Carboxymethyl chitosan as a polyampholyte mediating intrafibrillar mineralization of collagen via collagen/ACP self-assembly

1. Introduction

The optimal scaffold material for bone tissue engineering applications is one that can be regenerated perfectly and completely by mimicking the properties of the matrix of bone tissues. Natural bone has remarkable mechanical properties in terms of its hardness, toughness, strength, and fracture resistance because of its complex and well-organized composite structure [1,2]. These properties can contribute to synergistically generated intra- and extrafibrillar mineralized collagen fibrils [3,4]. Intrafibrillar mineralization occurs through the deposition of minerals within the gap zones of collagen fibrils and extending along the microfibrillar spaces within a collagen fibril, while extrafibrillar mineralization occurs through deposition within the interstitial spaces separating the collagen fibrils [5]. The prominent mechanical properties of mineralized collagen are mainly attributed to intrafibrillar mineralization rather than extrafibrillar mineralization, which could reproduce the mechanical properties of natural bone [3]. Therefore, there is great interest in developing biomimetic methods for the intrafibrillar mineralization of collagen in vitro for artificial bone grafts.

Thus far, a small number of non-collagenous proteins (NCPs) (10%-15%) have been identified in the extracellular matrix (ECM) of mineralized tissues, such as osteocalcin, osteopontin, bone sialoprotein, amelogenin, dentin matrix protein1 (DMP1), and dentin sialophosphoprotein (DSPP) [6,7]. Increasing evidence has indicated that NCPs such as fetuin, amelogenin, dentin phosphoprotein (phosphophoryn), and DMP1 can collectively direct the intrafibrillar mineralization of collagen in vitro [[8], [9], [10]]. These NCPs are acidic (anionic) polyelectrolytes that are attributed to the prevalence of polyaspartic acid and phosphorylated serine residues along the protein backbone [11,12]. The anionic polyelectrolyte nature of NCPs enables them to capture calcium ions from a supersaturated solution with respect to hydroxyapatite (HAP) in vitro via electrostatic interactions between mineral ions and functional groups to stabilize amorphous calcium phosphate (ACP) nanoparticles, which are thought to be the precursors of HAP [13]. Currently, several electrolytes are used as analogues of NCPs to stabilize ACP nanoparticles to facilitate intrafibrillar mineralization of collagen in vitro, such as polyaspartic acid (pAsp), polyglutamic acid (PGA), polyacrylic acid (PAA), poly(allylamine) hydrochloride (PAH), and carboxymethyl chitosan (CMC) [[14], [15], [16], [17], [18], [19]]. These ACP nanoparticles can enter into collagen fibrils via gap zones based on capillary forces [14,20], size exclusion [8,21], charge interactions [16] or a balance between the osmotic equilibrium and electroneutrality [17] and convert into HAP to accomplish the intrafibrillar mineralization of collagen.

In previous studies on the intrafibrillar mineralization of collagen, biomimetic intrafibrillar mineralization was generally achieved via the model of self-assembly of collagen preceding mineralization (MFS), in which collagen self-assembled completely before ACP entered, became oriented, and grew along the axes of collagen fibrils. In contrast, studies of one-step collagen fibril self-assembly and mineralization via the mixing of acidic collagen with calcium ions and phosphate-containing solutions have been reported [22,23]. Interestingly, it was recently reported that under acidic conditions, in the presence of aqueous calcium phosphate (CaP) precursors, nanoscale micellar fibrils (NMF) mimicking the nanostructure of type I collagen fibrils in a polymerised lyotropic liquid crystal (PLLC) matrix were obtained via the molecular self-assembly of di-acrylate modified, non-ionic, amphiphilic block co-polymers. By increasing the pH in the PLLC matrix, the aqueous domains of the PLLC can be mineralized with calcium phosphates for intrafibrillar mineralization. This work suggested that CaP precursors could be involved in the self-assembly process of synthetic fibrils such as NMF to achieve an analogous intrafibrillar mineralization [24]. Thus, the works mentioned above suggested that except for the MFS model, the mineralization of collagen, including intrafibrillar mineralization can be accomplished via the manner of synergistic collagen fibril self-assembly and mineralization. However, these in vitro models have not been used to evaluate ACP nanoparticles under acidic conditions or to comprehensively investigate ACP nanoparticles and their mechanism of regulating collagen mineralization mediated by NCP analogues.

