Withanolide-Type Steroids from Withania aristata as Potential Anti-Leukemic Agents

Leukemia is a blood or bone marrow cancer with increasing incidence in developed regions of the world. Currently, there is an ongoing need for novel and safe anti-leukemic agents, as no fully effective chemotherapy is available to treat this life-threatening disease. Herein, are reported the isolation, structural elucidation, and anti-leukemic evaluation of twenty-nine withanolide-type steroids (1–29) from Withania aristata. Among them, the new isolated withanolides, withaperoxidins A–D (1–4) have an unusual six-membered cyclic peroxide moiety on the withasteroid skeleton as a structural novelty. Their structures have been elucidated by means of spectroscopic analyses, including 2D NMR experiments. In addition, extensive structure–activity relationships and in silico ADME studies were employed to understand the pharmacophore and pharmacokinetic properties of this series of withasteroids. Compounds 15, 16, and 22 together with withaferin A (14) were identified as having improved antiproliferative effect (IC50 ranging from 0.2 to 0.7 μM) on human leukemia HL-60 cell lines compared with the reference drug, etoposide. This cytotoxic potency was also coupled with good selectivity index (SI 33.0–9.2) on non-tumoral Vero cell line and in silico drug likeness. These findings revealed that these natural withasteroids are potential candidates as chemotherapeutic agents in the treatment of leukemia.


Introduction
Leukemia is a blood or bone marrow cancer characterized by an uncontrolled hematopoiesis process [1]. Acute myeloid leukemia (AML) is a genetically heterogeneous disease that has a poor prognosis and is the most common leukemia in adults. Its incidence has increased dramatically in highly developed regions of the world. AML involves the abnormal proliferation and differentiation of a clonal population of the myeloid stem as a consequence of mutations in the genes implicated in these processes [2] and epigenetic changes [3]. In acute promyelocytic leukemia (APL), the formation of chimeric proteins such as RUNX1-RUNX1T1 and PML-RARA alter the normal maturation process of myeloid precursor cells [4]. Present-day standard intervention for leukemia consists of chemotherapy and stem cell transplantation [5]. Current chemotherapeutic drugs cause a range of side effects and health complications. Moreover, at a time when specific prevention efforts targeting these malignancies are non-existent [6], there is an ongoing need for novel and safe anti-leukemic agents.
Natural products represent a rich source of drug leads, and they have contributed greatly to anticancer drugs discovery and development [7,8]. In this regard, withanolides, naturally occurring C-28-steroidal lactones on an ergostane framework [9], have attracted considerable attention due to their diverse pharmacological properties [10], including anti-inflammatory, immunomodulatory, neurological disorders, and antitumor potential. Withaferin A (WA), the prototype of natural products with withanolide-type skeleton, was investigated for the first time in 1967 as an anticancer agent [11], and since then, a growing list of reports on WA's broad spectrum antitumor properties [12] and multifunctional molecular mechanism of action [13] have been disclosed. WA molecular biochemical targets against leukemia [14] have also been described. Thus, WA induces cell cycle arrest at G2/M phase [15]) and enhances oxidative stress triggering apoptotic cell death of human myeloid leukemia HL-60 cells [14]. WA induces apoptosis by activation of p38MAPk signaling cascade [16], through downregulation of Akt phosphorylation [17], and by acting as an agonist of liver X receptor a (LXR-α) [18] in leukemia cells.
In the search for new candidates for cancer treatment, we previously identified withanolide-type steroids from Withania aristata (Ait.) Pauq. (Solanaceae), an endemic medicinal plant from the Canary Islands [19], and designed a library of withaferin A-analogues as potent inducers in cancer cells [20][21][22][23]. Therefore, encouraged by previous works [14][15][16][17][18], the current study reports the isolation, structure elucidation and anti-leukemic evaluation of four new (1)(2)(3)(4) and twenty-five known  withanolides isolated from the acetone extract of W. aristata. Their structures were elucidated by detailed spectroscopic analyses, including 2D NMR experiments (COSY, HSQC, HMBC, and ROESY), and comparison with reported data of related compounds. The compounds were tested on HL-60 human cancer cell line, and normal Vero cells searching for selectivity. The structure-activity relationship was analysed, and in silico ADME studies were employed to understand the pharmacokinetic profile of this series of withanolides.

