Model Substrate/Inactivation Reactions for MoaA and Ribonucleotide Reductases: Loss of Bromo, Chloro, or Tosylate Groups from C2 of 1,5-Dideoxyhomoribofuranoses upon Generation of an α-Oxy Radical at C3 †

We report studies on radical-initiated fragmentations of model 1,5-dideoxyhomoribofuranose derivatives with bromo, chloro, and tosyloxy substituents on C2. The effects of stereochemical inversion at C2 were probed with the corresponding arabino epimers. In all cases, the elimination of bromide, chloride, and tosylate anions occurred when the 3-hydroxyl group was unprotected. The isolation of deuterium-labeled furanone products established heterolytic cleavage followed by the transfer of deuterium from labeled tributylstannane. In contrast, 3-O-methyl derivatives underwent the elimination of bromine or chlorine radicals to give the 2,3-alkene with no incorporation of label in the methyl vinyl ether. More drastic fragmentation occurred with both of the 3-O-methyl-2-tosyloxy epimers to give an aromatized furan derivative with no deuterium label. Contrasting results observed with the present anhydroalditol models relative to our prior studies with analogously substituted nucleoside models have demonstrated that insights from biomimetic chemical reactions can provide illumination of mechanistic pathways employed by ribonucleotide reductases (RNRs) and the MoaA enzyme involved in the biosynthesis of molybdopterin.

Abstraction of H3 from the ribosyl moiety of the substrate in A (Figure 1, X = OH) by a thiyl radical (•SCys439) to give B (X = OH) is the postulated initial step of the RNR-catalyzed deoxygenation of ribonucleotides [1]. Enzyme-assisted loss of water from C2 in B would produce the 3 -keto C2 The initial mechanistic steps are compatible with theoretical modeling studies [18,19], with the biomimetic chemical experiments reported by Giese [20] and Robins [21,22], and with Stubbe's enzymatic studies with gemcitabine [13][14][15][16][17] and 2'-deoxy-2'-fluoromethylenecytidine [23,24]. However, the detection of ribosyl-based radicals during RNR-catalyzed deoxygenation of substrates remained elusive [25]. Two sets of EPR signals were observed during kinetic studies with the E441Q mutant of E. coli class I RDPR and cytidine 5'-diphosphate [26]. Signals attributed to the initially detected radical were consistent with those from a disulfide radical anion [27], and those from the second were compatible with appearance of a nucleotide-based radical of the semidione type [28]. The latter-type of signals was observed during the inactivation of RNR by gemcitabine nucleotides [16].
Giese [20] and Robins [21,22] designed chemical models to simulate such initiation/elimination cascades that begin with the generation of a C3' radical during the reduction and mechanism-based inactivation mediated by RDPRs. Lenz  from C2 as predicted for base-promoted assistance by an enzymatic carboxylate group [20]. Robins et al. designed 6 -O-nitro-2 -substituted homonucleoside derivatives that produced 6 -oxyl radicals upon treatment with tributylstannane/AIBN. The O6 radicals were positioned to abstract H3 with the generation of C3 radicals containing a 3 -hydroxyl group. Fragmentation of a 2 -O-tosyl derivative occurred by anionic elimination [29], whereas a 2 -chloro analogue fragmented by a radical elimination pathway [30]. Nucleoside derivatives with a 2 -(azido, bromo, chloro, iodo, or methylthio) group also underwent elimination of a radical species upon generation of a radical center at C3 [31].
Herein, we report that 1,5-dideoxyhomoribofuranose derivatives containing a 3-hydroxyl group with a bromo, chloro, or tosylate substituent at C2 fragment with loss of bromide, chloride, or tosylate anions upon generation of a radical center at C3. The ionic fragmentations of both ribo and arabino epimers were essentially equivalent. These model reactions provide additional experimental data that allow further illumination of mechanisms employed by ribonucleotide reductases [18,19] and the MoaA enzyme [32] involved in molybdopterin biosynthesis.
Herein, we report that 1,5-dideoxyhomoribofuranose derivatives containing a 3-hydroxyl group with a bromo, chloro, or tosylate substituent at C2 fragment with loss of bromide, chloride, or tosylate anions upon generation of a radical center at C3. The ionic fragmentations of both ribo and arabino epimers were essentially equivalent. These model reactions provide additional experimental data that allow further illumination of mechanisms employed by ribonucleotide reductases [18,19] and the MoaA enzyme [32] involved in molybdopterin biosynthesis.
