Janus 3D printed dynamic scaffolds for nanode ection-driven bone regeneration


 The application of physical stimuli to cell cultures has shown potential to modulate multiple cellular functions including migration, differentiation and survival. However, the relevance of these in-vitro models to future potential extrapolation in-vivo depends on whether stimuli can be applied “externally”, without invasive procedures. Here, we report on the fabrication and exploitation of dynamic additive-manufactured Janus scaffolds that are activated “on-command” via external application of ultrasounds, resulting in a mechanical nanodeflection that is transmitted to the surrounding cells. Janus scaffolds were spontaneously formed via phase-segregation of biodegradable polycapolactone (PCL) and polylactide (PLA) blends during the manufacturing process and behave as ultrasound transducers (acoustic to mechanical) where the PLA and PCL phases represent the active and backing materials, respectively. Remote stimulation of Janus scaffolds led to enhanced cell proliferation, matrix deposition and osteogenic differentiation of seeded bone marrow derived stromal cells (BMSCs) via formation and activation of voltage-gated calcium ion channels.


Introduction
The eld of tissue regeneration emerged in the early '90s under the concept of designing active scaffold materials that can recapitulate several functions of the native targeted tissue 1 , promoting the formation of neo-tissues, which have been so far limited to mechanical 2 and structural properties 3,4,5 , and the implementation of bioactive factors 6,7,8,9 . However, cells within a tissue are subjected to dynamic (rather than static) stimuli that determine their function and fate 10 . Recent advances in scaffold design have thus been focused on tackling the "temporal" character of the native cell. Dynamic physical stimuli such as mechanical, sonic, magnetic or electrical stimuli on in vitro cell culture has proven effective in modulating multiple cell responses. However, extrapolation to in-vivo applications where the stimuli has to be externally applied has been limited 11,12,13,14,15 . Among these, ultrasound stimulation has found the most exploitation in clinical environments in detection 16 and therapeutic applications 17 . Lately, the use of ultrasounds for enhanced bone fracture healing has been studied both in vitro and in vivo with relative success, pointing at the need for more accurate systems that can e ciently transmit the applied waves. Moreover, the pulse of the ultrasound wave has also a great impact on the outcome of the stimulation 18 . In these studies, low intensity pulsed ultrasounds (LIPUS) with frequencies on the range of few MHz are directly applied to cells or to cells seeded on a construct that, at these frequencies, remain static 19,20,21 . Some examples of low frequency ultrasound stimulation of cell substrates have shown the relevance of these to induce osteogenic differentiation of cells, but the concept yet not reached functional 3D scaffolds, neither the possibility of extrapolation to realistic in-vivo situation 22,23 .
Additive manufacturing, commonly known as 3D printing, has shown great potential on tissue regeneration applications due to the ease of fabrication of patient-customized scaffolds. 24 Changes to the morphology of a 3D-printed object upon exposure to external triggers has been recently de ned as 4D printing. Examples have so far been limited to one-way, non-reversible morphological transformations that are induced via permanent crosslinking of the materials, heat or pH changes non-suitable for physiological environments 25,26,27 . Few examples exist of morphological transformations that are twoway, reversible and repeatable, and these are based on the use of magnetic, inorganic and nonbiocompatible materials that are not suitable for biological applications 28,29 .
To bring further the eld of tissue regeneration and render "dynamic" traditionally static 3D printed scaffolds, we demonstrate a new concept based on the use of ultrasounds as stimulus. Under low frequency ultrasound stimulation the 3D printed scaffolds deform, becoming dynamic on-command when ultrasounds are activated, similarly to 4D printed objects. To control the extent and the pulse at which the scaffolds deform we took an engineering approach, designing scaffolds as ultrasound transducers with a de ecting and a damping element. we rst developed a novel concept in additive manufacturing by inducing phase segregation of biopolymer blends, which led to the formation of scaffolds with spatially controlled chemistries. At a 50:50 PLA:PCL biopolymer ratio, we induced the formation of a Janus-like phase that occurs homogeneously along the 3D structure. We modelled the mechanical de ection of the designed scaffold materials upon remote stimulation with different sound frequencies and investigated their effect on cell proliferation. Further, we investigated the de ection of Janus scaffolds that showed a shorter pulse width, behaving as transducers, and used them as biomaterial platforms to enhance bone formation. Our data shows that ultrasound stimulation of Janus scaffolds led to a more pronounced osteogenic differentiation, compared to non-stimulated (static) cultures on the same materials, with higher expression of osteogenic markers and higher deposition of matrix proteins collagen I and bronectin. Moreover, we found that this enhanced differentiation correlated with increased formation and further activation of voltage-gated calcium ion channels, suggesting this biological pathway was activated.

