On the nature of the electronic effect of multiple hydroxyl groups in the 6-membered ring - the effects are additive but steric hindrance plays a role too

Research during the last two decades has shown a remarkable directional component of the substituent e ﬀ ects of hydroxy groups, which has a profound e ﬀ ect on the properties of hydroxylated compounds such as carbohydrates. While the epimerisation of a single hydroxyl function is well studied the conse-quence of multiple epimerisations is more speculative. In this work the e ﬀ ect of three epimerisations was investigated. To this end epimeric 2-phenyl iminoxylitols that have a phenyl group as a conformational anchor and thus hydroxyl groups in the axial or equatorial position, respectively, were synthesized and their p K a and conformation were studied. The results show that the large di ﬀ erence in the electronic e ﬀ ect between the axial and equatorial hydroxyls is partially cancelled by counteracting steric hindrance from 1,3-diaxial interactions. Hydrogen bonding does not appear to play any role in the electronic in ﬂ uence of the hydroxyl groups.


Introduction
The profound influence of substituents on the reactivity and properties of organic compounds has been recognized for many years. 1 The effect of substituents on the basicity of simple amines was nicely compiled and discussed more than half a century ago 2 where it was recognized that the hydroxyl was a particularly influential substituent, but was also fickle with respect to its base decreasing effect which could vary profoundly. 2a This was presumed to be due to hydrogen bonding. It was however not until recently that the highly directional and systematic aspect of this substituent effect was recognized and the importance of stereochemistry and conformation was revealed. 3 From the synthesis of diastereomeric pairs of hydroxylated piperidines we realized that 1-azafagomine and its galacto isomer have different pK a values, which reflect the different stability of the ammonium ions. It was found that the equatorial hydroxyl groups on the piperidine ring decreased the basicity much more than the corresponding axial stereoisomers. 3a,4 The substituent effects were also studied for other functional groups and found to be so consistent that the pK a can be calculated using the formula 10.7 − ∑σ s for piperidines or 7.3 − ∑σ s for hexahydropyridazines, where σ s is the substi-tuent contribution of the axial or equatorial substituent. The difference in pK a values can be attributed to several effects, but charge-dipole interactions are in common for all and provide a satisfactory explanation for the observations: each C-O bond constitutes a dipole while each ammonium group has a positive charge (mainly dispersed on the surrounding atoms)the closer the negative ends of the dipoles are to the positive charge the more favourable the structure. With this in mind it is not surprising that the stereoelectronic effect is coupled with a conformational effect, which is dependent on the protonation stage of the ring nitrogen, i.e. upon protonation it becomes even more favourable to have electro-negative substituents axially oriented as they effectively are less electron withdrawing than when equatorial due to the charge-dipole interaction. The fact that these stereoelectronic effects are strong enough to introduce conformational changes has been observed in a number of cases 5 e.g. the hexahydropyridazine derivative ( Fig. 1), which shifts conformation from 1 to 2 depending on pH 6 and also in bicyclic sulfonium salts like 3, which resides in an all axial conformation, which suffers from sterically unfavourable 1,3-diaxial interactions. 7 Generally, as the C-X dipole increase, with C-F as the extreme, 8 the conformational change placing the dipole in an axial orientation becomes increasingly important for the stabilization of the piperidinium ion. 9 In many cases the difference in stability between the two conformations is small and mixtures of conformations are present. The ratio of such mixtures influences the overall pK a value and can be determined directly by NMR from the intermediate J-couplings measured at room temperature 10 or even from the pK a itself. 6 The pK a values for the two conformational extremes 1 and 2 ( Fig. 1) can be calculated using the formula mentioned above giving pK a = 5.5 and 6.8, respectively, whereas the measured value is 6.2 as a result of both conformers being present. The connection between the substituent effects in nitrogen heterocycles, mainly piperidines, and other heterocycles has been thoroughly studied over the last decade, 11 and it is well accepted that the stereoelectronic effects observed in piperidines are linearly correlated with the anomeric reactivity in e.g. carbohydrates. 3b,12 As an example it has been revealed that there are linear correlations between the rate in glycoside hydrolysis or glycosylation reactions and the acidity of piperidinium salts, with analogous substituent patterns. 12a The generality of these effects is such that they with great success have been used to predict the behaviour and reactivity of carbohydrates such as anhydrosugars 13 and axial rich glycosyl donors 14 as well as hydroxymethyl conformers. 15 The pK a -calculations of piperidinium derivatives assume that substituent effects are additive, which for a significant number of reference compounds hold true with a relatively small margin of error. However extreme differences in the basicity of conformers are predicted by this method and such differences have so far not been confirmed experimentally due to the unavailability of appropriate reference compounds. The purpose of this work was to investigate whether such extreme differences in base strength would be observed in diastereoisomers. For this purpose we targeted the two epimers of 5-Cphenyliminoxylitol, 6 and 7, and determined their pK a values (Fig. 2). As model compounds 6 and 7 are ideally suited because the very bulky phenyl group in the 5-position (carbohydrate numbering) acts as a conformational anchor fixing the conformation so that the three hydroxyl groups will either be all equatorial or all axial, i.e. giving the two extremes. The differences in pK a values are therefore large and calculated to be 2.0 pK a units, with 7 being the most basic amine (Fig. 2).

