Saccharide Structure and Reactivity Interrogated with Stable Isotopes

Transcription

1 RESERVE THIS SPAE Saccharide Structure and Reactivity Interrogated with Stable Isotopes Wenhui Zhang, Reagan Meredith, Mi-Kyung Yoon, Ian armichael and Anthony S. Serianni* Department of hemistry and Biochemistry, University of Notre Dame, Notre Dame, IN USA * Several topics in saccharide chemistry and biochemistry, which were impacted by the work of Ernest Eliel and his contemporaries, are reviewed. We show how stable isotopic enrichment, NMR spectroscopy, and modern computational methods have been applied synergistically to reveal subtle and sometimes surprising properties of saccharides in solution. Examples include the use of stable isotopes to detect and quantify the cyclic and acyclic forms of reducing sugars in solution, and to investigate relationships between saccharide structure, conformation and the kinetics of anomerization. Thermodynamic and kinetics studies of cis-trans isomerization of the N-acetyl side-chains of saccharides are enabled by selective 13 -enrichment and saturation-transfer NMR methods. Redundant NMR spin-couplings sensitive to the same molecular torsion angle can be interpreted collectively to derive conformational populations of flexible fragments such as -acetyl side-chains. NMR studies of saccharide chemical transformations using stable isotopes reveal remarkable and stereospecific skeletal rearrangements such as 1 2 transposition that defied prior detection, opening the opportunity to develop new catalysts and/or to better understand catalytic mechanisms of chemical and biochemical processes involving saccharides. RESERVE THIS SPAE Serianni_Ms_Edited_3_6_17_with_Figures_ASEdits2.docx Printed 3/6/17 1

2 Introduction arbohydrates provide a unique and expansive playground on which to investigate the intra- and intermolecular forces that dictate conformational equilibria and dynamics of molecules in solution. This opportunity evolves from the enormous structural diversity of saccharides that is available for investigation with respect to carbon scaffold, configuration and substitution (1). This playground is particularly appealing because saccharides are biologically important molecules that are found in vivo in different degrees of polymerization (i.e., monosaccharides, oligosaccharides or polysaccharides) and in different modes of molecular conjugation (i.e., free in solution or appended to proteins, lipids and other biomolecules) (2). This biological relevance provides a compelling argument to investigate saccharide structure, which plays key roles in determining many important biological functions and processes, including diseases such as diabetes and cancer. Unraveling the relationships between saccharide structure and their chemical and biological functions cannot be achieved, however, by restricting studies to only those saccharides found in biological systems. Such an approach, while expedient from a biological perspective, samples only a fraction of the total structural space, space that, arguably, must be sampled generously in order to derive reliable relationships between saccharide covalent structure and higher-order structural features such as conformational equilibria and dynamics. The term structure is hierarchical (Scheme 1). In its simplest definition, it describes the atoms comprising the saccharide and the covalent bonds between them. Sequentially higher-order definitions include the absolute configuration of their constituent chiral carbons, the available conformational options (conformational equilibria), and the kinetics of exchange between accessible conformational states (dynamics). These features are influenced by solvation, be it by simple solvent molecules like water, or by functional groups present in the binding site of a biological receptor. If the saccharide contains ionizable functionality, solution ph may influence some or all of these properties (3). The impactful scientific achievements of Ernest Eliel in the field of organic stereochemistry benefitted from complementary studies of saccharides. Indeed, the book entitled onformational Analysis by Eliel, Allinger, Angyal and Morrison, published in 1965 (4), testifies to this fact, wherein many of the stereochemical principles articulated by Eliel from his studies of general organic systems were applied, tested and refined with the use of saccharides. The inclusion of Stephen Angyal as a coauthor of this seminal book was no accident; Eliel realized the central role of saccharides in confirming and amplifying the principles of stereochemistry (5) that he had worked to develop. Researchers who have subsequently built on the solid foundation provided by Eliel s seminal studies have benefitted from research tools and methods that 2

3 were unavailable, and perhaps unimaginable, in the mid-20 th century. These tools include, among others, very high field superconducting magnets (> 14 Tesla) in NMR spectroscopy to improve spectral dispersion and sensitivity (6), multi-dimensional NMR data collection to resolve and assign complex 1 H NMR spectra (7), polarization transfer methods to increase NMR sensitivity and selectivity (8), and routine access to highly enriched and pure stable isotopes such as 13, 15 N and 17,18 on large scales and at reasonable cost (9). A timely convergence of these tools enabled modern NMR structural studies of complex molecules having molecular weights in excess of 50 kd, a remarkable development considering that 1 H NMR spectrometers at MHz (1.41 Tesla) using permanent magnets and operating in continuous wave modes were just coming of age in the 1960 s when Eliel was conducting his research. In this chapter, several topics pertinent to the field of saccharide chemistry and biochemistry are discussed which were impacted by early work of Eliel and his contemporaries. We show how the interplay of isotopic enrichment and modern NMR methods, coupled to modern computational methods, has been used to reveal subtle and surprising properties of these important biomolecules, including unusual skeletal rearrangements. Saccharide Anomerization The spontaneous ring-opening and -closing of aldoses and ketoses in solution is known as anomerization (10). This process involves the acyclic aldehydo and keto forms of reducing saccharides (Scheme 2). The types and distributions of cyclic forms produced depend on aldose and ketose structure; typically only five- (furanose) and six - membered (pyranose) rings form, since larger and smaller rings have unfavorable enthalpies and/or entropies of activation (11). In addition to ring-opening and -closing, the acyclic carbonyl forms of aldoses and ketoses can also, in principle, react with solvent water to give acyclic hydrate forms (gem-diols) (Scheme 2). H H! = 13 Modern experimental measurements of anomerization equilibria are commonly made by NMR spectroscopy, and 13 NMR in conjunction with selective 13 -enrichment at anomeric carbons provides a superior approach to make these determinations (12 14). This application is illustrated in Figure 1, which shows a 13 { 1 H} NMR spectrum (150 MHz) obtained on an aqueous solution of D-[1-13 ]mannose (1) in which six monomeric forms are detected in equilibrium. Since 1 contains 99 atom-% 13 isotope at 1, the detection of the labeled carbons is ~100 times greater than for the remaining natural abundance carbons. Signals from the weak natural carbons can be observed between H 6 1 3! D-[1-13 ]mannose (1) 3