Results
Fabrication of 3D structures with spatially controlled chemistries via phase segregation.
Additive manufacturing allows for the fabrication of 3D scaffolds with well-de ned structures, but spatial control of the distribution of the deposited materials has so far been limited to the switch of materials between deposited layers (hundreds of microns) or to gradient structures at best 24 . To overcome this issue, we exploited basic concepts of physicochemistry, inducing phase-segregation of biopolymer blends (Fig. 1). We chose two widely used biodegradable and biocompatible polymers, polycaprolactone (PCL) and poly (D,L)-lactide (PLA), and blended them via extrusion to form laments that were then used to feed a traditional fused deposition modelling (FDM) printer (Fig. 1a). PCL and PLA are immiscible, and initial blending to form laments with PLA:PCL ratios of 20:80, 30:70, 40:60, 50:50, 60:40, 70:30 and 80:20 led to phase segregation ( Figure S1). PCL was able to crystallize from the melt, while the PLA was amorphous ( Figure S2). During the printing process, the polymer blends were molten and cooled down to room temperature (RT) while being deposited when PCL crystallized. Scanning electron microscopy images of the cross-section of the scaffolds showed that PCL and PLA phase-segregated during the printing process (Fig. 1b). A detailed observation of the formed phases via transmission electron microscopy (TEM, Fig. 1b) on the cross-sectional and longitudinal direction of the bers of the scaffolds showed the clear formation of PLA phases (light grey) within a PCL-rich matrix (dark grey) at a PLA:PCL ratio of 20:80. At this polymer ratio, the phase segregation followed the principle of nucleation, with PLA median particle cross-sectional and longitudinal areas of 0,152 µm 2 (CI 95,4%) and 0,216 µm 2 (CI 95,9%), respectively ( Figure S4 and S5). Increasing the PLA:PCL ratio to 30:70 led to a spinoidal decomposition and the formation of anisotropic phases that present maximum aspect ratios on the longitudinal (1,42; CI 95,6%) and transversal directions (2,77; CI 95,8%). At a PLA:PCL ratio of 50:50, the formation of well separated Janus structures was visualized by SEM, TEM, polarized light microscopy and light scanning microscopy (LSM), for which Rhodamine B and FITC were covalently attached to PCL and PLA, respectively, prior to scaffold fabrication (Fig. 1, Figure S3 and Figure S6). Moreover, the formation of Janus structures was observed to be homogeneous along the 3D structure, accounting for the same orientation of the phases along the depth of the structure (Fig. 1). At higher PLA:PCL ratios (60:40), the phases were inverted, forming PCL-elongated particles within a PLA matrix, with maximum particle sizes of 3,14 µm 2 (CI 95,3%) and 1,17 µm 2 (CI 96,01%) on the respective longitudinal and transversal directions.
From these data, it appeared evident that phase segregation occurs during printing and that different polymer ratios led to de ne structures, including a Janus structure at 50:50 PLA:PCL ratio.
Exploitation of Janus scaffolds as ultrasound-activated, dynamic mechanical scaffolds.
