University of Birmingham Thiol modification of silicon-substituted hydroxyapatite nanocrystals facilitates fluorescent labelling and visualisation of cellular internalisation

Calcium phosphates are used widely as orthopaedic implants and in nanocrystalline form to enable the transfer of genetic material into cells. Despite widespread use, little is known about their fate after they have crossed the cell membrane. Here we present a method of surface modi ﬁ cation of silicon-substituted hydroxyapatite (SiHA) through a silane group, which enables the engraftment of a ﬂ uorescent dye to facilitate real-time biological tracking. Surface modi ﬁ cation of the nanocrystal surface was undertaken using (3-mercaptopropyl)trimethoxysilane (MPTS), which presented a thiol for the further attachment of a ﬂ uorophore. Successful modi ﬁ cation of the surface was demonstrated using zeta potential measurements and ﬂ uorescence microscopy and the number of thiol groups at the surface was quanti ﬁ ed using Ellman's reagent. In vitro experiments using the ﬂ uorescently modi ﬁ ed particles enabled the discrimination of the calcium phosphate particulate from other biological debris following internalisation by a population of MC3T3 (pre-osteoblast) cells and the particles were shown to maintain ﬂ uorescence for 24 hours without quenching. The successful modi ﬁ cation of the surface of SiHA with thiol groups o ﬀ ers the tantalising possibility of the intracellular growth factor delivery.


Introduction
Calcium phosphate ceramics such as hydroxyapatite (HA) have been used widely for the restoration of function in diseased and damaged hard tissue. In addition, they have found application in a diverse selection of sectors as food additives, adsorbents in chromatography columns and even as substrates to enable absorption of pollutants from wastewater. 1,2 Within the biomedical sector, calcium phosphate salts have been used principally because of their similarity to the mineral component of bone and also since their dissolution products are non-toxic. Relatively recent work has seen calcium phosphate salts used for the delivery of biological materials into cells in the form of peptides, polymers and DNA sequences. [3][4][5][6][7][8] Calcium phosphate salts have a critical safety advantage over other vectors such as viruses in that they pose no risk of pathogenicity due to mutation. Although it is known that calcium phosphate-DNA complexes cause no apparent cytotoxicity, the fate of the particles upon internalisation is not yet known. The difficulty in tracking the particles can be related to the visual similarity to granulation within the cells.
The large crystal lattice of the apatites means that their structure may incorporate numerous substitutions, which can be used to tailor material chemistry to induce a particular biological reaction. [9][10][11][12] One main focus of research into the development of new calcium phosphate based materials has been the substitution of silicon into the hydroxyapatite lattice (SiHA) to enable additional biomolecule attachment and controlled release in vivo. Silicon substitution is of particular interest since silicon is well established to be an important factor in the production of new bone matrix and functions to assist in the production of collagen by osteoblasts. [12][13][14][15] Calcium phosphate composites have been used widely as implant materials due to the low toxicity of their ionic components and the intimate bond that they are able to form with a wide range of hard and so tissues. There have been proposed to be two sites in crystalline HA that may be exploited for functionalisation; a hydroxyl group and the phosphate group. [16][17][18] Previous studies have shown that the surface hydroxyl (OH) groups of HA may react with organic isocyanate groups and this has been exploited for the attachment of polymers such as poly(ethylene glycol) (PEG), poly(methyl methacrylate) (PMMA), poly(n-butyl methacrylate) (PBMA), poly(2-hydroxyethyl methacrylate) (poly-(HEMA)) to the crystal surfaces. 19,20 In addition, it has been found that substitution by alkylphosphonates causes calcium phosphate monolith structures to become mesoporous with high specic surface area, 21 which is of course crucial for biological applications where cell attachment, proliferation, bioresorbability, and tissue/interface regeneration rely on materials with specic surface areas approaching that of native biomineral. Of these reagents containing hydroxyl groups, phosphoric acid based reactants are favoured because the P-OH groups facilitate the adsorption of proteins, as demonstrated for the model protein, bovine serum albumin. 22 However, the limited availability and reactivity of OH groups can result in low numbers of biomolecules being graed on to the HA surface. 7,23 The Ca 2+ ions in calcium phosphates, including HA, have been exploited for the ionic bonding of various functional groups. For example, Lee et al. 7 have reported the thiol modication of HA using 3-mercaptopropionic acid and Ganesan et al. 8 attached porphyrins to the HA surface as a potential drug carrier. Silicon substitution has drawn much attention as a route to surface modication 24 with organosilanes, particularly those containing amino or thiol terminal groups. For example, previous work has demonstrated that osteoblast adhesion to HA is increased when functionalised with 3-aminopropyltriethoxysilane. 25 Although much attention has been given to the use of (3-mercaptopropyl)trimethoxysilane (MPTS) for the surface modication of silicon oxide, [26][27][28] there is scarcely any research on the surface modication of calcium phosphates with MPTS.
Here we report a new method for the surface modication of SiHA nanoparticles that enables the engrament of uorescent markers onto the surface of the nano particulate. The method involves the functionalisation of SiHA nano particles with the silane MPTS via covalent bonding, which in turn presents a thiol functional group from the particle surface. The inuence of the modication method on the physicochemical properties of the material were determined using X-ray diffraction (XRD), X-ray uorescence (XRF), Zeta potential (ZP) measurement, and Fourier Transform Infra-red Spectroscopy (FTIR). The surface modication of the SiHA particles was evaluated by the attachment of the commercially available thiol reactive probe uorescein-5-maleimide ( Fig. 1) and demonstrated by live cell confocal uorescence imaging of the particles aer internalisation by MC3T3 cells.

