.4 - 13.5 2.12 3.44 Zn75 46.1 2.6 6.7 24.4 - 20.2 2.12 3.91 Zn100 46.1 2.6 - 24.4 - 26.9 2.12 4.28 bone formation both in vitro and in vivo [15, 16] and is 2 Methods also a potent inhibitor of osteoclastic bone resorption in vitro [17]. Magnesium is an important element present in bone apatite [18], and magnesium (in addition to calcium) 2.1 Glass synthesis intake is considered necessary for preventing osteoporo- Glasses in the system SiO -P O -Na O-CaO/MgO/ZnO sis [19]. Magnesium also acts as a co-factor in about 300 en- 2 2 5 2 were prepared using a melt-quench route. CaO was re- zymes, thus playing a key role in energy metabolism and placed by MgO or ZnO in increasing amounts (0 to 100%; nucleic acid synthesis, and it is an essential element for Table 1) on a molar base to maintain a constant network the nervous system, being involved in conduction mecha- connectivity [28], assuming both act as modifiers. Mixtures nisms [20]. of SiO , CaCO , NaPO , MgCO and ZnO were sintered to- in vitro dissolution and cell culture experiments are 2 3 3 3 gether in a platinum crucible inside an electric furnace at usually performed at normal physiological pH (i.e. around 1250∘C for one hour and then melted for 1 hour at 1350∘C.A pH 7.4) [21]; however, low pH conditions are relevant dur- batch size of approximately 150 g was used. After melting, ing a range of clinical conditions, e.g. during bacterial in- the glasses were rapidly quenched into water to prevent fections [22]. Therefore, a better understanding of bioac- crystallisation. Glasses were crushed in a steel mortar and tive glass dissolution in various pH environments can help sieved using analytical sieves. us to improve bioactive glass performance [23, 24]. Partic- Glass monoliths were prepared by re-melting the glass ularly if used in GIC the dissolution and ion release be- frit at 1350∘C, pouring into a brass mould, placing into a haviour of bioactive glasses at acidic pH needs to be well pre-heated furnace set to 30 K below below glass transition understood to enable optimisation of glass degradation, temperature and allowing to cool to room temperature in GIC setting behaviour and, subsequently, GIC properties. the switched off furnace over night. The composition of se- Bioglassr 45S5 has been successfully used to regen- lected glasses was analysed on polished monoliths using erate bone, owing to its ability to degrade, release ions energy-dispersive X-ray spectroscopy (EDX; Jeol JSM-7001F and precipitate apatite [25, 26]. While usually alkali oxide- scanning electron microscope equipped with an EDAX Tri- free glasses are used in GIC (as the presence of low field dent analysing system). strength alkali ions may be expected to impair GIC mechanical properties), a recent patent describes the use of 45S5 in the preparation of GIC for use in human and vet- erinary medicine [27]. Here, we use Bioglassr 45S5 as a 2.2 Static dissolution experiments model glass system to study the influence of zinc or magne- Dissolution studies were performed at pH 7.4 (Tris buffer sium substitution for calcium on ion release at low (pH 4) solution) and pH 4 (acetic acid/sodium acetate buffer so- and physiological pH. lution). 0.062 mol L−1 tris(hydroxymethyl)amino methane (Tris) solution was prepared by dissolving 7.545 g of Tris in 800 mL of deionised water and adding 44.2 mL of 1 mol L−1 Influence of zinc and magnesium substitution on ion release... Ë 95

Table 2: Analysed glass composition (mol%)

Glass SiO2 P2O5 CaO Na2O MgO ZnO 45.9 2.4 27.9 24.3 45S5 ± 1.8 ± 0.2 ± 1.2 ± 1.5 45.6 2.4 16.2 23.7 12.1 Mg50 ± 1.8 ± 0.2 ± 0.8 ± 1.5 ± 0.7 45.7 2.3 1.1 25.0 26.0 Mg100 ± 1.8 ± 0.2 ± 0.3 ± 1.9 ± 1.6 47.7 2.5 12.1 21.1 16.6 Zn50 ± 2.3 ± 0.2 ± 0.5 ± 2.4 ± 1.0 45.3 2.4 0.0 26.3 26.1 Zn100 ± 1.8 ± 0.2 ± 0.0 ± 3.5 ± 1.4

tions were analysed using inductively coupled plasma optical emission spectroscopy (ICP-OES, Varian Liberty 150, Agilent Technologies, Böblingen, Germany). Experiments were performed in triplicates, and results are presented as percentage of ions present in solution relative to the amount of the respective ion present in the untreated glass (mean ± standard deviation, SD).