Having formed Janus scaffolds with distinct phases, we wished to investigate whether they could be exploited to induce mechanical stimulation to cells using biologically relevant stimuli. We selected ultrasounds at different frequencies to evaluate the optimal de ection of the scaffolds that could trigger a cell response. We modelled the de ection of single-layered Janus, PCL and PLA scaffolds xed on their extremes (mimicking implantation) and in liquid media under the application of continuous sound waves at different frequencies (Fig. 2b). The frequencies and wave intensities were de ned from experimental measurements of our homemade ultrasound set-up ( Figure S6, S7 and Table S1). The speed of sound on materials with similar densities, depends on the bulk stiffness of the media as per the equation: where ν is the speed of sound, Ε is the bulk Young`s modulus of the media and ρ is the density. At the same time, the speed of the wave is de ned by: , where λ is the wavelength and f, the frequency of the wave. Thus, for a given stimulation frequency, the wavelength of the transmitted wave (after crossing a material) will increase with increased material stiffness. Therefore, the exural modulus of the materials was measured under 3-point bending and used for the computational modelling. The exural modulus was 759 ± 62 MPa, 303 ± 70 MPa and 2326 ± 36 MPa for Janus, PCL and PLA, respectively ( Figure S8 and Table S2). For the selected frequencies of 10, 20 and 40 kHz, de ection amplitudes were v = λ ⋅ f determined: PLA displayed maximum de ections of 711, 112 and 20.7 nm, followed by Janus with 207, 79.3 and 19.8 nm and PCL scaffolds with 142, 47.1 and 14.2 nm for frequencies of 10, 20 and 40 kHz, respectively. These data indicate that de ection of the sound waves was attenuated at higher sound frequencies and by softer materials such as PCL. Knowing the predicted de ection amplitude for a given material and at the different sound frequencies, we then investigated its effect on hBMSCs proliferation.
We seeded hBMSCs on the different scaffolds and cultured them for 7 days applying 30 minutes of stimulation daily at either 0, 10, 20 or 40 kHz (Fig. 2c, d and e). At sound frequencies of 10 kHz when all PCL, PLA, and Janus scaffolds displayed the greatest de ection, the proliferation rate decreased, as compared to their respective counterparts in static culture (0 kHz). At ultrasound stimulation frequency of 20 kHz, a slight increase in the proliferation was measured for cells cultured on PCL and Janus scaffolds, which was more pronounced in PLA materials. Thus, lower de ection amplitudes (higher frequencies) resulted on higher proliferation rates when the materials were analyzed individually. Surprisingly, the cell proliferation rate was not only dependent on the amplitude of the de ection, but also on the material used. The de ection amplitude of PCL at 10 kHz was of 142 nm, comparable to the de ection of PLA when stimulated at 20 kHz, 142 nm. However, the proliferation rate on PCL scaffolds at 10 kHz was 26 ± 8% while a proliferation rate of 408 ± 185% was measured for hBMSCs cultured on PLA scaffolds at 20 kHz (Fig. 2d). Similarly, ultrasound stimulation at 40 kHz, that resulted on the smallest de ection in all the materials, led to an the highest increased cell proliferation in PLA and Janus scaffolds, with cell densities of 4492 ± 341 and 5219 ± 555 cell.cm − 2 and proliferation rates of 519 ± 111, 243 ± 70%, respectively. However, hBMSCs proliferation in stimulated PCL scaffolds was not signi cantly different to its counterpart in static conditions (0 kHz), with a cell density of 1044 ± 97 cell.cm − 2 and a proliferation rate 89 ± 18%, indicating that ultrasound stimulation was not effective in PCL materials. To decipher what was the reason for this differential behavior we further investigated the effect of ultrasound stimulation at 40 kHz on cell matrix deposition and differentiation.
Culture of hBMSCs on the scaffolds for 7 and 14 days under ultrasound stimulation (30 min/day at 40 kHz) resulted in the formation of a bronectin-rich ECM on Janus scaffolds, while cells on PLA scaffolds and on static cultures presented only intracellular expression of the protein ( Fig. 2f and Figure   S11). hBMSCs on PCL deposited bronectin with or without stimulation and populated the scaffold with a reduced cell number, indicating again that ultrasound stimulation was not effective in this particular material. Cells cultured in PLA and stimulated at 40 kHz had a higher proliferation rate, but their capability to deposit bronectin was unaltered. Cells cultured in Janus scaffolds, in the other hand, responded with an increased proliferation rate and matrix deposition when stimulated at 40 kHz as compared to static cultures. We therefore hypothesized that despite the similar de ection amplitude modelled for the different materials, these must had a different response in the pulse of the de ection that affected the cell response, as previously shown 18 .
