Background

Plants are used to cure various ailments by traditional medicinal practitioners. Despite the efficacy of synthetic pharmaceuticals, many people prefer plant-based treatments over synthetic drugs since they are more natural and have fewer side effects [1]. Herbal medicines contain a complex combination of phytochemicals with a wide range of structures. These compounds are responsible for the pharmacological properties displayed by plants [2]. Plants have produced many compounds of pharmacological importance. Examples include the anticancer drugs vinblastine and vincristine (from Catharanthus roseus (L.) G. Don), taxol (from Taxus brevifolia Nutt), etoposide (a semi-synthetic analogue of podophyllotoxin, isolated from Podophyllum species), and topotecan (an analogue of camptotethin, isolated from Camptotheca acuminata Decne) [3, 4]. Plants will continue to be sources of lead compounds for fuelling the drug development pipeline, given that only a small fraction of the world's biodiversity has been investigated for bioactivity so far.

Fast biological and chemical screening methods are used in the quest for novel bioactive chemicals from plants. Chemical screening involves the online characterization of bioactive compounds in crude extracts of plants using various hyphenated techniques such as high-performance liquid chromatography coupled with mass spectrometry (LC–MS or LC–MS-MS), high-performance liquid chromatography coupled with UV photodiode array detection (LC-DAD-UV), and gas chromatography coupled with mass spectrometry (GC–MS)[5]. The chemical screening approach provides preliminary information on chemical constituents and potential pharmacological properties of the plant under investigation. The GC–MS technique has various advantages. These include high chromatographic separation power, reliable quantification methodologies, and identification of metabolites with high accuracy [6]. GC separates low molecular weight metabolites that are either volatile or can be transformed into volatile and thermally stable molecules by chemical derivatization. To improve volatility and reduce the polarity of polar hydroxyl (-OH), amine (-NH2), carboxyl (-COOH), and thiol (-SH) groups, derivatization is often required [7]. In the biological screening approach, extracts of plants are subjected to various assays to determine their potential pharmacological properties.

Hepatocellular carcinoma (HCC) is the world's third leading cause of cancer death. Around 953,000 cases of liver cancer were reported worldwide in 2017, with 819,000 people dying from the disease [8]. In Sub-Saharan Africa, liver cancer was the second leading cause of cancer-related deaths in males and the fourth in women in 2020. The high prevalence of risk factors (such as alcohol consumption and cigarette smoking, aflatoxin B1 pollution, and hepatitis B and C infection) in Sub-Saharan Africa contributes to the region's high HCC rate. HCC has a poor prognosis in black Africans, with a median survival rate of 3–4 months [9]. Mortalin, an hsp70 chaperone protein, heightens liver cancer incidence by sequestering the p53 tumor suppressor protein into the cytoplasm, thereby inhibiting its cellular functions. p53 functional inactivation extends the longevity of normal human cells and accelerates the malignancy of cancer cells [10]. Inhibiting the mortalin–p53 interactions has therefore been proposed as a viable anticancer strategy against HCC. Many compounds isolated from plants are known inhibitors of p53–mortalin interaction. For example, Pham et al. [11] identified nine triterpenes (ailanthusin A, ailanthusin B, ailanthusin C, ailanthusin D, ailanthusin F, ailanthusin G, schisanlactone C, ailanaltiolide I, and ursolic acid) as promising p53–mortalin inhibitors, based on their binding affinity and drug-like and pharmacokinetic properties; campersterol, a constituent of black rice (Oryza sativa L.), was identified as an inhibitor of the p53–mortalin interaction [12]; solasonine, a steroidal glycoalkaloid from Solanaceae, is a potent inhibitor of p53–mortalin interactions [10]; embelin, a natural quinone found in the fruits of Embelia ribes Burm. f., inhibits mortalin–p53 interactions and activates p53 protein in tumor cells [13], and salvianolic acid B, a caffeic acid phenethyl ester analog, binds to mortalin [14].

