In Vitro Characterization of Hierarchical 3D Scaffolds Produced by Combining Additive Manufacturing and Thermally Induced Phase Separation

Abstract

This paper reports on the hybrid process we have used for producing hierarchical scaffolds made of poly(lactic-co-glycolic) acid (PLGA) and nanohydroxyapatite (nHA), analyzes their internal structures via scanning electron microscopy, and presents the results of our in vitro proliferation of MC3T3-E1 cells and alkaline phosphatase activity (ALP) for 0 and 21 days. These scaffolds were produced by combining additive manufacturing (AM) and thermally induced phase separation (TIPS) techniques. Slow cooling at a rate of 1.5 °C/min during the TIPS process was used to enable a uniform temperature throughout the scaffolds, and therefore, a relatively uniform pore size range. We produced ten different scaffold compositions and topologies in this study. These scaffolds had macrochannels with diameters of ~300 μm, ~380 μm, and ~460 μm, generated by the extraction of embedded porous 3D-plotted polyethylene glycol (PEG) matrices. The other experimental factors included different TIPS temperatures (−20 °C, −10 °C, and 0 °C), as well as varying PLGA concentrations (8%, 10%, and 12% w/v) and nHA content (0%, 10%, and 20% w/w). Our results indicated that almost all these macro/microporous scaffolds supported cell growth over the period of 21 days. Nevertheless, significant differences were observed among some scaffolds in terms of their support of cell proliferation and differentiation. This paper presents the results of our in vitro cell culture for 0 and 21 days. Our optimal scaffold with a porosity of ~90%, a modulus of ~5.2 MPa, and a nHA content of 20% showed a cell adhesion of ~29% on day 0 and maintained cell proliferation and ALP activity over the 21-day in vitro culture. Hence, the use of additive manufacturing and designed experiments to optimize the scaffold fabrication parameters resulted in superior mechanical properties that most other studies using TIPS.

1. Introduction

Each year, around 7.9 million fractures that require treatment occur in the United States, with 5–10% of these fractures progressing to delayed unions or non-unions [1]. The likelihood of fracture is increased in individuals with diseases that impact bone density. One example of such a disease is osteoporosis, which affects approximately one in three women and one in five men over the age of 50 [2]. Currently, the “gold standard” for treating bone non-unions is the autologous bone graft, due to this procedure’s low risk of rejection and a relatively high success rate (50–80%) compared to other treatments [3]. Other options to treat non-unions include allografts and external site stimulation [4], though both of these options [5] and autografts [6] have their limitations. Due to the high human and economic costs of fractures that result in non-unions, focus has shifted to the field of scaffold-based tissue engineering, which aims to create biodegradable scaffolds that can promote tissue regeneration.

Due to a demand for better treatment options, many different tissue engineering strategies are being examined as potential replacements for current treatments. These strategies largely focus on the use of a three-dimensional (3D) porous scaffold to both provide mechanical stability as the fracture heals and promote bone regeneration by providing a matrix onto which cells can adhere and mature. Studies hypothesize that scaffolds that mimic the chemical composition and mechanical properties of bone extracellular matrix would promote the best outcome [7, 8], because such a design would stimulate the patient’s own cellular response to repair the defect. Using computer-aided design and additive manufacturing (AM), researchers have been able to create scaffolds with well-defined pore size and pore interconnectivity. Porosity is a vital factor in the ability of cells to move into the construct and maintain cell viability, because a porous structure allows for cell migration into the scaffold, as well as vascularization to provide cells with the necessary nutrients and oxygen [8]. A defect that is too large for the body to completely bridge unassisted is known as a critically sized defect. Generally, critically sized defects occur in fractures greater than 1–2 cm in length, or with 50% loss of the circumference of the bone [9]. Fractures of this size generally require implantation of a scaffold material to promote healing due to the body’s limited capability to repair large defects in the bone.

Poly(lactic-co-glycolic) acid (PLGA) is commonly used as a scaffold material in tissue engineering due to its ease of processing in various conventional scaffold fabrication techniques and use in additive manufacturing techniques, as well as its controllable biodegradability and biocompatibility. PLGA degrades through hydrolysis of its ester linkages to produce lactic acid (LA) and glycolic acid (GA), and although these products are not harmful to the body on their own, buildup can cause accelerated degradation of the scaffold [10]. Degradation rates can be tuned through various methods, such as by changing the molecular weight of the polymer and by varying the ratio of GA to LA. PLGA with a higher percentage of LA is less hydrophilic, and therefore degrades more slowly due to the polymer absorbing less water [10]. Although PLGA is commonly used in tissue engineering, its poor mechanical properties and cell affinity mean that it is most often used with a ceramic component in a polymer/ceramic composite scaffold. Hence, composites of polymer matrices and biologically-active nanoparticles have gained interest in the biomedical field [11, 12].

The weight of dry bone is made up of 65 – 70% inorganic calcium phosphate and 30 – 35% organic matrix of collagen and fibrous protein [13]. Hydroxyapatite (HA) is a ceramic with chemical and structural similarities to the inorganic component of native bone [14, 15]. Studies have shown that HA holds great promise in biomedical applications due to its bioactivity, biocompatibility, and osteoconductivity. However, HA has found only limited use in clinical applications on its own due to its poor mechanical properties and difficulty of processing [16]. Our previous study has shown that hierarchical scaffolds made of PLGA and nanohydroxyapatite (nHA) can support the short-term growth of murine MC3T3-E1 cells when grown in static culture [17].

