2.1 Preparation and characterization of GMPD hydrogels
To construct GMPD hydrogels, gel precursors of GelMA and PEGDA are prepared separately and then mixed with the photoinitiator lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) in different ratios. The radicals induced by 405-nm light cleave the carbon-carbon bonds in GelMA and PEGDA, and then bond with each other to form stable interpenetrating networks (Fig. S3). Subaqueous printing is based on an appropriate initial viscosity as well as rapid crosslinking reactions so that the gel precursors are minimally diluted and dispersed, which are the obstacles that appeared in previous studies. Compressed samples of GelMA with different initial concentrations are prepared in air or water and then tested for compressive modulus. As shown in Fig. 2a, the modulus of hydrogel samples with concentrations above 15% shows no significant difference between preparations in air and water, while samples with lower concentrations are affected by their environment and are often unable to gel effectively in water. Through a rheological assay, gelation times are found to be 1.5 ± 0.6 s and 1.2 ± 0.5 s for 15% GelMA (abbreviated G15) and G20, respectively, with a light irradiance of 100 mW/cm2, whereas at 30 W/cm2 and 10 W/cm2, complete crosslinking of G15 takes 5.2 ± 0.8 s and 11.9 ± 1.5 s, respectively (Fig. 2b, c). These longer gelation times may increase the risk of dilution.
Although photocrosslinking is a strong and irreversible reaction, the thermosensitive properties of pure GelMA hydrogels cause their decomposition at human body temperature16 and further lead to sealing failure for subsequent application in the treatment of PROM. The introduction of PEGDA significantly delays the thermal decomposition of the hydrogels, and the residual weight of the hydrogel samples with only 5% PEGDA (w/v) increases from 60–90% (p = 0.0007492, G15 compared to G15 with 5% PEGDA (abbreviated G15P5)) after 14 days of immersion in 37°C water (Fig. 2d). On the other hand, the modulus of the hydrogel increases significantly with the introduction of PEGDA, while the stress–strain curve shows a corresponding decrease in the maximum strain, appearing as the hydrogel samples become more brittle, which is undesirable in tissue sealing (Fig. 2e, Fig. S5). To optimize the ratio of GelMA and PEGDA, scanning electron microscopy (SEM) analysis is performed on hydrogels with different ratios. The SEM images reveal that the hydrogel with a lower PEGDA ratio has an interconnected-pore structure with pore diameters of 91.5 ± 14.4 µm, while with increasing PEGDA, the walls of the hydrogels become thicker and the porosity decreases (Fig. 2g, h), which provides a microscopic explanation for the increase in the compression modulus and the decrease in the swelling ratio (Fig. 2f). When the concentration of PEGDA exceeds 15%, the porous structure disappears, and it is detrimental to the further application of the material.
Considering the resistance to thermal decomposition and the microstructure of the hydrogel, 5% final concentration of PEGDA is chosen for the hydrogel, which not only maintains an over 90% residual weight after immersion in Ringer’s solution at 37 ℃ for 2 weeks but still possesses a porous structure and does not affect the toughness of the hydrogel (Fig. 2i). The modulus of G15P5 is 81.2 ± 12.7 kPa, which resembles the modulus of native fetal membranes (FMs)4.
Rheological temperature sweep analysis of the hydrogels shows that the viscosity and modulus-temperature curves are essentially coincident and that the gel point temperature of different groups has no significant variation (Fig. 2j, k). This result indicates that the introduction of PEGDA has little impact on the thermosensitive properties of the gel precursors, which is important for subsequent regulation of the printability of the hydrogels.
2.2 Subaqueous printability evaluation of the GMPD hydrogel
Desired printability requires hydrogels with appropriate yield stress, shear thinning properties and rapid shear recovery behavior17. Rotational rheological tests are performed on proportionally optimized hydrogels, and the results suggest that different groups of hydrogels all possess shear thinning properties (Fig. 3a). Many non-Newtonian bioinks demonstrate viscoelasticity and are accurately represented by the Herschel-Bulkley (H-B) model, expressed as τ = τ0 + K·γn where τ0 denotes the yield stress, n denotes the flow behavior index and K represents the consistency index. The relationship between the shear rate and shear stress for G15, G15P5, and G15P10 are fitted with the H-B model (Fig. 3b), and all calculated H-B parameters are listed in Fig. 3c. This result indicates that the introduction of PEGDA increases the overall yield stress and has an impact on the flow parameters. The preferred G15P5 has a yield stress of 68.13 Pa, which is acceptable for printability, as noted in previous studies18.