2.3 Powder analysis

The retained powders after treatment in Tris or HAc/NaAc buffer were rinsed with acetone and dried. Treated andun- treated glass powders were characterised by powder X-ray diffraction (XRD; D5000, Siemens, Germany; CuKα; step ∘ Figure 1: XRD patterns of untreated glasses. time 31 s; step width 0.02 ; data collected at room temperature) and Fourier transform infrared spectroscopy (FTIR; Nicolet Avatar 370 DTGS, Thermo Electron Corporation, HCl solution. The solution was heated to 37∘C, the pH ad- Waltham, Massachusetts, USA). justed to 7.4 (pH meter HI 8314 with pH electrode HI 1217 D, HANNA Instruments, Kehl am Rhein, Germany) using −1 1 mol L HCl solution and the volume filled to 1000 mL. 2.4 Dynamic dissolution experiments 0.1 mol L−1 acetic acid/sodium acetate (HAc/NaAc) buffer was prepared by adding 6.005 mL of 100% acetic Ion release profiles for the glasses were measured using acid (Rotipuranr, Carl Roth GmbH Co. KG, Karlsruhe, Ger- a dynamic flow cell connected to an inductively coupled many) to 900 mL deionised water, adjusting the pH to plasma optical emission spectrometer (Optima 5300 DV, −1 pH 4 ± 0.05 using 1 mol L sodium hydroxide solution Perkin Elmer) as described previously [29, 30]. (Carl Roth) and filling the volume to 1000 mL. Additional The ion concentrations were measured on-line every HAc/NaAc buffer solution of pH 5 was prepared as a con- 12 s (one replicate per measurement), and the experimen- trol. tal set-up is described in detail elsewhere [30, 31]. The 75 mg glass (< 250 µm) was immersed in 50 mL Tris or sample cell was filled with glass particles (225 ±5 mg; ∘ HAc/NaAc solution at 37 C for 3, 6, 24 and 72 hours with < 212 µm), and random packing was assumed. A peri- additional time points of 15 min (HAc/NaAc only) and 168 staltic pump fed the medium vertically upwards through hours (Tris only). After each time period, the pH was mea- the bed of glass particles. The flow rate was adjusted to sured and samples were filtered through medium porosity 0.2 mL min−1 to achieve laminar flow [31]. Temperature filter paper (5 µm particle retention, VWR International) was kept at 37 ± 2∘C, and measurements were background and acidified using nitric acid (65%). Elemental concentra- corrected [31]. Results are presented as concentrations in 96 Ë M. Blochberger et al.

(a) (b)

Figure 2: pH changes over time after immersion of Mg (a,b) and Zn-substituted glasses (c,d) in HAc/NaAc (a,c) and Tris buffer (b,d). (Lines are visual guides only). solution normalised to the molar amount of the respective by contrast, only 45S5 and Mg-substituted glasses caused element in the glass. a noticeable pH rise (from pH 7.4 to pH 7.9 ± 0.1; Figure 2b), while pH remained relatively constant (not rising above pH 7.5) when Zn-substituted glasses were immersed (Fig- 3 Results ure 2d). In HAc/NaAc buffer, all glasses (shown for 45S5, Mg50 and Zn50 in Figure 3a,c,e) showed fast ion release, with 3.1 Glass formation up to 60% of ions being released within the first 3 hours of the experiment. In Tris buffer, 45S5 and Mg-substituted Glasses were obtained in an amorphous state as shown glasses (Figure 3b,d) showed the same trend; however, by XRD (Figure 1). Analysed glass compositions are given the overall ion release was significantly slower, with only in Table 2. up to about 30% of ions being released at 3 hours. Zn- substituted glasses (shown for Zn50 in Figure 3f), by contrast, showed very low ionic concentrations in Tris buffer, 3.2 Static dissolution experiments with no significant variation over time. Ionic concentrations vs. substitution in the glass are When immersed in HAc/NaAc buffer, all glasses gave apH presented at 3 days in Figure 4. While concentrations in rise from pH 4 to pH 4.4 ± 0.1 (Figure 2a,c). In Tris buffer, Influence of zinc and magnesium substitution on ion release... Ë 97

(a) (b)

(e) (f)

Figure 3: Results from static dissolution experiments on glasses 45S5 (a,b), Mg50 (c,d) and Zn50 (e,f) immersed in HAc/NaAc (a,c,e) and Tris buffer (b,d,f) showing percentages of ions present in solution (relative to the molar amount of ions present in the untreated glass)vs. time. (Lines are visual guides only).

Tris buffer were higher at 7 days (not shown), the trends were present in solution at this time point. In Tris buffer observed were essentially the same as at 3 days. Figure 4a (Figure 4b), concentrations of calcium, magnesium and shows that ionic concentrations in HaAc/NaAc buffer did sodium (about 40 to 65% present in solution) did not not change significantly with Mg for Ca substitution. About change dramatically with Mg substitution in the glass; 85 to 100% of calcium, magnesium and phosphate ions however, phosphate concentrations in solution increased 98 Ë M. Blochberger et al.

(a) (b)

Figure 4: Results from static dissolution experiments on Mg (a,b) and Zn-substituted glasses (c,d) at 3 days immersion in HAc/NaAc (a,c) and Tris buffer (b,d) showing percentage of ions present in solution (relative to the molar amount of ions present in the untreated glass)vs. Mg/Zn for Ca substitution in the glass. (Lines are visual guides only). from 0 to 55% with increasing Mg for Ca substitution. In phosphate (P-O) bending bands at 560 and 600 cm−1 and HAc/NaAc buffer, about 10% of silicon ions were found in the P-O stretch band 1020 cm−1 [32]), but they can also solution compared to 15 to 20% in Tris (Figs. 4a and b). give information on ion release and glass dissolution, e.g. For Zn-substituted glasses, calcium and silicon con- by disappearance of the non-bridging oxygen (NBO; Si— centrations in HAc/NaAc buffer remained constant at O− M+, where M+ is a modifier ion) bands at about 900 and about 95 and 13%, respectively, while zinc and phos- 850 cm−1. phate concentrations decreased with increasing substitu- FTIR spectra after immersion of Mg-substituted tion from 95% to 72 and 50%, respectively (Figure 4c). In glasses in HAc/NaAc buffer (of initial pH of 4) showed Tris buffer, however, concentrations for all Zn-substituted no pronounced differences with glass composition (Figure compositions were very low compared to 45S5 (Figure 4d), 5a), except for a weakly pronounced split band at 560 and with about 2% of phosphate, 3% of silicon, 1% of zinc and 600 cm−1 for 45S5, corresponding to calcium orthophos- 5 to 12% of calcium and 4 to 10% of sodium present in so- phates and indicating beginning apatite precipitation [32]. lution. In control experiments in HAc/NaAc buffer of a starting pH of 5 (Figure 5e) these bands were much more strongly developed for 45S5 but were still absent for Mg50 and Mg100. 3.3 Powder analysis In all cases, however, spectra showed only a low intensity band in the range for non-bridging oxygens (now at about FTIR spectra allow for detection of certain crystalline 960 cm−1), suggesting that a significant amount of ion ex- phases, including phosphates such as apatite (e.g. the Influence of zinc and magnesium substitution on ion release... Ë 99

