A research team led by Shaoning Cui and Han Du set out to test whether cellulose nanocrystals, or CNCs, could slow the retrogradation of pea starch. CNCs are rigid, rod-like crystalline fragments of cellulose, typically 100 to 500 nanometers long and 10 to 50 nanometers wide, produced by dissolving away the amorphous portions of microcrystalline cellulose with sulfuric acid. Unlike flexible polysaccharides such as chitosan or alginate, which behave like soft polymer chains, CNCs behave more like microscopic reinforcing bars. Their surfaces are studded with hydroxyl groups that can form hydrogen bonds with neighboring molecules, and their stiff geometry prevents them from tangling with one another. The team hypothesized that these rods would fight retrogradation in two ways at once: by competitively binding water that starch chains need to mobilize, and by physically blocking the realignment of amylose and amylopectin chains as the gel ages.

To test the idea, the researchers hydrolyzed wheat-straw-derived microcrystalline cellulose with 65 percent sulfuric acid under sonication, then purified the resulting rods by repeated centrifugation and dialysis. Transmission electron microscopy confirmed the classic whisker morphology, while dynamic light scattering revealed a zeta potential of −36.9 millivolts, a negative surface charge imparted by sulfate ester groups that keeps the particles stably dispersed in water. X-ray diffraction showed that the acid treatment had stripped out the disordered regions of the parent cellulose, raising the crystallinity index from 62.26 percent to 88.71 percent. These intrinsic properties—a high aspect ratio, strong surface charge, and near-perfect crystallinity—turned out to be critical to everything that followed.

The team then blended CNCs into pea starch at concentrations ranging from 2 to 8 percent by weight and subjected the mixtures to a battery of analytical tests spanning seven orders of magnitude in scale, from molecular vibrations to whole-gel texture. Rapid viscosity analysis showed that even small doses of CNCs changed how the starch behaved during cooking. At 4 percent CNCs, the peak viscosity of the paste dropped from 505 to 487 pascal-seconds, a sign that the nanorods were crowding the swelling granules and competing for water. More tellingly, the setback viscosity—a measure of how much the paste thickens as it cools and the chains begin to reassociate—fell steadily from 187 pascal-seconds in the pure starch to 159 pascal-seconds at 8 percent CNCs, indicating that the nanorods were already interfering with the earliest stages of chain reordering.

The most visually striking evidence came from simple refrigeration experiments. When gels were stored at 4 degrees Celsius for five and seven days, the untreated pea starch gel expelled progressively more water, a phenomenon called syneresis that anyone who has opened a container of old pudding will recognize. By day seven, the control gel had lost 30.63 percent of its weight as expelled water. The gel fortified with 8 percent CNCs lost only 27.49 percent, a reduction of more than three percentage points, and the effect grew stronger with increasing CNC loading. Low-field nuclear magnetic resonance revealed what was happening at the molecular level: the fraction of free, mobile water in the gels fell dose-dependently from 33.09 percent in the control to 27.79 percent at 8 percent CNCs after a week of storage, while the relaxation time of weakly bound water shortened, showing that the nanorods were literally pinning water molecules in place within the gel network.

Scanning electron microscopy provided the structural explanation. In the freeze-dried retrograded gels, the CNCs appeared as rigid rods uniformly distributed throughout the starch matrix, bridging adjacent fragments and pore walls like tiny struts in a scaffold. Gels containing the nanorods showed thickened fragment walls, smoother surfaces, and more homogeneous pore distributions than the untreated control, forming what the authors describe as a concrete-like supportive framework. The hydroxyl-rich surfaces of the rods competitively bind water while simultaneously forming hydrogen bonds with amylose and amylopectin chains, reinforcing the three-dimensional network and denying the starch chains the mobility they need to crystallize.

Spectroscopic and thermal measurements confirmed that the interaction is non-covalent but powerful. Fourier transform infrared spectroscopy showed no new absorption peaks and no peak shifts, only a pronounced broadening of the hydroxyl stretching band near 3300 wavenumbers—a fingerprint of extensive hydrogen bonding between the nanorods and the starch chains. Differential scanning calorimetry quantified the energetic consequence: after seven days of refrigeration, the retrogradation enthalpy of pure pea starch was 1.46 joules per gram, but only 0.60 joules per gram at 8 percent CNCs, cutting the degree of retrogradation nearly in half, from 14.07 percent to 7.23 percent. X-ray diffraction delivered perhaps the most dramatic number of all: the relative crystallinity of the stored gels fell from 34 percent to 16 percent across the CNC range, a 52.94 percent reduction, while the characteristic B-type crystal polymorph of the starch remained unchanged, confirming that the nanorods suppressed crystal growth without altering its fundamental nature.

