Finding the Ajoene Sweet‐Spot: Structure‐Activity Relations that Govern its Blood Stability and Cancer Cytotoxicity

Ajoene is an organosulfur compound found in crushed garlic that exerts its anti‐cancer activity by S‐thiolating cysteine residues on proteins. Its development is hampered due to limited bioavailability, so in this study, we synthesised analogues of ajoene to probe the significance of the ajoene vinyl disulfide/sulfoxide core with respect to cytotoxicity and blood stability. Polar side groups were also incorporated to improve aqueous solubility. It was found that derivatives containing a vinyl disulfide functional group (4–7, as in ajoene), were more cytotoxic compared to analogues in which the double bond was removed, although the latter showed superior blood stability. It was also found that the allyl‐S sulfur of the disulfide was more electrophilic to S‐thiolysis based on the global electrophilicity index (ω) and the condensed electrophilic Fukui function fk+ ${{ f}_{\rm{k}}^{\rm{ + }} }$ . S‐Thiolysis was found to be exergonic for the vinyl disulfides based on entropy and enthalpy computations with a deprotonated thiolate. Derivatisation to the dihydro (10, 12) and deoxydihydroajoenes (9, 11) produced analogues that were slightly less potent but with greatly improved blood stability. Taken together, the deoxydihydroajoenes present themselves as good candidates for further therapeutic development.


Introduction
Garlic is one of the oldest recorded medicines, with its health benefits being recognised since ancient times. [1]Garlic preparations have been commonly used to promote cardiovascular health, to stimulate the immune system, and to fight infections.Garlic is rich in low molecular weight organosulfur compounds, generated as a chemical defence against invasive threats when the clove is damaged.The biological chemistry behind garlic involves the coming together of the enzyme alliinase and its substrate alliin to form allicin (see Figure 1), a compound that was first isolated and identified by Cavallito and Bailey in 1944 through chemical synthesis as diallyl thiosulfinate from the oxidation of diallyl disulfide. [2]Two molecules of allicin are readily able to self-react to form a number of secondary thioallyl compounds which emerge in cooked or aged garlic preparations through reaction pathways that were proposed by Block in the 1980's. [3]Such compounds include the dialkyl polysulfanes, the dithiins and the ajoenes, all as oil-soluble constituents, as well as S-allylmercaptocysteine (SAMC) and S-allyl cysteine (SAC) as the water-soluble constituents (Figure 1).Allicin and its secondary rearrangement products are toxic to pathogens and to cancer cells. [4]he compound of interest in this study, ajoene, acts by Sallylating cysteine residues on proteins via thiolysis exchange [5] involving the ajoene vinyl disulfide.This reaction, akin to a post-translational modification, leads to a covalently modified protein that triggers downstream effects such as enzyme inhibition, altered protein function and cell signalling.5c] A correlation was found between cancer-cell cytotoxicity and the conjugate acid pK a of the leaving group, supporting thiolysis being driven by the stability of the leaving group.Thiolysis was also shown to be central to the cytotoxic mechanism of garlic organosulfur compounds (OSCs), [5b] with the most important members in this regard being allicin and its secondary rearrangement product ajoene.
There is evidence that allicin and ajoene undergo S-thiolysis prior to reaching the cancer cell. [6]For example, Lawson et al. failed to detect allicin in human blood, urine or stool even after consuming large amounts of garlic (25 g) or pure allicin (60 mg). [7]In a follow-up experiment, it was found that the addition of allicin to fresh blood results in a quantitative and rapid (half life < 1 min) production of allyl mercaptan. [8]Allicin also shows a very rapid first pass through the rat liver, with only its metabolites diallyl disulfide (DADS) and allyl mercaptan being detected in the liver perfusate. [9]The presence of allyl mercaptan is only transient, however, reaching maximal levels within 1-2 min and disappearing by 1 h [10] due to its further metabolism to allyl methyl sulfide (AMS). [11]AMS has been found to reach maximal levels at 4 h and to persist for over 30 h, indicating that it is the product of systemic metabolism. [12]he rate of maximum AMS production is of the order allicin > diallyl trisulfide (DATS), ajoene > DADS, which mirrors the trend observed for in vitro blood stability with a half life of < 1, 2-4 and 60 minutes respectively. [8]The blood instability therefore appears to be related to thiolysis, with the strength of the SÀ S bond being weaker for thiosulfinates and trisulfides than for disulfides. [13]The vinyl disulfide in ajoene can also be viewed as thiolysis reactive due to resonance stabilisation of the enethiolate leaving group. [14]S-Thiolysis regioselectivity in ajoene proceeds via the allyl-S atom (the sulfur at the nonsulfoxide end) of the vinyl disulfide, implying that it is the more electrophilic of the two.5a] If this capture takes place within the erythrocyte, it may contribute to their blood instability and poor bioavailability.For example, Freeman et al. found that allicin is rapidly metabolised by erythrocytes, [15] with no reaction occurring in serum. [8]Miron et al. proposed that this instability may be due to reaction with glutathione as evidenced by the very rapid (1 min) and abundant detection of S-allylmercaptoglutathione (GSSA) following in vitro erythrocyte incubation of allicin. [16]It is well known that a number of animal species are susceptible to haemolytic anaemia following ingestion of aliphatic sulfides from Allium vegetables. [17]In these animals, the ingested OSCs cause the depletion of GSH and the generation of hydrogen peroxide within the erythrocyte, which results in the oxidative denaturation of haemoglobin. [18]Upon examining species susceptible to OSC toxicity (i.e cattle, horses, sheep, cats, dogs, chickens, rabbits and rats), the formation of granules, termed Heinz bodies, could be observed within their erythrocytes.
In humans, garlic OSCs have shown promising anticancer activity, albeit with different OSC chemotypes varying their thiolysis reactivity and biological potency.While a more reactive thiophile may be more efficient at protein S-thiolation, it may also be less discriminate towards its targets, with the enhanced reactivity likely leading to a lowered metabolic stability and reduced bioavailability.In this study, we aimed to investigate the structure-activity and reactivity relationships in ajoene that underpin both its potency towards cancer cells and its poor bioavailability (blood instability), with the aim of finding whether a relationship exists between these effects and whether the "goldilocks sweet spot" within the ajoene chemotype may be identified.

