The hairless canary seed grains, corresponding to the CDC-Maria variety, were kindly supplied by Fitopal Company (Palencia, Spain). Samples were stored under refrigeration (4 ± 2 °C) until processing.

Isooctane, hexane and isopropanol (HPLC grade), NaOH, methanol, NaCl, BF₃, methyl tricosanoate, and tocopherol isomer standards (α, β, γ, and δ) were obtained from Sigma-Aldrich. Corn oil was supplied by Koipe Asua (Córdoba, Spain).

Lipid extraction using supercritical CO₂ was performed in a pilot-scale system, as illustrated in Figure 2a. The setup included a diaphragm pump (EH1, LEWA), a stainless-steel extractor (2 L, 70 MPa max pressure, 200 °C max temperature), and a separator (1.1 L, 30 MPa max. Pressure, 120 °C max. temperature). Throughout the extraction process, control parameters were monitored using a Measurement and Data Acquisition System (DAS-8000, Desin Instruments). Temperature and pressure profiles inside the extractor are shown in Figure 2b.

During the process, temperature and pressure in the extractor were maintained at 41 ± 1 °C and 40 ± 1 MPa, respectively. Once these conditions were reached, a 30 min equilibrium period was initiated. Afterwards, the SCCO2 extraction started maintaining a CO₂ flow rate of 12–15 kg/h. Throughout the extraction, the lipid fraction was collected in a separator maintained at 20 °C and 5 MPa. At the end of the extraction, the oil was stored at 4 °C until analysis protected from light to avoid oxidation. The extraction was considered complete when the weight of the lipid fraction stabilized. Due to the higher lipid content of WF compared to RF, the extraction durations were 540 and 360 min, respectively.

For comparison with SCCO2, RF and WF samples were defatted by Soxhlet extraction using n-hexane at 80 °C for 4 h, after which no additional oil was collected in the vessel. This temperature ensured continuous solvent recirculation. The lipid fractions obtained from this process were stored at 4 °C and protected from light until analysis.

Moisture content of the flours was measured following the Official AACC Method 44-19.01, lipids were quantified following the Official AACC Method 30-25.01, and ash content was determined with the Official AACC Method 08-01.01 (11). Proteins were determined by combustion following the UNE-EN ISO 16634-1 (12) applying a nitrogen-to-protein conversion factor of 5.7 (13). Total carbohydrates were determined by difference to 100%. Total fiber was measured using an enzymatic-gravimetric method in accordance with the Official Method AOAC 985.29 (14). All analyses were conducted in triplicate.

The PSD of the flours was assessed using a laser diffraction particle size analyzer (Mastersizer 2000, Malvern Instruments Ltd., Malvern, UK) paired with a Sirocco dry powder feeder. Results were reported as the diameters at which 10% (D₁₀), 50% (D₅₀ or median diameter), and 90% (D₉₀) of the particles have smaller size, and as the PSD dispersion ((D₉₀-D₁₀)/D₅₀). Experiments were performed in triplicate.

Water absorption capacity (WAC), oil absorption capacity (OAC), water absorption index (WAI), water solubility index (WSI), and swelling power (SP) were assessed on a dispersion of 5 g flour/100 mL, and foam capacity (FC) and foam stability (FS) were measured at a 2 g/100 mL, following the methods outlined by Abebe et al. (15). Emulsifying activity (EA) and emulsion stability (ES) were evaluated according to Vicente et al. (16). The results were expressed as follows: WAC and OAC in g_water/g_dried-flour and g_oil/g_dried-flou, respectively, WAI in g_sediment/g_dried-flour, WSI in g_{soluble-solids}/100 g_dried-flour, SP in g_sediment/g_{insoluble-solids}, FC as the foam volume in mL, FS as the percentage of foam after 60 min with respect to FC, EA as the percentage in volume of emulsion formed relative to the initial volume, and ES as the percentage in volume of emulsion after heating relative to the initial volume. All measurements were performed in triplicate.