Experimental
2.1 HA/SiHA synthesis, functionalisation and dye labelling HA and SiHA synthesis: HA was synthesised by a wet chemical precipitation method. A 250 mL solution of (NH 4 ) 2 HPO 4 (239 mM) and a 350 mL solution of (CaNO 3 ) 2 $4H 2 O (264 mM) were prepared with Millipore water, which was pre-boiled under reux for 2 hours to remove CO 2 and cooled to room temperature. The (CaNO 3 ) 2 $4H 2 O was transferred to a closed glass reaction vessel during magnetic stirring at 200 rpm and nitrogen bubbling. The pH of the solution was adjusted to 10 by the addition of NH 4 OH. The (NH 4 ) 2 HPO 4 solution was added to the reaction vessel drop wise by burette, while maintaining the pH of the reaction at 10 by addition of NH 4 OH, then le to age overnight. The resulting HA precipitate was washed 5 times with Millipore water by centrifugation at 4000 rpm for 10 minutes and resuspended in water for later use at a concentration of approximately 10 mg mL À1 . SiHA was synthesised using the method outlined above, but with a 250 mL aqueous solution of Si(OCOCH 3 ) 4 (41 mM) and (NH 4 ) 2 HPO 4 (203 mM) added drop wise to a 350 mL aqueous solution of (CaNO 3 ) 2 $4H 2 O (290 mM). Attachment of thiol functional groups via MPTS: 230 mg of HA/ SiHA in double distilled water was centrifuged at 4000 rpm for 10 min and the supernatant removed. The particles were then resuspended in 16.4 mL of absolute ethanol ltered with a 0.22 mm pore size syringe lter before adding 100 mL of 3-mercaptopropyltrimethoxysilane (MPTS). The samples were mixed using a ThermoMixer (ThermoMixer Comfort, Eppendorf UK Ltd., UK) at 37 C for 3 hours at a mixing speed of 1000 rpm and then washed 5 times with double distilled water by centrifuging at 4000 rpm for 10 minutes. Dye labelling with Fluorescein-5-Maleimide (F5M): A stock solution of 5.2 mg Fluorescein-5-Maleimide (Life Technologies Ltd, UK) in 1 mL PBS (without Mg 2+ and Ca 2+ ). 1 mL of MPTS functionalised SiHA/HA (SIHA/ HA-MPTS) (approximately 6 mg solid material in water) was centrifuged at 4000 rpm for 10 min. The supernatant was removed and the pellet resuspended in 500 mL of PBS before adding 470 mL of the uorescein-5-maleimide stock solution and mixed on a ThermoMixer for 2 h at 37 C. Finally, the SiHA-MPTS-F5M particles were washed twice with absolute ethanol and ve times with double distilled water (both ltered beforehand using a 0.2 mm pore size lter). As a control, SiHA particles, without thiol modication, were mixed with uorescein-5-maleimide using the above method.

