Divergent Behavior of Glycosylated Threonine and Serine Derivatives in Solid Phase Peptide Synthesis

Abstract

Solid phase peptide coupling of glycosylated threonine derivatives was systematically evaluated. In contrast to glycosylated serine derivatives which are highly prone to epimerization, glycosylated threonine derivatives produce only negligible amounts of epimerization. Under forcing conditions, glycosylated threonine analogs undergo β-elimination, rather than epimerization. Mechanistic studies and molecular modeling were used to understand the origin of the differences in reactivity.

Glycopeptides have been used extensively in basic and clinical research.^1–5 Glycopeptides can serve as structurally-defined models of complex glycoproteins, and several are in development as therapeutic agents. For example, a glycopeptide isolated from urine has antiproliferative activity,⁶ and glycosylation of an opioid peptide enhances its in vivo performance¹. Glycopeptides are also being developed as vaccines for cancer and HIV, and several have progressed into clinical trials.^2,7–9

Chemical synthesis is an important tool for obtaining structurally-defined glycopeptides, since homogenous glycopeptides are difficult to obtain from natural sources.^1,10 Chemical synthesis allows precise control of the chemical structure and access to large quantities of material. Moreover, one can obtain unnatural derivatives for studying structure-activity relationships.

The synthesis of glycopeptides can be significantly more challenging than standard peptides. Glycopeptides are commonly synthesized by coupling protected glyco-amino acids to a growing peptide chain via solid phase glycopeptide synthesis.¹ These couplings are often slow and inefficient. In addition, we recently demonstrated that many commonly used peptide coupling conditions produce high levels of epimerization for glyco-amino acids.¹¹ In fact, epimerization could produce as high as 80% of the unnatural epimer. At present, however, the factors that influence efficiency and epimerization are not well understood, and additional studies are needed to develop efficient and general peptide coupling conditions for glycopeptides.

Our previous study focused on coupling glycosylated serine derivatives to a growing peptide chain.¹¹ In this study, we evaluated efficiency and epimerization when coupling glycosylated threonine derivatives. Non-glycosylated threonine derivatives are known to react more slowly than the corresponding serine derivatives in solid phase peptide coupling reactions.¹² Since a slower coupling rate would allow more time for epimerization, it was possible that glycosylated threonine residues could be even more prone to side reactions.

We carried out a systematic comparison by coupling glyco-amino acids to a growing peptide chain, Pro-Gly-Hex (Hex: 6-aminohexanoic acid), on solid phase. Two commonly used glyco-amino acids, Fmoc-Thr(Ac₃GalNAcα)-OH and Fmoc-Thr(Ac₃GlcNAcβ)-OH, were selected for this study, along with the non-glycosylated counterpart as shown in Scheme 1. The desired glycopeptides, the unnatural epimers, and the truncated (uncoupled) peptides were chemically synthesized and used as standards for the development of HPLC assays to evaluate efficiency and epimerization (see Supporting Information). We selected seven coupling conditions that had previously been used to prepare glycopeptides (see Table 1 for references). They contain the most widely used activation agents, bases, and solvents, as well as variations in reaction times and equivalents of bases and reagents.

Scheme 1.

Coupling Threonine and Glycosylated Threonine Derivatives to ProGlyHex Resin

Table 1.

Summary of Yields, Epimerization and β-Eliminaion for Various Peptide Coupling Conditions

