Tough PEG‐only hydrogels with complex 3D structure enabled by digital light processing of “all‐PEG” resins

Digital light processing (DLP) of structurally complex poly(ethylene glycol) (PEG) hydrogels with high mechanical toughness represents a long‐standing challenge in the field of 3D printing. Here, we report a 3D printing approach for the high‐resolution manufacturing of structurally complex and mechanically strong PEG hydrogels via heat‐assisted DLP. Instead of using aqueous solutions of photo‐crosslinkable monomers, PEG macromonomer melts were first printed in the absence of water, resulting in bulk PEG networks. Then, post‐printing swelling of the printed networks was achieved in water, producing high‐fidelity 3D hydrogels with complex structures. By employing a dual‐macromonomer resin containing a PEG‐based four‐arm macrophotoinitiator, “all‐PEG” hydrogel constructs were produced with compressive toughness up to 1.3 MJ m−3. By this approach, porous 3D hydrogel scaffolds with trabecular‐like architecture were fabricated, and the scaffold surface supported cell attachment and the formation of a monolayer mimicking bone‐lining cells. This study highlights the promises of heat‐assisted DLP of PEG photopolymers for hydrogel fabrication, which may accelerate the development of 3D tissue‐like constructs for regenerative medicine.


INTRODUCTION
Hydrogels are three-dimensional (3D) polymer networks that contain a large amount of water. [1]Owing to their tunable physicochemical properties and, for some aspects, similarity to native extracellular matrices, hydrogels are widely used in biomedical applications for the preparation of 3D scaffolds in tissue regeneration, [2][3][4] drug-delivery vehicles, [5] soft robots, [6] and microfluidic devices. [7]For these applications, it is often important to create hydrogels with complex and detailed structures, especially in the case of tissue scaffolds and implants with customized geometries. [8]In recent years, 3D printing has emerged as a powerful tool for hydrogel fabrication, [9][10][11][12][13][14][15][16][17] with extrusion-based printing technologies most commonly used. [4,18,19]With shear-thinning inks, extrusion printing allows the fabrication of objects mimicking natural structures with controlled geometry and characteristics.Nevertheless, the extrusion technique is limited by low printing resolution (>200 μm) and poor structural fidelity. [8,20,21][24][25][26][27] Compared to SLA, DLP allows the faster fabrication of objects with complex geometries using a digital micromirror device.However, as opposed to water-free networks, the printing of hydrogels by DLP is challenging.This is attributed to a variety of factors.First, as the most widely used vat photopolymerization technique, DLP requires a liquid resin of low viscosity (usually less than 10 Pa s) to enable layer-by-layer stacking. [28]Consequently, existing water-borne resins for DLP printing usually have high water content (typically >70%), which results in low mechanical strength and toughness of the printed constructs.Second, the solidification of the DLP resin occurs through irradiation from the underneath of the resin, and the print platform moves vertically in order to allow the resin to flow back.The separation of every newly formed layer from the bottom plate subjects the structure to certain mechanical forces. [22,29]herefore, the crosslinked layers should have sufficient mechanical strength to remain intact during this process, and preserve the shape fidelity of the printed object. [30]As a result, an ideal DLP resin should have low viscosity and provide high mechanical resistance of the print, which is highly challenging to achieve with common water-based resins, especially for high molecular weight (MW) macromonomers.Lastly, light scattering generated from the photopolymerized layer that is not optically clear can further affect the printing resolution and shape fidelity, with the poly(ethylene glycol) (PEG) hydrogels as the typical examples. [31]][38][39][40][41] The SLA/DLP printing of PEG hydrogels is usually limited to low MW monomers (e.g., PEGDA 700), which generate brittle products.Although it is possible to print high-quality embedded channels or simple shapes using aqueous solutions of PEG (macro)monomers with higher MW (e.g., 6000 g mol −1 ), [40][41][42] high water content and low crosslinking degree result in low mechanical strength and make fabrication of free-standing PEG hydrogels with complex geometries difficult. [8,43,44]Therefore, high-resolution 3D printing of PEG hydrogels with potential for practical applications remains challenging.
[49] Even shape-memory thermoplastic photopolymers that are solid at room temperature have been successfully 3D printed by this technique. [50,51]To address the challenges in PEG hydrogel printing, we propose here direct heat-assisted DLP of macrophotoinitiator-based "all-PEG" resins that are not printable at room temperature, and the subsequent transformation into hydrogels by simple post-printing swelling in water (Figure 1).This strategy can circumvent the limitations of conventional 3D printing based on aqueous resins, in particular the weak mechanical properties of the final objects and light scattering from the crosslinked layers.High-resolution fabrication of elastic yet tough PEG hydrogels with complex geometries, excellent shape fidelity, and tunable mechanical strength can be realized by combining two PEG macromonomers of different MWs with a four-arm PEG-based macrophotoinitiator.Finally, the "all-PEG" resin was fabricated into bioactive bone-mimicking hydrogel scaffolds by post-printing surface functionalization.

