Structural architecture of Galdieria sulphuraria DCN1L

Introduction

NEDD8ylation is a major regulatory event for cullin-RING E3 ubiquitin ligases (CRL).^1,2 NEDD8ylation activates the ubiquitylation activity of CRLs,^3–5 which serves to regulate a wide variety of cellular processes⁶ by selecting specific targets for destruction in the proteasome.⁷ NEDD8 E3 ligases catalyze ligation of the ubiquitin-like protein NEDD8 (Rub1 in yeast) to the C-terminal winged-helix domain of the CRL “cullin” subunit.⁸ NEDD8ylation increases the ubiquitin (Ub) ligase activity of CRLs by enhancing binding of the CRL RING subunit to Ub-charged Ub-conjugating (E2) enzymes,⁹ and by driving conformational changes that permit the RING subunit to productively orient Ub-E2 conjugates with bound CRL substrates for catalysis.¹⁰ However, it is unclear how the NEDD8 E3 ligase catalyses its reaction, what its exact composition is, and how it is regulated. Moreover, current structural and functional models of the NEDD8 E3 ligase are based on the Saccharomyces cerevisiae system,⁸ in which, an integral NEDD8 E3 ligase subunit, DCN1, has low sequence identity to mammalian DCN1-like proteins, and several amino acid insertions, which together may impede understanding of NEDD8 E3 function in a mammalian system.

Protein NEDD8ylation is analogous to ubiquitylation.¹¹ NEDD8-activating enzymes (E1), conjugating enzymes (E2), and ligases (E3) are required for the ligation of NEDD8 to its target CRL(s).² Each enzyme catalyzes transfer of NEDD8 to its respective acceptor protein through NEDD8’s C-terminal glycine carboxylate. First, in an ATP dependent reaction, NEDD8 is covalently linked to the active site cysteine of a heterodimeric E1 enzyme.¹² NEDD8 is then transferred to the active site cysteine of an E2 enzyme (Ubc12).¹³ The E3 enzyme complex then serves to ensure reaction specificity and to promote chemical catalysis en route to NEDD8 transfer to a specific cullin lysine.⁸ The NEDD8 enzyme complex contains the cullin substrate, a RING domain containing subunit with bound NEDD8-E2 conjugate, and a DCN1-like protein (DCN1L).⁸

Mammalian cells utilize five DCN1L paralogs,¹⁴ in contrast to only two cullin-associated RING paralogs.⁸ The exact function of each DCN1L is currently unknown. Some DCN1Ls may work in a non-redundant manner as exemplified by human DCN1L3, which preferably interacts with human cullin 3 (Cul3), and localizes Cul3 to the plasma membrane.¹⁴ DCN1Ls generally contain two domains, an N-terminal domain that is unique for each paralog, and a conserved C-terminal domain, termed the PONY-domain.^15,16 The purpose of the N-terminal domain is not well understood, but likely aids in reaction specificity and cellular localization.¹⁴ The PONY-domain, however, is integral to protein NEDD8ylation, and is functional in CRL neddylation in the absence of the N-terminal domain.^8,16

Recently, Scott et al. solved the structure of yeast DCN1 in complex with the winged-helix domain of the S. cerevisiae cullin, Cdc53, and provided some clarification for the composition and mechanism of NEDD8 E3s.⁸ They identified DCN1 and the S. cerevisiae CRL RING subunit, Hrt1, as the major functional components of the yeast NEDD8 E3 ligase. Cdc53 appears to serve as the scaffold for its own NEDD8ylation, as it is already binds Hrt1 in the canonical CRL complex and binds tightly to DCN1.⁸ Hrt1 binds the NEDD8-charged E2, Ubc12, and enhances chemical catalysis.⁸ DCN1 facilitates catalysis, presumably by optimizing the orientation of the NEDD8-Ubc12 conjugate with the reactive Cdc53 lysine residue.⁸

