Conformational changes associated with post-translational modifications of Pro(143) in Skp1 of Dictyostelium--a dipeptide model system.

Prolyl hydroxylation and subsequent glycosylation of the E3(SCF) ubiquitin ligase subunit Skp1 affects its conformation and its interaction with F-box proteins and, ultimately, O2-sensing in the organism. Taking a reductionist approach to understand the molecular basis for these effects, a series of end-capped Thr-Pro dipeptides was synthesized, tracking the sequential post-translational modifications that occur in the protein. The conformation of the pyrrolidine ring in each compound was gauged via coupling constants ((3)JHα,Hβ) and the electronegativity of the Cγ-substituents by chemical shifts ((13)C). The equilibrium between the cis-trans conformations about the central prolyl peptide bond was investigated by integration of signals corresponding to the two species in the (1)H NMR spectra over a range of temperatures. These studies revealed an increasing preference for the trans-conformation in the order Pro < Hyp < [α-(1,4)GlcNAc]Hyp. Rates for the forward and reverse reactions, determined by magnetization transfer experiments, demonstrated a reduced rate for the trans-to-cis conversion and a significant increase in the cis-to-trans conversion upon hydroxylation of the proline residue in the dipeptide. NOE experiments suggest that the Thr side chain pushes the sugar away from the pyrrolidine ring. These effects, which depended on the presence of the N-terminal Thr residue, offer a mechanism to explain altered properties of the corresponding full-length proteins.

Introduction

The process whereby proteins are polyubiquitinated and delivered to the 26S proteasome for degradation involves several steps and is fundamental to the maintenance of healthy levels of all proteins in eukaryotic organisms. A prominent really interesting new gene (RING)-type third order (E3) ubiquitin ligase, referred to as the Cullin-RING-Ligase-1 (CRL1) or the Skp1-Cullin-F-box (SCF) complex, consists of several subunits including the S-phase kinase-associated protein 1 (Skp1).[1] In Dictyostelium discoideum, the proline residue at position 143 of Skp1 is hydroxylated by a cytoplasmic prolyl hydroxylase (PhyA).[2] We have recently confirmed that the hydroxylation, by analogy to the hypoxia inducible factor α (HIFα) prolyl hydroxylases, delivers (2S,4R)-4-hydroxyproline (Hyp).[3] Hydroxylation of Skp1 is important for the proper oxygen-dependence of differentiation of the organism into a fruiting body to disperse spores.[4,5] Polypeptide N-acetylglucosaminyl-transferase 1 (Gnt1) transfers N-acetylglucosamine (GlcNAc) from uridine diphosphate (UDP) to Hyp143 of Skp1.[6,7] The installation of the first GlcNAc residue renders Skp1 subject to further glycosylation by the PgtA (β3GalT/α2FucT) and AgtA (αGalT) glycosyltransferases,[8−10] which modulate oxygen-dependence conferred by prolyl hydroxylation. The hypothesized terminal structure, illustrated in Figure 1, has been validated for all but the linkage of the terminal αGal.[10] Biophysical studies, viz. circular dichroism (CD) and small-angle X-ray scattering (SAXS), indicate that the αGlcNAc residue promotes α-helical secondary structure and increased order of the polypeptide.[11] These post-translational modifications occur in other unicellular eukaryotes, including Toxoplasma gondii.(8,12,13)

Figure 1

Glycosylated region of Skp1.

While a fundamental impact of glycosylation on protein structure and function has been widely appreciated for many years, the empirical basis for structural changes is less well understood. Previous studies have focused on the more ubiquitous O-glycosidic linkages to serine and threonine residues.[14−18]

Prior to hydroxylation, the pyrrolidine ring of the Pro143 residue is likely to adopt a Cγ-endo conformation (Scheme 1). X-ray crystallographic studies show that the Thr-Pro143 peptide bond favors the trans conformation (ω = 180°) when the Skp1 proteins from Saccharomyces cerevisiae or Arabidopsis thaliana are complexed to an F-box protein.[19,20] This cis–trans isomerism might be important in Skp1 folding, perhaps mediated by a Thr-Pro prolyl isomerase.[21] By analogy to recent findings in the orthologous human HIFα,[22,23] upon hydroxylation by a trans-4-prolylhydroxylase, there could be dramatic changes in the topology of this region of Skp1, viz. pyrrolidine ring-flip to the Cγ-exo conformation and a reinforced preference for a trans amide bond (ω = 180°), which may in turn influence processing by Gnt1 and F-box protein binding. With a GlcNAc residue installed, torsion angles about the glycosidic bond will contribute to recognition by other proteins, including F-box proteins and the β3GalT/α2FucT that introduces the next two monosaccharides.[12,24,25] These interrelated conformational effects may provide the trigger for α-helix formation in Skp1.