4 ppm. The intense signals at ~95 ppm arise from the 1 carbons of the dominate aldopyranose forms, while the aldofuranose 1 signals appear slightly downfield of the aldopyranose 1 signals. The very weak signal observed at ~205 ppm arises from 1 of the acyclic aldehyde form, while that at ~91 ppm arises from 1 of the acyclic hydrate form (Tables 1 and 2). Integration of the six signals gives the following percentages of forms in solution at 30 o : α- pyranose, ± 0.05%; β-pyranose, ± 0.06%; α-furanose, 0.64 ± 0.04%; β-furanose, 0.25 ± 0.04%; aldehyde, ± %; hydrate, ± 0.001% (12). The large chemical shift dispersion of the 1 signals makes 13 NMR highly suitable for the detection of cyclic and acyclic forms of reducing saccharides in solution. When the reducing saccharide is a ketose, 13 NMR provides the only reliable means to determine anomeric equilibria, since these molecules lack anomeric hydrogens. A source of error in the measurements shown in Figure 1 is the potential for signal mis-assignment, especially that for the acyclic hydrate form. This problem can be partly addressed by measuring the J-coupling between 1 and its directly attached hydrogen ( 1 J 1,H1 ) (Table 1). These 1 J H values are sensitive to structure near the 1 carbon, and their values, in addition to chemical shift, can be used to make signal assignments. An example of this approach is shown in Figure 2, which shows the 1 carbon signal of the hydrate form of 1 when its directly attached hydrogen (and other hydrogens two- and three-bonds removed from 1) are decoupled and coupled to the carbon. The large splitting (164.2 Hz) is attributed to the one-bond 1 J 1,H1. 1 J 1,H1 values of Hz are typically observed for hydrate forms, and Hz for aldehyde forms (Table 1); significant deviations from these values constitute evidence that the assignment may be incorrect. 13 { 1 H} NMR spectra such as that shown in Figure 1 provide equilibrium constants for the component equilibria of aldose anomerization through signal integration, provided that the data were acquired under conditions that allow accurate quantitative analysis (12,13). The results of studies of aldopentoses and aldohexoses are summarized in Tables 1 4, which list the 1 chemical shifts of the various forms, 1 J 1,H1 values, and percentages in solution, for the cyclic and acyclic forms of aldopentoses and aldohexoses in solution. A comparison of the percentages of acyclic forms in these aldoses is shown in Figure 3A. Aldehydic content ranges from % in solution, with solutions of allose and glucose containing the smallest percentages and those of ribose and idose the largest. Hydrate percentages range from %, with solutions of allose, glucose and mannose containing very small percentages, and solutions of idose containing the largest (0.7 %; data not shown in Figure 3B). In some cases, anomerization equilibria include other acyclic forms in addition to the carbonyl and hydrate forms. This behavior is displayed by the 4

5 biologically important α-ketoacid, N-acetyl-neuraminic acid (2) (Scheme 3). The partial 13 { 1 H} NMR spectrum of [2-13 ]2 at ph 2 and 25 o is shown in Figure 4 (15). Labeled 2 signals arising from the pyranose forms of 2 appear at ~96 ppm. The β-pyranose (2βp) is most preferred (91.2 %), followed by the α-pyranose (2αp) at 5.8% (Scheme 3). The weak signal at ~94 ppm arises from the acyclic hydrate form (2h) (1.9%; Scheme 3). The spectral region between ppm contains signals arising from the acyclic keto form (2k) (198 ppm; 0.7%; Scheme 3) and, unexpectedly, the acyclic enol form (2e) D-threose (3) (143 ppm; 0.5%; Scheme 3). Natural abundance 1 H () signals from 2αp and 2βp appear as doublets in this region (these signals are split by the one-bond 1 J 1,2 ), as do the signals arising from the amide carbons in both pyranoses. Solution conditions affect anomerization equilibria, especially temperature and ph. For example, increasing the temperature of aqueous solutions of D-[1-13 ]threose (3) exerts little, if any, effect on the percentages of furanose forms, but the percentages of the acyclic hydrate and aldehyde forms decrease and increase, respectively, with increasing solution temperature (Figure 5) (16). In contrast, the ketopentose, D-threo-pentulose (D-xylulose) (4), anomerizes to potentially give solutions containing two cyclic ketofuranoses and two acyclic forms (Scheme 4), but no acyclic hydrate can be detected by 13 NMR even when 4 is labeled with 13 at 2 (17). The percentages of the three forms depend on solution temperature as shown in Figure 6. As observed for (3), the percentage of acyclic carbonyl form increases appreciably with increasing temperature, at the expense of the β-ketofuranose. ompared to (3), solutions of (4) contain much more acyclic carbonyl form (2.4% for (3) vs 24% for (4) at 50 o ). The kinetics for each component equilibrium in aldose and ketose anomerization is obtainable from NMR spectra of anomerizing systems at chemical equilibrium. Since the acyclic carbonyl forms of aldoses and ketoses are the presumed obligatory intermediates in the exchange of cyclic forms and the formation of hydrates, selective saturation of the well-resolved carbonyl carbon signals (or aldehydic hydrogens) results in the transfer of saturation to corresponding signals arising from the cyclic and acyclic hydrate forms due to chemical exchange (16,18). The resulting rate of loss in signal intensity is determined by the ring-opening rate constants, kopen, and the spin-lattice relaxation times of the signals. This application of saturation-transfer NMR spectroscopy (19) is illustrated for the anomerization of D-[1-13 ]erythrose (5), whose anomerization equilibrium is shown in Scheme 5. Note the significantly higher percentage of acyclic aldehyde and hydrate forms of 5 compared to systems in which pyranosyl rings can form (Table 1). For 5, only furanoses form upon ring closure of the acyclic aldehyde. If 5 is enriched with 13 at 1, 5