Here, we report the structure of a DCN1L from Galdieria sulphuraria (Gs-DCN1L) to 1.3 Å resolution. As all DCN1L structures deposited in the PDB to date are from S. cereviseae DCN1 (Sc-DCN1), which has low sequence identity and frequent amino acid insertions within its polypeptide chain as compared to mammalian DCN1Ls, we sought a DCN1L that would be more comparable with mammalian forms. Gs-DCN1L was identified in genomic sequences¹⁷ and expressed sequence tag data from G. sulphuraria.¹⁸ Gs-DCN1L was targeted for structural characterization when crystallization trials for the homologous protein, Mus musculus Dcn1D1 (Mm-Dcn1D1), failed to yield crystals. Gs-DCN1L shares higher sequence identity with mammalian DCN1Ls than Sc-DCN1, with a low frequency of amino acid insertions and gaps. Gs-DCN1L shares 25% amino acid identity with Sc-DCN1. At the cullin-binding region Gs-DCN1L and Sc-DCN1 are very similar in fold and in sequence, indicative of a conserved cullin-binding mode. Gs-DCN1L and Sc-DCN1 structures diverge markedly near the cullin interaction surface. In this region Gs-DCN1L has a disordered loop that exposes a hydrophobic pocket, whereas, the comparable polypeptide segment of Sc-DCN1 completely covers the analogous hydrophobic pocket, and is well ordered.^8,15,16 Sequence analysis indicates that this intriguing structural divergence is conserved among Gs-DCN1L and the mammalian DCN1L paralogs and may provide clues for additional DCN1L functionalities.

Materials and Methods

Expression and purification

The standard Center for Eukaryotic Structural Genomics (CESG) platform for cloning,¹⁹ protein expression,²⁰ purification,²¹ and bioinformatics management²² was utilized to produce Gs-DCN1L. Briefly, cDNA was cloned into a pEU-His-Flexi vector from cDNA provided by Andreas Weber.¹⁹ Cell-free expression was conducted on a 10-mL scale for 48 hours using WEPRO8240 extract and a Protemist XE protein synthesizer (CellFree Sciences, Yokohama, Japan). Selenomethionine was added to 0.6 mM, and the remaining 19 amino acids were added to 0.3 mM. His-tagged protein was purified by nickel affinity chromatography. The N-terminal His-tag was cleaved with tobacco etch virus protease, and tag-free protein was isolated by subtractive nickel affinity chromatography. Size-exclusion chromatography provided additional purification and permitted exchange of Gs-DCN1L into the final protein buffer (5 mM HEPES, 50 mM NaCl, 0.3 mM TCEP, pH 7.0). Gs-DCN1L was concentrated to 10 mg/mL, and flash frozen in liquid nitrogen.

Crystallization and structure solution

All crystallization experiments were set up in a sitting drop vapor diffusion format with a TTP Labtech Mosquito (Royston, UK). The UW192 screen (CESG) and the Index HT screen (Hampton Research, Aliso Viejo, California) were utilized to identify crystallization conditions. Crystallization experiments were monitored with and stored in Bruker AXS Crystal Farms (Madison, Wisconsin) operating at 4°C or 20°C. Precipitant solutions for crystal optimization were assembled with a Tecan Genesis RSP 150 (Männedorf, Switzerland).²²

MAD data were collected from a crystal grown in 27% PEG 1500, 150 mM LiCl, and 100 mM sodium acetate, pH 5.0. Data for structural refinement were collected from a crystal grown in 30% PEG 1500, 150 mM Li₂SO₄, and 100 mM sodium acetate, pH 5.0. The crystals were cryoprotected in fomblin 06/6 and fomblin 18/8, respectively, and flash frozen in a 100 K nitrogen stream. MAD and refinement data were collected at the General Medicine and Cancer Institutes Collaborative Access Team (GM/CA-CAT) 23-ID-D and the Life Sciences Collaborative Access Team (LS-CAT) 21-ID-D beamlines, respectively (Argonne National Laboratory, Argonne, Illinois). Data were indexed, integrated and scaled using HKL2000.²³ Phenix.hyss²⁴ and ShelXD²⁵ were employed to evaluate the selenium substructure. Refinement of Se positions, density modification and initial model building were conducted with AutoSharp.²⁶ Molecular replacement was conducted with Molrep,²⁷ using the Gs-DCN1L model produced by AutoSharp. Successive iterations of model building with Coot²⁸ and refinement with Phenix.refine²⁹ yielded the final model. Model validation was conducted with Molprobity³⁰ and Procheck.³¹ Superpositions were conducted with LSQKAB.³² Structures were presented using PyMOL (DeLano Scientific LLC, San Carlos, California).