Design and 3D printing of "All-PEG" resins
We selected methacrylate as the photo-crosslinking group due to its high crosslinking efficiency, convenient functionalization, and significantly lower cytotoxicity compared to the acrylate analog. [51]To tune the chain length of PEG macromonomers, we first synthesized a series of PEGDMAs by methacrylation of commercial PEGs with MWs of 2000, 4000, 6000, 8000, 10,000, and 20,000 g mol −1 .The functionalization degree of the resulting PEGDMAs 2000, 4000, 6000, 8000, 10,000, and 20,000 was 71%-93% based on 1 H-NMR spectroscopy (Figure 1A and Figure S1, Table S1).
To enable the water-free DLP printing of "all-PEG" resins, we synthesized a star-shaped PEG-based macrophotoinitiator that can replace conventional hydrophobic (e.g., bisacylphosphine oxide, BAPO), or water-soluble photoinitiators (e.g., lithium phenyl-2,4,6-trimethylbenzoyl phosphinate, LAP) (Figure 1B).54][55] The structure of the obtained 4♦-BAPO-PEG was confirmed by 1 H-and 31 P-NMR spectroscopies (Figure 1C and Figure S2).A customized DLP printer with temperature control of the resin tray and printing head (Figure 1D) [47] allowed for direct printing of "all-PEG" resins at elevated temperature (e.g., 90 • C).Therefore, the performance of the 3D-printed products could be evaluated without the interference of additives, especially reactive diluents (e.g., N-vinylpyrrolidone [NVP]) that are often used in SLA/DLP to reduce viscosity and dissolve the commercial photoinitiator. [56]nder sonication at 70 • C, 4♦-BAPO-PEG was easily mixed with the semicrystalline macromonomers (Figure 2A,B) and a UV absorber (Sudan I, 0.03 wt%) without the need for additional solvents.To evaluate the 3D printability of the PEG macromonomers, we first measured the viscosity of the polymers between 70 • C and 100 • C. The commercial PEGDA 700 was selected as a reference, as it has been widely used in 3D printing of PEG hydrogels.At 90 • C, the viscosity increased from 10 to 20,000 mPa s with increasing the PEG MW from 700 to 20,000 g mol −1 (Figure 2C).The viscosity of PEGDMA 2000-10,000 was in the printable range of DLP at 70 • C-100 • C, while it would be not possible to print PEGDMA 20,000 without the addition of diluents.We first printed the PEGDMA 6000 by heat-assisted DLP using 4♦-BAPO-PEG as the photoinitiator (2.0 wt%), and good printability was achieved, as shown in Figure S3H with ETH logo and a stent prototype.Given that the PEG macromonomers are semicrystalline with a melting point (T m ) in the range of 40 • C-60 • C (Figure 2B), their 3D-printed networks offered shape-memory properties with thermally triggered efficient shape recovery (Figure S3 and Videos S1-S3).
We then printed the other PEG macromonomers by heat-assisted DLP using 2.0 wt% 4♦-BAPO-PEG, and all of them displayed good printability.The high MW PEG networks exhibited enhanced mechanical properties, as demonstrated by their increased tensile stress (approximately 10 MPa) and Young's modulus (approximately 200 MPa) when using PEGDMA with a molecular weight greater than 4000 g mol −1 (Figure S3).This improvement can be attributed to their increased crystallinity, ranging from 64% to 80%.In contrast, lower crystallinity (38%-57%) was observed for PEGDMA 2000, and PEGDA 700 displayed no crystalline domains (Table S2).However, the PEGDMA 10,000 network did not exhibit a further increase in yielding stress (Figure S4).This is likely due to that as the PEG polymer chain length increased, the degree of crosslinking within the PEG networks declined.Note that the 3D-printed products from PEGDA 700 were very brittle and weak, similar to that printed from PEGDMA 2000 (Figure S3B).The good printability is the prerequisite for producing highresolution hydrogels in next step.We further examined the fidelity of the 3D printing of PEG networks, which demonstrated high dimensional stability and minimal size variations across all three dimensions (mostly <5%) (Figure S5 and Table S3).

