rECOBIOpack project: optimization of lipid extraction for coffee silverskin valorization

In the coffee year 2022/23, 10.092 million tons of coffee were produced generating waste at every step of the value chain. During coffee roasting process, the only generated by-product is silverskin (CS), the thin tegument covering green coffee beans. CS represents the 4.2% (w/w) of the beans and, from a circular economy perspective, it can be reused in other productive processes. CS has been mainly used for fuel or for composting so far. In this context, the rECOBIOpack project (funded by PRIN 2022) aims to develop a functional food packaging material from the main components of CS: cellulose as polymeric backbone and lipids, proteins and polyphenols as additives. In this work, the extraction of lipids, carried out using CO2 supercritical fluid extraction, was optimized using Design of Experiment (DoE). Moreover, the laboratory extraction was scaled up at ILSA S.p.A. (Arzignano, Italy) using an industrial supercritical CO2 apparatus. CS deriving from a mixture of Arabica coffee beans has been kindly supplied by the Italian coffee company Illycaffè S.p.A (Trieste, Italy). The lab-scale supercritical fluid extraction of lipids from CS was performed using the biomass as flakes, considering the following industrial application. A response surface methodology, based on DoE, was employed for the optimization of the extraction. The selected design was the Central Composite Face-centered (CCF), a design that supports a quadratic model, exploring the factors (variables having an influence on the system) at three levels (+1, 0, -1). Based on previous results, the selected factors were pressure (100, 300, 500 bar) and temperature (40, 60, 80°C). The extraction was performed in alternate static and dynamic cycles of 30 min and 10 min, respectively. To explore the experimental domain and to evaluate the experimental variability, a total of 11 experiments was performed. The optimized responses were extraction yield (Y%), free fatty acid content (FFA%), and percentage of unsaturated fatty acid (UNSAT%). The fatty acid composition of the samples was evaluated using GC-MS (Bruker Scion SQ Instrument, Milan, Italy), after derivatization (both transesterification and sylilation were employed). The significance of the resulting models was evaluated, afterwards the models were validated and the best extraction condition were selected. The fatty acid composition of the extracts obtained at the optimal conditions, were compared with the results of a conventional extraction using n-hexane. Finally, the industrial extraction was performed for 3 h at 400 bar and 60°C, in dynamic conditions. The biomass was extracted using two vessels (10 L) in series each one containing 0.7-0.9 kg of CS. Three DoE models were computed by multiple linear regression, however only the models obtained for Y% and FFA% resulted statistically significant. It is worth noting that the model coefficients related to pressure and to the interaction between pressure and temperature were significant for both the models but with opposite effect on each response. The extraction yields of the different DoE experiments (0.19-2.92%) permitted to highlight the influence of pressure on the CO2 solvent power. In particular, at lower value of pressure, the low density of CO2 resulted in lower yields due to its poor solvating power. On the other hand, higher FFA% were obtained for the experiments in which CO2 presents the lower value of density (100 bar and 60°C, and 100 bar and 80°C). This behaviour can be explained considering the lower solubility of lipids in these conditions, and the higher volatility of free fatty acids compared to glycerides. Due to the opposite effect of the factors on the investigated responses, the optimal extraction conditions were selected taking into account the Y% and were set at 400 bar and 50°C. Additional experiments were performed at the optimum and the model was successfully validated. Regarding the fatty acid composition, the GC-MS results showed a clear difference between sylilated and transesterified samples. The first present only the composition of free fatty acids, while the latter the total fatty acid composition. In particular, the sylilated samples presented a higher amount of long chain fatty acids. Compared to the conventional n-hexane extraction, supercritical CO2 extraction using the best extraction conditions did not show significant differences in terms of Y%, FFA% and characteristic fatty acids. The industrial extraction has been successfully carried out providing higher yields compared to the lab-scale, likely due to the higher (and more efficient) CO2-biomass ratio. The specificity of supercritical CO2 relies on its physical properties, which can be modulated by changing pressure and/or temperature. The use of DoE highlighted the influence of these parameters on lipid extraction and allowed to find the optimal conditions in order to maximum the yields. The application at industrial scale of supercritical CO2 has provided the lipidic extract and delipidized CS, necessary to the subsequent steps of the rECOBIOpack project, under the perspective of a zero-waste valorisation approach.