The pasting properties of both non-defatted and defatted flours were assessed using a Rapid Visco Analyser (RVA) model 4500 (Perten Instruments, Australia). The Standard 1 temperature profile from the Official Method AOAC 76-21.01 (17) was applied. From the pasting curves, generated in triplicate, the pasting temperature (PT), peak viscosity (PV), peak time (Pt), trough viscosity (TV), breakdown viscosity (BV), final viscosity (FV), and setback viscosity (SV) were obtained.

The rheological properties of gels prepared at a concentration of 3.5 g (based on 14% moisture content) in 25 g of distilled water following the procedure described in the Section 2.5.4 were determined using dynamic oscillatory tests. Tests were performed using a Kinexus Pro + rheometer (Malvern Instruments Ltd., Malvern, UK), equipped with parallel plate geometry (40 mm diameter) and a 1 mm working gap. Gels were placed between the plates, allowed to relax for 5 min at 25 °C. Strain sweeps were carried out from 0.1 to 1,000% strain (1 Hz) to determine the linear viscoelastic region (LVR), while frequency sweeps were performed from 10 to 1 Hz (1% strain). The frequency sweep data were fitted to the power-law model (18). All samples were measured by triplicate.

Flour thermal properties, including gelatinization and retrogradation transitions, were analyzed using a differential scanning calorimeter (DSC3, STARe-System, Mettler-Toledo, Switzerland) following the method described by Vicente et al. (16). A ratio of 30 g solid to 70 g water was used for the analysis. The first scan, on fresh samples, assessed the gelatinization and dissociation of the amylose-lipid complex. A second scan, conducted after samples being stored for 7 days at 4 ± 2 °C, evaluated the melting of recrystallized amylopectin and the dissociation of the reversible peak of the amylose-lipid complex. Both scans were performed from 0 to 120 °C, at a heating rate of 5 °C/min. For each scan, the enthalpy (∆H) was measured in J/g of dry matter, along with the peak temperature (T_p) in °C, and the gelatinization temperature range (∆T_gel), calculated as the difference between T_end-set and T_on-set. The degree of retrogradation was calculated as the area of retrogradation related to that of gelatinization and expressed as percentage. Experiments were conducted in duplicate.

The chemical composition of canary seed flours is presented in Table 2. Canary seed flours were primarily composed of carbohydrates, with values of 74.4 ± 1.7 g/100 g dry basis (db) for RF and 70.3 ± 2.4 g/100 g db for WF, being comparable to other cereals such as wheat and oats (29). Lower values were previously reported by other authors for different canary seed varieties (13, 29). RF contained 5.8% more carbohydrates than WF (p < 0.05), since RF consists mainly of starch-protein matrix from the endosperm, while WF includes also germ and bran, which contain other components (30). Canary seeds have been recognized as one of the most protein-rich cereals, comparable to legumes (31), and containing significant essential amino acids (13). In this study, the protein content was 17.2 ± 1.5 g/100 g db for RF and 19.0 ± 1.7 g/100 g db for WF. The higher protein level in WF could be due to the presence of the germ and bran, richer in protein, compared with RF (30). These results are consistent with literature indicating higher protein content in whole grains compared to refined flours (13, 30). Oil in canary seeds is primarily found in the germ, endosperm, and aleurone layer (30). Therefore, the lipid content in RF (6.1 ± 0.6 g/100 g db) was significantly lower (p < 0.05) than in WF (8.1 ± 1.6 g/100 g db). The oil content observed is consistent with previous reports (4, 6) and is higher than in pulses like black gram and chickpea, and cereals like rice, wheat, and millet (32). Regarding canary seed flours derived from defatting, no significant differences were observed between the two defatting methods. Both HX and SCCO2 reduced the lipid content to less than 1.7 g/100 g db in both refined and whole flours, with a mean oil extraction yield of 81 ± 2% relative to the initial oil content in the flours. Moisture content was affected by the defatting process, with non-defatted flours having a moisture content of 14.6 ± 0.9 g/100 g for RF and 13.6 ± 0.8 g/100 for WF. Defatting with hexane resulted in a significant reduction in moisture content, of −58% in RF and −73% in WF. In contrast, SCCO₂ caused only a small decrease of about 6–7% in both samples. Urbizo-Reyes et al. (2) reported a 27% reduction in moisture content after defatting by mechanical pressing. Other studies on soy (33) and quinoa (34) flours found moisture reductions of 16–18% with hexane, attributed to the solvent’s high evaporation temperature (69 °C) which contribute to drying. Fiber was the least abundant component in both RF and WF, which aligns with the results of Abdel et al. (3, 35), although, as could be expected, WF contained a significantly (p < 0.05) higher content (2.4 ± 0.5 g/100 g db) than RF (< 1.2 ± 0.2 g/100 g db).

