Hydroxyapatite formation on graphene oxide modified with amino acids: arginine versus glutamic acid

Abstract

Hydroxyapatite (HA, Ca₅(PO₄)₃OH) is the main inorganic component of hard tissues, such as bone and dentine. HA nucleation involves a set of negatively charged phosphorylated proteins known as non-collagenous proteins (NCPs). These proteins attract Ca²⁺ and PO₄³⁻ ions and increase the local supersaturation to a level required for HA precipitation. Polar and charged amino acids (AAs) are highly expressed in NCPs, and seem to be responsible for the mineralizing effect of NCPs; however, the individual effect of these AAs on HA mineralization is still unclear. In this work, we investigate the effect of a negatively charged (Glu) and positively charged (Arg) AA bound to carboxylated graphene oxide (CGO) on HA mineralization in simulated body fluids (SBF). Our results show that Arg induces HA precipitation faster and in larger amounts than Glu. We attribute this to the higher stability of the complexes formed between Arg and Ca²⁺ and PO₄³⁻ ions, and also to the fact that Arg exposes both carboxyl and amino groups on the surface. These can electrostatically attract both Ca²⁺ and PO₄³⁻ ions, thus increasing local supersaturation more than Glu, which exposes carboxyl groups only.

1. Introduction

The formation of human bone is one of the most well-known examples of biomineralization. Bone is an organic–inorganic hybrid material made of collagen and carbonated hydroxyapatite (HA, Ca₅(PO₄)₃OH) crystals. Collagen fibres provide a framework known as extracellular matrix (ECM) for HA nucleation and growth. HA nucleation and growth is directed by a set of non-collagenous proteins (NCPs) associated with the ECM [1,2]. To answer fundamental questions about biomineralization, researchers have investigated the effect of smaller biomolecules, such as amino acids (AAs) and peptides. AAs are the building blocks of proteins, and negatively charged AAs such as glutamic acid (Glu) and phosphoserine (PSer) are highly expressed in NCPs. Similar to NCPs [1,3], charged AAs can either inhibit or induce HA mineralization if they are dissolved in solution or bound to a surface.

While several researchers have analysed the inhibitory effect of AAs dissolved in solution on HA mineralization [4–18], the effect of AAs bound to surfaces has been the subject of just a few studies [19–25]; in most papers, researchers analysed the effect of molecules with functionalities simulating those found in protein [22–25]. Rautaray et al. investigated HA precipitation in the presence of aspartic acid (Asp)-capped gold nanoparticles [19] and showed that HA precipitation was promoted in the presence of Asp due to the interaction between the COOH groups from Asp and the Ca²⁺ ions. In addition to their mineralization activity, charged AAs, such as arginine (Arg), glutamic acid (Glu) or Asp bound to the surface of HA were also shown to promote protein absorption, osteoblast proliferation and alkaline phosphatase (ALP) activity [26–28]. Other researchers have focused on the effect of surface functional groups with different electrical charges on HA precipitation. Self-assembled monolayers (SAMs) of silanes on silicon [22], or of alkanethiols on gold [23] were used to investigate the effect of positively versus negatively charged surfaces on HA precipitation. Most of the works show that HA precipitation was faster on negatively charged than on positively charged SAMs, and that among the negatively charged groups, phosphonate groups are the strongest HA nucleators [23,24]. In one case, though, it was found that PO₄, COOH and NH₂ functional groups promoted the nucleation of calcium phosphate to a very similar extent [25]. However, the Ca/P ratio for the calcium phosphate formed in the presence of PO₄ was similar to that of HA (1.67) while lower Ca/P ratios were found for COOH (1.49) and NH₂ (1.60), indicating the presence of some amorphous calcium phosphate (ACP) on these samples. Interestingly, recent literature discussing the role of collagen on HA mineralization showed that the AAs present in the collagen sequence close to the hole zones (i.e. where HA nucleates intrafibrillarly) include both positive AAs such as Arg and negative ones such as Glu [29]. The presence of both negatively and positively charged AAs in these areas seems to be crucial for the retention of Ca²⁺ and ions which are necessary for HA nucleation and growth [29,30].