The arabino 2-(chloro, bromo, and tosyloxy) compounds (13, 14, and 18) were subjected to the conditions used with the ribo epimers to probe the effect of stereochemical inversion at C2. The same equilibrating mixtures of 34a/35a were produced with Bu3SnH (entries 7, 9, and 11) and the epimeric 2-deuterio derivatives 34b/35b were generated with Bu3SnD (entries 8, 10, and 12). Incorporation of deuterium into the ketone/hemiacetal products demonstrated that elimination of tosylate, chloride, and bromide anions occurred with all three of the arabino compounds. Enhanced yields of the tautomeric product mixtures were isolated with the chloro (13, entries 7 and 8) and tosylate (      Unchanged starting materials were recovered almost quantitatively, which confirmed their thermal stability under these conditions and excluded the possibility of initial dissociation of a substituent from C2. Compound 9 also was stable at 95 • C for 2.5 h in DMF. The arabino 2-(chloro, bromo, and tosyloxy) compounds (13, 14, and 18) were subjected to the conditions used with the ribo epimers to probe the effect of stereochemical inversion at C2. The same equilibrating mixtures of 34a/35a were produced with Bu 3 SnH (entries 7, 9, and 11) and the epimeric 2-deuterio derivatives 34b/35b were generated with Bu 3 SnD (entries 8, 10, and 12). Incorporation of deuterium into the ketone/hemiacetal products demonstrated that elimination of tosylate, chloride, and bromide anions occurred with all three of the arabino compounds. Enhanced yields of the tautomeric product mixtures were isolated with the chloro (13, entries 7 and 8) and tosylate (18, entries 11 and 12) substrates. Byproduct formation was not observed with 18, whereas it was formed (~19%) with the ribo substrate 9 (entries 1 and 2, footnote d). Some hydrodebromination also was detected with the arabino 2-bromo epimer 14 (entries 9 and 10, footnote e).
Samples of 9, 13, 14, and 18 were treated with Bu 3 SnH in deuterated toluene. Fragmentation of tosylate 9 in toluene-d 8 (Bu 3 SnH/AIBN) at 75 • C was 90-95% complete in 2.5 h (TLC and 1 H NMR) and produced 34a/35a (80-85%) as sole products with no incorporation of deuterium. Analogous treatment of 9 at 55 • C for 2.5 h resulted in~50% fragmentation. Treatment of the arabino tosylate 18 and chloride 13 substrates at 75 • C showed that the fragmentation of 18 was slightly faster than that of 13 ( Figure 2; 1 H NMR spectra were used for substrate/products ratios, see the Experimental Section). Fragmentation of the ribo tosylate 9 to give 34a/35a proceeded at a rate similar to that determined with the arabino chloride 13. Treatment of the arabino bromide 14 at 75 • C produced 34a/35a plus 2-deoxy byproduct mixtures with 1 H NMR spectra that were too complex for quantitative analysis.
Molecules 2020, 25, x FOR PEER REVIEW 7 of 22 entries 11 and 12) substrates. Byproduct formation was not observed with 18, whereas it was formed (~19%) with the ribo substrate 9 (entries 1 and 2, footnote d). Some hydrodebromination also was detected with the arabino 2-bromo epimer 14 (entries 9 and 10, footnote e).  Samples of 9, 13, 14, and 18 were treated with Bu3SnH in deuterated toluene. Fragmentation of tosylate 9 in toluene-d8 (Bu3SnH/AIBN) at 75 °C was 90%-95% complete in 2.5 h (TLC and 1 H NMR) and produced 34a/35a (80%-85%) as sole products with no incorporation of deuterium. Analogous treatment of 9 at 55 °C for 2.5 h resulted in ~50% fragmentation. Treatment of the arabino tosylate 18 and chloride 13 substrates at 75 °C showed that the fragmentation of 18 was slightly faster than that of 13 ( Figure 2; 1 H NMR spectra were used for substrate/products ratios, see the Experimental Section). Fragmentation of the ribo tosylate 9 to give 34a/35a proceeded at a rate similar to that determined with the arabino chloride 13. Treatment of the arabino bromide 14 at 75 °C produced 34a/35a plus 2-deoxy byproduct mixtures with 1 H NMR spectra that were too complex for quantitative analysis.

