ORGANIC CHEMISTRY

An enantioconvergent halogenophilicnucleophilic substitution(SN2X) reactionXin Zhang1*, Jingyun Ren1*, Siu Min Tan2, Davin Tan1,2,Richmond Lee2†, Choon-Hong Tan1†

Bimolecular nucleophilic substitution (SN2) plays a central role in organic chemistry. In theconventionally accepted mechanism, the nucleophile displaces a carbon-bound leavinggroup X, often a halogen, by attacking the carbon face opposite the C–X bond. A lesscommon variant, the halogenophilic SN2X reaction, involves initial nucleophilic attack ofthe X group from the front and as such is less sensitive to backside steric hindrance. Herein,we report an enantioconvergent substitution reaction of activated tertiary bromides bythiocarboxylates or azides that, on the basis of experimental and computational mechanisticstudies, appears to proceed via the unusual SN2X pathway.The proposed electrophilicintermediates, benzoylsulfenyl bromide and bromine azide, were independently synthesizedand shown to be effective.

Unimolecular nucleophilic substitution (SN1)and bimolecular nucleophilic substitution(SN2) are long-standing textbook reactionsin organic chemistry (Fig. 1A). Nonetheless,progress toward enantioconvergent nucleo-

philic substitutions of racemic tertiary electro-philes has beenmade only recently. SN1 reactionsgenerally require substrates that can form stabi-lized carbocation intermediates. In this regard,Jacobsen and co-workers reported the generationof nonheteroatom-stabilized carbocations for theenantioconvergent allylation of tertiary propargylacetates (Fig. 1B) (1), whereas Sun and co-workersreported oxygen-stabilized cations for the additionof indoles to racemic tertiary alkyl alcohols (2). SN2reactions, on the other hand, are more amenableto kinetic resolution, rather than enantiocon-vergent synthesis, owing to their stereospecificmechanism (3). Alternatively, enantioconvergentnucleophilic substitutions of racemic tertiaryelectrophiles can proceed through an SRN1 (uni-molecular radical-nucleophilic substitution) reac-tion initiated by single-electron transfer (4, 5). Theadvantage of the SRN1 mechanism is its insen-sitivity to steric influences: Fu and co-workerssuccessfully demonstrated the enantioconver-gent photoinduced coupling of racemic tertiaryalkyl chlorides with amines by using a coppercatalyst (Fig. 1B) (6).In the conventionally accepted SN2 mecha-

nism, the nucleophile displaces the leaving groupX, which is typically a halogen, by attacking frombehind the C–Xbond. Another substitution path-way that has only rarely been reported is the

halogenophilic SN2X mechanism wherein thenucleophile approaches X from the front (7–9).This mechanism has been posited for reac-tions such as the addition of thiol anions too-iodonitrobenzene (generating o-nitrophenylthioethers and nitrobenzene), as well as nucleo-philic displacement of halogens on 1-halo-1-alkynes (10, 11). In such SN2X reactions, thehalogen atom (X) of the electrophile interactswith the nucleophile (Nu) to generate a car-banion and a new electrophilic intermediate(Nu–X). The carbanion then displaces X fromthe Nu–X species to generate the desired sub-stitution product. Such reactions usually occurin cases where nucleophilic substitution at thecarbon atom is hampered. For sp3-hybridizedcarbon centers, bulky substituents may promotehalogenophilic attack. As such, they are ap-pealing targets for asymmetric constructionof sterically congested tertiary or quaternarystereocenters.Our interest in SN2X pathways follows from

our previous studies of the role of halogenbonding in catalytic reactions (12, 13). Herein,we report that two different nucleophiles, thio-carboxylates and azides, can formally displacebromide at tertiary stereocenters activated bytwo electron-withdrawing groups to generatethe substitution products with high yields andhigh enantioselectivities (Fig. 1C). A chiral cat-ionic pentanidium catalyst that our group hasdeveloped induces asymmetry (14, 15). Bothexperimental and computational mechanisticstudies support an SN2X mechanism ratherthan SN2 or radical-based SRN1 reactions.The activated tertiary bromide 1a was chosen

as a model substrate after extensive screeningexperiments of the enantioconvergent thiocar-boxylate substitution reaction (see table S1 forfull optimization details). A variety of thiocar-boxylate salts were tested in the presence of5mole% (mol%) of pentanidium salt as catalyst