Several Crinum species are used in ethnomedicine for treating various cancer types. Also, literature search shows that several Crinum species are active against many cancer cell lines. For example, bulb extract of Crinum ornatum (L.f. ex Aiton) Bury is used in traditional medicine to treat breast cancer [15], Crinum bulbispermum ((Brum. f.) Milne-Redhead and Schweickerdt) is used by Zulu, Sotho, and Tswana people to treat tumors, and in Madagascar, Crinum powellii Baker Handb. is used to treat abscesses and tumors [16]. Hot water extracts of C. asiaticum Linn leaves inhibited calprotectin-induced cytotoxicity in MM46 mouse mammary carcinoma cells [17]; palmilycorine, isolated from C. asiaticum, had an inhibitory effect on the viability of ascites tumor cells [18]. The polysaccharide, CAL-n, isolated and purified from dried fresh seeds of C. asiaticum had an IC50 value of 128.07 μg/mL when screened for anti-tumor activity against the HepG2 cell line [19]. Oral intake of hot aqueous extract of C. delagoense I. Verd. bulbs was reported as a cure for human cancer; crinamine, lycorine, and 6-hydroxycrinamine isolated from the C. delagoense bulbs showed activity against BL-6 mouse melanoma cells [20]. Shawky and co-workers [16] showed that Genapol X-80 and DES-3 extracts of Crinum powellii and C. bulbispermum bulbs had enhanced activities against HepG2 and HCT 116 cell lines. Perlolyrine, isolated from the leaves of Crinum latifolium L., showed significant cytotoxicity against five human cancer cell lines, including KB, HepG2, MCF7, SK-Mel2, and LNCaP with the IC50 values ranging from 22.12 ± 2.80 to 28.45 ± 3.75 μM [21].

Traditional healers in southwestern Nigeria use the bulb and leaves extracts of Crinum jagus (J. Thomp.) Dandy, to treat cancer. Ka et al. [22] reported the isolation of gigantelline, gigantellinine, gigancrinine, lycorine‐, cherylline‐, galanthamine‐, and crinine‐type alkaloids from C. jagus bulb. Another study identified the alkaloids: lycorine, hippadine, ambelline, 3-O-demethyltazettine, acetylambelline, crinanine acetate, crinine, acetylcaranine, caranine, O-methylmacronine, crinine acetate, epinorgalanthamine, trispheridine, 3-epimacronine, and voacangine in the bulb of C. jagus [23]. Previous studies indicate that C. jagus bulb extracts have hypoglycemic [15], antiplasmodial, and antimycobacterial [24, 25] properties. Methanol extract of C. jagus bulbs demonstrated cytotoxic and antiproliferative properties in brine shrimp, Allium cepa, and Sorghum bicolor radical growth inhibition assays [26]. While extensive studies have been carried out on the bulb of the C. jagus, no information exists in the literature on the phytoconstituents of the leaves of the plant. Also, there are no published studies on the anticancer potential of C. jagus leaf extracts. This research screens ethyl acetate and methanol extracts of the leaves of C. jagus for antiproliferative properties against HepG2 cell line. In addition, the phytoconstituents in the extracts would be characterized and analyzed for drug-like properties. Finally, molecular docking studies will be conducted on the phytoconstituents to identify potential p53–mortalin binding inhibitors.

Methods

Collection and extraction of plant samples

Fresh C. jagus leaf samples were collected at the Botanical Gardens of the University of Ibadan and identified by Mr. Kayode Owolabi, a taxonomist of the Garden. The plant sample was deposited at the University’s herbarium and was assigned the voucher number: UIH-22441. The leaves were dried under mild sunlight for three weeks and ground into powder. A 1 kg portion of the powdered material was de-fatted by maceration in 2.5 L of n-hexane for 72 h. The mixture was filtered, and the filtrate was discarded. The residue was used for the subsequent steps of the extraction process. The ethyl acetate extract was obtained by soaking the residue in 2.5 L of ethyl acetate for 72 h. The filtrate (extract) was separated from the residue by filtration and concentrated using a rotary evaporator. The ethyl acetate extract was assigned the code CJLEE. The methanol extract was obtained from the residue (obtained after extraction with ethyl acetate) by maceration in methanol for 72 h. The mixture was filtered, and the filtrate (extract) was concentrated using a rotary evaporator. The methanol extract was assigned the code CJLME. The solvents used for extraction (hexane, ethyl acetate, and methanol) were procured from Sigma-Aldrich.