The particle size of flours directly affects the techno-functional properties of food products. Particle size parameters and distribution curves for both non-defatted and defatted flours are shown in Table 3 and Figure 3, respectively. Non-defatted flours exhibited a bimodal PSD, with RF having a greater volume percentage of finer particles than WF. While both flours showed similar distributions in the coarse fraction, WF had significantly larger D₅₀ and D₉₀ values, due to the presence of small amounts of germ and bran. Bran in canary seeds appears as thin flakes with diameters ranging from 0.3 to 1.0 mm (3), which explains the larger particle sizes observed in WF compared to the bran-free RF sample. Between the defatted samples, the PSDs of RF-HX and RF-CO2 were similar (Figure 3), retaining the two peaks observed in RF but with reduced volume contributions. In RF defatted samples, a new peak appeared at smaller particle sizes (0.80 to 6.88 μm), also reflected in the particle size parameters (Table 3). The D₁₀, D₅₀, and D₉₀ values significantly decreased in defatted RF flours, in greater extent in RF-HX than RF-CO2 samples, meanwhile the size dispersion increased by 37 and 42%, respectively. For WF defatted samples, peak differentiation was greater (Figure 3). Unlike the RF defatted samples, only two peaks were identified in the WF-HX sample, with the peak corresponding to the largest particle sizes missing (−38% in D₉₀). In contrast, the WF-CO2 sample exhibited a three-peak distribution, although the peaks were softer and appeared as slight shoulders. These results are consistent with previous studies, such as Nahimana et al. (36), who found smaller particle sizes in defatted sweet yellow lupin proteins due to hot defatting processes disrupting protein aggregates. This could be an advantage when using defatted flours, as finer particle sizes improve dough rheological properties, while coarser flours result in higher baking losses, tougher crumbs and higher chewiness (3).

The hydration, foaming and emulsifying properties of canary seed flours are shown in Table 3. In general, the techno-functional properties exhibited higher values in the WF samples than in the RF samples. Regarding hydration properties, WF showed increases of 7–11% in WAC, WAI, and SP and of 64% in WSI compared to RF, suggesting a greater capacity for water interaction probably due to the higher levels of dietary fiber and other hydrophilic constituents. The ability of the flour to form emulsions and foams was strongly influenced by the type of flour and its composition. The EA and FC of WF were approximately 1 and two times higher than those of RF, respectively. However, none of the non-defatted flours were able to maintain emulsion or foam stability. Defatting process had a notably impact on the functional properties of the flours. The WAC increased - after defatting, with higher increases in HX samples (21–22%) than in those obtained by supercritical CO₂ extraction (10–16%). This may be attributed to the smaller PSD in hexane-defatted flours, as smaller particles tend to absorb more water due to their increased surface area, which enhances water interaction (37). An increase in WAC is related to better food processing by avoiding liquid loss and favoring the preservation of texture, nutrients, and bioactive compounds (8). Consistent with these findings, numerous authors have documented enhanced absorption capacities of flours following SCCO2 extraction of soy (33) and wheat (38). The WAI and SP values did not change significantly because of defatting, regardless of the method used. In contrast, the WSI decreased by 8–24%, with a greater reduction observed in HX flours. These defatted samples also exhibited the greatest variation in OAC values with reductions of 15 and 13% for RF-HX and WF-HX respectively, meanwhile RF-CO2 and WF-CO2 showed a slight/not significant increase with respect to the initial values of RF and WF samples. For all flours, regardless of flour type or defatting process, the OAC value was higher than the WAC. This indicates the lipophilic behavior of canary seed flour, as previously reported by Rikal et al. (31). The most significant change after defatting was the marked improvement in EA, FC and FS of defatted flours. The EA significantly increased (p < 0.05) although the effect was dependent on the type of flour, the defatting process and the double interaction flour x defatting process (p < 0.001). Hexane extraction led to the highest increase in RF samples (10-times higher EA values), while for WF samples, SCCO2 showed the highest effect (+72%). None of these samples led to stable emulsions. Foaming capacity significantly increased (p < 0.05) by 2–4 times in all defatted samples, with HX flours showing the highest values regardless of flour type. Removing fat from canary seed flours increased the FS from zero to 30–41%, regardless of flour type and defatting process, related to the higher protein and lower lipid content.