(a) (b)

(e)

Figure 5: FTIR spectra of Mg (a,b) and Zn-substituted glasses (c,d) at 3 days immersion in HAc/NaAc (a,c) and Tris buffer (b,d). (e) Control samples treated in HAc/NaAc buffer of initial pH of 5 rather than4. change (between modifier ions in the glass and protons In Tris buffer (Figure 5b), by contrast, sharp phosphate from the solution) had occurred. bands at about 560, 600 and 1020 cm−1 indicated that ap- 100 Ë M. Blochberger et al.

(a) (b)

Figure 6: XRD patterns of Mg (a,b) and Zn-substituted glasses (c,d) at 3 days immersion in HAc/NaAc (a,c) and 7 days immersion in Tris buffer (b,d). Labels correspond to apatite (#) and zinc phosphate+) ( . atite had formed for glasses 45S5 and Mg25. The shape of change with protons from the solution had occurred, in the spectrum of Mg50 suggested that the aforementioned contrast to the Mg-substituted glasses (Figure 5b). ion exchange had occurred and apatite formation may pos- XRD patterns confirmed the formation of small sibly have started, but for glasses Mg75 and Mg100, despite amounts of poorly crystalline apatite for 45S5 treated in ion exchange having occurred, apatite formation does not pH 4 HAc/NaAc buffer for three days, while no such forma- seem to have started, although formation of an amorphous tion could be detected for Mg50 or Mg100 (Figure 6a). The phosphate layer cannot be excluded for Mg75. shift in position of the amorphous halo compared to the FTIR spectra of Zn-substituted glasses treated in untreated glasses (Figure 1), however, confirmed changes HaAc/NaAc buffer (pH 4, Figure 5c, and pH 5, Figure 5e) through ion exchange, i.e. formation of an ion-depleted showed formation of bands at 1020, 600 and 560 cm−1 as glass or silica gel. At 7 days, glass Mg50 had clearly formed well as additional bands at 1060 and 630 cm−1, which in- apatite in Tris, but less than 45S5 (Figure 6b). Mg100, while dicate precipitation of phosphates such as possibly zinc not forming any crystalline precipitate, also showed a shift phosphates. Spectra of Zn-substituted glasses treated in in the position of the amorphous halo owing to ion ex- Tris buffer (Figure 5d) showed the typical bridging oxy- change having occurred [33]. gen band at 1000 cm−1 and well-developed NBO bands at XRD patterns of Zn-substituted glasses treated in 900 and 850 cm−1, indicating that no pronounced ion ex- HAc/NaAc buffer (Figure 6c) showed peaks correspond-

ing to zinc phosphate such as hopeite, Zn3(PO4)2·4H2O Influence of zinc and magnesium substitution on ion release... Ë 101

(a) (b)

(e) (f)

Figure 7: Results from dynamic dissolution experiments on glasses 45S5 (a,b), Mg50 (c,d) and Zn50 (e,f) immersed in HAc/NaAc (a,c,e) and Tris buffer (b,d,f) showing normalised concentration of ions (relative to the molar amount of ions present in the untreated glass) presentin solution vs. time.