Synthesis of Ajoene Analogues
For SAR purposes, a library of ajoene derivatives was synthesised varying the two key functionalities at the monosulfur (as sulfide or sulfoxide) and the vinyl disulfide (as is or as a saturated disulfide termed a dihydroajoene) to create ajoene/ deoxyajoene and dihydroajoene/deoxydihydroajoene variants respectively, see Figure 2. At the sulfoxide/vinyl disulfide end, the carbon group was retained as p-methoxybenzyl (PMB) since this had shown optimal activity for ajoene in a previous study, [5a] while at the other end, a 2-(p-hydroxyphenyl)ethyl group was attached with the hydroxyl group left free or functionalised into a carbamoyl derivative for increasing aqueous solubility.
Figure 2. Design of the ajoene library.The ajoene library varied the functionalities of the monosulfide (as sulfide or sulfoxide) and the vinyl disufide (as is or saturated) to create ajoene/deoxyajoene and dhydroajoene/deoxydihydroajoene variants, respectively.In addition, solubility was modulated through a parasubstituent of the phenyl ring at the vinyl disulfide end to produce a phenol and amide series in both of the above functional group variants.

Synthesis of the Vinyl Disulfides: Ajoenes and Deoxyajoenes 4-7
5a,19] Synthesis of 4 is shown in Scheme 1 involving first regioselective addition of thioacetic acid to the alkyne terminus of propargyl sulfide 1, the latter easily prepared from p-methoxybenzyl chloride using thiourea substitution followed by base hydrolysis to the thiol and in situ propargylation with propargyl bromide.The 1 H NMR spectrum of the radical addition product, vinyl thioacetate 2, as expected from previous additions, revealed a 2 : 1 Z : E mixture of vinyl sulfide stereoisomers from doubling of signals in the vinyl region of the spectrum (5.7-5.9 and 6.45-6.7 ppm for the β-and α-hydrogens respectively).Thereafter, 2 was committed to a low temperature base hydrolysis of the vinyl acetate to form the enethiolate, which was S-sulfenylated by thiosulfonate, 3 (with its phenolic hydroxyl group unprotected), to form vinyl disulfide, 4, in 72 % yield as a 5 : 2 Z : E mixture of stereoisomers as the first of the library members.The thiotosylate reagent, 3, could be prepared in three high-yielding steps (> 95 % yield for each) from commercially available methyl (4hydroxyphenyl)acetate via ester reduction (LiAlH 4 ), [20] Appel hydroxyl group iodination (imidazole, I 2 , and PPh 3 ) [21] and finally substitution with sodium p-toluenethiosulfonate salt.The pivotal vinyl disulfide, 4, was thoroughly characterised by 1 H and 13 C NMR spectroscopy, IR spectroscopy and HRMS, returning favourable data throughout, Scheme 1.
Vinyl disulfide 4 could then be used to access the other ajoene and deoxyajoene library members 5-7 via a series of conventional functional group conversions involving both sulfide and the phenolic hydroxyl group.The various steps are shown in Scheme 2 involving chemoselective sulfide monooxidation using m-CPBA and phenoxide protection with iodoacetamide (formed in situ from chloroacetamide and TBAI).Yields were generally low (40-60 %), and it was better to oxidise 4, and then alkylate rather than the other way round for accessing ajoene derivative 6.All compounds were new and fully characterised (see the SI).The chemoselective oxidation of the ajoene vinyl disulfide-sulfide motif to the sulfoxide is interesting and was first demonstrated by Block et al. in his seminal 1986 JACS paper [22] in which he used allyl o-(allylthiomethyl)phenyl disulfide containing a benzene ring as a model for the ajoene double bond.Block showed that overoxidation of ajoene is only achieved aggressively via KMnO 4 (to the sulfone retaining the vinyl disulfide) or excess (3 eq) peracetic acid for global oxidation of both sulfur functionalities.No mention was made of any other major oxidation products being formed in the relatively mild mono-oxidation step with m-CPBA, which is what we have observed in our various derivative synthesis over the years synthesising ajoene derivatives. [5,14,19]The nucleophilicity of the ajoene vinyl disulfide sulfurs towards oxidation is presumably sufficiently reduced by the one sulfur lone pair delocalising into the double bond.Moreover, in our case, the m-CPBA oxidation was carried out with only a slight excess of the reagent (1.5 eq., but the m-CPBA is only about 80-85% pure) and at low temperature (À 78 °C to rt).More recently, Wirth et al. have also used the allyl o-(allylthiomethyl)phenyl disulfide motif as part of their approach to producing ajoene derivatives via various novel synthesis strategies [23] in which the chemoselective m-CPBA oxidation was also demonstrated.