Mechanical properties of 3D-printed single-macromonomer hydrogels
The PEG hydrogels were then prepared by incubating the 3Dprinted PEG constructs in PBS pH 7.4 at 37 • C for 24 h to reach swelling equilibrium.As shown in Figure 2D,E, the swelling ratio increased from 150% to about 500% when the PEG MW increased from 2000 to 10,000 g mol -1 .This was confirmed by the change in size of the swollen samples.Subsequently, their mechanical properties were studied by performing tensile tests.An increase in elasticity was clearly observed when raising the MW of PEG.The Young's modulus (E t ) decreased from more than 30 MPa to less than 1 MPa, while the elongation at break (ε break,t ) increased from about 5% to around 70%, when PEG MW was increased from 700 to 10,000 g mol −1 (Figure 2F-I and Table S4).This is associated with the higher flexibility of long PEG chains as well as with the lower crosslinking density, and thus higher water uptake of hydrogels. [41]A similar  (F-I) Tensile stress-strain curves (F) of hydrogels prepared from 3D-printed specimens with PEG macromonomers, and the corresponding Young's modulus (G), maximum tensile stress (H) and elongation at break (I).The tensile mechanical properties are expressed as mean + SD (n = 3-6), except for the very weak PEGDMA 10,000 hydrogels (duplicates).(J and K) Compression stress-strain curves (J) of the hydrogel specimens and the corresponding compression modulus (K).The compression mechanical properties are expressed as mean + SD (n = 2-3), given that nearly identical stress-strain curves were obtained.
phenomenon was observed in compression tests (Figure 2J,K and Table S4).PEGDMA 10,000 hydrogel was very weak, likely due to its long polymer chain and relatively low methacrylation degree (74% vs. 82%-93% for other hydrogels).This is further supported by the significantly higher average chain length between crosslinks [57,58] for PEGDMA 10,000 (M c of 1860 g mol −1 ), compared to the other hydrogels (M c of 43-1048 g mol −1 , as shown in Table S4).In addition, The M c values for PEGDMA 6000 and 8000 hydrogels were found to be very similar (1048 vs. 998 g mol −1 ), which can be attributed to the interplay between the PEG chain length and the degree of methacrylation.This similarity in M c values explains why these hydrogels exhibited similar mechanical properties (Figure 2F-K).
To compare our approach with molding or conventional DLP printing, we prepared PEG hydrogels using aqueous resins with water-soluble LAP as the photoinitiator.By molding with PEGDMA 6000 at 70% water content via photo-crosslinking, the equilibrium swelling ratio of the obtained sample was 570%, which was much higher than that via heat-assisted DLP (approximately 360%).As a result, the former hydrogel showed Young's modulus of only 0.4 MPa, while the value for heat-assisted DLP printed hydrogel was 1.2 MPa (Figure S6).
We then printed PEG hydrogels using aqueous resin (80 wt% water content) by DLP at room temperature.To enhance the printability, NVP (8.0 wt%) was added to the resin, otherwise the curing could not occur (Figure S7A).As expected, the swelling ratio of the PEGDMA 10,000 hydrogels was much higher than that obtained by "waterfree" 3D printing (500%) and reached 1400% (Figure S7C).Clearly, the "water-free" approach may prevent the overswelling of the hydrogels that can negatively impact the mechanical strength.This is likely due to the much denser polymer networks formed by the direct crosslinking of the PEG macromonomers in the absence of large amounts of water molecules.This could be beneficial for biomedical applications as the hydrogels prepared from water-borne resins inevitably further swell under physiological conditions as a result of the difference in osmotic pressure, which deteriorates their mechanical properties. [1]In addition, the "waterfree" 3D printing avoids the aforementioned light scattering and layer separation issues that are commonly encountered in SLA/DLP-printed PEG hydrogels, [29][30][31] resulting in lowresolution prints (Figure S7B).Note that the swelling process is homogeneous in all dimensions, as evidenced by the size of 3D-printed cuboid structures (Figures S8 and S9).This highlights the high shape fidelity of the 3D hydrogels enabled by heat-assisted DLP.
To evaluate the hydrogel printability of our approach with a commercial photoinitiator, we further fabricated the PEG hydrogels using BAPO (Omnicure 819, formerly known as Irgacure 819), which is one of the most efficient molecular photoinitiators in DLP/SLA 3D printing. [56,59,60]In this case, BAPO was first dissolved in NVP (8.0 wt%) for resin preparation.A similar trend in MW-dependent variation of mechanical properties was observed in the 3D-printed PEG networks (Figure S3) and the corresponding hydrogels (Figure S10 and Table S5).Interestingly, PEGDMA 10,000 hydrogel showed much higher mechanical strength with BAPO in the presence of NVP than that with 4♦-BAPO-PEG (Figure S11).This can be explained by the fact that NVP can promote the photopolymerization and strengthen the network. [61]

Toughening of 3D-printed hydrogels using a dual-macromonomer resin formulation
Although heat-assisted DLP allowed the fabrication of PEG hydrogels with much stronger mechanical properties compared to traditional molding or "with-water" 3D printing, their maximal stress remained quite low for the most of elastic ones based on high-MW PEG (e.g., 0.6-0.7 MPa for PEGDMA 8000).To improve their mechanical properties, we employed a "dual-macromonomer" strategy [47] to combine the chain flexibility of high-MW PEG and the high amount of crosslinking sites of low-MW PEG.A series of resins containing varying ratios of PEGDMA 10,000 and PEGDA 700 were formulated, which showed reduced viscosity compared to that of PEGDMA 10,000 (Figure S12).Prior to printing, we performed a photo-rheological study of the dual-macromonomer resin at 70 • C using PEGDMA 10,000-PEGDA 700 (90/10, w/w) with single PEGDMA 8000 resin as the reference.Figure S13 shows the in situ photo-rheological data at 70 • C that mimic the heat-assisted DLP printing process.The resins were melted on top of the glass plate of the photo-rheometer setup at 70 • C.After 1 min of time sweep without light, UV-365 light was applied to induce in situ photo-crosslinking for 9 min.The modulus values were measured by time-lapsed small-strain oscillatory deformation.The results show that the two resins exhibit comparable photoreactivity and yielded networks with a storage modulus (G') of ca. 1 × 10 6 Pa.
Subsequently, all dual-macromonomer resins were printed into hydrogels after swelling in PBS (Figure 3A).The swelling ratio was still high enough to keep the water content more than 70% with the addition of PEGDA 700 up to 20 wt% (Figure 3B).As shown in Figure 3C-F, PEGDA 700 greatly improved the mechanical strength of the 3D-printed hydrogels, with maximal tensile stress up to approximately 1.5 MPa.With 10 wt% of PEGDA 700, the Young's modulus of the hydrogel was about 2.4 MPa, while the elongation at break was approximately 50%.On the other hand, the compression modulus (E c ) reached 3.2 MPa (Figure 3G,H) and the compressive toughness was nearly 1.0 MJ m -3 , which is two to four-folds higher than that of single-PEGDMA hydrogels with relatively high elasticity made from PEGDMA 4000-8000 (Figure 3I).As a result, the dual-macromonomer resin afforded relatively strong and tough "all-PEG" hydrogels via heat-assisted DLP (Figure 3K).64][65][66][67][68][69][70] The increase of PEGDA 700 fraction to 20 wt% generated an even higher compressive toughness (1.3 MJ m -3 ), but also reduced the elongation at break to 40% (Figure 3F).At 30 wt%, the objects became brittle similar to the PEGDMA 2000 hydrogel, due to the higher ratio of short network units, [47] with molar fraction of PEGDA 700 exceeding 80%.To further confirm the improvement of the mechanical properties brought by bimodal PEG networks, we compared the Young's modulus-polymer content relationships of the 3Dprinted products from different compositions, because it is known that the Young's modulus of hydrogels is highly dependent on the polymer content after swelling. [71]As shown in Figure 3J, the E t of the dual-macromonomer hydrogels as a function of crosslinked polymer content showed a strong positive correlation, with a slope higher than that obtained for single-macromonomer hydrogels (prepared with 4♦-BAPO-PEG or BAPO in the presence of NVP), which confirmed the improvement of the mechanical properties created by the bimodal network structure.Altogether, the dual-macromonomer resin combined high strength of short PEG chains with good flexibility of long PEG chains, and thus enabled the fabrication of PEG hydrogels with elastomeric properties.
We further characterized the hydrogel networks produced by heat-assisted DLP by scanning electron microscopy (SEM; Figure S14).As shown in cross-sectional SEM images, no porous structure was observed in the 3D-printed PEGDMA 8000 specimen before swelling (Figure S14A).After swelling in PBS for 24 h and subsequent lyophilization, a microporous structure with elongated pores was obtained (Figure S14B-D).Although the pore sizes (40-125 μm for most) were comparable to those of conventional PEG hydrogels, [10,27] the solid domain seemed much thicker.This is likely due to the much denser networks achieved by the direct 3D printing of PEG photopolymers.Similar phenomenon was observed in the hydrogels fabricated from PEGDMA 10,000-PEGDA 700 (90/10, w/w), which showed heterogeneous porosity (Figure S14F-H).This should be due to bimodal network based on the use of two macromonomers with highly different polymer chain lengths, in addition to the impact from the lyophilization process.