Chemistry
The air-dried leaves of W. aristata were powdered and extracted with acetone in a Soxhlet apparatus. The crude extract was subjected to multiple chromatographic steps on silica gel and Sephadex LH-20, including vacuum liquid chromatography, medium pressure liquid chromatography as well as preparative TLC and high performance TLC, to yield four new withanolides-type steroids (1)(2)(3)(4) with an unusual cyclic peroxide moiety, along with the known withanolides 5-29 (Figures 1 and 2). The structural elucidation was performed as follows.   Compound 1 was isolated as a colorless lacquer with a positive optical rotation [α] 20 D + 20.4. The molecular formula was established as C28H38O8 on the basis of the sodiated molecular [M + Na] + ion peak at m/z 525.2454 in its HRESIMS, suggesting ten degrees of unsaturation in the molecule. The UV spectrum exhibited a strong absorption at 216 nm, indicating the presence of an α,β-unsaturated carbonyl system, whereas the IR absorption bands revealed the presence of hydroxyl (3740 cm −1 ), carbonyl (1700 cm −1 ), and epoxide (1258 cm −1 ) groups. Its 1 H-NMR spectrum (Table 1)      In accordance with the molecular formula, the 13 C-NMR spectrum of 1 (Table 2) exhibited 28 carbon resonances, which were further classified as four methyls, four methylenes, ten methines, and ten quaternary carbons by an HSQC experiment. The downfield region of the 13 C-NMR spectrum displayed signals for an oxygenated methylene at δ C 57.4, three oxygenated methines (δ C 67.1, 78.6, and 79.2), two oxygenated tertiary carbons (δ C 84.2 and 84.9), two vinyl methines (δ C 126.5 and 141.4), and two olefinic quaternary carbons (δ C 125.1 and 154.3), together with two carbonyl carbons assigned to a δ-lactone (δ C 167.1) and a ketone (δ C 206.0). The aforementioned data suggest that 1 possesses a common tetracyclic withanolide-type steroid skeleton with a ketone, an α,β-unsaturated-δ-lactone, one disubstituted double bond, two oxygenated quaternary groups, two oxymethine groups, and one oxymethylene group (see Figure S1, Supporting Information).
The regiosubstitution of 1 was determined by an HMBC experiment (see Figure S2, Supporting Information), showing as the most relevant three-bond correlations those between the vinyl proton resonance at δ H 7.04 (H-4) and the signals at δ C 78.6 (C-2), 84.2 (C-5), 67.1 (C-6), and 48.2 (C-10), and correlation of the proton resonance at δ H 1.31 (H 3 -19) and signals at δ C 206.0 (C-1), 84.2 (C-5), 42.1 (C-9), and 48.2 (C-10), locating the ketone at C-1, a double bond at C-3/C-4, and the oxygenated groups at C-2, C-4, C-5, and C-6. Correlations observed from methyl resonances at δ H 1.06 (H 3 -21) and 0.88 (H 3 -18) to the signal at δ C 84.9 (C-17) confirmed the structure of the D ring, whereas the signal at δ H 2.05 (H 3 -28) was linked to the resonance at δ C 32.9 (C-23), 154.3 (C-24), and 125.1 (C-25), establishing the structure of the α,β-unsaturated-δ-lactone on the side chain. All these data accounted for nine out of 10 degrees of unsaturation established by its HRESIMS spectrum, indicating that compound 1 has an additional ring in the withanolide skeleton. The presence of a 1,2-dioxane system is the only option that satisfies this chemical shift requirements, which can be attributed to a six-or four-membered cyclic peroxide ring system at C-2/C-5 or C-5/C-6. This naturally occurring cycloendoperoxide would arise from a [4 + 2] or [2 + 2] cycloaddition reaction between the cyclohexadiene system ∆ 2,4 or double bond ∆ 5 on the withonolide skeleton and in situ formed singlet oxygen [24]. The regiosubstitution of the cyclic peroxide was resolved by chemical transformation. Thus, derivative 1a was prepared by standard acetylation of compound 1 with acetic anhydride. The main difference in its 1D NMR (see Figure S3, Supporting Information) was the replacement of the hydroxyl groups at C-6 and C-27 in 1 by two acetate groups in 1a, together with the downfield shift of the signals corresponding to H-6 and H 2 -27. These data suggest that 1a is the 6,27-diacetyl derivative of 1 and the six-membered cyclic peroxide ring system is located at C-2/C-5. Moreover, the structure of 1a was verified by 2D NMR spectroscopic analysis in combination with spectrometric studies.
Compound 4 showed spectral data resembling those of compound 1. Its molecular formula (C28H36O7), determined by HR-ESIMS, revealed 18 mass units less than the parent 1. A comparison of their NMR spectra showed the replacement of the signals assigned to the aliphatic H2-16 and the tertiary hydroxyl group at C-17 in compound 1 by resonances for a trisubstituted double bond [δH 5.55 (1H, s, H-16), δC 124.4 (CH-16) and 155.3 (C-17)] in 4. Even so, it should be noted that a set of 2D NMR spectra was performed for 4 with the aim of gaining the full and unequivocal assignment of the 1 H and 13 C-NMR resonances (Tables 1 and 2, and Figure S7, Supporting Information). The regiosubstitution of functional groups on the steroid skeleton was determined by an HMBC experiment and the relative stereochemistry by a ROESY experiment. This spectral evidence established the structure of withaperoxidin D (4) as 6β,27-dihydroxy-2α,5α-dioxan-1-oxo-witha-3,16,24-trien-26,22-olide.
The HRESIMS of withaperoxidin C (3) gave a molecular formula C 28 H 38 O 7 , indicating an oxygen atom less than 1. Its 1 H and 13 C-NMR data (Tables 1 and 2, and Figure S6, Supporting Information) showed that 3 has similar features to those of compound 1, with the only difference being the replacement of the tertiary hydroxyl group at C-17 (δ C 84.9, s) in 1 by a methine group [(δ H 1.16, m) and (δ C 51.9, d)] in 3, indicating that 3 is the 17-deoxy analogue of 1. Analysis of 2D NMR data allowed the complete and unambiguous assignment of chemical shifts, regiosubstitution, and relative configuration of compound 3. The HMBC experiment showed as the most relevant three-bond correlations those of the proton resonances of H 3 -18 (δ H 0.79) and H 3 -21 (δ H 1.05) with the signal of C-17 (δ C 51.9). Accordingly, the structure of withaperoxidin C (3) was defined as 6β,27-dihydroxy-2α,5α-dioxan-1-oxo-witha-3,24-dien-26,22-olide.

