The polymer blend of PLLA (M_v = 100,000 Da, DaiGang Biomaterial, Jinan, China) and PHBV (M_w = 520, 000 Da TianAn Biologic, Ningbo, China) at a mass ratio of 6:4 was electrospun into random fibrous mats (named random) and aligned fibrous mats (named aligned) by conventional electrospinning and stable jet electrospinning (SJES) methods (Yuan et al., 2012), respectively. In brief, 0.2 g of PHBV was added to 4.5 mL of trichloromethane (purity ≥99.0%, Changshu Yang-Park Chemicals, Changshu, China). The PHBV solution was subsequently heated to 65°C under stirring for 5 min, followed by adding 0.3 g of PLLA to obtain a homogeneous PLLA–PHBV solution. Last, a volume of 0.5 mL of N,N-dimethylformamide (purity ≥99.0%, Changshu Yang-Park Chemicals, Changshu, China) was introduced into the PLLA–PHBV solution to formulate the final polymer concentration of 10% w/v. Afterward, electrospinning was conducted as per the variables listed in Table 1. All the obtained fibrous mats were dried in vacuum at room temperature (23 ± 2°C).

Rectangular fibrous samples (named Original dimension: 20 × 10 × 0.06 mm) were divided into different groups and deformed by stretching in warm water (37°C and 46°C) to the predesignated strain (ε _deform = 30%, 50%, and 100%), followed by shape-fixing without releasing the applied stress in a 4°C refrigerator (thus, deformed samples were named deformed). The so-programmed fibrous samples can be actuated to recover in warm water (37°C and 46°C); thus, recovered samples were named recovered. Fiber morphology, structural analysis, and tensile properties of the original, deformed, and recovered samples were examined as follows.

Shape recovery ability under the variation of pre-designated programming parameters (i.e., T _prog and ε _deform shown in Figure 1) was assessed by DMA. In general, each test cycle was consisted of three steps in sequence: stretching to deform into a temporary shape at the deforming temperature, having the deformed shape fixed at low temperature, and applying thermal stimulus to trigger the shape recovery. Specific procedures for performing the shape memory tests regulated by T _prog and ε _deform are depicted as follows.

Shape memory tests regulated by T_prog: 1) At a high temperature T _high of 60°C, the fibrous mat sample was stretched using a strain ramp rate of 2%/min to 10% strain (ε _begin ); 2) the temperature was cooled down from T _high to T _prog (i.e., 37°C or 46°C) at 3°C/min; 3) at T _prog , the sample was stretched to 100% strain (ε _deform ) at a ramp rate of 4%/min; 4) at the 100% strain, the temperature was decreased to a low temperature T _low of 0°C at 3°C/min for vitrification; 5) at 0°C, the applied stress was released to 0, from which the fixed strain (ε _fix ) can be obtained; and 6) the recovered strain (ε _final ) was determined by reheating the temperature to the T _high of 60°C for activating shape recovery. Based on the generated strain−temperature curves, a particular T _sw can be determined by finding the inflection point (corresponding to the maximum recovery speed) through differentiation, and R _f and R _r were calculated according to Formulas 2, 3, respectively: R f % = ε f i x − ε b e g i n ε d e f o r m − ε b e g i n × 100 % ; (2) R r % = ε d e f o r m − ε f i n a l ε d e f o r m − ε b e g i n × 100 % . (3)

Shape memory tests regulated by ε _deform : 1) At T _high of 60 °C, the fibrous mat sample was stretched at 2%/min to 10% strain (ε _begin ); 2) the temperature was cooled down from T _high to T _prog of 37°C at 3°C/min; 3) at 37°C, the sample was stretched using a strain ramp rate of 4%/min to ε _deform of 30%, 50%, or 100%; 4) at the same designated strain, the temperature was decreased to the T _low of 0°C at 3°C/min for vitrification; 5) at 0°C, the stress was released to 0, from which the fixed strain ε _fix was obtained; and 6) the sample was reheated to T _prog of 37°C and then maintained for ca. 3 h, during which a plot depicting the kinetics of deformed strain versus time can be generated. Normalization to 100% strain percentage was performed for the cases where ε _deform =30%, 50%, and 100%.