Scheme 1

Pyrrolidine Conformation

Hydroxyproline glycosides are common in plants; however, no examples have been reported to-date in mammalian systems.[26] In 2007, Schweizer and co-workers reported on the thermodynamics and kinetics of cis–trans isomerization of a family of proline derivatives, Ac-Pro*-NHMe where Pro* was l-proline (Pro), trans-4-hydroxy-l-proline (Hyp), trans-4-tert-butoxy-l-proline, α-(1,4)-galactosyl-trans-4-hydroxy-l-proline [α-(1,4)-Gal-Hyp], and β-(1,4)-galactosyl-trans-4-hydroxy-l-proline [β-(1,4-Gal)-Hyp].[27] The surprising result of this work was that, within experimental error, the equilibrium constants and rate constants were similar for all compounds. We report herein that for a dipeptide that emulates the Thr142-Pro143 sequence in Skp1 there are noteworthy changes that correlate with recent findings on the F-box binding capacity of the isoforms of Skp1.

Results and Discussion

We prepared dipeptides 1–4 (Scheme 2) that track the initial post-translational modifications of Skp1 and incorporate the preceding Thr residue. We end-capped the dipeptides to mimic the amide backbone of the protein and to eliminate electrostatic effects that arise from ionizable groups.[28,29] At the outset, we did not know the configuration at Cγ of the hydroxyproline residue and thus included cis-4-hydroxy-l-proline (hyp) in our repertoire of proline analogues. Specifically compounds 2 and 3 were synthesized and used to confirm the 4R configuration of the hydroxproline residue in Skp1.[3] In addition, Ac-[α-(1,4)GlcNAc]Hyp-NHMe (5) was prepared. This compound represents the GlcNAc analogue of Ac-[α-(1,4)Gal]-Hyp-NHMe described previously by Schweizer and co-workers.[27] Full details of the synthesis of the compounds depicted in Scheme 2 are provided in the Supporting Information. In the case of glycosides 4 and 5, glycosylation was performed with thioglycoside donor 6 using NIS/AgOTf activation[30,31] and an appropriate carbamate protected hydroxyproline ester.

Scheme 2

Dipeptides Synthesized

Analysis of the 1H NMR 3J coupling constants Jαβ1 and Jαβ2 (α, β etc. as defined for compound 1 in Scheme 2) gave an indication of the conformational preference of the pyrrolidine ring in each compound (Table 1). According to the literature, in a Cγ-exo ring pucker both J-values are in the specific limiting range of 7–11 Hz, sometimes even converging to produce an apparent triplet.[32,33] In the Cγ-endo ring pucker there are two quite different dihedral angles between Hα and the neighboring Hβ protons, giving rise to distinct coupling constants in the ranges 2–3 and 6–10 Hz. In the case of a twist conformation, dynamically averaged values of ∼7 Hz are anticipated. Thus, the pyrrolidine rings in compounds 2, 4, and 5 adopt a Cγ-exo conformation. The Jαβ values could not be determined for compound 1 due to overlap between Pro Hα and Thr Hβ signals. For dipeptide 3 containing hyp, the J-values do not fit either profile but are in agreement with those reported for Ac-hyp-OMe by Owens et al.[34] and may signify a slightly different conformation, viz. Cβ-exo, Cδ-exo, or related twist confromations.[35]

Table 1

J-Values and Chemical Shifts Derived from 1D NMR Spectra of Compounds 1–5

compound	Jαβ1, Jαβ2 (Hz)	δ Cγ (ppm)
Ac-Thr-Pro-NHMe (1)	nda	23.6b
Ac-Thr-Hyp-NHMe (2)	9.7, 7.8	69.7c
Ac-Thr-hyp-NHMe (3)	9.2, 4.6	69.5 (67.5)d
Ac-Thr-[α-(1,4)GlcNAc]Hyp-NHMe (4)	9.7, 8.1	75.7c
Ac-[α-(1,4)GlcNAc]-Hyp-NHMe (5)	8.5, 8.5	74.8 (74.0)d

Not determined due to overlap between Pro Hα and Thr Hβ.

While two species are evident, the two Cα signals appear to be coincident.

On the time scale of the 13C NMR experiment a single, averaged species was observed.

The value in parentheses represents a signal due to the minor conformation.