6 three signals are observed in the anomeric carbon region (Figure 7A): α- furanose (αf), β-furanose (βf) and hydrate (h). The 1 signal of the acyclic aldehyde (not shown) is observed at ~205 ppm. Saturation of the aldehyde signal for increasing amounts of time causes significant loss of signal intensity for 1 of the αf and βf forms (Figure 7, B and ). Linearizing the data (Figure 7D) allows k open values for cyclic forms, and k dehydration for the hydrate (not shown), to be determined. Determinations of the individual K eq values for the component equilibria in Scheme 5 allow k close and k hydration values to be calculated, thus providing complete characterization of the anomerization kinetics under a specific set of solution conditions. This method is generally applicable to measure rate constants in the range s -1 ; values >10 s -1 are obtained from quantitative treatments of line-broadening in the presence of chemical exchange (Gutowsky-Holm treatment) (19c,21). The effect of phosphate group ionization on the anomerization kinetics of pentose phosphates is shown in Figure 8 for D-[1-13 H ]ribose 5-phosphate (R5P) (6) (18). This 2 3 P system is similar to that shown in Scheme 5 for 5 in that only two cyclic furanose and two acyclic forms of R5P are possible in solution. The effect H D-ribose 5-phosphate (6) of phosphate differs for both anomers, with the α- furanose more prone to ring-opening than the β-furanose at all solution ph values studied. Saturation-transfer experiments were conducted to measure k open values at ph 2.3 and 4.0, and line-broadening experiments were conducted to make k open measurements at the remaining ph values. In general, the presence of phosphate in the saccharide increases anomerization rate constants relative to the same molecule devoid of phosphate, suggesting a potential role for intramolecular catalysis in the anomerization of phosphorylated sugars in vivo (18). Kinetic studies of anomerizing systems involving pyranosyl rings have also been reported, and data for the aldohexose, D-[1-13 ]talose (7), are summarized in Scheme 6. From a thermodynamic perspective, this system is similar to that of D-mannose (1) (Figure 1), with pyranose forms dominating over furanose forms. Under the solution conditions indicated, k open values range from s -1, and k close values range from 3 43 s -1. Interconversions of talopyranoses with the acyclic aldehyde occur more slowly compared to corresponding furanose interconversions. Thus, while talofuranoses are not favored thermodynamically, they are favored kinetically (23). 6

7 Relationships Between Saccharide Structure and Anomerization Kinetics Ring opening is rate limiting for the interconversion of cyclic forms of aldoses in solution in most, if not all, cases. This behavior is demonstrated for D-talose (7) in Scheme 6, where k close values are ~1000-fold larger than k open values, and the smallest k close (2.7 s -1 ) is ~60-fold larger than the largest k open (0.046 s -1 ). However, exceptions to this behavior are likely to exist, especially in ketoses, as discussed below. In simple aldofuranoses such as those shown in Scheme 7, the anomer bearing hydroxyl groups in a cis arrangement at 1 and 2 opens more rapidly to the acyclic aldehyde than does the anomer bearing the same groups in a trans arrangement, although relative configuration at 2 and 3 affects the size of the difference (see below). This behavior (hereafter referred to as the cis-1,2 effect ) is illustrated by the kinetics data for (5) in Figure 7, and in Table 5 where kopen values for D-[1-13 ]erythrose (5), D-[1-13 ]threose (3), and several 5-modified aldopentoses, measured under identical solution conditions, are compared. The cis-1,2 effect is most apparent in furanose rings having 2 and 3 trans, as found in the threo, arabino and xylo ring configurations. A model explaining this behavior invokes anchimeric assistance by 2 as a facilitator of proton extraction at 1, the latter stimulating ring-opening to the acyclic aldehyde (Scheme 8). In the erythro and ribo rings, the cis-1,2 effect is reduced considerably, however, and in some cases it is abolished (e.g., lyxo rings). Although the data are limited, deoxygenation at 5 appears to increase kopen slightly under the given solution conditions (H 2 -catalyzed region) (10b). The effect of 4 substitution on k open appears small; for example, k open values for erythrose (5) and threose (3) range from s -1, whereas those for the 5-modified aldopentoses range from s -1, under the solution conditions given. The cis-1,2 mechanism shown in Scheme 8 may not be the only potential structural explanation for the observed differences in k open between furanose anomers. Arguing from the principle of microscopic reversibility, furanose ring conformation may also influence k open, with some conformations more prone to ring opening than others. The enhanced k open values H 2 observed in the threo, arabino and xylo ring configurations might result from preferred ring conformations that also favor ring opening. This argument evolves from ring-closure arguments where H trajectories of hydroxyl oxygen attack on the carbonyl D-erythro-pentulose (8) (D-ribulose) carbon are highly constrained (25), thus leading to a small subset of ring conformations as the immediate products of closure. This 7