XRD
X-ray diffraction was used to determine the crystalline composition of the samples. HA/SiHA in solution was centrifuged at 4000 rpm for 10 min. The pellet was removed, dried in an oven at 65 C overnight and ground into a ne powder. XRD patterns of the powder samples were obtained with a X-ray diffractometer (D5000, Bruker Corp. USA.) using the Cu K a line. Data were collected from 2q ¼ 5 to 60 with a 0.02 step-size and a step time of 0.5 s/ . These as-precipitated samples are referred to as HA and SiHA herein. To allow a better comparison between HA/SiHA and the ICDD reference patterns, a proportion of the precipitated materials were sintered at 650 C prior to XRD analysis. Sintered samples are referred to as HA-650 and SiHA-650 herein.

TEM
HA and SiHA samples were diluted from 10 mg mL À1 to 400 mg mL À1 and a 5 mL drop placed on mesh-400 copper TEM grid (Agar Scientic). Samples for TEM were imaged using a JEOL JEM 1200EX microscope with a beam energy of 80 kV.

XRF
HA and SiHA solution was dried in an incubator set to 85 C. The elemental composition of each sample was also determined using an X-ray uorescence spectrometer (S8 TIGER, Bruker Corp., U.S.A). Powder forms of the samples were prepared as described above. 500 mg of the sample powder was mixed with 2.5 grams of wax and pressed into a pellet.

FTIR
HA, SiHA, HA-MPTS and SiHA-MPTS powders were prepared as previously described and heated overnight at 85 C to remove moisture prior to analysis. 2 mg of sample powder was mixed with 198 mg of KBr (1%w/w), milled and then pressed into a pellet. FTIR spectra were acquired using a ThermoScientic Nicolette 380 FTIR instrument (ThermoScientic, UK) and represent an average of 64 runs corrected with a background measurement of a 200 mg pure KBr pellet.

Zeta-potential measurements
HA, SiHA, HA-MPTS and SiHA-MPTS particle dispersions were mixed with 10 mM KCl solution at a concentration of 0.05 mg mL À1 . The pH of the sample solutions was adjusted by the addition of 100 mM HCl and 100 mM KOH at 25 C and le overnight to equilibrate. Zeta potential proles as a function of cell z-position were collected using a Beckman Coulter Delsa Nano C. At each of the 9 z-positions 10 accumulations were performed repeated over three runs before repeating the whole process with a fresh sample three times. The pH of each sample was checked and recorded immediately before analysis, corrected (if necessary) and recorded again aer analysis.