Results and discussion
The two target piperidines have either a D-gluco or L-ido stereochemistry if numbered from the CH 2 group using carbohydrate numbering. In view of this, diacetone glucose is an obvious starting material as the stereogenic centers at C2, C3 and C4 can be maintained unchanged while phenyl and nitrogen are installed at C5. Using this approach aldehyde 8 was prepared in 3 steps from glucose acetonide 16 and a Grignard addition using phenyl magnesium bromide was carried out to give an approx. 12 : 1 ratio of the epimers (9 and 13), with L-ido (13) being the major product. 17 Since both were required, a twostep conversion of 13 involving PDC oxidation and NaBH 4 reduction provided the D-gluco derivative 9 as the main product (approx. 10 : 1 over 13). The L-ido isomer 13 was readily tosylated followed by nucleophilic substitution using NaN 3 in DMF to give the fully protected azide 14. Removal of the isopropylidene group and hydrogenation and benzyl hydrogenolysis over Pd/C afforded the 5-C-phenyl-D-gluco-1,5-iminopentitol 6. The same approach could however not be applied for the epimeric alcohol 9 as the tosylation did not proceed even under more forcing conditions. This is probably due to steric hindrance. To circumvent this, the 3-O-benzyl group was removed by hydrogenolysis and a cyclic sulfite 11 was prepared using thionyl chloride. Treating this with NaN 3 in DMF gave the desired azide 12. The acidic removal of the isopropylidene protective group followed by hydrogenolysis using Pd/C and triethylsilane (TESH) gave the target compound L-2-phenyl-iminoxylitol 7. The hydrogenation/hydrogenolysis reactions were generally sluggish and attempts to use more forcing conditions mainly gave the pyridine derivative 15. Changing the catalyst to Pearlman's catalyst (Pd(OH) 2 ) or Raney® Ni did not improve the reaction.
In order to study the influence of internal hydrogen bonding four additional derivatives were synthesized. The H-bond donating capability of the 3-OH group was investigated by preparing the 3-O-methyl derivatives 18 and 20 (Scheme 2). The D-gluco epimer 18 was synthesized by 3-O-methylation of diacetone glucose followed by selective removal of the 5,6-Oisopropylidene group, periodate cleavage and addition of phenylmagnesium bromide to obtain alcohol 16 predominantly as the L-ido isomer. Tosylation and subsequent S N 2 substitution using sodium azide gave azide 17, which was treated with AcOH (75%) to successfully remove the isopropylidene group from where hydrogenolysis to 18 was attempted. Neither hydrogenolysis with H 2 (g) in the presence of Pd/C, Pearlman's  catalyst (Pd(OH) 2 ) or RANEY® Ni gave satisfactory results nor did a two-step procedure of the Staudinger (Ph 3 P) reaction and NaBH 4 reduction give satisfactory yield. However, changing the hydrogen source in the catalytic reduction solved the problem and hence using triethylsilane (TESH) in combination with Pd/C gave the 3-O-methyl 2-phenyl-L-iminoxylitol 18 in reproducible good yields. In order to prepare 20 it was necessary to start with the late intermediate 12 (Scheme 1) where the azide was already present. Compound 12 was methylated with MeI/ NaH, deprotected with acetic acid and subjected to reductive amination according to the protocol above (Pd/C, TESH) to give 20 (Scheme 2).
The D-gluco epimer 23 and the L-ido epimer 27 (Scheme 3) having 2-and 4-hydroxyls protected with methyl groups were prepared starting from the late stage intermediates 14 and 12 (Scheme 1). Compound 12 had to be benzylated to the 3-O benzyl derivative 24 (Scheme 3). Then both epimers followed the same reaction sequence of removal of isopropylidene with acetic acid followed by reductive amination, which could be performed using Pd/C and TESH without any loss of benzyl ether. The obtained piperidines 21 and 25 were Boc-protected, methylated and the benzyl groups removed using Pearlman's catalyst and hydrogen. The treatment of the obtained piperidines 22 and 26 with TFA gave the desired compounds 23 and 27 as their corresponding piperidinium trifluoroacetates (Scheme 3).

