Introduction
Tomato farming is a huge industry especially in southern European countries with Spain, Italy and Portugal producing more than half (Almeida et al., 2024). However, large quantities of waste by-products are generated such as rotten tomatoes, unsaleable (misshapen and unripe) tomatoes and tomato pomace which is comprised of peels and seeds when creating tomato paste. The current use of tomato waste by-products undergoes composting to form a fertilizer with a high protein content of up to 20% (Fritsch et al., 2017). However, there is potential to upscale this biomass to focus on protein from tomato seeds. In a previous study, an examination of two different tomato seed cultivars at different stages in the growth cycle revealed a protein content associated with non-defatted seeds to be 23% (g per 100 g dry weight) and 27% with defatted seeds (Brodowski & Geisman, 1980). The effect of pH on protein recovery from tomato seeds has been found to be important factor when a pH range from 2–12 was investigated, revealing a significant increase in yield from 5% to 14% between pH 7 and pH 9 followed by a gradual increase from pH 9 to 12 (Piyakina et al., 1995). In addition, protein recovery yields from defatted seeds were higher at about 20% than from non-defatted seeds at about 14%. Consequently, most studies have focused on using an alkaline extraction process to recover the protein from defatted tomato seeds at pH 8 in 0.1 M NaCl at 50 °C (Sarkar and Kaul, 2014), in 1% NaOH (Sogi et al., 2005) and 1 N NaOH at pH 9 (Latlief and Knorr, 1983). While protein recovery might be higher at pH 12, this will result in loss of the three-dimensional structure of the protein which in turn could result in changes of the functional properties such as colour (Rana et al. 2025).
Sedimentation has often been used to separate the seeds from tomato pomace (Kaur et al., 2005; Meshkani et a., 2016; Miklavčič Višnjevec et al., 2021) and from whole tomatoes (Sarkar and Kaul, 2014). However, it is possible to recover the seeds from whole tomatoes by filtration through sieves with holes sufficiently large enough to allow the passage of the tomato seeds. The presence of a gelatinous layer surrounding the tomato seeds allows the seeds to readily pass through these holes whereas the tomato peel remains trapped on the surface of the sieve. This was demonstrated with 20 kg whole cherry tomatoes to recover 53 g of seeds (Baker et al., 2022) and from 100 kg salad-type whole tomatoes to recover 110 g of seeds (Baker, unpublished). Whether this process would be effective at a much larger scale remains to be determined as the larger quantity of seeds could prevent passage through the holes. However, the process to extract tomato juice from tomatoes has resulted in tomato pomace containing seeds where the gelatinous layer has been removed, and these seeds could only be separated from the remaining tomato pomace by sedimentation (Miklavčič Višnjevec et al., 2021). The seed content associated with the tomato pomace (at 53%) based on of dry weight (Liadakis et al., 1995) is high in contrast to regular tomatoes at 2% (Baker, unpublished). Furthermore, compositional analysis of tomato pomace compared with whole tomatoes has revealed differences in the fibre content resulting in a lower hemicellulose content and higher lignin and lipid contents were higher, often by two-fold or more (Almeida et al., 2024).
The aim of this study was to scale-up from previous smaller scale laboratory experiments to pilot scale using the fixed equipment available at the pilot plant. Previous studies have made comparisons in protein recovery at laboratory scale to pilot-scale from pre-treated pea revealing a slight improvement in protein yield from 60% to 94% (Schmidt et al., 2022). More recently, pilot scale protein extractions have also been performed with rapeseed (Ahlström et al., 2022), lentil (Joehnke et al., 2021) and hempseed (Helstad et al., 2022). In our study, modifications were necessary during upscaling to account for the increased volume during the process. The pilot facilities offered the opportunity to concentrate the protein using ultrafiltration and to recover the protein using spray drying instead of acid precipitation. The logistics of transporting tonnes of waste tomatoes from Spain to the UK was considered and found to be unviable. Consequently, a supplier of tomato pomace was found in Ukraine requiring the transport of only 50 kg which was logistically possible possessing a higher seed density.
Materials
Tomato pomace (50 kg) formed by pressing Roma tomatoes grown on farms by the conglomerate, Agrofusion in Kherson, Ukraine and sedimentation and was shipped frozen from Agrofusion, Kherson, Ukraine once separated from the tomato juice. Deseeding was achieved by immersion in hot water at 50 °C and skimming off floating pomace and skins from the surface and the resuspended seeds were collected on a sieve (Miklavčič Višnjevec et al., 2021).
