Comprehensive Characterisation of Photocurable PEGDA/Gelatine Hydrogels for Extrusion-Based 3D Printing

TL;DR

This study develops and characterizes photocurable hydrogel inks for 3D printing, showing they can be tuned for mechanical properties and are safe for cell growth after washing.

Contribution

A novel PEGDA–gelatine–collagen ink formulation with optimized rheology, tuneable stiffness, and low cytotoxicity after washing is introduced.

Findings

The Herschel–Bulkley model best described the inks' flow behavior.

Washing significantly reduced residual photoinitiator and improved cell viability.

PeGeCol_10_2 ink enabled high-fidelity printing of large anatomical models.

Abstract

The development of photocurable hydrogel biomaterial inks with suitable rheology, low cytotoxicity, and tuneable mechanical properties is essential for reliable biofabrication. This study aimed to formulate PEGDA–gelatine–collagen inks using lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) as photoinitiator. Rheological characterisation and flow-model fitting were performed, mechanical stiffness modulation under different light intensities was evaluated, complex structures were printed using direct extrusion and FRESH methodologies, and PEGDA/LAP extractables were quantified by NMR after controlled washing procedures. In vitro assays assessed cell viability and proliferation on the resulting scaffolds. The Herschel–Bulkley model best described the flow behaviour across formulations; while viscoelastic measurements showed that increasing light intensity progressively enhanced hydrogel stiffness, enabling fine control over final mechanical properties. NMR analysis showed that washing removed a substantial fraction of residual LAP, in agreement with the biological findings: fibroblasts failed to survive on unwashed scaffolds but exhibited robust proliferation and recovered their characteristic elongated morphology on washed constructs. Among all inks, PeGeCol_10_2 provided the best combination of shear-thinning behaviour, structural integrity, low residual photoinitiator, and tuneable mechanics. Using this formulation, we successfully printed large anatomical models with high fidelity and excellent handling properties, underscoring its potential for soft-tissue prosthetics and broader tissue-engineering applications.

Figures

Full text

1. Introduction

Tissue engineering, a major field within biomedicine, focuses on creating biomimetic structures for studying tissue models in vitro [1,2]. This approach offers an alternative to organ transplantation [3], decreases dependence on animal testing [4], and enables personalised medicine using patient-derived cells [5]. A key enabling technology for tissue engineering is three-dimensional (3D) printing, which allows the fabrication of well-defined scaffolds by the controlled deposition of hydrogels and bioactive components in a layer-by-layer manner [1,6].

The successful use of 3D-printed scaffolds strongly depends on the physicochemical and rheological properties of the materials employed. Parameters such as viscosity, gelation mechanism, and crosslinking kinetics are critical to ensure adequate printability, structural fidelity, and long-term stability of the constructs [7,8]. The use of sacrificial support strategies such as Freeform Reversible Embedding of Suspended Hydrogels (FRESH) enables the correct deposition of soft hydrogels that cannot be printed by conventional extrusion alone. In the FRESH approach, the ink is extruded into a support bath that behaves as a solid at rest but locally fluidises under shear during nozzle movement [7,9,10]. This temporary mechanical confinement prevents filament collapse during deposition and can be fully removed after printing, typically by mild temperature changes or dissolution. As a result, FRESH allows the printing of low-stiffness materials and the fabrication of complex three-dimensional geometries, including hollow structures, overhangs, and unsupported curved features that would otherwise be unattainable using direct extrusion methods [11]. While enabling the processing of very soft hydrogels, the development of formulations suitable for direct extrusion printing remains a major challenge, since printed filaments may collapse, fuse, or lose resolution due to insufficient mechanical support [7,12,13]. Achieving formulations that combine good extrusion behaviour with post-printing stability is therefore essential for the broader applicability of 3D printing.

To address these requirements, different strategies have been proposed, including the use of natural thermoresponsive polymers such as collagen or gelatine [14,15,16], which provide reversible sol–gel transitions and improve handling during the printing process. In parallel, photocurable hydrogels have attracted significant attention, as light-induced polymerisation allows reinforcement of the printed structures with high spatial control [16,17,18]. These systems typically incorporate short-chain polymers, such as poly(ethylene glycol) diacrylate (PEGDA), together with a photoinitiator. Upon irradiation, the initiator generates free radicals that promote polymerisation, yielding mechanically stable scaffolds with tuneable internal architectures. Among photoinitiators, lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) has gained relevance due to its activation under blue light (405 nm), which is less harmful to cells compared to UV-based systems like Irgacure [17,18,19].

An additional advantage of photocurable formulations is the possibility to modulate the degree of crosslinking by adjusting parameters such as light intensity, exposure time, or polymer concentration. This tunability enables the fabrication of scaffolds with a wide range of mechanical properties, which can be tailored to mimic the stiffness of different native tissues—from soft brain or liver tissue to stiffer cartilage matrices [20,21,22]. Since cellular behaviour is highly sensitive to the mechanical characteristics of the surrounding environment, such as substrate stiffness or viscoelasticity, fine-tuning the crosslinking conditions is essential to promote appropriate cell adhesion, proliferation, and differentiation [23,24]. Despite these advantages, residual unreacted compounds (e.g., PEGDA oligomers and LAP) may compromise cytocompatibility if not adequately removed [25,26]. Hence, proper characterisation of both the printing performance and the biological compatibility of photocurable formulations is crucial [18,27].