Antiproliferative Activity
The twenty-nine compounds (1-29) isolated from the acetone extract of W. aristata were assayed for their in vitro antiproliferative activity against the human promyelocytic leukemia cell line HL-60, and the Vero (African green monkey kidney) non-tumoral cell line used for selectivity purposes. Among the isolated withasteroids, except for withaferin A (14), previously investigated on various leukemia cell lines [14][15][16][17][18], there is only one earlier report from the 80 s on the anti-leukemic activity of the known isolated compound 22 [41]. Thus, we included both compounds because the cell lines and procedures were different from those used herein.
We assume that a selectivity index (SI) value higher than two indicates good selectivity for inducing cytotoxicity in tumor cell lines as compared to those in noncancerous cells, according to Suffness [42]. Thus, among the tested compounds, 21 of them showed selectivity to some extent (SI > 2) in the non-tumoral Vero cell line with respect to the HL-60 cell line. The overall results of the biological assays identified withasteroids 15, 16, and 22, in addition to the well-known anti-cancer lead compound, withaferin A (14), as having significantly improved activity profiles compared with the reference drug, etoposide. Moreover, their increased selectivity index makes these withasteroids suitable for further studies. The results have expanded the structural diversity of withanolides and opened up new possibilities to identify new lead compounds based on the WA skeleton.