Analysis of Gs-DCN1L oligomeric state

Analytical size exclusion gel chromatography was conducted using a 24 ml Superdex 200 GL column (GE Healthcare, Piscataway, New Jersey) with an ÄKTA FPLC chromatographic system (GE Healthcare, Piscataway, New Jersey) at 25°C. 25 μl of sample were loaded per run. Protein elution was monitored by UV absorbance at 280 nm. The elution buffer comprised 200 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine, and 50 mM HEPES, pH 7.5 at 25°C. The column was calibrated with blue dextran, ferritin, conalbumin, carbonic anhydrase, RNase A.

Results and Discussion

Since Mm-Dcn1D1 failed to produce crystals in our high-throughput platform, Gs-DCN1L was selected as a candidate for structural characterization. Gs-DCN1L has higher sequence identity with mammalian DCN1Ls than Sc-DCN1, which is the only other structurally characterized DCN1L to date (Figure 1A). Pairwise comparisons of Gs-DCN1L and human DCN1L proteins 1–5 (Hs-DCN1L) yielded 36–46% amino acid identity over the 199 amino acid Gs-DCN1L PONY-domain, while Sc-DCN1 and human pairs share 22–26% identity.

Figure 1. Sequence comparison of DCN1Ls and overall Gs-DCN1L structure.

(A) A sequence alignment of G. sulphuraria (Gs), human (Hs), and S. cerevisiae (Sc) DCN1Ls is presented. Secondary structural elements correspond to the Gs-Dcn1L structure. Positions of residues involved in cullin-binding as described by Scott et al.⁸ are indicated with an asterisk, where red or black indicates interactions that were mediated through main chain or side chain atoms, respectively. The location of Ile244 of Sc-Dcn1 is indicated with a dagger. (B) The Gs-Dcn1L structure is rendered as a cartoon in this stereo pair. α-helices and the N– and C–termini are labeled. For clarity the 9–10 loop is circled and the side chains of residues that are bounded by unmodeled regions of the loop are shown. The protein is colored by B-factor using the visible spectrum, where blue and red indicate relatively low or high B-factors, respectively. (C) Gs-DCN1L was superposed with the Sc-DCN1 of the Sc-DCN1/Cdc53 complex (PDB 3O2P). Gs-DCN1L is colored as in (B), Sc-DCN1 and Cdc53 are colored gray and black, respectively. Selected α-helices and loop-regions are labeled. An asterisk is used to show the position of the disordered Gs-DCN1L residues that correspond with the Cdc53 binding region of Sc-DCN1

Gs-DCN1L was expressed based on gene models from the G. sulphuraria genome project.^17,18,33 PONY-domain containing proteins generally comprise two domains, a C-terminal PONY-domain,¹⁶ and a variable N-terminal domain, which may function in regulation, substrate specificity, or cellular localization.¹⁴ Notably, Gs-DCN1L lacks an N-terminal domain. Since the PONY-domain of Sc-DCN1 is sufficient to stimulate cullin neddylation in vivo and in vitro,¹⁶ we reasoned Gs-DCN1L would provide results of general interest for understanding PONY-domain linked cullin neddylation.

The structure of Gs-DCN1L was solved to a resolution of 1.3 Å. MAD phasing of data collected from a single crystal and automated model building yielded an initial model, which was utilized to phase a high-resolution dataset by molecular replacement. The latter dataset was used for structural refinement. Statistics for data collection and structural refinement are provided in Table I. The final Gs-DCN1L model contained one selenomethionine, and one oxidized cysteine. Residues 1–3, 165–166, 174–176, and 193–199 were not modeled for lack of interpretable electron density.