In this context, the development and comprehensive characterisation of photocurable hydrogels represent key steps toward designing printable materials that balance structural integrity with biological functionality. Understanding how rheological and crosslinking properties influence both printability and cytocompatibility will ultimately enable the creation of 3D scaffolds that more accurately replicate native microenvironments and support the formation of physiologically relevant tissue models.

In this study, we focus on the systematic evaluation of a PEGDA/gelatine/LAP hydrogel, optimised for extrusion-based 3D printing and subsequent photo-crosslinking. By analysing the rheological, mechanical, and crosslinking behaviour of the system, we aim to establish correlations between material properties, printability, and biological response. This approach underlines the importance of material characterisation as a key step toward developing photocurable hydrogels that combine structural fidelity with the capacity to support cell growth and tissue formation.

2. Results and Discussion

2.1. Viscoelastic Properties

To systematically evaluate how composition influences printability, stiffness, and potential cytocompatibility, six hydrogel formulations were prepared combining PEGDA, gelatine, and type I collagen at different concentrations. Two base families were defined: PEGDA–gelatine inks without collagen (PeGeLap_5, PeGeLap_10, and PeGeLap_10_2) and type I collagen-enhanced inks (PeGeCol_5, PeGeCol_10, and PeGeCol_10_2), the latter designed to promote cell adhesion and support biological performance. Within each family, formulations at 5% and 10% PEGDA were produced to compare network density, while a second set with the best performing inks (10% PEGDA) reduced the LAP content (from 11% to 2% v/v) to minimise potential photoinitiator-related cytotoxicity while maintaining acceptable photocuring behaviour. This design allowed us to independently assess the effects of collagen addition, PEGDA concentration, and LAP reduction on the final properties of the hydrogel.

The viscoelastic properties of the six indicated hydrogels were studied. As can be observed in Figure 1, before the light is turned on (t = 30 s), all tests display almost the same value of G′ (kPa).

Once the light is activated, the polymerisation reaction of PEGDA begins due to the activation of photoinitiator LAP, leading to markedly different final G′ values among the various biomaterial inks and light intensities. This demonstrates that, by adjusting the ink composition and the applied light, a wide range of mechanical stiffness values can be achieved. From the data, the formulation showing the highest G′ at 100% light intensity is PeGeLap_10 with approximately 120 kPa, whereas the lowest value, around 12 kPa, corresponds to PeGeCol_5. These results confirm that the ink formulation has a direct influence on the resulting storage modulus, allowing the tuning of mechanical properties across a broad range.

In most cases, inks containing collagen exhibit lower G′ values at any light intensity. This behaviour can be attributed to interference of collagen fibres with the internal crosslinking of PEGDA and gelatine, leading to a final material of reduced stiffness. However, the opposite trend is observed with PeGeLap_10_2 and PeGeCol_10_2. In these formulations, the collagen-containing ink shows higher stiffness. This is likely due to the lower concentration of LAP; with fewer LAP molecules, fewer free radicals are generated upon irradiation, resulting in slower and more limited PEGDA polymerisation. Consequently, the structural and physical reinforcement provided by collagen dominates over the curing process, yielding a stiffer material compared to the control (PeGeLap_10_2). Additionally, it is noteworthy that, for all materials, reducing light intensity causes a noticeable time lag between light activation and the onset of stiffness increase. This can be explained by the reduced photon flux at lower intensities, which decreases the number of LAP molecules activated and therefore delays free radical generation and PEGDA polymerisation. A similar delay was also observed in inks with lower LAP or PEGDA concentrations, as the initiation of the crosslinking reaction requires more time under these conditions.

Only results obtained under 405 nm light exposure are discussed in this section, as those collected using 365 nm illumination exhibited unstable measurements for some of the biomaterial inks. This instability is likely related to the much faster activation of LAP molecules at that wavelength, which may cause uncontrolled or random curing of the inks. In agreement with this interpretation, reducing the LAP concentration led to more reproducible and stable behaviours.

2.2. Rheological Analysis and Mathematical Modelling of Flow Behaviour

The rheological behaviour of the developed inks was evaluated to assess their suitability for extrusion-based 3D printing [28]. All formulations exhibited pronounced shear-thinning behaviour, as shown in Figure 2a, where viscosity decreased steadily with increasing shear rate. This non-Newtonian trend is characteristic of hydrogel-based biomaterial inks and ensures easier extrusion through the printing nozzle under applied shear stress. Among all formulations, PeGeLap_10 and PeGeCol_10 showed clearly distinct viscosity profiles compared to the rest of the inks, exhibiting significantly higher viscosity values across the entire shear rate range. This behaviour is consistent with their higher polymer content, which increases chain entanglement and intermolecular interactions within the hydrogel matrix. The presence of collagen in PeGeCol_10 further enhanced this effect, slightly increasing viscosity relative to its non-collagen counterpart. More generally, collagen-containing formulations exhibited higher viscosity in the shear-rate range relevant for extrusion (100–1000 s^−1^) compared to their collagen-free analogues [29]. This behaviour is consistent with previous studies reporting that collagen fibrillogenesis and physical interactions reinforce hybrid hydrogel networks [30]. In contrast, the formulations with reduced LAP concentration (PeGeLap_10_2 and PeGeCol_10_2) displayed overlapping viscosity curves. A plausible explanation is that higher LAP contents may favour a small degree of inadvertent pre-gelation during handling under ambient light, subtly increasing viscosity before testing. When LAP is reduced to 2% v/v, this effect becomes negligible, resulting in more fluid inks and minimising compositional differences such as collagen addition.