In vitro, collagen molecules self-assemble from the initial acidic environment in a quarter-staggered manner to form collagen fibrils, resulting in the formation of collagen fibres and bundles [25]. Furthermore, identifying the analogues of NCPs that can stabilize ACP under acidic conditions to achieve a combination of collagen self-assembly and remineralization would be meaningful. Fortunately, CMC, as a polyampholyte, has been found to have both carboxyl and amino groups and can therefore act as an either anionic or cationic polyelectrolyte below or above its isoelectric point (IP = 3.5), respectively [[26], [27], [28]]. In our previous study, CMC was proven to be able to stabilize ACP at pH 7 and promote intrafibrillar mineralization of collagen via the MFS model [18,19].

Here, we hypothesized that positively charged CMC (the cationic polyelectrolyte) can stabilize ACP nanoparticles by capturing negatively charged phosphate ions from a supersaturated solution with respect to HAP in vitro under acidic conditions (pH < 3.5). The inclusion of ACP nanoparticles stabilized by CMC under acidic conditions in the process of collagen self-assembly could allow synergistic collagen fibril self-assembly and mineralization. This mode of mineralization is a biomimetic model and is different from established mechanisms in which fibril self-assembly precedes mineralization (the MFS model).

2. Materials and methods

2.1. Chemicals

CMC (Qingdao Honghai Bio-tech, Shandong, China) was applied to stabilize the CaP mineralization solution. Other chemicals, including CaCl₂•2H₂O, K₂HPO₄, HCl, NaOH, acetic acid, Tris-HCl-NaCl buffer and 4-(2-Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer, were all purchased from Energy (Shanghai, China) and Sigma-Aldrich (Beijing, China).

2.2. Preparation of CMC/ACP mineralization medium

The CMC solution was prepared by dissolving 240 mg of CMC powder into 38 ml of deionized water with stirring (500 rpm) for 15 min. The pH of the CMC solution was adjusted to approximately 2 or 7. To obtain the A solution, 0.042 g of K₂HPO₄ was dissolved in 1 ml of deionized water. To obtain the B solution, 0.059 g of CaCl₂•2H₂O was dissolved in 1 ml of deionized water. The B solution and A solution were successively added into the CMC solution (pH 2) in a dropwise manner with stirring (500 rpm) for 5 min to form the CMC/ACP nanocomplex (pH 2) mineralization medium. In contrast, the CMC/ACP nanocomplexes (pH 7) were obtained by adding the A solution followed by the B solution to the CMC solution (pH 7) in a dropwise manner with stirring (500 rpm) for 5 min.

2.3. Preparation of type I collagen

Rat tail tendon fascicles from Sprague-Dawley (SD) rats (7 weeks old) were placed in Tris-HCl-NaCl buffer (0.05 M, pH 7.4) for 24 h and dissolved in acetic acid (0.1 M) for 3 d. They were then centrifuged at 3000 rpm for 30 min at 4 °C. The collagen gel was collected from the supernatant and lyophilized. Before use, a type I collagen solution (8 mg/ml) was obtained by dissolving the freeze-dried collagen in 0.1 M acetic acid at 4 °C.

2.4. Synchronous self-assembly/mineralization of collagen (SSM)

The synchronous self-assembly/mineralization of collagen was conducted by mixing 2 ml of acidic CMC/ACP solution (pH 2) with 1 ml of collagen solution in dialysis bags (8000-14,000 Da). The dialysis bags were placed in 20 ml of HEPES buffer (10 mM) for 2 d, and the pH was approximately 6.0. To accelerate the transformation of ACP to HAP, the pH value of the contents in the dialysis bags was adjusted to 8 by adding NaOH (1 M) dropwise. Then, the dialysis bags were placed in 20 ml of HEPES buffer (10 mM) and cultivate for 2 d. The HEPES buffer was changed every day. Mineralized collagen gel was obtained via centrifugation at 3000 rpm for 5 min and then washed with deionized water 3 times (5 min each time). The collagen scaffolds were prepared by lyophilizing the mineralized collagen gel.

