Biomimetic cellulose-based superabsorbent hydrogels for treating obesity

Preparation of CB-SAHs with tunable elasticity from CMC backbones

Among cellulose derivatives, CMC is an optimal candidate for the synthesis of CB-SAHs that are expected to function in the wide range of pH and ionic strength conditions found along the GI tract. CMC is modified cellulose containing a given number of carboxymethyl groups per glucose unit, as defined by the degree of substitution (DS) (Fig. 2a). As well known, carboxymethyl groups dissociate in aqueous solution, thus providing the CMC chains with anchored negative electrostatic charges. These fixed ions make a CMC-based crosslinked material not only superabsorbent, due to the Donnan effect³⁰, but also responsive to pH and ionic strength variations, as these environmental variables control the degree of ionization of the CMC chains and the number of mobile counterions in the bathing solution³⁰. However, the chemical crosslinking or stabilization of single CMC in water solution is hindered by the electrostatic repulsion among the CMC chains, which leads to the formation of intramolecular rather than intermolecular crosslinks²⁸. Intramolecular crosslinking gives rise to weak polymer networks with poor structural integrity and low elasticity.

However, in case of DS ≤ 1, physical crosslinking of concentrated CMC solutions may occur under static conditions (i.e. at rest), due to the establishment of hydrophobic associations among the unsubstituted cellulose blocks (Fig. 2b)³². Such physical associations or entanglements, which are responsible for the observed thixotropy of CMC solutions (i.e. time-dependent flow behavior, characterized by shear thinning and structural recovery after given time at rest)^27,32, are favored when the CMC has a higher molecular weight and/or is less uniformly substituted (‘blocky’ CMCs exhibit a higher tendency to form hydrophobic aggregates than regularly substituted ones)^32,33.

In the search of a method that could promote the formation of stable CMC networks, without the use of any additional ‘reinforcing’ agents (e.g. non-polyelectrolytes^28,29 or cyclodextrins³⁴), we focused on the optimization of CA-based crosslinking³⁵ (Fig. 2c, d). Since the crosslinking reaction (i.e. an esterification between the carboxylic groups of CA and the hydroxyl groups of cellulose) occurs upon preliminary casting and drying of a CA/CMC solution³⁵, we hypothesized that a stable CMC network could form as a result of both chemical and physical crosslinks, with the latter adding more elastically effective junction points to the covalent network (Fig. 2e). In this approach, physical crosslinks are controlled by the use of given CMC types, e.g. having given molecular weight distribution and/or uniformity of substitution, while chemical crosslinks are modulated by the CA crosslinking.

Here, we used two commercially available CMC types, referred to as CMC-L and CMC-H, to produce different classes of CB-SAHs with variable elasticity, namely GelA and GelB (see Table 1 and Methods). Although both CMCs have a nominal DS of 0.7, we found that they greatly differ for the molecular weight distribution and the thixotropic behavior of their aqueous solutions (see Supplementary Methods and Supplementary Figs. 1, 2). In particular, CMC-H shows a higher molecular weight (M_w), a lower polydispersity index (PDI) and a larger thixotropy than CMC-L. This suggested the potential of CMC-H chains to form a more entangled structure compared to CMC-L ones (Fig. 2b).

GelA and GelB samples were prepared in granular form, with a particle size range of 100–1000 μm in order to rapidly hydrate and create gel pieces similar in size to ingested raw vegetables (see Methods and Fig. 3a, Supplementary Movie 1). Right after the synthesis, we evaluated the actual formation of CB-SAHs by assessing the dynamic-mechanical properties (Fig. 3b) and the absorption of the granular materials in two different media, i.e. a mixture of simulated gastric fluid (SGF) and water, with SGF/water 1/8 v/v (diluted SGF, pH = 2.1; Supplementary Table 1) and phosphate buffered saline (PBS) (pH = 7.4).