Thermally induced phase separation (TIPS) can be used to create highly porous scaffold structures with well-defined pore size and interconnected pores that allow for vascularization. Pore size and interconnectivity can be controlled by changing the TIPS process parameters, such as solvent/polymer ratio, polymer type, solvent composition, concentration, and quenching rate and temperature [18–20]. In the TIPS method, 1,4-dioxane is a commonly used solvent due to its relatively low boiling point (101.1 °C) and high melting temperature (11.8 °C), which makes it easy to remove via the freeze-drying method [21]. Once PLGA is dissolved in 1,4-dioxane, the solution is cooled below its solubility temperature to induce phase separation into a polymer-rich and a polymer-lean phase [18]. The resulting scaffold is highly porous with high pore interconnectivity. Compared to other scaffold fabrication methods such as solvent casting and particulate leaching, gas foaming, and electrospinning, the TIPS method is relatively fast and controllable. Combining TIPS with AM has been shown to create scaffolds with macroscopic channels to guide bone ingrowth and vascularization as well as micropores to enable cell adhesion and differentiation [17, 22].

While the TIPS method generates micropores for oxygen and nutrient supply, the 3DP technique produces macrochannels for cell migration and bone ingrowth. In our previous study, we demonstrated that these hierarchical scaffolds are highly porous (> 85%) with a modulus exceeding 5 MPa [23]. In this study, these scaffolds were seeded with MC3T3-E1 preosteoblastic cells. This paper reports on the hybrid process we have used for producing these scaffolds, analyzes their internal structures via scanning electron microscopy, and presents the results of our in vitro cell proliferation and alkaline phosphatase activity (ALP) for 0 and 21 days.

2. Materials

PLGA (RESOMER^® LG 824 S; I.V. = 1.7–2.6 dl/g; with 82:12 lactide:glycolide ratio) was purchased from Evonik Biomaterials. Poly(ethylene glycol) (PEG, 35,000 Da), nHA (< 200 nm particle size (BET), ≥ 97%, synthetic), ethanol (EtOH), and 1,4-dioxane (ACS reagent, ≥ 99.0%) were purchased from Sigma-Aldrich (St. Louis, MO, USA).

3. Methods

3.1. Rheological Measurements

The shear viscosity of PEG was measured using a twin-bore Rosand RH2000 capillary rheometer equipped with an 8-mm-long straight die with an inner diameter of 0.5 mm. The measurements were performed at shear rates ranging between 50 and 600 s⁻¹. The shear viscosity data as a function of shear rate were generated at three temperatures (85 °C, 100 °C, and 115 °C). The relationship between the shear viscosity (η) and shear rate ($\dot{\gamma }$) for a non-Newtonian polymer melt obeying a power-law model can be shown as [24].

$$\eta =K{\dot{\gamma }}^{n-1}$$ (1)

In this equation, K is the consistency index and n is the power-law index. The Weissenberg-Rabinowitsch correction [25] was made by the equipment to calculate the true wall shear rate.

3.2. Thermogravimetric Analysis (TGA)

A small PEG granule (~15 mg in weight) was heated to 800 °C at a rate of 10 °C/min using a thermogravimetric analyzer (TGA Q500, TA Instruments), and the sample weight loss over time was recorded. The TGA measurements helped to verify the thermal stability of PEG within the temperature ranges targeted in this study. In addition, TGA measurements were conducted on the porous scaffolds to verify the retention of nHA at the end of the scaffold fabrication process. A sample ~7 mg in weight was heated to 800 °C at a rate of 10 °C/min, and the sample weight loss was recorded over time.

3.3. Scaffold Design

Table 1 lists the compositions of the 3D scaffolds investigated in this study as well as the TIPS temperatures used for preparing these constructs. In our previous study, we used an I-optimal design of experiments (DoE) to design hierarchical scaffolds with a modulus exceeding 5 MPa while keeping the porosity over 85% [23]. The experimental factors in our design were the strand diameter of 3D-plotted PEG (D), PLGA concentration (C₁), nHA content (C₂), and the TIPS temperature (T), leading to 18 designs. Scaffold characterizations in our previous study enabled us to eliminate some of these designs. For example, a TIPS temperature of −20 °C led to defects in some scaffolds during phase separation and upon scaffold removal from the beaker. Table 1 shows the ten scaffold designs that we have chosen for our in vitro cell culture in this study. Eight of these 10 scaffolds were produced at either −10 °C or 0 °C, whereas two scaffolds were produced at −20 °C. The latter were defect-free due to the relatively high concentrations of PLGA and nHA in the formulation of these scaffolds.

Table 1.

The scaffold compositions and TIPS temperatures investigated in this study.

Scaffold Number	Strand Diameter, D (μm)	Polymer Concentration, C1 (% w/v)	nHA Content, C2 (% w/w)	TIPS Temperature, T (°C)
1	460	8	20	0
2	380	12	10	0
3	300	10	10	−10
4	460	12	20	−20
5	300	8	10	0
6	380	8	0	−10
7	380	10	20	−10
8	380	10	10	−20
9	380	10	10	−10
10	460	10	0	0

3.4. Fabrication of 3D-plotted PEG

Table 2 lists the 3D-plotting parameters used in this study, and Fig. 1a shows the schematics of the 3D-plotting process as well as the descriptions of the plotting parameters (D, L, and h). The strand diameter (D) represents the diameter of the macrochannels in the final PLGA/nHA or PLGA scaffolds prepared using the TIPS method. Polyethylene glycol (PEG) was melt-processed using a 3D-Bioplotter (EnvisionTEC^™, Germany) to produce porous PEG constructs, as we have previously reported [23]. These PEG constructs were eventually extracted using deionized (DI) water to generate interconnected macrochannels within our PLGA/nHA or PLGA scaffolds prepared using the TIPS method as described below.