Anticancer screening of selected extracts in HepG2 cell line

The HepG2 cell line was purchased from American Type Culture Collection (ATCC) (USA). The cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (supplemented with 10% fetal bovine serum (FBS) and 1% of penicillin–streptomycin) and were incubated at 37 °C under 5% CO2 in a humidified atmosphere. Cell culture, at a concentration of 2 × 103 cells/mL, was prepared and plated (100 µL/well) onto 96-well plates. Different concentrations (1000 µg/mL, 500 µg/mL, 250 µg/mL, 125 µg/mL, 62.5 µg/mL, 20 µg/mL, and 1 µg/mL) of the extracts (diluted in FBS-free medium) and the Standard (cisplatin) were introduced, as appropriate, into each well. After 72 h, MTT solution was added followed by incubation for 3 h. Dimethyl sulfoxide (DMSO) (100 µl) was used to dissolve the purple formazan crystals formed. Thereafter, the optical density of the plant extract was measured at a wavelength of 570 nm on a microplate reader (Tecan, Switzerland). The percentage cell viability was determined using the expression:

$${\text{Cell}}\;{\text{viability}} = \frac{{{\text{Absorbance}}\;{\text{of}}\;{\text{sample}}\;({\text{mean}})}}{{{\text{Absorbance}}\;{\text{of}}\;{\text{Control}}\;({\text{mean}})}} \times 100 \%$$

The percentages of viable cells were plotted against different extract concentrations. Cytotoxicity was recorded as the IC50 (the drug concentration causing 50% growth inhibition of the tumor cells) [27]. The MTT reagent and DMSO were purchased from Sigma-Aldrich.

Gas chromatography–mass spectrometry analysis of extracts

Derivatization of extract into trimethyl silyl (TMS) forms

The extracts were derivatized into Trimethyl Silyl (TMS) form as described by [28]. A 3 mg portion of each extract was treated with dry pyridine (60 μL) and bis(trimethylsilyl)trifluoroacetamide in 1% trimethylchlorosilane (100 μL). The mixture was heated at 70 °C for 30 min. GC–MS is used to analyze an aliquot of the resultant solution. Pyridine and bis(trimethylsilyl)trifluoroacetamide in 1% trimethylchlorosilane were obtained from Sigma-Aldrich.

Derivatization of extract into fatty acid methyl ester (FAME) forms

The extracts were derivatized into methyl ester forms using procedures described by [29]. A 5 mg portion of each extract was weighed into test tubes. Each extract was heated with a CH3OH/H2SO4/CHCl3 (1.7:0.3:2.0 v/v/v, 4 mL) mixture for 90 min at 90 °C. After cooling, 1 mL of water was added and vigorously shaken. Using a Pasteur pipette, the CHCl3 layer (lower part) was gently removed and transferred to GC vials. GC–MS was used to analyze the prepared FAME derivatives. Methanol, chloroform, and tetraoxosulfate (VI) acid were obtained from Sigma-Aldrich.

GC–MS analysis of derivatized extracts

An Agilent 7890 Gas Chromatograph coupled to an Agilent 5975C mass selective detector in electron impact mode (ionization voltage, 70 eV) was used to analyze the derivatized samples. An Agilent Chrompack CP-Wax 52 CB capillary column was used to separate the sample constituents. The column had an internal diameter of 0.32 mm, a column length of 30 m, and a film thickness of 0.25 m, respectively. A 1.0 μL portion of the diluted sample was injected splitless into the chromatograph at a temperature of 250 °C. Helium, flowing at 5 mL/min, was the carrier gas. The inlet pressure was 12.936 p.s.i. The column oven temperature was gradually increased from 50 to 240 °C at 8 °C/min. The total time spent on hold was 5 min. The sample contents were identified by comparing their spectra to the Mass Spectral Library of the National Institute of Standards and Technology (NIST) on the GC–MS database (NIST 14L).

Prediction of drug-likeness of phytoconstituents

The ADME-related physicochemical properties and the drug-likeness of the bioactive compounds detected in C. jagus extracts were predicted by pasting the SMILES format of each compound in the SwissADME online web server (https://www.swissadme.ch). For drug likeliness, the molecular parameters such as molecular weight, hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), lipophilicity log (log P), and topological polar surface area (TPSA) were assessed. The cutoff values of the properties were set by Lipinski’s rule of five (ROF). The physicochemical properties analyzed include gastrointestinal (GI) absorption and permeability of the blood–brain penetration barrier [30].