Texture analysis tied the physics back to the eating experience. After a week of cold storage, the untreated pea starch gel had hardened to 93.76 grams of resistance, while every CNC-fortified gel was measurably softer, with springiness preserved near 1.0 throughout. Interestingly, the relationship between CNC concentration and gel properties was not strictly linear. The inhibitory effect rose with loading and then plateaued between 6 and 8 percent, and above that threshold the nanorods began to agglomerate, losing their ability to disperse and undermining the very interfacial contact on which the mechanism depends. This non-linear, dispersion-governed behavior distinguishes rigid rod-like CNCs from both flexible hydrocolloids and spherical nanoparticles, and it establishes a practical formulation window for food developers.

The implications reach beyond the laboratory. Because CNCs are derived from abundant agricultural byproducts and work through physical rather than chemical modification, they sidestep the toxicity concerns of chemical starch additives and the limited controllability of enzymatic treatments. A dosing guideline of 6 to 8 percent CNCs could extend the refrigerated shelf life of pea-based noodles, jellies, and meat analogues without any synthetic preservatives. The authors caution that their study covered only one week of storage at 4 degrees Celsius, leaving room-temperature stability, freeze-thaw cycling, digestibility, and real-food matrices for future work. But the core message stands: the answer to keeping plant-based foods fresh may lie in scattering billions of impossibly small cellulose rods through the gel, where they lock up water, brace the network, and quietly stop starch from turning back against itself.

Subject of Research: Using cellulose nanocrystals to inhibit the retrogradation of high-amylose pea starch in food gels

Article Title: Effects of cellulose nanocrystals on the physicochemical and retrogradation properties of pea starch

Article References: Cui, S., Cui, S., Xie, W., Li, H., Li, F., Ge, S., Fu, Z., & Du, H. (2026). Effects of cellulose nanocrystals on the physicochemical and retrogradation properties of pea starch. Food Chemistry: X, Article 104569. https://doi.org/10.1016/j.fochx.2026.104569

Image Credits: AI Generated

DOI: 10.1016/j.fochx.2026.104569

Keywords: pea starch, cellulose nanocrystals, starch retrogradation, food chemistry, amylose, syneresis, nanomaterials, gelatinization, X-ray diffraction, LF-NMR, plant-based foods, shelf life

The team’s choice of almond protein isolate was deliberate. Almonds already have a reputation in the plant-milk world for mild flavor and smooth texture, avoiding the beany off-notes and allergenic baggage that complicate soy. The isolate is roughly 97 percent protein, most of it the water-soluble globulin amandin, which makes up 65 to 70 percent of the total. Crucially, almond protein shares an isoelectric point with casein, in the pH 4.5 to 5.5 range, and retains about 70 to 80 percent solubility even at pH 4. That means it can undergo heat- and acid-driven structural changes at almost exactly the point where yogurt fermentation ends, mimicking the acid coagulation that builds dairy gels. The researchers hypothesized that varying the almond protein concentration within a xanthan gum-stabilized millet matrix would tune protein–polysaccharide interactions and dictate whether the final gel held together or fell apart.

The experimental design was straightforward but rigorous. Millet milk was prepared by soaking, blending, and filtering proso millet grains, then supplemented with almond protein isolate at 0, 0.5, 1, 1.5, and 2 percent weight per volume, alongside 0.5 percent xanthan gum, 3 percent inulin as a prebiotic, and 3 percent sucrose. After hydration, pasteurization at 90 degrees Celsius for 30 minutes, and inoculation with a vegan starter culture containing Streptococcus thermophilus, Lactobacillus delbrueckii subsp. bulgaricus, L. acidophilus, and Bifidobacterium animalis subsp. lactis, the mixtures were fermented to pH 4.6 and chilled. The resulting yogurts were then put through an unusually complete battery of tests: proximate composition, fermentation kinetics, electron microscopy, infrared spectroscopy, zeta potential, free sulfhydryl measurement, selective-solvent interaction analysis, colorimetry, water-holding capacity, texture profiling, rheology, and a trained sensory panel.

The compositional results alone tell a striking story. Millet milk starts with a meager 1.23 percent crude protein, and the unsupplemented control yogurt contained just 0.38 percent. Adding almond protein raised protein content in a concentration-dependent manner, reaching 2.23 percent at the 2 percent dose, roughly a six-fold increase over the control. Fat stayed below 0.5 percent across all formulations, comfortably meeting the European Union criterion for a fat-free claim. But the more consequential effects appeared during fermentation. Yogurts with 1 percent almond protein or more acidified faster and reached the pH 4.6 endpoint earlier than the control, and viable lactic acid bacteria counts climbed from 5.29 log CFU per gram in the control to 7.00 log CFU per gram at the highest dose. The likely explanation is nitrogen: the o-phthaldialdehyde assay showed that trichloroacetic acid-soluble amino groups rose significantly during fermentation, from 2.39 to 3.42 millimolar glycine equivalents in the 2 percent formulation, providing the peptides and free amino acids that fuel bacterial metabolism and acid production.