Similar to the procedure described in 2.3, the fibrous mat samples deformed to the predesignated strain (ε _deform = 10–100% at a 10% interval) were prepared in 37°C warm water, followed by shape-fixing in a 4°C refrigerator. Thereafter, upon being triggered to recover in 37°C warm water, the maximal recovery stress σ _max of the deformed samples with their two ends fixed by clamps can be measured using a 5-N force gauge.

Quantitative data are generally presented as mean ± standard deviation and checked by normality tests. Statistical analysis was performed using the Origin software (OriginLab, Northampton, MA, USA). One-way ANOVA with post hoc Tukey’s HSD test was used to make pairwise comparisons between groups. *p < 0.05, **p < 0.01, or ***p < 0.001 is considered to be statistically significant.

Both random and aligned fibers of the PLLA–PHBV blend with comparable diameters (1,274 ± 209 nm vs. 1,207 ± 370 nm) could be readily produced using the conventional electrospinning and SJES methods, respectively (Figures 2A1, B1). The FFT output images (graphical depiction of FFT frequency, Figures 2A2, B2) and the derived pixel intensity plots (Figures 2A3, B3) indicate that, compared to the random group, a distinctly higher degree of fiber anisotropy is noted in the aligned group by showing concentrated pixels mainly in a cross-like style and a sharp peak (Ayres et al., 2006; Ayres et al., 2007). At the molecular level, the P-FTIR results (Figure 2C) show strong molecular orientation of the polymer chains within the aligned fibers, as there were pronounced differences in absorbance intensity in the directions parallel and perpendicular to the fiber alignment. This is true either for the characteristic C-O-C symmetric stretch at 1,086 cm⁻¹, OC-O asymmetric stretch at 1,179 cm⁻¹, and C=O vibration at 1,759 cm⁻¹ in PLLA (Zhang et al., 2010) or the C=O stretch at 1,722 cm⁻¹ ascribed to PHBV (Furukawa et al., 2005). However, the polarized measurements in the ⊥ and // directions were almost identical in the case of random fibers. The quantitative dichroic ratios (D _r ) with respect to the aligned group, showing values of higher deviation from the data of 1.00 (Supplementary Table S1), support the observed enhancement of molecular orientation in the aligned fibers (Chan et al., 2009; Wang et al., 2006). Such an enhancement in molecular orientation can be attributed to the SJES technique itself (Yuan et al., 2012; Zhang & Zhang, 2016) and the strong stretching exerted by the high-speed rotating drum. A higher extent of molecular orientation benefits the formation of more crystalline structures (Figure 2D). This was also evidenced by the appearance of reflection peaks at 2θ = 16.9 and 2θ = 13.6, corresponding to the (110) and (020) crystal planes of PLLA (Sun, Yu, Zhuang, Chen, & Jing, 2011; Zembouai et al., 2013) and PHBV (Zembouai et al., 2013; Montanheiro et al., 2019), respectively. Obviously, the characteristic peak intensities in the XRD pattern of the aligned group are higher than those in the random group. As anticipated, due to the macroscopic and molecular orientation, aligned fibers exhibited superior mechanical properties (Figures 2E–H) than the random counterpart. Despite a slightly lower fracture strain, the aligned fibers are about 4-fold stronger in tensile strength (i.e., 8.5 ± 1.1 vs. 2.1 ± 0.1 MPa, p < 0.001) and Young’s modulus (i.e., 397.4 ± 56.3 vs. 98.8 ± 7.6 MPa, p < 0.001) than the random fibers.