Table 2.

The 3D-plotting parameters used to produce porous PEG constructs with three target strand diameters.

Needle Diameter (μm)	Plotting Speed (mm/s)	Dispensing Pressure (bar)	Melt Temperature (°C)	Target Strand Diameter (μm)	Layer (Slicing) Thickness (μm)	Distance Between Strands (μm)		Number of 3D-Plotted Layers
Dn	V	P	Tm	D	h	L	L + D	N
300	1.7	2.5	115	300	240	1,000	1,300	15
400	2.1	1.5	115	380	304	1,000	1,380	14
400	1.7	1.5	115	460	368	1,000	1,460	12

Figure 1.

(a) Schematics and description of the 3D-plotting parameters, (b) The scaffold fabrication process using the 3DP-TIPS technique. Reproduced with permission from Ref. [23].

3.5. Scaffold Fabrication

Figure 2b shows the schematics of the scaffold fabrication using the TIPS method [23]. Briefly, PLGA was cryogenically ground using a model 6770 Freezer/Mill (SPEX SamplePrep^™) and a stainless steel grinding mill operating at 10 cycles/s. We used a precool time of 5 min and a grinding time of 10 min. The PLGA powder was then blended with nHA particles, when applicable, inside a 50-ml beaker and then mixed with 6 ml of 1,4-dioxane to prepare PLGA and PLGA/nHA solutions, with PLGA concentrations of 8%, 10%, and 12% w/v (see Table 1). After heating for 2.5 h at 60 °C inside a water bath, the mixture was sonicated three times (10 min each) at an amplitude of 20 μm using a Fisherbrand model 50 Sonic Dismembrator^™. Then, four 3D-plotted PEG constructs were placed side-by-side at the bottom of the beaker, covered using parafilm, and transferred inside a model ZP-16–1.5-H/AC^™ Cincinnati Sub-Zero environmental chamber. A flask containing ethanol was also placed inside the environmental chamber, and then the chamber was heated to 40 °C at a rate of 2 °C/min, maintained at this temperature for 5 min, and then cooled to the target temperatures at a rate 1.5 °C/min (see Table 1). After 90 min of phase separation, ~40 ml of ethanol was transferred to the beaker to extract 1,4-dioxane. Ethanol was refreshed 1 h before the end of the freezing period, and then the chamber was heated up to 20 °C at a rate 1.5 °C/min. The scaffold was then air-dried for 45 min and punched using 10-mm biopsy punches. Then, the scaffolds were washed at 40 °C inside biopsy meshes in a DI water bath equipped with a stir-bar. After 48 h of PEG extraction by water, the scaffolds were kept for 1 h inside a −80 °C freezer, and then freeze-dried inside a Labconco 1-L FreeZone^™ for 3 days (see Fig. 1).

Figure 2.

Shear viscosity as a function of shear rate for PEG at three different melt temperatures obtained using a capillary rheometer.

3.6. Porosity Measurements

As described in our previous study [23], the following equation was used to calculate the porosity of the punched scaffolds (n=15):

$$\phi = \frac{V_a-V_t}{V_a}\times 100$$ (2)

In this equation, V_a is the apparent volume of the scaffold, estimated using thickness and diameter measurements, while V_t is the true volume of the scaffold material calculated based on scaffold matrix density (ρ_m) and scaffold mass (m) as follows:

$$V_t=m/{\rho }_m$$ (3)

Therefore, the scaffold density (ρ) and porosity (φ) are related as:

$$\rho ={\rho }_m\left(1-\phi \right)$$ (4)

For each scaffold, the matrix density (ρ_m) in (3) and (4) was estimated using the rule of mixtures based on the PLGA and nHA percentages:

$${\rho }_m={\rho }_{\mathit{\text{PLGA}}}\left(1-\frac{C_2}{100}\right)+{\rho }_{nHA}\left(\frac{C_2}{100}\right)$$ (5)

where ρ_PLGA = 1.3 g/cm³ and ρ_nHA = 3.16 g/cm³ [14] are the PLGA and nHA densities, respectively.

3.7. Mechanical Characterization

The compressive modulus of the scaffolds was measured using a model 3344 single-tower mechanical tester (Instron^™) featuring a 100 N load cell [23]. Unconfined ramp compression tests were performed at a displacement rate of 1 mm/min and a preload of 0.1 N. The slope of the stress vs. strain graph at 5% strain was calculated for three independent measurements (n=3).

3.8. MC3T3-E1 Cell Growth

Murine MC3T3-E1 cells were purchased from the American Type Culture Collection (ATCC.org). All cells were previously frozen at passage 4 and grown to passage 7 for seeding. Cells were grown to ~80% confluency and cultured according to the supplier’s specification. Briefly, cells were cultured in proliferation media consisting of alpha modified minimum essential media without ascorbic acid (α-MEM, Gibco) supplemented with 10 % fetal bovine serum (FBS, Gemini), 1% Glutamax, 1% amphotericin B, and 1 % penicillin/streptomycin (Gibco) and incubated in a 37 °C, 5 % CO₂ incubator. The media was replaced every 2–3 days and the subconfluent cells were harvested using 0.25 % trypsin EDTA (Gibco). After cells fully detached from the bottom of the flask, they were centrifuged at 3000 rpm for 5 min, re-suspended with 5 ml of α-MEM complete media and counted using a hemocytometer.