3D printing of structurally complex hydrogels
To further test the printability of our approach, we produced PEG hydrogels with different complex structures by heat-assisted DLP (Figure 4).We first printed 3D hydrogel models of the human meniscus and ear, which are representative cartilage-based structures. [71]Both of them displayed high fidelity and a smooth surface (Figure 4A,B).Furthermore, complex objects with gyroid structure or Dprime surface were 3D printed and fabricated into hydrogels (Figure 4C-G).No shape deformation that commonly occurs in conventional hydrogel printing was observed.The 3Dprinted structure with D-prime surface was characterized by both optical microscopy and scanning electronic microscopy (SEM) before swelling.An average layer thickness of the object printed from PEGDMA 10,000-PEGDA 700 (90/10, w/w) before swelling was about 45-55 μm, which is very close to the set value (50 μm).After swelling, the hydrogel layer thickness increased to 75-85 μm, with a size expanding ratio of approximately 1.6 (Figure 4E,G).Based on our observations and calculations, we have found no substantial difference in size expansion between the x+y axes and the z-axis.This is consistent with the results obtained in the swelling test of the cylinder and cuboid specimens (Figure 2D and Figure S9).If needed, the layer thickness could be set to lower values (e.g., 25 μm) to further improve the zresolution.The minimal horizontal feature size was about 160 μm (Figure 4F).Thanks to the "all-PEG" resin design, free-standing PEG hydrogels with complex structures can be obtained, which are neither brittle as PEGDA 700 hydrogels nor weak as long PEG-based "with water" hydrogels (Table S6).
We further fabricated small-sized complex architectures using a trabecular model (ϕ 3.20 mm × h 2.03 mm) with microchannels obtained by the computed tomography (CT) scan of a trabecular bone core from a horse femur as described previously (Figure 5A). [69]It was found that the microstructure of the bone scaffold was effectively printed, with the majority of structural features accurately reproduced.The minimal feature size was determined to be 50-60 μm (Figures S15 and S16).The average print size deviation was calculated to be ϕ −7.3% × h −0.3%.In addition, the 3D-printed structures also showed high reproducibility regardless of their size (Figure S17).After equilibrium swelling, the hydrogel model showed excellent surface quality, and the minimal feature size was determined to be 80-90 μm (Figure 5B), as well as homogeneous size expansion in three dimensions.Such high-resolution complex hydrogels would be hard to obtain by extrusion 3D printing or conventional "with-water" vat photopolymerization, especially when the water content is relatively high (>70%).
[74][75][76] Therefore, we tested the printing of PEGDMA 8000 at 50% water content on a tomographic volumetric printer (Video S4).Prior to the printing, the photo-crosslinking efficiency of PEGDMA water-borne resins was evaluated by in situ photo-rheology under UV irradiation at 365 nm (Figure S18).Although volumetric printing is much faster, when compared to heat-assisted DLP, it was not possible to achieve good printing quality of the bone model.The model was further modified to expand the size of the microchannel by approximately four folds to eventually enable the printing, as shown in Figure 5C.This test further confirmed the great advantage of heat-assisted DLP of "all-PEG" resins for high-resolution 3D printing of structurally complex hydrogels.