Discussion
The mechanism proposed [1] for conversion of ribonucleoside 5 -diphosphates (A, X = OH) to 2 -deoxynucleotides (F) by RDPRs (Figure 1, inside the boxes) is supported by biochemical, chemical, and theoretical modeling studies. However, the enzymatic processing of 2 -chloro analogs (A, X = Cl) could cause inactivation by different chemistry. Incubation of a 2 -chloro-2 -deoxynucleoside 5 -diphosphate with RDPR produced 2-methylene-3(2H)furanone (G), a Michael acceptor that could cause time-dependent enzyme inactivation. Stubbe rationalized [1] that spontaneous elimination of a chloride anion (and a proton) at the active site (rather than the enzyme-assisted removal of HOH with substrates) could cause active site changes resulting in dissociation of the 2 -deoxy-3 -oxo intermediate from D (Figure 1). Successive β-eliminations of pyrophosphate from C5 and the base from C1 could generate G in solution. However, the identical 3 -keto intermediate in D was postulated in the substrate to product sequence, which makes the presence of chloride and a proton the only difference for the inactivation sequence.
We reasoned that elimination of a chlorine atom from the initial C3 radical was more likely. The electronegative character of C1 would make the elimination of chloride (with generation of positive character at C2 ) unfavorable, whereas loss of a chlorine atom with generation of a C2 radical was well precedented [38], and generation of an enol would be energetically advantageous. Elimination of a chlorine atom (rather than chloride) at the active site could have serious consequences. The chlorine radical could attack a sulfhydryl group and the resulting sulfenyl chloride could react with nucleophilic groups in the enzyme or undergo hydrolysis to a sulfenic acid. Chlorine-atom abstraction of hydrogen from an amino acid residue (and resulting radical processes) and other chlorine-radical reactions would be possible, whereas such events would not occur with a ground state chloride anion. Radical-induced disruption of active-site architecture provides a more plausible explanation for dissociation of the 2 -deoxy-3 -oxo intermediate from D.
We have shown [31] that leaving-group radical stability is crucial for substituent elimination from C2 upon generation of a radical at C3 . Treatment of 2 -(azido, bromo, chloro, iodo, or methylthio)-2 -deoxy-3 -O-phenoxythiocarbonyl nucleosides (Ha, Figure 3) with BuSnH resulted in generation of C3 radicals Ia that underwent elimination of a radical from C2 with formation of 2 ,3 -unsaturated derivative J. In contrast, treatment of the 2 -(fluoro, mesyloxy, or tosyloxy)-3 -thionocarbonates Hb generated C3 radicals Ib that abstracted hydrogen from the stannane to give 3 -deoxy-2 -(fluoro, mesyloxy, or tosyloxy) derivatives Kb. Elimination of a radical from C2 of Ia gave J, whereas homolytic scission to release a high-energy fluorine atom or a mesyloxy or tosyloxy radical is energetically prohibitive-and did not occur. In those cases, abstraction of hydrogen from the stannane by the C3 radical Ib gave the reduced products Kb.

Discussion
The mechanism proposed [1] for conversion of ribonucleoside 5´-diphosphates (A, X = OH) to 2´-deoxynucleotides (F) by RDPRs (Figure 1, inside the boxes) is supported by biochemical, chemical, and theoretical modeling studies. However, the enzymatic processing of 2´-chloro analogs (A, X = Cl) could cause inactivation by different chemistry. Incubation of a 2´-chloro-2´-deoxynucleoside 5´diphosphate with RDPR produced 2-methylene-3(2H)furanone (G), a Michael acceptor that could cause time-dependent enzyme inactivation. Stubbe rationalized [1] that spontaneous elimination of a chloride anion (and a proton) at the active site (rather than the enzyme-assisted removal of HOH with substrates) could cause active site changes resulting in dissociation of the 2´-deoxy-3´-oxo intermediate from D (Figure 1). Successive β-eliminations of pyrophosphate from C5´ and the base from C1´ could generate G in solution. However, the identical 3´-keto intermediate in D was postulated in the substrate to product sequence, which makes the presence of chloride and a proton the only difference for the inactivation sequence.