In this work, we try to shed some light on the effect of positive versus negative charges on HA nucleation using a positively charged (Arg) and a negatively charged (Glu) AA bound to a substrate made of graphene oxide (GO). Arg and Glu are present in NCPs involved in bone mineralization [1] and have been found to play a key role in collagen mineralization [29–31]. Several other substrates, such as gold nanoparticles [19], silicon [22] and titanium foil [24] have been used before to study HA precipitation. We choose GO as it contains many different surface functional groups, making it easy to modify [32]. Its high surface area resulting from its unique two-dimensional structure is ideal to speed up HA precipitation [32]. In addition, the graphene families of nanomaterials (GFNs) are able to promote osteogenic differentiation in mesenchymal stem cells [33] and are potential candidates for bone tissue engineering applications [32,34,35]. A few papers have started exploring the formation of HA on these materials [36–42]. Also, a few researchers have explored the modification of GFNs with biomolecules and investigated the effect of such modifications on in vitro HA mineralization [20,21,36,38,40,42]. All these studies showed that HA precipitation was improved in the presence of biomolecules. Our work clearly shows that GO modified with a positively charged AA, Arg, is the best substrate to nucleate HA in vitro. As discussed in the paper, this result furthers our fundamental understanding of the effect of positively and negatively charged AAs bound to surfaces on HA mineralization, while providing an example of a highly promising GO-based substrate modified with a simple biomolecule for bone tissue engineering applications.

2. Methods

Details on the materials and purity used in this study can be found in the electronic supplementary material, SI. 1.

2.1. Surface modification of graphene oxide flakes

The AAs can be grafted onto the GO surface by coupling the NH₂ group from the AAs with the carboxyl groups that are naturally present on the surface of GO flakes. However, to improve this process, more carboxyl groups were introduced on the GO surface using a previously described technique [43] with a few modifications (electronic supplementary material, SI. 2). The AAs were then grafted onto carboxylated GO (CGO) using the well-known EDC coupling technique (electronic supplementary material, SI. 3).

2.2. Precipitation experiment

Simulated body fluid (SBF) was prepared according to Kokubo et al. [44]. 1.9 mg of GO, CGO, CGO-Glu or CGO-Arg were dispersed in 40 ml of SBF in centrifugation tubes. The tubes were placed in an incubator at 37°C for 15 days. The samples were then removed from the incubator, washed three times with DI water (three cycles of centrifugation at 4000 r.p.m. for 15 min and re-suspension in DI water) and dried using a VirTis freeze-drier.

2.3. Characterization

GO-modified samples were characterized by X-ray photoelectron spectroscopy (XPS), Fourier transform infrared (FT-IR) spectroscopy, thermogravimetric analysis (TGA), zeta potential, SEM and TEM. The concentration of Ca and P during the precipitation experiments was measured by inductively coupled plasma atomic emission spectroscopy (ICP-AES). Details of all techniques can be found in the electronic supplementary material, SI. 4.

3. Results

To study the effect of positively and negatively charged AAs bound to GO, we first transformed GO into carboxylated GO (CGO), thus increasing the number of carboxylated groups on its surface. Then, we bound a positive (Arg) and a negative (Glu) AA on it via an amidation process achieved with the aid of EDC and NHS. We then immersed the AA-modified CGO in SBF, to study how AAs influence heterogeneous nucleation of HA on CGO.

3.1. Characterization of graphene oxide surface modification

3.1.1. X-ray photoelectron spectroscopy analysis

Figure 1a shows O, C, S and N atomic % on GO, CGO and CGO-AA samples obtained from XPS survey spectra. C and O mainly originate from GO on all samples; however, all CGO samples show a significant decrease in O and increase in C compared with GO. This suggests that the amount of O in the O-containing groups that are removed from GO (hydroxyl, epoxy and carbonyl groups) is larger than the amount of O that is introduced by carboxylation. No significant differences in C and O are observed between CGO samples before and after AA coupling. However, while CGO does not show any N and S, as expected, CGO-AA samples show the presence of some N and S. S can originate from the physisorbed coupling components, NHS and MES, and N can originate from both the physisorbed coupling components, EDC, NHS and MES, or from the AAs. The unexpected S and N observed on the GO sample may be attributed to the precursors used in Hummer's method for GO production [45]; these are removed after the extensive washing performed during the carboxylation stage, since no S and N are observed on the CGO sample. A control sample prepared by immersion of CGO in EDC/NHS solution without further AA coupling showed a similar amount of S as the CGO-AA samples, but much less N (see the electronic supplementary material, table S2), thus confirming that while the presence of S on CGO-AA samples is due to physisorbed components from the coupling solution, the N is mainly related to the presence of AAs on the surface of the CGO-AA samples. While Arg has three times more N atoms than Glu in its molecular structure, CGO-Arg shows only two times more N than CGO-Glu. This may indicate that less Arg than Glu was presented on the CGO-AA samples. However, XPS data are affected by the orientation of the AAs in the coating and by the coating thickness, and thus this interpretation needs to be corroborated by results obtained with other techniques. Results obtained by FT-IR, which also confirm the presence of AAs on the surface, and zeta potential, are discussed in the electronic supplementary material, SI. 6 and SI. 7, respectively.