The flow behavior is critical for steady and continuous printing to achieve ideal fidelity. However, it is difficult to directly monitor the flow behavior within the nozzle; thus, the H-B fluid model is used here to analyze the flow of the hydrogel to predict the shear rate and velocity distribution during extrusion. At a flow rate of 0.25 mL/min, 80.4% of the radius (R0) demonstrates plug flow where the hydrogel has merely internal shear stress that does not exceed the yield stress and behaves like a solid19. The hydrogel in the plug flow is extruded forward at the same velocity, while the shearing is confined to a narrow region along the extruder walls (Fig. 3f). In addition, the viscosity of G15P5 corresponding to the shear rate distribution ranges from 1.7 to 32.5 Pa·s and has been revealed to be suitable for extrusion-based bioprinting20. These properties greatly improve the fidelity of the printed hydrogel filaments.
The GMPD hydrogel presents a typical shear-induced gel-fluid transition behavior and rapid self-healing. The strain sweep rheological assay on G15P5 exhibits a gel-fluid transition at 283% strain (Fig. 3d), and this transition is reversible. A cyclic shear applied to G15P5 at high (300%) and low (1%) strains shows that the hydrogel responds rapidly to shear and recovers immediately once the shear disappears (Fig. 3e). This shear thinning and self-healing can be repeated many times without sacrificing modulus.
MIS performed in the womb normally involves a fetoscope that can be abstracted as a rigid tubule with a length of 300 mm. In extrusion-based bioprinting within long needles, the state of the hydrogel may change due to long-term shear effects and may be detrimental to the printing fidelity. However, with the favorable printability of the GMPD hydrogel, it forms a robust, stable, and continuous hydrogel filament after extrusion from a 300 mm nozzle, and the length of the filament increases linearly during printing (Fig. 2g), even when the length extends to approximately 50 mm. These results indicate that the GMPD hydrogel possesses adequate printability and is suitable for subaqueous in vivo bioprinting based on the long-range extrusion procedure. In addition, the velocity of the extruded filament (Vink) and the nozzle feed velocity (Vnoz) should be matched to obtain the desired printed pattern; when Vink is lower than Vnoz, the continuity of the hydrogel filaments will not be guaranteed, and instead, the filaments will accumulate randomly, which is not conducive to 3D construction in water (Fig. S9a, b). As shown in Fig. 3h, filaments with diameters of 1.5 to 2.2 mm hold fidelity and 3D accumulation capacity, where the corresponding ranges of Vink and Vnoz are delineated.
For the treatment of PROM, compact hydrogel patches with uniform thickness are the most desirable 3D structures, and thus the fusion of filaments in the patches needs to be evaluated. Herein, the fusion index Fu is proposed and defined as Fu = L/(a + b) where L denotes the perimeter without a bottom edge, and a and b denote the height and length of the ideal gel patch, respectively. In overfusion where Fu < 1, filaments aggregate and appear as a droplet under surface tension and go against multilayer printing, while in underfusion where Fu > 1, insufficient fusion between filaments leads to imperfect patch surfaces and easily causes seal failure at the filament connections. Only under proper-fusion conditions can a dense and uniform hydrogel patch with Fu = 1 be obtained (Fig. 3i). With the printing parameters guided by the fusion index, a dense bilayer hydrogel patch with Fu = 0.9984 is printed subaqueously, and it maintained adequate mechanical properties to be picked up with tweezers (Fig. 3j, Supplementary Video 1).
2.3 Evaluation of robot-assisted bioprinting under MIS
MIS is a surgical approach in which operations are performed with long, narrow surgical tools that are held by a robotic arm and inserted into the patient through a small incision21. Surgical tools must pivot around these incisions which are called RCM constraints22 (Fig. 4a). Due to the redundancy associated with the additional DoF, the 7-axis robotic arm is capable of MIS at any spatial point as the RCM constraint actively, which provides greater flexibility compared to passive RCM approaches, including the da Vinci system produced by Intuitive Surgical, Inc23. Under the guidance of obstetricians, an end effector with a length of 300 mm and a diameter of 4 mm (similar to commercial fetoscopes) is designed, and a puncture mechanism, a syringe pump and optical fibers are integrated (Fig. 4b) for subaqueous bioprinting.
The active RCM approach is implemented through the manipulator’s software controller; thus, infinite available RCM locations exist within the manipulator’s workspace that need further optimization to acquire improved motion accuracy and efficiency. Yoshikawa24 proposed a “manipulability measure”, and Salisbury25 proposed the condition number of Jacobian matrixes as a measure of kinematic quality, but they could evaluate only a specific configuration of the manipulator but not the kinematic performance within the entire workspace under one RCM constraint. Inspired by Stucco and Salcudean26, the global isotropy index (GII) is used to provide a measure of available RCM locations. The GII can be calculated as
where σmin(X1) is the smallest singular value of the Jacobian J at workspace position X1, and σmax(X2) is the largest singular value of the Jacobian J at workspace position X2. A GII value of 1 indicates perfect isotropy, while a value of 0 indicates a singularity.