Conformational studies
With the 6 compounds (6,7,18,20,23 and 27) prepared, their respective conformations were studied using NMR. Since the conformation can be strongly dependent on the pH, as mentioned in the Introduction, the compounds were studied in both their amine and ammonium forms by adding either NaOD or DCl to the NMR sample. As Table 1 shows there is practically no conformational change upon protonation and hence the role of the 2-phenyl group as a conformational anchor that keeps the conformation fixed in one chair form is confirmed. In the D-gluco series (6,18 and 23) the compounds, regardless of pH, show large trans vicinal couplings in the range 9-12 Hz for J 23 , J 34 and J 45 showing that these compounds are in the 4 C 1 chair conformation (Table 1). The compounds in the L-ido series (7, 20 and 27) display very small couplings (broad singletsbs) for J 23 , J 34 and J 45 , which is consistent with a 1 C 4 chair conformation (Table 1) and inconsistent with a 4 C 1 chair or B 1,4 boat conformation. In both cases this is as anticipated.

Titrations
The pK a values of the 6 compounds (6, 7, 18, 20, 23 and 27) were determined using potentiometric titration. At least two titrations were conducted per compound and the average results are presented in Table 2 with the uncertainty being within 0.1 units. From the results we see some general and interesting trends. Firstly, the difference in the pK a value between the epimeric pairs, 6/7, 18/20 or 23/27, is 1.1 to 1.3 pH units, respectively, with the all axial epimer being more basic. While this qualitatively is the anticipated result it is considerably less than is predicted from calculations as these predict a difference of approximately 2 pK a units. Secondly O-methylation of the 3-OH has no significant influence on the pK a value as seen from the essential identical values obtained from the pairs 6/18 and 7/20. Thirdly the 2,4-di-O-methyl derivatives are found consistently less basic than the hydroxylated counterpart as seen from the 0.4 unit lower pK a value compared to 6 and the 0.6 unit lower pK a value of 27 compared to 7.