Pilot scale extraction from tomato seeds
The seeds (2.815 kg wet weight equivalent to 809 g in dry weight) were resuspended in 10 L water, were added and stirred at 160 rpm. An outline of the procedure is shown in Figure 1. The suspension was at pH 5.5 and at about 20 °C at the start of the experiment. The seeds were transported to Bangor University laboratory and high-shear mixed for five minutes at 8000 rpm each using each of the different workheads, (General purpose disintegrating head, axial flow head and square hole high shear screen, Silverson, UK), beginning with the workhead with the largest holes (5 mm) and completing with the workhead with the smallest holes (1 mm). High shear mixing was used initially instead of colloidal wet milling because the large quantity of seeds was too much for the wet mill to initially process. After high shear mixing, the suspensions were transported to the pilot scale facilities and another 10 L water was added to the suspension which passed from one 150 L stirred tank through a wet mill (model no. MZ50, FrymaKoruma) into another 150 L stirred tank over 30 min where the cone size of the wet mill was set at 0.5 mm at the beginning and decreased to 0.1 mm. The remaining suspension was adjusted to pH 8 with 2 M sodium hydroxide (500 g, Fisher Scientific, UK, product code 15433117) and the suspension was stirred at 160 rpm for 1 h at room temperature. The suspension was then filtered through 250 μm sieve and the filtrate was retained. The sieved material was resuspended in 20 L water and adjusted to pH 8 with 2 M sodium hydroxide. The process of resuspending the material collected on the sieve in 20 L water, adjusting to pH 8, stirring for 30 min and sieving again to collect another filtrate was repeated altogether three times. The filtrates were pooled together, stirred and adjusted to pH 5 with 3 M phosphoric acid (2.5 L, Fisher Scientific, UK, product code 10698133) and left overnight at room temperature. Acidification to pH 4 was avoided to prevent protein precipitation. The filtrate was separated by a stack disc centrifuge (model no. SB 7–06–576, GEA Westfalia, Germany) into a centrate and an enriched solids suspension that was stored at 4 °C. The centrate was ultrafiltered through a 10 kDa polyether sulphone, ultrafiltration membrane (Axium Process Ltd., UK) at 30 °C resulting in a permeate and a retentate containing the concentrated protein and a permeate containing soluble compounds without protein. This retentate was high shear mixed with the emulsor screen for 5 min at 7000 rpm in 5 L batches and then passed through the spray dryer (Spray dryer model no. S1, Anhydro, Denmark) at a temperature between 90–100 °C. A sample of this crude protein (1 g) was resuspended in 200 ml hexane for 1 h at 70 °C and then the 95% n-hexane (2.5 L, Fisher Scientific, product code 10783601) was decanted. The pellet was completely dried by rotary evaporation (Rotavapor R100, Buchi, Switzerland), resuspended in deionized water and freeze-dried.
Figure 1. Schematic diagram of pilot process to extract protein from tomato seeds and pomace.
The process was similar with both substrates except where shown in brackets.
Pilot scale extraction from tomato pomace
Frozen tomato pomace (2.806 kg equivalent to 0.703 kg in dry weight) was thawed in 20 L hot tap water and then high shear mixed in the Bangor University laboratory using each of the workheads for 5 min at 7000 rpm in batches of 5 L as described in the previous section. The resuspended pomace was transported to the pilot scale facilities and was colloidal wet milled using a cone setting of 0.1 mm for 30 minutes at room temperature and then passed through a 250 μm sieve. The pellet was resuspended in another 20 L tap water and colloidal wet milled for 30 min, then sieved and filtrate was retained. The pellet was resuspended in another 10 L tap water and colloidal wet milled for 30 min, sieved and filtered. This was repeated a further two times resulting in two filtrates. The filtrates were combined and stirred. The pooled filtrate was passed through the stack disc centrifuge to remove the solids which was stored at 4 °C. The centrate was concentrated by ultrafiltration using a 10 kDa membrane resulting in a large volume of permeate waste. This suspension was high shear mixed with the emulsor screen attached for 5 min at 7000 rpm in 5 L batches and then passed through the spray dryer at temperatures between 90–100 °C.