(JCPDS 01-074-2275). By contrast, XRD patterns of Zn- any peaks at 7 days, suggesting that no crystalline precip- substituted glasses treated in Tris (Figure 6d) did not show itate had formed. In addition, no shift in position of the 102 Ë M. Blochberger et al. amorphous halo was observed compared to the untreated ion release, which enables the GIC to set by formation of glasses (Figure 1), confirming that no pronounced ion ex- ionic bridges between carboxylate groups by metal cations change had occurred. such as Al3+. Magnesium and zinc have similar ionic radii (Mg2+ 0.072 (0.057) nm; Zn2+ 0.074 (0.060) nm; 6 (4)- 3.4 Dynamic dissolution experiments coordinated [7]), and both have field strengths which place them at the boundary between modifiers and intermedi- Results of dynamic dissolution experiments (Figure 7) ates [35]. Intermediates can switch their role, allowing are shown as normalised concentrations, i.e. the molar them to either act as modifiers (creating NBO) or enter the amount of an element detected in solution divided by silicate network (forming Si—O—M bonds, with M being the molar amount of the same element present in the an intermediate cation). untreated glass. Sodium concentrations were generally Their structural role in bioactive glasses has thus above the detection limit (≥ 130 mg L−1 [29]) and are thus been the subject of debate, with 29Si MAS NMR results not displayed in the figures. Results showed much higher suggesting presence of Si—O—Mg bonds, indicating that ion release in HAc/NaAc buffer than in Tris. This was par- magnesium partially entered the silicate network [38], ticularly noticeable for 45S5. Here, phosphate concentra- while molecular dynamics simulation did not find such ef- tions were up to 30 times higher in HAc/NaAc buffer (up fects [39]. Lusvardi et al. observed an increase in silicate −1 −1 to 33.2 L ) than in Tris (up to 1.6 L ), while calcium con- network polymerisation as ZnO replaced Na2O in SiO2- −1 centrations were nearly twice as high (7.4 vs. 4.3 L ). Nor- Na2O-CaO-ZnO glasses, indicating that zinc entered the malised ionic concentrations were much higher for 45S5 silicate network [40]. Line broadening in 29Si MAS NMR than for Mg50 or Zn50. Release patterns of Mg50 and Zn50 spectra of zinc-substituted bioactive glasses made it chal- in HAc/NaAc buffer were quite similar (Figure 7c,e), both lenging to gain information about the structural role of in shape of the curves and normalised concentrations, zinc; however, deconvolution of the spectra suggested the which were between 3.3 and 13.2 L−1, depending on the presence of Si—O—Zn bonds [41]. element. Release patterns of these two glasses in Tris, by Despite the similarities between magnesium and contrast, showed pronounced differences. Zn50 showed zinc, the dissolution behaviour of magnesium and zinc- release of phosphate (up to 4.4 L−1) and calcium (up to containing glasses at physiological pH differed dramati- 0.8 L−1) mostly (and sodium, with the concentration be- cally in the present study, which may indicate that the ing above the detection limit, as mentioned above), while structural role of magnesium and zinc is not that sim- normalised concentrations of silicon and zinc in solution ilar after all. This was first noticeable when observing were negligible. Mg50, by contrast, showed pronounced pH changes during immersion of glass powders in Tris release of all components, with magnesium, calcium and buffer (Figure 2). 45S5 and its Mg-substituted compositions phosphate concentrations being between 1 and 2.8 L−1. Sil- showed the typical pH rise [26], while the correspond- icon concentrations were higher in Tris than in HAc/NaAc ing Zn-substituted compositions did not. As this pH rise buffer for 45S5 and Mg50 (45S5: 1.0 vs.− 0.3L 1; Mg50: is an indicator for the overall ion exchange occurring (as 0.5 vs. 0.2 L−1), while Zn50 gave higher concentrations in cations from the glass are exchanged for protons from HAc/NaAc buffer (0.2 to 0.7− L 1) than in Tris (≤ 0.05 L−1). the solution, thereby resulting in a pH increase [23]), this suggests that the ion exchange was significantly reduced when Zn (rather than Mg) was substituted for Ca. This 4 Discussion agrees with findings on 45S5 with lower Zn contents, which also showed a drastic decrease in ion release with increasing Zn content [42]. Dental GIC set by an acid-base reaction between an acid- In Tris buffer no pronounced change in ionic concen- degradable (fluoro-) aluminosilicate glass and a polymeric trations was observed as magnesium was substituted for acid such as poly(acrylic acid), PAA [34]. In aluminosili- calcium, and magnesium-containing glasses showed com- cate glasses, aluminium, an intermediate element accord- parable ion release profiles as those observed for 45S5 (Fig- ing to Dietzel’s rules [35], has been shown to be present ure 3b,c), even if ion release from Mg-substituted glasses in four-fold coordination mostly, forming Si—O—Al bonds seemed slightly slower. Differences in ionic concentra- and AlO− groups, which are charge-balanced by modifier 4 tions in solution (e.g. higher phosphate concentrations cations [36]. These Si—O—Al bonds are readily hydrolysed for highly Mg-substituted glasses compared to 45S5; Fig- at low pH [37], allowing for rapid glass degradation and ure 4b) were mostly caused by significantly slower apatite Influence of zinc and magnesium substitution on ion release... Ë 103 formation, i.e. concentrations for 45S5 were lower owing to By contrast, the results seem to