3.9. Cell Seeding on the Scaffolds

The scaffolds were placed inside a sterile 50-ml conical tube, and then ~40 ml of 70% ethanol was added to the tube. The sealed tubes were incubated for ~3 h. Following ethanol aspiration, 25 ml of sterile DPBS (with Ca⁺⁺/Mg⁺⁺) was added, then DPBS was removed by aspiration, and the scaffolds were washed with DPBS (with Ca⁺⁺/Mg⁺⁺) twice. Subsequently, 25 ml of DPBS (with Ca⁺⁺/Mg⁺⁺) was added, and the tube was incubated for 1.5 h. The scaffolds were washed with DPBS (with Ca⁺⁺/Mg⁺⁺) two more times. Finally, 25 ml of α-MEM complete media was added to the tube and incubated overnight. The scaffolds were then transferred to non-tissue culture treated 24-well plates under a biosafety cabinet. Then, 3×10⁶ cells suspended in 100 μl of α-MEM complete media were seeded onto each scaffold, then the plate was covered and transferred to a 37 °C, 5% CO₂ incubator. After 2 h of incubation, 1 ml of α-MEM complete media was added to each well, and the scaffolds were incubated at 37°C, 5% CO₂ for ~24 h. After the 24 h time period (designated as day 0), the proliferation media was replaced with 1 ml of differentiation media, which was α-MEM complete media supplemented with 50 μg/ml ascorbic acid (Sigma) and 10 mM β-glycerophosphate (Sigma) [17]. Then, the scaffolds designated as day 0 were harvested and the remaining scaffolds were kept in culture. The media was replaced every 2 days, and the cell-seeded scaffolds were harvested on day 21.

3.10. DNA Content Assay

The 10-mm scaffolds were cut into four equal pieces, and one of the four pieces from each scaffold was used for the DNA content assay (n=3). The number of cells adhered on the scaffolds was quantified by analyzing their DNA content using a CyQUANT cell proliferation assay kit (Life Technologies). Briefly, the cell-seeded scaffolds were washed with DPBS (with Ca⁺⁺/Mg⁺⁺) twice and then stored at −80 °C. Following the manufacturer’s instructions, RNAse/lysis buffer was composed of 1X cell-lysis buffer supplemented with 180 mM NaCl, 1 mM EDTA, and 1.35 Kunitz units/ml DNase-free RNase. Dye/cell-lysis buffer solution was prepared by diluting CyQUANT® GR dye 400 fold into 1X cell-lysis buffer. RNAse/cell-lysis buffer (250 μl) was added to each scaffold, the scaffolds were vortexed, sonicated, incubated for 1 h on ice, and then sonicated again. Sonications were performed using the Branson Sonifier 250 to break apart the scaffold and lyse the cells. Samples were sonicated at a duty cycle of 60%, with the timer on hold for 25–30 bursts. Following sonification, samples were vortexed and then centrifuged at 13,000g for 3 min. Volumes of 25 μl, 50 μl, and 100 μl were taken from the top of dilution so that each sample could be run in triplicate. All wells were brought to 100 μl with RNAase/cell-lysis buffer, and then 100 μl of dye/cell-lysis buffer was added to each sample. The fluorescence of the samples was measured using a NOVOstar cell-based fast kinetic microplate reader with a 482/50 excitation filter and a 528/20 emission filter. The standard curves were generated using known numbers of MC3T3-E1 cells sonicated in RNA/lysis buffer [17].

3.11. Alkaline Phosphatase (ALP) Activity

ALP activity is a measure of osteogenic differentiation and was used in this study to quantify the differentiation state of the cell population on each scaffold. We used this extensively documented marker as a way to determine if our scaffolds permit osteogenic differentiation of the MC3T3-E1 cells in culture. One quarter of each scaffold was placed in a 1.5 ml microcentrifuge tube with 250 μl of ALP assay buffer (BioVision) and stored at −80 ℃ until assayed (n=3). The tubes were thawed on ice and then sonicated as described above. Pierce bicinchoninic acid (BCA) protein assay kit (Thermo Scientific) was used to determine the concentration of protein in supernatant. The absorbance was measured using a SpectraMax Plus384 plate reader at 405 nm. Then, 5 μg of total protein was assayed for alkaline phosphatase (ALP) activity spectroscopically by measuring the conversion of p-nitrophenyl phosphate to p-nitrophenol using an ALP assay kit (BioVision). The ALP activity is quantified by measuring the production of p-nitrophenol using a SpectraMax Plus384 plate reader at an absorbance of 405 nm. Only five out of 10 scaffolds were used for the ALP assay. These scaffolds were selected based on the DNA content assay, while taking their moduli into consideration. Scaffolds 1, 4, and 8 had a modulus below 1 MPa, and therefore, were excluded from the ALP assays.

3.12. Scanning Electron Microscopy (SEM)

The scaffold samples were sputter-coated with 20 nm gold using Denton Desk II Sputter Unit and imaged using a Zeiss Supra^™ 35V microscope at an accelerating voltage of 4–15 kV. The cell-seeded scaffolds were fixed in 1% glutaraldehyde/2% paraformaldehyde in Dulbecco’s phosphate-buffered saline (DPBS) for 1 h, and then washed in DPBS four times at five minutes each. Following a serial dehydration at 25, 50, 75, 95 and 100%, the scaffolds were critical-point dried in a Tousimis Samdri 780A. The dried samples were then gold coated and imaged at an accelerating voltage of 4 kV.