(Fig. 2A). Potassium thiophenyl thiocarbox-ylate 2b and pentanidium PN4, bearing 3,5-bis(trifluoromethyl)benzyl groups, were foundto produce the best results. Under these optimizedconditions, the substrate scope of the reactionwas then examined (Fig. 2B). Phenyl, thiophenyl,and furyl thiocarboxylate salts reacted with highyields and high enantiomeric excess (ee) (3ato 3c). 2-Alkyl–substituted methyl 2-bromo-2-cyanoacetates with primary and secondaryalkyl groups afforded their respective products ingood yields and ee (3d to 3h). The reaction wasalso effective for substrates bearing allylic (3ito 3k) and propargylic (3l) substituents. Forbenzyl-substituted substrates, both electron-withdrawing and electron-donating aryl groupswere tolerated (3m to3q).Moreover, both benzyl-and naphthyl-substituted diethyl bromocyanome-thylphosphonates reactedwith good yields and ee(5a to 5h). Substrates bearing 2-(methyl)furanand 2-(methyl)thiophene groups also producedfavorable results, albeit with lower enantiose-lectivities (5i and 5j).The successful preparation of highly enantio-

enriched tertiary thioesters with our catalystprompted us to explore the use of other nucleo-philes. Organic azides act as valuable buildingblocks in synthetic chemistry, but the prepara-tion of chiral tertiary azides remains nontrivial.Two examples of the formation of chiral tertiaryazides were reported via SN2 substitution, butthese reactions required the use of enantiopuretertiary halide precursors (16, 17). After extensiveinvestigations, we discovered that 3 mol % ofPN5 and 2.0 equivalents of NaN3, in the pres-ence of a two-phase mixture of diisopropyl ether(iPr2O, 1.0 ml) and saturated K2CO3 aqueoussolution (0.2 ml), were found to be essential forthe enantioconvergent transformation of bulkyactivated tertiary bromide 6a to tertiary azide 7asmoothly over 3 to 4 days at low temperature(−40°C)withgoodyield (86%) andee (94%) (Fig. 2C;see table S2 for full optimization details). Thereason for this slow reactivity could be at-tributed to low solubility of NaN3 in the two-phase mixed solvent system at very low reactiontemperature. 2-Alkyl substituted tert-butyl2-bromo-2-cyanoacetates, such as methyl- andphenylethyl-substituted bromides, were suit-able substrates and provided good results (7band 7c). For tertiary bromides with bulky sub-stituents, slightly lower yields were obtained,owing to the formation of protonated sideproducts (7d to 7g). Substrates containingalkenyl and alkynyl substituents also affordedtheir respective products (7h to 7j and 7m) withgood yields and ee. When cyclopropyl- andpentenyl-substituted substrates were used asradical probes, substitution products wereobtained in desirable yields and ee withoutformation of any ring-opening or ring-closingproducts (7e and 7i, respectively). This outcomeindicated that free radical species were not likelyto be involved in the reactions. Substrates bear-ing t-butyldimethylsilyl– or acetyl-protected alco-hols showed no negative effects on the yields andenantioselectivities (7k and 7l); however, free

RESEARCH

Zhang et al., Science 363, 400–404 (2019) 25 January 2019 1 of 5

1Division of Chemistry and Biological Chemistry, School ofPhysical and Mathematical Sciences, Nanyang TechnologicalUniversity, 21 Nanyang Link, 637371 Singapore. 2Scienceand Mathematics Cluster, Singapore University of Technologyand Design, 8 Somapah Road, 487372 Singapore.*These authors contributed equally to this work.†Corresponding author. Email: richmond_lee@sutd.edu.sg (R.L.);choonhong@ntu.edu.sg (C.-H.T.)