The pasting parameters are presented in Table 4, and the pasting curves in Figure 4. The PT, or temperature at which the viscosity of starch slurries begins to increase during heating, and the Pt, representing the time to reach maximum viscosity, were significantly higher in WF than RF. For the PT, significant differences (p < 0.05) were observed depending on the flour type (F1), but not on the defatting process (F2). The PT and Pt values were higher than those reported for other cereals, such as waxy maize, wheat, tapioca, corn starch, and pea starch (3, 21, 39, 40). This may be explained by the high protein content of the canary seed flours examined in this study, which limits the swelling of starch granules and increases the temperature at which viscosity begins to rise.

PV, determined by the maximum swelling of starch granules, was always reached well after the start of the 95 °C hold period. The RF sample exhibited a PV value that was 21% above that of WF. This could be explained by the lower lipid content in RF (Table 2), as the lipid fraction limits water absorption and granule swelling (41). The PV values obtained for RF and WF were comparable to those previously reported for CDC Maria, C05041 (39), CDC Calvi, and CDC Cibo (21) canary seed varieties. After defatting, PV significantly increased (p < 0.05) for both RF and WF flours, demonstrating the role of lipids in hindering the swelling process. Nevertheless, when compared to other cereals, defatted flours from RF displayed similar PV values to corn starch and higher than wheat (21, 39, 40).

Both RF and WF, showed similar BV values to other reported canary seed varieties, but higher than wheat starch (39), and notably lower than those of waxy maize, tapioca, and pea starches (40). BV values, in general low, significantly (p < 0.05) increased after defatting, with the highest values observed in the HX samples, accounting for 45 and 42% of increase in RF-HX and WF-HX compared to RF and WF, respectively. These findings indicate that the great capacity of canary seed flours to withstand stress and heat is partially decreased after the defatting process.

SV indicates the starch gelling ability or retrogradation tendency (39). WF samples exhibited significantly lower values than RF (p < 0.05) due to its higher fat and fiber content. The highest values were found after SCCO2 defatting, surpassing those of the corresponding original flours, being especially notable the increase in WF-CO2 sample (26%) with respect to WF. On the contrary, HX reduced significantly SV values in both types of flours. These findings are consistent with previous studies, where the SV of wheat flour was higher in the SCCO2-defatted sample compared to hexane-defatted and non-defatted flour (38). The differences between the two defatting processes could be due to the higher temperatures used during hexane defatting process compared to SCCO2 (80 °C vs. 41 °C), which may promote a physical modification involving changes in starch structure (34). Similar results were reported by Irani et al. (39), who found comparable SV values for two canary seed varieties (CDC Maria and C05041) to those obtained in this study. Compared to other cereals, only pea starch showed similar SV values, while waxy maize, tapioca, corn, and wheat starch exhibited lower values, indicating lower gelling and retrogradation tendencies (21, 40). This rise in viscosity and gelling capacity in defatted flours may make them useful as thickening agents in food dispersions.