3.13. Statistical Analysis

The experimental data were presented as mean ± standard deviation. Analysis of Variance model with post hoc analysis was used to verify the statistical significance of differences between each set of samples and our optimal scaffold, and p<0.05 was considered statistically significant.

4. Results

4.1. Rheological Measurements

Figure 2 shows the shear viscosity (η) of PEG as a function of shear rate ($\dot{\gamma }$) at the three temperatures used for rheological measurements. The dotted lines represent the trend lines based on the power-law model, as shown by the equations on the chart. The shear-thinning behavior is clearly visible at these temperatures. An average shear-thinning index of n ≈ 0.139 suggests a highly non-Newtonian behavior (n = 1 for a Newtonian fluid). This shear-thinning behavior indicates that the dispensing pressure and the needle diameter during 3D-plotting could highly influence the viscosity of PEG melt. Hence, the dispensing pressure was kept within a relatively narrow range (1.5–2.5 bar) and the needle diameter varied only between 0.4–0.5 mm. Then, the strand diameter was fine-tuned by manipulating the plotting speed (see Table 2).

4.2. Thermogravimetric Analysis (TGA)

The TGA thermogram of PEG is presented in Fig. 3a. The slope of the sample weight vs. temperature shows a peak, indicating that the thermal degradation of PEG starts at ~200 °C. The PEG grade used in this study had a melting endotherm below ~70 °C. Therefore, a temperature of 115 °C enabled a quick thermal equilibrium inside the 3D-plotting cartridge without posing any major risk of early thermal degradation. Figure 3b and 3c show the TGA thermograms of the scaffolds containing 10% w/w nHA and 20% w/w nHA (scaffolds 2 and 1, respectively). The vertical axis (weight %) in both cases demonstrate that the remaining sample weight at the end of the TGA experiment closely matches the nHA content of these scaffolds.

Figure 3.

(a) A TGA thermogram showing the weight loss for PEG as a function of temperature. The slope of sample weight vs. temperature shows a peak that starts at ~200 °C (onset of thermal degradation). (b & c) The TGA thermograms of the scaffolds containing 10% w/w nHA and 20% w/w nHA (scaffolds 2 and 1, respectively).

4.3. Scanning Electron Microscopy (SEM)

The side-view SEM images of a 3D-plotted PEG construct is shown in Fig. 4a. The strand diameter (D) reveals some variation in the zones of contact with the underlying layer. Nevertheless, the fine-tuned 3D-plotting parameters allowed generating strand diameters approaching the target stand diameters listed in Table 2 (i.e., ~290 μm, ~388 μm, and ~457 μm, respectively [23]). Figure 4b–d show the SEM images of scaffold 9 prepared in this study (Table 1). This scaffold contained 10% PLGA and 10% nHA, and was prepared at a TIPS temperature of −10 °C using a PEG strand diameter of 380 μm. The side-view image in Fig. 4b shows well-defined macrochannels, generated after the extraction of 3D-plotted PEG. These macrochannels are surrounded with a microporous matrix created after 1,4-dioxane extraction, showing micropores ranging between 20 – 40 μm (Fig. 4c). Figure 4d shows the surface topography generated during phase separation, with submicron structural features that could potentially favor cell adhesion. Figure 4e and 4f show a similar microporous structure and sub-micron surface topography for scaffold 7, which is one of the two desirable scaffolds based on our previous study.

Figure 4.

(a) The cross-sectional view of a 3D-plotted PEG construct as seen by SEM, (b) cross-sectional SEM image of a PLGA/nHA scaffold produced by the 3DP-TIPS method (scaffold #9 in Table 1), (c & d) SEM micrographs showing the micropores and sub-micron structural features of scaffold #9, (e & f) SEM micrographs showing the micropores and sub-micron structural features of scaffold #7.

Looking at the SEM images of the cell-seeded scaffold on day 21 (Fig. 5), it appears that these scaffolds support cell adhesion and spreading. The scaffolds in Fig. 5 were prepared at a TIPS temperature of 0 °C or −10 °C, using the PEG strand diameters of ~300, ~380, and ~460 μm. Different combinations of PLGA and nHA concentrations (0, 10, and 20%) were used to prepare these scaffolds (Table 1).

Figure 5.

SEM images of MC3T3-E1 cell growth on the surface of 3DP-TIPS scaffolds on day 21 (a) Scaffold #1, (b) Scaffold #2, (c) Scaffold #3, (d) Scaffold #6, (e) Scaffold #7, and (f) Scaffold #10.

4.4. Scaffold Porosity and Modulus

Figure 6 compares the modulus and the porosity of the ten scaffolds. The porosity varied between 89 – 93% for these scaffolds. As anticipated, the formulations with higher PLGA concentrations generated denser scaffolds. The modulus of these scaffolds varied within an order of magnitude, ranging between 0.53 – 5.2 MPa [23]. Scaffold 9 (10% nHA) was the optimal output of our designed experiment due to its relatively high modulus (~5.1 MPa) and a porosity of 89%, while showing well-defined macrochannels and micropores (Fig. 4b–4d). These characteristics were closely matched by scaffold 7 (20% nHA) with a modulus of ~5.2 MPa and a porosity of 90% (Fig. 4e–4f). These two scaffolds had the same macrochannel diameter (~380 μm), PLGA concentration (10% w/v), and TIPS temperature of −10 °C. Given the higher nHA content of scaffold 7 compared to scaffold 9 (20% vs. 10%), we have used scaffold 7 as our optimal for the statistical analysis using both one-way and two-way ANOVA.