Synthesis of the Dihydroajoenes: Dihydro-and Deoxydihydroajoenes 9-12
For the dihydro-and deoxydihydroajoene series, a straightforward linear synthesis from 1,3-propanedithiol gave access to compounds 9-12 as shown in Scheme 3. Removal of the vinyl disulfide double bond meant that no stereoisomers were formed.
Hence, chemoselective mono S-protection of the dithiol with PMB was achieved by slowly adding PMBCl as limiting reagent, dropwise into a solution of the dithiol of propane-1,3dithiol in methanol with a slight excess of KOH.The reaction was swift even at 0 °C, gratifyingly producing the monoprotected thiol, 8, in 88 % yield after work-up and chromatography.The latter was subjected to a second alkylation using thiotosylate, 3 from before, to form disulfide, 9, as the first library member of this series.Here, low temperature (À 78 °C) reaction with triethylamine as base, ensured a good yield of 9 in 77 % after work-up and chromatography.Thereafter, chemoselective m-CPBA oxidation of 9 as in the first series, accessed dihydroajoene, 10, in 70 % yield, while alkylation of 9 with iodoacetamide as before except, using caesium carbonate as base (K 2 CO 3 led to decomposition) afforded the deoxdihydroajoene derivative, 11, in 77 % yield.Finally, compound 11 could be oxidised by m-CPBA to the protected dihydroajoene derivative 12 in 72 % yield.In this series, phenolic hydroxyl group alkylation preceding sulfide oxidation was found to work better than the other way round (the reverse of the case for ajoenes 5 and 6).Yields in this series were better than in the first one, and as new compounds, 9-12 were fully characterised (see the SI).Finally, sulfide oxidation to the sulfoxide (as opposed to the sulfone) in both series could be discerned by virtue of a diastereotopic AB pair for the benzylic (PMB) protons flanking the chiral sulfoxide moiety in the 1 H NMR spectra.Such coupling would have been absent in the achiral sulfone.
5a] These findings are rationalised by thiolysis driving the cytotoxicity, the former improving as the stability of the enethiolate leaving group increases.More specifically, removal of the double bond going from ajoene to the dihydroajoene series results in a loss of leaving group resonance stabilisation (enethiolate to sulfide) with reduced cytotoxicity.5b] This implies a very important role for the sulfoxide and the aromatic side groups, the former presumably imparting an enhanced leaving ability to one of the disulfide sulfur atoms through a long range inductively withdrawing effect of the sulfoxide.Applying the same principles, this trend is similarly observed for partners 4 vs 9 and 7 vs 11, with a 30-and 3-fold loss in activity respectively.
Regarding the role of the sulfoxide in the dihydroajoene series, its removal in 9 and 11 caused a 3-and 2-fold loss in activity for the phenol and amide analogues respectively, presumably due to the inductive effects already mentioned.However, the sulfoxide overall probably plays a minor role, as the presence of the sulfide alone (in the deoxyajoenes 4 and 7) resulted in analogues that are strongly active, and in some cases (4 and bisPMB deoxy-analogue [5a] ), the sulfide is the more active compound.It has been previously shown, [25] that the sulfoxide can hydrogen bond to an adjacent amino acid residue in the protein microenvironment.However, according to the present data, this aspect bears little weight to the thiolysis argument.
On in vitro cytotoxicity specifically, while ajoene has previously been shown to be cytotoxic to oesophageal cancer cells, [5a] it has also been shown to be similarly cytotoxic to other cancer cell lines including lymphoma, mammary, bladder, colorectal, hepatic, nasopharyngeal, gastric, prostrate, lung, pancreatic, lymphoma, leukaemia, and skin. [24,26]The more active phenol series (4, 5, 9 and 10) was therefore assessed against a panel of breast-cancer cell lines.The compounds were tested on SUM159 cells (basal-like subtype) at a concentration of 20 μM for 48 h, and viable cells were quantified by Trypan blue staining (Figure 3A).E-ajoene (EA) significantly inhibited the cancer cell growth relative to the vehicle-treated control.On the other hand, ajoene analogues deoxy-4, dihydro-10 and deoxydihydro-9 were all more active than EA with inhibitory activities in the range 50-70 %.Moreover, phenol ajoene 5, with the complete vinyl disulfide/sulfoxide core, was the most active against this cell line with only 6% of viable cells remaining following the same incubation period.This analogue was further evaluated on MCF-7 (luminal-type), SK-BR-3 (HER2enriched), and MCF-10 A (normal) cells.A dose-response growth inhibitory effect was observed on all the cell lines in the concentration range of 5, 10, and 20 μM (Figure 3B), although the compound was most active on the SUM159 cell line overall.Phenol ajoene 5 was also evaluated for its ability to inhibit colony formation in MCF-7 cells.At concentrations as low as 0.5-1 μM, phenol ajoene 5 exhibited a 20-30 % inhibitory effect on colony formation.These results indicate the cytotoxicity of phenol ajoene 5 is not subtype-dependent but displays the highest activity in basal-type breast cancer cells.It also reaffirms the superior cytotoxicity of the vinyl disulfide pharmacophore in ajoene over the deoxy-or dihydroajoenes.