Protease treatment of tomato seed fractions
Crude protein was extracted from tomato seeds (10.5 g wet weight equivalent to 3 g in dry weight) as previously described (Miklavčič Višnjevec et al., 2021). Briefly, the seeds were immersed in 500 ml deionized water and high shear mixed (Silverson mixer, UK) for 15 min at 7000 rpm. The seed hull solid was collected by sedimentation, and the crude protein suspension was decanted into another vessel. The pellet was resuspended in deionized water (50 ml) and a sample (1 ml) was removed to determine protein and peptides content before protease treatment. The pH of the remaining suspension was checked to ensure that it was between pH 5.5 to 7.5, the optimal pH range of the enzyme, and Neutrase (100 μl) (100 ml, NCBE, UK) was added. A sample (1 ml) was removed from the supernatant (100 ml) was checked to determine protein and peptide content and then Neutrase (20 μl) was added. Both the resuspended pellet and supernatants were incubated at 50 °C on the bottle roller (Fisher Scientific, UK, product code 15258854) at 80 rpm for 1 h and then incubated at 85 °C for 15 min with occasional shaking to deactivate the enzyme. Finally, the suspensions were cooled to room temperature in order to determine the protein and peptide concentrations using the Bradford reagent (500 ml, Fisher Scientific, UK, product code 15424109) and modified Kaiser test kit (50 ml of each of three components, Merck, product code 12161503). The resuspended seed hulls were centrifuged (Avanti JXN-30 centrifuge, Beckman Coulter, Germany) at 4320 xg for 5 min. The protein concentrations were determined using the Bradford assay while peptide and amino acid concentrations using the modified Kaiser assay. Controls were performed using Neutrase (100 μl) immersed in 100 ml in water only. The protein, peptide and amino acid concentrations were calculated by subtracting the appropriate controls depending on whether protease was used.
Collection and analysis of samples
The volume of effluent was measured at each stage and was used in determining the protein contents. Samples (10 ml) were periodically collected during the process steps (
Table 1) for soluble protein analysis and determination of bacterial populations. Protein contents were determined using the Bradford assay and measured in a plate reader (BioTek Epoch microplate spectrophotometer, Agilent Technologies, USA) at 595 nm. Bacterial populations were determined by serial dilution in 0.1% (w/v) sodium chloride (500 g, Fisher Scientific, UK, product code 10616082) solution and plating onto nutrient agar (500 g, Fisher Scientific, UK, product code 10219252). The plates were incubated at 25 °C for one week and colonies were counted. The modified Kaiser assay was used to determine the presence of amino acids and peptides in the resuspended fraction and in the filtrate revealing that considerably more peptides were found in the filtrate. The latter assay was performed using 910 μl of sample, neat or diluted, with the addition of components of the Kaiser test {30 μl 0.2 mg potassium cyanide and 66% (v/v) pyridine, 30 μl 5% ninhydrin in n-butanol, and 30 μl 200% phenol in n-butanol]. The suspension was then incubated at 105 °C for 5 min and then cooled to room temperature. Samples of 100 μl were measured in the plate reader at 570 nm. A standard curve was created using proteose peptone which had been enzymatically treated was prepared at 0, 0.2, 0.4, 0.6, 0.8 and 1.0 mg/ml.
Table 1. Protein concentrations and bacterial populations during each of the processing steps from tomato seeds and protein concentrations from tomato pomace comprising both seeds and skins.