			Fmoc-Thr(R)-OH, where R =
no.	coupling conditionsa		Trt	Ac3GalNAcα	Ac3GalNAcα
1	AAs: 1.5 equiv	yields (%)b	98.3	98.2	98.8
	HATU/HOAt: 1.2/1.2 equiv	epimerization (%)b	<0.2	0.4	0.3
	NMM: 2.4 equiv	β-elimination (%)b	<0.2	<0.2	0.5
	0/8 h in DMF13
2	AAs: 4.4 equiv	yields (%)b	99.9	84.1	40.4
	HATU: 4.4 equiv	epimerization (%)b	1.5	0.9	0.8
	NMM: 8.8 equiv	β-elimination (%)b	<0.2	<0.2	59.4
	3/12 h in NMP14
3	AAs: 2 equiv	yields (%)b	98.6	99.1	94.1
	HBTU/HOBt: 4.5/4.5 equiv	epimerization (%)b	0.3	0.6	<0.2
	DIEA: 9 equiv	β-elimination (%)b	<0.2	0.4	5.0
	0/1 h in NMP15
4	AAs: 3 equiv:	yields (%)b	99.8	99.2	97.9
	DCC/HOBt: 18/18 equiv	epimerization (%)b	0.2	0. 7	0.3
	1/64 h in NMP16	β-elimination (%)b	<0.2	0.7	1.9
5	AAs: 1.5 equiv	yields (%)b	92.5	99.3	95.3
	HBTU/HOBt: 1.5/1.5 equiv	epimerization (%)b	<0.2	0.5	0.6
	DIEA: 1.5 equiv	β-elimination (%)b	<0.2	0.3	2.2
	0/0.33 h in DMF 17
6	AAs: 2.5 equiv	yields (%)b	98.3	99.0	98.6
	BOP/HOBt: 2.5/2.5 equiv	epimerization (%)b	0.2	0.4	0.3
	DIEA: 2.5 equiv	β-elimination (%)b	<0.2	0.3	1.1
	0/4 h in DMF18
7	AAs: 2.0 equiv	yields (%)b	99.1	99.0	98.2
	HATU/HOAt: 2.0/2.0 equiv	epimerization (%)b	<0.2	0.4	0.5
	TMP: 2.0 equiv	β-elimination (%)b	<0.2	0.3	0.3
	0/2 h in DMF

^aGlyco-amino acids were coupled to Pro-Gly-Hex resin. Preincubation time and coupling times are listed as x/y h (e.g. 3/12 h = 3 h pre-incubation followed by 12 h reaction time).

^bAll data have an error of less than 0.3%. The yield refers to the percentage of D+L products relative to the total peptide (D+L products, β-elimination product, and truncated peptide).

Abbreviations are as follows: 2-(7-Aza-1H-benzotriazol-1-yl)-N,N,N’,N’-tetramethyl aminium hexafluorophosphate (HATU), 2-(1H-benzotriazol-1-yl)-N,N,N’,N’-tetramethylaminium hexafluorophosphate (HBTU), N,N'-dicyclohexylcarbodiimide (DCC), benzotriazol-1-yloxy)tripyrrolidinophosphonium hexafluorophosphate (BOP), N,N-diisopropylethylamine (DIEA), N-methylmorpholine (NMM), 2,4,6-trimethylpyridine (TMP), dimethylformamide (DMF) and N-methyl-2-pyrrolidone (NMP).

The results are summarized in Table 1. Little or no truncated peptide was observed in any of the reactions. Moreover, only negligible epimerization (≤1.5%) was detected for all of the coupling reactions. This result was in stark contrast to our previous results with glycosylated serine analogs. For example, coupling of Fmoc-Ser(Ac₃GalNAcα)-OH using condition 2 produced approximately 70% epimerization, whereas less than 1% epimerization was observed for glycosylated threonine derivatives. Under several conditions, however, an unknown peptide was produced accounting for up to 60% of the mass. Further analysis revealed the side product to be Fmoc-dehydroaminobutyrate-Pro-Gly-Hex-OH, which results from β-elimination of the glycan moiety (see Scheme 2, route b). Based on this result, we assessed β-elimination for all three amino acids and all seven conditions. β-Elimination primarily occurred for Fmoc-Thr(Ac₃GlcNAcβ)-OH.

Scheme 2.