We reasoned that elimination of a chlorine atom from the initial C3´ radical was more likely. The electronegative character of C1´ would make the elimination of chloride (with generation of positive character at C2´) unfavorable, whereas loss of a chlorine atom with generation of a C2´ radical was well precedented [38], and generation of an enol would be energetically advantageous. Elimination of a chlorine atom (rather than chloride) at the active site could have serious consequences. The chlorine radical could attack a sulfhydryl group and the resulting sulfenyl chloride could react with nucleophilic groups in the enzyme or undergo hydrolysis to a sulfenic acid. Chlorine-atom abstraction of hydrogen from an amino acid residue (and resulting radical processes) and other chlorine-radical reactions would be possible, whereas such events would not occur with a ground state chloride anion. Radical-induced disruption of active-site architecture provides a more plausible explanation for dissociation of the 2´-deoxy-3´-oxo intermediate from D.
We have shown [31] that leaving-group radical stability is crucial for substituent elimination from C2´ upon generation of a radical at C3´. Treatment of 2´-(azido, bromo, chloro, iodo, or methylthio)-2´-deoxy-3´-O-phenoxythiocarbonyl nucleosides (Ha, Figure 3) with BuSnH resulted in generation of C3´ radicals Ia that underwent elimination of a radical from C2´ with formation of 2´,3´unsaturated derivative J. In contrast, treatment of the 2´-(fluoro, mesyloxy, or tosyloxy)-3´thionocarbonates Hb generated C3´ radicals Ib that abstracted hydrogen from the stannane to give 3´-deoxy-2´-(fluoro, mesyloxy, or tosyloxy) derivatives Kb. Elimination of a radical from C2´ of Ia gave J, whereas homolytic scission to release a high-energy fluorine atom or a mesyloxy or tosyloxy radical is energetically prohibitive -and did not occur. In those cases, abstraction of hydrogen from the stannane by the C3´ radical Ib gave the reduced products Kb. Because the radicals I did not contain a 3´-hydroxyl group, we prepared model compounds that were more closely related to the natural substrates for RDPR-catalyzed 2´-deoxygenations. Treatment of the adenine 6´-O-nitro-2´-O-tosyl L and uracil 6´-O-nitro-2´-chloro M analogues with Bu3SnD/AIBN/Δ generated the 6´-oxyl radicals N by attack of a stannyl radical at nitrate oxygen ( Figure 4). Intramolecular abstraction of H3´ from N gave the C3´ radicals O or S. Loss of the proton from the 3´-hydroxyl group of O and elimination of the 2´-tosylate (with shift of the unpaired electron from C3´ to C2´ and generation of the O=C3´ double bond) would drive the overall elimination of Because the radicals I did not contain a 3 -hydroxyl group, we prepared model compounds that were more closely related to the natural substrates for RDPR-catalyzed 2 -deoxygenations. Treatment of the adenine 6 -O-nitro-2 -O-tosyl L and uracil 6 -O-nitro-2 -chloro M analogues with Bu 3 SnD/AIBN/∆ generated the 6 -oxyl radicals N by attack of a stannyl radical at nitrate oxygen ( Figure 4). Intramolecular abstraction of H3 from N gave the C3 radicals O or S. Loss of the proton from the 3 -hydroxyl group of O and elimination of the 2 -tosylate (with shift of the unpaired electron from C3 to C2 and generation of the O = C3 double bond) would drive the overall elimination of toluenesulfonic acid from O to produce P. Deuterium transfer from the stannane to C2 of P followed by elimination of the trans proton/deuteron and adenine from Q would give the observed 2-(2-hydroxyethyl)-3(2H)-furanone (R) containing deuterium at C4 (C2 from the nucleoside). In contrast, loss of the chlorine atom from S would give enol T, which could undergo 1,4-elimination to U with no incorporation of deuterium (as observed). This demonstrated the distinct differences between TsO-C2 bond cleavage by a two-electron heterolysis (in O) and cleavage of the chlorine-C2 bond by a one-electron homolysis (in S) [21,29,30].