Molecular docking analysis

Ligand preparation

After evaluating the bioactive molecules identified in the extracts for drug-likeness, crinamine, galanthamine, tangeretin, hexadecanoic acid, and hexadecanoicacid methyl ester were selected for docking with the target proteins. Withaferin A and withanone, which are known inhibitors of mortalin, were chosen as standards against which the binding properties of the phytoconstituents were measured [31]. Structures of crinamine, galanthamine, tangeretin, hexadecanoic acid, hexadecanoic acid methyl ester, withaferin A, and withanone were obtained from the PubChem database and saved in the SDF format. Before docking, these structures were converted into mol2 format using the Open Babel software [32].

Protein preparation

There are two possible binding sites of mortalin into p53: the peptide binding domain, matching with PDB ID: 3N8E (residues 439–597), and the residues 253–282 domain that matched with PDB ID: 4KBO [33, 34]. These were the two models selected for the current study which were obtained from the Protein Data Bank.

Blind docking

To investigate the interaction between crinamine, galanthamine, tangeretin, hexadecanoic acid, and hexadecanoicacid methyl ester and the target proteins, docking was carried out using the SwissDock Server (http://www.swissdock.ch/) [35]. The proteins and ligands, (in the PDB and mol2 formats, respectively) were uploaded as appropriate. A link containing each docking result was sent to the user’s email. All the possible binding modes for each ligand were generated by SwissDock. The most favorable binding modes at a given pocket were clustered and saved in an output file referred to as the prediction file which contains information such as estimated binding free energy, cluster rank, element, and the full fitness score. The lowest energy model was considered to be the most favorable interaction. After docking, Chimera was used to visualize the interactions between the receptors and the ligands. The amino acids interacting with the ligands, the specific atoms involved, and the hydrogen bonds formed were identified [36].

Results

To investigate the extracts for antiproliferative properties, the percentage of viable cells was plotted against different extract concentrations, as shown in Fig. 1. From the graphs, IC50 values of 5 μg/mL, 5 μg/mL, and 70 μg/mL were obtained for the standard (cisplatin), CJLME, and CJLEE, respectively. The error bars represent the standard deviation of the data sets.

Fig. 1
figure 1

Plots of percentage viability of HepG2 cells against different extract concentrations

Full size image

The compounds identified in the ethyl acetate and methanol extracts of C. jagus leaves are neophytadiene (1), hexadecanoic acid methyl ester (2), n-hexadecanoic acid (3), 3′,4′,5,6,7,8-hexamethoxyflavone (nobiletin) (4), eicosane (5), 4′,5,6,7,8-pentamethoxyflavone (tangeretin) (6), stearic acid (7), eicosanoic acid (8), docosanoic acid (9), hexadecanoic acid ethyl ester (10), galanthamine (11), and crinamine (12). The peak areas and the elution times of each component are shown in Tables 1, 2, 3, 4. The GC–MS chromatograms of the derivatized ethyl acetate and methanol extracts are shown in Fig. 2. The structures of the compounds are shown in Fig. 3.

Table 1 Compounds detected in TMS-derivatized CJLEE
Full size table
Table 2 Fatty acids in FAME-derivatized CJLEE (detected as methyl esters)
Full size table
Table 3 Compounds detected in TMS-derivatized CJLME
Full size table
Table 4 Fatty acids in FAME-derivatized CJLME (detected as methyl esters)
Full size table
Fig. 2
figure 2

GC–MS chromatograms of the derivatized ethyl acetate and methanol extracts

Full size image
Fig. 3
figure 3

Structures of phytoconstituents detected in ethyl acetate and methanol extracts of Crinum jagus Leaves

Full size image

Tables 5 and 6 show the physicochemical and pharmacokinetic parameters of the compounds detected in the ethyl acetate and methanol extracts.

Table 5 Physicochemical parameters computed for bioactive compounds identified in Crinum jagus extracts
Full size table
Table 6 Pharmacokinetic properties of phytoconstituents identified in Crinum jagus extracts
Full size table

Based on the physicochemical and pharmacokinetic properties of the compounds, hexadecanoic acid, methyl ester, n-hexadecanoic acid, tangeretin, galanthamine, and crinamine were selected for docking with the mortalin residues, 3N8E and 4KBO, using the SwissDock Web server (http://www.swissdock.ch). The aim was to determine if any of the compounds could play a role in disrupting the mortalin–p53 interaction.

