Synthesis of INU-PLA copolymers
The synthesis of INU-PLA copolymers was carried out in very mild conditions (Fig. 1). By using CDI as coupling agent, the imidazolide derivative of PLA (PLA-im) can be obtained  and used in one pot alcoholysis by adding INU, previously treated with a base. CDI has been preferred over other activating agents, such as DCC, because pure water hydrolyzes it at room temperature in short time, with evolution of CO2, and the reaction gives imidazole as the sole side product. The water solubility and non-toxic nature of such compound represent a good opportunity for a green option, coherently with the choice of INU as a renewable substrate. The purification from unreacted PLA was carried out by means of precipitation and washings in diethyl ether (EtOEt). Since PLA5k has a lower solubility in EtOEt compared to PLA1k, 15% of acetone was added to completely eliminate any unreacted PLA. Considering earlier literature  and predictable reactivity, it is likely that functionalization occurs at the primary hydroxyls of inulin with the activated end-carboxylic acid group of PLA.
Four derivatives, named INU-PLA1-4, were synthesized by using two different PLA with Mn of 1041 Da and 5017 Da, here named PLA1k and PLA5k, respectively. In turn, those two polyesters were grafted onto INU by applying two different molar ratios between PLA and INU repeating units (R1), i.e., 0.06 and 0.12. The reaction conditions and the names attributed to the four obtained derivatives are summarized in Fig. 1.
Characterization of INU-PLA copolymers
Products obtained after purification were analyzed by mono- and bidimensional 1H, 13C NMR analysis in DMSO-d6. 1H NMR analysis spectra, reported in Fig. 2, showed the coexistence of the characteristic shifts of both INU and PLA and allowed to calculate the molar degree of derivatization (DDmol%), i.e., the average mmoles of PLA conjugated every 100 mmoles of INU repeating units.
DDmol% was calculated by comparing the integral of the signals comprised between 1.15 and 1.78 ppm accounting for the PLA methyl groups, with the integral of the signals comprised between 3.38 and 4.14 ppm assigned to the protons of inulin repeating unit with exception of hydroxyls (7H). DDmol% of the obtained copolymers are reported in Table 1.
Noteworthy, comparison of the 1H NMR spectra of PLA and PLA grafted INU showed that the signals for the 3 protons of the methyl group adjacent to the PLA terminal carboxylic group (HOCH(CH3)C(O)O-PLA-CH(CH3)C(O)OH) undergo a shift, from ~ 1.37 to ~ 1.3 ppm, as a result of esterification (Fig. S1).
2D NMR spectra allowing direct or long-range correlation between 1H and 13C signals provided many useful information, especially when signals are overlapping, as often in polymer spectra, or in cases where there are characterizing heteroatoms or carbon atoms not bearing a bounded H. Figure 3 presents the 1H-13C HSQC and 1H-13C HMBC spectra of INU-PLA1.
1H-13C HSQC spectrum of INU-PLA showed the primary correlations from INU and the PLA side chains. In particular, the strong δC/δH correlations at 16.1/1.45, 19.9/1.29, 65.3/4.21, and 68.3/5.21 ppm were assigned to CP3/HP3, CP6/HP6, CP5/HP5, and CP2/HP2, respectively. For INU 1H-13C HSQC and full attribution of INU-PLA correlations, please refer to Fig. S2 and Table S1 in the Supplementary Materials. Moreover, new δC/δH correlations (highlighted in green in Fig. 3) at 65.3/5.16 and at 28.4/1.24 ppm appeared in the 1H-13C HSQC of INU-PLA that cannot be assigned to native polymers. Those correlations were assigned to the CH2 of INU-PLA in position 6′ adjacent to the substituted primary OH and to the CH3 of the corresponding substituted end of the PLA chain (P3′) respectively showing that PLA side chains were successfully grafted onto INU.
Observing the 1H-13C HMBC spectra of INU-PLA (Fig. 3), it is possible to appreciate specific correlations by atoms involved in the new formed bonds between INU and grafted PLA, along with the characteristic correlations between signals of the two blocks. In particular, the signal of the protons at position 6′ shows long-range correlations with the signal of the carbon P1′ δC/δH at 161.6/5.16, and the same protons were found to correlate with another carbon atom which in turn correlated with H4 and, therefore, recognized as the C5′. The two signals are highlighted in green in Fig. 3 with δC/δH at 77.9/5.16 and at 77.9/3.79 ppm, respectively. For INU 1H-13C HMBC and full attribution of INU-PLA correlations, please refer to Fig. S3 and Table S2 in the Supplementary Materials.