Molecules 2020, 25, x FOR PEER REVIEW 9 of 22 toluenesulfonic acid from O to produce P. Deuterium transfer from the stannane to C2´ of P followed by elimination of the trans proton/deuteron and adenine from Q would give the observed 2-(2hydroxyethyl)-3(2H)-furanone (R) containing deuterium at C4 (C2´ from the nucleoside). In contrast, loss of the chlorine atom from S would give enol T, which could undergo 1,4-elimination to U with no incorporation of deuterium (as observed). This demonstrated the distinct differences between TsO-C2´ bond cleavage by a two-electron heterolysis (in O) and cleavage of the chlorine-C2´ bond by a one-electron homolysis (in S) [21,29,30]. Barton´s nitrite ester [39] and Wagner's δ-substituted aryl ketone [40] photolysis studies showed a strong preference for six-membered transition states for hydrogen abstraction by oxyl radicals. Fraser-Reid employed that [1,5]-hydrogen shift with oxyl radicals generated from carbohydrate nitrate esters and Bu3SnH [41]. Radical-induced loss of -OTs and H + (Figure 4, path a) is also analogous to an ionic LiEt3BH-promoted rearrangement of 2´-O-tosyladenosine. Removal of the 3´hydroxyl proton by Et3BH -initiated a [1,2]-hydride shift from C3´ to C2´ with displacement of tosylate from C2´ [42,43]. That rearrangement occurred also with 2´-chloro-2´-deoxyadenosine, but at a lower rate with the poorer leaving group (chloride).
Theoretical modeling [18,44] of RDPR-catalyzed 2′-deoxygenation invoked hydrogen bonding from the 3′-OH to a carboxylate group and from H-donors to 2′-OH. Analogous attraction between the cis 3-OH in 9 and a tosylate oxygen atom was considered for possible assistance of the heterolytic cleavage of the TsO-C2 linkage. However, treatment of the arabino tosylate 18 (no cis 3-OH) gave the same 34b/35b mixture in higher yield (93%) than with 9 (~70%). Byproduct with a hydroxyl group at C6 and tosylate at C2 was isolated (~19%) from treatment of 9, but no such arabino byproduct was observed with 18. These results are more consistent with a greater population of the C2-endo/C3-exo conformation range in 9 (reduction of strain with the 2-tosylate group) that would make a sixmembered transition state for intramolecular abstraction of H3 by the 6-oxyl radical less favorable. Enhanced abstraction of deuterium from the stannane by the 6-oxyl radical would then occur to produce more of the byproduct. Greater C2-exo/C3-endo populations in 18 (reduction of strain between the arabino tosylate at C2 and the side chain at C4) would enhance the approach of the 6oxyl radical for H3 abstraction and elimination of tosylate from the C3 radical to produce 34b/35b in higher yield (93%).
Thus, noteworthy mechanistic changes were observed within our model series. Anionic elimination of tosylate from C2´ of a nucleoside occurred (Figure 4, pathway a), whereas radical elimination of a chlorine atom (pathway b) was preferred. However, treatment of our ribo anhydroalditol tosylate 9, chloro 20, or bromo 22 substrates produced the same 2-deuterio-3-ketone 34b and hemiacetal 35b mixture. The arabino chloro 13 and bromo 14 epimers also gave 34b/35b (plus some hydrodebromination with 14). In every case, the loss of bromide or chloride anions from an intermediate C3 radical resulted in formation of deuterium-labeled materials rather than homolytic loss of a halogen atom to give the unlabeled alkene. The absence of an electronegative Barton's nitrite ester [39] and Wagner's δ-substituted aryl ketone [40] photolysis studies showed a strong preference for six-membered transition states for hydrogen abstraction by oxyl radicals. Fraser-Reid employed that [1,5]-hydrogen shift with oxyl radicals generated from carbohydrate nitrate esters and Bu 3 SnH [41]. Radical-induced loss of -OTs and H + (Figure 4, path a) is also analogous to an ionic LiEt 3 BH-promoted rearrangement of 2 -O-tosyladenosine. Removal of the 3 -hydroxyl proton by Et 3 BHinitiated a [1,2]-hydride shift from C3 to C2 with displacement of tosylate from C2 [42,43]. That rearrangement occurred also with 2 -chloro-2 -deoxyadenosine, but at a lower rate with the poorer leaving group (chloride).