The thixotropic behaviour, evaluated through the three-step shear recovery test in Figure 2b, confirmed that all formulations underwent a reversible viscosity drop under high shear (700 s^−1^) followed by partial recovery when returned to low shear conditions. This indicates the ability of the inks to flow during extrusion and recover their internal structure post-deposition. Notably, PeGeLap_10 showed the highest viscosity recovery, consistent with its higher baseline viscosity, whereas the remaining formulations displayed more limited structural recovery. It is important to note that these rheological tests were conducted at 27 °C, a temperature at which gelatine remains in a predominantly liquid state. Consequently, the inks behave more as viscous fluids rather than self-supporting gels under these conditions. During the printing process, this property is deliberately exploited; the thermoreversible nature of gelatine enables in situ gelation when the material is deposited onto a printing bed maintained at a lower temperature, promoting rapid solidification and shape retention despite the limited viscosity recovery observed in rheometry tests. Overall, these results demonstrate that the combination of high polymer and photoinitiator concentrations yields a more viscous and structurally robust ink, while formulations with reduced LAP are more fluid and compositionally insensitive. Moreover, the thermoreversible behaviour of gelatine provides a practical strategy to control printability and structural stability through temperature management during the printing process.

To gain further insight into the flow behaviour of the inks and quantitatively describe their rheological profiles, the experimental flow curves were subsequently fitted to different mathematical models. As part of the analysis workflow, for the flow curve mathematical modelling, after the initial fitting of the replicate data, a second fitting was performed using the optimal parameter estimates obtained from the first run as starting values. This iterative approach aimed to refine parameter convergence and improve model accuracy. However, the second fitting did not yield any meaningful improvement in the coefficient of determination (R^2^) in more than 80% of the cases; therefore, it was omitted in the subsequent analyses. In those instances where the second iteration provided a better fit, the improved results were retained. When comparing the best fits obtained for each replicate between TRIOS and pyRheo [31,32], pyRheo outperformed TRIOS in over 90% of the cases. This difference is largely due to the fact that TRIOS occasionally identifies models with rheologically inconsistent parameters, such as negative flow behaviour index (n < 0). These unphysical results were automatically discarded and, in every such case, the pyRheo-derived model was selected instead, as pyRheo enforces physically meaningful parameter boundaries. After averaging the replicate data, the mean curves were also fitted using pyRheo, and the results were compared against the averaged model parameters and the TRIOS estimations. In approximately half of the cases, the best-fitting model was obtained by directly fitting the averaged data with pyRheo, while, in the remaining cases, the averaged model and the TRIOS estimation performed comparably. It is worth noting, however, that pyRheo consistently produced models of sufficient quality and rheological coherence, whereas parameter averaging often led to unrealistic approximations, and TRIOS results were, as previously mentioned, frequently not physically plausible. These observations suggest that direct fitting of the average data using pyRheo constitutes a reliable and efficient approach. Although the multi-step optimisation process described above could potentially yield slightly better estimations, it requires considerable additional time and computational effort, which may not be justified in practical terms.

Using all pyRheo’s available functions, the integrated machine learning predictor (MLP) was also employed to identify the most suitable model for each ink. Although the integrated MLP tool identified the Carreau–Yasuda (CY) model as the most representative for the experimental data (Figure 3), the quantitative comparison based on the coefficient of determination (R^2^) indicated that the Herschel–Bulkley (HB) model generally provided a better fit across formulations. The corresponding parameters obtained from both models are summarised in Table 1.

In all cases, the high R^2^ values (>0.87) confirmed a good agreement between experimental data and the theoretical models, with the HB model performing particularly well for formulations exhibiting a measurable yield stress. The flow index (n) values obtained from both models were below 1 for all formulations, confirming a pronounced shear-thinning behaviour typical of printable hydrogels. Formulations with lower n values exhibited stronger pseudoplastic behaviour, which favours extrusion and shape retention after deposition. In contrast, PeGeLap_10_2 and PeGeCol_10_2, which contain lower LAP concentrations, displayed slightly higher n values (0.58–0.70), consistent with a less crosslinked and more fluid-like behaviour. Notably, formulations containing collagen (PeGeCol series) showed much higher yield stress (σ0) values compared to their LAP-based counterparts, suggesting an additional contribution of the collagen network to the overall viscosity and yield stress. The Carreau–Yasuda parameters (η0, η_∞_, a) provided complementary information, with PeGeCol_5 and PeGeCol_10 showing the highest zero-shear viscosities. Conversely, formulations with lower polymer content (PeGeLap_5) exhibited smaller η0 value, consistent with its poorer printing performance. Overall, these results highlight that both polymer composition and photoinitiator concentration play a key role in tuning the rheological properties of PEGDA-based hydrogels, determining their printability and structural fidelity during extrusion.