The parameters obtained from the rheological tests performed on gels with both non-defatted and defatted flours are presented in Table 4. This table presents the maximum stress value that the gel could withstand before rupture within the LVR, τ_max, and the crosspoint (G′ = G′′) obtained from strain sweep tests, as well as the parameters G₁’, G₁,” (tan δ)₁, and the exponents a, b, and c derived from fitting the power law model to the frequency sweep data within the LVR (0.898 < R² < 0.997, data not shown). The corresponding frequency sweep and the strain sweep plots are shown in Figure 5.

To date, there is little evidence concerning the rheological properties of canary seed flour. In general, both the type of flour (F1) and the defatted process (F2), significantly influenced the gel’s viscoelasticity (p < 0.05), apart from F2 for the (tan δ)₁ parameter, which showed no significant effect (Table 4). Additionally, significant interaction effects (F1 × F2) were observed for all rheological parameters except G₁’, indicating that the impact of defatting process varied depending on the type of flour (Table 4). All gel samples exhibited predominantly elastic behavior within the LVR, with G₁′ > G₁′′ and (tan δ)₁ values well below 1, implying a well cross-linked network structure. They also showed high stability, with very small dependence of G’ on frequency as can be concluded from the very low values of the exponent “a,” well below 0.1 (Table 4), and the small slope of the mechanical spectra (Figure 5b). No significant differences were observed in the elastic modulus (G₁’) between RF and WF flours. However, G₁’ significantly increased (p < 0.05) in RF-CO2 sample (13%), while no significant differences were observed between WF-CO2 and WF. By contrast, defatting with hexane resulted in reduced gel elasticity, with G₁’ decreasing significantly by 6% for RF-HX and 18% for WF-HX compared to RF and WF, respectively. This reduction is associated with the decrease in FV and may be explained by the higher temperatures applied during the hexane defatting process compared with the SCCO2 process and the annealing process associated as described above. These results suggest that the rheological properties of the gels were influenced by both the presence and composition of the lipid fraction in flours, which could inhibit amylose leaching during gelatinization and its retrogradation by forming amylose-lipid complexes, thereby weakening the continuous phase and leading to a reduction in the viscoelastic moduli. The removal of most of the lipids can increase the friction among the starch granules, potentially enhancing G’ (42), as observed for RF-CO2 flour. The gel made with WF exhibited a significantly higher viscous modulus, G₁,” than those made with RF (44 Pa vs. 42 Pa, respectively); however, this difference was not able to change significantly the value of (tan δ)₁ that was mainly affected by the high values of G₁’ in both RF and WF samples. Lipid removal led to a significant reduction in the viscous modulus (G₁”) especially in the HX-defatted flours (−12% for RF-HX and −26% for WF-HX with respect to the original samples). In general, the gels made with defatted flours showed a significantly lower (tan δ)₁ which indicates a better cross-linked network structure and a stronger solid-like character.

The low values of a exponent for flours, 0.015 for RF and 0.018 for WF, indicated minimal frequency dependence of the elastic modulus, G₁’, suggesting the formation of stable gel structures. The b exponent, associated with the frequency dependence of viscous modulus (G₁”), was significantly lower for WF (0.259) than for RF (0.278).

Regarding the maximum stress tolerated by samples within the LVR (τ_max), the value was higher for RF- than WF-samples, increasing significantly after SCCO2-defatting, with increases of 60 and 19% in RF-SCCO2 and WF-SCCO2 samples with respect to HX-defatted counterparts. The higher τ_max values denote greater gel stability against shear, as higher stress levels are required to disrupt their structure and transition to a viscous state (8). The crossover point, which indicates the stress for the transition from predominantly elastic-like to viscous-like behavior, followed a parallel evolution to τ_max and was also significantly higher for RF than WF, probably due to the disruption and weakening effect that fibre can cause to the gel structure. The most stable gels were also obtained from samples defatted by SCCO2, which stresses at the crossing point increased by 15%, in RF-SCCO2, and 52%, in WF-SCCO2, with respect to their original RF and WF corresponding samples.