2.5. Mineralization following self-assembly of collagen (MFS)

Three millilitres of the collagen solution were packed into each dialysis bag, and the dialysis bags were then placed in 20 ml of HEPES buffer (10 Mm) for 3-5 d to obtain the self-assembled collagen gel. The HEPES buffer was changed every day. The single-layer collagen model was prepared by dipping gold grids into the collagen gel. Mineralization was conducted by floating the grids upside down in CMC/ACP solution (pH 7.4) for 3-5 d.

2.6. Kinetic measurements

The kinetic self-assembly and mineralization of collagen can be speculated by the turbidity change of the solution. The turbidity of samples was monitored by a UV-vis spectrophotometer (Thermo Scientific Multiskan GO, MA, USA) at 400 nm. Pure collagen solution, CMC solution, CMC/ACP solution, a mixture of collagen and CMC/ACP solution or a mixture of collagen and CMC solution was packed into a dialysis bag, respectively. The dialysis bags were placed in 20 ml of HEPES buffer (10 mM), and then the pH value and turbidity of each sample were measured at different intervals.

2.7. Characterization

The IP of CMC was determined based on the change in turbidity at different pH levels. The turbidity of CMC was detected on a UV-vis spectrophotometer (Thermo Scientific Multiskan GO, MA, USA) at 800 nm. The pH of the prepared CMC solution (5 mg/ml) was adjusted with 1 M HCl and NaOH at room temperature.

The zeta potential of the CMC and CMC/ACP solutions was measured using a Malvern Nano ZS system (Malvern, England). This system employed the laser Doppler electrophoresis technique using a folded capillary cell equipped with gold electrodes. The optic unit contained a 4-mW He-Ne laser with a wavelength of 633 nm.

The morphology of the collagen and CMC/ACP nanocomplexes was examined via transmission electron microscopy (TEM, JEOL 2100 F, Japan) at 80 kV. Selected-area electron diffraction (SAED) was used to determine the crystallinity of the minerals. A TEM-energy dispersive X-ray analysis (TEM-EDX) was performed to measure the calcium (Ca) and phosphorus (P) contents of the mineralized collagen. The spatial distribution of calcium and phosphorus within the collagen fibrils was measured via TEM-EDX elemental mapping. High-resolution quantitative TEM-EDX imaging and bright-field imaging were performed. Spectrum acquisition and elemental mapping were conducted using an EDAX INCA X-sight detector.Partial samples were stained with uranyl acetate for 15 s, washed with deionized water, and then dried.

The attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR) spectra of the collagen scaffolds were recorded in reflection mode using an infrared spectrophotometer (SHIMADZU 8400S, Japan). Spectra were collected in the range from 700 to 4000 cm^-1 at a resolution of 16 cm^-1 and scanned 100 times.

3. Results

3.1. Isoelectric point (IP) of CMC

The IP of CMC was approximately pH 3.5 (Fig. 1). The transmittance of the CMC solution was lowest during pH 3 to pH 4 (greatest precipitation and lowest concentration of CMC). Near pH 3, the CMC began to precipitate and the solution became turbid (transmittance approach to minimum). When the pH is close to IP, the CMC was separated out more and precipitated out significantly at IP (about pH 3.5), leading to a decrease in turbidity of supernatant instead. And thus, the transmittance of the supernatant increased somewhat at IP. Next, with the increasing of pH, the deposited CMC began to dissolve, which caused increasing in turbidity (decreasing in transmittance). After pH 4, the CMC dissolved significantly, so that the transmittance raised obviously again (Fig. 1(a)). The net charge of the CMC solution at pH 3.5 was almost zero, which was consistent with the result measured based on transmittance (Fig. 1(b)). Solutions of ACP nanoparticles stabilized by CMC at pH 2 and pH 7 were characterized via TEM (Fig. 2) and zeta potential measurements (Fig. S1 in Supplementary information). These solutions were designated as CMC/ACP nanocomplexes for the mineralization of collagen. The zeta potentials of nanocomplexes of CMC/ACP were 22.2 mV at pH 2 and -20.7 mV at pH 7, respectively (Fig. S1).