| Step in procedure | Concentration | Bacterial population associated with seeds |
|---|
| Seeds (mg/g DM seeds) | Pomace (mg/g DM pomace) | Pomace* (mg/g DM seeds) | (cell counts per g DM) |
|---|
| At start | 0.0 | 0.5 | 0.7 | 6.48 |
| After high shear mixing and wet milling | 23.8 | 0.6 | 0.9 | 6.87 |
| Adjustment to pH 8 | 23.8 | ND | ND | ND |
| First filtrate | 22.4 | 0.8 | 1.1 | 6.82 |
| Second filtrate | 1.9 | 0.3 | 0.5 | 6.15 |
| Third filtrate | 0.9 | 0.1 | 0.2 | 6.90 |
| Forth filtrate | 0.5 | ND | ND | 6.59 |
| Pooled supernatants at pH 5 | 52.4 | 1.7 | 2.5 | 6.18 |
| Effluent from stack disc centrifuge | 45.0 | 2.0 | 2.9 | 4.66 |
| Solids from stack disc centrifuge | 8.0 | 0.1 | 0.2 | 8.54 |
| Permeate after ultrafiltration | 0.2 | ND | ND | 8.32 |
| Retentate after ultrafiltration | 4.0 | ND | ND | 7.55 |
Results and discussion
The suspension of tomato seeds was at pH 5.5 at the start and almost no protein was detected (
Table 1). Protein concentrations in the filtrate had increased that corresponded with the disintegration of the seeds. High shear mixing was necessary in order to reduce the particle size of the seeds because the large quantity of seeds was too much for the wet mill to process that resulted in a back-up of material. The protein concentration in the filtrate had increased as expected and resuspension of the remaining material collected on the sieve was at pH 7.31 which was adjusted to pH 8. The adjustment to pH 8 resulted in no change in the concentration of soluble protein in contrast to a previous study that revealed the quantity of protein extracted increased from pH 7 to pH 9 (Piyakina et al., 1995). Furthermore, tomato protein extract has been shown to interact with sunflower oil to form oil emulsions consisting of fine droplets measuring 533 nm at pH 5 (Sarkar et al., 2016). Much larger droplet sizes of 8000 nm were formed at pH 3. Consequently, the presence of oil, albeit at low quantities, will result in some of the protein becoming associated with the oil. The resuspensions of the remaining material collected on the sieve were at pH 7.51 on the third cycle and pH 7.67 on the fourth cycle and were readjusted to pH 8 each time. The protein concentrations recovered after the first and subsequent stirring cycles revealed considerably lower quantities were being recovered. The lower quantities of protein being recovered in subsequent cycles indicated that stirring was ineffective, and that wet milling was required during the subsequent steps. It would be expected that a further reduction in particle size would increase surface area supporting the transfer of proteins closer to the surface into the suspension. The laboratory study revealed that a high liquid-to-seed material of 500 ml/g was necessary to recover most of the protein during the first step (
Table 2). However, the use of larger quantities of liquid becomes difficult to handle during processing, necessitating the use of multiple extractions. A previous experiment has shown that three consecutive extractions increased protein yield recovery by 42% compared with the first initial extraction (Piyakina and Yunusov, 1955). The first step in both laboratory experiments (20–200 g of tomato seed material) and the pilot scale extraction was similar involving high shear mixing and similar ratios of liquid to solid. However, the yield of protein was two-fold lower in the pilot scale extraction compared with the laboratory extraction indicating that another factor had affected the recovery of protein.
Table 2. Protein extraction from wet tomato seeds using consecutive high shear mixing extractions using different starting quantities.
| Protein concentration (mg per g DM of original seeds) |
|---|
| Sample | 1 g 1 | 20 g 1 | 200 g 2 | 800 g 3 |
|---|
| Extraction volume (ml) | 500 | 1000 | 4000 | 20,000 |
| Liquid/solid ratio (ml/g) | 500 | 50 | 20 | 25 |
| Hulls | 16.3 | 62.52 | 32.9 | ND |
| First extract | 86.0 | 78.69 | 52.5 | 22.4 |
| Second extract | ND | 2.98 | 17.7 | 1.9 |
| Third extract | ND | ND | 7.4 | 0.9 |
The bacterial population during the pilot-scale processing steps remained low and appeared to decrease during the subsequent extraction steps as any bacteria remaining with the equipment was washed away (