Both raw INU and PLA and INU-PLA 1 to 4 were characterized by FT-IR spectroscopy (Fig. 4). The characteristic bands around 3300, 1027, and 934 cm−1 were assigned to INU backbone. The FT-IR spectrum of INU-PLA also revealed an intense absorption band from the carbonyl group (C = O) at 1748–1749 cm−1, confirming the presence of the PLA polymer. The latter band appeared shifted of 5 to 6 cm−1 if compared to the same band in raw PLA. The band for O–H stretching vibration band was shifted to a higher wave-number domain, probably due to the decreased hydrogen bonding force in INU after partial modification of its hydroxyls with PLA.
INU-PLA copolymers were also characterized by gel permeation chromatography (GPC) in DMSO at 60 °C. GPC traces are shown in Fig. S4. The elution peaks showed the complete removal of unreacted PLA homopolymer and further confirmed the successful grafting of PLA onto INU. More importantly, the products appear narrow dispersed with a PDI within 1.5, thus showing a good uniformity of the functionalization along the molecules of the carbohydrate. An exception is represented by INU-PLA4 for which a PDI of 1.7 was found.
Substances with amphiphilic properties have often the advantage to be soluble in a wider variety of aqueous and organic media compared to isolated blocks. This potentially improves their workability in a bigger scale and broads the solvent selection during the preparation of drug loaded nanoparticles, taking into account toxicity and polymer/drug solubility aspects. INU has a limited dispersibility in organic solvents, being DMSO almost the only one in which the oligo-fructans are freely soluble at high concentration at room temperature. Despite the high hydrophilicity, the solubility of INU in water at room temperature is poor as well, being impaired by the dense network of H-bonds between sugars.
The affinity of the amphiphilic INU-PLA copolymer with organic solvents compared to INU increased. This was shown by estimating the amount of DMSO needed to dissolve a suspension of INU-PLA1-4 in a fixed amount of acetone (a very poor solvent for INU) (Fig. 5a–b). Increasing both DDmol% and the molecular weight of the grafted PLA, the affinity for acetone increased, with INU-PLA4 needing only 0.15 ml DMSO, i.e., the 3.7% of the DMSO needed to bring INU in solution from acetone dispersion (4.1 ml). Noteworthy, grafting of PLA1k drives the copolymer towards a full dispersibility in water at room temperature (Fig. 5c). By repeating the same experiment but starting from water instead of acetone, the amount of DMSO needed to solubilize the solid decreased in the order INU-PLA4 > INU-PLA3 > > INU-PLA2 = INU-PLA1 (Fig. 5d), with INU-PLA1 and INU-PLA2 no needing DMSO to be added, but gave place to a homogeneous, clear dispersion at once after water addition.
Thermal analysis of INU-PLA copolymers
Thermal analysis allows correlation of thermal properties of copolymers (crystallinity, Tg, decomposition temperature) with drug release behavior, stability of formed micelles, and affinity with a drug. Thermograms of unmodified INU, INU-PLA copolymers, and PLA, obtained by DSC, are presented in Fig. 6. To remove residual solvents and erase the thermal history of the sample, results of second heat cycle were mostly taken, except for neat PLA and INU-PLA2, for which second-order transitions were appreciable only at higher heating rate (30 K min−1) and in the first step (thermograms obtained by DSC of PLA1k and PLA5k recorded at the first heating cycle are reported in Figs. S5 and S6).