on February 12, 2020

http://science.sciencem

ag.org/D

ownloaded from

alcohol groupswere notwell tolerated because ofthe formation of a tetrahydrofuran through intra-molecular cyclization.We initially considered SRN1 as a possible

mechanistic pathway for these reactions andwere hence intrigued when cyclopropyl- andpentenyl-substitutions on the alkyl cyanoesterswere unperturbed by the azidation reaction (Fig.2C; 7e and 7i). In addition, neither the radicaltrap TEMPO [(2,2,6,6-tetramethylpiperidin-1-yl)oxyl] nor the redox trap m-dinitrobenzene sub-stantially affected the substitution reaction (fig. S1).A systematic investigation was then carried out byadding proton donors (MeOH, H2O, and PhOH)or hydrogen-atom donors (1,4-cyclohexadiene,fluorene, xanthene, and Et3SiH) to the reactionmedium containing tertiary bromide 1a andpotassium thiophenyl thiocarboxylate 2b (Fig. 3A).Whereas hydrogen-atom donors did not affectthe reaction, proton donors facilitated forma-tion of the protonated product 1a-H in goodyield. We speculated that if 1a-H was generatedvia protonation of a carbanion intermediate,we would be able to experimentally observe sub-

stituent effects of this carbanion in a Hammettstudy. Indeed, a linear correlation with a po-sitive slope (r = +2.57) was obtained througha Hammett plot (Fig. 3B). This result impli-cated accumulation of negative charge in therate-determining transition state, which wasstabilized by electron-withdrawing groups. Acarbanion-exchange experiment was also con-ducted to provide experimental evidence forthe proposed carbanion intermediate. Whena mixture of tertiary bromide 1l (bearing a4-methylphenyl group) and methyl 2-cyano-3-(4-fluorophenyl)propanoate 1m-H was subjectedto the standard reaction conditions, a mixtureof tertiary thioester 3n (53% yield, 86% ee) andtertiary thioester 3o (45% yield, 77% ee) (Fig. 3C)was obtained. This result suggested that the initialcarbanion generated from 1l had abstracted aproton from 1m-H to generate the correspondingcarbanion. Both carbanions reacted with thesulfenyl bromide to generate the mixture of ter-tiary thioesters 3n and 3o.At this point, our experimental observations

indicated an SN2X pathway for these reactions.

The key feature of SN2X reactions is the genera-tion of a new electrophilic intermediate (Nu–X)from the attack of the nucleophile (Nu) on thehalogen atom (X). To support this pathway,benzoylsulfenyl bromide 8a (18), the proposednew electrophilic intermediate, was preparedand treated with carbanion 1a-A, derived from1a-H, in the presence of pentanidium PN4(Fig. 3D). This experiment afforded a mixture ofthe tertiary bromide 1a (12% yield, 0% ee) andtertiary thioester 3a (66% yield, 82% ee; comparewith Fig. 2B, 3a). Similarly, bromine azide (BrN3),the proposed new electrophilic intermediate inthe azidation reaction, was prepared via a knownprocedure with sodium azide and bromine (19).Reaction of BrN3 with carbanion 6a-A, derivedfrom 6a-H, afforded a mixture of tertiary bro-mide 6a (32% yield, 0% ee) and tertiary azide7a (36% yield, 78% ee). These experimentsindicate that both benzoylsulfenyl bromide 8aand BrN3 are plausible intermediates in thesereactions. Although the tertiary thioesters ortertiary azides were obtained in high ee, thetertiary bromides were obtained as racemates.

Zhang et al., Science 363, 400–404 (2019) 25 January 2019 2 of 5

Fig. 1. Nucleophilic substi-tutions. (A) Nucleophilicsubstitutions: SN1, SN2, andSRN1. SET, single-electrontransfer. (B) State-of-the-artenantioconvergent nucleo-philic substitution of tertiaryelectrophiles. TMS, trime-thylsilyl; Me, methyl; Et,ethyl; Ph, phenyl; Ar, aryl;tBu, tert-butyl; OTf, triflate;h, Planck’s constant; n, fre-quency; LED, light-emittingdiode. (C) This work:Pentanidium-catalyzedenantioconvergenthalogenophilic nucleophilicsubstitution (SN2X).