Table 1). However, the use of mild acidification to prohibit microbial growth during overnight storage was unsuccessful resulting in an increase in the bacterial population. The suspension showed a minor increase from pH 5.03 to pH 5.37 on the following day. Ideally, continuous processing would have been preferential and instead the protein extracts were stored at 4 °C during the subsequent trial with tomato pomace. The soluble protein content at this stage was higher compared with the sum of the soluble protein in the four separate filtrates which was perhaps caused by protease activity associated with the bacteria releasing insoluble protein that was not measured. The bacterial population remained relatively stable thereafter, but the soluble protein concentrations appeared to have decreased by the end of ultrafiltration. There are several possible explanations for the decrease in protein such as degradation caused by bacterial contamination or the presence of oils causing the proteins to adhere to the surface of the ultrafilter membrane. The protein concentrations associated with the solids enriched suspension collected by the stack disc centrifuge were also low because it is likely that most of these proteins were insoluble. The predicted soluble protein yield was highest with the pooled filtrates at 7.7% per g dry mass of tomato seeds and subsequently decreased to 0.6%. However, this calculated protein yield did not take into account insoluble protein associated with solids enriched suspension collected by the stack disc centrifuge, which was not measured. Rather than using ultrafiltration to concentrate the protein, acid precipitation followed by centrifugation has been previously used in the recovery of protein from rapeseed (Helstad et al., 2022). Consequently, a protein yield of 57% and a protein content of 91% was obtained when using alkaline extraction at pH 10.5. The crude protein yield after spray drying in our study was 21.88 g (2.7% of seeds DM) and appeared light yellow (
Figure 2a). The powder contained a protein content of 40.3% as determined by Kjeldahl analysis and 35% as determined using the Bradford assay. A higher protein content of 92% (w/w) was recovered from defatted tomato seeds using a similar procedure except seed hulls were centrifuged at a higher speed (Sarkar and Kaul, 2014) although it is anticipated that a lower quantity of protein would be recovered. Protein recovery from lupin seeds appeared to be more effective due to the ease of dehulling resulting in kernels having a much higher initial protein content of 60% (Burgos-Díaz et al., 2019). Consequently, the use of more alkaline conditions and similarly high centrifugation speeds to remove remaining seed particles resulted in a protein content of 82% (w/w). While it is important to recover as much of the protein from tomato seeds, the low protein content of 40% (w/w) in this study would have limited applications in foods at that concentration (Sari et al., 2015) and would require further processing to remove associated components. This yield was lower than expected compared with laboratory trials (
Table 2) where the yields appeared to be three-fold higher. The failure to use additional wet milling during subsequent extraction cycles was likely to be one of the contributing factors. In addition, it was evident in a previous study that the presence of plant oils during the recovery of dietary fibres from olive pomace had a profound negative effect in pilot scale extraction compared with the laboratory experiments (Baker et al., 2023). Hexane extraction of the crude protein to remove fats and oils resulted in an increase in the protein content from 35% to 47% as determined using the Bradford assay. The protein extract before hexane extraction appeared pale yellow whereas another study revealed that alkaline extraction of tomato seed protein may appear darker possibly caused by the presence of phenolics despite having a higher protein content of 84% (Rana et al., 2025).
Figure 2. Final spray dried protein extracts from (a) tomato seeds and from (b) tomato pomace.
The processing of the whole tomato pomace was initiated using 2.806 kg in wet weight (833 g in dry weight) that comprised 70.4% seeds and 29.6% pomace. The pomace pellet collected on the sieve after the third and final filtration weighed 3 kg which had been squeezed by hand to remove as much water as possible. The tomato pomace after hand pressing had a dry weight content of 13%. Therefore, the remaining dry weight of material leftover after three successive extractions was 0.39 kg representing 47% of the original material at the start. The pooled filtrate was determined to be pH 4.20 and was stored in the fridge which decreased to pH 4.03 when left overnight. The protein content in the extracts was considerably lower than the extracts recovered from the separated seeds even after accounting for the proportion of seeds in the tomato pomace (
Table 1). The spray dried protein extract weighed 42.2 g (6.0%) and contained a protein content of only 7% as determined using the Bradford assay, which reflected the lower protein content in tomato pomace. Furthermore, the spray dried crude protein appeared dark green (
Figure 2b). Therefore, tomato seeds which contain most of the protein compared with the rest of the tomato pomace (Baker et al., 2022) must be separated from the remaining tomato pomace to efficiently recover protein in terms of yield and colour.