In the current work, the inflection point instead of midpoint was chosen for the determination of the glass transition temperature (Tg), to minimize the onset-/endset-related variation. No melting temperature (Tm) was observed in the curve of unmodified PLA1k and PLA5k, suggesting that the polyesters were amorphous. This was predictably ascribed to the random distribution of D- and L-chiral centers along the polymer backbone, which disturbs the regular crystalline lattice . INU exhibited, instead, a Tm and a Tg of approximately 174 °C and 93 °C, respectively, revealing the semicrystalline nature of the material. For the INU-PLA copolymers, a reduced crystallinity was found, and each curve exhibited a single Tg. Thus, modification resulted in a material with thermoplastic behavior and a reduced Tg to about 40–50 °C (Table 2). For details of DSC curves showing Tg of INU, PLA, and INU-PLA1-4, please refer to Figs. S5–S8). This indicates two possible phenomena, occurring simultaneously: first, that grafted PLA side chains expanded the distance between polysaccharidic chains and increased their degree of freedom, thus playing a plasticizer-like role (this effect was previously observed by Zhang et al. for xylan-g-PLA ), and second, the breaking of non-covalent interactions (hydrogen bonding) within the dangling fructose units due to functionalization of hydroxyls led to augmented mobility of polymer chains, thus decreasing the Tg. This is substantiated by the shift of the hydroxyls’ absorption band around 3300 cm−1 in the ATR FT-IR spectra of INU-PLA copolymers compared to raw INU (compare Fig. 4). Changings due to the lost capacity of fructose to H-bond one another are not new in INU chemistry: it is frequently seen that after functionalization, even with hydrophobic small molecules, the dispersibility in water at room temperature increases when compared with raw INU, which is poor in water at room temperature .
Noteworthy, when PLA is grafted onto INU, the degradation of the polyester chain occurred at a higher temperature: while degradation peak was observed between 294 and 301 °C for neat PLA or PLA in the physical mixture with INU, in the case of grafted PLA, an increase up to 40 °C was observed. Data, the whole set of which is reported in Table 3, indicate the improvement in thermal stability, caused by the modification of carboxyl functional end group by grafting onto INU. An increase in the degradation temperature is compatible with the modification of the end group of a PLA chain as described earlier . These results suggest that INU could interfere with one of the chemical processes involved in the thermal degradation of PLA, such as intra- or intermolecular transesterification, cis-elimination, and radical and concerted nonradical reactions .
The compartmentalization of new amphiphiles: preparation and characterization of INU-PLA-based nanosystems
Once the amphiphilic copolymers were obtained and characterized, the next step was to study their ability to self-assemble into nanostructured systems in aqueous environment and to load doxorubicin (Doxo) as model anticancer drug.
Firstly, a comparison between the 1H NMR spectra of INU-PLA in D2O and in DMSO-d6 was carried out. Fig. S9 shows the spectra of INU-PLA2 in both solvents. The analysis of both spectra revealed that the signals which belong to -CH3 and -CH protons of PLA grafted chains have a lower resolution in D2O in comparison with the corresponding signals detected in DMSO-D6. Indeed, the ratio between the integrals of the signals of INU backbone and the signal of PLA methyls reduced from 0.45 in DMSO-D6 to 0.19 in D2O. Such an attenuation of signals together with disappearance of the splitting in 1H NMR spectra indicates that the hydrophobic part of the graft copolymer has a lower mobility. The reduced degree of freedom is suggestive of the PLA chains being in a confined structure such as the inner core of a water dispersed colloid [29, 30].
Among different methods used to produce nanoparticles from amphiphilic copolymers, two of them were selected to explore the behavior of INU-PLA1-4 copolymers in water and to optimize the preparation of drug-loaded nanoparticles: the thin film rehydration method and the nanoprecipitation method. The methods are schematically represented in Fig. 7.
In the thin film rehydration method, the addition of water to the casted thin film of INU-PLA3 and INU-PLA4 caused the detachment of large film fragments, and only 10 to 30% of the starting weight was recovered after prolonged sonication, filtration, and freeze-drying. On the contrary, INU-PLA1 and INU-PLA2 were readily dispersible giving a clear phase, and the yields of the obtained solids were calculated to be over 85%. With the aim to characterize the morphology and dimensions of the obtained systems, freeze-dried samples were visualized by SEM and analyzed by DLS after dispersion in double distilled water and filtration over 0.45-µm syringe filters. The two analyses gave different results: while DLS of filtered samples revealed homogeneous population of particles in the range 160–220 nm with a zeta potential in the range of − 18 to − 31 mV in all the cases, and on the other side, at the SE microscope, a completely different scenario emerged. As can be seen in Fig. 8a and b, the solid appeared mostly formed by nonuniform microparticulate, showing a very high heterogeneity of samples produced by the thin film formation-rehydration method. Surprisingly, rehydration of INU-PLA4 from the film gave rise to stable giant polymersomes, a fraction of them still integer and not collapsed after freeze-drying (Figs. 8b and S8). Detailed information on the size of systems prepared by both the thin-film hydration and nanoprecipitation methods at different stages of particle preparation are reported in Table S3.
