Discarded and uneatable parts of vegetables and fruits were collected from various roadside markets (mandis), in Delhi/NCR region, India. These included cauliflower/cabbage stubs, peels of peas, the skin of banana/oranges/sweet lime, and fruit parts. The waste was frozen after collection and segregation, to prevent the loss of sugar. The carbon (C) and nitrogen (N) content of the waste material was determined individually so that they could be mixed in appropriate proportions to achieve the desired C/N ratio of 20:1, which has been standardized by our earlier study (Sinha et al., 2020). After weighing the waste (30 g), the peels were chopped, added to water (100 ml), and steamed under pressure in an autoclave, to prepare about 100 ml of the hydrolysate as reported in detail earlier (Singh et al., 2018).

The organism used in this study was R. toruloides ATCC 204091 (Sinha et al., 2020). Peels of various seasonal fruits and inedible parts of vegetables (stubs of cabbage and cauliflower, pea pods) were collected from the local vegetable markets (“mandis”), to prepare a hydrolysate Waste Extract (WE) and used as fermentation medium. Optimal nitrogen C:N ratio of 20:1 was maintained in the WE medium by mixing appropriate amounts of high carbon containing fruit waste. After filtering and cooling to room temperature, the hydrolysate was used as a culture medium. Fermentation of yeast was carried out in the 500-ml shake flasks, each containing 100 ml cultivation media, and was inoculated with a starter culture (10% v/v) and incubated at 30°C for 120 h, as described earlier (Sinha et al., 2020).

The carotenoid was extracted from yeast cells by solvent extraction following the protocol given by Hiscox and Israelstam (1980), after suitable modifications (Figure 1A; Sinha et al., 2020). Optical density (OD) was determined via Systronics UV-Vis spectrophotometer at 400 nm against hexane. Quantitative analysis was done by means of a standard curve of β-carotene (Sinha et al., 2020). Thin-layer chromatography (TLC) analysis was carried out using silica gel 60 TLC plates (Merck, Germany) and petroleum ether and acetone (80:20 v/v) as a mobile phase and Rf values were determined as described earlier (Marova et al., 2010). The carotenoids were also separated and analyzed by high-performance liquid chromatography (HPLC) on Agilent, C₁₈ column, and a gradient solvent system of water: acetonitrile was used for separation of pigments, followed by detection at 250 nm (UV-Visible Detector).

The antimicrobial property was evaluated by the agar diffusion method where wells on a Nutrient Agar plate were loaded with 20 μL of carotenoid extract (10 mg/ml). The plate was previously spread with 10⁸ CFU/ml of microbial suspension. After incubation at 37°C for 24 h, inhibition zones around the spot were measured in millimeters. Wells contained only hexane (negative control) and 20 μL of 1.0 mg/ml antibiotic ciprofloxacin (positive control). Both gram-positive and gram-negative bacteria were used as mentioned in the table (Supplementary Table 1) for testing antibacterial activity. Two yeast strains, i.e., Candida albicans (MTCC 183), and Saccharomyces cerevisiae (MTCC 170), and fungal strains, i.e., Fusarium oxysporum (MTCC 2773) and Aspergillus niger (MTCC 281) were used to test antimicrobial activity. For positive control in the case of yeast and fungal strains, amphotericin B was used (20 μg/ml). The % inhibition was calculated by taking the diameter of the antibiotic inhibition zone as 100%.

2,2-Diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity was determined by taking various concentrations of the carotenoid extract (10–50 μg/ml), exposing it to 1.5 ml freshly prepared solution of DPPH in methanol for 20 min, and then measuring the absorbance at 515 nm. The percentage inhibition of DPPH in the medium was calculated by comparing it with the control (untreated DPPH). Percentage inhibition was calculated as below:

Cell culture reagents were procured from Gibco (Thermo Fischer Scientific, Inc., USA). Two human breast cancer cell lines, MCF7 and MDA-MB-231, and one human normal cell line HEK 293T were procured from NCCS, Pune, India. Breast cancer cells were cultured, and cellular cytotoxicity was evaluated against carotenoid extracts treated in increasing concentrations, as described earlier (Selvendran et al., 2021). Graphical plotting of concentration vs. OD values was analyzed, and inhibition of proliferation by the extract was compared to vehicle-treated control. IC₅₀ values (concentration showing 50% proliferation inhibition) for carotenoid extract were calculated for both MCF-7 and MDA-MB-231 cells, and the extract was further evaluated in normal cells (HEK 293T) with corresponding IC₅₀ concentrations to check whether the anti-proliferative potency is restricted to cancer cell line only (de Souza et al., 2021; Fatima et al., 2021; Kaur et al., 2021).

