Manufacturing Method for Customizable Collagen Membrane Scaffolds for Use in Tissue-Engineering Applications

Purpose Collagen are used as tissue engineering tool. The manufacturing methods often lack of reproducibility which limits their application to regenerative medicine. We aimed to set a simple and reproducible method for the production of collagen scaffolds for tissue engineering.


Matryoshka system
The temperature inside the system remained constant at 40°C (SD of ± 0.19°C at the stabilization phase and SD of ± 0.64°C in the presence of the samples). The RH at the stabilization phase ranged from 47-42% (SD, ± 1.68%), whereas it ranged from 51-33% (SD, ± 5.52%) during the sample-desiccation period (Fig. 1b, c). A volume of 50 mL of the saturated salt solution lasted for ~ 100 days when the samples were inside the matryoshka system. Shaker function was turned on 48 h after the gel was placed inside the matryoshka system, to avoid gel tilting.

Desiccation
The transformation of gels (Fig. 2a) into membranes occurred when the thickness of the collagen decreased to ~ 1 mm. Gels at 1x required 10-13 days to reach the membrane state, whereas 2x and 3x required approximately 15 and 20 days, respectively. We also discovered that phenol red must be removed during the gel-to-membrane transformation, to avoid further interference with transparency ( Fig. 2a, b). At the end of the 37-day desiccation period, the membranes were completely transparent (Fig. 2c).

Membrane rehydration
Membrane rehydration for ~ 20 min was necessary to achieve a good retrieval from the cast. First, the borders of the membrane were lifted, then the border was scratched gently with a rounded spatula. No support was needed to retrieve the membranes from the casts. A malleable, transparent material was observed in water suspension. The membranes were placed in a water drop on plastic lm, the borders were unfolded, and water was removed using a pipette, to straighten the membranes, which were left to dry again at room temperature (Fig. 2c).

Confocal microscopy
Confocal microscopy of the material obtained was performed at a width of 3.65 µm for 1x collagen membranes, 4.8 µm for 2x membranes, and 7.2 µm for 3x membranes (Fig. 3). Our method produced homogeneous surfaces. The deposition of a greater volume of collagen increased membrane width.
Optical microscopy and SEM surface structure A homogeneous surface was observed using optical microscopy; however, transparency was enhanced when the membranes were wet (Fig. 4a, b). Moreover, the surface structure changed over the desiccation process: collagen bers were observed at day 7 but had disappeared from the surface at day 37 ( Fig. 4c, b, e, f). The average ber diameter was 1.3 µm (SD ± 0.23). Surface homogeneity and lack of roughness also increased with increasing collagen concentration (Fig. 4d, e, f). SEM transversal-view laminar structure Membrane width increased with collagen concentration and lamina density (Fig. 5a, b, e, f). Laminar structure was conserved at the three concentrations (Fig. 5b, e, f). Hydration conditions led to lamina fusion (Fig. 5c, d).
In vitro cell adherence, viability, and cytotoxicity NIH3T3 mouse broblasts showed adherence to 1x, 2x, and 3x membranes. A uorescence analysis indicated that the 10 x 10 3 cells that were seeded initially proliferated to ~ 31.7 x 10 3 cells in wells with no membranes. The populations of cells cultured on 1x, 2x, and 3x membranes increased to ~ 35.8 x 10 3 , ~ 29.6 x 10 3 , and 29.5 x 10 3 cells, respectively. The populations of cells increased by 3-fold within 48 h.
One-way ANOVA detected no signi cant differences between the means of all four samples (P = 0.323, n = 3). However, compared with controls, the populations of cells cultured on 1x membranes increased by 12%. Consistently, we can assume the absence of a cytotoxic effect because the Cell Titer Blue reaction exhibited an inverse correlation with the LDH assay.