It is already known for block copolymers that they can self-assemble from spheres to polymersomes by control over the amphiphilicity. The key to this phenomenon is the hydrophobic/hydrophilic ratio. Here, the molecular weight ratio of hydrophobic to hydrophilic segments (fH/H) for INU-PLA1-4 and hydrophilic-lipophilic balance (HLB) is reported in Table 4. The lower the hydrophilicity, the lower is the interfacial curvature of the copolymer, which pushes the assembly towards vesicles instead of micelles or nanoparticles . Therefore, it is not particularly surprising that in our case (even if INU-PLAs are graft and not block copolymers), once the right fH/H is reached, a change in the morphology of the obtained system occurred. This was not observable by changing the method to nanoprecipitation. Indeed, in this second case, SEM analysis showed in all cases population of small nanoparticles with diameter around 70 nm (Fig. 8c and d). Moreover, the obtained yields were high, and after DLS analysis, we found results in accordance with the two techniques.
The above results indicate that in conditions where the formation of lyotropic phases (such as a lamellar phase) occurs after hydration (such as in the case of film rehydration method), the hydrophobic PLA segments in INU-PLA interact and freeze in a local minimum of energy from where bending and fitting in the interior of a core/shell nanoparticle is highly disfavored, which is why the copolymer tends to form bigger particles and vesicles, such as polymersomes in case of INU-PLA4 which has the highest fH/H.
On the contrary, during nanoprecipitation, a process that allows the formation of the nanosystem via nucleation in dependence on the balance between the assembled state and unimers in dispersion, the final architecture is not matured from an intermediate state, but it is “trapped” in the nucleus shape, which is maintained during growing eventually . Information on the size of systems prepared by both the thin-film hydration and nanoprecipitation methods at different stages of particle preparation is reported in Table S3.
HLB values determined for INU-PLA1-4 follow in the range 6–15. Some ester derivatives of inulin [32,33,34] with HLB values comprised between 7 and 17 have been found to be good emulsifiers, potentially employable in pharmaceutical as well as in other technological field such as cosmetics, food chemistry, detergents, and bioplasticizers.
INU-PLA-based nanosystems as drug carrier
Following the above preliminary results, we chose to use nanoprecipitation method for further investigation on drug loading ability of the nanosystem. Thus, we explored the effect of the method parameters (polymer concentration, theoretical drug loading, and homogenization method) on the size distribution and actual drug loading (DL%).
In a first set of experiments, we used INU-PLA1 and the homogenization by ultra turrax at 0 °C (to avoid excessive foaming and to contrast heat production due to metallic part rotation). The polymer concentration in the feed solution ([INU-PLA1]feed) was changed in the range 10–50 mg/ml, and the hydrodynamic diameter was followed by DLS measurements of samples redispersed in water after freeze-drying. Changing [INU-PLA1]feed appeared to have little or no influence on the hydrodynamic diameter (Fig. 9a). The only exception occurred when the concentration was below 20 mg/ml: in that case, the size statistically decreased by about 50 nm, and a bimodal distribution characterized the intensity vs size plot, with the appearance of a narrow population around 20 nm (Fig. 9b).
Under the same conditions, Doxo was added to the feed polymer solutions at a Doxo/INU-PLA1 of 20% wt/wt. Although the effect of Doxo encapsulation on particle diameter appeared negligible in the range 20–40 mg/ml, the repercussion of polymer concentration on drug loading seems not to be trivial, with a surprising decrease in DL% with increasing [INU-PLA1]feed from 1.90 to 0.87% (Fig. 9c). By keeping fixed the [INU-PLA1]feed (40 mg/mL), and increasing the Doxo/INU-PLA1, a progressive increase in the amount of drug entrapped was found. In particular, results show that an increase in Doxo/INU-PLA1 feed from 5 to 20% by weight led to an increase in DL% from 0.1 to 1.4% (Fig. 9d). Clearly, higher theoretical loadings correspond to a greater number of drug molecules available to be trapped inside particles, leading to the greater DL% observed.