Structure-Activity Studies of Ajoene: Blood Stability
The ajoene analogues were investigated for their in vitro stability in fresh mouse blood involving incubation at 37 °C for up to 120 min.At specified time points, an aliquot of the reaction mixture was quenched in acetonitrile and the percentage compound remaining was quantified using the peak area from an LCMS chromatogram.Table 1 gives the % remaining at the 20 min timepoint taken from the curves shown in Figures 4A and B. The parent ajoene derivatives containing the vinyl disulfide/sulfoxide core, i. e. bisPMB, phenol 5, and amide 6 were all found to have extremely poor blood stability with nothing detected at 20 min, which is in line with that reported for ajoene. [8]Fractions of the other analogues were all detected at 20 min indicating that removal of the sulfoxide and/or the vinyl group enhances blood stability.What is striking is that the deoxydihydroajoenes (both phenol 9 and amide 11) were now quite stable in blood with 77 � 3.8 % and 53 � 5.3 % remaining at this timepoint.
An inspection of the blood stability results revealed that a decrease in reactivity of the disulfide (achieved by removal of the vinyl and/or sulfoxide) results in an increase in the blood stability.Thus, a crucial trend emerged: that reducing the reactivity towards thiolysis results in an improved blood stability, an observation with implications for drug development.Indeed, a plot of IC 50 vs blood stability at 20 min shows this correlation, in which reduced cytotoxicity (i.e. higher IC 50 value) correlates well with increased blood stability.While a loss of cytotoxicity is not desirable for drug potency, the deoxydihy-droajoenes were only 3-5 fold less active than the ajoenes, and still remain more active than E-ajoene itself (Figure 3A).Interestingly, when comparing the cytotoxicity IC 50 to the blood stability at 20 min, there is an inflection point between the two series (Figure 4C, grey arrow), with the more lipophilic phenol analogues having lower IC 50 values but improved stability compared to their more hydrophilic amide counterparts.Thus, lipophilicity may favour cytotoxicity, while blood instability may be favoured for the more hydrophilic ajoenes.These opposing reactivity preferences may in turn speak to the different thiolysis targets of ajoene in blood vs cancer cells.Overall, these SAR studies on ajoene have uncovered analogues that appear to be stable in blood while retaining good cancer cell cytotoxicity.The deoxydihydroajoenes may be particularly attractive pre-clinical candidates as they are stable in blood with 80 % and 35 % remaining for 9 and 11 respectively at 120 min, and with good activity, as their IC 50 s are still in the low micromolar range.

Molecular Modelling of Ajoene Reactivity and Thiolysis Exchange
To further probe structure-activity relations of ajoene and its derivatives of importance to the cytotoxicity -blood stability interrelationship, the two libraries (phenols and amides, 8 compounds) were modelled for electrophilicity towards thiolysis exchange.The two electrophilicity parameters calculated as kinetic indicators were the global electrophilicity index (ω) [27] and the condensed electrophilic Fukui function f þ k [28] in the context of nucleophilic S-attack on the allylic sulfur atom (phenethyl-S) or the vinylic sulfur atom (vinyl-S) of the disulfide.The global electrophilicity index provides insight into the ability of the molecule to accept an electron based on its chemical potential and hardness, while the Fukui function describes the response to a variation in the number of electrons, which in its condensed (or local) form can be used to identify the most electrophilic sites.Both these definitions of electrophilicity can be reconciled with frontier molecular orbital theory and consider the initial approach of the incoming nucleophile; therefore, thermodynamic considerations are not included and the role of the transition state structure and leaving group in driving this reaction is not taken into account.The atomcondensed Fukui function f þ k was calculated using several related methods (see computational details).While the results using the orbital-weighted Fukui function (OW) f þ k [29] was chosen for discussion, all of the data and a comparison of trends are included in the SI.At most, 10 low-energy conformers of each derivative were considered in calculating the Boltzmann-averaged < ω > , whereas for the calculation of f þ k , only the lowest energy conformer was used.For the sulfoxides, the R stereoisomer was arbitrarily selected, and the ajoene vinyl disulfide-containing analogues were modelled as both E-and Zstereoisomers.
Based on ω and f þ k (Figure 5A and 5B), the Z-vinyl disulfides were found to be more electrophilic than the corresponding Evinyl disulfides.This finding agrees with the approximate 1.5fold enhancement of cytotoxicity which has been reported for Z-ajoenes over E-ajoenes against a number of cancer cell lines, [5a,19,24] supporting that electrophilicity towards thiolysis is a driver of cancer cell cytotoxicity.In addition, the active deoxyajoene and ajoene analogues respectively, 4 and 5, were the most electrophilic according to both ω and f þ k , particularly for the Z-isomers.In the case of ω, the least active analogue, deoxyajoene 7, was also the least electrophilic.The dihydroajoenes were overall less electrophilic than the vinyl disulfides, with the deoxydihydro phenol 9 being the least electrophilic.Using the Fukui function, the results were less clear cut, although the dihydroajoenes were overall less electrophilic than the Z-vinyl disulfides.Again, 4 and 5 Z-vinyl disulfides were the most electrophilic.
5a] We rationalised this observation based on the resonance stabilisation of the expelled enethiolate leaving group, which directs the regioselectivity.In this context, f þ k supports regioselectivity, with allyl-S (more specifically phenethyl-S in this case) returning a larger value than vinyl-S for the vinyl disulfides (vinyl disulfides and deoxyajoenes), although this difference is far more pronounced for the Z-compared to the E-vinyl disulfides (Figure 5B).In general, the dihydroajoenes were also less electrophilic than the vinyl disulfides, which correlates with their decreased cancer cell cytotoxicity and increased blood stability.
5b] To better understand this process, overall reaction enthalpies and free energies of a model reaction were calculated using DFT at the ωB97XÀ D/ def2-SVP level of theory [30] (see computational details).N-acetyl cysteine methyl ester as its deprotonated thiolate was used as a nucleophile to generate the cysteine disulfide and leaving group (either as an enethiolate for the vinyl disulfides or a thiolate for the dihydroajoenes) (Figure 6).Only the lowest energy conformers for all reactants and products were considered without any Boltzmann averaging as done for the electrophilicities.For the vinyl disulfides in both the ajoene and deoxyajoene series, except for the amide deoxyajoene 7, the Z-isomer reactions were more exothermic and exergonic at 298.15 K (i.e., both Δ r H and Δ r G) for the Z-vinyl disulfides than the E-vinyl disulfides in keeping with cytotoxicity trends.The dihydroajoenes on the other hand, all exhibited relatively endergonic reactions of ~40 kJ/mol.
Overall, defining electrophilicity as the ability of a molecule to accept an electron or its response to variation in the number of electrons at a particular site, the Z-vinyl disulfides are more electrophilic than their E-counterparts.Regarding atoms, however, the phenethyl-S of the disulfide was also found to be more electrophilic than the corresponding vinyl-S, although this was more pronounced in the Z-series.However, thermodynamics may provide a better description of thiolysis Thus taking into consideration the transition state and the expulsion of the enethiolate leaving group, the vinyl disulfides emerge as more favourable over the dihydroajoenes.5a,19,24] Ajoene Targets βCys-93 in Bovine Haemoglobin Having identified that the vinyl disulfides are stable in serum but not in blood, an investigation was conducted to identity the putative erythrocyte S-thiolysis target of ajoene; namely, haemoglobin (Hb).Two samples of human Hb were prepared by incubation with 100 μM Z-ajoene or vehicle alone.The samples were subsequently trypsin-digested to afford peptide fragments followed by LC-ESI-MS analysis (Figure 7A).From visual inspection of the spectra, there are two isotope clusters with the mass to charge ratios (m/z) at 1421 and 711 in the untreated sample that are absent in the treated sample.This fragment originates from the peptide sequence GTFATL-SELHCDK as [M + H] + (m/z = 1421.6729)and [M + 2H] 2 + (m/z = 711.3401)ions (Figure 7B and C).Each of these peaks are shifted by 73.01 Da in the treated sample to two new clusters at 1493.67863 and 747.34462 respectively (calculation in S1).This mass increase corresponds to a thioallyl fragment originating from the S-allyl group of ajoene, corroborating the regioselectivity results from before.The identified fragment corresponds to the cysteine residue βCys-93, a known redox-reactive cysteine [31] demonstrating that Z-ajoene modifies bovine Hb at βCys-93.This cysteine residue is in close proximity to the haem moiety of human haemoglobin.