2.3. 3D Printing

As previously mentioned, most materials designed for 3D printing are tailored for a specific technique and are generally unsuitable for alternative approaches. In this study, most of the hydrogels prepared were sufficiently versatile to enable the fabrication of various 3D structures using both direct semi-solid extrusion and the FRESH technique.

The printability of the hydrogels was assessed employing a range of STL models that encompass both printing methods and diverse geometric characteristics (see Supplementary Materials for printing parameters). Variations in component concentration did not produce any discernible difference in the macroscopic appearance of the inks. Nevertheless, print fidelity, resolution, and the ease with which the final 3D constructs could be manipulated were found to depend significantly on the ink composition.

A square-shaped thin scaffold (0.50 mm thickness) was printed by direct semi-solid extrusion to evaluate both the resolution and the variation in stiffness among inks when photopolymerised under different wavelengths (405 nm and 365 nm, 30 s exposure). When assessing the printing resolution prior to photocrosslinking (Figure 4), the influence of temperature and composition on ink printability became evident. Increasing the printing-head temperature enhanced ink fluidity but at the expense of printing resolution in some cases, as filament spread and fusion between adjacent strands became more pronounced. A clear compositional trend was also observed: inks with higher polymer content (10% gelatine, PEGDA, and collagen) exhibited markedly improved structural definition compared with their 5% counterparts. PeGeLap_5 and PeGeCol_5 required lower extrusion temperatures (20 °C) to achieve any degree of printability. However, due to the poor filament resolution and extensive merging between layers, PeGeLap_5 was not further considered in subsequent printing trials. Reducing the LAP concentration in the formulations did not impair printability; on the contrary, these inks maintained well-defined filament paths and good layer alignment. Moreover, the presence of collagen appeared to enhance print fidelity and interlayer cohesion, likely by reinforcing the physical integrity and elastic response of the ink prior to photocuring. Overall, PeGeCol_10 and PeGeCol_10_2 displayed the best printing performance, combining continuous filament deposition and good shape retention before photocuring.

Regarding the handling of the thin samples, as can be seen in Figure 5, inks with higher stiffness (PeGeLap_10 and PeGeCol_10) exhibited pronounced structural integrity and mechanical stability post-printing and curing. These constructs retained a well-defined geometry and could be lifted and repositioned cleanly without deformation or adhesion issues. In contrast, PeGeLap_5 and PeGeCol_5, which were less stiff and exhibited high self-adhesion, tended to bend under their own weight and stick to themselves when lifted, indicating a softer material behaviour. Formulations with lower LAP concentration (PeGeLap_10_2 and PeGeCol_10_2) demonstrated intermediate behaviour. These inks showed reduced stiffness, reflecting a lower crosslinking density, but also low self-adhesion. Although they also bent when lifted, they could still be repositioned without losing structural integrity. When comparing constructs cured under both light sources, those photocrosslinked at 365 nm were, in every case, slightly more rigid than those cured at 405 nm. This increased rigidity rendered them more brittle and challenging to remove from the plate but also improved shape retention.

To further investigate the internal morphology of the printed hydrogels, printed scaffolds were lyophilised to remove water while preserving the 3D microstructure. This approach allowed a qualitative assessment of porosity, filament definition, and the effect of formulation and light wavelength on the crosslinked network (Figure 6). Lyophilisation revealed clear differences in the internal architecture of the scaffolds depending on ink composition and photocuring conditions. Formulations with lower concentrations of gelatine, PEGDA, and collagen displayed more voids and irregular pores, likely due to reduced viscosity and crosslinking density. In contrast, inks containing collagen produced more coherent and continuous filaments, particularly evident for PeGeCol_10 cured under 405 nm light. Moreover, scaffolds photocrosslinked with 405 nm exhibited a looser, less compact internal matrix compared to those cured at 365 nm, which appeared denser and more uniform. The influence of LAP concentration was also evident; formulations with lower LAP content showed fewer interconnections and less structural cohesion than their higher-LAP counterparts.

After the initial printing and curing tests, the PeGeCol_10_2 formulation was selected for the fabrication of larger structures due to its good printability, reduced LAP content—minimising potential cytocompatibility concerns—and its ability to produce fully cured, mechanically stable constructs with high shape fidelity, as shown in Figure 7. Given its tactile properties and stiffness range relevant for soft-to-intermediate tissue-mimetic applications, two anatomical models—a nose and an ear—were chosen to evaluate the performance of this biomaterial in printing complex, clinically relevant geometries. Both models measure approximately 4 cm in size. Detailed printing parameters for these demonstrator pieces are provided in the Supplementary Materials.