Cell seeding on 3D-printed bone-like hydrogel scaffold
To explore the potential of our 3D-printing approach in biomedical applications, we tested the in vitro cytotoxicity of the 3D-printed PEG networks using different resin formulations using 4♦-BAPO-PEG.The 3D-printed discs were incubated with A549 cells for 48 h using Transwell inserts, and the cell viability was determined by the MTS assay and compared to negative control.As shown in Figure 5D, approximately 100% cell viability was observed for the hydrogel disks, indicating the cytocompatibility of the tested PEG networks.
Motivated by these results, we further produced the bone-mimicking hydrogel scaffolds printed from PEGDMA 10,000-PEGDA 700 (90/10, w/w), and evaluated their bioactivity for tissue engineering applications (Figure 5E).The scaffolds were scanned by micro-CT at high resolution (17 μm).A biomimetic porous architecture (Figure 5F) with a pore size in the range of 100-500 μm was achieved.We reasoned that a small amount of the residual methacrylate groups (5%-10%) on a pre-formed hydrogel should be enough for covalent fixation of cell adhesive motifs, despite the high double-bond conversion of PEGDMA (Figures S19 and S20).To this, fibronectin-derived cysteinecontaining arginylglycylaspartic acid (RGD) peptides (N-C: CGRGDS) were conjugated to the scaffold surface via Michael addition between thiols and the residual methacrylates at a concentration of 10 mM (Figure S21), using a reported method. [77]After surface functionalization, the scaffolds were thoroughly washed to remove unreacted RGD, as soluble RGD has been shown to inhibit cell adhesion. [78]or cell-seeding experiments, human mesenchymal stem cells (hMSC) were selected as the precursor for in vitro bone tissue formation. [79,80]However, initial attempts for direct hMSC seeding atop PEG scaffolds were unsuccessful as the cells rapidly sedimented to the bottom through the pore space due to gravity.To better control the spatial distribution of cells in the scaffolds, we used type I collagen hydrogel as a temporal supportive matrix.Notably, we found that an optimal concentration of collagen is the key to efficient cell penetration and spatial distribution within the scaffold.A cell seeding using a soft collagen (1.0 mg mL -1 ) enabled efficient cell penetration into the pore space.Initially, cells were embedded in a 3D environment.Over time, they migrated out and attached to the scaffold surface (Figure 5G,H).Moreover, increasing the cell-seeding density from 0.7 to 3 Mio mL -1 promoted faster migration of the cells from the collagen matrix.A live-dead assay confirmed that the cells after seeding at such high density were highly viable at Day 7 (95.3%± 4.4%, Figure 5H), indicating excellent cell-compatibility of the scaffolds.Some of the cells spread on the scaffold surface, exhibited osteoblast-like morphologies, and lined up to form a monolayer that mimics bone-lining cells (Figure 5I).This was also observed in the confocal microscopic image of seeded cells stained for actin and nuclei (Figure 5J).The lining cell-like morphologies can be attributed to the higher mechanical stiffness of the printed PEG matrices than the collagen matrix.

CONCLUSION
We introduced "all-PEG" macrophotoinitiator-based DLP resins for the 3D printing of geometrically complex PEGbased networks and the subsequent hydrogels after swelling.Using BAPO-conjugated four-arm PEG as the photoinitiator, a series of PEG macromonomers with MW ranging from 2000 to 20,000 g mol -1 were printed by DLP at elevated temperature.Allowing easy vertical detachment of the crosslinked layer from the vat film in DLP, and avoiding the use of aqueous resins that can cause strong light-scattering, this approach enabled robust 3D printing of PEG hydrogels with high resolution and shape fidelity.Importantly, a dual-macromonomer strategy allowed the 3D printing of mechanically tough hydrogels by combining high-and low-MW PEG macromonomers.Using this strategy, tough cytocompatible hydrogels with bone-like structures were fabricated and seeded with cells for potential application in tissue engineering.
We envisage that PEG-based biodegradable hydrogels may also be obtained by heat-assisted DLP using various PEGpolyester block photopolymers, [81][82][83] which would provide even stronger mechanical properties.Moreover, the 3Dprinted PEG hydrogels may be further toughened by introducing a secondary network post-printing, depending on the specific application.In principle, most photo-crosslinkable synthetic or natural polymers with a suitable meting point (e.g., 40 • C-70 • C) would be suitable for hydrogel fabrication using heat-assisted DLP, which may greatly expand the scope of hydrogel additive manufacturing.Our approach brings new perspectives for the high-resolution 3D printing of geometrically complex PEG hydrogels, which may advance the fabrication of personalized bioactive scaffolds and medical implants.

Polymer synthesis
Synthesis of PEGDMA [84] : PEG polymers were first dried under vacuum at 70 • C overnight prior to the reaction.Take PEGDMA 6000 as a representative example: PEG 6000 (60.0 g, 0.01 mol) was dissolved in 200 mL anhydrous DCM in a round-bottom flask, followed by the addition of Et 3 N (4.2 mL, 0.03 mol).The solution was purged with nitrogen for 15 min.Next, methacryloyl chloride (3.0 mL, 0.03 mol) was added dropwise to the solution under stirring and cooling in an ice bath.The ice bath was removed after the heating discontinued and the reaction was running for 3 days at room temperature.Afterwards, the reaction mixture was filtered and the supernatant was purified by alumina column.With vitamin E (0.18 g) added to prevent premature crosslinking, the solution was concentrated and precipitated in diethyl ether twice.After drying under vacuum overnight, 51.0 g of a white powder was obtained.The methacrylation conversion was determined to be 82% by 1 H-NMR spectroscopy.In some cases (e.g., PEGDMA 2000 and 4000), extraction was used instead of alumina column purification.