In agreement with the expected formation of crosslinked networks, all hydrated samples showed a prevalently elastic behavior, with the elastic modulus G′ (in the following also referred to as elasticity) being one order of magnitude higher than the viscous modulus G″ over the tested frequency range (Fig. 3c). For the sake of comparison, values of G′ and G″ (at 10 rad/s), together with the corresponding medium uptake ratio (MUR), are reported in Fig. 3d, e for hydration in diluted SGF and PBS, respectively.

Interestingly, GelB02 (based on CA-crosslinked CMC-H, Table 1) showed increased elasticity over GelA (based on CA-crosslinked CMC-L), regardless of the hydration medium. In detail, the G′ values of GelA and GelB02 were respectively 1066 ± 92 Pa and 1846 ± 143 Pa in diluted SGF (Fig. 3d; p < 0.01), and 479 ± 44 Pa versus 1292 ± 68 Pa in PBS (Fig. 3e; p < 0.01). The significant increase of elasticity for GelB02 is directly ascribable to the use of CMC-H instead of CMC-L, which favors the formation of a more entangled polymer structure. However, the MUR of GelB02, although high, was reduced compared to GelA (GelA vs. GelB02: 85.3 ± 1.1 g/g vs. 77.0 ± 1.0 g/g in diluted SGF, p < 0.0001; 59.7 ± 1.1 g/g vs. 55.0 ± 0.6 g/g in PBS, p < 0.0001). We also observed that GelA and GelB02 showed increased elasticity at higher MUR levels, i.e. in diluted SGF. The extension of the polymer chains resulting from the medium uptake promoted the elastic response of the crosslinked network. Moreover, the significantly higher MUR values achieved in diluted SGF compared to PBS (p = 0.02 for GelA; p = 0.03 for GelB02) highlighted the marked sensitivity of GelA and GelB02 to the ionic strength of the bathing solution (13 mM of diluted SGF vs. 162.7 mM of PBS). In this regard, the presence of additional ionizable groups (-COOH) in the CA-crosslinked networks (Fig. 2d) could further enhance the environmental responsiveness of the CMC backbone.

On the contrary, GelB01 (which was not stabilized by CA crosslinking, Table1) showed elasticity and absorption values that were not substantially affected by the hydration medium (Fig. 3d, e). As for GelB03 and GelB04 (obtained by progressively increasing the CA crosslinking time), they showed lower MUR and higher G′ values with respect to GelB02, in both hydration media, in accordance with the expected formation of increasingly crosslinked covalent networks (Fig. 3d, e). In particular, GelB03 and GelB04 showed higher elasticity (as well as lower absorption) in PBS (Fig. 3e), compared to SGF (Fig. 3d). In PBS, the G′ values of GelB03 and GelB04 were respectively 6550 ± 372 Pa and 13,482 ± 653 Pa, while being 5216 ± 147 Pa and 7318 ± 620 Pa in diluted SGF.

Overall, we found that the CB-SAHs elasticity was tunable, particularly by changing the CMC type and/or adjusting the CA crosslinking. Among the tested samples, GelA and GelB02 showed different elasticity levels, coupled with high hydration in diluted SGF (about 70–80 times their dry weight). These samples were thus selected for further characterization and development.

CB-SAHs absorption and elasticity in simulated GI conditions

An in vitro GI model was specifically designed to simulate the passage of the CB-SAHs, as well as several references (e.g. raw vegetables), through the different parts of the GI tract (i.e. stomach, small intestine and colon). The designed model allowed to assess, in a reproducible system, the expected variations of MUR (absorption) and G′ (elasticity) of the tested materials through the GI tract (see Methods and Supplementary Table 2 for details).