Table I.

Crystal Parameters, Xray Data Collection, and Refinement Statistics

Dataset	Refinement	Peak	Inflection	High remote	Low remote
Crystal parametersa
Unit-cell parameters (Å)	36.7 × 52.4 × 49.1, β=111.2°	36.2 × 51.5 × 48.5, β=110.5°	36.4 × 51.7 × 48.7, β=110.2°	36.4 × 51.6 × 48.7, β=110.3°	36.4 × 51.8 × 48.7, β=110.4°
Data collection statisticsb
Wavelength (Å)	0.97625	0.97949	0.97965	0.96427	0.99522
Resolution range (Å)	50.0–1.30 (1.32–1.30)	50.0–1.81 (1.84–1.81)	50.0–1.81 (1.84–1.81)	50.0–1.79 (1.82–1.79)	50.0–1.84 (1.87–1.84)
No. of reflections	42243(1849)	15091(671)	15189(680)	15703(767)	14482(615)
Completeness (%)	98.5(87.2)	97.7(89.7)	97.5(88.1)	98.0(94.5)	97.4(84.2)
Rmergec	0.040(0.388)	0.075(0.231)	0.074(0.275)	0.074(0.228)	0.076(0.326)
Redundancy	4.0(3.0)	7.5(6.3)	7.5(5.8)	7.5(6.7)	7.4(5.2)
<I>/σ(I)	23.1(2.6)	25.4(10.2)	27.1(8.4)	27.0(11.1)	28.3(5.7)
Refinement and model statisticsb
Resolution range (Å)	21.0–1.30 (1.33–1.30)
No. of reflections (work/test)	40050/2111
Completeness (%)	98.5(88)
Rcrystd	0.156(0.183)
Rfreee	0.178(0.197)
RMSD bonds (Å)	0.018
RMSD angles (°)	1.566
B factor-overall/ protein/waters (Å2 )	20.2/18.6/32.2
No. of protein molecules/ all atomsf	1/1714
Rama. plot by MOLPROBITY (%)
Favored region	98.3
Additionally allowed region	1.7
Outliers	0
PDB code	3KEV

^aAll crystals belonged to the P21 space group.

^bValues in parentheses are for the highest resolution shell.

^cR_merge=Σ_hΣ_i|I_i(h)− <I(h)> |/ Σ_hΣ_iI_i(h), where I_i(h) is the intensity of an individual measurement of the reflection and <I(h)> is the mean intensity of the the reflection.

^dR_cryst=Σ_h||F_obs|−|F_calc||/Σ_h|F_obs|, where F_obs and F_calc are the observed and calculated structure-factor amplitudes, respectively.

^eR_free was calculated as R_cryst using the randomly selected unique reflections (5.01%) that were omitted from structural refinement.

^fThese include atoms from 176 water molecules, acetate ions, and 1 sulfate ion.

The overall fold of Gs-DCN1L comprises ten α-helices, and a largely disordered loop made up of residues 164–178 that is located between Helices 9 and 10 (Figure 1B). This loop, which we have termed the “9–10 loop,” contains some helical character. The analogous segment of Sc-DCN1 (residues 237–254) is well-ordered and contain an additional helix (Sc-DCN1 Helix 10). Despite low amino acid sequence identity (25% over 199 amino acid residues), Gs-DCN1L and Sc-DCN1 superposed with an RMSD of 1.6 Å over 152 matching α-carbons (see Figure 1C). Like Sc-DCN1,^15,16 Gs-DCN1L contains two EF-hand pairs. The N-terminal EF-hand pair comprises Helices 1–4 and specifically binds Helix 5, while the C-terminal EF-hand pair is composed of Helices 6–9. The solution size of Gs-Dcn1L as measured by size exclusion chromatography was 33 kDa. Given that Gs-Dcn1L is rod-shaped, and has a molecular weight of 23 kDa, the measured molecular weight is appropriate for Gs-DCN1L monomers.