In silico molecular docking studies were performed to understand possible receptor mediated anti-breast cancer pathways through which the active molecules in the carotenoid extract may exert anti-proliferative activity. Torularhodin, torulene, and β-carotene structures were docked into the active site of several human breast cancer receptors via estrogen receptor (ERα, ERβ), progesterone receptor (PR), human epidermal growth factor receptor 2 (HER2/ErBB2), vascular endothelial growth factor receptor 1 and 2 (VEGFR1 and VEGFR2), fibroblast growth factor receptors (FGFR1 and FGFR2), and EGFR. Protein structures were downloaded from Research Collaboratory for Structural Bioinformatics (RCSB)¹ protein data bank, followed by a protein preparation wizard where pre-process (assigning bond order, adding missing hydrogen, filling missing loop, and missing side chain) and refinements (optimization of H bond, minimization using OPLS4 force field) were done. Ligand structures were downloaded from the PubChem database and prepared using the LigPrep wizard (charge fixation, orientation of groups, 2D to 3D conversion, corrected bond lengths and bond angles, adding hydrogens) using the OPLS4 force field. Receptor grid and task grids were generated on proteins and ligands were docked using the ligand docking task of the GLIDE application (Maestro 12.8, Schrödinger Suite- 2021-3).

To evaluate drug likeness and pharmacological profile, the QikProp application was used, which reports a 95% range for known drugs with several descriptors and properties of Schrödinger workflow to evaluate in silico ADME properties. The probability of toxicity was also predicted using admetSAR online tool (Cheng et al., 2012).

All assays were performed in triplicates and expressed as mean ± standard deviation (SD). Data analysis was performed with the software package Microsoft Excel, version 2016. Statistically significant difference was determined using one-way Analysis of Variance (ANOVA). and p ≤ 0.05 was used as a limit to indicate statistical significance.

To grow the red yeast strain, a medium was prepared using “mandi” waste consisting of inedible peels of fruits and unused parts of vegetables. Fruit and vegetable peels having a high content of carbon were added to waste containing appropriate nitrogen content, to maintain the C/N ratio to 20:1 in the WE that promotes carotenoid accumulation over lipid production (Sinha et al., 2020). The maximum amount of carotenoid (62 ± 0.93 mg/L) was produced by R. toruloides at 88 h of growth in the above medium. The extracted pigments were separated using acetone:petroleum ether (20:80) as mobile phase (Figure 1A). TLC results showed that the Rf value of a band from the extract (0.90) was similar to that of a standard β-carotene spot (Yehia et al., 2013). Three main pigments (i) β-carotene, (ii) torulene, and (iii) torularhodin were visually separated as also shown earlier (Park et al., 2007; Minyuk and Solovchenko, 2018). HPLC analysis confirmed the above finding where β-carotene appeared at 3.9 min and torulene and torularhodin at 1.8 and 1.7 min (Figure 1A), whereas other smaller peaks remained unidentified. Several studies on carotenoid production from yeast strains on different mediums have been reported previously, showing that the waste of different types, e.g., cassava-bagasse, sago-starch hydrolysate, and whey, (Sinha et al., 2020) is suitable for cultivation of oleaginous red yeast for carotenoid production. Here, we developed a culture medium using discarded vegetable parts and grown oleaginous red yeast Rhodosporidium sp. Carotenoid extract analysis revealed mainly three major compounds, i.e., β-carotene, torulene, and torularhodin, similar to HPLC analysis observed for R. glutinis grown under external stress (Marova et al., 2010).