8 same subset of conformations would then be favored for ring opening, but pseudorotation may not favor them at equilibrium. When favored ring conformations in solution coincide with those favored for ring opening, kopen will be enhanced. Rates of ring pseudorotation relative to k open will affect the potency of this factor. The effect of converting aldoses to ketoses on anomerization kinetics can be seen by comparing D-erythrose (5) and D-threose (3) with D-[2-13 ]threopentulose (D-xylulose) (4) and D-[2-13 ]erythro-pentulose (D-ribulose) (8) (Table 6). Ketopentoses 4 and 8 are essentially alkylated derivatives of aldotetroses 3 and 5, respectively, in which H 2 groups replace the anomeric hydrogens in the aldoses. This type of ring alkylation affects k open only to a small extent; for example, the average k open value for 3 and 5 is 0.45 ± 0.17 s -1, whereas that for 4 and 8 is 0.23 ± 0.08 s -1 (Table 6). However, the effect of this alkylation on k close is significant, with the aldoses ~30 times more reactive than the ketoses (average k close of 0.34 ± 0.27 s -1 for the ketoses vs 10.8 ± 3.7 s -1 for the aldoses). The reduced rate of ring closure in the ketoses is likely due to the greater steric demands of the keto group relative to the aldehyde group and/or to the greater electrophilicity of the aldehyde. The former factor imposes significant constraints on the reaction trajectory required for productive ring closure, and thus the conformational dynamics of the acyclic keto form may affect k close. The practical implications of the data in Table 6 are that, under identical solution conditions, ketoses anomerize more slowly than aldoses having related structures, largely because of reduced rates of ring closure. The cis-1,2 effect is also observed in ketose 4 in which 3 and 4 are trans, with ring-opening occurring more rapidly for the β-furanose than for the α-furanose (Table 6). As observed in the some aldoses (Table 5), this effect is essentially abolished in ketose 8 in which 3 and 4 are cis. Anomerization rates are affected by solution ph, with acid-catalyzed, H 2 -catalyzed, and base-catalyzed regions (10b). The role of acid catalysis is apparent in k open values for 5-deoxy-L-lyxose and 5--methyl-D-lyxose measured at ph (24). Under acidic solution conditions, the rate constant for acid catalysis, kh3+, is determined from eq. [1], k obs = k H2 + k H3+ [H3 + ] eq [1] where k obs is the observed rate constant and k H2 is the rate constant for the water-catalyzed reaction. A plot of k obs vs [H 3 + ] gives a line with slope equal to k H3 +. These plots are shown in Figure 9 for the four anomers, from which the following k H3 + values for ring-opening were determined: for 5-deoxy-Llyxose, 79 ± 3 s -1 M -1 (α) and 184 ± 6 s -1 M -1 (β); for 5--methyl-D-lyxose, 27 ± 8

9 2 s -1 M -1 (α) and 69 ± 4 s -1 M -1 (β). In both aldopentoses, k H3 + values are ~2- fold larger for the β-anomer, that is, unlike behavior under water-catalyzed conditions (Table 5), the cis-1,2 effect is evident in acidic solution, attesting to the key role that solution conditions play in determining the relative reactivity of anomers. In addition, k H3 + values are ~ 3-fold larger in the 5-deoxy derivative, consistent with a -H 3 substituent being less electron-withdrawing than - H 2 H 3, thereby promoting protonation at either 1 or 4 during catalysis. H 3 H 3 H 2-deoxy-D-glycerotetrose (9) H 3 3-deoxy-DL-glycerotetrose (10) H 3 2 H 3--methyl-DLerythrose (11) H 3 2 H 3--methyl-DLthreose (12) H 3 3-deoxy-3,3-di--methyl- DL-glycero-tetrose (13) H 5--methyl-D-ribose (14) H 2-deoxy-5--methyl-Derythro-pentose (15) The effects of furanose ring deoxygenation and alkylation on the percentages of acyclic aldehyde form in aqueous solutions of aldofuranoses are illustrated by the data shown in Table 7. Deoxygenation at 2 increases the percentage of H 2 3 P H H 2 3 P H 2 3 P H H D-arabinose 5-phosphate (16) D-lyxose 5-phosphate (17) D-xylose 5-phosphate (18) acyclic aldehyde in solution (3 vs 9; 14 vs 15), whereas deoxygenation at 3 exerts only a minor effect (3 vs 10). Ring alkylation at 3 and 5 reduces the percentage of acyclic aldehyde form appreciably (3 vs 11 14). The latter shift towards cyclic forms is a manifestation of the Thorpe-Ingold effect (27 29) (gem-dialkyl effect) that has been attributed to enthalpic and entropic factors. The ratio, [hydrate]/[aldehyde], varies between within the group of compounds shown in Table 7; this ratio is influenced by steric factors in the acyclic hydrates, which vary with substitution pattern. For example, the very low percentage of hydrate form in solutions of 13 is caused, in part, by steric interactions between the two groups at 1 and the two H 3 groups at 3. The latter interactions are partly relieved in 11 and 12, and largely eliminated in 10, resulting in progressively higher percentages of hydrate in solution. 9

10 The effects of furanose ring deoxygenation and alkylation on anomerization kinetics are illustrated by the data in Table 8. The cis-1,2 effect (Scheme 8) on k open is maintained in 10 12, with 12 showing the greatest effect as expected, since 2 and 3 in 12 are trans. Interestingly, k open for the α- and β-furanoses of 15 are essentially identical, whereas those of 9 differ, with the a- furanose opening more rapidly than the β-furanose. This behavior may be attributed to different preferred ring conformations of the α-furanoses of 9 and 15, with the former favoring 2 E and the latter 3 E based on 3 J HH analysis (Scheme 9) (20). The former ring conformation orients both 1 and 3 quasiaxially, which presumably allows anchimeric assistance as shown in Scheme 8. In contrast, the preferred ring conformation of 15 orients 1 and 3 quasiequatorially, thus disallowing anchimeric assistance and rendering similar kopen values for both anomers. The apiofuranoses behave like 11 and 12 with respect to relative values of k open (30). Furanose ring alkylation enhances k close values significantly; for example, k close values range from s -1 for 3 and 5, but values from s -1 are observed for This behavior is a further manifestation of the Thorpe- Ingold effect in saccharides (27 29). Ring-opening rates constants for the pentose 5-phosphates (Table 9) depend on ring configuration, with ribo (6) showing the greatest reactivity. The effect of ph on k open is significant, with ~100-fold increases observed as ph is raised from 4.2 (mono-anion) to ph 7.5 (di-anion). A comparison of k open values for the four 5--methyl aldopentoses in Table 5, obtained at ph 4.0 and 60 o, to those found for the four pentose phosphates in Table 9, obtained at ph 4.2 and 40 o, provides indirect evidence for catalysis by phosphate. The H H HH H D-arabinuronic acid (19) H D-riburonic acid (21) D-lyxuronic acid (20) H H D-xyluronic acid (22) average value of the eight k open values in Table 5 is 0.21 ± 0.11 s -1 compared to the average of 0.44 ± 0.23 s -1 for the pentose phosphates, despite the 20 o lower temperature used in the pentose phosphate measurements. Unlike neutral (uncharged) furanoses, the cis- 1,2 effect (Scheme 8) is not observed in pentose phosphates; the α-anomer gives the larger k open at ph 4.2 and 7.5 in all ring configurations except lyxo (17). In 17, k open values for both anomers are essentially the same at either ph value. A possible mechanism that attempts to explain these observations invokes the phosphate group as a source of protonation of the ring oxygen in α-anomers, either directly (Scheme 10), or indirectly via H-bonding to a participating water 10