Crinamine, galanthamine, tangeretin, hexadecanoic acid, and hexadecanoic acid methyl ester bound with the mortalin residues 439–597 (PBID:3N8E) with energy of − 7.35, − 7.06, − 7.91, − 8.21, and − 8.21 kcal mol−1, respectively (Table 7). The binding energy of tangeretin, hexadecanoic acid, and hexadecanoic acid methyl ester was very close to that of withaferin A (which had a binding energy of − 8.29 kcal mol−1). Hexadecanoic acid and hexadecanoic acid methyl ester bind better to the protein residue (3N8E) than withanone (with a binding energy of − 8.14 kcal mol−1). Also, from the full fitness scores, tangeretin, hexadecanoic acid, and hexadecanoic acid methyl ester had more favorable binding modes than withaferin A and withanone. The docking results showed hydrogen bonds and other weaker interactions exist between the ligand and 3N8E. The docking poses of crinamine (A), galanthamine (B), hexadecanoic acid (C); hexadecanoic acid methyl ester (D), tangeretin (E), withaferin A (F), and withanone (G) against 3N8E are shown in Fig. 4. The hydrogen bonds are shown in purple while yellow bonds represent other clashes/interactions between the ligands and protein. Table 7 shows the docking results of the ligands with the protein residue.

Table 7 Docking results of mortalin residue 3N8E with selected phytoconstituents
Full size table
Fig. 4
figure 4

Docking poses of phytoconstituents against mortalin residue, 3N8E

Full size image

Crinamine, galanthamine, tangeretin, hexadecanoic acid, and hexadecanoic acid methyl ester had binding energies of − 6.99, − 7.47, − 7.25, − 7.66, − 7.45 kcal mol−1, respectively, with 4KBO. Withaferin A and withanone had binding energy of − 7.68 kcal mol−1 and − 7.59 kcal mol−1, respectively (Table 8). The binding energy of galanthamine, hexadecanoic acid, and hexadecanoic acid methyl ester was very close to those of withaferin A and withanone. Hexadecanoic acid binds better with the protein than withanone. In addition, galanthamine, hexadecanoic acid, and hexadecanoic acid methyl ester (with full fitness scores of − 1947.2039, − 2018.7064, and − 2003.8347 kcal mol−1, respectively) had better binding modes than withaferin A and withanone (with full fitness scores of − 1702.6406 and − 1665.646 kcal mol−1, respectively). The ligands had hydrogen bonds (and other weaker) interactions with the protein residue. Different docking poses of crinamine (A), galanthamine (B), hexadecanoic acid (C), hexadecanoic acid methyl ester (D), tangeretin (E), withaferin A (F), and withanone (G) against 4KBO are shown in Fig. 5. The hydrogen bonds are shown in purple while yellow bonds represent other clashes/interactions between the ligands and protein.

Table 8 Docking results of mortalin residue 4KBO with selected phytoconstituents
Full size table
Fig. 5
figure 5

Docking poses of phytoconstituents against mortalin residue, 4KBO

Full size image

Discussion

After de-fatting, the dried leaf samples were extracted sequentially using solvents of varying polarities (ethyl acetate is less polar than methanol), thereby separating the phytoconstituents in the extracts based on their polarities. The extracts under investigation will, therefore, exhibit varying antiproliferative properties. The graph obtained by plotting percentage of viable cells against different extract concentrations showed that the cell viability decreased with the increasing concentration of the extracts (Fig. 1). The results obtained from previous studies have shown similar trends [37]. The IC50 values obtained indicate that the antiproliferative property of CJLME is comparable to that of the standard drug (cisplatin). The National Cancer Institute Guidelines (USA) stated that extracts with IC50 < 30 μg/mL are antiproliferative agents [38]. CJLME possesses significant antiproliferative properties.