Quantitating thiol groups of the silane functionalised HA/SiHA
Thiol presentation on the particle surface was quantied based on an assay using 5,5 0 -dithio-bis-(2-nitrobenzoic acid) (DTNB), also known as Ellman's Reagent, that binds to free -SH groups to form the yellow-coloured product 2-nitro-5-thiobenzoic acid (TNB). The absorbance of the assay solution is proportional to the concentration of free -SH groups in the sample. First, a 'Reaction Buffer' was made consisting of 100 mM Na(PO) 4 and 1 mM EDTA in deionised water set to pH 8.0 by drop wise addition of Na(OH). 50 mL of Ellman's reagent solution (4 mg Ellman's reagent in 1 mL Reaction Buffer) was added to 2.50 mL of Reaction Buffer in a centrifuge tube with a separate tube for each sample plus an additional tube for a control sample. 250 mL of the functionalised HA/SiHA particle dispersion was then added to the tube, mixed using vortex mixer and incubated at room temperature for 15 minutes to form an 'assay solution'. For the control, an addition 250 mL of reaction buffer was added to the tube instead of HA/SiHA particle solution. Aer incubation, 1 mL of the assay solution was transferred to a clean cuvette and the absorbance measured at 412 nm using a spectrophotometer (Cecil CE7500, Buck Scientic, US) zeroed on the control sample. The relationship between molar absorptivity, E (M À1 cm À1 ), and concentration of TNB, c (moles per litre), can be dened as follows: where A ¼ measured absorbance and b ¼ path length of the cuvette in centimetres. The concentration of TNB (and hence concentration of free -SH groups) in the solution in the cuvette was then calculated by solving eqn (1) for c and substituting b ¼ 1 cm and E ¼ 14 150 M À1 cm À1 . The number of moles of -SH groups in the assay solution, m assay , was then calculated using eqn (2): where the factor '2.80 mL' represents the total volume of the assay solution when 250 mL of sample and 50 mL of Ellman's reagent is added to the 2.5 mL of reaction buffer. Given that the -SH groups were contributed solely by the 250 mL fraction of the assay solution from the addition of the particle solution, the nal molar concentration of free -SH groups in the original particle solutions, C sample was determined to be: The cells were seeded at a density of 3 Â 10 4 cells per quadrant in a 4-segmented live cell imaging dish (code: 627870, Greiner-Bio One Ltd., Gloucester, UK) and incubated overnight at 37 C in the supplemented media. HA and SiHA in supplemented media was prepared at concentrations of 0 mg mL À1 and 300 mg mL À1 , mixed using a vortex mixer and sonicated for 5 min before being kept in a water bath at 37 C until required. Aer 24 h, the media was removed from each well and replaced with 1 mL of supplemented media before adding 1 mL HA or SiHAmedia solutions to each quadrant while gently agitating the dish to ensure even distribution of the particles across the surface. The media was removed 24 hours aer exposure to the particles, the cells in each quadrant washed three times with 1 mL PBS and replaced with 1 mL per quadrant of cell imaging media.
Fluorescein stained samples were imaged with a Zeiss LSM 710 ConfoCor 3 confocal system (Carl Zeiss Ltd, U.K.) attached to a Zeiss Axio Observer.Z1 inverted microscope and equipped with a Zeiss EC Plan-Neouar Â63 NA ¼ 1.40 oil objective lens, 488 nm laser diode, a 458 nm/488 nm beam splitter and a 34channel spectral detector, which was used to divert uorescence between 500 nm and 650 nm to a photomultiplier tube detector. Bright eld images were obtained simultaneously with the uorescence images by detecting the transmitted excitation laser light with a second photomultiplier tube detector.