11 molecule. Steric hindrance between the cis-oriented phosphate group and 1 in β-anomers presumably weakens this mode of catalysis. Studies of k close in the pentose phosphates show that D-ribose 5P (6) is the most reactive, followed by 16 and 17/18 (18). Therefore, with respect to k open and k close, the ribo ring (6) is the most reactive pentose phosphate. This enhanced reactivity may have played a role in its evolutionary selection as a key sugar phosphate in biological metabolism (18). The different ring-opening behaviors of pentose phosphates compared to neutral furanoses stimulated interest in the anomerization kinetics of penturonic acids (Table 10) (31). At ph 1.5, where the carboxyl group is mostly protonated, the cis-1,2 effect (Scheme 8) is observed, that is, the relative values of k open mimic those found in neutral furanoses. A comparison of k open values for the four 5--methyl aldopentoses in Table 5, obtained at ph 4.0 and 60 o, to those found for the four penturonic acids in Table 10, obtained at ph 4.5 and 50 o, provides indirect evidence for catalysis by the group. The average value of the eight k open values in Table 5 is 0.21 ± 0.11 s -1 compared to 1.38 ± 0.83 s -1 for the penturonic acids, despite the 10 o lower temperature used in the penturonic acid measurements (the latter average could be enhanced somewhat by the slightly higher ph used for these measurements). At ph 1.5, intramolecular catalysis by the protonated carboxyl group and intermolecular catalysis by H + enhance k open values relative to those measured in neutral furanoses. A potential role for the group in intramolecular catalysis involves protonation of 4 (Scheme 11, A). The effects of this catalysis may be offset by the electron withdrawing character of the group. The latter factor would render 4 less prone to protonation and decrease k open. In contrast, k open values for at ph 4.5, where the carboxyl group is largely ionized, favor β-anomers (Table 10), that is, anomers in which 1 and the - group are cis. While anomerization at ph 4.5 is largely water-catalyzed for neutral furanoses, k open values may be enhanced in by an intramolecular mechanism involving deprotonation of 1 (Scheme 11, B). The electrondonating property of the - group (relative to ) may also promote protonation of the ring oxygen. 11

12 N-Acetyl Side-hain is-trans Isomerization in Aminosugars and onformational Properties of -Acetyl Side-chains Monosaccharides contain different types of substituents and side-chains that influence their chemical and biological properties (32). For example, the rotation of bonds involving hydroxyl groups influences H and bond lengths, which in turn influence the magnitudes and signs of NMR spinspin (scalar; J-coupling) coupling constants such as J HH, J H and J (33). These J-couplings, which are both abundant and redundant (i.e., multiple values report on the same conformational element) are valuable structural constraints to determine the conformational and dynamics properties of saccharides in solution (34 36). Examples of exocyclic groups such as hydroxymethyl ( H 2 ), - acetyl ( H 3 ), N-acetyl ( NH H 3 ), -phosphate monoesters ( P3-2 ), -sulfate monoesters ( S 3-1 ), N-sulfamides ( NH S 3-1 ), - lactoyl ( - HR H 3 ), and -glyceryl ( H H H 2 ) are shown in Scheme 12 (32). Here we limit discussion to N-acetyl and -acetyl side-chains to illustrate recent work that aims to better understand their structural properties. A. N-Formyl and N-Acetyl Side-chains. The conformational behaviors of N-acyl groups in saccharides are characterized by two factors: (1) rotation about the x NH bond θ 1 that attaches the group to x of the saccharide, and (2) cistrans isomerization of the amide bond θ 2 in the acyl substituent (Scheme 13). Signals arising from the cis and trans forms of the N-formyl and N-acetyl groups in methyl 2-[ 13 ]formamido-2-deoxy-d-glucopyranosides (23) (α) and (24) (β) and 2-[1-13 ]acetamido-2-deoxy-d-[2-13 ]glucopyranosides (25) (α) and (26) (β), respectively, can be observed by 1 H and/or 13 NMR, with signal assignments facilitated by selective 13 -enrichment at 2 of the H H H H " H 3 saccharide and/or the carbonyl carbon # NH NH H of the side-chain (37). For example, !! the 1 H NMR spectrum of methyl 2- H H! = 13 [ 13 ]formamido-2-deoxy-β-dglucopyranoside (24) at 600 MHz H contains two sets of formyl hydrogen! H! H H " H 3 signals at ~8.15 ppm (trans) and # NH NH H 3 ~7.93 ppm (cis), and two sets of 25 26!! anomeric hydrogen signals at 4.45 H 3 H 3 ppm (trans) and 4.42 ppm (cis) (Figure 10). The former are split by the one-bond 1 J H of Hz (cis) and Hz (trans) that, along with their characteristic 1 H chemical shifts, confirms their identity. In 24, ~74% of the amide bond exists in the trans 12