The electron-withdrawing nature and R-configuration of substituents at Cγ are attributed with favoring the Cγ-exo conformation of the pyrrolidine ring and inter alia the trans conformation of the preceding prolyl peptide bond. The 13C NMR chemical shift for Cγ has been used to assess the strength of the electron-withdrawing substituents; our results appear in Table 1. Hydroxylation at Cγ leads to a significant downfield shift from 23.6 to 69.7 ppm. Glycosylation further deshields the Cγ nucleus from the magnetic field, leading to an additional 6 ppm shift downfield upon glycosylation of hydroxyproline, in accord with the observations of Schweizer and co-workers upon α-d-galactosylation.[27] The J-values and Cγ chemical shifts reported in Table 1 are very similar to those reported for the Ac-Pro*-NHMe series,[27] implying that the additional Thr residue does not affect pyrrolidine ring conformation.

The thermodynamics of the cis → trans isomerization were studied by NMR as described previously.[36,37]K values were determined at 5° temperature intervals from 25 to 85 °C, by integration of well-resolved signals for protons associated with the cis and trans conformations. The derived van’t Hoff plots are depicted in Figure 2. All peptides demonstrated a decrease in K with an increase in temperature, and the linear nature of the plots indicates that ΔH° and ΔS° are independent of temperature.

Figure 2

Van’t Hoff plots for compounds 1 (■), 2 (▲), 4 (◆), and 5 (●).

In the case of the dipeptides, one of the most useful signals in these studies, vis-à-vis resolution and intensity, was the doublet at ∼δ1.25 ppm, attributable to the methyl group (Hγ) of the Thr residue. At lower temperatures, there was excellent resolution between the upfield (cis) and downfield (trans) signals. An increase in temperature was accompanied by a move toward coalescence of the two signals; however, complete coalescence was not observed within the temperature range accessible in D2O.[38] Broadening and partial coalescence of the signals in the case of glycosylated dipeptide 4 led to omission of the 85 °C data point (Figure 2), since integration of the two signals was not possible.

The incidence of cis prolyl peptide bonds, when preceded by a Thr residue, has been estimated at 2.9% based on crystallographic data for proteins.[39] In shorter peptides, without secondary structure, the percentage is higher.[40] Larger acyl groups N-terminal to the Pro residue have a larger steric bias toward the trans conformation. This notwithstanding, K for dipeptide 1 is low in the context of other dipeptides we have studied: Ac-Gly-Pro-NHMe (K 5.5 in D2O at 298 K) and Ac-Phe-Pro-NHMe (K 2.2).[36] The latter is expected to favor the cis conformation due to an aromatic–proline interaction.

As anticipated,[37,38,41] hydroxylation of the proline residue led to a significant shift in favor of the trans peptide bond (Scheme 3, Table 2). This arises from stereoelectronic effects that favor a Cγ-exo conformation of the pyrrolidine ring that, in-turn, aligns the peptide backbone for a stabilizing n → π* interaction[42,43] in the trans conformation. Formation of the α-glycoside further accentuates this effect. This second result contrasts with earlier reports from Schweizer’s group that showed both α- and β-galactosylation of Ac-Hyp-NHMe[27] or Ac-Hyp-OMe[34] to have essentially no impact on the thermodynamics or kinetics of the cis → trans equilibrium.

Scheme 3

cis → trans Isomerization of Prolyl Amide Bond

Table 2

Thermodynamics Parameters for Compounds 1, 2, 4, 5, and 7–9

compound	Kt/c (298 K)	ΔH° (kcal mol–1)	ΔS° (cal mol–1 K–1)	ΔG°(298 K) (kcal mol–1)
1	2.3	–0.64	–0.52	–0.49
2	8.7	–1.52	–0.80	–1.28
4	13.2	–1.73	–0.72	–1.52
5	3.3	–0.68	+0.06	–0.70
7a	3.2	–0.95	–0.87	–0.65
8a	3.9	–1.33	–1.76	–0.81
9a	3.7	–1.27	–1.64	–0.78

Data from ref (27).

To investigate the mechanism behind the differences in equilibrium constants for compounds 1–5, kinetics of the prolyl peptide bond isomerization were studied by magnetization transfer NMR experiments.[44,45] Experiments were performed at elevated temperatures (60–80 °C) because at lower temperatures the rates are too slow to be determined by this method. Related experiments have been performed in unbuffered D2O,[46−48] and Reimer et al. demonstrated that for end-capped pentapeptide Ac-Ala-Thr-Pro-Ala-Lys-NH2, the rate of cis–trans isomerization was independent of pH across the 1.8–10.8 range and a variety of temperatures.[39] This was shown to be a general phenomenon across a series of pentapeptides without ionizable side chains. The CIFIT program[49] was used to estimate the rate constants for the trans → cis process. The reverse reaction rates (cis → trans) were derived therefrom in conjunction with the equilibrium constants. Data were fitted according to the Eyring equation[50] to afford the activation energy data presented in Table 3.