Figure 6.

The modulus and porosity of the 10 scaffolds listed in Table 1.

4.5. Cell Proliferation

As described earlier, the 10-mm scaffolds were cut into four equal pieces, and one of the four pieces from three scaffold replicates was used for the DNA content assay. Therefore, the seeded cells onto each piece was ~750,000. These results indicated that between 26 – 41% of the cells were retained on the scaffolds after seeding, except for scaffolds 1 and 9 that had low cell retention of 11% and 18%, respectively. The normalized cell proliferation was calculated by dividing the number of cells at each time point by the cell retention on day 0 for each scaffold. Statistical analysis was performed on cell proliferation and ALP activity measurements as described below. Table S1 in the supporting information provides the details of our statistical analysis, including the estimated differences between two group means, the standard errors, and the 95% confidence interval (CI) of the estimated differences.

4.5.1. One-way ANOVA of day 0 measurements

We have compared the raw measurements among scaffold groups in each day using boxplots (Fig. 7a and 7b). We fitted two one-way ANOVA models, the first one with day 0 number of cells for each scaffold as the response, and the second one with day 21 number of cells in each scaffold as the response, and both using the scaffold group as the treatment variable. Post-hoc analysis was carried out for both models using Dunnett’s procedure [26] to adjust the p-values of pairwise comparison when we have multiple pairs of comparison simultaneously. When compared with scaffold 7, scaffold 1 was the only scaffold that showed a significantly lower mean number of cells on day 0, while the other scaffolds did not show statistically significant differences (Fig. 7c). However, this changed in day 21. When compared with scaffold 7, scaffolds 2, 3 and 10 showed significant differences: Scaffold 2 had a significantly higher mean number of cells than scaffold 7, while scaffolds 3 and 10 both had significantly lower mean DNA content than scaffold 7 on day 21. All other scaffolds did not reveal statistically significant differences in the mean when compared to scaffold 7.

Figure 7.

(a & b) Boxplots of number of cells on day 0 and day 21, respectively; (c) Number of cells on day 0 and day 21 compared for the 10 scaffolds; (d) Number of cells compared between day 0 and day 21, (e) Boxplot of the normalized number of cells measurements on day 21, and (f) normalized cell number on day 21 comparing the 10 scaffolds. (* indicates p<0.05).

4.5.2. ANOVA Model with two factors (scaffold and day) and their interaction

We also looked at the two-way ANOVA model using both scaffold design (1 – 10) and time point (0 vs. 21), as well as their interaction in the comparison, so that both factors were considered in a single model. The results indicated that when compared with day 0 cell count of scaffold 7, only day 21 measurement of scaffold 2 had a statistically higher mean, whereas all the other groups of day-scaffold combinations were not statistically different in the post hoc analysis. When compared with day 21 cell count of scaffold 7, day 21 measurements of scaffold 2 had a significantly higher mean, while day 0 measurement of scaffold 1 and day 21 measurement of scaffold 10 had a significantly lower mean. All the other groups of day-scaffold combinations were not statistically different in the post hoc analysis. When we compare the day 21 vs. day 0 difference in mean within each scaffold, only scaffold 2 had significantly higher mean on day 21 than day 0, and scaffold 10 had significantly lower mean on day 21 than day 0 (Fig. 7d).

4.5.3. One-way ANOVA of day 0 measurements (normalized data)

We have compared the normalized data among scaffold groups using boxplots (Fig. 7e). The data in this figure are cell numbers on day 21 normalized to the number of cells at day 0 for each scaffold. This gives a clear representation of the cell number increase for each scaffold over the 21-day culture period. We fitted a one-way ANOVA model for day 21 normalized number of cells as the response, again with the scaffold as the treatment group variable. Post-hoc analysis was carried out for the model using Dunnett’s procedure to adjust the p-values of pairwise comparison when we have multiple pairs of comparison simultaneously. When compared with scaffold 7, scaffolds 1, 2, and 10 showed significant differences: Scaffolds 1 and 2 had significantly higher means of normalized number of cells than scaffold 7, while scaffold 10 has significantly lower mean normalized number of cells than scaffold 7 on Day 21 (Fig. 7f). All the other scaffolds did not show statistically significant differences in the mean when compared to scaffold 7.

4.6. Alkaline Phosphatase (ALP) Activity

4.6.1. One-way ANOVA of day 0 measurements

We have compared the ALP activity measurements among scaffold groups in each day using boxplots (Fig. 8a and 8b). We fitted two one-way ANOVA models, the first one with day 0 ALP activity in each scaffold as the response, and the second one with day 21 ALP activity in each scaffold as the response, and both using scaffold as the treatment variable. Post-hoc analysis was carried out for both models using Dunnett’s procedure to adjust the p-values of pairwise comparison when we have multiple pairs of comparison simultaneously. For the day 0 measurements, when compared with scaffold 7, scaffold 6 had a higher mean ALP activity that was statistically significant, while the others were not statistically different from scaffold 7 (Fig. 8c). For the day 21 measurements, when compared with scaffold 7, scaffolds 6 and 9 showed significant differences: Scaffold 6 had a significantly higher mean than scaffold 7, while scaffold 9 had a significantly lower mean than scaffold 7 on day 21. All other scaffolds were not showing statistically significant differences in the mean when compared to scaffold 7.