Literature reports show that many of the compounds identified in the ethyl acetate and methanol extracts possess anticancer properties. Nobiletin and tangeretin have antiproliferative activities against gastric cancer, leukemia (HL-60), squamous cell carcinoma (HBT43), B16 melanoma, human lung carcinoma (A549), and T-cell leukemia (CCRF-HSB-2) cell lines [39]. In addition, n-hexadecanoic acid had an IC50 of 0.8 μg/mL against human colorectal carcinoma (HCT-116) cells [40]. Similarly, Crinamine exhibited significant cytotoxic against human breast cancer (BCA-1), human fibrosarcoma (HT-1080), human lung cancer (LUC-1), human melanoma (MEL-2), and human colon cancer (COL-1) and cervical cancer (SiHa) cell lines [41, 42]. These compounds could be partly responsible for the anticancer properties demonstrated by the leaves of C. jagus.

A potential drug candidate must possess a good absorption, distribution, metabolism, and excretion (ADME) profile. With advances in science, it is now possible to predict the ADME properties of new drug candidates in silico [43]. Lipinski and co-workers [44] proposed that a compound must satisfy certain conditions (referred to as the rule of 5 (Ro5)) before it can be a viable drug candidate: the hydrogen bond donors in the compound ≤ 5, the hydrogen bond acceptors ≤ 10, the octanol/water coefficient (log P) ≤ 5, and the molecular weight must be less than 500. When taken together with Lipinski’s Ro5, polar surface area (PSA) can indicate the candidate’s bioavailability [45]. Orally bioavailable compounds with a PSA less than 140A2 exhibit better intestinal absorption [46], while compounds with PSA values less than 70A2 can penetrate the blood–brain barrier [45]. From Table 5, Compound 4 (nobiletin) violated two of Lipinski’s Ro5 requirements (molecular weight and HBA) and therefore has poor potential as a drug candidate. The compound (nobiletin) would also likely possess a low bioavailability since its PSA value is above 140 A2, and was therefore excluded from further study.

Table 6 shows the results of the pharmacokinetics investigations of the selected phytoconstituents detected in the extracts. Drugs must be able to transit across biological membranes seamlessly [46]. A sufficient intestinal absorption of a drug has a bearing on its action and affects its absorption, distribution, and elimination in the body [46]. In addition, a drug must be able to cross the blood–brain barrier (BBB) to exert therapeutic actions on the brain. BBB acts as a sieve that prevents the access of polar molecules to the brain. From the results obtained (Table 6), only compounds hexadecanoic acid methyl ester (2), n-hexadecanoic acid (3), tangeretin (6), galanthamine (11), and crinamine (12) have high gastrointestinal absorption and could transit across the blood–brain barrier. Galanthamine (11) is a substrate for Permeability glycoprotein (Pg-p), an efflux transporter pump found in the cell membrane which is responsible for conveying drugs away from the cell membrane and cytoplasm which causes therapeutic failure when the concentration of the drug is reduced [47]. An optimal clinical drug/ derivative should have high gastrointestinal permeability and low P-gp efflux liability. However, galanthamine has been retained for further study because it is a known drug used to treat certain stages of Alzheimer’s disease (AD) [48]. A bioactivity score of 0.55 shows that the compounds have excellent pharmacokinetic properties [49]. Hexadecanoic acid methyl ester (2), n-hexadecanoic acid (3), tangeretin (6), galanthamine (11), and crinamine (12) are potential drug candidates for the treatment of hepatocellular carcinoma. From Tables 7 and 8, tangeretin, hexadecanoic acid, and hexadecanoic acid methyl ester showed very great potentials as inhibitors of the mortalin residue, 3N8E (Table 7), while galanthamine, hexadecanoic acid, and hexadecanoic acid methyl ester showed the best potentials as inhibitors of 4KBO.

Conclusion

The results obtained from the present study showed that extracts from Crinum jagus possess antiproliferative properties against HepG2 cell lines. The methanol extract of the leaves demonstrated an anticancer activity comparable to that of cisplatin. Of the compounds detected in the sample by GC–MS analysis, hexadecanoic acid methyl ester, n-hexadecanoic acid, tangeretin, galanthamine, and crinamine possess druglike properties. Literature reports indicate that crimanine, tangeretin, and hexadecanoic acid detected in the extracts have anticancer properties against several human cancer cell lines; these compounds could be partly responsible for the cytotoxic properties of the extracts against the HepG2 cell line. Furthermore, n-hexadecanoic acid, hexadecanoic acid methyl ester, galanthamine, and tangeretin were identified as potential inhibitors of the p53–mortalin interaction.