Fourier-transformed infrared spectra
The transmittance peak at wavelength 3281 ( Fig. 6, arrow 1) was decreased in 3x collagen membranes, indicating the formation of O-H bonds. Moreover, the ngerprint area corroborated the identity of collagen in membranes. Peaks 1660 (Amide I band), 1627, 1635 (β-sheet 2ry structures of Amide I), 1637 (triple helix), and 1679 (stretching C = O vibrations that are H bonded) were fused into a peak with least transmittance at 1635 (arrow 2). Peak 1635 indicating β-sheet secondary structures were correlated with the laminar structure shown in transversal SEM sections. The peak at 1535 (arrow 3) indicated Amide II, whereas the peak at 1240 (arrow 6) was characteristic of collagen. The contribution of peaks at 1446, 1396, and 1202 is seldom (or not at all) reported in the literature (arrows 4, 5, and 7). Peaks 1084 and 1029 (arrows 8 and 9) are related to nucleic acids. Peak 1084 was reported by Movasaghi as corresponding to the phosphodiester bonds of the phosphate/sugar backbone of nucleic acids, whereas peak 1029 was reported as corresponding to collagen and the phosphodiester groups of nucleic acids. Finally, the peaks at 565 and 527 (arrow 10) corresponded to phenyl group torsion. All peaks were intensi ed as collagen volume increased.

Transmittance analysis
The membranes produced using our method were ~ 90% transparent in the visible light spectrum (Fig. 7). The average transmittance for 1x membranes was 92.6%, whereas it was 94% for 2x and 89.21% for 3x membranes. Lower transmittance values were recorded at violet wavelengths (ranging from 380 to 450 nm); 1x membranes, 80.4%; 2x membranes, 82.98%; and 3x membranes, 77.56%. Moreover, 1x membranes yielded a transmittance peak of 99.6% at 560 nm and 2x membranes had a peak of 99% at 540 nm, both of which are in the green wavelength spectrum. In contrast, 3x membranes exhibited no peak but yielded higher transmittance values (94.7%) at 650 nm. Percent transmittance increased progressively and stabilized at 510 nm (1x) ad 590 nm (2x and 3x). Increasing collagen concentration decreased transmittance by ~ 2.5% per added volume, with 1x membranes showing the highest percent transmittance. ANOVA indicated the presence of signi cant differences (P ≤ 0.001) for 3x compared with 1x and 2x membranes.

X-ray diffraction analysis
The XRD patterns of pure collagen type 1 (1x, 2x, and 3x membranes) are shown in Fig. 8. All samples exhibited a typical broad hump around 22°, indicating that collagen was in an amorphous phase [34,35].
A high-intensity peak was observed at 25.9°, which was related with the hydroxyapatite (HA) 002 plane. A shoulder of the HA 320 plane in the hump was shown at 39.8° compared with datasheet 9-432 from the Joint Committee of Powder Diffraction Standards. All samples showed a similar pattern, regardless of the content of collagen.