suggest that zinc en- precipitation of phosphate-containing crystalline species. ters the silicate network to a much larger extent, with When zinc was substituted for calcium, by contrast, most or possibly all zinc forming Si—O—Zn bonds, thereby the ion release into Tris buffer decreased dramatically, significantly increasing the silicate network polymerisa- with ionic concentrations being negligible compared to tion. This would impede water penetration (and subse- 45S5 [23] or Mg-substituted compositions (Figs. 3f and 4d). quent ion release) even more effectively than for the Mg- As no crystalline species were detected when immersing substituted glasses, thus explaining the very low concen- Zn-substituted glasses in Tris buffer, these lower concentrations found in solution. Results from static dissolution trations were most likely caused by a lower (or slower) ion experiments (pH and ionic concentrations) suggest that release from the glass. this results in very low overall ion release, as relative FTIR spectra support these results. In Zn-substituted amounts of ions found in solution were low for all ele- glasses at 3 days immersion in Tris buffer, the NBO band (at ments in the glass. Dynamic dissolution experiments sug- 940 cm−1) was still well-pronounced, while it had largely gested something different, however, as they showed that disappeared for 45S5 and Mg-substituted glasses. This in- phosphate and modifiers (Ca and Na) were still released dicates that in the latter compositions, a large percent- from the glass, while only the release of zinc and silicon age of modifier ions (ionically bound to NBO) had already was negligible. This suggests that zinc may enter the sil- been released from the glass and been replaced by pro- icate network here, thus behaving more like silicon than tons, forming silanol groups. For zinc-substituted glasses, like a modifier. this process did not yet seem to have occurred. The effect of zinc or magnesium fully or partially en- Results from dynamic dissolution studies further con- tering the silicate network can be illustrated by network firmed these trends. Dynamic ion release studies have sev- connectivity (NC; [28]) calculations. Table 1 shows NC val- eral advantages over static ones, and the set-up employed ues calculated assuming that magnesium or zinc act as a here allows for very early time points to be studied. The typical modifiers, forming NBO (equation 1). constant flow of medium through the cell impedes any 4[SiO ] + 6[P O ] − 2([Na O] + [CaO] + [MO]) build-up of ions released from the glass, and thus allows NC = 2 2 5 2 (1) [SiO ] the ion release to be studied without ionic concentrations 2 in solution being affected by precipitation. In addition, a [SiO2], [P2O5], [Na2O] and [CaO] are the concentrations of dynamic system can, to some extend, simulate fluid trans- the components in mol%. [MO] refers to the concentration port in vivo, and is such more comparable to a living sys- of either MgO or ZnO. Here, NC remains constant through- tem. Dynamic release patterns in Tris buffer showed lower out the series, i.e. silicate network polymerisation does not ion release from Mg50 than from 45S5 (Figure 7b,d). Zn50 change. In this case, we would expect glass solubility and showed release of phosphate and calcium only, with re- ion release to be affected by variations in packing density lease of silicon and zinc being negligible. Fagerlund et (through variations of ionic radii and coordination num- al. [29] showed that in dynamic dissolution studies in Tris bers) and strength of ionic Si—O− M+ bonds only. 2− buffer, initial ion release patterns of various glasses corre- If we assume, however, that Mg and Zn form MO4 lated well with their bioactivity, and allowed for glasses to units exclusively [38, 40], thus requiring modifiers for be arranged in four groups (rapid, medium, slow and no charge-balancing purposes, in analogy to the role of Al in − bioactivity). It is interesting to note that the patterns ob- aluminosilicate glasses (forming AlO4 groups, [36]), net- served in the present study for 45S5, Mg50 and Zn50 place work connectivity (NC*) can be expected to increase ac- them in the groups for rapid, medium and slow/no bioac- cording to equation 2: tivity, respectively, which corresponds to their apatite for- 4([SiO ] + [MO]) + 6[P O ] mation in static studies, as discussed below. NC* = 2 2 5 (2) ([SiO2] + [MO]) These results may suggest that in the present glass sys- 2([Na2O] + [CaO] − [MO]) tem magnesium acted as a typical modifier, and that only − ([SiO2] + [MO]) a small percentage of magnesium ions entered the silicate network (forming Si—O—Mg bonds), if any. The slower ion The values for NC* shown in Table 1 show a dramatic release from Mg-substituted glasses may be explained by increase, moving from a typical silicate polymerisation the higher field strength of Mg2+ ions compared to Ca2+, of bioactive glasses (network connectivity around 2) to resulting in a more tightly packed network, in which wa- a highly cross-linked network. (These values are theoret- ter penetrates less easily [43], thus resulting in slower ion ical calculations based on the aforementioned assump- exchange with protons from the dissolution medium. tions only, with the aim of indicating a possible trend. The 104 Ë M. Blochberger et al. highest calculated network connectivity of 4.28 would, of iological pH. At physiological pH, silicon and zinc release course, not be expected to exist in a silicate glass.) were very low, suggesting