Thermal-responsive SMPs usually rely on the thermal transition (e.g., glass transition temperature T _g ) to determine the temperature window for shape-programming. Figure 2I shows the tan δ−temperature curves of the random and aligned fibers non-isothermally scanned via DMA. Clearly, either the random or the aligned fibers exhibited a pretty broad thermal transition region with the maximum peak at ca. 65°C, which can be ascribed to T _g of the dominant constituent PLLA (T _{g, PLLA} ) within the PLLA–PHBV blend (Nanda et al., 2011; Jaszkiewicz et al., 2013; Li et al., 2015). A further subtle scrutinization below T _{g, PLLA} revealed the presence of an indistinct peak around 18°C attributed to T _g of PHBV (T _{g, PHBV} ) (Nanda et al., 2011; Jaszkiewicz et al., 2013) and a relaxation process with the peak maximum at 43.7°C in the range from 36°C to 54°C, indicative of T _g of the mixed phase (T _{g, mix} ) (Nanda et al., 2011), especially in the case of random. Although the much higher T _{g, PLLA} does not allow adaptation of T _sw for shape recovery up to the physiologically relevant temperature, choosing T _prog in the vicinity of T _{g, mix} of PLLA–PHBV would afford an alternative strategy to easily adjust T _sw due to the well-documented temperature-memory effect (TME) (Miaudet et al., 2007; Kratz et al., 2011; Kratz et al., 2012). In this sense, 37°C and 46°C (slightly above body temperature) in the range of ΔT _{g, mix} for the mixed phase (i.e., 36–54°C) can be chosen as T _prog for our subsequent shape-memory response studies.

For shape-memory polymer fibers, stretching deformation to generate a temporary shape usually results in a reduction in fiber fineness together with improved fiber orientation along the loading direction; and upon being triggered for shape recovery, the morphological features of the original fibrous structure can be generally recovered (Figures 3A, B). This was verified in our current study by varying T _prog and ε _deform . As shown in Figures 3C, D summarized from the SEM quantification data (Supplementary Figures S1A–F), under the deforming conditions of stretching in warm water (37°C and 46°C), for instance, 100% of ε _deform , it gave rise to fiber diameter attenuation by 16–21% in random, slightly larger than that of aligned (15–18%). Also, a higher magnitude of improvement in fiber orientation was observed in the random group (16–18%) than in the aligned group (<1%). In terms of shape recovery, while the fiber thickness and orientation features of the aligned fibers could be largely reverted, small differences in morphological restoration can be noted in the random fibers.

Mechanistically, apart from the macroscopic differences in their fibrous architectures (Figures 2A, B), the initial status of molecular orientation (Figure 2C) and the stretching deformation resultant enhancement in crystallization may also be responsible for the different responses in the previously noted morphological changes. Taking the stretching deformation for shape-programming at 37°C for illustration (Figures 3E, F), due to the strain-induced crystallization effect (Tabatabaei et al., 2009; Tcharkhtchi et al., 2014; Nissenbaum et al., 2020), the increments of polymer crystallites formed in overall crystallinity at 100% of ε _deform could be increased by 22.9% and 110.9% (both mainly contributed by the PLLA crystalline phase, Table 3) for the random and aligned fibers, respectively. As a result, it also gave rise to the progressively strengthened elastic modulus and stiffness (Figures 3G–I; Supplementary Figure S2) if increasing ε _deform of both the random and aligned fibers during the stage of shape-programming. Collectively, these results demonstrated that the morphological changes were not dramatically influenced by T _prog but responded more to changes in ε _deform . The stretching deformation and recovery process could, therefore, be applied to control the diameter and orientation of the shape-memory-capable electrospun fibrous structures, which may provide a dynamic topography for modulating the cellular behavior (Ahn et al., 2014; Madl et al., 2021; Song et al., 2013).