Discussion
Various types of equipment, often specialized and expensive, are required for the production of collagen membrane scaffolds. The main synthesis methods used in this eld involve electrostatically charged trays and vacuum systems [36], electrospinning and lyophilization apparatuses [5], and even dialysis equipment [37]. Methods that require a desiccation process include equipment such as humidity chambers and clean air chambers [6,8] with controlled RH; however, no model or further description is provided for these systems [38,39]. In addition, the use of patented materials has been reported [15].
The matryoshka system is a closed desiccation device that provides a stable and gradual decrease in RH inside the desiccation bubble regardless of the environmental RH. This ensures system and synthesis stability. Hence, the characteristics of the scaffolds produced using this system are reproducible.
The system described here can be assembled using common laboratory equipment and materials (a desiccator, salts, and a shaking incubator), thus avoiding additional production costs. Our desiccation system exhibited an RH variation of 5.52%, which was similar to the 5-8% variation reported for commercial options of higher prices.
Based on desiccation conditions established previously [8], our synthesis process was carried out at 40°C and 40% RH over 37 desiccation days and produced collagen-based membranes that were comparable with previously described materials. Differences in width were addressed rst. A width of 20-50 µm is obtained after 2 weeks of desiccation using a collagen concentration of 2 µl/mm 2 [6,7], whereas our membranes were 3-7 µm thick, depending on collagen volume deposition (Figs. 4,5). The width of the scaffolds was also smaller than that of other materials of different origins and compositions, such as amniotic membranes [7], human puri ed type I collagen membranes [40], collagen-silk-broin [38], crosslinked collagen-gelatin-hyaluronic acid biomimetic lms [41], and decellularized and decalci ed sh-scale-derived extracellular matrix [42]. A literature review revealed that oriented-ow cast collagen laminates alone attained a thickness of 2-5 µm thickness per stacked lamina [43].
Our SEM data pertaining to surface and transversal ultrastructure agreed with those reported by other authors, regardless of the use of additional collagen volume in the production of the membranes. Randomly arranged brils are observed at the surface of the membranes after short desiccation periods [7,8]. Nevertheless, the bers observed here were larger by an order of magnitude, 1.3 µm (SD, ± 0.23), compared with the sizes of collagen membranes reported previously [8,44] and resembled the early brils reported for collagen gels [45]. This phenomenon is attributed to the different order of the neutralization and blending steps used for gel formation, which triggers the formation of collagen brils [45,46]. Moreover, the incubation conditions affect the increase in the thickness of brils: the lateral fusion of discrete (∼4 nm) subunits leads to an increase in bril diameter, followed by longitudinal growth [47]. Fibril bundling and total uni cation (greater bril density and homogeneity) were observed when drying time was increased and high temperatures were used, leading to the production of the homogeneous and smooth surface observed in 1x, 2x, and 3x membranes. A similar non-brillar surface was observed in collagen membranes crosslinked with 1-ethyl-3-(3-dimethyl aminopropyl) carbodiimide (EDC) and Nhydroxysuccinimide (NHS) [38,48] and UV [40]. Heat energy is involved in the process of bril uni cation. The transversal laminar arrangement is also conserved across pure collagen-based membranes produced using similar systems, although desiccation time, RH, and temperature vary [38,39].
The highly organized bril laminae observed transversally were responsible for the high transmittance of our collagen membranes, which was higher than that of the human cornea [40] and those of membranes obtained using comparable conditions (40% RH, 40°C, 1 week and 5 weeks), as reported by Calderon-Colon et al. Similar conditions reportedly improved the transmittance of Takezawa's rst collagen membrane. However, even at similar collagen concentrations, the transmittance pattern of 1x and 2x membranes was different from those obtained using the different experimental conditions described by Calderon-Colon et al. and from those of all materials mentioned previously, because of the peak detected at the green wavelength (around 550 nm, at which L and M cones are stimulated). In contrast, 3x membranes exhibited a transmittance pattern that was similar to that of other materials, with no peak observed at 550 nm. Transmittance patterns similar to that observed for our 3x membranes have been reported for other collagen membranes [38,40], combinations of silk-broin [38], chitosan [15], gelatin and hyaluronic acid [41], and sh-scale collagen scaffold [42]. The increase in collagen deposition (i.e., membrane width) sets a light transmittance threshold [43] and pattern. Therefore, the tailoring of the ultrastructure of these membranes by changing desiccation conditions and collagen amount allows the convenient adaptation of their optical transmittance [8].
Furthermore, because of our desiccation conditions and based on previous knowledge [8], we should expect a change in the bulk refractive index, because it depends on the composition and water content of the material. We analyzed the water content of our membranes using FTIR (Fig. 1, peak 3281, arrow 1); we found that it increased with the addition of collagen volume, which was most evident in the case of 3x membranes. It is likely that that additional water molecules are trapped because of the increase in the collagen matrix, which cannot be compensated, even by prolonged desiccation period. Moreover, in contrast with the report of Calderon-Colon et al., we did modify the composition of membranes by adding collagen volume and maintained the desiccation parameters. Therefore, we preserved the structural motifs by increasing the number/density of the brils (the peak indicated the presence of β-sheets at FTIR 1635 nm), which increased light scattering, thus explaining the lower transmittance observed for 3x membranes.
Desiccation time could set a volume decay tendency. We understand that the specimens reported by Calderon-Colon et al. had a variable thickness; therefore, they had to be normalized to 50 µm to enable comparisons (but the same amount of collagen was used for all conditions). Thus, we con rmed that membrane thickness is a function of the different desiccation conditions [8]. However, our 1x membranes at 1 week of desiccation were not functional: they were too weak even to resist hydration for separation from the cast. Similar observations were reported previously by other authors [40]. Therefore, the collagen membrane manufacturing process should be standardized across the various types of equipment.
Collagen membrane scaffolds have been widely reported as being biocompatible in vivo and in vitro [7,10,15,36,38,40,[49][50][51]. In vitro, our membranes were not cytotoxic and allowed cell adhesion and growth despite of their smooth surface. Cell interactions with collagen might favor and compensate for adhesion and growth despite the presence of a smooth surface [52,53]. Of note, we used NIH3T3 broblasts from mouse embryos that had lost contact inhibition. Thus, the assessment of proliferation beyond full con uence was not useful for obtaining a clear idea of speci c cell-type behavior and, hence, of tissue regeneration. Nevertheless, the experimental design for such purposes using membranes produced with the desiccation system described here should take into account cell size (smaller cells have smaller pores and lower membrane porosity) and strati cation behavior caused by the increase in lamina density and width that accompanied the addition of collagen volume in our scaffolds [4,32,54,55]. For example, if a 3D tissue model is intended, the use of a combination of 1x, 2x, and 3x membranes as different basal lamina could be an option for the adaptation of factors such as cell nutrients and oxygen supply.
Further study of various aspects of our experiments may help solve several common biosafety and quality issues. FTIR peaks corresponding to the presence of the sugar backbone of nucleic acids [56] (Fig. 6, arrows 8 and 9) could be a useful indication of the presence/absence of genomic material through a non-destructive test. However, similar peaks have been reported as corresponding to an ester band (1100 cm − 1) and were indicative of crosslinking [3]. Moreover, the complete FTIR spectrum is not presented often in similar reports of structural characterization (and quanti cation if absorbance values are provided) and comparisons among samples from different manufacturing rounds. A similar situation exists for the results of 3D confocal microscopy. Finally, ndings from the XRD analyses of amorphous collagen and HA provide data pertaining to the origin of the collagen and desirable properties regarding biological responses, such as enhancement of mineralization, long-term degradation, and tissue integration [57,58].