that zinc had entered the sili- The actual network connectivity depends on the cate network. Release rates at low pH were much higher for amount of Mg or Zn entering the network as well as on the both ions (which was in contrast to 45S5 and Mg50 where actual structural units being formed, as they may not nec- silicon release decreased with decreasing pH), suggesting 2− essarily be MO4 units. The results of these calculations degradation of the silicate network through acid hydroly- (Table 1) do show, however, that the change in network sis of Si—O—Zn bonds at low pH. connectivity may be dramatic, depending on the structural Such a mechanism is known to occur for aluminosili- behaviour of the element. cate glasses and is the key mechanism in glass dissolution It is known that glass dissolution and ion release de- for the setting of dental GIC [37]. The results here also agree crease with increasing network connectivity [26, 28] owing with previous findings in studies on (phosphate-free) zinc to a less open silicate network, smaller numbers of NBO silicate glasses [45], which also showed pronounced differ- and lower concentrations of modifier ions, which are all ences in ion release with pH of the dissolution medium. key elements of the ion release mechanism observed in Previous studies showed that GIC properties, includ- bioactive glasses [43, 44]. The dramatic decrease in ion re- ing hydrolytic stability, differed greatly, depending on lease at physiological pH with zinc-substitution may thus whether they were prepared using a Mg-containing glass be caused by an increase in network connectivity owing to or a Zn-containing one [6]. We hypothesised that this was zinc entering the silicate network, assuming the role of a caused by differences in ion release behaviour of the glass, network former [45]. resulting in different concentrations of ions being avail- We have previously shown that the pH of the dissolu- able for crosslinking of carboxylate groups and, subse- tion medium can play a significant role in bioactive glass quently, differences in GIC mechanical properties. The re- ion release [23, 24, 46]. In the present study, the ion release sults presented here, however, showed that pronounced patterns from static dissolution experiments on Zn- and differences in ion release were noticeable at physiologi- Mg-substituted glasses at pH 4 (Figure 3a,c,d) were more cal pH only, while Zn- and Mg-substituted glasses released similar than those at physiological pH. At pH 4, all glasses comparable amounts of ions at acidic pH, which is the rel- showed fast ion release. Differences with zinc substitution evant pH region for GIC setting. There must, therefore, be (e.g. phosphate and zinc concentrations) were caused by another influencing factor causing GIC from Zn-containing precipitation of zinc phosphates during static dissolution glasses to show better mechanical properties than those of experiments rather than by differences in ion release. Mg-containing glasses, and this topic requires further in- Dynamic dissolution studies (Figure 7a,c,d), again, vestigation. confirmed this trend, as ion release here was also much Some of the differences in ionic concentrations in faster at low pH than at physiological one. 45S5 showed static dissolution experiments observed with changes in particularly fast phosphate release, with normalised cal- glass composition were caused by precipitation of crys- cium concentrations found for 45S5 being lower than those tal phases rather than (or at least not only) by differ- for Mg50 and about as high as for Zn50. Mg50 and Zn50 ences in glass solubility. 45S5 is known to form ap- showed very similar release patterns here, both in shape atite (thought to be a carbonate-substituted hydroxyap- and in normalised concentrations, suggesting that at low atite [47]) in simulated physiological solutions including pH both glasses released ions at comparable rates. Tris buffer, and this was confirmed here by FTIR and We have previously explained the faster ion release at powder-XRD results. With increasing Mg-substitution, the low pH (compared to physiological pH) with larger concen- typical apatite-related features in FTIR (e.g. the split P-O trations of protons being available in solution for ion ex- bending band at 560 and 600 cm−1 and the P-O stretch change with modifier cations [23]. The pronounced differ- band at about 1020 cm−1 [32]; Figure 5b) and XRD (Fig- ences observed here for Zn-substituted bioactive glasses ure 6b) appeared more weakly pronounced and fully dis- (with very low ion release at physiological pH and very appeared for Mg100. This suggests that with increasing fast one at low pH) do suggest, however, that faster ion ex- Mg-substitution in the glass, apatite formation was de- change owing to more protons being available cannot be layed or inhibited. This explains the higher phosphate the only explanation. It is most likely that besides simple concentrations detected in static Tris buffer dissolution ex- ion exchange (protons for modifier ions) a second mech- periments with increasing Mg-substitution (Figure 4b), as anism, involving acid hydrolysis of Si—O—Zn bonds, may less phosphate was consumed for apatite precipitation as play a role at low pH. This is particularly indicated by com- the magnesium content in the glass increased. While there paring continuous ion release from Zn50 at low and phys- is evidence for magnesium incorporation into the apatite Influence of zinc and magnesium substitution on ion release... Ë 