Synthesis of 4♦-BAPO-PEG:
The four-arm PEG 10,000 was first methacrylated to PEG 10,000 tetramethacrylate (PEGTMA) using the same method as for PEGDMA.The four-arm PEGTMA (4.0 g, 0.4 mmol) was dissolved in 30 mL ethanol in a Schlenk flask, followed by the addition of bis(mesitoyl)phosphane (BAP-H, 0.63 g, 1.9 mmol) and KOtBu (0.02 g, 0.18 mmol).The phospha-Michael addition reaction was running for 5 days at 60 • C under argon.The reaction mixture was neutralized with 2 M HCl solution and oxidized with 0.6 mL tert-butyl hydroperoxide (TBHP) solution (5.5 M in decane) at room temperature.From this stage onwards, all following steps were conducted under exclusion of light.The mixture was purified by alumina column using DCM as an eluent.After drying under vacuum, 3.5 g of light-yellow powder was obtained.

4.3
Polymer characterization 1 H-NMR spectra were recorded on Bruker AV400 spectrometer at 400 Hz using CDCl 3 as a solvent. 31P-NMR spectrum was recorded on Bruker 300 spectrometer operating at 300 MHz, with chemical shifts δ were measured according to IUPAC and given in parts per million (ppm) relative to H 3 PO 4 .Differential scanning calorimetry (DSC) analysis was performed using TA Q200 DSC (TA Instruments-Waters LLC).The samples (ca. 10 mg) were placed on Tzero hermetic pans (TA Instruments-Waters LLC) and exposed to heat-cool-heat cycles from −80 • C to 200 • C under nitrogen flow (50 mL min −1 ) using heating and cooling rates of 10 • C min −1 .Data were analyzed using TA Instruments Universal Analysis 2000 software (5.5.3).Fourier-transform infrared (FTIR) spectra were recorded on a Perkin-Elmer Spectrum 65 (Perkin-Elmer Corporation) in transmission mode in the range of 600-4000 cm −1 .Viscosity measurements were performed using HAAKE RheoStress 600 rotational rheometer (Thermo Electron Corporation) with cone and plate geometry (35 mm/2 • ).Viscosity was determined at a shear rate of 100 s −1 in the temperature range of 70 • C-100 • C, applying temperature ramp of 0.05 • C s −1 or −0.05 • C s −1 , with Thermogap function enabled.

DLP 3D printing
The STL files for 3D printing were downloaded from Thingiverse and Allevi3d or designed with Tinkercad from Autodesk, Inc.A commercial DLP 3D printer (Asiga PICO2) comprising the LED light source of 405 nm with customized resin tray and printing head with heating functions [47] was used to fabricate all the objects.Post-curing of the 3D-printed products was conducted using a UV chamber from CL-1000 Ultraviolet Crosslinker from UVP (Ultra-Violet Products Ltd.).DLP resins were prepared by mixing the macromonomers (PEGDMA or PEGDA), 4♦-BAPO-PEG (2.0 wt%) or BAPO (1.0 wt%), Sudan I (0.03 wt%), and vitamin E (0.3 wt%).The resins were sonicated at 70 • C until homogenous mixture was obtained.The printing was performed at temperature of 90 • C (40 • C for PEGDA 700), with layer thickness of 50 μm and exposure time of 3-6 s.As it is difficult to directly measure the curing depth of solid PEG photopolymers on a DLP printer, we relied on our previous experience with DLP printing of biodegradable photopolymers based on four-arm poly(CL-co-DLLA) MA (1k-15 kDa). [47]This experience informed our selection of an exposure time range of 3-4 s.However, as single-PEG resins with higher molecular weight PEGDMA (especially 10,000) have lower crosslinking degree, we increased the exposure time to 6 s.For dual-macromonomer PEG resins, we opted for a printing exposure time of 4 s.The light intensity of the printer LED was 25.67 mW cm -2 .After the printing, the printed objects were cleaned in acetone and 2-propanol, and then cured in the UV chamber for 15 min.In the case of "with-water" printing, LAP (1.0 wt%) was used as the photoinitiator, and NVP (8.0 wt%) was used to dissolve Sudan I before adding it to the photopolymers.

In situ photo-rheology
Rheological measurements of the resins were performed on a modular photo-rheometer MCR 302 (Anton Paar) using a parallel plate with 20-mm diameter.To mimic the volumetric printing conditions, the temperature was set to 4 • C.During time sweep measurements with an interval of 6 s for 5 min, UV curing of the resins was induced after 60 s by illumination with an UV-LED lamp (Thorlabs, λ = 365 nm, light intensity = 10 mW cm -2 ) with 45 μL of resin and the gap set to 0.1 mm.The storage moduli (G′) and loss moduli (G″) were recorded to assess the crosslinking ability and terminal stiffness of the used resins.To prevent drying, wet tissue paper was placed within the temperature chamber.For the photo-rheology of water-free resins, the resins were first melted on top of the glass plate of the photo-rheometer setup at 70 • C.After 1 min of time sweep without light, UV-365 light was applied to induce in situ photo-crosslinking for 9 min.