Discussion
Research on allicin has demonstrated its role as a potent redox toxin within living cells and blood, where it can deplete low molecular weight thiols like GSH through the formation of Sallyl disulfide conjugates. [32]While the S-thioallylation process in allicin is akin to a thiol-disulfide exchange, the presence of the polarised sulfinyl group (allylSO) serves as a superior leaving group compared to a disulfide because it can localise the charge onto oxygen in the sulfenate ion.Consequently, allicin exhibits increased reactivity and serves as a more potent oxidant.In light of the findings from our present study, this heightened electrophilic nature aligns extremely well with allicin's remarkably short half-life in the liver and blood, given the high concentrations of free GSH, ranging from 2-5 mM in blood to up to 10 mM in hepatocytes. [33]Similarly, our studies and those of others indicate that ajoene also exhibits instability in blood, explained here by the same trends in structure-activity relationships linking to thiolysis exchange.As a result, thiolcontaining components in the bloodstream are responsible for the rapid degradation of ajoene. [34]However, ajoene engages in more selective thiol-disulfide exchange reactions due to its reduced chemical reactivity when compared to allicin.
The βCys-93 residue is positioned in a solvent accessible region close to the iron centre in Hb. [31b] It is susceptible to oxidation, [31a,c,35] and its modification plays a role in oxygenlinked conformational changes.Owing to their role in oxygen transport, erythrocytes contain elevated levels of antioxidants (e. g.GSH) to protect both Hb and the RBC membrane from oxidative stress.Under oxidative stress, glutathione is able to Sglutathionylate reactive cysteines to form protein-SSG mixed disulfides.31b] In support of Hb oxidation occurring in vivo and in response to garlic, it has been reported that ingesting garlic or onions may lead to haemolytic anaemia in several mammals. [37]While anaemia has been reported in animals, this is not a known side effect of garlic consumption in humans.
5b] The presence of the vinyl disulfide was necessary for achieving high cancer cell cytotoxicity, but also resulted in an impaired bioavailability.From a chemistry theory perspective, the increased stability of the resonance-stabilised enethiolate compared to a thiolate leaving group of a disulfide is proposed as an important driver of thiolysis in ajoene.This was supported by modelling of the reaction enthalpies in which vinyl disulfides were found to be exergonic compared to dihydroajoenes which were endergonic.Of these, the Z-vinyl disulfides were the most exergonic, which aligned with the cancer cell cytotoxicity findings.Similarly, from a kinetic perspective, it was found that both the global electrophilicity index and the Fukui function provided further insights.For ajoene and its deoxy derivatives, cancer cell cytotoxicity was found to relate to the S-thiolating ability of the allyl-S (non-vinyl) sulfur of the Z-vinyl disulfides.Consistent with these findings were that the dihydroajoenes were less electrophilic, with endergonic thiolysis outcomes translating to a reduced cytotoxicity, although with improved bioavailability.The findings regarding S-thiolysis susceptibility agree with previous cytotoxicity findings [5a,19,24] in which Zajoenes were shown to be more cytotoxic than E-ajoenes.
The presence of the sulfoxide group enhanced cytotoxicity in the dihydroajoenes but reduced it in the vinyl disulfides, also reflected in the electrophilicity parameters.The phenol and methoxy functionalities had similar effects on cytotoxicity, whereas incorporation of the amide caused the compounds to be both less active against cancer cells, and also less stable in blood, likely because of a significant decrease in lipophilicity and the ability to penetrate cells.Thus, the more lipophilic phenol analogues had improved IC 50 values (and an improved stability compared to their amide counterparts), suggesting that lipophilicity (phenol analogues) may favour cytotoxicity.Conversely, blood instability may be linked to more hydrophilic ajoenes (amide analogues), and these opposing reactivity preferences may in turn speak to the different thiolysis targets of ajoene in blood and cancer cells.