Conclusion
The speci cation of type of equipment, gel formation, and desiccation conditions is vital for the reliable, replicable, and uniform production of transparent collagen-based membrane-like scaffolds. This information is useful for customization purposes when addressing the various necessities of TE and regenerative medicine, such as transparency and stacking (for 3D culture). We intended to report detailed information for the use of these applications by other researchers. Moreover, information about the manufactured collagen material may clarify the effects of the width and other properties of membranelike collagen-based materials. We synthesized a biocompatible, transparent, sandwich-like and laminar collagen-based membrane ultrastructure with a smooth surface and a thickness of 3-7 µm. Future work should address the effects of these characteristics in different cell lines, primary cultures, and tissues.

Matryoshka system
The matryoshka system consisted of an incubator with a steady temperature (T) bearing a desiccator with a saturated K2CO3 solution, to control RH. First, 50 mL of a K2CO3 solution were prepared using 1.15 g of K2CO3 (Sigma-Aldrich, P5833, St. Louis, Missouri, USA) per milliliter of bidistilled water. The salt solution was placed in a 200 mL beaker and xed onto a 330 × 246 × 262 mm desiccator with a clear plastic dome (Thermo Scienti c Nalgene, 5310-0250, Cleveland, Ohio, USA). A closed system was created by placing the sealed desiccator inside a shaking-plate incubator (I5211DS; Labnet International, Edison, New Jersey, USA) (Figure 1 a). The temperature and shaking speed of the incubator were set to 40°C and 30 rpm, respectively. T and RH were measured with a Monitoring Traceable Hygrometer (4040CC; Traceable® Products, Webster, Texas, USA) placed inside the desiccator.
System-stabilization phase T and RH were registered daily until a temperature of 40°C and an RH of 40% were recorded consistently for ~7 days. During this period, no sample was placed inside the system.
Twelve-well plates were used as casts for 2.2 mL of the collagen mix (2.8 l/mm2 of collagen; 1x volume). Gel casts were also prepared at volumes of 2x and 3x. The plates containing the collagen mix were placed in an MCO-18AIC incubator (Sanyo, Osaka, Japan) and kept at 37°C, 5% CO2 for 2 h, until geli cation. No support ring was included.

Desiccation
Plates containing collagen gels were placed inside the matryoshka system. The temperature and shaking speed of the incubator were set at 40°C and 30 rpm, respectively. The system remained closed for daily T and RH registration. Around day 10-13, the volume of the collagen gels decreased and membranes were formed. The matryoshka system was then opened and membranes were rinsed with bidistilled water until elimination of the phenol red in the medium. The samples were placed back inside the system, to complete a period of 37 days of incubation.

Membrane rehydration
Plates were removed from the matryoshka system. Bidistilled water was poured into the wells of the plates containing the collagen membranes, left for ~20 min, then removed. The borders of each membrane were lifted using water pressure. Rounded-tip forceps were used to gently pull the remainder of the membranes away from the bottom of the well.