Using vortexing at room temperature as a method of homogenization, a sharp increase in drug loading was achieved up to 5%. The reason why switching from ultra turrax to vortex cause such an increase in the ability of the system to entrap the drug could be ascribed to the different rate Doxo is able to diffuse in the water phase while the nucleation occurs (higher for ultra turrax), as well to the higher temperature during vortexing. The data also showed that INU-PLA is able to load Doxo to the same extent as PEG-PLA. In particular, for INU-PLA3 and PEG-PLA having the same fH/H of 1.1, we found a DL% of 4.7 ± 1.8% and 5.6 ± 1.3%, respectively. Results are summarized in Fig. 10a.
The dispersion of the drug in the hydrophobic core alters to some extent the overall hydrophilic-hydrophobic balance [35, 36]. The effect increases with increasing of the drug loading, and it is reflected on the size difference between empty and loaded nanosystems. For systems obtained starting from INU-PLA1 and INU-PLA2, in which the amount of Doxo entrapped is averagely 24.4 ± 4.2 µg/mg and 20.3 ± 2.7 µg/mg, respectively, the difference between empty and loaded particles is not statistically relevant. Going to INU-PLA3 and INU-PLA4, which were able to entrap almost double the amount of Doxo compared to the analogues 1 and 2 (47.0 ± 18.2 and 44.6 ± 16.0 µg/mg, respectively), a more pronounced effect on the drug dispersed in the core was observed as a difference in size between empty and loaded systems (Fig. 10b).
While the dispersed drug has stabilizing activity in the case of INU-PLA graft construct assembly, the contrary occurred for PEG-PLA, in which case the colloidal stability was almost lost after drug incorporation and freeze-drying (Fig. 10c). The better re-dispersibility of INU-PLA, without the need of a cryoprotectant, represents an advantage in the processability and translatability of the systems. It is not surprising that INU gave such an effect, since it is already known its ability to act as a cryo- and lyoprotectant .
Differential scanning calorimetry (DSC) showed that Doxo is in the nanosystems in a non-crystalline form, as no melting transition characteristic of Doxo could be measured, suggesting that the drug is molecularly dispersed in the systems (Fig. S10). Moreover, we cannot exclude the drug that is partially adsorbed on the outer hydrophilic shell, being the zeta potential after dialysis purification and freeze-drying slightly less negative for Doxo loaded INU-PLA nanoparticles compared to empty ones (Fig. 10d).
Surprisingly, when the INU-PLA copolymer was used in the formation of particle systems (INU-PLA/Doxo), a double second-order transition corresponding to two different Tg were observed: one at about 50 and one at about 70 °C (Fig. S11). The phenomenon, although not entirely clear, is suggestive of a segregation between the two polymer components and has been observed as a consequence of nanoparticle confinement in core–shell systems [38, 39].
Finally, in vitro Doxo release was evaluated in TRIS buffer (5 mM, pH = 7.4) at 37 °C. During experiments aiming to estimate the amount of free Doxo diffused with time, only around 75% of drug was released after 24 h, and no further release was detected by UV–VIS analysis (Fig. S12). This behavior was due to the degradation of Doxo as well as its retention after membrane adsorption. As a consequence of its poor stability, once released from the nanoparticles, Doxo was indeed partially degraded in the release medium, thus making analysis impossible for the longer time points. Within 24 h, it is possible to observe a different release kinetic for INU-PLA1-4 (Fig. 11).
In particular, increasing the fH/H, the Doxo release slowed down with the maximum amount of Doxo released after 24 h being that of INU-PLA1/Doxo nanoparticles (56 ± 6% of the loaded Doxo). An exception is represented by INU-PLA4-based nanoparticles, showing a lower repeatability and a big deviation on average results. This could be due to the higher heterogeneity of the starting copolymer in terms of molecular weight distribution (PDI = 1.7, cfr Table 2) leading to nanoparticles with a higher heterogeneity in the composition.
Drug release is apparently in contrast with DSC data. For a polymeric micellar type, a lower Tg should correspond to a lower CMC, a higher core dynamicity, and an easier exchange between the core and the external medium, with a consequent faster drug release. Therefore, it is evident that the new INU-based materials are able to self-assemble in a frozen state (not micellar) with a very slow rate of exchange (kinetically frozen aggregates) that can be slowed down by the length of the hydrophobic PLA core. Block INU-PLA copolymers with the same H/H ratios would all normally bring to a micellar system.