Conclusions
The cancer cell cytotoxicity of the ajoene chemotype may be explained in terms of the ease of thiolysis exchange.This in turn appears to be governed by both kinetic and thermodynamic parameters related to the electrophilicity of the allyl-S sulfur of the vinyl disulfide, as well as the overall thermodynamics of the thiolysis.An important observation from this study is that a trade-off between cancer cell cytotoxicity and blood stability is achieved for the dihydroajoenes.These are slightly less cytotoxic than their vinyl disulfide counterparts, with IC 50 values still within an acceptable low micromolar range, yet with a retention of blood stability.With their improved bioavailability and good cancer cell cytotoxicity, the deoxydihydroajoenes present themselves as promising leads for future therapeutic development.

General Reagents and Cell Lines
Tissue culture media and fetal bovine serum (FBS) were purchased from GIBCO (Grand Island, NY).MeOH was from Fisher Scientific (Waltham, MA) and crystal violet, trypan blue, dimethyl sulfoxides (DMSO) were from SigmaAldrich (St. Louis, MO).Human breast cancer cell lines MCF-7 and SK-BR-3, and a normal mammary epithelial cell line MCF-10 A were obtained from American Type Culture Collection (ATCC, Manassas, VA).The SUM159 cell line was purchased from Asterand (Detroit, MI) and maintained according to the supplier's instruction.Monolayer cultures of MCF-7 cells were maintained in MEM supplemented with 0.1 mmol/L nonessential amino acids, 1 mmol/L sodium pyruvate, 10 % FBS, and antibiotics and SK-BR-3 cells were maintained in Macoy's 5 A media supplemented with 15 % FBS and antibiotics.The MCF-10 A was cultured in Mammary Epithelial Growth Medium (Clonetics, San Diego, CA) supplemented with 100 ng/mL cholera toxin (Calbiochem, La Jolla, CA).MCF-7, SK-BR-3, and SUM159 cell lines were authenticated in 2015 and 2017.Each cell line was maintained in a humidified atmosphere of 95 % air and 5 % CO 2 at 37 °C.

Cell Viability: Trypan Blue Assay
Each cell line (1×10 5 ) in complete medium was plated in twelvewell plates and allowed to attach overnight.The following day, the medium was replaced containing different concentrations of the desired E-ajoene (EA) or its analogues, and the plates were incubated for 24 h or 48 h at 37 °C.Stock solutions of EA or its analogue was prepared in DMSO and diluted with complete media immediately before use and an equal volume of DMSO (final concentration < 0.1 %) was added to the control wells.At the end of the incubation, both floating and adherent cells were collected by trypsinization and centrifugation, and cell pellets were resuspended with 200 μL of 0.1 % trypan blue solution.Live cells only were counted under an inverted microscope.

Cell Viability: MTT Assay
The oesophageal cancer cell-line WHCO1 was derived from a biopsy of primary oesophageal squamous cell carcinoma of South African origin. [38]Cells were all incubated at 37 °C under 5 % CO 2 and cultured with antibiotics in DMEM (Dulbecco's Modified Eagle Medium) containing 10 % FBS (Gibco).For experiments, cells were plated at the specified cell density and allowed to recover overnight before treatment with compounds (dissolved in DMSO).Cytotoxicity of compounds was evaluated using the standard 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide cellular viability assay.Briefly, WHCO1 or Het-1 A cells were seeded at a density of 2.5 and 5.0×10 3 cells per well respectively (90 μL) in 96 well plates and allowed to recover overnight.Compounds in DMSO were added to the cells (10 μL) at relevant concentrations (2-fold dilutions) and incubated for 48 h.Control cells received 0.1 % DMSO or media alone.To confirm that the compound itself does not interfere with the assay, compound in the absence of cells was included as a control.Thereafter 5 mg/mL MTT (10 μL) was added to each well and incubated for 4 h at 37 °C, followed by addition of 100 μL of 10 % SLS in 0.01 M HCl to solubilise the formazan crystals.The absorbance was read at 595 nm on a Multiscan FC plate reader (Thermo Scientific), and data was analysed using Graphpad Prism 4 software using sigmoidal dose-response variable slope curve fitting.

Clonogenic Assays
MCF-7 Cells (5×10 2 cells) were seeded in triplicate in a six well plate and allowed to adhere overnight.The following day they were treated with indicated doses of 5 or vehicle (DMSO) followed by incubation for 9 days.Cells were then fixed with 100 % MeOH and stained with 0.05 % crystal violet and colonies (> 50 cells) were manually counted.

In vitro Blood Stability Assay
Fresh mouse blood (6-8 weeks old, BALB/c, female) was collected and kept at 4 °C in K 2 EDTA coated tubes.The blood (400 μL) was transferred to a glass vial and heated to 37 °C in a water bath under stirring.The blood was spiked with 4 μL of 100 mM ajoene analogue (in DMSO) to give a final concentration of 1 mM ajoene and 1 % DMSO.At specified time points (0, 5, 10, 15, 20, 60 and 120 min), 50 μL of blood was removed and added to 125 μL of CH 3 CN.The sample was sonicated for 10 min at 37 °C, centrifuged for 10 min at 12500 rpm and filtered through a 22 μm nylon syringe filter (Sigma) into a 2 mL HPLC vial with 0.2 mL glass insert (Supelco).Each ajoene analogue was tested in triplicate where each repeat was performed on blood from a different mouse.The samples were run on an Agilent 6410 TQ MS, with an Agilent 1200 HPLC using a Kinetex 1.7 μm EVO C18 column (LC column 50×2.1 mm), 1 μL injection volume, flow 1.2 mL/min; gradient: t = 0 min: 95 % A-5 % B; t = 0.2 min: 95 % A-5 % B; t = 1.5 min: 0 % A-100 % B; t = 1.9 min: 0 % A-100 % B; t = 2.20 min: 95 % A-5 % B; t = 2.70 min: 95 % A-5 % B. Mobile phase A: Water, 0.1 % formic acid; Mobile phase B: Acetonitrile, 0.1 % formic acid ESI + , UV chromatogram (range 210-400 nm).The areas under the absorbance curve (as validated from the M + ion of the MS read-out) were normalised (where area of t 0 = 100 %) and plotted against time.The respective ajoene compound half-life was calculated using the one phase decay equation in Graph-Pad Prism 5 software (GraphPad Software, La Jolla, CA).