Structure-Activity Relationship Analysis
The influence of the substitution pattern of the withanolide-type skeleton and its connection with anti-leukemic activity was examined, revealing the following trends of this series of natural withanolides: (a) In ring A, a double bond at the C-2 position was highly favorable (14, IC 50 0.2 µM versus 24, IC 50 > 40 µM). Notably, the cytotoxic effect of the 27-acetylwithanolide 25 (IC 50 1.5 µM) was 26-fold more potent than its corresponding hydroxyl derivative 24, suggesting that the enhancement of potency could be correlated to the lipophilicity of the molecule. No straightforward conclusion can be drawn from the presence of hydroxyl group or ketone group at C-4, given that their existence displayed increases (5 versus 8) and decreases (15 versus 22) in antiproliferative activity. The stereochemistry of the hydroxyl group at C-4 has a moderate effect on activity (12 resulted 2-fold more potent than the β-disposition (13) in contrast to previous findings [43]. (b) Analysis of the role of the functional group at B ring showed that withanolides with an epoxy group at C-5/C-6 displayed more potency than those with a double bond moiety (14, 18, 19, 20, and 23 versus 7, 9, 5, 6, and 13, respectively). By contrast, dihydroxylation at C-5/C-6 has a detrimental effect on the activity (14 vs. 26). In agreement with previous results [22], the cytotoxic effect of the 6-chlorowithanolide 27 was 12-fold more potent than its corresponding hydroxyl derivative 26, indicating that the enhancement of potency appears to be correlated to the lipophilicity of substituents at C-6. (c) In general, substituents on the five member D ring play a moderate role according to previous results [20]. Thus, a detrimental effect was observed in ∆ 16 -withanolides over the corresponding congeners (4, 7, 19, and 20 vs. 3, 5, 14, and 15, respectively) and those with a hydroxyl group at C-17 (3 vs . 1 and 14 vs. 23), whereas withasteroids with a hydroxyl group at C-16 and double bond at C-17/C-20 exhibited a dramatic decrease in cell proliferation (9)(10)(11). (d) Regarding the lactone ring, results indicated that the substituent at C-27 on the core skeleton plays a notable role on activity. Compounds with a hydroxyl-methylene group (14 and 19) showed better in vitro activity on the HL-60 cell line than those with a methyl group (15 and 20), suggesting that it may be involved in H-bonding with the target. Moreover, a ketal group at C-27 resulted in a significant drop of potency (17 vs. 14). (e) SAR analysis of withanolides 1-4 showed that the replacement of the typical system enone and epoxy at the A and B rings for the unusual cycloendoperoxide moiety was unfavorable, leading to a broad range of activities (IC 50 values ranging from 3.3 to 26.7 µM), depending largely on the substituent at C-17 (1 and 2 vs. 3 and 4), suggesting the enhancement of potency appeared to be correlated to the lipophilicity of the substituents on the D-ring and, to a lesser extent, on the stereochemistry of the endoperoxide (1 vs. 2). Furthermore, the structure-selectivity relationship analysis is in agreement with the anti-leukemic activity. Thus, the double bond at C-2 (14 vs. 24), the C-5/C-6 epoxy group (14,18,19, and 23 vs. 7, 9, 5, and 13, respectively) and the hydroxyl-methylene group at C-27 ( In summary, simple modifications of both essential groups, a double bond at C-2/C-3, and an epoxy group at C-5/C-6, produced a clear decrease in activity, as occurs in endoperoxides 1-4, Michael adduct 24, and withanolides 28-29 with an A-ring rearrangement. These results agree with previous studies [21,22,43], which define these functions as essential structural requirements in the withanolide skeleton for optimum antiproliferative activity. Moreover, substituents on D and lactone rings open up new challenges to modulate the antileukemic activity ( Figure 4).
Molecules 2020, 25, x FOR PEER REVIEW 9 of 16 12-fold more potent than its corresponding hydroxyl derivative 26, indicating that the enhancement of potency appears to be correlated to the lipophilicity of substituents at C-6. (c) In general, substituents on the five member D ring play a moderate role according to previous results [20]. Thus, a detrimental effect was observed in Δ 16 -withanolides over the corresponding congeners (4, 7, 19, and 20 vs. 3, 5, 14, and 15, respectively) and those with a hydroxyl group at C-17 (3 vs . 1 and 14 vs. 23), whereas withasteroids with a hydroxyl group at C-16 and double bond at C-17/C-20 exhibited a dramatic decrease in cell proliferation (9)(10)(11). (d) Regarding the lactone ring, results indicated that the substituent at C-27 on the core skeleton plays a notable role on activity. Compounds with a hydroxyl-methylene group (14 and 19) showed better in vitro activity on the HL-60 cell line than those with a methyl group (15 and 20), suggesting that it may be involved in H-bonding with the target. Moreover, a ketal group at C-27 resulted in a significant drop of potency (17 vs. 14). (e) SAR analysis of withanolides 1-4 showed that the replacement of the typical system enone and epoxy at the A and B rings for the unusual cycloendoperoxide moiety was unfavorable, leading to a broad range of activities (IC50 values ranging from 3.3 to 26.7 μM), depending largely on the substituent at C-17 (1 and 2 vs. 3 and 4), suggesting the enhancement of potency appeared to be correlated to the lipophilicity of the substituents on the D-ring and, to a lesser extent, on the stereochemistry of the endoperoxide (1 vs. 2). Furthermore, the structure-selectivity relationship analysis is in agreement with the anti-leukemic activity. Thus, the double bond at C-2 (14 vs. 24), the C-5/C-6 epoxy group (14,18,19, and 23 vs. In summary, simple modifications of both essential groups, a double bond at C-2/C-3, and an epoxy group at C-5/C-6, produced a clear decrease in activity, as occurs in endoperoxides 1-4, Michael adduct 24, and withanolides 28-29 with an A-ring rearrangement. These results agree with previous studies [21,22,43], which define these functions as essential structural requirements in the withanolide skeleton for optimum antiproliferative activity. Moreover, substituents on D and lactone rings open up new challenges to modulate the antileukemic activity ( Figure 4).