Collagen-based membrane characterization
Optical microscopy The surface of collagen membranes was observed with an Axiovert 40 CFL contrast microscope (Carl Zeiss Microscopy, Jena, Germany) and photographed. Membranes were humidi ed for 20 min and the water excess was then removed.

Scanning electron microscopy (SEM)
Samples were coated with gold using a Quorum QR150 ES sputtering system (Quorum Technologies, Laughton, United Kingdom). The surface and transversal ultrastructure of 37-day samples was observed and measured with an EVO MA Scanning Electron Microscope (Carl Zeiss Microscopy). Moreover, the bers of samples that were desiccated for 7 days (4 samples, 34 bers per sample) were photographed and measured using NIH ImageJ. Brie y, image calibration to microns was performed, followed by selection of linear regions of interest and accumulation using the Analysis>>Measure route.

3D confocal microscopy
Surface regularity and sample thickness were assessed in triplicate using an Axio-CSM 700 50x objective (Carl Zeiss Microscopy). Samples were hydrated for 20 min prior to data collection.
In vitro cell adherence, viability, and cytotoxicity Collagen membranes were cut into circles with a diameter of 5 mm, to t a 96-well plate. Membranes were sterilized with Microdacyn (Oculus Innovative Sciences, CA, USA) and rinsed with sterile water. Water was removed and membranes were left to dry and adhere to the bottom of the wells. NIH3T3 cells were seeded (~10,000) in 100 µl of DMEM F12 (12491-015; Gibco) and incubated overnight at 37°C, 5% CO2.
A cytotoxicity and viability test was performed using Cell Titer Blue (Promega, Madison, WI, USA) according to the manufacturer's instructions. Brie y, 20 µl of Cell Titer Blue were added to each experimental, assay control, and sample control well, followed by gentle resuspension and incubation at 37°, 5% CO2 for 2 h. The same conditions were used to perform a 2-50 x 103 cell uorescence ladder comparison. The Cell Titer Blue uorescent signal is produced by resazurin reduction, a viability indicator, which exhibits an inverse correlation with the release of lactate dehydrogenase, a cytotoxicity indicator. Fluorescence was recorded on a Synergy HT spectrophotometer (BioTek, Winooski, VT, USA). Triplicates were tested for each membrane concentration in the experimental condition (membrane + cells), the sample control condition (cells alone), the negative control condition (membrane alone), and the assay control condition (medium + Cell Titer Blue).

Fourier-transformed infrared spectra (FTIR)
Functional groups were identi ed on each membrane sample using an infrared spectrophotometer coupled with a Fourier transform Spectrum 400 apparatus (Perkin Elmer, Waltham, MA, USA) and recorded in the wavenumber range of 4000-400 cm−1 at room conditions.

Transmittance analysis
Absorbance was acquired throughout UV-VIS spectra (380-700 nm) using a Synergy HT spectrophotometer (BioTek, Winooski, VT, USA) and transformed to transmittance using the Beer-Lambert law equation. Experiments were performed in triplicate for all collagen volumes.

X-Ray diffraction analysis (XRD)
We used XRD to assess the crystallinity of the collagen membranes. XRD was recorded in the 2θ range, between 10° and 85°, with a step size of 0.026 using a PANalytical Empyrean diffractometer (PANalytical, Almelo, Netherlands) and CuKα radiation (λ = 1.5406 Å). The voltage applied was 45 kV and the current was 40 mA. For measurements, whole samples were xed to the corresponding support.

Statistical analysis
Statistical comparison of the experimental groups was carried out using paired Student's t-tests or analysis of variance (ANOVA). Signi cance was set at P < 0.05. Microsoft Excel (2013; Redmond, WA, USA) and Systat Sigma Plot (V. 11; San Jose, CA, USA) were used for data processing, statistical analysis, and graph generation.  Desiccation process: (a) gel (upper well) and membrane with phenol red; (b) rinsed membrane at day 37; and (c) membrane removed from the cast and dried at room temperature.  Contrast microscopy (10x image) of 37-day desiccated 1x membrane (a) with 20 min of hydration or (b) dry conditions. Surface micrograph of (c) a 7-day desiccated membrane and 37-day desiccated (d) 1x, (e) 2x, and (f) 3x membranes.  FTIR spectra of (a) pure collagen type 1, (b) 1x collagen membranes, (c) 2x collagen membranes, and d) 3x collagen membranes.