Figure 1.

XPS quantitative analysis: (a) O, C, S and N atomic % on GO, CGO, CGO-Glu and CGO-Arg obtained from survey spectra; (b,c) percentages of the different components used to fit high-resolution XPS C 1s (b) and O 1s (c) spectra shown in figure 2. (Online version in colour.)

Further confirmation of the presence of AAs can be obtained by analysing the high resolution C1s and O1s spectra (figures 1b,c and 2). The C 1s spectra can be deconvoluted into four components, centred at 284.9, 286.9 ± 0.2, 288 ± 0.05 and 289 ± 0.05 eV (figure 2a). On all samples, the 284.9 eV component originates from C–C/C=C bonds from the benzene rings in GO [46,47]. The components at 287, 288 and 289 eV originate from the oxygenated functional groups, namely epoxy/hydroxyl (C–O), carbonyl (C=O) and carboxyl (O–C=O) groups, respectively [46,47]. Figure 1b shows a drastic decrease in C–O and C=O bonds and a significant increase in O–C=O bonds on CGO compared to GO. This confirms the replacement of epoxy, hydroxyl and carbonyl groups on GO by carboxyl groups during the carboxylation stage.

Figure 2.

XPS C 1s (a) and O 1s (b) high-resolution spectra for GO, CGO, CGO-Glu and CGO-Arg samples. (Online version in colour.)

Both CGO-Arg and CGO-Glu show a similarly drastic decrease in the 287 and 288 eV components, and an increase in the 285 and 289 eV component compared with GO (figure 1b). However, on these samples, the 287 eV component is significantly more intense than on CGO; this is related to the fact that on these samples, this component can also correspond to C–N [48] and C=N [49] bonds, which can originate from the AAs present on their surface. The higher intensity of the 287 eV component on CGO-Arg compared to CGO-Glu can be explained by the fact that Arg has more C–N/C=N bonds than Glu. The peak at approximately 289 eV (figure 2a) on these samples can also be assigned to amide groups (N–C=O) originating from the coupling between AAs and carboxyl groups during EDC coupling [50]. The CGO-AA samples do not show significant changes in this component intensity compared with CGO (figure 1b); this is to be expected, since the amide bond formed by EDC coupling substitutes for carboxylate groups already present on CGO.

The changes in C=O and C–O bonds are also evident from the high-resolution O 1s spectra shown in figure 2b (components centred at 531.4 ± 0.2 and 533.1 ± 0.1 eV, respectively) [51]. Consistent with the C 1s results, the C–O component of O 1s is significantly lower for the CGO and CGO-AA samples in comparison with the GO sample, while their C=O component is significantly higher (figure 1c). This further confirms the replacement of oxygen-containing functional groups on GO by carboxyl groups during the carboxylation stage. The small component at 535 ± 0.1 eV observed on all samples (figure 2b) can be attributed to water molecules [52]. Its higher intensity (figure 1c) on CGO and CGO-AA samples compared with GO confirms the presence of more charged functional groups on these samples, making them more hydrophilic. CGO showed somewhat higher water content than CGO-Arg, but no significant differences between CGO and CGO-Glu were found.

3.1.2. Thermogravimetric analysis

Two main decomposition temperatures are observed for GO and CGO, and three for CGO-AA samples (figure 3 and table 1). The weight loss (WL) observed at T1 (approx. 55°C) is attributed to the evaporation of water molecules [53] trapped inside GO. There are no significant differences in WL% at this temperature among all samples except for CGO-Arg, which shows somewhat less water content than the others. While no significant differences in water content between CGO and GO are found by TGA, a significant difference was found by XPS (see figure 1c and its discussion); this discrepancy may relate to the fact that the differences observed by XPS are related to water adsorption occurring at the surface of the samples, and this amount of water is not detectable by TGA. The WL at T2 observed on all samples can be ascribed to the pyrolysis of oxygen-containing functional groups, such as epoxy, carbonyl, hydroxyl and carboxyl groups, present on the GO surface or in the AA structure [53,54]. T2 is lower for CGO (163 ± 3°C), CGO-Glu (167 ± 2°C) and CGO-Arg samples (165 ± 1°C) than for GO (206 ± 9°C). This can be attributed to the higher concentration of carboxyl groups on CGO and CGO-derived samples, because carboxyls have lower decomposition temperatures than carbonyl, epoxy or hydroxyl groups [55,56]. The WL at this temperature is significantly higher for GO (38 ± 1%) than on CGO (20 ± 1), CGO-Glu (19 ± 1) and CGO-Arg (21 ± 0). This once again confirms that there are more hydroxyls/epoxy/carbonyl groups on GO than carboxylic groups on CGO and CGO-derived samples. The WL observed at T3 only on CGO-AA samples can be attributed to the decomposition of CN bonds, either formed during EDC coupling [57], or present in the AAs, as shown by the presence of similar temperatures on the TGA graphs of Arg and Glu powders (electronic supplementary material, figure S4). As the WL observed at T3 is about the same on both samples (approx. 11%), we can conclude that there are probably fewer Arg molecules on CGO-Arg than Glu molecules on CGO-Glu because Arg (i) has more CN than Glu; (ii) can form more amide bonds than Glu and (iii) has a higher MW than Glu (174.2 and 147.1 g mol⁻¹, respectively) [57]. This is in agreement with XPS survey data discussed before (figure 1a).