Figure 8.

(a & b) Boxplots of day 0 and day 21 ALP measurements, respectively; (c) ALP activity on day 0 and day 21 comparing the 5 scaffolds. (* indicates p<0.05).

4.6.2. ANOVA Model with two factors (scaffold and day) and their interaction

We also looked at the two-way ANOVA model of the ALP activity using both scaffold design and time point (0 vs. 21), as well as their interaction in the comparison. When compared with day 0 of scaffold 7, both day 0 and day 21 ALP activity of scaffold 6 had a statistically higher mean, whereas all the other groups of day-scaffold combinations were not statistically different in the post hoc analysis. When compared with day 21 ALP activity of scaffold 7, both day 0 and day 21 ALP activity of scaffold 6 had a statistically higher mean, whereas all the other groups of day-scaffold combinations were not statistically different in the post hoc analysis. When we compare the day 21 vs. day 0 difference in mean within each scaffold, none of the scaffolds showed significant differences. Table S2 in the supporting information provides the details of our statistical analysis, including the estimated differences between two group means, the standard errors, and the 95% confidence interval (CI) of the estimated differences.

5. Discussion

For a tissue engineering scaffold to be successful, it must be able to support cell adhesion and differentiation. Highly porous scaffolds can help to overcome the challenge of supplying nutrients and oxygen to the cells inside of a 3D construct. We have prepared hierarchical 3D scaffolds made of PLGA and nHA by combining TIPS and 3D-plotting (3DP), which is an AM technique [17, 23]. The high metabolic activity of bone cells limits the size of scaffold-tissue constructs that can be cultured in bone tissue engineering [27]. The role of scaffold pore size on the in vivo vascularization of tissue-engineered bone has been investigated in several studies. In general, there is a consensus that bone growth in vivo is favored via scaffold pore sizes ranging between 300 – 400 μm, and the formation of capillaries has been reported in newly formed bone [28, 29]. In addition, a pore size exceeding 350 μm has been shown to enhance cell migration and angiogenesis [28]. Hence, the internal architectural features of 3D scaffolds have a key role in the success of a bioengineered bone graft.

Characteristics that increase favorable conditions for cells to adhere and differentiate are important for transplantation success of tissue engineered scaffolds [30, 31]. Cells interact with the ECM surface topography guiding adhesion, proliferation and differentiation [30, 32–34]. Cells have surface receptors called integrins that can sense the surface topography and biomolecules on the surface of the ECM [14], and the integrins can up and downregulate specific genes [35]. Surface roughness and topography as well as surface chemistry are of importance to biological interactions with the ECM because of the integrin receptors mechanism [36, 37], and both characteristics can be manipulated [37, 38]. Some studies have explored ideal surface roughness leading to increased cell interactions and protein creation [39–41]. In addition, mimicking the physico-mechanical properties of bone tissues has been emphasized in the development of scaffolds for bone tissue regeneration [42, 43].

Although a limited body of research has looked into how specific TIPS parameters influence surface topography for maximal cell adhesion, increased ALP activity, indicating increased osteogenic differentiation, has been measured in the presence of sub-micron topography created by TIPS [44]. In addition, the TIPS method can generate 3D scaffolds with nanotextures, with a surface roughness that promote protein adsorption and cell adhesion [44]. The SEM results presented in this study (Fig. 4d) revealed a sub-micron surface topography produced by phase separation and 1,4-dioxane extraction. This microporous matrix surrounded the macrochannels produced by the extraction of 3D-plotted PEG.

We used a ramp temperature profile at a low cooling rate (1.5 °C/min in the present study), which has shown to produce a relatively uniform micropore size throughout the scaffolds [23, 45]. Compared to −20 °C and −10 °C, the scaffolds prepared at 0 °C experienced a shorter time to reach the target phase separation temperature. In light of this, a dense microporous matrix was generated at a high TIPS temperature of 0°C (e.g., scaffold 1 and 2 in Table 1), whereas larger micropore sizes were produced at a TIPS temperature of −10 °C (scaffold 6 and 7) and −20 °C (scaffold 4 and 8) [23].

Scaffold 9 (10% nHA) was the optimal output of our designed experiment in our previous study, due to its relatively high modulus (~5.1 MPa) and a porosity of 89%, while showing well-defined macrochannels and micropores (Fig. 4b–4d). These characteristics were closely matched by scaffold 7 (20% nHA) with a modulus of ~5.2 MPa and a porosity of 90% (Fig. 4e–4f). These two scaffolds had the same macrochannel diameter (~380 μm), PLGA concentration (10% w/v), and TIPS temperature of −10 °C. Given the higher nHA content of scaffold 7 compared to scaffold 9 (20% vs. 10%), we have used scaffold 7 as our optimal for the statistical analysis.

MC3T3-E1 subclone 4 cells were used in the present study. This subclone exhibits high levels of osteoblastic differentiation when grown in the presence of ascorbic acid and phosphate donor like β-glycerophosphate [46]. MC3T3-E1 has been used extensively in studies ranging from osteogenic differentiation to bone cell physiology to bone tissue engineering. In addition, our previous study used this cell line [17]. This allows the comparison of our findings with those previously published studies using this cell line.