In Silico ADMET Predictions
In silico prediction of pharmacokinetic properties such as drug absorption, distribution, metabolism, excretion, and toxicity (ADMET) has become an important tool to select new lead/drug candidates and help in anticipating clinical suitability [44]. Although withanolides have been extensively investigated for their anticancer properties, there are only three reports on their ADMET studies [23,45,46]. In order to increase the success rate of compounds reaching further stages of development, the QikProp module of Schrödinger software [47] can be used as a computational method for analyzing the pharmacokinetic descriptors of the compounds (Table 4). Overview of the structural requirements of isolated withanolides based on structure-anti-leukemic activity relationship analysis.

In Silico ADMET Predictions
In silico prediction of pharmacokinetic properties such as drug absorption, distribution, metabolism, excretion, and toxicity (ADMET) has become an important tool to select new lead/drug candidates and help in anticipating clinical suitability [44]. Although withanolides have been extensively investigated for their anticancer properties, there are only three reports on their ADMET studies [23,45,46]. In order to increase the success rate of compounds reaching further stages of development, the QikProp module of Schrödinger software [47] can be used as a computational method for analyzing the pharmacokinetic descriptors of the compounds (Table 4).  The parameters analyzed provide insights into key aspects such as drug likeness, permeability, solubility, bioavailability, oral absorption, metabolism, etc. A detailed account of these parameters for compounds 1-29 is given in Table S1, Supporting Information. The first parameter that is taken into account is #stars. This identifies the number of properties of each compound that fail to remain within the recommended ranges; therefore, a lower number of #stars represents a better drug-like molecule [47]. Chemo-informatics analysis showed that all the compounds complied with this descriptor with excellent values. Thus, taking into consideration the IC 50 values of the assayed withanolides with higher cytotoxic effect than etoposide (IC 50 < 2.4 µM on HL-60 cell line), compounds 14-16, 19, 21, 22, 25, and 27 were selected to analyze their predicted pharmacokinetic properties (Table 4). Predicted values for all compounds fit the lipophilicity parameter (QP logPo/w), an important physicochemical property requirement for a potential drug, as it plays a crucial role in absorption, bioavailability, hydrophobic drug-receptor interactions, metabolism, excretion, toxicity, and in vivo pharmacological properties [48]. Distribution depends on various parameters such as the amount of binding of the drug to plasma proteins (log Khsa), the volume of body fluid that is required to dissolve it, and central nervous system permeability (QP log S). In this sense, all compounds had values in accordance with standard values. A descriptor of metabolic reactions (#metab) determines how efficiently a drug is converted to metabolites after being absorbed, distributed, and excreted, and the compounds showed ideal ranges. In this way, the predicted values of selected compounds for properties such as molecular weight, number of hydrogen-bond donor, number of hydrogen-bond acceptor, number of rotatable bonds, solubility in water, surface area, octanol/water partition, gut-blood barrier permeability, human serum albumin binding, and percentage of human oral absorption lie well within the recommended ranges. Therefore, the above data indicate that the most potent compounds, withasteroids 15, 16, and 22, in addition to WA (14), exhibited very good drug-likeness, since they obeyed Lipinski's rule of 5 and did not inflict adverse effects, as well as meeting all the pharmacokinetic criteria. Therefore, they can be considered as candidate leads.

General
Optical rotations were measured on a Perkin Elmer 241 automatic polarimeter (Waltham, MA, USA) in CHCl 3 at 20 • C, and the [α D ] are given in 10 −1 deg cm 2 g −1 . UV spectra were obtained on a JASCO V-560 spectrophotometer (JASCO-Europe, Cremelle, Italy), and IR (film) spectra were measured on a Bruker IFS 55 spectrophotometer (Bruker Co. Billerica, MA, USA). NMR experiments were performed on a Bruker Avance 400 spectrometer (Bruker Co. Billerica, MA, USA), and chemical shifts are shown in δ (ppm) with tetramethylsilane (TMS) as internal reference. EIMS and HREIMS were recorded on a Micromass Autospec spectrometer, and ESIMS and HRESIMS (positive mode) were measured on an LCT Premier XE Micromass Electrospray spectrometer (Micromass, Manchester, UK). Purification was performed using silica gel 60 µM for column chromatography (particle size 15-40 and 63-200 µm), POLYGRAM SIL G/UV 254 used for analytical and preparative TLC, and HPTLC-platten Nano-Sil 20 UV 254 were purchased from Panreac (Barcelona, Spain). Sephadex LH-20 for exclusion chromatography was obtained from Pharmacia Biotech (Pharmacia, Uppsala, Sweden). The spots in the TLC were visualized by UV light and heating silica gel plates sprayed with H 2 O-H 2 SO 4 -AcOH (1:4:20). All solvents used were analytical grade from Panreac (Barcelona, Spain). Reagents were purchased from Sigma Aldrich (St Louis, MO, USA) and used without further purification.