TME is the capability of a shape-memory polymer to memorize its thermomechanical history, in particular the T _prog at which it was deformed before, by reversing the deformation at a characteristic T _sw roughly identical to the previously applied T _prog . While the TME concept has been previously demonstrated in different material systems (Cui et al., 2011; Kratz et al., 2011; Miaudet et al., 2007; Xie et al., 2011), how well T _sw coordinates with T _prog in SMPfs remains to be understood. It has been known that temperature memory polymers usually exhibit a broad thermal transition temperature region ΔT _trans within which the T _prog applied for deformation can be memorized (Cui et al., 2010; Cui et al., 2011) during a specific SMCP. Based on the identified ΔT _{g, mix} for the mixed phase (Figure 2I), the effects of applying T _prog with physiological relevance (37°C and 46°C) at 100% of ε _deform on T _sw , R _f , and R _r were first examined (Figure 4A). A quantitative summary on the comparison of shape-memory properties between random and aligned (Table 4) indicates an impressive fixation of the deformation with R _f > 95.0, while R _r > 70% could be obtained. It is to be noted that the characteristic T _sw , defined by the temperature with the maximum recovery speed under stress-free conditions, was always slightly higher than T _prog (Figures 4B, C). As a proximity indicator, the much lower T _sw /T _prog ratios of 1.02–1.06 compared to other forms of SMPs (e.g., films) with T _sw /T _prog ≈ 1.3 at 37°C reported in the literature (Cui et al., 2010) suggest superior thermal sensitivity of the fibrous structures due to their intrinsic attribute of high specific surface area (Lu et al., 2017). When making a comparison between random and aligned, the characteristic T _sw is closer to the applied T _prog in aligned. This can be ascribed to the better thermal conductivity in aligned fibers with structural features that include not only the high degree of macroscopic fiber alignment (Figures 2A, B) but also the increased molecular orientation and crystallinity (Figures 3E, F) (Shen et al., 2010; Lu et al., 2017; Bai et al., 2018). These results confirmed that the characteristic T _sw of TME can be more accurately correlated with T _prog for shape-memory polymer fibers, which could therefore permit a facile adjustment of T _sw according to specific biological application requirements (Kratz et al., 2012).

To examine whether varying ε _deform (30%, 50%, and 100%) would regulate the recovery ability of the two types of fibrous structures at body temperature, a further comparison between random and aligned fibers was then performed (Figure 4D). As shown in Figures 4E, F, depicting the recovery kinetics for random and aligned which were prior-deformed at different ε _deform , and upon being triggered at T _prog of 37°C, the shape recovery events can be activated for both. Notably, occurrences of the maximum recovery rates determined by differentiation to the normalized ε _deform curves in aligned fibers were shorter than those in random ones (aligned 8.5–10.7 min vs. random 9–11 min), and the recovered strain of aligned fibers during the testing time frame was always higher than that of random counterparts at the same ε _deform . Again, the observed better shape recovery performance in aligned fibers can be attributed to its high thermal conductivity and the underlying conducive molecular structures in connection. Consistent with previous investigations (Wong and Venkatraman, 2010; Wu et al., 2018; Zhao et al., 2017), in both cases, fibrous structures with increased ε _deform during the SMCP gave rise to decreased shape recovery efficiency.

The shape recovery process upon being triggered is usually accompanied by the release of the strain energy stored during deformation, which accordingly enables the generation of recovery stress (σ _rec ) and other shape-memory responses. To compare the impact of varying ε _deform on the generated σ _rec between random and aligned, a pair of custom-made grips was used to fasten the fibrous mat sample for σ _rec measurement under constrained recovery condition (Figure 5A). This allowed producing a stress versus time curve showing a maximum stress point (σ _max) (Figure 5B). The presence of such a peak recovery stress under constrained recovery condition has also been observed previously for other SMPs (Cui et al., 2010; Kratz et al., 2011; Kratz et al., 2012). By plotting the σ _max versus ε _deform graphs (Figures 5C, D), it was revealed that σ _max could be systematically adjusted by variation of ε _deform , and the σ _max versus ε _deform correlations for random and aligned could be mathematically well-fitted in polynomial and linear regressions, respectively. As the stress generated by shape recovery is a growing function of the strain energy stored during deformation at a high temperature (Miaudet et al., 2007), it is reasonable to observe continuous increases in σ _max with increasing ε _deform . In comparison, σ _max generated by aligned is significantly higher than that generated by the random counterpart. Specifically, when the two types of fibrous mats with 30%, 50%, and 100% ε _deform were recovered in 37°C warm water, the σ _max generated from aligned was correspondingly 3.4-fold, 2.1-fold, and 2.3-fold stronger than that generated by random.