The thermal properties of both non-defatted and defatted flours, as determined from the first gelatinization and the second retrogradation scans, are presented in Table 5. Thermograms are shown in Figure 6. During the gelatinization scan, two endothermic peaks were observed, as previously reported for canary seed flours (35, 39). The first peak, associated with starch gelatinization, appeared at a T_p of 66.0 ± 0.2 °C, with no significant effect of the type of flour or defatting process. However, the gelatinization temperature range (∆T_gel) differed significantly between RF and WF. In WF, ∆T_gel was 1.0 °C wider because T_on-set decreased (−1%) while T_end-set increased (+1%), expanding the interval at both ends. This behavior can be attributed to the higher fiber content of WF, which absorbs and restricts the amount of water available for gelatinization and delays the melting of amylopectin crystallites in starch granules (43). The gelatinization temperatures found in the present study are higher than those of wheat flour, and are in line with previous reports on canary seed flours (35, 39). However, our gelatinization temperature range was half, indicating more uniform crystals formation in our canary seed samples. This could be attributed to differences in the canary seed cultivar and chemical composition, such as lipid and fiber content. The gelatinization enthalpy (∆H_gel) significantly increased (p < 0.05) after defatting. In RF-CO2 and WF-CO2 samples ∆H_gel increased by 18 and 13% with respect to RF and WF, while for RF-HX and WF-HX this increase was smaller, 12 and 9%, respectively. This denoted a more stable starch structure in defatted samples which require more energy to destroy the crystalline area (44), highlighting the crucial role of lipid, amylose, and amylopectin interactions within the starch structure in determining the gelatinization properties of the flours. The differences between SCCO2- and HX-defatted flours could be attributed to the higher temperatures applied during conventional (hexane) defatting compared with the SCCO2 process. These higher temperatures may have promoted an “annealing” or slight “pre-gelatinization” effect, leading to a lower gelatinization peak area, as previously observed in defatted quinoa samples treated with hexane compared with those treated with SCCO2 (34). The small second peak, that appeared at 95–97 °C with a small ∆H, of ~ 3 J/g dry matter regardless of the type of flour and the presence/absence of fat, was probably due to the disruption of the amylose-lipid complex (43). It showed for all samples a small ∆H, around ~3 J/g dry matter, regardless of the type of flour and the presence/absence of fat. This suggests that the lipids involved in the amylose-lipid complex could not be removed during defatting processes, indicating that they were already bound to the starch within the granule.

In the second scan, two peaks were also observed in all flours. The first peak, linked to the melting of amylopectin retrograded during storage, occurred at 46 ± 5 °C, while the second peak, associated with the reversible amylose-lipid dissociation, appeared at 99 ± 1 °C. The higher temperature of the amylose-lipid dissociation peak observed in the second scan compared to the first indicates that, following gelatinization, this complex formed a more stable structure (45). The lower melting temperature of the retrograded amylopectin compared to that observed during gelatinization is due to the amylopectin crystals formed during storage are smaller and less perfect than those in present in native granules (46). The degree of retrogradation, that reflects starch re-association during storage, varied between 3% (for WF) and 12% (for RF). These low values indicate slow staling kinetics and are like those recorded for other cereals, such as wheat (10%) (47), and some pseudocereals, such as quinoa (9–11%) and amaranth (5%) (48). However, they were significantly smaller than the values reported for other gluten-free cereal flours such as rice (60%) (49), sorghum (46%) (50) or tef (25%-34) (47). Among the samples studied, WF showed the lowest retrogradation enthalpy, four times smaller than RF. However, no significant differences were observed because of the defatting process. These findings indicate that the fiber in whole flour may hinder the reorganization of amylopectin molecules during storage, making it more difficult to form an ordered structure. The ∆H_am-lip increased after the second scan compared to the first by 43, 23 and 20% for the RF, RF-HX and RF-SCO2 samples, respectively. However, WF and the two defatted flours derived from it hardly varied in the second scan with respect to the first. The increase in ∆H_am-lip values after the second scan has been previously reported in other flours, attributed to the leakage of amylose from starch granules when temperatures exceed the gelatinization range during the first scan (44, 45). The presence of fiber in WF could hinder the formation of new or more ordered amylose-lipid complex structures after gelatinization.