Figure 3.

TGA curves collected under N₂ atmosphere on GO, CGO, CGO-Glu and CGO-Arg samples. The empty circles correspond to the temperatures at which peaks were observed on the δmass/δT derivative curves. (Online version in colour.)

Table 1.

WL temperature (°C) and corresponding weight loss WL% observed on the TGA curves (figure 3) for GO, CGO, CGO-Glu and CGO-Arg samples. Asterisk and plus signs indicate values that are statically significantly different from the correspondent values measured on GO and CGO samples, respectively, with p < 0.05.

samples	GO	CGO	CGO-Glu	CGO-Arg
WL (T1)	5 ± 1 (57 ± 1)	6 ± 1 (57 ± 2)	4 ± 1 (53 ± 3)	3 ± 0*,+ (53 ± 6)
WL (T2)	38 ± 1 (206 ± 9)	20 ± 1* (163 ± 3)	19 ± 1* (167 ± 2)	21 ± 0* (165 ± 1)
WL (T3)	0 ± 0 (n.a.)	0 ± 0 (n.a.)	11 ± 1*,+ (340 ± 15)	11 ± 1*,+ (302 ± 1)

3.2. Mineralization assay

We immersed all samples in SBF to evaluate the mineralization potential of the modifications introduced on GO. The initial degree of supersaturation in SBF solution has been previously calculated by several authors [58,59]; for an initial total [Ca] of 2.5 mM, total [P] of 1 mM and pH of 7.4, at 37°C, the initial supersaturation with respect to HA is 1.33 E12, corresponding to a thermodynamic driving force for HA precipitation (ΔG) of approximately −8 kJ mol⁻¹ [59]. During our experiment, the pH is maintained constant due to the presence of tris buffer (this was confirmed by pH measures performed throughout the experiment). Changes in total [Ca] and [P] measured by ICP are shown in figure 4. The initial negative ΔG value implies that HA precipitation is thermodynamically favoured in the starting SBF. However, the sample labelled as ‘SBF’ in figure 4 (grey bars), shows that no significant changes in total [Ca] and [P] were measured in the absence of GO substrates. This implies that the initial supersaturation degree was low enough to prevent homogeneous nucleation over the period of time considered in our experiment. The sample labelled as ‘HA’ (blue bars) is, instead, a positive control sample, i.e. HA particles that act as nucleation seeds to promote HA precipitation. The rapid decrease in both [Ca] and [P] concentrations measured after 3 days of immersion on this sample clearly shows that the presence of HA seeds strongly favours heterogeneous nucleation of HA. On the GO and CGO samples, a decrease in total [Ca] was observed after 3 days of immersion in SBF (although lower than on the positive control, HA sample); and a decrease in total [P] was detected only after 15 days on the CGO-Arg sample. These results show that heterogeneous nucleation on GO and CGO samples happens starting from day 3. Significant differences related to the presence of AAs are visible after 15 days of incubation, when all CGO-AA samples show significantly lower [Ca] and [P] than both GO and CGO. CGO-Arg shows the strongest decrease.

Figure 4.

[Ca] (a) and [P] (b) measured by ICP on negative control sample (SBF, grey bars) and in SBF containing GO (solid black), CGO (striped black), CGO-Glu (solid red), CGO-Arg (striped red) and positive control (HA powder, blue bars) after 0, 3, 8 and 15 days of immersion in SBF. Asterisk and plus signs indicate values that are statically significantly different from the correspondent values measured on GO and CGO samples, respectively, with p < 0.05. (Online version in colour.)

The amount of precipitates formed on the samples after 15 days of immersion in SBF was measured by TGA (figure 5a and table 2). These TGAs are performed in air to fully burn the organic components and leave only the inorganic component as residues at the end of experiment. The WL at T1, T2 and T3 correspond to the removal of water molecules, C/O bond destruction and carbon combustion to CO and CO₂, respectively [54]. The amounts of residues in table 2 indicate that CGO-Arg (44 ± 4%) contained the highest amount of precipitate, followed by CGO-Glu (21 ± 2%). The GO (16 ± 2%) and CGO (14 ± 1%) samples showed comparable amounts of inorganic component as residual mass. The residual mass (9 ± 1%) found on GO not soaked in SBF (GO-0d) is attributed to contamination.