Table 3

Activation Parameters for trans → cis and cis → trans Isomerization about the Prolyl Amide Bonda

	trans → cis				cis → trans
compound	ktc (s–1)	ΔH‡ (kcal mol–1)	ΔS‡ (cal mol–1 K–1)	ΔG‡ (kcal mol–1)	kct (s–1)c	ΔH‡ (kcal mol–1)	ΔS‡ (cal mol–1 K–1)	ΔG‡ (kcal mol–1)
1b	0.38e	9.86	–32.6	19.57	0.72f	9.86	–30.6	18.98
2b	0.22e	18.12	–8.3	20.59	1.33f	16.96	–8.0	19.34
5b	0.34e	18.9	–5.8	20.63	0.94f	17.90	–6.8	19.93
7b,d	0.31	21.3	1.4	20.9	0.81	20.6	1.2	20.2
8b,d	0.25	21.1	0.5	21.0	0.73	20.2	0.1	20.2
9c,d	0.27	20.6	1.0	21.3	0.83	22.4	6.2	20.6

The italic values are data determined in the current work; data below those in italics come from the literature[27] for comparative purposes.

Unbuffered D2O.

Phosphate buffer, pH 7.4.

Reference (27).

Measured by magnetization transfer.

Calculated from K and k.

Upon hydroxylation there is a decrease in the rate of conversion of trans → cis ∼20% in the case of Ac-Pro-NHMe (8 vs 7)[27] and ∼40% at the dipeptide level (2 vs 1). The cis → trans rates do not follow the same trend; there is a ∼10% decrease in rate from 7 to 8,[27] but a ∼80% increase from 1 to 2. This accounts for the more pronounced increase in K in the dipeptide series. Indeed, Larive and Rabenstein also concluded that large K values for prolyl peptide bond isomerization in oxytocin and arginine–vasopressin were also due to an increase in the rate of the forward reaction, rather than an inherent increase in the stability of the trans conformation, relative to smaller molecules.[46] With the increase in electronegativity of the 4R substituent, there is presumably an increase in the sp3-character and degree of pyramidalization of the pyrrolidine nitrogen that facilitates rotation about the C–N amide bond.[47,51,52] As originally demonstrated by Fischer et al.[51], a hydrogen bonding interaction between the pyramidalized nitrogen and the NH of the residue C-terminal to Pro (mimicked in our case by the NHMe) helps to lower the energy of the transition state. Model compounds in which this amide is substituted for by an ionizable carboxylate or an ester cannot emulate this feature of the peptide backbone.

In the proline methyl amides 7–9, the free energy barriers to rotation about the amide bond are dominated by enthalpic factors. While entropy values are positive and close to zero for these amino acid derivatives, significant negative entropies of activation are observed for compounds 1, 2, and 5. This is generally true of larger peptides[46] and reflects a higher degree of organization in the transition state for an extended peptide molecule.[28] Compound 1 shows very different activation enthalpy and entropy contributions, relative to compounds 2 and 4. This must be attributable to the influence of the electron-withdrawing substituent on the energy and degree of organization of the transition state.

In the case of glycodipeptide 4 there was no cis/trans pair of resonances in the 1H NMR spectrum with adequate resolution for the signal of one conformation to be selectively inverted without also inverting the signal of the other conformation. Shi et al. have previously acknowledged that not all peptides are amenable to rate determination in this manner.[47] Lubell and co-workers have used inversion of the Cα trans signal in 13C NMR to measure rates of trans → cis isomerization.[38,53] In the present case, a single set of resonances was observed for compound 4, even at 70 °C. These experimental facts precluded magnetization transfer experiments that would have given insight into the impact of glycosylation on cis–trans isomerization kinetics at the dipeptide level.

The importance of N-acetyl groups of sugars in hydrogen bonding and peptide conformation has been commented on by Wong and co-workers[16] and also by Live et al.[54,55] The differences in K, k, and k between compound 5, bearing a GlcNAc residue, and the Gal derivative 9 reported by Schweizer and co-workers[27] are almost within experimental error. The only notable difference is in the entropies of activation: the GlcNAc residue in 5 seems to incur a higher degree of order in the transition state for peptide bond isomerization than the Gal residue in 9.