Janus scaffolds are a combination of a de ecting and a damping material.
To study the potential differences on pulse width and pulse repetition frequency (PRF), we rst measured the wave transmitted by PLA, PCL and Janus scaffolds when stimulated at 40 kHz in liquid media with the use of a hydrophone detector ( Figure S9, S10 and Table S3). A Fast Fourier Transform (FFT) of the recorded data decoupled this response onto a wave with the characteristic frequency of the input signal ( 40 kHz) and a secondary wave (the pulse) that was dependent on the material used. Next, we measured the de ection of the scaffolds upon ultrasound activation at 40 kHz with a nanoindenter (Fig. 3). The tip of the cantilever was approached to the surface and engaged (500 nm indentation), after which the ultrasound was switched on and the response on the de ection of the scaffolds was measured (Fig. 3, indentation). PLA scaffolds de ected with a pulse amplitude of 53 ± 2 nm, a pulse width of 2 cycles (0.49 s) and a PRF of 1.1 Hz. PCL scaffolds, however, barely reacted and behaved as damping materials, with a response that was not pulsed (rather sinusoidal) and with an amplitude of de ection of 4 ± 1 nm. The combination of these two materials, PLA as active and and PCL as damping material, in Janus scaffolds resulted in a reduced amplitude of 36 ± 4 nm and pulse width (1 cycle, 0.1 s), and an increased PRF of 2.17 Hz as compared to PLA scaffolds. The amplitude of the de ection, measured by nanoindentation correlated well with the trend observed on the models showing higher de ection for PLA, followed by Janus and PCL scaffolds. The differences we observed in cell proliferation might be related to the pulse of the scaffold de ection, being the pulse width highest in PLA scaffolds, followed by Janus and PCL scaffolds. It is noteworthy that the PCL response to the applied 40 kHz frequency resulted on a continued rather than pulsed wave. We hypothesized that the continuous mechanical stimulation, although smaller in amplitude might result on a stress to cells rather than a sudden mechanical stimulation, thus reducing (or at least maintaining) the proliferation rate and not affecting the cell matrix deposition.
Janus scaffolds promote osteogenic differentiation of BMSCs under remote ultrasound stimulation.
Ultrasound-mediated external activation of scaffolds has mostly been reduced to the use of piezoelectric composite materials with susceptible particles, or of crystalline piezopolymers that result on an electrical signal 30 . However, the mechanical activation of scaffolds for tissue regeneration purposes via ultrasounds has yet not been investigated. To evaluate the potential of external ultrasound stimulation of 3D printed scaffolds for bone regeneration, we cultured BMSCs on Janus, PCL and PLA scaffolds for 21 days in osteogenic media and under static (0 kHz) and stimulated (40 kHz, 30 min/day) conditions. After 21 days of culture, a dense collagen I network was formed on stimulated Janus scaffolds (Fig. 4a) while stimulated PCL, PLA or any of the static culture conditions showed only intracellular expression. An upregulation of collagen I, RunX2 and osteocalcin was also detected at a gene level on cells cultured on stimulated Janus and PCL materials (Fig. 4b) compared to stimulated PLA or any of the static conditions, with the increase being more pronounced on Janus scaffolds. The cell number was also signi cantly increased on Janus and PLA materials with respect to their counterparts in static culture, in agreement with earlier observations (Fig. 4c and Fig. 2c). Alizarin red staining for Ca 2+ deposits was maximum on PCL and Janus scaffolds in stimulated culture (Fig. 4d). However, a detailed observation of the deposited minerals by SEM revealed that BMSCs cultured on PCL underwent what has been de ned as "chemical differentiation", depositing minerals with a smooth and big crystal-like morphology characteristic of Ca 2+ salts 31 . Minerals deposited on Janus scaffolds, however, showed a rounded and porous morphology characteristic of amorphous hydroxyapatite 31,32 , indicating bone formation potential. Osteocalcin release increased with ultrasound stimulation for cells cultured on Janus and PLA scaffolds but not for PCL, with cells cultured on the last having the lowest release ( Figure S12). Contrary to this, and as shown previously 33 , alkaline phosphatase activity was similar for cells with and without ultrasound stimulation but highest for cells cultured on Janus scaffolds ( Figure S13). ATP release, a key regulator of osteoblast response upon mechanical stimulation, was 4-fold higher in BMSCs cultured on stimulated Janus scaffolds compared to any of the other stimulated or static conditions. ATP release has also been shown to increase upon direct low-intensity ultrasound stimulation, and consequent depolarization of the cell membrane, during osteogenic differentiation of MC3T3-E1 cells and in 2D cultures of mesenchymal stem cells, which is in agreement with what we observed here (Fig. 4e) 33,34 .