3.2. Characterizations of CMC/ACP at pH 2 and pH 7

The solutions of both CMC/ACP (pH 2) and CMC/ACP (pH 7) nanocomplexes showed calcium phosphate (CaP) aggregates composed of discernible nanoparticles (electron-dense aggregates), which were amorphous, as indicated by SAED (Fig. 2). Individual ACP nanoparticles were observed in the solutions of CMC/ACP nanocomplexes (pH 2 and 7) (Fig. 2). These aggregates of the ACP nanoparticles are solute CaP precursors that form via liquid-liquid phase separation and have been identified as prenucleation clusters of CaP [29,30]. At pH 7 (>IP), further aggregation of ACP nanoparticles leading to precipitation is inhibited because of the ionized carboxyl groups of CMC, which interact with the positively charged calcium ions of ACP to stabilize ACP nanoparticles. By comparison, at pH 2 (<IP), ACP essentially cannot form; however, since the amino groups of CMC are protonated, CMC shows a positive charge and can therefore stabilize ACP nanoparticles via electrostatic interactions with the phosphate ions of ACP. Therefore, because of the presence of ACP nanoparticles in acidic conditions, under which the in vitro assembly of collagen began, collagen mineralization can be accomplished simultaneously with collagen assembly in vitro (synchronous self-assembly/mineralization of collagen, SSM) rather than following collagen self-assembly (mineralization following self-assembly, MFS).

3.3. Characterizations of SSM and MFS models

To investigate the SSM and MFS modes of mineralization with solutions of CMC/ACP nanocomplexes at pH 2 and pH 7, a time-resolved analysis was conducted beginning in the early stages of mineral formation. During the process of SSM, although bands characteristic of collagen self-assembly were not observed after 2 h (Fig. 3(a)), spindle-shaped calcium phosphate nanoparticles with a length of 40 nm were found arranged in parallel in a particular direction (Fig. 3(a)). After 12 h of mineralization, the shape of the collagen fibrils and the characteristic band were clearly observed, indicating that collagen self-assembly had been achieved (Fig. 3(b) and (d)). A simultaneous view of collagen fibrils under the pre-assembly and post-assembly parts is shown in Fig. 3(b), and the co-occurrence of collagen microfibrils and ACP nanoparticles is shown in the pre-assembly parts. ACP nanoparticles homogenously covered the surface of the self-assembled collagen fibrils (Fig. 3(c) and (f)), and both round and spindle-shaped CaP nanoparticles appeared in the gap zones (Fig. 3(c)). The exact location of the spindle-shaped ACP nanoparticles in self-assembled collagen fibrils was investigated by staining the fibrils (Fig. 3(d)). The main bodies of the nanoparticles were mostly located in gap zones, although some extended to the overlap zones (Fig. 3(d)). In fact, these spindle-shaped ACP nanoparticles were aggregates of smaller nanoparticles (Fig. 3(b)). Thus, loose flexible aggregates of ACP nanoparticles could be moulded into spindle-shaped aggregates inside collagen microfibrils via the template effect of collagen, which is driven by the self-assembly of collagen.