Mass-Spectroscopic Analysis of Ajoene-Treated Haemoglobin
Purified human haemoglobin protein (Sigma-Aldrich) was suspended in PBS (1 mg/mL) and treated with 100 μM Z-ajoene in DMSO or DMSO alone (control) and incubated at 37°C for 2 hours.The protein samples were then purified by SDS-PAGE, the gel was stained with Coomassie and the identified bands were excised.Gel pieces were treated with trypsin (Promega) at a final trypsin:protein ratio of 1 : 20 made up to 50 μL with 50 mM NH 4 HCO 3 (Sigma-Aldrich).Samples were digested for 18 hours at 37 °C.Peptides were then dried by vacuum centrifugation and resuspended in 0.1 % formic acid (Sigma-Aldrich) and 2.5 % acetonitrile (Anatech) to a final concentration of 500 ng/μL.Samples were then stored at À 80 °C until analysis.Nano-RP LC chromatography was performed using a Dionex Ultimate 3000 nano-HPLC system.LC-MS/MS analysis was conducted with a Q-Exactive quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) coupled with a Dionex Ultimate 3000 nano-HPLC system.The mobile phases consisted of solution A (0.1 % formic acid in water) and solution B (100 % CH 3 CN, 0.1 % formic acid).The HPLC fractionated peptides were dissolved in sample loading buffer (2.5 % CH 3 CN, 0.1 % formic acid) and loaded onto a C18 trap column (100 μm×20 mm×5 μm).Chromatographic separation was performed using a C18 column (75 μm×250 mm×3.6 μm).The mass spectrometer was operated in positive ion mode with a capillary temperature of 250 °C and an applied electrospray voltage of 1.95 kV.Database interrogation was performed by the Centre for Proteomic and Genomic Research (CPGR, Cape Town, South Africa) with the Mascot algorithm using the MSDB database on a GPS workstation.

UV-Vis Spectroscopy
Stock solutions of 200 mM Z-ajoene or bisPMB were prepared in DMSO.Fresh blood was collected from a Balb/C mouse and placed on ice in a heparin coated tube.A 10 μL aliquot of the blood was diluted into 2 mL of PBS buffer, pH 7.4 in a 2 mL quartz cuvette.The background spectrum was recorded from 700-230 nm in a Shimadzu UV-1800 UV-Spectrophotometer against a blank containing PBS alone.To both test sample and the blank was added increasing concentrations (0-200 μM) of Z-ajoene or bisPMB to reach a maximum concentration of 0.1 % DMSO.