Cell membrane depolarization, as a consequence of mechanical stimuli, results in the activation of voltage-gated Ca 2+ ion channels (VGCC) to regulate calcium in ux into the cell. Some reports have shown the presence of L-type VGCC in BMSCs and suggested that they play a pivotal role in cell attachment, proliferation and osteogenic differentiation 35,36 . To determine whether L-VGCCs were involved in the osteogenic differentiation, we evaluated the expression of CACNA1c gene that encodes the Ca v 1.2, a subunit of L-VGCC; we found that CACNA1c was upregulated 3-fold on stimulated Janus scaffolds but not on stimulated PCL or PLA scaffolds (Fig. 5a). L-type VGCC are activated via physical coupling of the Ca v 1.1 subunit of the dehydropyridine receptor (DHPR) to the Ryonodine receptor (RyR) on the endoplasmic reticulum of cells 37 . Staining of DHPR revealed the formation of L-VGCC on cells cultured in PCL and Janus scaffolds and only the coupling of these to RyR on cells cultured on Janus scaffolds ( Fig. 5b and Figure S14), proving the direct effect of mechanical de ection on Janus scaffolds via ultrasound stimulation. Indeed, blocking of L-VGCC with 1 µM nifedipine during the differentiation process resulted in a decreased cell number in all materials and culture conditions, but was more pronounced in stimulated conditions. The difference in cell number between stimulated and static conditions was not signi cant in PCL and PLA scaffolds. However, the cell number measured in stimulated Janus scaffolds was signi cantly lower to their counterparts in static culture. Blocking of L-VGCC also resulted in the downregulation of collagen I, RunX2 and osteocalcin gene expression, resulting in no signi cant differences between stimulated and static culture conditions ( Fig. 5c and d). Thus, when L-VGCC were blocked, ultrasound stimulation no longer in uenced cell differentiation or proliferation, proving their direct correlation.
In summary, we presented two new alternative routes to 4D printing: in-situ phase segregation to control spatially the composition of the printed structure, and ultrasound stimulation to remotely activate the de ection of the scaffolds. Varying the ratio of the polymer blend allowed controlling the phase formation to give rise to particles, ellipsoidal phases or Janus structures that were spontaneously formed during the printing process. The formation of such phase-segregated structures provides additional control for 3D printing strategies, developing the technology towards 4D printing. Particularly, the control over the chemistry could be exploited for the selective functionalization of the different formed phases and thus further control the cell phenotype.
Theoretical models proved the use of the scaffolds feasible as dynamic implants that could be externally activated "on-command". PCL and PLA scaffolds responded to ultrasound as damping and de ecting materials, respectively. Combinations of these as Janus structures resulted in shorter pulse widths and smaller de ection. These properties of the scaffolds directly affected cell proliferation, matrix deposition and osteogenic differentiation of cultured BMSCs. Ultrasound stimulation of PCL scaffolds showed no signi cant effect in cell proliferation or matrix deposition, but a small increase in gene expression of some osteogenic markers and the deposition of Ca 2+ salts, ascribed to a chemical differentiation. Stimulation of PLA scaffolds resulted on an increased cell proliferation, but no effect on cell differentiation. Stimulation of Janus scaffolds affected hBMSCs with increased cell proliferation and expression and deposition of osteogenic markers. Moreover, we showed that L-VGCC were activated on cells cultured in stimulated Janus scaffolds, and that blocking these cancelled the cellular effect of ultrasound stimulation. Taken altogether, we propose that remote activation of Janus scaffolds could present an ideal alternative to traditional "static" implants, providing "on-command" stimulation of cells.