13 configuration and ~26% in the cis configuration at 42 o (K trans/cis = 2.84 ± 0.02) in aqueous ( 2 H 2 ) solution (37). By comparison, ~81% trans and ~19% cis are observed in 23 (K trans/cis = 4.22 ± 0.03), showing that anomeric configuration affects the trans/cis ratio. At temperatures of o, aqueous solutions of 24 consistently contain more cis isomer than do those of 23 (K trans/cis ranges from for 23, while K trans/cis ranges from for 24), with increasing temperature increasing the percentage of cis form in solution in both anomers (Figure 11A). The detection of cis forms in aqueous solutions of 25 and 26 is challenging because significantly less of this form is present compared to aqueous solutions of 23 and 24. By incorporating selective 13 -enrichment (~99 atom-% 13 ) at both 2 and the carbonyl carbon of the side-chain, not only is the detection of the labeled 2 and signals enhanced by ~100 fold, but J-coupling between the labeled carbons can be measured and used to confirm the assignments (assuming a non-zero value of the J-coupling). Note that 25 and 26 lack a characteristic 1 H signal in the side-chain, unlike 23 and 24, thus favoring 13 NMR as the method of analysis. The carbonyl carbon region of the 13 { 1 H} NMR spectrum of 26 at 22 o contains a strong (labeled) signal at ~177 ppm and a weak (labeled) signal at ~180 ppm (Figure 12A) (37). Both signals are split into doublets by 1.0 Hz and 0.8 Hz, respectively. The 2 region of the same spectrum contains a strong (labeled) signal at ~58 ppm and a weak (labeled) signal at ~63 ppm, and both are split into doublets by 1.0 Hz and 0.8 Hz, respectively (Figure 12B). The fact that identical signal splittings are observed for the paired weak and 2 signals, and the paired strong and 2 signals, confirms their assignments to the cis and trans forms of 25, respectively, with the splitting attributed to 2 J 2,. K trans/cis values for 25 range from over the temperature range o, and values of were observed for 26 over the same temperature range (Figure 11B) (37). Thus, while significantly less cis form is observed in aqueous solutions of 25 and 26 compared to 23 and 24 (the latter have K trans/cis values of 2 4; see above), solutions of β-anomer 26 consistently contain more cis isomer than do those of α-anomer 25 at any given temperature, thus mimicking the behavior of the N-formyl compounds 23 and 24. van t Hoff plots of the data shown in Figure 11 give slightly negative values for ΔH o (~ 1 3 kcal/mol) and ΔS o ( cal/k/mol) for the conversion, cis amide trans amide, showing that the process is enthalpically favored but entropically disfavored. onformation about θ 1 in is believed to favor structures in which H2 and the NH hydrogen are antiperiplanar (θ 1 = 180 o ) based on the magnitudes of 3 J H2,NH values (Scheme 14) (38,39). However, geometries in which H2 and NH are eclipsed (θ 1 = 0 o ) give 3 J H2,NH values similar in magnitude to those in 13

14 anti geometries (Figure 13) (40), so conformational conclusions on θ 1 based solely on a single 3 J H2,NH are not unequivocal. Efforts to provide more definitive assessments of θ 1 await the application of redundant J-couplings such as 3 J 1,NH, 3 J 3,NH, 3 J H2,, 3 J 1, and 3 J 3, that have been parameterized recently using density functional theory (DFT) (40). The observation of distinct signals arising from the cis and trans forms of implies slow exchange on the NMR time-scale, rendering the system amenable to study by saturation-transfer (19) to measure the first-order rate constants, k cis trans and k trans cis. As described above in studies of anomerization kinetics (Figure 7), selective 13 saturation of the carbonyl signal of the cis form gives k trans cis while saturation of the carbonyl signal of the trans form gives k cis trans ; saturation of the 2 signals could also be performed since they, like the carbonyl carbons, are reasonably well resolved at 150 MHz (Figure 12) (37). This application is shown in Figure 14 for 2-[1-13 ]acetamido-2-deoxy-α-d-[2-13 ]glucopyranoside (25), which shows the loss in carbonyl carbon signal intensity of the trans form with increasing saturation time of the carbonyl carbon in the cis form at different temperatures. Linearizing the data, as discussed in Figure 7, gives k trans cis values at each temperature. Side-chain cis-trans isomerization (TI) rate constants in determined by this method are summarized in Figure 15. Data show that TI kinetics depends on anomeric configuration, with β-anomers more kinetically favored than α-anomers in both the N-formyl and N-acetyl compounds. Within the series 23 26, βglcnac structure 26 is most reactive, exhibiting the largest k cis trans and k trans cis values at any given temperature. Energies of activation range from and kcal/mol for the cis trans and trans cis reactions, respectively (37). The biological implications of side-chain TI remain to be established, but these data show that TI equilibria and kinetics are both affected by side-chain structure (N-formyl vs N-acetyl) and anomeric configuration when the N-acyl side-chain is an equatorial orientation and when appended to 2 of an aldohexopyranosyl ring. It remains to be determined how these behaviors are influenced by other local environments and structural contexts (e.g., site of substitution; axial vs equatorial orientation; presence in larger oligosaccharides), and whether TI influences the binding of N-acylated saccharide substrates to biological receptors. B. -Acetyl Side-hains. -Acetyl side-chains resemble N-acetyl sidechains in that analogous θ 1 and θ 2 torsion angles characterize their conformations (Scheme 15). Recent work has shown (35) that the conformational behavior of θ 1 can be investigated using redundant NMR J- couplings and circular statistics. For example, in the mono--acetylated compounds 27 29, selective 13 -labeling incorporated into the -acetyl side- 14

15 chain allows the measurement of two or three J values (for example, in 28: 3 J 2,, 3 J 4,, 2 J 3, ) and simplifies the measurement of one or two J H values (for example, in 28: 3 J H3, ). These J-couplings are sensitive to θ 1, and DFT calculations (41) on model structures give parameterized equations that relate each to θ 1. Pertinent equations derived for the 3--acetylated compounds 28α/β take the following forms (35): 3 J H3, = cos (θ) sin (θ) cos (2θ) sin (2θ) 0.04 cos (3θ) sin (3θ) rms 0.22 Hz eq. [2] 3 J 2, = cos (θ) 0.73 sin (θ) 0.75 cos (2θ) 1.22 sin (2θ) 0.14 cos (3θ) sin (3θ) rms 0.13 Hz eq. [3] 3 J 4, = cos (θ) 0.74 sin (θ) 0.68 cos (2θ) 1.21 sin (2θ) 0.14 cos (3θ) sin (3θ) rms 0.04 Hz eq. [4] 2 J 3, = cos (θ) 0.14 sin (θ) cos (2θ) sin (2θ) 0.29 cos (3θ) 0.03 sin (3θ) rms 0.10 Hz eq. [5] Similar equations were derived for compounds 27α/β and 29α/β (35). These equations and the experimental J-couplings measured from 1 H and 13 { 1 H} NMR spectra were treated with a circular statistics package, MA AT, in which different 2-parameter continuous circular probability distributions were used to model θ 1 in This modeling gave mean values of θ 1 and circular standard deviations (SD) for each compound. The mean identifies the most abundant θ 1 torsion angle in aqueous solution, and the SD provides a measure of the librational disorder about the mean, analogous to the order parameter S 2 derived from NMR spin-relaxation measurements (42). This treatment is possible because four different J-couplings are available that display different functional dependencies on θ 1 (redundancy). The results of this data analysis are shown in Figure 16. The position of the -acetyl side-chain on the aldohexopyranosyl ring affects the mean value of θ 1 and the SD. The preferred conformation of the -acetyl side-chain in 28α/β is shown in Scheme 16. In the trans configuration of the ester (θ 2 ) shown, the = bond of the ester eclipses the 3 H3 bond in the preferred geometry about θ 1. This situation contrasts with the behavior of θ 1 in 29α/β, where the carbonyl carbon of the side-chain is roughly gauche to both H6R and H6S, and the = bond bisects the H6R 6 H6S bond angle in the trans configuration of the ester (Scheme 17). The SD is appreciably greater for 29α/β than for 28α/β, which indicates greater disorder about θ 1 for -acetyl groups appended to the primary alcoholic carbon. The 15