Figure 5.

Analysis of precipitates formed on GO, CGO, CGO-Glu and CGO-Arg after 15 day immersion in SBF: (a) TGA curves collected in air. GO 0d is a GO sample not immersed in SBF. The empty circles correspond to the temperatures at which peaks were observed on the δmass/δT derivative curves. (b) P, Ca and Mg atomic % measured from XPS survey spectra. Complete survey results are shown in the electronic supplementary material, table S5. Asterisk and plus signs indicate values that are statically significantly different from the correspondent values measured on GO and CGO samples, respectively, with p < 0.05. (Online version in colour.)

Table 2.

WL percentages and temperatures (°C) for GO not immersed in SBF (GO 0d), and for GO, CGO, CGO-Glu and CGO-Arg samples immersed in SBF for 15 days. Asterisk and plus signs indicate values that are statically significantly different from the correspondent values found on GO 15d and CGO 15d, respectively, with p < 0.05.

samples	GO 0d	GO 15d	CGO 15d	CGO-Glu 15d	CGO-Arg 15d
WL% (T1)	3.0 ± 0.3 (45)	2.0 ± 0.2 (53)	2.0 ± 0.2 (61)	2.0 ± 0.2 (60)	2.0 ± 0.2 (60)
WL% (T2)	40 ± 4 (202)	27 ± 3 (190)	18 ± 2* (184)	19 ± 2* (182)	13 ± 1*,+ (170)
WL% (T3)	48 ± 5 (516)	55.0 ± 5.5 (455)	66 ± 7 (491)	57 ± 6 (525)	41 ± 4*,+ (482)
residue	9 ± 1	16 ± 2	14 ± 1	21 ± 2*,+	44 ± 4*,+

We analysed the nature of the precipitates by FT-IR (electronic supplementary material, figure S5 and table S4) and XPS (figure 5b and electronic supplementary material, table S5) spectroscopy, and X-ray diffraction (XRD; electronic supplementary material, figure S6). Both FT-IR and XRD can detect the formation of HA only on CGO-Arg (electronic supplementary material, figures S5 and S6). A more detailed discussion of these data is found in the electronic supplementary material, SI. Using XPS, a small amount of Ca can be detected on all samples, but CGO-Arg is the only sample where both Ca and P are found (figure 5b); the Ca/P ratio is 1.9 ± 0.2, i.e. not significantly different from what is expected for HA (1.67). The higher value may be due to carbonate-to-phosphate substitutions in HA [60,61]. No P is found on samples other than CGO-Arg; this is due to the fact that since the Ca levels are very low on these samples, the P level, expected to have a similar ratio to Ca than what is found on CGO-Arg, goes below the XPS detection limit. This confirms what was previously observed at ICP (figure 4), i.e. a larger amount of calcium phosphate precipitates on CGO-Arg.

The amount of Ca detected on CGO and CGO-Glu is comparable and significantly higher than on GO, which again is in line with the observed depletion of Ca in solution measured by ICP. Some Mg is observed on GO and CGO. Other researchers noticed the presence of Mg in HA precipitating from SBF on titanium [62]. Some N is found on GO and CGO after immersion in SBF (between approx. 1.4 and 1.7%; electronic supplementary material, table S5). Since almost no N was present on these samples before SBF immersion (figure 1a), this may indicate the incorporation of tris, used as a buffer in SBF, in the precipitates formed on these samples.

Back-scattered electron (BSE) images obtained by SEM are shown in figure 6. Before immersion in SBF, randomly oriented GO flakes are observed (figure 6a), with no impurities or particles on their surface (figure 6b,c). Similar images were collected on functionalized GO samples before immersion in SBF (electronic supplementary material, figure S7). After immersion in SBF for 15 days, some particles are observed on all samples (figure 6d–q). The largest amount of precipitation on CGO-Arg is clearly confirmed by the low-magnification SEM images shown in figure 6a,d,g,j,m. On GO, CGO and CGO-Glu irregularly shaped aggregates are found (average size of 1.5 ± 1, 1.0 ± 0.5 and 2.0 ± 1 µm, respectively). The aggregates consist of spherical particles with an average diameter of approximately 100 ± 20 nm on GO and CGO (figure 6f,i), and of a more wide variety of particles on CGO-Glu (figure 6l). On CGO-Arg, instead, two different sets of particles are observed (figure 6n–q): many micrometre-sized spherulites of 3.5 ± 0.5 µm diameter made of nano-sized platelets (20 ± 3 nm) (figure 6n,o), and a few spherical particles of 60 ± 5 nm diameter (figure 6p,q). We discuss this in more detail below when we analyse the TEM results for this sample.