Structure and composition analysis
X-ray diffraction patterns of as-precipitated HA and SiHA along with HA and SiHA sintered at 650 C (HA-650 and SiHA-650 respectively) were obtained and are shown in Fig. 2. The broad diffraction peaks of the as-precipitated HA and SiHA samples suggested that they were of low crystallinity and composed of sub-micron sized crystals, which was expected from a room temperature precipitation method without thermal treatment. 17,29,30 Peaks were identied at around 26 , 32 , 40 and between 47 and 53 and appeared to align well with the most intense peaks from reference data for HA (ICDD PDF card no. 00-009-0432). However, it was evident that the broad HA peak at 32 in Fig. 2 was an envelope of the three most intense peaks in the reference pattern data and thus imposing a limit on quality of the match between the measured data and the HA reference. Two shoulder peaks were identied within the main peak of SiHA, but were still poorly resolved. Furthermore, it is known that the XRD patterns of various apatites could resemble those of as-precipitated HA and therefore an assessment of the chemical/structural changes in sintered samples is required in order to conrm the phase composition and structure of the original as-precipitated samples. Fig. 2 shows the XRD pattern of HA sintered at 650 C for 2 hours. Dominant peaks were identied at 25.83 , 31.81 , 32.20 , 32.88 and were attributed to the (002), (211), (112) and (300) planes of HA respectively.
Sharper and more intense peaks were observed in sintered HA (HA 650) compared to as-prepared HA, indicating an increase in the crystallinity of the sample. In contrast, sintering did not increase the crystallinity of SiHA since no change in the sharpness of the diffraction peaks and little or no change in the peak intensity was observed when compared to as-precipitated SiHA. These observations are consistent with other works that have explored the effect of sintering temperature on Si doped HA where the lack of change (or decrease in cases of Si doping above 2-3%wt) in crystallinity has been attributed to Si incorporation into the HA crystal lattice. 13,31,32 Silicon substitution did not appear to change the angular position or relative intensity of the peaks when compared to HA and HA-650. Secondary phases consisting of a-TCP and b-TCP were not observed in the diffraction patterns of the HA or SiHA samples when compared to ICDD reference patterns 00-009-0348 and 00-009-0169 respectively. TCP phases were not detected in HA even aer sintering at 900 C, but SiHA decomposed into a mixture of HA and a-TCP (data not shown). The TEM images of HA and SiHA shown in Fig. 3 displayed the needle like morphology commonly reported with wet chemical precipitation methods. 7,33,34 Additional phases such as a-TCP, b-TCP and CaCO 3 , which normally have a plate-like morphology, were not observed in either the HA or SiHA samples. The SiHA crystals appeared smaller and their edges were not as clearly dened compared with HA which may indicate that the SiHA sample was of lower crystallinity compared to HA. This observation corresponds well with the broad peaks observed for SiHA in the XRD patterns.
XRF was used to determine the %wt of Ca and P oxides and corresponding mole ratios of the elements are presented in Table 1. The Ca : P ratio of the HA sample was 1.64 and compares well with the theoretical Ca : P ratio of HA of 1.67 to within experimental error using this technique. Ca : P ratio of HA-MPTS (1.95) was higher than expected and may have been due to CO 3 2À ions substituting the PO 4 3À ions in the crystal structure, thus reducing the phosphorous content of the sample and hence increasing the Ca : P ratio. Fig. 4 shows the FTIR spectra of HA, SiHA along with thiolterminated silane functionalised HA and SiHA (HA-MPTS and SiHA-MPTS respectively). A summary of the peaks identied from the spectra, along with the chemical bond and mode the peak was assigned to, is presented in     . Evidence of surface modication was obtained from zeta potential measurements of HA, SiHA, HA-MPTS and SiHA-MPTS at pH 7.4, shown in Table 1. The zeta potential of HA and SiHA was measured to be À1.97 mV and À1.66 mV respectively. Most notably, the zeta potential of SiHA SiHA-MPTS was measured to be À11.66 mV and the change in the zeta potential suggested that this sample exhibited a different surface chemistry to both SiHA and HA. The result compares well with the work of Shyue et al. which also reported a decrease in the seat potential due to the presence of thiols. 35 HA-MPTS did not show the same change in zeta-potential as SiHA-MPTS, which may further suggests that either none or signicantly less thiol groups were attached to the HA surface in comparison to SiHA-MPTS. The presence of thiol groups was conrmed and quantied using an Ellman's reagent assay. For SiHA-MPTS, the thiol content was estimated to be 1.60 Â 10 À5 moles per mg of solid material. DNA binding efficiency experiments showed that 100 mg of SiHA-MPTS particles could completely bind up to 10 mg DNA whereas only 85% and 65% of this DNA mass was bound to 100 mg of unfunctionalised SiHA and HA respectively. At higher DNA masses of 50-100 mg, 20% of the DNA was bound to SiHA-MPTS while only 1-10% of the DNA was bound to SiHA and HA.
3.2 Demonstration of application: live cell imaging of SiHA internalisation by MC3T3 cells using a thiol reactive uorescent probe Using confocal uorescence microscopy, unmodied SiHA treated with uorescein-5-maleimide showed very weak uorescence (Fig. 5b) in the regions of the image where particles could be seen in bright eld (Fig. 5a and c), which demonstrated the effectiveness of the washing procedure in removing non-specically bound dye molecules. The uorescence from SiHA modied with MPTS (SiHA-MPTS) shown in Fig. 5e was brighter than the uorescence from unmodied SiHA. Furthermore, the intensity of the uorescence was even across the vast majority of the structures shown in Fig. 5d and f. The increased uorescence intensity was attributed to the proportion of dye molecules bound to the thiols of the SiHA-MPTS since unbound molecules were expected to be removed during the washing procedure.
The images in Fig. 6 are combined confocal uorescence and brighteld images of MC3T3 cells aer a 24 hour exposure to thiol-terminated silane functionalised SiHA particles conjugated to uorescein-5-maleimide dye (SiHA-MPTS-F5M). Fluorescence could not be clearly seen in the low magnication image (Fig. 6a), which would most likely have been due to the low particle concentration and low numerical aperture of the objective lens used to acquire the image. However, some uorescence was detected in multiple cells at a higher magnication (Fig. 6b). SiHA-MPTS-F5M was not found to be cytotoxic at a concentration of 0.6 mg mL À1 over 24 hours aer performing a live-dead assay (Fig. S1 †). In Fig. 6(c-e), the SiHA-MPTS-F5M particles appeared to form small aggregates, approximately 500-1000 nm in diameter, which aligned along the cell membrane. These structures could not be removed from the cell membrane despite repeated washing in PBS, indicating a strong affinity for the cell membrane. When focusing on an image plane through the middle of the cell (Fig. 6f-j), bright uorescence was detected in new locations within the boundary of the cell as shown in the composite image. Furthermore, uorescence from the few particles on the coverslip could no longer be detected at this new focal plane, demonstrating that any detected uorescence was not an integration of uorescence from particles at the bottom of or underneath the cell. Fluorescence spectroscopy of the sample during imaging revealed an emission prole matching that of the dye with an emission peak between 519 and 529 nm (Fig. S2 †). The internalised structures measured approximately 400-500 nm in diameter and appeared to localise at various points within the cytoplasm of the MC3T3 cell, but could not determine whether this was the result of either (i) individual particles being internalised and then concentrated within the cell, or (ii) the aggregates being internalised as a whole. However, the localisation and strong intensity of the uorescence did indicate that the uorescent labels were still attached to the particle surface and photoactive aer 24 hours in culture media and post internalisation. Internalised particles were observed again while focusing up towards the top of the cell.