For the thermal-responsive SMP systems on the molecular level, it is the netpoints (e.g., crystallites, rigid segments, or chemical crosslinks that determine the permanent shape) and the switching segments (e.g., amorphous fractions in a polymer acting as a reversible phase) within the SMPs that constitute the two structural elements responsible for the SME (Salaris et al., 2022). Also, applying a stretch deforming strain ε _deform at a high temperature (i.e., T _prog ) and then cooling down cause the polymer chains to become oriented and trapped in the temporarily fixed shape, thus storing strain energy. When being reheated in the vicinity of T _trans (e.g., T _g ) associated with the switching domains, the polymer chains become mobile enough to move back to their random coil-like conformation, thus leading to recovery of the original shape. Since the σ _max generated during the shape recovery process is associated with the netpoint/switching segment density or hard segment weight content (Cui et al., 2010; Kratz et al., 2012) and considering the fact that increasing ε _deform by stretching gave rise to the polymer chains highly oriented (Yano et al., 2012; Youm et al., 2016) accompanying with the formation of more crystallites (Figure 3F), the observed higher recovery stress in aligned than in random can be reasonably explicated as illustrated in Figure 5E.

The previously presented results suggest that the TME allows us to judiciously choose T _prog from ΔT _trans for shape-programming and triggering shape recovery at T _sw with physiological relevance (e.g., 37°C) and that the shape recovery stress generated under constrained recovery condition may be utilized to endow the fibrous scaffolds with mechanoactivity for in situ modulating cellular behavior (e.g., osteogenic differentiation of stem cells). To prove this concept, an aligned fibrous mat was chosen for stretch deforming the aligned fibers to 10% of ε _deform (without affecting cell survival (Altman et al., 2001)) at T _prog of 37°C, from which a mechanically dynamic culture system based on the SME-enabled mechanoactive fibrous scaffold can be constructed to direct the fate of BMSCs (Figure 6A). Cytocompatibility assessments based on the planned experimental scheme (Figure 6B) indicated that both the plain aligned 0% and the mechanoactive aligned 10% could support BMSCs to attach and grow along the fiber direction (Figure 6C), and the cell proliferation capacities of both were comparable during the 12 days of observation (Figure 6D). These results suggest that the integrated dynamic shape-programming and recovery processes during cell culture had no detrimental effect on cell growth, which is in good accordance with previous observations (Guo et al., 2022; Tseng et al., 2016; Xing et al., 2016).

Both the ALP staining and calcium quantification assay were performed to evaluate the efficacy of using the mechanoactive scaffold for promoting osteogenic commitment of BMSCs in the absence of any osteogenic induction factors. Obviously, the cells on the mechanoactive aligned 10% stained positively for endogenous ALP activity, whereas the expression of ALP in cells cultured on the aligned 0% was comparably quite low during the 21 days of culture (Figure 6E). The quantified content of calcium deposits, as another key indicator of osteogenic differentiation of BMSCs, similarly revealed statistically significant higher levels of calcium production in the aligned 10% group than in the aligned 0% group after 14 and 21 days of culture (Figure 6F). At the gene level, expressions of representative osteogenic markers including Runx2, Ocn, and Alp were also detected (Figures 6G–I). The mRNA expression patterns generally show that while the early time of expression (e.g., at days 1 and 8) is comparable, higher levels of markers at day 21 can be detected. These results consistently corroborated that applying in situ mechanical cues provided by the SME-enabled mechanoactive scaffold per se indeed resulted in achieving enhanced efficiency in directing osteogenic differentiation of stem cells.

In bone tissue engineering, applying appropriate means to efficiently drive osteodifferentiation of mesenchymal stem cells (MSCs) is crucial to osteogenesis. While utilization of various potent osteogenic factors (e.g., bone morphogenetic protein, dexamethasone, and transforming growth factor) has been demonstrated to be effective, given the mechanosensitive nature of bone, an attractive strategy for regulating osteogenic differentiation of MSCs has been the use of mechanical stimuli. Amongst, substrate stiffness has been identified to play a key role in directing stem cell osteodifferentiation (Gazquez et al., 2016; Lee et al., 2018; Song et al., 2021). In our current study, since shape recovery under constrained recovery condition can generate a stress-stiffening effect in the fibrous substrate (Guo et al., 2022), directing stem cell differentiation into an osteogenic lineage based on the SME-enabled mechanoactive scaffold itself, rather than using external bioreactors for exerting mechanical stimulus, was proved to be successful. The feasibility demonstrated here provided a fresh paradigm for developing mechanically active scaffolds for regulation of osteogenic differentiation in MSCs without using any biological supplements and/or bioreactors to provide osteogenic induction cues.