A

B

C

RESEARCH | REPORTon F

ebruary 12, 2020

http://science.sciencemag.org/

Dow

nloaded from

This suggests that stereoinduction does notoccur through the halogen abstraction stepinvolving the C–Br bond. Next, we separatedthe two enantiomers of tertiary bromide 1i byusing preparative high-performance liquid chro-matography (HPLC) and subjected them to

enantioconvergent thiocarboxylate substitutionseparately (Fig. 3E). We found that both en-antiomers were transformed to the same en-antiomer of thioester, (+)-3k (87% ee), and therecovered 1i was racemized. From these results,we propose that the sulfenyl bromide and BrN3

are generated through the SN2X mechanismand are ambident electrophiles. Positing thatthe C–Br bond cleavage step is reversible (Fig.3F) explains the racemization of enantioenrichedtertiary bromide (Fig. 3E). This mechanismalso explains the formation of homocoupleddisulfide side products observed in the thio-carboxylate substitution reaction (10). Sim-ilar mechanistic studies were conducted forthe azidation reaction and support the SN2Xmechanism (fig. S2).Density functional theory (DFT) calculations

were conducted to provide more mechanisticinsights and support the experimental studies(vide supra). The modeling involved a simplifiedpentanidium catalyst, in which phenyl andbenzyl groups were truncated into either H ormethyl groups, to reduce computational costs.Two possible pathways to achieve the thioester3d, via SN2 or halogenophilic SN2Xmechanisms,were modeled (Fig. 4). Our studies revealed thatthe SN2 pathway involving rear addition of thethiocarboxylate to the tertiary carbon of 1b viaTS-A, required overcoming a relatively highenergy barrier of 27.1 kcal/mol. This conclusionis unsurprising, as it is well known that SN2reactions are sterically dependent. For the SN2Xpathway, the tertiary bromide 1b is held in placeby S–Br intermolecular halogen bonding withthe thiocarboxylate, forming int-B (solutionGibbs free energyDGsol = 11.2 kcal/mol) (20). TheBr possesses a s hole that is enhanced by theproximity of two covalently bonded electron-withdrawing functional groups (21), namely thecyanide and ester moieties. This intermediateis thus primed for C–Br bond cleavage due tohalogen bonding. The calculated S···Br atomicdistance in int-B is 3.04 Å, which is well withinthe reported range for halogen bonds and theC–Br bond of tertiary bromide 1b is slightlyelongated to 2.00 Å (Fig. 4). Indeed, mappingof noncovalent interaction surfaces (22, 23) forint-B (fig. S9) showed strong positive interactionbetween the s hole of Br and S of the thio-carboxylate. Formation of such intermolecularhalogen bonds has been widely established as astrategy to prepare cocrystals in the field ofcrystal engineering (24), and it has been recentlyadopted for the rational design of reactions andcatalysts (25).From int-B, the most kinetically feasible

outcome would be the stepwise Br abstrac-tion by the thiocarboxylate via TS-B (DGǂ

sol =20.9 kcal/mol) (26–28). The process throughTS-B resulted in the simultaneous formationof the S–Br bond and breakage of the Br–Cbond, leading to the sulfenyl bromide/enolatecomplex int-C (DGsol = 17.0 kcal/mol), which isheld together by halogen bonding (Br···O atomicdistance of 2.55 Å) (Fig. 4). The dissociation ofthe sulfenyl bromide intermediate from the cat-alyst, forming a sulfenyl bromide and the catalyst–enolate ion pair separately, is thermodynamicallyless endergonic (DGsol = 7.3 kcal/mol). The ex-periments, vide supra, revealed that theC–Br bondscission-reformation equilibrium is not enantio-determining, which suggests that the catalytic

Zhang et al., Science 363, 400–404 (2019) 25 January 2019 3 of 5

A

B

C

Fig. 2. Pentanidium-catalyzed enantioconvergent halogenophilic nucleophilic substitution(SN2X). (A) Optimization of reaction condition. (B) Enantioconvergent thiocarboxylate substitutionof tertiary bromides. (C) Enantioconvergent azidation of tertiary bromides. Isolated yields arereported, and ee values were determined using chiral HPLC or gas chromatography. See thesupplementary materials for detailed reaction conditions.