Figure 6.

BSE-SEM images of GO sample not immersed in SBF (a–c) and GO (d–f), CGO (g–i), CGO-Glu (j–l) and CGO-Arg (m–q) sample immersed in SBF for 15 days. (Online version in colour.)

EDS analysis showed the presence of C, N, O, P, Ca and Mg in all samples. This discrepancy with XPS, which showed P only on CGO-Arg, is likely related to the lower amount of sample analysed by XPS, and thus its overall lower sensitivity. A small amount of S was also present on the smaller particles observed on CGO-AA samples, possibly due to the S originally present on these samples (figure 1a). The Ca and Mg content in each sample is normalized to P and shown in table 3. The Ca/P ratio measured on the micrometre-sized particles on CGO-Arg (figure 6o) is 1.8 ± 0.3, which is very close to that of HA (1.67). This confirms that these particles are HA, as hypothesized before based on both XPS and IR (figure 5b and electronic supplementary material, figure S5). Both we and other researchers previously reported the formation of micrometre-sized HA spherulites in the presence of AAs with morphology very similar to that shown in figure 6 [7,44,63]. The Ca/P ratios measured on the small spherical particles observed on CGO-Arg (figure 6q) have large variability (1.7 ± 1.2), probably due to the fact that it was difficult to isolate them during the analysis. More definite results on these particles are provided by TEM, discussed below. Much higher Ca/P ratios are found on GO and CGO (10.1 ± 3.3 and 9.8 ± 1.2, respectively). These values indicate that on these samples there is no significant amount of HA, and the high amount of Ca may be related to physisorbed Ca ions or Ca(OH)₂ derivatives [64]. A Ca/P ratio closer to HA is found on CGO-Glu (4.0 ± 1.8). Similarly to what found by XPS, EDS shows the presence of significant amounts of Mg only on GO and CGO (error bars on the Mg/P found on the other samples make them not statistically significant). Finding Mg only on the samples that show the highest Ca/P ratios and lowest amounts of HA may be related to the role of Mg as an inhibitor of HA crystallization [65].

Table 3.

Mg/P and Ca/P ratio calculated from the SEM-EDS spectra on GO, CGO, CGO-Glu and CGO-Arg after 15 day immersion in SBF.

samples	Ca/P	Mg/P
GO	10.1 ± 3.1	0.64 ± 0.1
CGO	9.8 ± 1.2	9.13 ± 3.18
CGO-Glu	4.0 ± 1.8	0.64 ± 0.45
CGO-Arg (spherulite)	1.8 ± 0.3	0.04 ± 0.02
CGO-Arg (sphere)	1.7 ± 1.2	0.14 ± 0.17

CGO-Arg immersed in SBF for 15 days was further analysed by TEM (figure 7). Again, we observed both spherulitic particles with diameters of approximately 3 ± 1 µm (figure 7a) and much smaller nanoparticles with diameters of approximately 50 ± 5 nm (figure 7d). Both types of particles contained mainly Ca and P (see EDS spectra shown in the insets). Traces of Mg are detected on the larger particles only, although its absence on the nanoparticle spectrum may be due its overall lower intensity. The selected area electron diffraction (SAED) pattern shown in figure 7b shows that the spherulitic particles are poorly crystalline and mostly composed of HA, consistent with IR and XRD results and previous SEM discussion based on Ca/P ratios. The main diffraction rings observed on the TEM SAED (figure 7b) are in agreement with the most intense peaks observed by XRD on this sample (electronic supplementary material, figure S6d), overall similarly to what was observed by Zhou et al. [66]. These SAED on the nanoparticles, instead, show mostly amorphous material; only weak diffraction rings corresponding to the (004) and (211) reflections are visible. Most probably, then, these smaller particles are mostly ACP precursors not yet transformed into HA. The presence of these particles on CGO-Arg may be interpreted in two different ways. Nucleation may be continuously occurring on this sample. The ACP nanoparticles are formed first, and with time they conglomerate and reorganize into the micrometre-size spherulitic crystalline HA particles observed in figures 6o and 7a. Throughout SBF immersion, ACP nanoparticles keep nucleating, and thus after 15 days some of them are still visible. Alternatively, it is possible that some precipitates remain in the form of ACP nanoparticles and never transform into HA. A study performed at different time points will be the subject of a forthcoming publication, to help elucidate which of these mechanisms is correct, and to attempt to explain the similar shape but different composition observed for the nanoparticles found on CGO-Arg and all other samples.