Scaffold 1 displayed reduced initial attachment as evidenced by significantly fewer cells in the scaffold on day 0 compared to scaffold 7. On day 21 (Figure 7c), scaffold 2 had the greatest number of cells and was significantly different that scaffold 7 and displayed a significant increase in cell number after 21 days (Figure 7d day 0 vs. day 21). Scaffolds 3 and 10 had significantly fewer cells than scaffold 7 on day 21 (Figure 7c) and scaffold 10 displayed a significant decrease in cell number after 21 days in culture (Figure 7d day 0 vs. day 21). Finally, comparing normalized cell numbers (Figure 7f), scaffolds 1 and 2 displayed significant increases in cell numbers over the 21-day culture period than scaffold 7 whereas scaffold 10 displayed a significant reduction in cells during 21 days in culture.

These results suggest that some of these scaffold compositions decreased initial cell adhesion but still supported cell growth (scaffold 1), most supported initial attachment, some supported increased proliferation (scaffold 2), and others demonstrated a reduction in cells over the 21-day culture period (scaffold 10). Scaffold 7 and 9, designated as our two desirable scaffolds, showed a cell adhesion of ~29% and ~18% on day 0, respectively. These results might point to the effect of higher nHA content (20% for scaffold 7 vs. 10% for scaffold 9) on cell adhesion and growth. Scaffold 2 showed a cell adhesion of ~33% on day 0 and provided a favorable matrix for cell proliferation over the 21-day culture. Sub-micron structural features tend to dominate at a TIPS temperature of 0 °C and lead to a surface topography favoring cell retention on scaffold 2. Nevertheless, this scaffold topology did not have well-defined macrochannels and micropores, due to the limited crystal growth during phase separation at 0 °C that interfered with PEG extraction [23]. This could reduce its favorability for our future in vivo studies.

Scaffold 6 supported increased osteogenic differentiation as measured by ALP activity at both day 0 and day 21 as the ALP activity for scaffold 6 was significantly greater that scaffold 7 at day 0 and day 21 (Fig. 8c). This scaffold had no nHA and a low concentration of PLGA (Table 1), where both factors are expected to reduce the density of the scaffold. Osteogenic differentiation was significantly decreased in scaffold 9 on day 21 compared to scaffold 7 (Fig. 8c). The trend towards increased ALP activity on all scaffolds at day 21 compared to day 0 (with the possible exception of scaffold 9) suggests that most of these scaffolds support osteogenic differentiation of MC3T3-E1 cells in culture. Further investigation on the effect of scaffold micro/macro-architecture and composition is necessary to draw a conclusion on the ALP activity of these scaffolds and their ability to promote bone formation. In our future studies, which will be done on a few optimal scaffold formulations, we intend to extend the studies to longer time points and additional measures of osteoblast differentiation that will include mineralization. Ultimately, we intend to utilize our optimal scaffolds in in vivo studies of bone formation in animal models.

Fabrication of polymer/ceramic scaffolds through additive manufacturing has been well studied. Combining additive manufacturing with conventional methods has also been considered by others [47, 48]. The scaffold fabrication technique used in our study does not require direct deposition of the polymer/solvent mixture, thereby reducing the risk of high exposure to organic solvents during scaffold fabrication. The combination of TIPS and 3D-printing used by our team aims to produce composite scaffolds with controlled macro and micropore size. The use of designed experiments to optimize the fabrication parameters has resulted in superior mechanical properties compared to most other studies using TIPS. In studies using TIPS, but no additive manufacturing, the scaffolds of PLGA and nHA showed a compressive modulus ranging between 0.7 to 2.53 MPa [22, 49]. Lao et al. [50] produced electrospun scaffolds composed of nHA and PLGA, which resulted in 22.5% cell adhesion after 24 hours measured with a DNA assay. This fits within the range of cell adhesion measured for our scaffolds. Overall, this study generated scaffolds with a higher compressive modulus than scaffolds composed of similar materials produced via TIPS while supporting cell viability and osteogenic differentiation.

6. Conclusions

A limited body of research has looked into how specific TIPS parameters influence surface topography for maximal cell adhesion and increased ALP activity. This paper reports the cell proliferation and ALP activity on the hierarchical scaffolds made of PLGA and nHA. These scaffolds were produced by combining additive manufacturing (AM) and thermally induced phase separation (TIPS) techniques. We produced ten different scaffold compositions and topologies in this study. Our results indicated that, besides scaffold 10, all these macro/microporous scaffolds supported cell viability over the period of 21 days; there were fewer cells on scaffold 10 at day 21 compared to day 0 whereas all other scaffolds contained the same, or an increased, number of cells at day 21 compared to day 0. Nevertheless, significant differences were observed among these scaffolds in terms of their ability to support cell proliferation and differentiation. The number of cells on scaffolds 1 and 2 increased, the number of cells on scaffold 10 decreased, whereas the remaining scaffolds supported cell viability without significant changes in cell number over the 21 days in culture. After 21 days in culture, all scaffolds supported osteogenic differentiation (measured as ALP activity) and scaffold 6 displayed increased and scaffold 9 decreased differentiation support compared to scaffold 7. Our optimal design (scaffold 7) with a porosity of ~90%, a modulus of ~5.2 MPa, and a nHA content of 20% supported cell adhesion of ~29% on day 0 and maintained cell proliferation and osteogenic differentiation over the 21-day in vitro culture. In our future study, we will choose a small number of scaffold formulations to examine for extended times in culture.