105 crystal lattice in small amounts [48], Mg for Ca substitu- when they are needed to perform their antibacterial action in the apatite lattice generally is difficult, as the Mg2+ tion. ion is much smaller than the Ca2+ ion. In addition, Ca In previous studies using aluminium-free GIC glasses is 7- and 9-fold coordinated in the apatite crystal lattice, with very high magnesium [3] or zinc contents [5, 6] were while Mg prefers 6-fold coordinated sites (e.g. in whitlock- used. As the ion release of calcium-, zinc- and magnesium- ite) [18]. Similarly, Zn, which tends to be 4- or 6-fold co- containing glasses is very similar at low pH, results from ordinated, does not easily substitute for calcium either, the present study suggest that such high magnesium or and amounts of zinc incorporated into apatite tend to be zinc contents may not be necessary for GIC setting, as small [49]. Furthermore, both magnesium and zinc have calcium ions were released similarly fast. Owing to their been found to inhibit apatite crystallisation [50]. In the larger field strength compared to calcium ions, however, present study, however, the absence of apatitic features magnesium or zinc ions may still be necessary in GIC to in FTIR spectra (Figure 5d) and XRD patterns (Figure 6d) allow for good mechanical properties through stronger for Zn-substituted glasses can also be explained simply by ionic linkages between carboxylate groups of the poly- very low ion release from these glasses, resulting in ionic meric component. concentrations being too low for precipitation of apatite or other crystal phases, an effect which has also been shown Acknowledgement: The authors acknowledge funding for to occur at lower Zn contents in 45S5 [42]. a bilateral exchange project between Jena and Turku sup- 45S5 has previously been shown to form apatite fast in ported jointly by the German Academic Exchange Service HAc/NaAc buffer solution of a starting pH of 5 [23]. Inthe (DAAD) and the Academy of Finland. DS Brauer acknowl- present study, FTIR spectra of 45S5 immersed in HAc/NaAc edges research funding by the Carl Zeiss Foundation, Ger- buffer of a starting pH of 4 showed very weakly pro- many. nounced apatite bands only (Figure 5a,c). XRD patterns, however, confirmed the formation of very small amounts of apatite (Figure 6a). Mg-substituted glasses, by contrast, References did not seem to form any apatite at this low pH, as indicated by FTIR and XRD results. While 45S5 immersed in [1] Wilson A.D., Prosser H.J., Powis D.M., Mechanism of adhesion a control solution of initial pH 5 did show the typical ap- of poly-electrolyte cements to hydroxyapatite, J Dent Res 1983, atite bands in FTIR (Figure 5e), Mg50 (and Mg100) did not 62, 590–592. seem to form any apatite here either, most likely caused by [2] De Barra E., Grifln S., Henn G., Hill R., Devlin J., Johal K.et the inhibitory effects of magnesium ions. Zn-substituted al., The mechanism of fluoride release from glass (ionomer) polyalkenoate cements, J Dent Res 1995, 74, 833–833. glasses precipitated zinc phosphates, as confirmed by XRD [3] Brauer D.S., Karpukhina N., Kedia G., Bhat A., Law R.V., Radecka results (Figure 6c), which resulted in lower Zn and phos- I. et al., Bactericidal strontium-releasing injectable bone ce- phate concentrations observed in static dissolution experiments based on bioactive glasses, J Roy Soc Interface 2013, 10, ments in HAc/NaAc buffer with increasing Zn-substitution 20120647. in the glass (Figure 4c). [4] Blades M.C., Moore D.P., Revell P.A., Hill R., in vivo skele- Taken together, the results presented here show that tal response and biomechanical assessment of two novel polyalkenoate cements following femoral implantation in the fe- despite their similarities in ionic radius and size, magne- male New Zealand White rabbit, J Mater Sci-Mater M 1998, 9, sium and zinc affect the dissolution behaviour of bioactive 701–706. glasses in very different ways. Particularly the behaviour [5] Boyd D., Clarkin O.M., Wren A.W., Towler M.R., Zinc-based glass of zinc is interesting, with very low solubility at physiolog- polyalkenoate cements with improved setting times and me- ical or neutral pH but very fast ion release at acidic pH. chanical properties, Acta Biomater 2008, 4, 425–431. [6] Brauer D.S., Gentleman E., Farrar D.F., Stevens M.M., Hill R.G., Together with the antibacterial properties of zinc ions, this Benefits and drawbacks of zinc in glass ionomer bone cements, makes zinc-containing bioactive glasses of interest for use Biomed Mater 2011, 6, 045007. as controlled release devices, which release zinc at low [7] Shannon R.D., Revised effective ionic radii and systematic stud- pH only, thereby providing antibacterial action when it is ies of interatomic distances in halides and chalcogenides, Acta needed, such as during bacterial infection or caries, where Cryst 1976, A32, 751–767. pH tends to be low [22]. As high Zn concentrations released [8] Balasubramanian P., Strobel L.A., Kneser U., Boccaccini A.R., Zinc-containing bioactive glasses for bone regeneration, dental from bioactive glasses [51] or from other biomaterials [6] and orthopedic applications, Biomedical Glasses 2015, 1, 51– have been shown to be cytotoxic in vitro, it is desirable to 69. release very low amounts only during normal physiologi- [9] Diba M., Tapia F., Boccaccini A.R., Strobel L.A., Magnesium- cal pH conditions, and only release larger concentrations containing bioactive glasses for biomedical applications, Int J 106 Ë M. Blochberger et al.