In an independent part of the study, the proportion of amino acids and peptides was assessed using a modified version of the Kaiser test. This test was originally designed to determine whether amino acids were present in a sample during protein synthesis (Kaiser et al., 1970). However, more recent studies have applied this method to quantify amino acids revealing most had similar levels of detection except cysteine and lysine which were barely detectable (Stauß et al., 2024). Detection is based on the reaction of ninhydrin with primary amines (Jarre et al., 2014) and therefore peptides would have only one amine group despite having a much larger molecular weight. These semi-quantitative results revealed the presence of amino acids and peptides were associated with the seed extract. It was assumed as shown with the soluble proteins that most of peptides would be soluble and found in the seed extract rather than remaining with the seed hull pellet. It was unclear about the proportion of amino acids and peptides. Proteolysis revealed that the total content of amino acids and peptides had appeared to have increased assuming that the proportion of amino acids to peptides remained unchanged. This increase was accompanied by a decrease in the soluble protein content within the seed extract (
Table 3). Likewise, amino acids and peptides were also found associated with the seed hull pellet after proteolysis, but soluble protein content associated with the seed hulls appeared to have shown a minor increase. It is highly probably that some insoluble protein remained with the seed hull and proteolysis may have solubilized some of this protein. A further reduction of the soluble protein associated with the seed extract could have been achieved by increased proteolysis incubation time. The recovery of soluble and insoluble proteins, and peptides from the suspension would require ultrafiltration through a 10 kD filter membrane followed by spray drying the retentate. The use of proteases in the recovery of protein from plant biomass appears to have the highest success in yields compared with the use of other enzymes (Baker and Charlton, 2020).
Table 3. Treatment of tomato seed extract and tomato seed hulls remaining after extraction with a protease and measurement of proteins and peptides.
| Neutrase proteolysis | Sample | Concentration (mg/g dry seeds) | Yield (% of dry weight) |
|---|
| | Protein | peptides | protein | peptides |
|---|
| Before | seed extract | 52.5 | 51.4 | 18.2 | 18.8 |
| seed hull pellet | 1.2 | ND | 0.4 | |
| After | seed extract | 26.4 | 173.2 | 9.2 | 60 |
| seed hull pellet | 1.4 | 18.8 | 0.5 | 6.1 |
It is perhaps worth considering the value of an industry that could be created using the protein recovered from tomato pomace. This would be based on the values described by Almeida et al. (2024) within Europe of 16.9 million tonnes of tomatoes of which 40% are processed. If the dry weight of the average tomatoes is 18% and the seed composition is 3.5% (Baker and Charlton, 2020), this would amount to 42,500 tonnes of seeds. The recovery of about 4000 tonnes of crude protein could be achieved assuming a recovery yield of just under 10% was possible based on current estimates. It was found that expected purchasing prices by different types of industry of protein flour, concentrates and isolates would be €3, €4–€5, €10–€14, respectively (Pinilla and de la Fuente, 2018). These values could be used to calculate the total income that could be generated. The low protein content of the crude tomato seed protein would be classified as a protein flour and therefore the combined European market value of protein from tomato seeds could be worth €12 million during 2018. However, this highly approximate income does not take into account investment and operational costs.
Despite the relatively low income associated with this protein source, one of the beneficial factors associated with tomato seed protein is the presence of a higher concentration of lysine, an essential amino acid, compared with proteins recovered from other major sources such as soy and corn (Rana et al., 2025; Mechmeche et al., 2017). In addition, tomato seed protein has been shown to have high oil absorption (Liadakis et al., 1998; Sogi et al., 2002). A further investigation into the oil emulsion property of tomato seed protein revealed that it was stable in the presence of sodium chloride, at above pH 4, at a temperature below 80 °C, and in calcium chloride below 50 mM (Sarkar et al., 2016). These properties could make tomato seed protein more desirable in the market and that proteins recovered from other plant seeds such as rapeseed showed poor solubility in water (Nasrabadi et al., 2021).
Conclusions
A full pilot scale protein extraction from tomato seeds was performed which revealed problems within the process leading to low yields compared with initial laboratory experiments. Suggestions have been considered that could significantly increase protein yields such as the use of high shear mixing instead of wet milling. Furthermore, the presence of oils and fats associated with the seeds appeared to significantly impact protein concentration during ultrafiltration and it appears that protein precipitation by acidification would be more effective. At his stage, it would be possible to treat the crude protein extract with hexane to remove any remaining oils and fats before progressing to enzyme hydrolysis with protease, followed by ultrafiltration. This process would shift hexane extraction from the initial stage with tomato seeds to much later resulting in a major reduction in solvent usage. A comparison with pilot scale protein extraction from tomato pomace revealed a lower yield even when taking into account the lower proportion of seeds.
Ethics and consent
Ethical approval and consent were not required.