Curing test
The 3D printing resin was irradiated with a round spot shape LED light (405 nm) on the 3D printer with different exposure times ranged from 2 to 20 s, and the thickness of the crosslinked layer was measured using a caliper.Penetration depth (D p ) was calculated according to the Jacobs' equation based on Beer-Lambert law [85] (Equation 1): where C d is the depth/thickness of cured resin, E 0 is the energy of light at the surface, and E c is the "critical" energy required to initiate polymerization.A semilog plot of C d versus E 0 produces a straight-line curve with a slope of D p and an x-intercept of E c .Exposure time was chosen based on D p and E c related to the desired part properties.

Post-printing swelling
Hydrogel swelling was conducted in PBS pH 7.4 for 24 h using an incubator shaker (200 rpm, 37 • C) from Infors AG.The swelling ratio was calculated according to Equation (2): where wt 0 is the weight of the dry sample and wt 1 is the weight of the sample reaching equilibrium after 24 h of swelling.The water content was calculated according to Equation (3):

Determination of number average molecular weight between crosslinks (M c )
The hydrogel swelling depends on the number of effective chains per unit volume V e according to Equation (4) [57,58] : where  is the specific volume of bulk PEG in the amorphous state (0.893 cm 3 g −1 ), V 1 is the molar volume of water (18 cm 3 mol −1 ), and V 2,r and V 2,s are the polymer volume fraction of the hydrogel in the relaxed and swollen state, respectively, V 2,r is 1 in this work due to the direct printing, and V 2,s was calculated using the dry weight of the 3D-printed network and the water content of the hydrogel.The value of the Flory-Huggins polymer-solvent interaction parameter (μ) used was 0.426, [86] as determined by Merrill et al.From V e and the number average MW of PEG (M n(0) ), the number average molecular weight between crosslinks (M c ) can be determined using Equation ( 5) [87] : The V 2,s can be determined using Equation ( 6): where polymer density () for PEGDA and PEGDMA is 1.12 and 1.11 g cm −3 , respectively, and the density of water ( H 2 O ) is 1.0 g cm −3 .Q is swelling ratio.

SEM imaging
The original 3D-printed samples for SEM images were imaged following printing on their surfaces without further processing.The hydrogel samples were prepared by lyophilization of the dog-bone shape tensile hydrogel specimens or cylinder compression hydrogel specimens, followed by peeling the surface of the samples.The samples were mounted on a carbon tape.Images were obtained using a Quanta 200F environmental SEM (FEI, Oregon, USA) at high vacuum, a secondary electrons detector, and an accelerating voltage of 10 kV.

Volumetric printing
For volumetric printing, a Tomolite printer from Readily3D and the updated Apparite software were used.To optimize printing parameters toward clearly defined constructs, a laser dose test was conducted for each resin.Defined spots were created on the wall of a quartz cuvette filled with a resin by the laser beam (λ = 405 nm) for varying exposure times ranging from 4 to 16 s and varying average light intensities ranging from 4 to 16 mW cm -2 .The light dose threshold (mJ cm -2 ) required for precise photopolymerization and minimal off-target exposure was calculated by multiplying exposure time with the average light intensity of the weakly visible polymerized spots in the cuvette.Construct printing was performed in a glass vial with 10-mm diameter in 1 mL of resin.The printed objects were washed by PBS pre-warmed to 37 • C.

Mechanical test
Tensile and compression tests were performed using TA.XTplus texture analyzer (Stable Micro Systems).Tensile tests were carried out on dog bone-shaped 3D-printed specimens (ASTM 638 type IV) with a gauge length of 13 mm at a rate of 0.15 mm s −1 .Every material was tested in at least triplicate.Compression tests were performed on 3Dprinted cylinders (H 8.0 mm and Ø 5.0 mm before swelling) at a rate of 0.17 mm s −1 with three compressions per sample.The compressive toughness was calculated as the area under stress-strain curves.Young's modulus (E) was determined as follows (Equation 7): where σ is the engineering stress and ε is the engineering strain.

In vitro cytotoxicity test
The test was performed in triplicates, with three samples of 3D-printed discs (H 0.8 mm, Ø 5.5 mm) for each replicate.The 3D-printed discs were washed with acetone for 30 min, followed with PBS pH 7.4 overnight, dried under vacuum for 24 h at room temperature, cured in UV chamber for 20 min, and soaked in medium for 20 min before the incubation.A549 cells were seeded in a 24-well plate with seeding density of 50,000 cells per well.The 3D-printed discs were then put into the wells on the top of Transwell inserts.As positive control, cells were incubated in medium with 100 mM H 2 O 2 while medium alone was used as a negative control.Cell viability was determined by the MTS assay after 48 ± 1 h of incubation (37 • C ± 1 • C, 5% CO 2 ).The Transwell supports and medium were removed, the wells were washed with PBS, and MTS reagent was added into the wells.Absorbance was measured after 45-60 min of incubation (until A = 0.6-0.8)using a spectrophotometric plate reader (490 nm, TPP24fT, without lid, linear shaking 10 s, amplitude 1.5 mm, number of flashes 25, settle time 5 ms).Cell viability was calculated as a percentage of the negative control. [47]The absorbance was recorded on spectrophotometric plate reader infinite M200 pro from Tecan (Switzerland).

Macroscopic imaging
Macroscopic images of 3D-printed complex structures, ear and meniscus models, and femur scaffolds were acquired on a stereomicroscope (Leica MZ6) coupled with a Leica DFC320 CCD camera or stereomicroscope (Leica M205 FA) coupled with a Leica DFC550 CCD camera for real color imaging.