The pink lipid soluble carotenoid extract was found to be very effective against both gram-negative and gram-positive strains and yeast. It was, however, not effective against fungal cultures. Comparing diameters of the inhibition zone (Supplementary Table 1) showed that it was most active against Escherichia coli and Pseudomonas sp. as compared to Staphylococcus aureus, Micrococcus, Candida sp., etc., The carotenoid extract was able to scavenge stable free radical DPPH and its IC₅₀ was 25 ± 5 μg/ml (Supplementary Figure 1). Earlier reports, using DPPH assay, showed comparable IC₅₀ values for astaxanthin 79.32 ± 18.10, lutein 35, and zeaxanthin 10 μg/ml (Neumann et al., 2019).

We were further compelled to evaluate the anti-proliferative properties of the extract as many antioxidant natural compounds have shown anticancer activities and are being pursued as anticancer drug candidates (Kaur et al., 2021). When screened against human breast cancer cell lines, in vitro, the carotenoid extract showed dose-dependent inhibition of proliferation and IC₅₀ in MCF7 and MDA-MB-231 cells were calculated as 29.11 and 7.82 μg/ml, respectively (Supplementary Figure 2). The carotenoid extract was also found to be biocompatible as we obtained insignificant cytotoxicity in normal human cells HEK293T when treated with these IC₅₀ concentrations separately (Figure 1B). These results indicated the selective anti-proliferative potency of carotenoid extract in breast cancer cells.

The major chemotherapeutic drugs used nowadays to suppress breast cancer cells are developed for targeted delivery of chemo-agents against several molecular and cellular signaling pathway proteins, mainly, for example, ER, PR, HER2, VEGFR1, VEGFR2, and EGFR (Sarkar and Mandal, 2009; Jin and Ye, 2013). We conducted in silico molecular docking to evaluate whether the active compounds of the extracts, i.e., β-carotene, torulene, and torularhodin can bind or interact with the breast cancer-specific major molecular targets. After their molecular docking with several breast cancer-specific cellular targets, binding energy was calculated (ΔG, reported as kcal/mol) and listed (Table 1). Among all, molecular interaction between VEGFR2 and torularhodin reflects the highest binding energy (−10.092 kcal/mol) followed by VEGFR1 > ERα > ERβ > HER2. Torularhodin formed a hydrogen bond with asparagine 923 at the active site of the protein VEGFR2 and the carboxyl group moiety of the molecule. A similar type of interaction was also found between torularhodin and VEGFR1 and between lysine 831 and tyrosine 911 of the VEGFR1 active site (Figures 1C,D). The binding energy of torulene was highest with VEGFR2 (−9.217 kcal/mol) followed by VEGFR1 > ERα. The binding energy for β carotene in decreasing order was found as HER2 > VEGFR2 > VEGFR1 > ERα. Our results indicated that the selective anti-proliferative activity may be governed through binding with cellular tyrosine kinase receptor VEGFR (VEGFR1 and VEGFR2). Interestingly expansion and regulation of the breast cancer stem cell population directly depend on VEGFR2 expression (Zhao et al., 2014; Farzaneh Behelgardi et al., 2020), and inhibitors against VEGFR1 and VEGFR2 have been designed to suppress tumor growth and under the clinical application. According to a recent study, it has been demonstrated in animal breast cancer models, that carotenoids as phytochemicals showed anti-proliferative properties through VEGFR down-regulation (Metibemu et al., 2021).

Several properties were evaluated in silico for drug likeness and toxicities of torularhodin, torulene, and β-carotene (Table 1). All showed good molecular weight (∼500 kDa), with donor and accept HB values within suitable range (below 5) and not violating the rule of five (maximum being 4) and rule of three (maximum being 3). Among the three compounds, the predicted brain/blood partition coefficient (QPlogBB) of torularhodin was good. The probability of crossing the blood–brain barrier and gut–blood barrier was high for torulene and β carotene while torularhodin was moderate. Torulene and β-carotene were shown to be active compounds in central nervous system (CNS). The probability of human oral absorption percentage (HOA %) was high for all, besides all being non-toxic (AMES toxicity negativity probability > 75%).