When employing the FRESH technique, a tubular structure was printed to evaluate the capability of the hydrogel formulations to form elongated constructs with thin walls, representative of blood-vessel- or stent-like geometries. Such structures are of particular interest in the field of regenerative medicine, where the fabrication of perfusable and mechanically compliant tubular constructs remains a major challenge for tissue engineering applications [33]. This approach demonstrated that complex, tall, and delicate structures—typically unachievable by direct extrusion due to gravitational collapse—could be successfully fabricated and recovered from the Pluronic bath after photocrosslinking with light. A prolonged curing time of 2 min was adopted for FRESH prints to account for light scattering within the support matrix. The only formulation unsuitable for FRESH 3D printing was PeGeCol_5. As illustrated in Figure 8, the tubular constructs (5 mm diameter) produced with this formulation lacked well-defined contours and structural integrity. This failure is attributed to the elevated temperature necessary to solidify the sacrificial bath, which caused the ink to become excessively fluid during printing. In contrast, PeGeLap_10_2 and PeGeCol_10_2 produced successful FRESH prints, yielding stable tubular structures that were slightly softer and more flexible than those generated with PeGeLap_10 and PeGeCol_10. Aside from this, no substantial differences were detected among the latter formulations.

2.4. In Vitro Cell Viability

To assess the biocompatibility of the biomaterial inks with normal cells, the PeGeCol_10_2 formulation was selected due to its combination of appropriate stiffness, well-defined filament structure, and a compact, interconnected internal network—features that favour cell attachment and proliferation. L929 mouse fibroblasts were seeded onto 3D-printed PeGeCol_10_2 scaffolds, and cytotoxicity was assessed through complementary quantitative and qualitative approaches. Quantitative analysis was performed using two independent assays (LIVE/DEAD and HS-AlamarBlue), while qualitative evaluation focused on cell morphology, including signs of lysis, vacuolisation, or abnormal spreading. Considering the potential cytotoxicity derived from residual LAP and other hydrogel constituents, four experimental post-printing washing conditions were examined: (1) non-washed discs, (2) discs washed for 24 h (24 h-W), (3) discs washed for 24 h, with a medium change after 8 h (24 h-W + MC), and (4) discs washed for 48 h under continuous immersion (48 h-W).

Cell viability was analysed by staining live (green) and dead (red) cells with calcein and BOBO-3 iodide respectively. As shown in Figure 9a, fibroblasts cultured on all washed discs remained highly viable after 7 days, with more than 90% of cells exhibiting green fluorescence. In contrast, non-washed discs displayed pronounced cell death and poor adhesion from the earliest time points, suggesting cytotoxic effects probably associated with unremoved photo-initiator or other residual components, like PEGDA. Notably, fibroblasts cultured on washed hydrogels spread and proliferated throughout the scaffold, adopting the characteristic elongated morphology typical of this cell line in two-dimensional culture, as seen in the lower panels of Figure 9a.

Consistent with the LIVE/DEAD results, metabolic activity measured using the HS-AlamarBlue assay (Figure 9b and Table S7) revealed a significant increase in proliferation in all washed conditions, with approximately a 30-fold rise at day 7 relative to day 0 (**** p* ≤ 0.001). No significant proliferation was observed in the non-washed group. These findings demonstrate that pre-washing the scaffolds effectively eliminates cytotoxic residues, thereby enhancing fibroblast survival and proliferation within the PeGeCol_10_2 hydrogel matrix. Overall, these results demonstrate that the PeGeCol_10_2 formulation provides a highly biocompatible environment that supports cell attachment, viability, and proliferation, provided that residual cytotoxic compounds are removed through appropriate washing.

2.5. NMR Analysis

In line with the in vitro cell viability results, which highlighted the importance of removing residual components, the quantification of unreacted PEGDA and LAP remaining after photocrosslinking was carried out by ^1^H NMR spectroscopy. PeGeCol_10_2 hydrogel discs were immersed in deionised water under the same post-printing washing conditions applied in the biocompatibility assays—namely 24 h in static water (24 h-W), 24 h with a medium change after 8 h (24 h-W + MC), and 48 h under continuous immersion (48 h-W). After washing, the eluates were lyophilised, and the dry residues were analysed to determine the residual monomer and photoinitiator content. The corresponding results are summarised in Table 2.

The results indicate that prolonged immersion (48 h) resulted in the highest extraction of both LAP and PEGDA, which is consistent with extended diffusion time and additional network relaxation. In contrast, the two 24 h washing protocols showed very comparable values, with only minor differences between samples washed with or without medium renewal. Given the small magnitude of these variations, they likely fall within experimental variability and do not suggest a meaningful effect of medium replacement on extraction efficiency. Overall, the only clear trend observed was the increased recovery of both components following 48 h of continuous washing. Notably, these findings align with the in vitro assays, where cells failed to proliferate on unwashed scaffolds but exhibited adhesion and growth on all washed samples—regardless of whether they were washed for 48 h or for 24 h with or without medium renewal.

2.6. Swelling Behaviour of Hydrogels

The swelling behaviour of PeGeCol_10_2 was evaluated over a 72 h period. As shown in Table 3, the hydrogel exhibited a progressive increase in mass during the first 48 h of immersion, reaching a maximum average swelling value of 19.01 ± 0.24. This initial phase reflects effective water uptake and network expansion. After 72 h, a decrease in the swelling ratio was observed (14.45 ± 1.81), suggesting that the system reached a swelling equilibrium followed by partial network relaxation. Importantly, no uncontrolled swelling was detected, indicating good dimensional stability of the photocured network. The low variability observed at 48 h highlights the reproducibility of the swelling behaviour at equilibrium and is in good agreement with the NMR extractables analysis, supporting the idea that network relaxation at equilibrium facilitates the diffusion of residual, loosely bound components from the hydrogel matrix.