Through structural and functional analyses Scott et al. identified several Sc-DCN1 PONY-domain residues that participate in binding of the winged helix domain of the yeast cullin, Cdc53. Matching or analogous Gs-DCN1L residues are located in Helices 9 and 10 and in the 9–10 loop (See Figure 1A). This result in addition to conservation of the protein fold (Figure 1C) indicates a conserved cullin-binding mode among DCN1Ls. At the cullin-binding site, Gs-DCN1L differs from Sc-DCN1, as residues 174–176 are unstructured (Figure 1C). Disorder within this portion of the polypeptide chain may be common among mammalian homologs, as all Hs-DCN1Ls except DCN1L3 have a glycine in the position analogous to Gly176 of Gs-DCN1L rather than the alanyl residue present in Sc-DCN1. Hs-DCN1L3 has a glutamate at this position. In this region, Sc-DCN1 utilizes the main chain atoms corresponding to Gs-DCN1L Asn175 to interact with Cdc53.⁸ Consequently, cullin-binding may stabilize this region in Gs-DCN1L to resemble Sc-DCN1 more closely. Regarding conservation of cullin-binding site residues among the human paralogs, there are some variations (Figure 1A). Hs-DCN1L1 and Hs-DCN1L2 are practically identical and similar to Gs-DCN1L and Sc-DCN1. Hs-DCN1L4 and Hs-DCN1L5 differ at the center of Helix 9, while Hs-DCN1L3 varies in the 9–10 loop. Each Hs-DCN1L paralog should display subtle differences in their respective cullin-binding surfaces that may provide some reaction specificity, but the overall main chain fold within the cullin-binding face should be quite similar, notwithstanding differences elsewhere within the five N-terminal helices of the PONY-domain.

Since EF-hand pairs generally serve as protein binding modules and the N-terminal EF-hand of the PONY domain binds Helix 5 (Figure 1B and Yang et al.¹⁵), we employed the Dali search engine³⁴ to identify analogous EF-hand pairs complexed with an α-helix. Figure 2A shows three cases in which an EF-hand pair binds an α-helix in trans (troponin C-troponin I complex, PDB 1J1E;³⁵ caltractin-Kar1 complex, PDB 1OQP;³⁶ Cdc31p-Sfi1p complex, PDB 2GV5³⁷). Thus, Gs-DCN1L utilizes a common EF-hand conformation to bind Helix 5. This structural feature participates in the juxtaposition of the two EF-hand pairs. The conservation of loop and α-helix length from Helix 1 of the PONY-domain to the loop bridging Helices 5 and 6 is high among DCN1Ls, suggesting that the relative orientation of the two EF-hands has been important in the evolution of DCN1Ls. The importance of the relative positions of the EF-hand pairs has some support as residues of the N-terminal EF-hand pair influence the ability of Sc-DCN1 to facilitate Cdc53 NEDD8ylation through interaction with the N-terminal α-helix of S. cerevisiae NEDD8-Ubc12 conjugates.⁸ Moreover, alteration in the shape of Ubc12’s N-terminal α-helix reduces Sc-DCN1’s catalytic efficiency. This interaction appears to be important for the proper orientation of the RING-bound NEDD8-Ubc12 conjugate with respect to cullin’s reactive lysyl residue.⁸ However, it should be noted that human Ubc12 may not form an N-terminal α-helix,⁸ and the interactions between NEDD8-Ubc12 and Hs-DCN1Ls have not been characterized experimentally.

Figure 2. Comparisons of Gs-DCN1L with EF-hand containing proteins and Sc-DCN1, and modeling of the Cul1-Rbx1-DCN1L complex.