Figure 4c shows the GII distribution in the y = 0 plane. The best GII appears at (0.45, 0, 0.26) (meters) with a value of 0.1192, and the white background indicates the RCM locations at which part or all of the workspace is unreachable. Different RCM locations correspond to different control properties. Trajectory tracking experiments are conducted at 17 RCM locations where the actual trajectories are captured with two orthogonally placed cameras and compared with the desired trajectories. The RMS deviation is displayed in Fig. 4c. The best RCM location possesses a minimal deviation of 0.177 ± 0.011 mm, and the zigzag, star and circle trajectories performed at the best RCM location are listed in Fig. 4e, showing the coincidence between theoretical and actual trajectories (Supplementary Video 3). The same zigzag trajectory is performed with a hand-held procedure under the RCM constraint (Fig. S11b), and the RMS deviation reaches 0.896 ± 0.172 mm, which is 5-fold larger than that of the robotic arm (Fig. 4d). Meanwhile, the instability of the velocity in trajectory tracking may lead to nonuniformity of the hydrogel filament diameter, which makes the constructed hydrogel patch poorly integrated and prone to seal failure (Fig. 4f, g).
The constraint of the printing angle (the angle between the nozzle and normal orientation of the fetal membrane) is important when determining the incision on the body. As revealed in our previous study27, hydrogel filaments tend to slip off the substrate when the printing angle decreases. The FM and abdomen are abstracted as an ellipsoid and a plane, respectively, where the optimal incision is calculated by taking the printing angle as the optimization parameter, as shown in Fig. 4h.
An in vitro model is fabricated in accordance with the size of the human amniotic sac, and it is placed and punctured according to the calculated optimal RCM and incision location. Ringer’s solution at 37 ℃ is poured into the model to simulate amniotic fluid, and agarose gel is laid on the part to be printed to mimic soft tissue, such as the endometrium. Our bioprinting robot performs subaqueous bioprinting under RCM constraints, and the printed hydrogel patch is compact with a gel-rivet structure embedded in the agarose substrate (Fig. 4i, Supplementary Video 2). These results further demonstrate the feasibility and future application potential of GMPD hydrogels, the subaqueous printing process and the robot-assisted MIS approach.
2.4 Sealing evaluation of subaqueous bioprinting on in vitro models
To investigate the adhesion properties of the GMPD hydrogel, burst pressure tests and standard lap-shear tests are conducted. Figure 5a shows the schematic diagram, representative images and results of standard as well as modified burst pressure tests. The burst pressure obtained by the standard procedure solely relying on the adhesion of hydrogel is 37.0 ± 2.1 kPa, which is slightly lower than the 45.1 ± 4.2 kPa of the intact FM. However, the burst pressure recorded in the modified tests reaches 52.5 ± 7.2 kPa and shows no significant difference from that of the intact FM, indicating favorable sealing performance underwater. The same tests conducted on the intestinal mucosa of pigs presents similar results (Fig. S12d).
Lap shear tests are conducted according to the ASTM F2255-05 standard, and the shear strength between the hydrogel and the tissues is measured using two glass slides coated with human FM as the substrate. The shear strength of the crosslinked GMPD hydrogel reaches 33.2 ± 2.5 kPa which is higher than that of other commercial surgical sealants, including fibrin glue (18.2 ± 1.1 kPa) and PEG glue (21.9 ± 5.0 kPa) (Fig. 5b). To further improve the reliability of patch sealing, we propose a strategy in which the intermolecular forces of the material and mechanical forces of the specific structure combine to strengthen the adhesion of the hydrogel patch, namely the hydrogel rivets (Supplementary Video 4). With the ability to fabricate structures with 3D shapes, the in vivo bioprinting approach prepares hydrogel patches with rivets that are embedded in the tissue and complement the interfacial adhesion of the hydrogel (Fig. S16).
In evaluating the adhesive strength of hydrogel rivets, hydrogel patches are printed on agarose gel with or without gel rivets, and two specially designed agarose substrates are connected with a patch. The tensile test results show that the adhesive strength of the patch with rivets improves approximately 6-fold in comparison to samples without rivets, from 2.43 ± 0.65 kPa to 14.53 ± 2.49 kPa (Fig. 5c, Supplementary Video 5). To simulate fluid sloshing in a real amniotic sac, we use a flow-controlled water stream to scour the patch with or without rivets, and the results indicate that the ability of patches with rivets to resist stream impact improves by approximately 7-fold (Fig. S12a, b, Supplementary Video 6).