RESEARCH | REPORTon F

ebruary 12, 2020

http://science.sciencemag.org/

Dow

nloaded from

Zhang et al., Science 363, 400–404 (2019) 25 January 2019 4 of 5

Fig. 3. Mechanisticstudies. (A) Effects ofhydrogen-atom donorsand proton donors.(B) Hammett plot oflog(kAr/kPh) for theformation of 3m to3q versus thecorresponding s value(k, reaction rate).(C) Carbanion-exchange experiments.(D) Reactions usingthe proposed SN2Xintermediates benzoyl-sulfenyl bromide 8a andBrN3. (E) Reactionswith enantioenrichedtertiary bromides.(F) Proposed SN2Xpathway and sidereactions.

y = 2.57 x + 0.13 R² = 0.97

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

-0.3 -0.2 -0.1 0 0.1 0.2 0.3

log(

kA

r/kP

h)

Substitution Constant σ

OMe

Me

H

F

Cl

A

C

D

E

B

F

Fig. 4. DFT calculations. DFT wasused to calculate relative freeenergies of the intermediates andtransition states of the SN2and SN2X mechanistic pathways.Images of the geometricallyoptimized structures int-B andint-C, depicting key atomic bonddistances, are shown with H atomsomitted for clarity.

RESEARCH | REPORTon F

ebruary 12, 2020

http://science.sciencemag.org/

Dow

nloaded from

stereocontrol in orientation and binding of sub-strates is important for int-D (DGsol = 16.0 kcal/mol) to give rise to enantioenriched product.Subsequently, C–S bond formation via TS-D(DGǂ

sol = 21.3 kcal/mol) leads to the formationof preproduct complex int-E (DGsol = 0.3 kcal/mol) and then product complex cat-pdt (DGsol =−8.3 kcal/mol). The activation barriers TS-B andTS-D relative to cat-enolate/sulfenyl bromideare nearly isoenergetic, consistent with exper-imental evidence, suggesting that sulfenyl bromideis an ambident electrophile.Whereas halogen bonding has been well ex-

ploited in the field of supramolecular chemistryand crystal engineering, its role in reaction de-velopment and catalysis is still in its infancy. Theresults presented herein open the door for fur-ther exploration along those lines.

REFERENCES AND NOTES

1. A. E. Wendlandt, P. Vangal, E. N. Jacobsen, Nature 556,447–451 (2018).

2. W. Zhao, Z. Wang, B. Chu, J. Sun, Angew. Chem. Int. Ed. 54,1910–1913 (2015).

3. S. V. Pronin, C. A. Reiher, R. A. Shenvi, Nature 501, 195–199(2013).

4. R. A. Rossi, A. B. Pierini, A. B. Peñéñory, Chem. Rev. 103,71–168 (2003).

5. G. C. Fu, ACS Cent. Sci. 3, 692–700 (2017).6. Q. M. Kainz et al., Science 351, 681–684 (2016).7. N. S. Zefirov, D. I. Makhon’kov, Chem. Rev. 82, 615–624 (1982).

8. P. K. Sazonov, G. A. Artamkina, I. P. Beletskaya, Russ. Chem.Rev. 81, 317–335 (2012).

9. J. F. Bunnett, Acc. Chem. Res. 5, 139–147 (1972).10. S. Montanari, C. Paradisi, G. Scorrano, J. Org. Chem. 58,

5628–5631 (1993).11. M. C. Verploegh, L. Donk, H. J. T. Bos, W. Drenth, Recl. Trav.

Chim. Pays Bas 90, 765–778 (1971).12. L. Zong, X. Ban, C. W. Kee, C.-H. Tan, Angew. Chem. Int. Ed. 53,

11849–11853 (2014).13. W. He, Y.-C. Ge, C.-H. Tan, Org. Lett. 16, 3244–3247 (2014).14. L. Zong, C.-H. Tan, Acc. Chem. Res. 50, 842–856 (2017).15. T. Ma et al., J. Am. Chem. Soc. 133, 2828–2831 (2011).16. K. Shibatomi, Y. Soga, A. Narayama, I. Fujisawa, S. Iwasa,

J. Am. Chem. Soc. 134, 9836–9839 (2012).17. D. Kim et al., J. Org. Chem. 82, 4936–4943 (2017).18. S. Kato, K. Miyagawa, S. Kawabata, M. Ishida, Synthesis 1982,

1013–1014 (1982).19. A. Hassner, F. Boerwinkle, A. B. Lavy, J. Am. Chem. Soc. 92,

4879–4883 (1970).20. G. Cavallo et al., Chem. Rev. 116, 2478–2601 (2016).21. T. Clark, M. Hennemann, J. S. Murray, P. Politzer, J. Mol. Model.