Cell Viability
Viable cells were assessed using the colorimetric MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide reduction assay [49]. A cell suspension (1 × 10 4 /200 µL/well) in log-phase of growth seeded in DMEM, supplemented with 5% fetal bovine serum in microtiter well plates (96 wells, Iwaki, London, UK) were added to compounds predissolved in dimethylsulphoxide (DMSO) at different concentrations. After 48 h of incubation, 20 µL/well of the MTT solution (5 mg/mL in phosphate buffered saline, PBS) were added and plates incubated for 4 h at 37 • C. Subsequently, the medium/MTT solutions were removed carefully by aspiration and replaced with 150 µL/well of DMSO to dissolve the formazan crystals. Absorbance was measured in a microplate reader (Infinite M200, Tecan) at a wavelength of 550 nm. Solvent control cultures were considered 100% and percentage of viable treated cells were plotted against compound concentrations. The 50% cell viability value (IC 50 ) was calculated from the curve. Each experiment was performed at least three times in triplicates. Data are given in arithmetic means ± SD. Selectivity ratio was defined as IC 50 value for the tumoral cell line divided by IC 50 value for the Vero cell line.

Plant Material
Leaves of W. aristata were collected in Icod de los Vinos, Tenerife, Canary Islands (Spain), in May 2005. A voucher specimen (TFC 48.068) is deposited in the Herbarium of the Department of Botany, University of La Laguna, Tenerife, and identified by Leticia Rodríguez-Navarro.

Extraction and Isolation
The air-dried powdered leaves of W. aristata (1.8 kg) were exhaustively extracted with acetone in a Soxhlet apparatus, and the solvent was evaporated at reduced pressure. The residue (80 g) was fractioned by vacuum-liquid chromatography on silica gel and eluted with hexanes/EtOAc mixtures of increasing polarity (from 100:0 to 0:100) affording thirty-four fractions, which were combined on the basis of their TLC profiles in six fractions (F1 to F6). Preliminary 1 H-NMR analysis revealed that fractions F3 to F5 were rich in withanolides and were further investigated. Each of these fractions was subjected to column chromatography over Sephadex LH-20 (hexanes/CHCl 3 /MeOH, 2:1:1) and silica gel (CH 2 Cl 2 /acetone of increasing polarity). Preparative and high performance thin-layer chromatography developed with CH 2 Cl 2 /acetone (8.5:1.5) and CH 2 Cl 2 /EtOAc (8:2) were used to purify the new compounds 1 (3.9 mg), 2 (3.

Conclusions
In the current study, and as a continuation of our efforts to find new drug candidates for leukemia treatment, four new withanolide-type steroids, named withaperoxidins A-D, with an unusual cyclic peroxide moiety, and twenty-five known compounds were identified from W. aristata. Cyclic-endoporoxide natural products are very rare, and this is the first report of withasteroid-endoperoxides from Withania genus, suggesting that the enzymatic system in W. aristata is quite different to other Withania species. Biological evaluation and in silico ADME exploration revealed that four compounds from this series are promising anti-leukemic agents, showing higher effectiveness than the known chemotherapeutic drug etoposide. This potency is also coupled with a good selectivity index on non-tumoral cells and drug-likeness profile. SAR analysis provided new, valuable information on the pharmacophore for withanolide-type compounds, suggesting that chemical optimization of B, D, and lactone rings will be helpful in the rational design of anti-leukemic drugs. Therefore, the compounds identified deserve further studies to unravel their potential as a therapeutic alternative against leukemia. These results reinforce the potential of Withania species as a source of lead compounds and chemical diversity. Funding: This was founded by RTI2018-094356-B-C21 Spanish MINECO project, co-funded by European Regional Development Fund (FEDER).

Conflicts of Interest:
The authors declare no conflict of interest.