Figure 7.

TEM images (a,c), and corresponding SAED patterns (b,d), and EDS spectra (insets) for CGO-Arg sample immersed in SBF for 15 days. (Online version in colour.)

4. Discussion

Overall the results presented before show that both a positive (Arg) and a negative (Glu) AA increase HA precipitation rate compared to CGO and GO; however, despite the fact that we probably had fewer Arg molecules bound on CGO-Arg than Glu on CGO-Glu, Arg increased the HA precipitation rate much more than Glu, and much more HA was found on CGO-Arg than CGO-Glu after 15 days of immersion in SBF. This result contradicts many previous studies, which showed that the interaction between Ca²⁺ and negatively charged residues of biomolecules, such as carboxylate and phosphorylated groups play a key role in HA precipitation [20,22–24,38,40]. In this section, we first attempt to provide an explanation for this result, and then we compare the effectiveness of CGO-Arg on HA precipitation with previously reported GFN substrates.

4.1. Effect of Arg versus Glu bound to carboxylated graphene oxide on hydroxyapatite precipitation

Precipitation happens when the concentration of precursor ions in solution increases above a critical level. This can be achieved locally on a surface, if there are some functionalities that can strongly interact with the ions. Our results show that both Arg and Glu are able to locally increase Ca²⁺ and surface concentration, and induce early nucleation of HA compared to what is observed on both GO and CGO. As we had somewhat fewer Arg molecules on CGO-Arg than Glu on CGO-Glu, Arg must be able to interact with the ions more strongly than Glu. Figure 8 shows a few possible ways in which Arg and Glu may bind to CGO, and the subsequent interactions between the remaining functional groups exposed by the AAs and Ca²⁺ and ions in solution. With the help of this schematic, we discuss here two reasons why Arg is a more effective HA nucleator than Glu.

Figure 8.

Schematic showing possible interactions of Glu and Arg with CGO surface and Ca²⁺ (circle) and (triangle) ions. The ions are shown in red or blue if they are interacting with the AAs by forming complexes or electrostatically, respectively. (Online version in colour.)

As shown in figure 8b, after binding to CGO, Glu exposes two carboxylate groups, one alpha and one gamma, which can interact with the ions present in solution. While Ca²⁺ ions can interact with these groups (stability constants of log K = 1.4 [67] and 1.7 [67] are reported for Glu/Ca²⁺ complexes at pH 7.4), ions do not have strong interactions with carboxylate groups [63]. Therefore, Glu/ complexes are not shown in figure 8. Arg can bind to CGO through both its α-amino (figure 8c) and guanidyl group (figure 8d), which implies that it exposes its α-carboxyl group and either its guanidyl group (figure 8c) or α-amino group to the solution (figure 8d). Arg can also form more than one amide bond to CGO (figure 8e shows an example of this). At pH 7.4, Arg can interact with Ca²⁺ ions through its α-amino group, forming complexes whose stability constant (log K) is 2.21 [68]. Arg can interact with ions through its guanidyl side chain, forming a complex with a reported stability constant of log K = 1.9 [69]. The stability constants of complexes formed between Arg and both Ca²⁺ and are higher than between Glu and Ca²⁺. Thus, the faster precipitation of HA on CGO-Arg than on CGO-Glu may be partially explained by the stronger Arg/Ca²⁺ and Arg/ interactions leading to more stable complexes formed on CGO-Arg, which result in a higher concentration of Ca²⁺ and ions on the CGO-Arg surface.

In addition to complex formation, electrostatic interactions between surface functional groups and Ca²⁺ and ions could play a role in HA precipitation. As shown by zeta potential measures (electronic supplementary material, figure S3), all surfaces are overall negatively charged. However, locally, CGO-Arg exposes both positive and negative charges (figure 8c,d), whereas CGO-Glu exposes only negative charges (figure 8b). This implies that locally, CGO-Arg is more likely to attract both Ca²⁺ and ions than CGO-Glu, which is likely to attract only Ca²⁺ ions. This factor may contribute to a higher increase in local supersaturation with respect to HA for CGO-Arg than the CGO-Glu sample, and thus a faster precipitation of HA on CGO-Arg. This hypothesis is confirmed by the EDS (table 3) and XPS (figure 5b) results, which showed higher Ca/P ratios on CGO-Glu than on CGO-Arg, thus confirming the greater tendency for CGO-Glu to attract positively charged ions than CGO-Arg. Thus, the presence of both positively and negatively charged groups close to each other on CGO-Arg may be another justification for the stronger HA-nucleating effect of this sample compared with CGO-Glu.