With increasing time, additional parallel CaP nanoparticles appeared in the gap zones of the collagen fibrils after 2 d (Fig. 4(a)-(c)). These CaP nanoparticles transformed into bundles of needle-like HAP crystals after 4 d and filled the collagen fibrils, and SAED showed a diffuse band characteristic of HAP (Fig. 4(d)-(f)). ATR-FTIR spectra showed the presence of a peak corresponding to the v₃PO₄ stretching vibration from 1021 to 1030 cm^-1, indicating the formation of HAP crystals in the mineralized collagen fibrils. With increasing mineralization time, the content of crystals increased in the mineralized collagen fibrils (Fig. 5). However, without the restriction of collagen templates, disordered needle-like HAP crystals could be observed outside of the fibrils, whereas well-ordered crystals were observed inside the collagen fibrils (Fig. 4(e)). Elemental maps also confirmed that Ca and P were well distributed inside the collagen fibrils (Fig. 6). Moreover, certain self-assembled collagen fibrils do not have trapped ACP nanoparticles, which could result in poor intrafibrillar mineralization of collagen (Fig. 4(f)).

3.4. Kinetic characterizations of SSM model

The turbidity of acidic pure collagen solution increased with dialysis time, which was shown in the turbidity-time curve (i.e., kinetic curve) with a logarithmic phase and a plateau, indicating the self-assembly of collagen fibrils induced by increased pH value (Fig. 7(a)). The TEM image of self-assembled collagen was taken at 120 min and 1800 min (Fig. 7(d) and (e)). As shown in Fig. 7(a), the pattern of the kinetic curve of the mixture of collagen and CMC, without a significantlogarithmic phase and a plateau was different from that of pure collagen, indicating that CMC significantly affected the kinetics of collagen self-assembly which was supported by the TEM image taken at 120 min and 1800 min (Fig. 7(f) and (g)); in the presence of CMC, the rate of collagen self-assembly became slow. However, the kinetic curve pattern of the mixture of collagen and CMC/ACP was similar to that of pure collagen, which indicates that CMC/ACP did not influence the process of collagen self-assembly much and induced intrafibrillar mineralization of collagen fibrils (Fig. 7(b) and (c)). The kinetic curve patterns of CMC/ACP and CMC were similar, with a rapid rise in the kinetic curve when approaching the IP of CMC (pH 3.5).

During the process of MFS, the calcium phosphate nanoparticles homogenously coated the surfaces of the collagen fibrils after 2 d (Fig. 8(c) and (d)), and these nanoparticles had also infiltrated the collagen fibrils via gap zones and were regularly observed in the gap zones (Fig. 8(c) and (d)). Subsequently, apatite crystals began to develop within a bed of ACP nanoparticles in gap zones and then elongated and fused with each other and filled the fibrils (Fig. 8(e) and (f)).

4. Discussion

CMC is a polyampholyte (Fig. 1) and thus presents characteristics that are similar to those of proteins. Generally, the Debye-Hückel (DH) theory for electrolytic solutions of simple ions, has been used to describe the electrostatic interaction among the fixed charges of a polyampholyte [28]. In particular, the tendency of polyampholyte chains to assume a globular conformation and reduced solubility at IP has been attributed to an intramolecular charge screening effect. Accordingly, repulsive electrostatic interactions between fixed charges are responsible for the expansion of polyampholyte chains that have a large net charge [28]. Outside of the IP, the CMC chains spread out and repel each other gradually because of the increase in the repulsive interactions between the —NH³⁺ cations or —COO^- anions on CMC chains [[31], [32], [33]]. This rationale was described in the upper picture in Fig. S2 in Supplementary information. It should be noted that the role of counter-ions and added electrolytes on the conformation of polyampholytes is also related to charge screening according to the DH model [28]. With the addition of calcium and phosphate ions, the pH of deposition of CMC/ACP slightly decreased. Calcium and phosphate ions added were different from Na⁺ and Cl^-, which could be ionization completely in water, while the Ca²⁺ and PO₄^3- tend to deposit. Therefore, at pH 2 if PO₄^3- couldn’t neutralize fixed positive charges of CMC chains completely, it could maintain a “balance” between the positively charged CMC chains and calcium ions inside the CMC/ACP nanocomplexes, which provide an appropriate condition for the formation of ACP. That is, PO₄^3- will screen negative charged CMC and lead to coils of chains, but Ca²⁺ weaken the screening effect of PO₄^3- due to the thermodynamics prediction of precipitation reaction. Therefore, Ca²⁺ and PO₄^3- prevent CMC chains from a globular conformation causing CMC deposition to a certain degree, whereas CMC chains inhibit the further aggregation of Ca²⁺ and PO₄^3- by separating them and thus stabilize ACP nanoparticles. At pH 7, the rationale is similar. This rationale is described in the lower picture of Figs. S2 and S3.