Theoretical modeling [18,44] of RDPR-catalyzed 2 -deoxygenation invoked hydrogen bonding from the 3 -OH to a carboxylate group and from H-donors to 2 -OH. Analogous attraction between the cis 3-OH in 9 and a tosylate oxygen atom was considered for possible assistance of the heterolytic cleavage of the TsO-C2 linkage. However, treatment of the arabino tosylate 18 (no cis 3-OH) gave the same 34b/35b mixture in higher yield (93%) than with 9 (~70%). Byproduct with a hydroxyl group at C6 and tosylate at C2 was isolated (~19%) from treatment of 9, but no such arabino byproduct was observed with 18. These results are more consistent with a greater population of the C2-endo/C3-exo conformation range in 9 (reduction of strain with the 2-tosylate group) that would make a six-membered transition state for intramolecular abstraction of H3 by the 6-oxyl radical less favorable. Enhanced abstraction of deuterium from the stannane by the 6-oxyl radical would then occur to produce more of the byproduct. Greater C2-exo/C3-endo populations in 18 (reduction of strain between the arabino tosylate at C2 and the side chain at C4) would enhance the approach of the 6-oxyl radical for H3 abstraction and elimination of tosylate from the C3 radical to produce 34b/35b in higher yield (93%).
Thus, noteworthy mechanistic changes were observed within our model series. Anionic elimination of tosylate from C2 of a nucleoside occurred (Figure 4, pathway a), whereas radical elimination of a chlorine atom (pathway b) was preferred. However, treatment of our ribo anhydroalditol tosylate 9, chloro 20, or bromo 22 substrates produced the same 2-deuterio-3-ketone 34b and hemiacetal 35b mixture. The arabino chloro 13 and bromo 14 epimers also gave 34b/35b (plus some hydrodebromination with 14). In every case, the loss of bromide or chloride anions from an intermediate C3 radical resulted in formation of deuterium-labeled materials rather than homolytic loss of a halogen atom to give the unlabeled alkene. The absence of an electronegative nucleobase on C1 (C1 is a -CH 2 -moiety in the anhydroalditol models) allowed elimination of an anion from C2 when loss of the 3-hydroxyl proton could generate an O = C3 double bond.
Ramos questioned our choice of toluene as solvent and lack of a basic residue in our prior nucleoside model studies and stated: "we concluded that the nature of the leaving substituent can be controlled; it will be anionic or radical depending on the presence or absence of a basic residue capable of deprotonating the 3 -HO group. In the enzyme, such functionality does exist, and so it can be concluded that the enzyme indeed eliminates anions, and not radicals" [19].
Clearly, the elimination of an anion is possible without a basic residue present in our current anhydroalditol models. Siegbahn [18] calculated a dielectric constant of ε~4 in the region of the enzyme active site, which is approximated much more closely by that of toluene (ε~2.4) than that of the aqueous methanol solutions (ε > 40) used by Lenz and Giese [20]. Thus, the intrinsic electronic status of the anomeric carbon C1 (inductively negative in nucleosides) or C1 (inductively positive in anhydroalditols) as well as the nature of the leaving substituent (X• or A − ) are major factors that control heterolytic elimination of an anion or homolytic elimination of a radical in biomimetic model reactions-and such factors also likely play a key role in enzyme-initiated inactivation processes.
Begley [32] also has invoked our elimination of a chlorine atom [30] upon abstraction of H3 from 2 -chloro-2 -deoxyguanosine triphosphate by a 5 -deoxyadenosyl radical cofactor in MoaA. His 2 -deoxy-3 -ketone underwent the sequential β-eliminations [30] of triphosphate and guanine to give the same 2-methylene-3(2H)furanone (G, Figure 1 Figure 5) might aid the loss of HOTs with generation of a 1,2-double bond. An analogous interaction involving the β-proton on C1 and a tosylate oxygen with arabino epimer 30 also is possible. Abstraction of hydrogen from C4 of the resulting resonance hybrid would produce furan 37. The observed similar yields of 37 plus the respective 2-O-tosyl byproducts 38 and 39 are consistent with parallel processes for the ribo 25 and arabino 30 epimers. Tosylate 25 was stable in toluene at 95 °C, which confirmed that generation of a C3 radical was necessary for the elimination of tosylate and production of 37.

Experimental Section
The 1 H (400 or 500 MHz) and 13  Merck kieselgel 60 (230-400 mesh) was used for column chromatography. Reagent grade chemicals were used, and solvents were dried by reflux over and distillation from CaH 2 (except for THF/potassium) under argon or by passing the solvents through activated alumina cartridges using a solvent purification system.