Micro-computed tomography
3D-printed scaffolds were scanned at high resolution (17 μm) on a micro-CT 45 (Scanco Medical AG) operated at an energy of 45 kilovolt peak (kVp), an intensity of 177 μA with an integration time of 595 ms and frame averaging of 1.Using the Image Processing Language (IPL) software (Scanco Medical AG), scaffolds were segmented from background using a global threshold.After reconstruction, a 3D Gaussian filter (sigma, 1.3; support, 1) was applied to all images as described elsewhere. [88]

Surface functionalization and cell seeding
Small 3D-printed femur models (d = 3.20 mm, h = 2.03 mm; n = 3) were sequentially sterilized by UVA irradiation and 70% ethanol for 15 min each.After washing three times in PBS, they were functionalized with a fibronectin-derived arginylglycylaspartic acid (RGD) peptide (China Peptides, N-C: CGRGDS) at 10 mM in PBS (pH 7.94) to promote cell attachment.Samples were incubated in this solution at 37 • C overnight before washing three times with PBS to remove unreacted RGD.For 3D culture, a collagen type I hydrogel was prepared from an 8.91 mg mL −1 stock solution (rat-tail, Corning) as described by Shin et al. [89] Briefly, the collagen stock solution was mixed with 10% of 10 × PBS and cells in osteogenic medium to obtain a final collagen concentration of 1 mg mL −1 and a cell concentration of 3 × 10 6 mL −1 .The pH was adjusted to 7.4 using 0.54 M NaOH.The final gel precursor was then pipetted into the 3D-printed femur models and incubated for 30 min at 37˚C to polymerize.Cells were cultured in the osteogenic medium until Day 7.

Cell staining and confocal imaging
For live/dead cell imaging, two scaffolds were washed in PBS and stained for 15 min in a solution containing Calcein Green AM (CaAM, 1:500, Sigma-Aldrich) and Ethidiumhomodimer-1 (EthD-1, 1:1000, Sigma-Aldrich) in PBS.After washing three times in PBS, imaging was performed on an upright confocal laser scanning microscope (CLSM, Zeiss LSM 780).Viability was calculated as the percentage of live cells among all counted cells in six z-stack images from

F I G U R E 2
3D-printed hydrogels based on single PEG macromonomer and 4♦-BAPO-PEG.(A) Schematic illustration of the single-macromonomer resin formulation and the 3D-printed networks.(B) DSC curves from the second heating cycle of PEG macromonomers.(C) Viscosity of PEG macromonomers at various temperatures.(D) Photograph of the 3D-printed cylinders before and after swelling.(E) Swelling ratio of the 3D-printed PEG hydrogels with PEG macromonomers.Swelling ratios are expressed as mean + SD (n = 3 or 6).

F
I G U R E 3 3D-printed hydrogels based on dual-macromonomer resin.(A) Schematic illustration of the dual-macromonomer resin formulation and the 3D-printed networks.(B) Swelling ratio of the dual-macromonomer hydrogels.Swelling ratios are expressed as mean + SD (n = 6).(C-F) Tensile stress-strain curves (C) of dual-macromonomer hydrogels with different feed ratios of PEGDMA 10,000 and PEGDA 700 in comparison with single-PEG hydrogels, and the corresponding Young's modulus (D), maximum tensile stress (E) and elongation at break (F).The tensile properties are expressed as mean + SD (n = 3-6), except for the very weak PEGDMA 10,000 hydrogels (duplicates).(G and H) Compression stress-strain curves (G) of the hydrogel specimens, and the corresponding compression modulus (H).(I) Compressive toughness of the three groups of hydrogels.The compression mechanical properties are expressed as mean + SD (n = 2-3), given that nearly identical stress-strain curves were obtained.(J) Correlation between the Young's modulus of the hydrogel specimens and the equilibrium crosslinked polymer content.(K) Photographs of the hydrogel sample prepared from 3D-printed PEGDMA 10,000-PEGDA 700 (90/10, w/w) before and after bending.

F
I G U R E 5 3D-printed bone-mimicking PEG hydrogels.(A) Computed tomography scan model of a trabecular bone core from a horse femur.(B and C) Photographs and microscopy images of 3D-printed bone-like hydrogel scaffold from PEGDMA 10,000-PEGDA 700 (90/10, w/w) by heat-assisted DLP (B) or tomographic volumetric printing (C).(D) A549 cell viability after 48-h incubation (37 • C) with 3D-printed discs from PEGDMA 10,000-PEGDA 700 (green) and PEGDMA 6000 mixed with carboxylic acid-functionalized PEG 6000 (yellow) determined by the MTS assay relative to the negative control (cell culture medium, blue).Positive control is presented using 100 mM H 2 O 2 .Mean + SD (n = 3).(E) Schematic of human mesenchymal stem cell (hMSC) seeding on RGD-functionalized PEG scaffolds using a soft collagen hydrogel as the supportive matrix.The covalently bound RGD motifs promote cell attachment and differentiation under osteogenic culture.(F) 3D reconstruction of the scaffold by micro-CT imaging prior to swelling.(G-J) Microscopic images of hMSC seeding on RGD-functionalized hydrogel scaffolds.Bright field microscopic images of the samples following 2 h after seeding (G) and confocal microscopic maximum intensity projection images of live (green)/dead (red) stained cells (H and I) and actin (red)-nuclei (blue) stained cells (J) at Day 7 after seeding.