In vitro intrafibrillar mineralization of collagen has generally been conducted after the self-assembly of collagen, in which ACP nanoparticles stabilized by analogues of NCPs enter the collagen fibrils via gap regions and are converted into HAP crystals [13]. In this study, self-assembled collagen fibrils were intrafibrillarly mineralized by CMC/ACP nanocomplexes (pH 7) through the MFS model, which is consistent with previous studies [18,19]. In the MFS model, collagen molecules self-assembled into fibrils as a template for the ACP nanoparticles to infiltrate the collagen via gap zones to cause intrafibrillar mineralization of collagen (Fig. 9). By comparison, the intrafibrillar mineralization of collagen, accomplished by combining the self-assembly of collagen fibrils with the incorporation of ACP nanoparticles to fibrils, was particularly notable. Fig. 3(b) captures the collagen fibrils prior to assembly and the post-assembly collagen fibrils with incorporated ACP nanoparticles, further indicating that the intrafibrillar mineralization of collagen could be a synergistic process involving collagen fibril assembly and ACP nanoparticles. This process started at pH 2, at which ACP nanoparticles could be stabilized by CMC, as mentioned above. As the pH gradually increased, the collagen fibrils began to self-assemble from collagen molecules. Simultaneously, the ACP nanoparticles adopted spindle shapes with a length of approximately 40 nm, which corresponded to the gap region of the collagen, and they were arranged along the long axis of the collagen microfibrils (Fig. 3). The TEM images suggested that collagen microfibrils may wrap around the ACP nanoparticles and orient them in gap zones, and then collagen fibrils with embedded ACP nanoparticles form. Thus, this mode of mineralization involves the synergistic action of collagen fibril assembly and the transport of ACP nanoparticles. Next, the ACP nanoparticles fuse, extend and transform into HAP crystals along the pores between collagen microfibrils to achieve intrafibrillar mineralization of collagen. This process, called SSM, is described in the illustration shown in Fig. 9.

As shown in the results of the kinetic investigation and corresponding TEM characterization, in the dialysis process, the CMC may block collagen self-assembly in a similar manner to other electrolytes, such as polyaspartic acid (pAsp) [23], but CMC/ACP did not significantly inhibit the process of collagen self-assembly. This could be attributed to that ACP weakened the interaction between collagen and CMC due to the electrostatic interaction between CMC and calcium and phosphate ions. Within the first interval of sampling (30 min), the change in turbidity of the collagen-CMC/ACP was inconspicuous compared to that of collagen-CMC (Fig. 7(a)); this phenomenon indicates that ACP nanoparticles could completely blend and interact with self-assembling collagen molecules for a considerable time. Notably, after the first interval of sampling, the change in turbidity of the collagen-CMC/ACP rapidly reached the logarithmic phase compared to that of the collagen-CMC (Fig. 7(a)), indicating that CMC/ACP significantly accelerated the collagen self-assembly. Accordingly, the kinetic and TEM results revealed the synergistic feature of collagen fibril self-assembly and ACP nanoparticles in the SSM model.