GelA reached a higher hydration in SGF/water 1/8 v/v (pH 2.1) than GelB02, with MUR values, at 60 min of incubation, of about 86 g/g and 74 g/g for GelA and GelB02, respectively (p < 0.0001, Fig. 4a). Following incubation in SGF/water 1/4 v/v (pH 1.8) up to 120 min and then in pure SGF (pH 1.1) for an additional hour, the MUR of both CB-SAHs gradually diminished (down to 20 g/g and 22 g/g for GelA and GelB02, respectively; p = 0.47), consistently with the lower hydration expected at decreasing pH values³⁰. Then, the further pH increase upon immersion in simulated intestinal fluid (SIF, pH 6.8) led to the re-hydration of both CB-SAHs: at 300 min, the MUR of GelA and GelB02 increased up to the 80–85% of the initial value in SGF/water 1/8 v/v (p < 0.0001 for GelA at 300 min vs. GelA at 60 min; p < 0.01 for GelB02 at 300 min vs. GelB02 at 60 min). The final incubation of the CB-SAHs in simulated colonic fluid (SCF) then induced a rapid decrease of the MUR, due to concurrent degradation of the CMC network.

In accordance with the hydration kinetics, the G′ modulus of both GelA and GelB02 (Fig. 4b) reached its maximum value upon incubation in SGF/water 1/8 v/v, with GelB02 showing a significantly higher modulus than GelA (980 Pa and 1990 Pa for GelA and GelB02 respectively, at 60 min of incubation; p < 0.0001). The pH decrease resulting from the subsequent incubation in SGF/water 1/4 v/v and then in pure SGF, which led to a lower hydration, also caused a gradual decrease of the elasticity for both CB-SAHs. In particular, at 180 min of incubation, a reduction of G′ of about 28% and 35% was detected for GelA and GelB02, respectively (p < 0.05 for GelA at 180 min vs. GelA at 60 min; p < 0.0001 for GelB02 at 180 min vs. GelB02 at 60 min). A gradual reduction of G′ (down to approximately 260 Pa at 300 min) was then found for GelA upon incubation in SIF. In particular, G′ diminished from 180 to 240 min (p < 0.01) and then remained quite stable up to 300 min (p = 0.72). On the contrary, GelB02 was quite stable in SIF, with its elastic modulus at 300 min (~ 1200 Pa) being almost the same at 180 min (p = 0.58). Finally, consistently with the more extensive backbone degradation expected in the colon, the G′ modulus of both CB-SAHs rapidly diminished in SCF.

Mechanical properties comparison with raw vegetables and functional fibers in simulated GI conditions

After assessing the performance of GelA and GelB02 in simulated GI conditions, we used the same model to compare the elasticity of the CB-SAHs to two reference raw vegetables, minced cucumber and mixed salad, and three functional fibers, glucomannan, guar gum and psyllium, through the GI tract simulation (Fig. 5a, b). We monitored the G′ of the two CB-SAHs, as well as the reference vegetables and fibers as a function of the changing environmental GI conditions. The measured values of G′ variation generated distinctive profiles of behavior of the tested materials through time in the simulated conditions.

As reported in Fig. 5a, the elasticity profiles of the CB-SAHs and the raw vegetables were very similar. In detail, throughout the incubation in SGF the G′ of GelB02 (about 1300–2000 Pa) was maintained in the range defined by the modulus of mixed salad (as lower limit) and that of cucumber (as upper limit), while the G′ of GelA (about 600–950 Pa) was kept below the modulus of mixed salad. However, in the time frame between 180 and 240 min (i.e. between the final stage of SGF digestion and the initial stage of SIF digestion), both GelA and GelB02 showed very close similarity with the tested vegetables. More precisely, GelA showed G′ values very similar to those of mixed salad (at 180 min, about 705 Pa vs. 698 Pa for GelA and salad respectively, p = 0.95; at 240 min, about 302 Pa vs. 544 Pa for GelA and salad, p = 0.05). As regards GelB02, although at 180 min it showed an elastic modulus still lower than that of cucumber (about 1300 Pa vs. 1687 Pa for GelB02 and cucumber, p = 0.002), at 240 min the G′ moduli of GelB02 and cucumber were comparable (about 994 Pa vs. 816 Pa for GelB02 and cucumber, respectively; p = 0.15). It is also worth noting that, upon incubation in SIF, the G′ of both vegetables tended to decrease quite rapidly; at 300 min, cucumber and salad showed similar G′ values (~ 815 Pa for cucumber and ~ 544 Pa for mixed salad, p = 0.78). On the contrary, the elasticity of GelA and GelB02 appeared quite stable in SIF. However, upon incubation in SCF, both vegetables and CB-SAHs lost their elasticity, due to degradation.