Potential Routes for Epimerization (a) and β-Elimination (b)

The difference in behavior of glycosylated serine and threonine derivatives in peptide coupling reactions could be due to many factors. First, we anticipated that glycosylated threonine derivatives would react more slowly than serine derivatives; however, the relative rates had not been previously measured. If glycosylated threonine derivatives actually react faster, then there would be less time for epimerization. To test this possibility, we measured the relative coupling rates of other amino acids against threonine in a competition assay. Fmoc-Thr(Trt)-OH was mixed with the competing amino acid at an equivalent amount. The mixture was activated and then captured with the resin. The relative amounts of each peptide produced after 5 minutes was measured via an HPLC assay. As summarized in Table 2, the overall coupling rates of threonine-based amino acids were slower than serine derivatives. Therefore, a faster rate of peptide coupling does not account for the large difference in epimerization levels.

Table 2.

Relative Overall Coupling Rate Versus Threonine

amino acida	relative reaction rate AA/Thr
Fmoc-Thr(Trt)-OH	1
Fmoc-Ser(Trt)-OH	3.33 ± 0.53
Fmoc-Thr(Ac3GalNAcα)-OH	0.70 ± 0.03
Fmoc-D-Thr(Ac3GalNAcα)-OH	0.90 ± 0.02
Fmoc-Thr(Ac3GlcNAcβ)-OH	2.62 ± 0.06
Fmoc-D-Thr(Ac3GlcNAcβ)-OH	1.01 ± 0.04

^aAll reactions were carried out by mixing Fmoc-Thr(Trt)-OH (2 equiv) and the listed amino acid (2 equiv) with HATU (4 equiv), HOAt (4 equiv) and TMP(4 equiv) in DMF. The assays were conducted in triplicate.

A second possible explanation is that for glycosylated threonine derivatives, the equilibrium between the natural and unnatural epimer lies heavily in favor of the natural epimer. If this was the case, epimerization could be occurring in the reaction, but it would not produce significant amounts of the unnatural epimer. To test this hypothesis, we pre-incubated the unnatural epimers of Fmoc-D-Thr(Trt)-OH and D-glyco-amino acids with HATU/NMM for three hours and then coupled them to ProGlyHex resin (condition 2). The long pre-incubation step permits epimerization. If the natural epimer is energetically favored and the rate of epimerization is sufficient, extensive amounts of the natural epimer would be formed in these reactions. In actuality, very little epimerization was observed in these reactions. Fmoc-D-Thr(Trt)-OH gave approximately 4% epimerization, and Fmoc-D-Thr(Ac₃GalNAcα)-OH gave <0.2% epimerization. Since epimerization was not observed for either the natural or the unnatural epimer, we concluded that the rate of epimerization was too slow to reach equilibrium. Interestingly, Fmoc-D-Thr(Ac₃GlcNAcβ)-OH produced 92.4% of the β-elimination product under these conditions. Epimerization may be occurring for this substrate, but rapid β-elimination prevents measurement of the equilibrium.

To better evaluate the equilibria, we next attempted to increase the rate of epimerization by varying the type of base and increasing the number of equivalents of base. As shown in Figure 1, we pre-incubated Fmoc-Thr(Ac₃GlcNAcβ)-OH (a) or Fmoc-Thr(Ac₃GalNAcα)-OH, and HATU for three hours with different equivalents of NMM(a) or DIEA(b), respectively. Even under forcing conditions, little D-epimer was detected. Instead the β-eliminated peptide was observed as the major product for both glyco-amino acids. For example, incubation of Fmoc-Thr(Ac₃GlcNAcβ)-OH with twelve equivalents of NMM produced about 90% β-eliminated product. Although we were unable to measure the equilibrium ratio of epimers, it is clear that glycosylated serine and glycosylated threonine analogs produce different side products in these coupling reactions.

Figure 1.

Epimerization and β-elimination as a function of base. a) Fmoc-Thr(Ac₃GlcNAcβ)-OH; b) Fmoc-Thr(Ac₃GalNAcα)-OH. Glyco-amino acids were coupled using the following coupling agents and ratios: AAs/HATU/base = 1/1/x. Each amino acid was pre-incubated for three hours and then the extent of epimerization and β-elimination were measured as described in supporting information.