3. Conclusions

This study shows the effective creation of photocurable PEGDA–gelatine–collagen biomaterial inks with low cytotoxicity, strong printability, and adjustable mechanical characteristics. Rheological study showed that the Herschel–Bulkley model was the best descriptor of flow behaviour, therefore verifying the great shear-thinning profile necessary for extrusion-based printing. Non-destructive mechanical measurements demonstrated that hydrogel stiffness can be precisely modulated by adjusting light intensity during photocuring, with the selected formulation (PeGeCol_10_2) reaching stiffness values of approximately 100 kPa. Swelling experiments revealed a controlled, time-dependent water uptake with no uncontrolled expansion, indicating good dimensional stability of the photocured network. NMR quantification of extractables showed that prolonged immersion (48 h) led to the highest recovery of mobile PEGDA and LAP fractions, while the in vitro results demonstrated that a 24 h washing step was already sufficient to reduce residual components to levels compatible with cell survival. Accordingly, fibroblasts cultured on washed scaffolds—independently of washing duration—exhibited high viability (>90%) after 7 days, whereas non-washed constructs showed extensive cell death and poor adhesion. PeGeCol_10_2 showed the most favourable balance of printability, structural stability, and low residual photoinitiator content among all formulations. This composition allowed the creation of large, high-fidelity anatomical models like ear and nose structures, therefore emphasising its usefulness in soft-tissue prosthetic applications as well as its potential as a flexible platform for future tissue-engineering methods.

4. Materials and Methods

4.1. Materials

Gelatine from porcine skin (CAS no. 9000-70-8), poly(ethylene glycol)diacrylate (PEGDA) (CAS no. 26570-48-9), lithium phenyl-2,4,6-bimethylbenzoylphosphinate (LAP) (CAS no. 85073-19-4), and Pluronic F-127 (CAS no. 9003-11-6) were purchased from Merck Life Science, Madrid, Spain. Sodium bicarbonate solution 7.5%, Dulbecco’s Phosphate Buffered Saline (DPBS), collagen type I from rat-tail and 3 mL printer compatible Luer-Lok™ tip syringes were purchased from Fisher Scientific, Madrid, Spain.

4.2. Biomaterial Ink Formulation

Biocompatible thermoreversible inks were prepared in six different compositions (Table 4). All formulations followed the same general procedure, with a single additional step for those containing collagen. For inks without collagen, gelatine was hydrated by adding half the total amount of DBPS volume into an empty syringe (Part 1), followed by the complete amount of gelatine. The syringe was placed then in a 37 °C water bath until the gelatine was dissolved. Meanwhile, the remaining components—except the PEGDA—were combined in a second syringe (Part 2). Once the gelatine solution was ready, PEGDA was added into Part 1. The contents of both syringes were then mixed using a Luer-Lok™ syringe-to-syringe connector, applying gentle manual agitation to ensure homogeneity. Finally, the syringes were adequately labelled and stored at 4 °C in the dark until use. For collagen-containing inks, the protocol was identical except for one step: cold collagen was added to Part 2. To facilitate the homogeneous incorporation of LAP, a 2.25% (w/v) stock solution was prepared, from which the required aliquot was taken for each formulation.

4.3. Viscoelastic Properties Analysis

The viscoelastic properties of the inks were evaluated using an ElastoSens™ Bio mechanical tester (Rheolution Inc., Montreal, QC, Canada). This instrument is equipped with an ultrasonic volume probe, three integrated UV and visible light sources, a laser optical probe, and a Peltier-based temperature control system. Measurements were performed using µ-volume sample holders, which require 250 µL of sample per test. Each test was performed by triplicate.

Prior to testing, formulations were incubated in a 30 °C bath for 60 min to ensure complete melting. Then, 250 µL of each sample was poured into the µ-volume sample holders and stored at 4 °C to allow gelation for 30 min. The holders were subsequently placed in the ElastoSens™ Bio, pre-cooled to 20 °C. The test protocol consisted of a 30 s baseline measurement in the absence of light, followed by 30 s of illumination and a final 60 s measurement after the light was turned off. Each formulation was analysed in triplicate under two different wavelengths (365 nm and 405 nm). Data were processed using the instrument’s software, and results were expressed as elastic modulus (G′) versus time curves.

4.4. Rheological Analysis

Rheological analysis of the ink samples was performed using a controlled-stress rheometer (Discovery HR-2, DHR, TA Instruments, New Castle, DE, USA) equipped with a parallel plate (25 mm diameter, 0.5 mm gap) and a controlled convection/radiant heating chamber (Environmental Test Chamber, ETC, TA Instruments, New Castle, DE, USA) for stable temperature control. Prior to measurements, the inks were incubated in a 27 °C water bath for 30 min to obtain a flowable state allowing easy handling while minimising structural damage and preventing air entrapment. Shear-viscosity tests were conducted in flow ramp mode, with the shear rate increasing from 0.1 to 1000 s^−1^ over 300 s at a temperature of 27 °C. Thixotropic behaviour was evaluated through a three-step shear recovery test at the same temperature: low shear rate of 10.0 s^−1^ for 120 s, high shear rate at 700 s^−1^ for 30 s and low shear rate of 10.0 s^−1^ for 120 s. Each test was performed in triplicate, and the average values were used to generate the corresponding curves. Data acquisition and analysis were carried out using TRIOS software (Version 5.7.2.101, TA Instruments, New Castle, DE, USA).