Discussion
The pK a values of the phenylpiperidines in the 4 C 1 conformation (6, 18 and 23) are 6.1, 6.1 and 5.7 and are very much in line with the values found in similar compounds or are at least readily explained. A comparison of the pK a value of 6.1 for compound 6 with that of 1-deoxynojirimycin (28) of 6.7 ( Fig. 3) 3a reveals that the value is reasonable since the only difference between these molecules is that the hydroxymethyl group in 28 has been replaced with a phenyl group. This should decrease the basicity since the phenyl group is considerably more electron withdrawing as reflected by its higher inductive substituent constant (0.94 versus 0.66), 18 and this is seen as the α-phenyl group reduces the base strength by 1.3 19 whereas hydroxymethyl only reduces it by 0.7. 3a The influence on the pK a of O-methylation in compounds 18 and 23 is also in line with what was previously found with 1-deoxynojirimycin derivatives where the tetra-O-methyl derivative 29 11b  m is multiplet, s is singlet and bs is broad singlet. ( Fig. 3) was less basic ( pK a 6.0) than 28 (pK a 6.7). The reason for this reduced basicity is the slightly higher electron withdrawing power of O-alkyl over OH, i.e. when 4-OH groups are replaced with OMe pK a drops to 0.7. However a single methylation on the most remote hydroxyl group has little influence as 18 that has the same pK a as 6. But a larger base reducing effect of 0.4 is seen when the closer 2-and 4-OH-groups are methylated (23). Altogether this shows that 6, 18 and 23 behave as we would expect from earlier findings: the pK a is mainly a result of the inductive and electrostatic effects of the phenyl group and the 3 equatorial OH or OMe groups, with OMe being slightly more electron withdrawing compared to an OH group. However the pK a values of the phenylpiperidines in the 1 C 4 conformation (7, 20 and 27) are significantly smaller than would be anticipated from earlier findings. If we as above set the base reducing influence of the α-phenyl group to 1.3, 19 we find that the pK a of 7 should be 8.2 ( pK a = 10.7-1.3-0.5-0.5-0.2) if only inductive and electrostatic effects played a role and not the observed 7.4 ( Table 2). The fact that the basicity is remarkably lower is also seen from many literature examples some of which are shown in Fig. 3. Epimerisation of a single benzyloxy group from the equatorial to axial position in piperidines 30 and 31 increases the pK a by 1.0, 11b which is almost the same as the difference between the pairs 23 and 27 despite the fact that there are two methoxy groups and one hydroxyl group changing from equatorial to axial. Also galactodeoxynojirimycin 32 (Fig. 3) has a pK a of 7.5 and is thus slightly more basic than 7 despite the fact that it has only 1 and not 3 axial OH groups. Thus the all axial derivatives 7, 20 and 27 are clearly much less basic than anticipated from previous data except for one example: the 3,6-anhydro derivative 33 (Fig. 3) was previously prepared and its pK a value was determined to be 7.2, 11b which is also an extraordinary low value. This extraordinary, according to electronic effects, low basicity that compounds 7, 20, 27 and 33 display must be due to the second important factor that influences the base strength of aminessteric hindrance. 2a These compounds differ from the other piperidines in that they all have two axial non-hydrogen substituents β to the nitrogen. Therefore the protonation of the nitrogen will lead to two 1,3-diaxial steric interactions between a hydrogen and a hydroxyl or methoxy group, which obviously is unfavourable enough to reduce the pK a of these compounds with 0.8 to 1.0 units (Fig. 4). The fact that the influence of the 1,3-diaxial interactions from hydrogen is high is clearly seen in literature examples: while tert-butylcyclohexane (34) is well known to have the large tert-butyl group, equatorial cis-5-tert-butyl-2-methyl-1,3-dioxane (35) has the tert-butyl group predominantly at the axial position simply due to the lack of 1,3-diaxial interactions with hydrogen (Fig. 4). 20 The effect is seen directly on the pK a in base pairs 36/37 and 38/39 (Fig. 4). 21 The compounds having the ethylene bridge β to the amine (37 and 39) are 0.87 and 0.75, respectively, less basic than their analogues having the ethylene bridge α (36 and 38). As the electronic effects from the carbon substituents are negligible 2a the lower base strength must be due to the steric hindrance caused by the 1,3-diaxial interaction between the β-bridge and ammonium hydrogen. Therefore 7, 20 and 27 do not become as basic as one would expect based on electronic effects because they are more difficult to protonate due to hindrance. The base lowering effect from the steric hindrance of the two axial hydroxyl groups is about 0.8, which is very similar to the pK a difference in base pairs 36/37 and 38/39. This similarity is reasonable as the methylenes in 37 and 39 should be comparable in size to OH and less bulky than methyl.

Conclusions
This work has shown that an all-axial 2,3,4-trihydroxypiperidine has a significantly lower base strength than can be accounted for by the electronic effect. The corresponding all equatorial trihydroxypiperidine, on the other hand, has the pK a one would expect based on the electronic effect. The reason for the low pK a of the axial isomer is the two unfavourable 1,3-diaxial interactions that is present when the amine is protonated, which makes it more difficult to protonate. Thus the electronic effects are additive but pK a is reduced by the counteracting steric effect. Partial O-methylation has no significant effect other than reducing the base strength slightly due to the somewhat higher electron-withdrawing effect of OMe compared to OH.