16 results of aqueous molecular dynamics simulations of are in good agreement with these experimental models (35), which testifies to the reliability of the method used to fit the redundant J-couplings and provides important experimental validation of the MD results. Future applications of this type of J- coupling analysis are anticipated in structural studies of other types of saccharide side-chains, saccharide rings, and -glycosidic linkages in oligosaccharides. Stereospecific arbon-skeleton Rearrangements in Saccharides: 1 2 Transposition Reactions A. Molybdate-atalyzed 2-Epimerization of Aldoses. In 1973, Bilik and coworkers reported that sodium molybdate catalyzes the 2-epimerization of aldoses in aqueous solution (43), and proposed a mechanism involving hydrogen shift to explain the reaction stereochemistry (44). However, unexpectedly, when the reaction was conducted with [1-13 ]aldoses as reactants, it was found that 1 2 transposition accompanies 2-epimerization (45). This remarkable 2Mo 3 + H 2 Mo 2 7 H 2 (bimolybdate) eq. [6] Mo 2 7 H H 14 7 [Mo H 10 7 ] H 3 + eq. [7] skeletal rearrangement is believed to involve initial complexation of bimolybdate with the hydrate (1,1-gem-diol) form of the aldose (Scheme 18). The reaction apparently occurs in two steps described by eqs. [6] and [7]. In the first step, monomeric molybdate dimerizes to form the bimolybdate species, Mo 2 7 H 2. The latter subsequently binds with the aldose acyclic hydrate to give a negatively charged bimolybdate-aldose complex. This complex is catalytically active, yielding a putative transition state containing partial covalent bonds between 1 2, 2 3 and 1 3. The ratio of starting aldose and its 2- epimeric product observed after equilibration is determined by their relative thermodynamic stabilities, since the reaction is freely reversible. In most cases, equilibration is reached in ~3 h at ~85 o, and few if any by-products are observed. Studies have shown that hydroxyl groups at 1, 2 and 3 of the aldose reactant are required for molybdate-catalyzed 2-epimerization (ME), while an group at 4 is not required but increases the reaction rate and reduces the formation of by-products (45). From a practical standpoint, ME provides a powerful complement to cyanohydrin reduction (R) reactions in the synthesis of stable isotopically labeled saccharides (Scheme 19) (46). Stable isotopes ( 13, 2 H) are introduced at 1 of aldoses by reacting K 13 N (a relatively cheap, commercially available labeled precursor) with a starting aldose electrophile under solution conditions that stabilize the initially formed α-hydroxynitriles (cyanohydrins). The latter are hydrogenolyzed using a heterogeneous metal 16

17 catalyst (typically Pd/BaS 4 ) and H 2 gas to give a pair of 2-epimeric [1-13 ]aldoses in >80% yield, each containing one more carbon than the starting aldose (chain extension). The 2-epimeric products are purified by chromatography (47). If the hydrogenolysis is conducted with 2 H 2 gas in 2 H 2 solvent, the product [1-13 ]aldoses will also contain 2 H at 1 (Scheme 19) (46b,48). These [1-13 ]-labeled aldoses can then be subjected to ME to transfer the 13 and/or 2 H to 2 of the 2-epimeric products. R and ME reactions have been effectively integrated into synthetic reaction pathways to provide access a wide range of selectively, multiply and/or uniformly labeled saccharides and their derivatives (e.g., nucleosides) (Scheme 20) (49). Inspection of the bimolybdate complexes shown in Scheme 18 shows that the space enveloping H1 of the aldose reactant is largely unobstructed, such that replacement with a larger R-group (to give a 2-ketose reactant) should be possible without affecting reactivity. This expectation is realized in practice. Studies show that ME interconverts 2-ketoses with 2--substituted aldoses with high stereospecificity, providing a convenient route to branched-chain aldoses (50 52). Two examples of this application are shown in Scheme 21. Reaction B demonstrates the high tolerance of the reaction to relatively bulky R- groups appended to 2 of the 2-ketose reactant. B. Molybdate-atalyzed onversion of sones to Aldonates. The 1 2 transposition that accompanies ME can be informally viewed as an internal redox process wherein the oxidation states of 1 and 2 are exchanged during the transformation. This mental construct for the reaction leads to the expectation that aldonates should be produced when 1,2-dicarbonyl sugars such as D-arabino-hexos-2-ulose (D-glucosone) (30) are used as reactants. Recent unpublished work from this laboratory indicates that the reaction of [1-13 ]30 with molybdate at 90 o gives D-[2-13 ]gluconate (31) and D-[2-13 ]mannonate (32) in a 85/15 ratio (Scheme 22). zone 30 presumably binds bimolybdate in its dihydrate form to satisfy the hydroxyl group requirements discussed above. By analogy to the complexes that form with aldoses (Scheme 18), two different complexes with 30 are possible. ne complex gives D-[2-13 ]31, and the other D-[2-13 ]32. Unlike the aldose reactions, however, the reaction with 30 is not reversible; the aldonates apparently cannot be converted to the osone, and an aldonate cannot be used to generate its 2-epimer. The negatively charged aldonates do not form bimolybdate complexes, presumably because of electrostatic repulsion (both partners are negatively charged). Since the reaction is irreversible, the ratio of 2-epimeric aldonates is not determined by their relative stabilities, but rather by the relative stabilities of the two bimolybdate complexes (binding phase) and/or the relative catalytic efficiencies of the two complexes (catalytic phase). The rates of release of aldonate products from their complexes are assumed to be identical. It is interesting to note that, by analogy to the 2-ketose reactants shown in Scheme 21, 2,3-17