In contrast, glucomannan, guar gum and psyllium, showed a completely different elasticity profile (Fig. 5b). All the tested functional fibers demonstrated orders of magnitude lower G′ modulus (about two orders of magnitude) than the CB-SAHs, in both SGF and SIF. Glucomannan, which was the fiber with the highest G′ among the tested functional fibers, showed G′ values in the range 27–50 Pa through the different media. The G′ modulus of guar gum showed a similar trend to glucomannan, although its values were kept in the lower range 6–10 Pa. Interestingly, in SCF, both glucomannan and guar gum were quite stable, compared to the CB-SAHs. With regard to psyllium, it showed very low values of G′, lower than 5 Pa, and was found to undergo a rapid degradation in SIF.

Ex vivo organ culture

The potential impact of the CB-SAHs on the gut tissue was studied by utilizing an ex vivo organ culture (EVOC) system (as described in the Methods)³⁶. Since weight management interventions are typically long term, understanding the effects of the CB-SAHs on the gut health could further teach us about the potential safety and efficacy of these products. In particular, we wanted to investigate whether the gut tissue would interact differently with hydrogels having similar composition but different elasticity levels, using as a reference the same vegetables and functional fibers studied in the GI simulation. For this purpose, we tested 4 types of CB-SAHs with variable elasticity, namely GelB01, GelB02, GelB03 and GelB04 (Table 1). While these CB-SAHs are all based on CMC-H, their elasticity progressively increases from GelB01 to GelB04, as highlighted in Fig. 3d, e.

A schematic representation of the intestinal epithelium, the mucus layer and the proliferating cell niche is represented in Fig. 6a (yellow arrows), while Fig. 6b shows the EVOC system, with the compounds in direct contact with the explanted tissues. When testing the CB-SAHs (upon hydration in PBS), it emerged that the mucus layer as well as the tissue architecture were preserved, in a way comparable to the medium or even better (the blue staining represents the mucus layer, in Fig. 6c), when the tissue was in contact with GelB02 or GelB03, i.e. the CB-SAHs having an intermediate elasticity level, approximately between 1300 and 6500 Pa (as shown in Fig. 3e). Conversely, GelB01 (i.e. the CB-SAH with the lowest elasticity, about 800 Pa) and GelB04 (i.e. the CB-SAH with the highest elasticity, about 13,000 Pa) failed in maintaining the mucus integrity, similarly to PBS (Fig. 6c). In agreement with the effect on mucus layer, GelB02 and GelB03 performed better in preserving the tissue proliferative ability, similarly to the medium (brown nuclei correspond to proliferating cells, in Fig. 6c; total nuclei are counterstained with hematoxylin). On the contrary, proliferative cells became rare in tissues exposed to GelB01 or GelB04, similarly to PBS (Fig. 6c; the quantification of Ki67 staining is reported in Fig. 6d).

We then compared GelB02, the hydrogel behaving more similarly to the medium condition (positive control), to the functional fibers and vegetables in the same conditions. Only the tested vegetables (both cucumber and salad) and glucomannan, which is the soluble fiber having the highest elasticity in PBS among those tested (Supplementary Fig. 3), were able to preserve the mucus (Fig. 6e) and partly the tissue architecture, similarly to GelB02; conversely, both psyllium and guar gum failed in fully preserving tissue homeostasis ex vivo (Fig. 6e). Moreover, while cucumber and salad were able to maintain tissue proliferation as GelB02, none of the fibers were able to do so; indeed, when the tissue was exposed to soluble fibers, the tissue damage was such that the Ki67 staining was of bad quality and could not be used reliably.