Computational Details
Conformational space was explored using the Conformer-Rotamer Ensemble Sampling Tool (CREST 2.12). [39]The default metadynamics-based iMTD-GC workflow was used with the GFN2-xTB semiempirical Hamiltonian. [40]The aqueous solvent was included using the analytical linearised Poisson-Boltzmann (ALPB) model. [41]From the full set of conformers obtained, the fifty with the lowest energy were extracted and further refined by geometry optimisation at the ωB97XÀ D/def2-SVP [30] density functional level of theory (DFT) with the solvation model based on density (SMD) to describe the water solvent. [42]Duplicate conformers were then excluded based on similar values of the electronic energy, RMSD (root mean square deviation) in atomic positions from the global minimum conformer and calculated rotational constants.The nature of each stationary point as a local minimum was confirmed using computed harmonic vibrational frequencies, which were also used to calculate zeropoint energies as well as thermal and entropic corrections to the enthalpy and free energy, at 298.15 K. Vibrational entropy was calculated using the quasi-RRHO (rigid rotor harmonic oscillator) model of Grimme in which vibrational frequencies below ω 0 = 100 cm À 1 are treated as free rotors and for those above, the RRHO expression is retained; a damping function is used to interpolate between these two expressions. [43]r the thiolysis exchange reactions, reaction enthalpies (Δ r H) and reaction free energies (Δ r G) were calculated as the difference between the enthalpies or free energies of the most stable conformer of the reactant and product species (found as explained above, using CREST and further DFT refinement), respectively, computed at the ωB97XÀ D/def2-SVP level of theory.All DFT computations were done using Gaussian 16, [44] with the default integration grid and convergence criteria for the SCF iterations and optimisations.
Electrophilicities were calculated using the global electrophilicity index ω, defined by Parr et al. [27] as where m ¼ À IþA 2 is the electronic chemical potential and h ¼ I À A is the chemical hardness, and I and A are the ionisation energy and electron affinity, respectively. [45]One-electron Kohn-Sham frontier molecular orbital energies were used to estimate I and A, i. e., I � À e HOMO and A � À e LUMO , [46] to consequently give m ¼ e HOMO þe LUMO 2 (2) and h ¼ e LUMO À e HOMO : (3) To assess the effect of conformation on reactivity, electrophilicities were calculated for each of the ten lowest energy conformers identified in the conformational search, and a Boltzmann weighted average was determined according to w h i¼ where p i are the Boltzmann probabilities calculated at 298.15 K, in which G i are the Gibbs free energies obtained as described above, R is the molar gas constant and T = 298.15K.
Within the framework of conceptual DFT, [47] the regioselectivity of reactive sites can be quantified using the Fukui function, f r ð Þ, which Parr et al. [28b] defined in terms of the following two equivalent partial derivatives, where m is the chemical potential, 1 r ð Þ is the electron density, N is the total number of electrons and u r ð Þ is the external potential, which for an isolated molecule derives from the potential exerted by the nuclei.The Fukui function therefore describes the response of a molecular system to a variation in the number of electrons.28b] However, because the derivative is discontinuous [48] it must be approximated, either from the left (by addition of an electron) or the right (by removal of an electron).Addition of an electron then leads to the electrophilic or acceptor Fukui function, which measures reactivity towards a nucleophilic reagent.The electron densities with and without the added electron are given by 1 Nþ1 r ð Þ and 1 N r ð Þ, respectively.A nucleophilic Fukui function f À ðrÞcan be defined analogously using 1 N r ð Þ and 1 NÀ 1 r ð Þ, f À ðrÞ ¼ 1 N ðrÞ À 1 NÀ 1 ðrÞ: Furthermore, under the frozen orbital approximation, f þ ðrÞ � 1 LUMO ðrÞ and f À ðrÞ � 1 HOMO ðrÞ, [28b,46] which conveniently connects the Fukui function to concepts in frontier molecular orbital theory.
28c] However, since atomic partial charge is not a quantum mechanical observable its definition is not unique, and many other schemes exist [49] that might also be used in Equations 9 and 10.The use of Mulliken charges has been questioned. [50]28a] Rather than explicitly computing the electron densities (and atomic partial charges) of the N and N + 1 species, so-called orbital-weighted (OW) electrophilic (and nucleophilic) Fukui functions [29] can conveniently be calculated using the electron density of the unperturbed molecule only, where w i are Gaussian-based weighting factors that depend on the chemical potential, orbital energies, and the width of the Gaussian function, Δ, which determines the range of molecular orbitals that contribute to this reactivity index, The value of Δ was set to 0.1 E h as recommended by Pino-Rios et al. [29b] The corresponding atom-condensed values were then calculated by numerical integration of the OW Fukui function inside the Laguerre-Voronoi cell of the specified atom, determined using van der Waals radii from Mantina et al. [53] OW atom-condensed electrophilic Fukui functions were calculated using AaronTools [54] and the SEQCROW plugin [55] to ChimeraX. [56]r calculation results see the Supplementary Tables: Table S1.Electronic energies, Gibbs free energies, HOMO and LUMO energies, chemical potential, absolute hardness and Parr electrophilicities for each of the 10 lowest energy conformers, as well as the Boltzmannaveraged electrophilicities at 298.15 K for each conformer ensemble.Table S2.Reaction enthalpies and Gibbs energies for the thiolysis exchange reactions at 298.15 K. Table S3.Electrophilic atom-condensed Fukui functions calculated using minimum basis set Mulliken (MBS), Hirshfeld and iterative Hirshfeld (Hirshfeld-I) population analysis, as well as using the orbital-weighted (OW) procedure.

Statistical analysis
A two-tailed, unpaired t-test was used to ascertain significant differences between untreated and treated groups.Statistical significance was defined as a p-value of less than 0.05, where * pvalue < 0.05; ** p-value < 0.01; *** p-value < 0.005.

Figure 1 .
Figure 1.Overview of the OSCs found in garlic.Top: The thiosulfinates are generated upon tissue damage and rearrange/decompose to second generation OSC's.Bottom: chemical structures of the main OSC chemotypes in crushed and processed garlic.

Figure 3 .
Figure 3. Cytotoxicity of the phenol series against cancerous and normal breast cell lines.(A) Cytotoxicity of E-ajoene (EA) compared to phenol ajoenes 4, 5, 9, and 10 on the survival of SUM159 cells determined by the Trypan blue dye exclusion assay.Tested at 20 μM for 48 h.(B) Cytotoxicity of 5 against MCF-7, SK-BR-3, and MCF-10A cells after 24-h treatment in the concentration range 0, 5, 10 and 20 μM.(C) Phenol ajoene analogue 5 inhibits colony formation in MCF-7 cells.Results are analysed by one-way ANOVA followed by Dunnett's multiple comparison test and presented as mean � SD (n = 3).The P value less than 0.05 is considered as significant and symbolized with asterisk.Veh, vehicle; EA, E-Ajoene.

Figure 4 .
Figure 4.In vitro stability of ajoene analogues in mouse blood.(A) Blood stability of phenol series 4, 5, 9 and 10 as percentage remaining vs incubation time.(B) Blood stability of amide series 6, 7, 11 and 12 as percentage remaining vs incubation time.(C) Comparison of the percentage compound remaining at 20 min vs the cytotoxicity IC 50 against WHCO1 cells.(D) LC/MS quantification of percentage bisPMB remaining following incubation with mouse blood for 120 min at 37 °C: whole blood (WB), plasma fraction, red blood cell (RBC) fraction.(E) UV-vis spectra of whole mouse blood following incubation with bisPMB at 25-200 μm.

Figure 5 .
Figure 5.In silico modelling of the thiolysis electrophilicity of the E-and Zvinyl disulfides 4-7 and dihydroajeones 9-12.(A) The calculated global electrophilicity index (w); and the (B) condensed electrophilic Fukui function calculated for vinyl-S and phenethyl-S.

Figure 7 .
Figure 7. Z-ajoene modifies haemoglobin at Cys-93.(A) Workflow of sample preparation for proteomics analysis of S-allylated hamoglobin.(B) Mass spectra of untreated (UT) and Z-ajoene treated (T) bovine haemoglobin samples.Boxed showing the peaks of interest for (M + H]*.(C) Mass difference detected on the Sallylated peptide.

Table 1 .
Structure-activity relations of ajoene probing the vinyl disulfide/sulfoxide core and solubility in relation to cancer cytotoxicity (WHCO1 cells) and blood stability.