4.2. Arg-carboxylated graphene oxide as a strong nucleator of hydroxyapatite

In this section, we compare our precipitation results with the few previous studies on in vitro precipitation of HA on GFNs [20,38,40]. In general, GFNs are good HA nucleators due to their high surface area, and the presence of biomolecules always increase the amount and rate of HA precipitation [20,21,38,40,42].

Liu et al. [20,38] showed the formation of HA on reduced GO (rGO) modified with a layer of in situ polymerized dopamine (polydopamine, PDA) after one week and gelatin-modified GO after two weeks of immersion in 1.5 × SBF. On rGO-PDA, a layer of HA nanoparticles was formed, in an amount corresponding to approximately 50% in weight of the composite, as evaluated by TGA [20]. This result is comparable with what we observed after two weeks of immersion in SBF (figure 5a); however, we used 1.5 less concentrated SBF than Liu et al. Thus, we conclude that Arg and PDA are similarly effective in promoting HA nucleation, despite their large differences in size. On rGO-gelatin, Liu et al. show spherulitic particles similar to what we found on CGO-Arg [38]. SEM images show a more extensive coverage on this sample after two weeks of immersion in 1.5 × SBF than what we observed, but no quantitative TGA results are reported.

Fan et al. showed that the surface of casein phosphopeptide (CPP)-modified CGO was partially covered with HA nanoparticles after only 1 day incubation in 1.5 × SBF and was completely covered with HA after 3 days, corresponding to an HA content of approximately 80 wt% by TGA [40]. This amount of precipitation achieved in such a short time is significantly larger than what we observed after 15 days (figure 5a), and thus cannot be solely related to the higher concentration of SBF used in [40]. Evidently, the high concentration of phosphonate groups on CPP makes these peptides much stronger HA nucleators than Arg.

Overall, this comparison shows that although a small, positively charged molecule like Arg cannot compete with highly phosphonated biomolecules, its effectiveness at promoting HA nucleation once bound to GO or CGO is not drastically different from that of large, negatively charged biomolecules such as PDA or gelatin. This confirms the importance of the large local supersaturation achieved in the presence of Arg, due to both complex formation and electrostatic reasons, as discussed in the previous section.

5. Conclusion

We studied the effect of a positively (Arg) and a negatively (Glu) charged AA bound to GO flakes on the precipitation of HA at physiological conditions. Both AAs were bound to GO after transforming oxygenated GO functionalities in carboxyl groups (CGO). We showed that while both AAs increased the HA precipitation rate on CGO compared with GO and CGO, Arg increased the GO mineralization rate much more than Glu, giving rise to larger amounts of HA precipitates, including both micrometre-sized spherulitic aggregates and smaller nanoparticles. We explained these results based on two factors: Arg can form more stable complexes with Ca²⁺ and ions than Glu; and, the presence of both carboxyl and amino groups exposed to the solution on CGO-Arg may favour local supersaturation with respect to HA by electrostatically attracting both Ca²⁺ and ions. This result may not be generalized to all positive and negative functionalities; for example, groups able to form even stronger complexes with either ion, such as phosphonate groups, which can attract Ca²⁺ ions very strongly, may change this balance. Indeed, when comparing our results with previous work studying the mineralization of surface-modified rGO, we showed that the amount of HA deposited on CGO-Arg was much lower than that observed in the presence of heavily phosphonated peptides like CPP.

Overall, this report provides new insights into the effect of single AAs bound to surfaces on HA mineralization in physiological conditions, and some guidelines on how to improve the mineralization of GO for bone regeneration applications. If our insight on the importance of both complex formation and the presence of positive and negative groups to increase local supersaturation is correct, future attempts to modify material surfaces to improve HA mineralization should include similar amounts of both negatively and positively charged groups, e.g. by combining carboxylates or phosphonates and amino groups. If nothing else, this would allow for the production of a surface that better mimics the composition of the AAs present in the collagen sequence close to the hole zones.

Authors' contributions

As the first author of the paper, M.T. designed and conducted all the experiments. In addition, material preparation and testing, data collection and analysis, and writing the manuscript have been her responsibility. M.C. was the project's supervisor and guided the whole process and extensively reviewed the manuscript. R.G. provided SEM access and gave advice on the SEM analysis presented in this paper. N.B. collected the SEM images and SEM-EDS results shown in figure 6 and table 3, and helped with the analysis of such data.

Competing interests

We declare we have no competing interests.

Funding

This research project is supported by the McGill Engineering Doctoral Award, the Canada Research Chair foundation, the Natural Science and Engineering Research Council of Canada, the Canada Foundation for Innovation, the Center for Self-Assembled Chemical Structures and the Fonds Quebecois de la Recherche sur la Nature et les Technologies.