18 dicarbonyl sugars in their acyclic dihydrate forms should also form productive bimolydate complexes, leading to branched-chain aldonates (Scheme 23). This potential transformation, however, remains to be tested in the laboratory.. Phosphate-Mediated onversion of sones to 2-Ketoses. As discussed above, osones are reactive substrates in molybdate-catalyzed reactions where 1 2 transposition occurs to give a pair of 2-epimeric aldonates. Recent work has shown, however, that this type of transposition in osones is not confined to molybdate-mediated reactions. Prior work has shown that D-glucosone (30) undergoes spontaneous degradation in dilute phosphate buffer at ph 7.4 and 37 o to give D-ribulose (33) (Scheme 24) (53). Recent NMR studies conducted with D-[2-13 ]glucosone (30) confirm this behavior, with unlabeled formate and D-[1-13 ]ribulose observed as the major degradation products (54). The reaction pathway presumably involves the formation of 2,3-enediol and 1,3-dicarbonyl intermediates, the latter undergoing attack at 1 by - with subsequent 1 2 bond cleavage and protonation to give the 2-ketopentose and formate. Additional studies of this degradation pathway using other 13 - isotopomers of 30, however, indicated that the pathway shown in Scheme 24 is incomplete, and that, surprisingly, 1 2 transposition also occurs during degradation. Initial indications of this transposition were found in the reaction shown in Scheme 24 in that a small amount of [ 13 ]formate was observed by 13 NMR in the reaction mixture even though the mechanism shown does not explain its formation. A more definitive experiment was conducted in which D- [1, ]glucosone (30) was used as the substrate for degradation. Under these reaction conditions, the detection of D-[1, ]ribulose (33) in the reaction mixture would constitute clear evidence that 1 2 transposition occurred during degradation. The 13 { 1 H} NMR spectrum of the products of this reaction is shown in Figure 17. These data show that most of the D-[1, ]30 degrades as shown in Scheme 24, giving D-[2-13 ]33 and H 13 - as the primary end-products. However, closer inspection of the 2 signals arising from D-[2-13 ]33 reveals weak satellites on each signal. The upfield region of the spectrum contains the 1 signals arising from each of the three forms of D- [1, ]33 present in solution (keto and two furanose forms). Each of these signals is split by one-bond J-couplings that are identical to those measured in authentic D-ribulose (17) and to the splittings measured from the 2 satellites (αf, 51.8 Hz; βf = 51.3 Hz; keto, 41.5 Hz). These and other lines of evidence indicate that during the degradation of 30, most of the carbon (~90%) flows down the pathway involving direct 1 2 bond cleavage to give 33 and formate. However, approximately 10% of 30 undergoes 1 1 transposition during degradation (Scheme 25). Potential mechanisms for this transposition involve inorganic phosphate as a catalyst in the initial formation of 18

19 a 1,3-dicarbonyl cyclic phosphate intermediate (Scheme 26) and subsequently as a tether during 1 2 transposition (54). It is noteworthy that arsenate also appears to substitute for Pi in these reactions (54). The preceding discussion serves to illustrate that 1 2 transposition may be a more common skeletal rearrangement in saccharides than currently appreciated. These rearrangements are remarkable, but their detection requires the use of 13 -labeling in conjunction with NMR and other analytical methods to trace the fates of individual carbons during the reaction. In the original studies of molybdate-catalyzed 2-epimerization of aldoses (43,44), and of glucosone degradation (53), 13 -labeling was not employed, leading to erroneous or incomplete mechanisms for these reactions. It is interesting to note that the transfer of two-carbon fragments is a common occurrence in saccharide metabolism. For example, thiamine pyrophosphate promotes reactions catalyzed by the pentose phosphate pathway enzyme, transketolase, wherein the coenzyme functions as a carrier of a negatively charged acylium anion formed from the 1 2 fragment of a 2-ketose, with the inherently unstable anion resonance-stabilized when covalently attached to the coenzyme (55 57). In principle, this carrier might also enable 1 2 exchange during the two-carbon exchange as shown in Scheme 27, although, like the glucosone degradation pathway, only a small percentage of the catalytic cycles may follow this pathway. Studies with 13 -labeled substrates would be needed to test this possibility. Molybdenum-catalyzed skeletal rearrangements mimic enzyme-catalyzed reactions in their simplicity and high stereospecificity. Whether enzymes have evolved to exploit the inherent catalytic properties of molybdate in this fashion remains to be determined, as is the potential role of molybdate in chemical evolution. ther elements of the Periodic Table that lie in the vicinity of Mo have not shown an ability to catalyze 1 2 transposition in aldoses. The one element that has not yet been tested is technetium, whose oxides have solution properties similar to those of molybdate (58), but whose rarity and radioactivity thus far have discouraged studies of its reactivity. oncluding Remarks As discussed in the foregoing paragraphs, studies of the structures and reactivities of saccharides are enabled and/or strengthened when isotopically labeled substrates, especially 13 -labeled, are used to increase the information content of laboratory experiments. We have shown how these isotopes can be used to detect and quantify the cyclic and acyclic forms of reducing saccharides in solution and to investigate relationships between saccharide structure, conformation and the kinetics of tautomer exchange. With the use of 13 - labeled compounds, redundant NMR spin-couplings sensitive to the same 19