The SSM model is different from those described in previous studies involving the one-step self-assembly and mineralization of collagen fibrils by mixing acidic collagen with calcium ions and/or phosphate-containing solutions (Pompe’s model and Nassif’s model, respectively) [22,23]. In Pompe’s model, the reactions of collagen self-assembly and calcium formation were combined in one step by increasing the pH value immediately, and fibril formation was thought to occur before calcium phosphate precipitation because the collagen fibrils were assumed to act as templates for mineralization [23]. The kinetic investigation suggested that the addition of polyaspartic acid retarded the transformation from ACP to HAP, but no obvious evidence of intrafibrillar mineralization was found in this model. In the absence of polyelectrolytes, intrafibrillar and interfibrillar mineralization was accomplished in Nassif’s model, in which the pH of the mixture of high concentrations of calcium and phosphate ions and acidic collagen molecules in a dialysis chamber was increased by ammonia gas diffusion. Nassif’s model leads to the co-precipitation of collagen molecules into fibrils and apatite minerals, which was defined as collagen/apatite self-assembly [19]. In contrast, in the SSM model, ACP nanoparticles and collagen molecules could coexist simultaneously for a considerable time and the SSM model emphasizes a synergistic assembly behaviour involving collagen fibrillogenesis and ACP nanoparticles anchoring in the gap zones of collagen fibrils (i.e., collagen/ACP self-assembly).

To accomplish the SSM process, the rate of collagen self-assembly in the synergistic process must be controlled, and the ACP nanoparticles must be stabilized by CMC under acidic conditions. The process of collagen self-assembly is highly dependent on the external pH because the charged amino acids of the collagen molecules are balanced and favour an elongated configuration, resulting in fibrillation at a pH close to the IP of collagen (9.3) [34,35]. Therefore, a higher external pH will likely cause collagen self-assembly to occur more rapidly. Reports have indicated that at pH 3, the fibrillogenesis of collagen is not observed over a period of several days, and at pH 5, fibril formation is sufficiently slow to allow fibrils to grow almost linearly over time [35]. In this study, dialysis ensured a slowly increasing environmental pH of the system of collagen molecules and ACP nanoparticles. With a slow increase in pH, the SSM model provides sufficient time for the assembly of collagen molecules and ACP nanoparticles, compared to the mineralization model with an immediate rise of pH, and thus could cause significant intrafibrillar mineralization of collagen (Fig. S4(a)). In Nassif’s model, ammonia gas diffusion also neutralized the acetic acid for 4-8 d to increase pH value of the mineralizing system slowly [22]. Consequently, increasing the pH value at a certain speed is key to accomplish intrafibrillar mineralization in the one-step model of self-assembly and mineralization of collagen fibrils.

The SSM model, which is based on the non-classical crystallization pathway, is a biomimetic method because of the presence of an analogue of NCPs (i.e., CMC) and the nanocomplexes of the analogues and ACP. In the non-classical crystallization pathway, amorphous inorganic primary particles coated/stabilized with organic molecules as analogues of NCPs can form larger mesocrystals via self-assembly and crystallographic alignment, which is called crystallization by particle attachment (CPA) [36]. In the SSM model, small nanoparticles of ACP assemble into a spindle-shaped aggregate and finally transform into HAP crystals via CPA in collagen fibrils (Fig. 3(g)), which is consistent with the non-classical crystallization pathway. Furthermore, in this SSM model, the transformation of ACP clusters into HAP in collagen fibrils, captured in Fig. 3(g), exhibited the typical spindle-shaped mineral area. Consistent with the previous report about mineralization of the NMF model, in the SSM model crystalline areas in collagen fibrils with faster crystal growth could incorporate or consume smaller crystallites, leading to the initial deviation from a spherical morphology to the typical spindle-shaped features. Hence, this step-flow cluster/dissolution-growth mechanism and the collagen fibrils with the confined domains both regulate the crystal growth along the c-axis of the collagen fibrils, which ultimately show a single orientation [24]. Accordingly, to some extent, the rationale of mineralization in the NMF model was verified in the SSM model.