An in vitro uterus model is established and punctured to mimic PROM for evaluating the sealing performance of various hydrogels. Different groups of hydrogels are printed subaqueously as sealants at the rupture site, and fluid leakage is recorded periodically. Figure 5d shows that hydrogel sealants without PEGDA are prone to seal failure due to partial dissolution at 37 ℃, resulting in accelerated fluid leakage. However, the optimal G15P5 sealant maintains a leakage profile similar to that of the intact FM during the 21-day observation period (Supplementary Video 7), which validates the effectiveness of the subaqueous bioprinting sealing approach for treating PROM.
2.5 Sealing evaluation of subaqueous bioprinting on mid-gestation rabbit models
The biocompatibility of bioinks is essential for intracorporal applications. We prepare cell culture plates coated with G15P5 hydrogel and culture rabbit amniotic epithelial cells on the surface of the hydrogel for 7 days. The cell viability assay indicates that the cells maintain over 90% viability in the 7-day culture in both the experimental and control groups (normal 2D culture) (Fig. S13a), and the representative images of Calcein AM/propidium iodide immunofluorescent staining shows that the cells could extend to spindle or diamond shapes in the appearance of the G15P5 hydrogel (Fig. S13c). The results of the CCK-8 assay in Fig. S13b also show that the cells proliferate well in the experimental groups.
Pregnant rabbits at mid-gestation are considered a suitable model to study FM defects and to validate PROM sealing techniques and are therefore selected for intracorporal research and evaluation. The average duration of a rabbit’s pregnancy is 32 days, with litters of approximately eight offspring. The amniotic sac reaches a size of 4–5 cm at a gestational age (GA) of 22 days, which is compatible with FM repair tests. According to previous studies28, 29, one to two amniotic sacs in one rabbit are selected as the experimental group and positive control group (Fig. 6a). Sacs in the experimental group are punctured and sealed with an I-shaped hydrogel patch that could be embedded at the incision, whereas sacs in the positive control group are simply punctured. At a GA of 31 days, the rabbits are euthanized for a second hysterotomy to evaluate the sealing performance (Fig. 6b). One rabbit died due to surgical infection, while the other eight survived without herniation of the fetus into the maternal abdomen at sacrifice. After myometrial dissection and opening of the gestational sac, no intraamniotic adhesions or amniotic bands are observed. Further data on the outcome are presented in Fig. 6d.
The fetal survival rate is 72.7% in the sealing group, similar to 81.3% in the native control group. Among them, 10 out of 11 hydrogel patches could be traced, and 8 out of 10 remained stably mounted on the FM, and these eight sacs remain full of amniotic fluid. The remaining two patches separated from the FM and are found freely in the uterus. In the positive control group, amniotic fluids almost vanish in all sacs, and only one fetus survived. The average weight of fetuses in the experimental group is 33.2 ± 4.7 g compared to 8.6 ± 3.1 g in the positive control group, indicating that fetal development stops shortly after FM puncture and that fetuses are partially resorbed. Both gel sealant and FM could be identified from sealed tissues processed for histological evaluation. From HE-stained sections, it could be seen that the GMPD hydrogel is tightly attached to FM. At larger magnifications, the GMPD hydrogel is observed to be well integrated with FM, completing the restoration of membrane integrity and even exhibiting re-epithelialization. (Fig. 6c).
In subcutaneous implantation experiments in rats, the tissue around the implanted GMPD hydrogel undergoes further histological and immunofluorescence assays at 7, 14 and 21 days after implantation to verify immunogenicity. When a hydrogel is implanted, the body’s immune system recognizes such implants as foreign and encapsulates them in dense collagen30. The morphology and thickness of the collagen capsule can be considered metrics of immunogenicity. Masson’s trichrome staining of the implanted tissue after three weeks shows that the collagen capsule is discontinuous and not compact with a thickness of ~ 50 µm, indicating that the hydrogel does not elicit a severe substantial foreign-body reaction (Fig. 6e). HE staining on day 21 reveals the presence of the G15P5 hydrogel and shows favorable integration with the native tissue (Fig. 6f). In addition, immunofluorescence staining for CD68 (Fig. 6g) shows few macrophages (stained in green) after 21 days. Together, these assays indicate that GMPD hydrogels present low immunogenicity and favorable biocompatibility that can be considered an ideal biomaterial for intracorporal applications.