Conclusions
In this study, we have prepared nano crystalline SiHA using a wet chemical precipitation technique at room temperature without additional thermal treatment. Presentation of thiol functional groups on the surface of SiHA via silane functionalisation was achieved without perceptible change in the chemical composition of the HA or SiHA itself as determined by XRF. Zeta potential measurements showed a signicant change in the surface chemistry of the SiHA particle surface, but not that of HA subjected to the functionalisation process. The presence of SiO 4 4À in SiHA and SiHA-MPTS was identied by two different Si-O vibrational modes together with the characteristic reduction in the intensity of the OH peak in the FTIR spectra. The presence of MPTS could not be conrmed by FTIR alone as the Si-O-C stretching mode could not be reliably resolved from the phosphate peaks. However, thiols were detected using Ellman's reagent and showed that the using the functionalisation method on SiHA produced approximately 10 À5 moles of thiols per milligram of solid material above the base line measurements for HA and HA-MPTS alone. Thiol group modication was visually demonstrated by the detection of uorescence from uorescein-5-maleimide specically bound to the thiol groups of the modied SiHA surface despite extensive washing of the samples, suggesting that the thiol groups were covalently bonded to the SiHA surface via a Si-O-Si surface network. This was further demonstrated in the confocal uorescence and bright eld images of SiHA-MPTS-F5M internalised by MC3T3, where the bright uorescence allowed the particles to be discriminated from other cellular material with similar morphology. Furthermore, the uorescent labels remained attached to the particle and photoactive at for least 24 hours aer internalisation. This method of functionalisation could allow time course tracking of internalisation of calcium phosphates by various bone cells in order to understand their localisation and fate during bone formation and resorption.