Appl Glass Sci 2012, 3, 221–253. [31] Fagerlund S., Ek P., Hupa L., Hupa M., Dissolution kinetics of a [10] Underwood E.J., Trace elements in human and animal nutrition. bioactive glass by continuous measurement, J Am Ceram Soc Academic Press, New York, 1971 1971. 2012, 95, 3130–3137. [11] Hsieh H.S., Navia J.M., Zinc-deficiency and bone-formation in [32] Jones J.R., Sepulveda P., Hench L.L., Dose-dependent behavior guinea-pig alveolar implants, Journal of Nutrition 1980, 110, of bioactive glass dissolution, J Biomed Mater Res 2001, 58, 1581–1588. 720–726. [12] Oner G., Bhaumick B., Bala R.M., Effect of zinc deficiency on [33] Aina V., Bertinetti L., Cerrato G., Cerruti M., Lusvardi G., serum somatomedin levels and skeletal growth in young rats, Malavasi G. et al., On the dissolution/reaction of small-grain Endocrinology 1984, 114, 1860–1863. Bioglass 45S5 and F-modified bioactive glasses in artificial [13] Lansdown A.B.G., Mirastschijski U., Stubbs N., Scanlon E., saliva (AS), Applied Surface Science 2011, 257, 4185–4195. Agren M.S., Zinc in wound healing: Theoretical, experimental, [34] Wilson A.D., A hard decade’s work: Steps in the invention of the and clinical aspects, Wound Repair Regen 2007, 15, 2–16. glass-ionomer cement, J Dent Res 1996, 75, 1723–1727. [14] Prasad A.S., Clinical manifestations of zinc-deficiency, Annu [35] Dietzel A., Structural chemistry of glass, Naturwissenschaften Rev Nutr 1985, 5, 341–365. 1941, 29, 537–547. [15] Yamaguchi M., Oishi H., Suketa Y., Stimulatory effect of zinc on [36] Neuville D.R., Cormier L., Massiot D., Al coordination and speci- bone formation in tissue culture, Biochem Pharmacol 1987, 36, ation in calcium aluminosilicate glasses: Effects of composition 4007–4012. determined by Al-27 MQ-MAS NMR and Raman spectroscopy, [16] Yamaguchi M., Yamaguchi R., Action of zinc on bone metabolism Chem Geol 2006, 229, 173–185. in rats - Increases in alkaline phosphatase activity and DNA con- [37] Grifln S.G., Hill R.G., Influence of glass composition onthe tent, Biochem Pharmacol 1986, 35, 773–777. properties of glass polyalkenoate cements. Part I: Influence of [17] Holloway W.R., Collier F.M., Herbst R.E., Hodge J.M., Nicholson aluminium to silicon ratio, Biomaterials 1999, 20, 1579–1586. G.C., Osteoblast-mediated effects of zinc on isolated rat osteo- [38] Watts S.J., O’Donnell M.D., Law R.V., Hill R.G., Influence of mag- clasts: Inhibition of bone resorption and enhancement of osteo- nesia on the structure and properties of bioactive glasses, J clast number, Bone 1996, 19, 137–142. Non-Cryst Solids 2010, 356, 517–524. [18] Elliott J.C., Structure and chemistry of the apatites and other cal- [39] Pedone A., Malavasi G., Menziani M.C., Computational insight cium orthophosphates, 1st ed. Elsevier, Amsterdam, New York, into the effect of CaO/MgO substitution on the structural prop- London, Tokyo, 1994 1994. erties of phospho-silicate bioactive glasses, J Phys Chem C [19] Aaseth J., Boivin G., Andersen O., Osteoporosis and trace ele- 2009, 113, 15723–15730. ments – An overview, J Trace Elem Med Bio 2012, 26, 149–152. [40] Lusvardi G., Malavasi G., Menabue L., Menziani M.C., Segre U., [20] Fawcett W.J., Haxby E.J., Male D.A., Magnesium: physiology and Carnasciali M.M. et al., A combined experimental and compu- pharmacology, Brit J Anaesth 1999, 83, 302–320. tational approach to (Na2O)1−x·CaO·(ZnO)x·2SiO2 glasses char- [21] Cannillo V., Pierli F., Ronchetti I., Siligardi C., Zaffe D., Chemi- acterization, J Non-Cryst Solids 2004, 345, 710–714. cal durability and microstructural analysis of glasses soaked in [41] Linati L., Lusvardi G., Malavasi G., Menabue L., Menziani M.C., water and in biological fluids, Ceram Int 2009, 35, 2853–2869. Mustarelli P. et al., Qualitative and quantitative structure- [22] Punnia-Moorthy A., Evaluation of pH changes in inflammation property relationship analysis of multicomponent potential bio- of the subcutaneous air pouch lining in the rat, induced by glasses, J Phys Chem B 2005, 109, 4989–4998. carrageenan, dextran and staphylococcus aureus, J Oral Pathol [42] Aina V., Malavasi G., Pla A.F., Munaron L., Morterra C., Zinc- Med 1987, 16, 36-44. containing bioactive glasses: Surface reactivity and behaviour [23] Bingel L., Groh D., Karpukhina N., Brauer D.S., Influence of dis- towards endothelial cells, Acta Biomater 2009, 5, 1211–1222. solution medium pH on ion release and apatite formation of [43] Tilocca A., Cormack A.N., Modeling the water-bioglass interface Bioglassr 45S5, Mater Lett 2015, 143, 279–282. by ab initio molecular dynamics simulations, ACS Appl Mater [24] Shah F.A., Brauer D.S., Desai N., Hill R.G., Hing K.A., Fluoride- Inter 2009, 1, 1324–1333. containing bioactive glasses and Bioglassr 45S5 form apatite [44] Tilocca A., Cormack A.N., The initial stages of bioglass disso- in low pH cell culture medium, Mater Lett 2014, 119, 96–99. lution: a Car-Parrinello molecular-dynamics study of the glass- [25] Jones J.R., Review of bioactive glass: From Hench to hybrids, water interface, P Roy Soc A-Math Phy 2011, 467, 2102–2111. Acta Biomater 2013, 9, 4457–4486. [45] Chen X., Brauer D.S., Karpukhina N., Waite R.D., Barry M., [26] Brauer D.S., Bioactive glasses—structure and properties, McKay I.J. et al., ‘Smart’ acid-degradable zinc-releasing silicate Angew Chem Int Edit 2015, 54, 4160-4181 and Angew Chem Ger glasses, Mater Lett 2014, 126, 278–280. Ed 2015, 127, 4232–4254. [46] Shah F.A., Brauer D.S., Wilson R.M., Hill R.G., Hing K.A., Influ- [27] Miller C., Hatton P.V., Mirvakily F., inventors; The University of ence of cell culture medium composition on in vitro dissolu- Shefleld, assignee. A novel glass-ionomer cement. UK patent tion behavior of a fluoride-containing bioactive glass, J Biomed WO 2014/102538 A1. 3 July 2014. Mater Res A 2014, 102, 647–654. [28] Hill R.G., Brauer D.S., Predicting the bioactivity of glasses using [47] Brauer D.S., Karpukhina N., O’Donnell M.D., Law R.V., Hill R.G., the network connectivity or split network models, J Non-Cryst Fluoride-containing bioactive glasses: Effect of glass design Solids 2011, 357, 3884–3887. and structure on degradation, pH and apatite formation in sim- [29] Fagerlund S., Hupa L., Hupa M., Dissolution patterns of bio- ulated body fluid, Acta Biomater 2010, 6, 3275–3282. compatible glasses in 2-amino-2-hydroxymethyl-propane-1,3- [48] Mayer I., Schlam R., Featherstone J.D.B., Magnesium- diol (Tris) buffer, Acta Biomater 2013, 9, 5400–5410. containing carbonate apatites, J Inorg Biochem 1997, 66, [30] Fagerlund S., Ek P., Hupa M., Hupa L., On determining chemical 1–6. durability of glasses, Glass Technol 2010, 51, 235–240. Influence of zinc and magnesium substitution on ion release... Ë 107

[49] Mayer I., Apfelbaum F., Featherstone J.D.B., Zinc ions in syn- [51] Aina V., Perardi A., Bergandi L., Malavasi G., Menabue L., thetic carbonated hydroxyapatites, Arch Oral Biol 1994, 39, 87– Morterra C. et al., Cytotoxicity of zinc-containing bioactive 90. glasses in contact with human osteoblasts, Chem-Biol Interact [50] Kanzaki N., Onuma K., Treboux G., Tsutsumi S., Ito A., Inhibitory 2007, 167, 207–218. effect of magnesium and zinc on crystallization kinetics ofhy- droxyapatite (0001) face, J Phys Chem B 2000, 104, 4189–4194.