13, 291–296 (2007).22. J. Contreras-García et al., J. Chem. Theory Comput. 7, 625–632

(2011).23. E. R. Johnson et al., J. Am. Chem. Soc. 132, 6498–6506

(2010).24. A. Mukherjee, S. Tothadi, G. R. Desiraju, Acc. Chem. Res. 47,

2514–2524 (2014).25. D. Bulfield, S. M. Huber, Chemistry 22, 14434–14450 (2016).26. Transition states TS-B and TS-D are flat saddle points;

i.e., points with small values of the modulus of imaginaryfrequencies. Because intrinsic reaction coordinatecalculations were not successful, we further carried outrelaxed potential scans perturbing the transition statesalong the direction of the vibrational modes to confirm thatthe transition states are connected to the minimumstructures. See (27).

27. M. W. Lodewyk, D. Willenbring, D. J. Tantillo, Org. Biomol.Chem. 12, 887–894 (2014).

28. M. N. Glukhovtsev, A. Pross, H. B. Schlegel, R. D. Bach,L. Radom, J. Am. Chem. Soc. 118, 11258–11264 (1996).

ACKNOWLEDGMENTS

Funding: We acknowledge support from Nanyang TechnologicalUniversity (M4011663 and M4080946); Ministry of Education,Singapore (MOE2016-T2-1-087); and Singapore Universityof Technology and Design (T1MOE1706 and IDG31800104)as well as computational resources from NationalSupercomputing Centre (Singapore). Author contributions:C.-H.T. conceived of the project; C.-H.T., X.Z., and J.R. designedthe research; X.Z. carried out the thiocarboxylate substitution;J.R. performed the azidation reaction; R.L. designed thecomputational studies; and D.T. and S.M.T. conducted theDFT calculations. All authors took part in writing and reviewingthe manuscript. Competing interests: The authors declareno competing interests. Data and materials availability:Crystallographic data are available free of charge fromthe Cambridge Crystallographic Data Centre under referencenumbers CCDC 1845290, 1845291, 1845292, and 1851709.All other data are available in the main text or thesupplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/363/6425/400/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S12Tables S1 to S3Spectral DataReferences (29–35)

14 July 2018; resubmitted 14 October 2018Accepted 11 December 201810.1126/science.aau7797

Zhang et al., Science 363, 400–404 (2019) 25 January 2019 5 of 5

RESEARCH | REPORTon F

ebruary 12, 2020

http://science.sciencemag.org/

Dow

nloaded from

2X) reactionNAn enantioconvergent halogenophilic nucleophilic substitution (SXin Zhang, Jingyun Ren, Siu Min Tan, Davin Tan, Richmond Lee and Choon-Hong Tan

DOI: 10.1126/science.aau7797 (6425), 400-404.363Science 

, this issue p. 400Scienceback to form a carbon-sulfur or carbon-nitrogen bond enantioselectively.carbon centers activated by electron-withdrawing groups. A chiral cationic catalyst then directed the carbon fragmentattack on the halogen from the front. Specifically, nitrogen and sulfur nucleophiles stripped bromine from a variety of

now present an asymmetric catalytic substitution reaction that flips the script with anet al.opposite direction. Zhang theincoming reactive group will attack the carbon from behind its bond with the halogen, causing the halogen to depart in

Biomolecular substitution reactions are widely applied to compounds with carbon-halogen bonds. Typically, anAttack from the front

ARTICLE TOOLS http://science.sciencemag.org/content/363/6425/400

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/01/23/363.6425.400.DC1

REFERENCES

http://science.sciencemag.org/content/363/6425/400#BIBLThis article cites 35 articles, 1 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Science. No claim to original U.S. Government WorksCopyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of

on February 12, 2020

http://science.sciencem

ag.org/D

ownloaded from