On the basis of GOESY experiments conducted for compound 9, Owens et al. described a conformation in which the hydrophobic face of the galactose residue sits over the top of the pyrrolidine ring[27] (Figure 3). There were few crosspeaks in the NOESY spectrum of compound 5, in which Gal is substituted by GlcNAc. Weak crosspeaks potentially support a conformation similar to that of 9 but are ambiguous due to severe peak overlap in the 3.5–4.0 ppm region of the 1H NMR spectrum. A more defined conformation for dipeptide 4 was discernible, based on NOESY crosspeaks summarized in Figure 3. One-dimensional GOESY experiments supported these conclusions (see Supporting Information). As depicted in Figure 3, it appears that the strong preference for the Cγ-exo pyrrolidine ring and trans peptide bond (with alignment for the n → π* backbone interaction) renders the threonine side chain partially over the pyrrolidine ring, pushing the sugar residue away.

Figure 3

Conformation of compounds 4, 5, and 9. Double-headed arrows indicate NOE correlations.

To better understand the role of the threonine residue, we considered key chemical shift and coupling constant data for this residue. In compounds 1, 2, and 4, the Thr Cα and Cβ chemical shifts did not show much variation: 57.1–57.2 and 67.1–68.8 ppm, respectively. These are somewhat upfield of statistical average values,[56−58] as expected when the (i + 1) residue is Pro.[59] Amide proton (vide infra) and Hα chemical shifts likewise showed little change upon Cγ substitution: 8.15–8.26 and 4.58–4.61 ppm, respectively. These values are closer to those observed for random coil peptides than for α-helix or β-sheet. motifs.[55−57] There was an increase in 3J across the series at 295 K: 4.8 Hz (1), 6.9 Hz (2), and 7.8 Hz (4) reflecting a change in the theta (ϕ) dihedral angle.

1H NMR spectra were acquired in 90% H2O, 10% D2O at neutral pH. Amide protons were assigned on the basis of TOCSY spectra. The C-terminal methyl amide (NHCH3) gave rise to a broad signal with a similar chemical shift in all compounds. The GlcNHAc amide in compounds 4 and 5 showed a typical 3J coupling contant of 9–10 Hz to H2.[60] The AcNHThr amide (in compounds 1, 2, and 4) displayed a chemical shift of δ 8.1–8.2 ppm. A significant difference in chemical shift was observed for the Hδ that is trans to Hγ in compounds 4 (δ4.17 ppm) and 5 (δ3.79 ppm). This reflects the relatively deshielded environment in compound 4 where the polar Thr side chain is in close proximity, compared to compound 5 where Hδ is shielded by the hydrophobic face of the sugar. In all cases, variable temperature experiments showed typical temperature dependence for non-hydrogen bonded amides (Δδ < −4 ppb K–1). Thus, the strong conformational preferences appear to be due to steric and stereoelectronic effects and thereby provide a true baseline for moving forward to study larger peptides.

Conclusion

Reorganization of the pivotal Pro143 region of Skp1 is a recurring theme during interactions with F-box proteins. Recent CD and SAXS studies show that the secondary structure content of free Skp1 of Dictyostelium is sensitive to its glycosylation status which correlates with binding to a model F-box protein.[11] Crystal structures have demonstrated a well-folded state for Skp1 complexed with F-box proteins in other organisms.[19,20] The equivalent of Pro143 is highly conserved throughout phylogeny except for chordate animals. Even in organisms with multiple Skp1 genes, at least one copy of the Pro-containing Skp1 is retained. The amino acid at this position initiates the C-terminal α-helix. Even in Skp1 proteins where the equivalent of Pro143 is substituted by another amino acid, the theme of reorganization persists. A study by Tan et al. of OCP2 (an orthologue of Skp1 from the organ of Corti) concluded that binding to OCP1 (bearing an F-box motif) is accompanied by a substantial degree of refolding.[61] The C-terminal region is alternatively folded in a loop in one complex.[62]

Thus, a picture is emerging that Skp1 exists as an ensemble of conformations whose suitability to dock with F-box protein partners is under environmental regulation via effects of post-translational modifications to Skp1. Indeed, Zondlo and co-workers have recently uncovered the role of threonine/serine phosphorylation and GlcNAc attachment in mediating neighboring secondary structure.[63,64] The model we propose herein is analogous to the mechanism by which the Skp1 homologue elongin c selects its binding partners in a related E3 polyubiquitin ligase[65] and may allow for regulation of the rate or selectivity of polyubiquitination of target proteins with impact on development. Milner-White et al. observed that Pro residues at the N-terminus of α-helices always adopt trans prolyl peptide bonds and demonstrate a strong preference for the Cγ-exo pyrrolidine conformation.[66] The post-translational modifications to Pro143 enhance preferences for these conformational attributes and thereby offer an attractive mechanism to trigger global changes in Skp1 conformation and deserve further mechanistic studies.