4.5. Mathematical Modelling of Flow Behaviour

The rheological data obtained from the shear-viscosity tests were analysed using two different tools: the proprietary TRIOS software (TA Instruments, New Castle, DE, USA) and the open-source Python package pyRheo (https://github.com/mirandi1/pyRheo, accessed on 30 July 2025). pyRheo is an open-source framework designed to assist in the selection and fitting of mathematical models for rheological data. It uses machine learning algorithms to identify the most suitable model for a given dataset and performs parameter estimation by fitting the experimental data to the selected rheological equation. Model fitting is based on minimising the weighted residual sum of squares using optimisation algorithms from the SciPy library. Up to three minimisation algorithms can be employed, and the package allows for different strategies to generate initial parameter guesses—either randomly within predefined bounds, through a Bayesian estimation method, or manually (the latter not recommended for initial fitting). The choice of starting parameters is critical, as poor initial guesses may lead to convergence to local minima or fitting failure; therefore, multiple random and Bayesian initialisations were explored in this study. Among the various rheological models implemented in pyRheo contemplates, five were selected for analysis, where $[eqn]$ is the shear rate and $[eqn]$ the apparent viscosity:

Herschel–Bulkley:

$$[eqn]$$

Bingham:

$$[eqn]$$

Power Law:

$$[eqn]$$

Carreu–Yasuda:

$$[eqn]$$

Casson:

$$[eqn]$$

For each of the six formulations, flow ramp data were collected in triplicate. Each replicate was first analysed independently in both TRIOS and pyRheo. In pyRheo, 128 different initial guesses were tested per replicate, combining three minimisation algorithms with both random and Bayesian initialisation strategies to maximise the likelihood of achieving optimal fitting. Consequently, each replicate yielded six fitting results per model. The best fit for each model was selected based on the coefficient of determination ( $[eqn]$ ), calculated according to the standard definition:

$$[eqn]$$

where $[eqn]$ is the experimental value, $[eqn]$ is the model prediction, and $[eqn]$ is the mean of the measured data. Because TRIOS does not use the standard definition of $[eqn]$ —and the software does not disclose the exact formulation it applies—we recalculated the $[eqn]$ values for all fitted models using the equation described above. This ensured full consistency across analyses, as both the TRIOS-generated fits and the Pyrheo fits were evaluated using the same $[eqn]$ definition. The best estimations from both tools (TRIOS and pyRheo) were compared for each replicate and model, and the superior fit was retained for further analysis. To obtain representative parameter values for each hydrogel and model, the viscosity data from the three replicates were averaged across identical shear-rate points to generate a “merged” dataset. This dataset was then analysed using three complementary approaches. First, the averaged data were fitted directly in pyRheo following the same fitting procedure applied to individual replicates. Second, the optimal parameters obtained from each replicate were averaged, and the corresponding $[eqn]$ value was recalculated for this averaged model to assess its fit quality. Finally, the merged dataset was analysed using TRIOS, and the resulting parameter estimations were compared with those obtained from pyRheo. For each hydrogel and model, the best estimation among these three approaches was selected for further analysis.

4.6. 3D Printing Process

The 3D-printed structures were fabricated using a syringe-based extrusion printer (bIDO-I, Idonial Technological Centre, Gijón, Spain). The corresponding 3D models (STL files) were created using Autodesk^®^ TinkerCAD™ (online CAD tool, https://www.tinkercad.com/, accessed on 13 May 2025) and Solidworks (version 2012). These models were then processed using the open-source slicing software Slic3r (version 1.2.9), which converted STL files to G-code format, compatible with the printer. Two different printing approaches were evaluated. The first consisted of direct semi-solid extrusion, in which the inks were preheated at their corresponding temperatures prior to printing to ensure appropriate viscosity and homogeneity. For formulations containing 10% polymer content, printing temperatures between 25 °C and 27 °C were evaluated, as a slight increase in temperature improved filament deposition. In contrast, formulations with 5% polymer content required lower printing temperatures due to their reduced viscosity; while PeGeCol_5 could be printed at temperatures down to 20 °C, PeGeLap_5 remained excessively fluid and was therefore excluded from further printing experiments. After printing, the structures were photocrosslinked using the ElastoSens^TM^ Bio apparatus (Rheolution Inc., Montreal, QC, Canada), UV–visible light sources at 365 nm and 405 nm (100% intensity, corresponding to 25.545 and 23.854 mW/cm^2^, respectively) for 30 s. The second approach was FRESH printing. In this method, a 25% (w/v) Pluronic F-127 solution in deionised water served as sacrificial support bath. The baths were prepared by dispensing 7 mL of liquid Pluronic into the macro-holder samplers of the ElastoSens^TM^ Bio apparatus. As in the previous method, all materials were preheated before the printing process started. After printing, the constructs were photocured for 2 min with the corresponding light source. Then, the samples were stored at 4 °C until the support matrix liquefied. Finally, the printing constructs were washed with cold DPBS for 10 min at 4 °C to remove residual Pluronic.