Microscopy revealed exactly why the control yogurt failed. Under focused ion beam scanning electron microscopy, the unsupplemented sample showed a fragmented, poorly connected structure of irregular particles that remained fluid and flowed when the cup was inverted. At 0.5, 1, and 1.5 percent almond protein, however, the gels formed compact, homogeneous networks with honeycomb-like pores, smooth cohesive surfaces, and the ability to hold their shape upside down. At 2 percent, the picture reversed: pores collapsed, fractures appeared, the surface turned rough and aggregated, and visible syneresis, or water separation, set in. The lesson is that gel quality is not a simple dose–response curve. Moderate protein supplementation builds a continuous network, while too much pushes the system past the point where electrostatic repulsion can keep protein molecules from clumping into a coarse, grainy mass.

The molecular explanation lies in the balance of intermolecular forces. Selective buffer experiments showed that electrostatic interactions were the dominant force in most almond-supplemented yogurts, rising from 0.17 milligrams per milliliter at 0.5 percent protein to 0.49 at 1.5 percent before falling back to 0.27 at 2 percent, where hydrogen bonding took over. Zeta potential measurements reinforced this interpretation: all formulations were negatively charged, consistent with the anionic xanthan gum, but the potential became less negative as protein increased, dropping sharply to minus 26.0 millivolts at 2 percent. That collapse in electrostatic repulsion coincided with the aggregation and structural failure seen under the microscope. Free sulfhydryl content did decline during fermentation, indicating some disulfide bonding, but the values were far below those reported for soy yogurt, reflecting almond protein’s limited cysteine content. Disulfide bridges helped, but the electrostatic glue between charged protein regions and anionic polysaccharides was the real architect of the gel.

Physical performance tracked the molecular story closely. Water-holding capacity, the single best indicator of physical stability in set yogurt, jumped from a dismal 17.65 percent in the control to 99.00 percent at 0.5 percent almond protein and 97.92 percent at 1 percent, then slid to 89.47 percent at 1.5 percent and crashed to 59.77 percent at 2 percent. Firmness in the back-extrusion test rose from 0.19 newtons to 21.46 newtons across the same range, and the consistency index climbed from 4.56 to 437.18 newton-seconds. Notably, the 0.5 and 1 percent formulations landed nearest the firmness and consistency values typical of dairy set yogurt, around 2 to 3 newtons and 50 to 60 newton-seconds, while the two highest doses produced gels far harder than anything a consumer would recognize as spoonable. Rheology confirmed the pattern: storage modulus exceeded loss modulus in every sample, confirming solid-like behavior, but the 2 percent gel reached moduli of 10,000 pascals, orders of magnitude above the 1 to 100 pascal range of conventional yogurt.

Perhaps the most counterintuitive finding came from the three-interval thixotropy test, which measures how well a gel rebuilds itself after being sheared apart. The 1 percent formulation showed a significantly higher recovery rate than the 1.5 and 2 percent samples, and the 2 percent gel recovered worst of all. Pearson correlation analysis made the point explicit: firmness and storage modulus were negatively correlated with structural recovery, meaning that stiffer gels bounced back more poorly after mechanical deformation. Strength, in other words, is not the same as resilience. A dense, rigid network built on excessive protein aggregation and hydrogen bonding cannot absorb stress the way a balanced, electrostatically stabilized, porous network can. For product developers, this decoupling of hardness from quality is a critical design principle: the goal is not the strongest gel but the most organized one.

The sensory panel of twenty trained tasters agreed. The millet yogurts outscored a commercial soy set yogurt on appearance, and the 0.5 and 1 percent formulations earned the highest texture scores, consistent with their yogurt-like firmness and rheology. Almond protein improved taste and flavor across the board, but the 1 percent sample achieved the highest overall acceptability at 4.85 on the nine-point hedonic scale, a judgment that weighed color, water retention, microstructural uniformity, texture, and flavor together. The authors conclude that heat-induced protein unfolding and fermentation-driven acidification enabled almond protein to mediate protein–polysaccharide interactions that convert a weakly gelling cereal matrix into a self-supporting, low-fat, soy-free set yogurt whose gelation pattern resembles conventional dairy yogurt. They caution that the detailed molecular events of aggregation were not directly observed and will require further characterization. Even so, the study offers a practical recipe for the next generation of plant-based yogurts: a climate-hardy grain, a mild tree nut protein, and a dose window narrow enough that precision, not abundance, makes the gel.

Subject of Research: Using almond protein isolate to modulate protein–polysaccharide gel networks and improve the physicochemical stability of millet-based set yogurt

Article Title: Modulating protein–polysaccharide gel networks via almond protein to improve physicochemical stability of millet-based set yogurt

Article References: Modulating protein–polysaccharide gel networks via almond protein to improve physicochemical stability of millet-based set yogurt. (n.d.). https://doi.org/10.1016/j.fochx.2026.104535

Image Credits: AI Generated

DOI: 10.1016/j.fochx.2026.104535

Keywords: plant-based yogurt, almond protein isolate, proso millet, protein–polysaccharide interactions, xanthan gum, gel network, water-holding capacity, rheology, lactic acid fermentation, food texture, syneresis, dairy alternatives