In our previous study on peptide couplings of glycosylated serine analogs, we found that the mild base, TMP, provides high yields with little or no epimerization. To further test the utility of this base, we evaluated the use of higher equivalents of TMP, analogous to the studies above with NMM and DIEA. Remarkably, epimerization and β-elimination were less than 5% with up to 8 equivalents of TMP for all three amino acids. Therefore, TMP produces at least a 10 fold lower level of side products as compared to NMM and DIEA. Based on these results and our previous results, we consider condition 7 (2 equivalents of glyco-amino acid, 2 equivalents of HATU and HOAt, and 2 equivalents of TMP) to be the best conditions we have examined for solid phase peptide couplings involving glycosylated amino acids. Other mild peptide coupling conditions may also be useful for making glycopeptides.¹⁹

While the above experiments provided interesting insight into the reaction pathways, it was still not clear why glyco-threonine analogs have much lower levels of epimerization than glyco-serine analogs. Previous NMR studies have shown that glyco-serine and glyco-threonine can have significantly different preferred conformations in water,²⁰ but little information was available for protected derivatives in organic solvents. Therefore, we carried out molecular modeling to determine if differences in conformational preferences might contribute to differences in reactivity. To mimic the reaction solvent, a distance dependent dielectric constant of 38 was used. Since epimerization and β-elimination are thought to proceed through an oxazolone intermediate²¹ (see Scheme 2), the corresponding oxazolone intermediates of Fmoc-Thr(Ac₃GalNAcα)-OH and Fmoc-Ser(Ac₃GalNAcα)-OH were modeled using methods reported previously.¹¹

The preferred conformations of the glyco-serine and glyco-threonine oxazolones showed distinct differences (see Figure 2). In particular, the value of Hα-Cα-Cβ-Oβ dihedral angle for the threonine derivative was approximately 60° (Figure 2, C), placing the β-methyl and the O-GalNAc groups gauche to the α-hydrogen. In contrast, the value of Hα-Cα-Cβ-Oβ dihedral angle for the serine derivative was approximately 180° (Figure 2, C), placing the O-GalNAc group anti to the α-hydrogen. The difference has two key effects. First, the alpha hydrogen of glyco-threonine analogs is more sterically shielded than that of glyco-serine derivatives, which could significantly hinder abstraction of the alpha hydrogen. Second, overlap of C_β-O σ*-antiboding orbital with the C_α-H σ bond in the glyco-serine derivative, but not the threonine derivative, could facilitate abstraction of this hydrogen. Therefore, the combination of steric and stereoelectronic effects may contribute to the low epimerization of glyco-threonine analogs compared with the glyco-serine ones. When abstraction of the α-hydrogen is possible in a threonine derivative, formation of a more highly substituted alkene relative to serine derivatives likely contributes to the preference for β-elimination over epimerization.

Figure 2.

Low-energy models of the oxazolone intermediates derived from (A, C) Fmoc-Ser(GalNAcα)-OH and (B, D) Fmoc-Thr(GalNAcα)-OH generated from the Conformational Search utility in Macromodel 2011. Panels (C) and (D) focus on the oxazolone ring and the Hα−Cα−Cβ−Oβ dihedral angle.

In summary, we demonstrate that serine and threonine based glyco-amino acids display significantly different behaviors in solid phase peptide coupling reactions. Glyco-threonine derivatives produce little or no epimerization, but occasionally give rise to β-elimination. In contrast, glyco-serine derivatives are highly prone to epimerization. The difference in reactivity is likely due to distinct differences in conformational preferences of the reactive intermediates. Fortunately, use of TMP as a base in the peptide coupling reactions significantly reduces both epimerization and β-elimination side reactions. Finally, our studies show that even small structural changes to glyco-amino acids, such as the presence or absence of a single methyl group, can have a dramatic effect on side reactions. Therefore, a more detailed understanding of the factors that contribute to epimerization and β-elimination will significantly improve our ability to synthesize glycopeptides with high efficiency and predictability.