4.7. In Vitro Cell Viability Assays

In vitro biocompatibility studies were conducted using L929 mouse fibroblast cells (ATCC^®^ CCL-1™, American Type Culture Collection, Manassas, VA, USA) cultured on 3D-printed hydrogel discs. The selected formulation corresponded to PeGeCol_10_2, printed as disc-shaped scaffolds with a 15 mm diameter, 0.5 mm height and 80% infill density. The constructs were photocrosslinked for 30 s under 405 nm light at 100% intensity using the ElastoSens™ Bio apparatus (Rheolution Inc., Montreal, QC, Canada). Each disc was seeded with 25,000 L929 cells in 200 µL of complete culture medium. After seeding, the plates were kept inside a laminar flow cabinet for 30 min to allow cell attachment, followed by the addition of 400 µL of complete medium per well.

Four experimental conditions were evaluated: (1) non-washed printed discs, (2) discs washed for 24 h (24 h-W), (3) discs washed for 24 h, with a medium change after 8 h (24 h-W + MC), and (4) discs washed for 48 h under continuous immersion (48 h-W). Each condition was tested in duplicate. After photocrosslinking, hydrogel discs were carefully transferred individually from the printing plate to six-well culture plates. For washing conditions, each disc was immersed in 5 mL of culture medium (DMEM) per well, sealed with Parafilm and protected from light. For the 24 h-W + MC condition, the washing medium was completely replaced after 8 h with fresh DMEM. For the 24 h-W and 48 h-W condition, samples remained immersed under static conditions for the entire period. Non-washed samples were kept at 4 °C until further analysis.

Cell viability, proliferation, and cytotoxicity on the hydrogels were evaluated using the LIVE/DEAD^TM^ Viability/Cytotoxicity 488/570 kit (Invitrogen™, Thermo Fisher Scientific, Waltham, MA, USA) and the HS-alamarBlue^TM^ assay (Invitrogen™, Carlsbad, CA, USA) at day 0 (4 h after seeding) and day 7 of culture, following the manufacturer’s instructions. For the Live/Dead assay, cells were incubated with the dye at room temperature for 30 min. Fluorescence images were acquired using a ZEISS Axiovert 5 digital microscope (Carl Zeiss, Oberkochen, Germany). For the HS-alamarBlue assay, 10% v/v of the reagent prepared in complete culture medium without phenol red was added to each disc and incubated at 37 °C in a humidified atmosphere containing 5% CO_2_ for 3 h. Subsequently, 100 µL of the solution from each well was transferred to a 96-well plate. Absorbance was measured at 570 nm with 620 nm as a reference wavelength using a Multiskan™ FC Microplate reader (Thermo Fisher Scientific, Waltham, MA, USA). Cell viability was evaluated according to the widely accepted criterion that a reduction greater than 30% relative to control values indicates cytotoxicity.

4.8. Sample Lyophilisation

Lyophilisation of the printed samples was carried out using a Christ Alpha 1-2 LSCbasic lyophiliser (Martin Christ GmbH, Osterode am Harz, Germany). Prior to freeze-drying, the samples were ultra-frozen at −80 °C for 45 min. The lyophilisation process consisted of a primary drying step at 0.6 mbar for 2 h, followed by a secondary drying step at 0.1 mbar for 4 h. After completion, the samples were stored in the dark at room temperature until further use.

4.9. Nuclear Magnetic Resonance Analysis

The determination of unreacted PEGDA and LAP present in final samples was performed using proton nuclear magnetic resonance (^1^H NMR) spectroscopy (Bruker AV-600 spectrometer, Bruker Española S.A., Rivas-Vaciamadrid, Spain) at room temperature. Solvent peaks (HDO, D_2_O) were referenced to 4.7 ppm using Tetramethylsilane (TMS) as an internal standard. Quantification of PEGDA and LAP was achieved by adding 10 µL of t-butanol to each NMR sample as a reference compound. Photocured hydrogel samples (750 µL per sample), cured under 405 nm light at 100% intensity, were immersed in distilled water at room temperature for 24 h or 48 h and protected from light. Subsequently, the water used for hydrogel washing was collected, ultra-frozen with liquid nitrogen and lyophilised using a program consisting of a primary drying step of 14 h at 0.3 mbar followed by a secondary drying step of 4 h at 0.1 mbar.

4.10. Swelling Behaviour Analysis

Hydrogel samples (750 µL), prepared following the same protocol described for the NMR analysis, were weighed at time zero ( $[eqn]$ ) and then immersed in 5 mL of distilled water to evaluate their swelling behaviour over time. Samples were maintained at room temperature, and their mass was recorded after 8 h, 24 h, 48 h, and 72 h of immersion.

The swelling ratio (SR%) was calculated as follows:

$$[eqn]$$

where $[eqn]$ and $[eqn]$ represent the initial weight and the weight after water absorption, respectively. All swelling measurements were performed in triplicate using independent hydrogel samples.