Recent in vitro studies suggest that ACP can enter self-assembled collagen fibrils via capillary forces [14,20], size exclusion [8,21] or charge interactions between the polymer-mineral and fibril complexes at specific sites (bands a1-a3) in the gap zones of collagen [16,37]. In addition, the most recent study on this topic suggested that the balance between osmotic equilibrium and electroneutrality provides the driving force for the infiltration of polyelectrolyte-stabilized ACP nanoparticles in the water compartments of collagen to initiate intrafibrillar mineralization [17]; that is, ACP nanoparticles are deposited within collagen fibrils by replacement of the free and loosely bound intrafibrillar water in the self-assembled collagen [38,39]. These theories elucidate the phenomenon of intrafibrillar self-assembled collagen mineralization. In MFS, at pH 7, the negatively charged ACP nanoparticles stabilized by CMC do not selectively bind to the gap zones of collagen fibrils and instead coat almost the entire surface of the collagen fibrils, including the zones with positive and negative charges (Fig. 8(d)). In addition, the co-occurrence of ACP nanoparticles on the surface of collagen fibrils and mineral bands located in the gap zones of the collagen fibrils suggested that the ACP nanoparticles enter collagen fibrils via gap zones (Fig. 8(d)). These results indicate that the electrostatic interaction between ACP and collagen may not be the predominant factor mediating the infiltration of ACP nanoparticles into collagen fibrils. According to the size of the gap zones and the size-exclusion characteristic of collagen, only nanoparticles with a certain diameter (<40 nm) and molecules with a certain molecular weight (<40 kDa) can smoothly diffuse into collagen fibrils [21]. Therefore, the theories referenced above indicate that at pH 7, CMC cannot enter assembled collagen fibrils, and the dispersed nanoparticles with a diameter of 5 nm can be dislodged from the CMC matrix and driven into the collagen fibrils via gap zones (Fig. 8). More interestingly, during SSM, we observed the formation and arrangement of spindle-shaped aggregates of ACP nanoparticles with a length of approximately 40 nm. This finding indicated that the size-exclusion characteristic of collagen also plays a role in the SSM model, unlike in the MFS model (Fig. 8).

In addition, since the incorporation of free or bound water in the triple helix or on the collagen surface is followed by collagen self-assembly [40], competition between ACP and free or bound water was speculated to exist at gap zones or on the surface of collagen in SSM. In both MFS and SSM, the final peeling away of these localized hydration layers facilitates the transformation of ACP into HAP [41]. How CMC affects the self-assembly of collagen through the interaction between collagen and CMC and regulates the transformation of ACP into HAP is poorly understood. These points will be further investigated in the future.

With the aid of NCPs and their analogues, the infiltration of ACP nanoparticles into assembled collagen fibrils via gap zones has become a common method for accomplishing in vitro intrafibrillar collagen mineralization [42,43]. In contrast, research on in vivo mineralization of the collagen matrix induced by intracellular vesicles storing calcium phosphate has not distinguished between extrafibrillar and intrafibrillar mineralization of collagen [[44], [45], [46]]. Moreover, the detailed mechanism through which ACP nanoparticles penetrate into collagen in vivo has not been elucidated. Notably, however, procollagen molecules are clearly secreted from osteoblasts and extracellularly assembled into fibrils and then fibres because their large size does not permit formation within cells [47]. This scenario provides a spatiotemporal condition for in vivo early intrafibrillar mineralization of collagen via SSM model, which would be mediated by NCPs. After this process, the ACP nanoparticles would still enter into self-assembled collagen fibrils to further mineralize collagen via the MFS model as depicted in Fig. 10. Therefore, although the mineralization of collagen originating in an acidic environment is not consistent with the in vivo situation, the SSM model suggests that synchronous collagen/ACP self-assembly may represent a mechanism underlying early in vivo intrafibrillar mineralization. Furthermore, SSM provides a new strategy for the development of materials from mineralized collagen scaffolds.

5. Conclusion

CMC as an analogue of NCPs can stabilize ACP under acidic conditions due to its typical characteristics as a polyampholyte. Therefore, CMC can mediate the intrafibrillar mineralization of collagen via collagen/ACP self-assembly (i.e., synchronous self-assembly/mineralization of collagen, SSM). SSM provides a new strategy for in vitro intrafibrillar mineralization and the development of materials from mineralized collagen scaffolds for tissue engineering.

Acknowledgements

The authors gratefully acknowledge the financial support of the project from the National Natural Science Foundation of China (Nos. 31870947 and 81571016).

The authors have declared that no competing interests exist.

---