Methods
Fabrication of scaffolds PCL and PLA were purchased from Sigma-Aldrich and NatureWorks, both with a molecular weight of 80 kD. The two polymers were blended at the de ned ratios using a twin-extruder working at 150 °C and 100 rpm. The laments were collected, chopped down to approximately 5 mm height and used to feed a fused-deposition modelling Bioscaffolder SYSENG at a feed rate of 500 mm/min and a dispensing volume of 30 rpm. The structure had a length of 30 mm, a strand distance of 0.6 mm and a layer thickness of 0.25 mm using a needle of 0.4 mm inner diameter. For cell culture experiments, singlelayered scaffolds were glued from the extremes to petri dishes with Norland optical adhesive 68 and cured by 30 s of 395 nm UV exposure.

Imaging of phase segregation by SEM, TEM and POM
For SEM and polarized optical microscopy (POM), scaffolds were cut with a razor blade perpendicularly to the sample height and directly imaged with a Nikon Ti-S with crossed polarizers. Samples for SEM were sputter-coated with gold (Cresignton 108 Auto) and imaged on a FEI/Philips XL-30 microscope under both secondary and backscattered electrons with an acceleration voltage of 10 keV. For TEM, samples were embedded on Epon LX112 resin (Hexion) and slowly cured at RT for 21 days. Blocks were then sliced on an ultra-cryomicrotome Leica EM FC6 at temperatures of -100 °C and -45 °C for the chamber and for the knife, respectively. The samples were collected on TEM carbon-supported Cu grids from the 60% DMSO bath. The grids were imaged on a TEM FEI/Tecnai G2 Spirit BioTWIN at a typical acceleration of 80 keV.

Computational models of scaffold de ection
Simulations of the scaffolds' de ection under different sonication frequencies was performed using COMSOL Multiphysics modelling software to compute the pressure variation for propagation of acoustic waves in the scaffolds when submerged on a cylinder of aqueous media at quiescent background conditions. As input parameters, the density of the materials and the calculated exural modulus were used. The applied wave sounds were circular, with an intensity of 60 Pa and at de ned frequencies of 10, 20 or 40 kHz.

Ultrasound stimulation and detection set-up
A Kemo M048N ultrasound generator (Kemo Electronic) with a frequency range of 2-40 kHz was connected to four Kemo P5123 mini piezoelectric tweeters (Kemo Electronic) with frequency range of 2.5-45 kHz and a pressure of 118 dB. The petri dishes containing the scaffolds were place on top of the loud speakers that were mounted on at surfaces. To characterize the waves after transmission through the petri dishes (empty controls and containing scaffolds) and in aqueous media, a 0.2-mm needle hydrophone (NH0200, Precision Acoustics) was connected to a hydrophone preampli er and a DC coupler that were then connected to a Rigol DS1102 oscilloscope.

Measurement of scaffold de ection
The de ection of the scaffolds was measured in a Piuma nanoindenter (Optics11) with a MN2-TN1-ST probe of 0.5 N/m spring constant (Optics11) and in aqueous media using the same sonication set-up described above. The samples were visualized with the built-in camera, and the cantilever probe was brought to the center of one of the scaffold bers. After engaging, an indentation maximum 500 nm was applied and held for 10 s. Immediately after indentation, an ultrasound frequency of 40 kHz was applied.
Cell culture.
BMSC were isolated from a 22-year-old male donor by aspiration at Texas A&M Health Science Center after ethical approval from the local and national authorities and written consent from the donor.
Cryopreserved vials at passage 2 were plated at a density of 1,000 cells/cm 2 and cultured in basal media consisting of alpha-MEM media supplemented with Glutamax (Gibco) and 10% fetal bovine serum (FBS) (Sigma-Aldrich).
Cell culture in 3D printed scaffolds under sonic stimulation.
3D printed scaffolds were sterilized in 70% ethanol, rinsed with PBS and coated for 1 h with human recombinant vitronectin (Thermo Fisher) at a surface density of 1 ug/cm -2 and assuming 3 cm 2 per sample.
BMSCs were seeded at a density of 15000 cell/cm 2 from cell-concentrated dispersions of 1.5·10 6

Competing interests
The authors declare no competing interests.