(A) The N-terminal EF-hand pair of Gs-Dcn1L (black) was superposed with the following EF-hand pair containing proteins: troponin C (green), caltractin (red) and Cdc31p (blue) and rendered as ribbons in this stereodiagram. Helix 5 of Gs-Dcn1L and the binding partners of the other EF-hand pairs, Troponin I, Kar1 and Sfi1p, respectively, were not included in the superposition. (B) Gs-DCN1L (slate) and Sc-DCN1 (gray) were superposed and rendered as cartoons in this stereodiagram. Using the Gs-DCN1L naming convention, Helices 8–10 were labeled as H8-H10. Residues lining the hydrophobic pocket formed by Helices 8–10 are displayed as sticks. Except for Ile244 of Sc-DCN1, all labeled residues correspond to the Gs-DCN1L model. The major anchor point of the Gs-DCN1L 9–10 loop, Tyr172, is included. A line was drawn between Asp168 of Gs-DCN1L and Ile244 of Sc-DCN1 to indicate the positional disparity of these sequence related residues, and the gross differences in polypeptide chain morphology between the two respective DCN1Ls. Note that the 9–10 helix of Gs-DCN1L was not fully modeled due to its high mobility. (C) Gs-DCN1L (colored as in Figure 1B) was modeled onto the Cul1-Rbx1 complex (gray and black, respectively), and shown as a stereodiagram. This was accomplished by superposition of Gs-DCN1L onto the Sc-DCN1/Cdc53 complex (PDB 3O2P), and through superposition of Cdc53 upon Cul1 of the SCF ubiquitin ligase complex (PDB 1LDK). Atoms of the NEDD8 reactive Cul1 residue, Lys720, are shown as spheres.

Enlarged loop size in Sc-DCN1, including six additional residues between Helices 7 and 8 (the “7–8 loop,”), and three additional residues within the 9–10 loop (Figure 1A), constitute the most obvious differences in the overalls folds of Sc-DCN1 and Gs-DCN1L/Hs-DCN1Ls. The 7–8 loop differences lead to structural divergence between residues 196–207 of Sc-Dcn1 and residues 130–135 of Gs-DCN1L. The additional Sc-DCN1 residues within the 7–8 loop provides a disparate surface between Helix 5, the C-terminal EF-hand pair adjacent to the 7–8 loop. Effectively, a groove found in this region of Gs-DCN1L is absent in Sc-DCN1. The structural variations between Gs-DCN1L and Sc-DCN1 at the 9–10 loop (Helix 10 in Sc-DCN1) are points of interest. Gs-DCN1L is relatively disordered in this region, while Sc-DCN1 is well ordered and has a prominent α-helix (Figure 1C). In Gs-DCN1L the 9–10 loop is stabilized by the interaction of Tyr172 with Phe161 of Helix 9 and Trp178, Ile182, Asp183, and Val186 of Helix 10 (Figure 2C). Though these interactions are predominantly hydrophobic in nature, Tyr172 appears to interact with Asp183 via hydrogen bonding. Asp183 maps to Asp259 of the Sc-DCN1, which has demonstrable importance in NEDD8ylation in the S. cerevisiae system and hydrogen bonds to Lys790 of Cdc53.^8,14,16 Thus, the fact that the highly conserved Tyr172 is the only residue of the 9–10 loop that is noncovalently tethered to the remainder of the PONY domain is likely of functional significance. We hypothesize that Tyr172 is important for the proper positioning of the aspartate with cullin. Ile244 of Sc-DCN1 contributes to the stability of the helix corresponding to the Gs-DCN1L 9–10 loop, and its interaction with the hydrophobic pocket formed by Helices 8–10 (Gs-DCN1L numbering). Gs-DCN1L and the Hs-DCN1Ls have an aspartate residue in this position (Asp168 in Gs-DCN1L, Figure 1A). In Gs-DCN1 the corresponding aspartate is solvent exposed and the helical region is rotated away from the comparable position in hydrophobic pocket occupied by Ile244 of Sc-DCN1 (Figure 2B). A direct result of the primary sequence differences in the 9–10 loop is that a portion of the hydrophobic cavity formed by Helices 8–10 is solvent exposed in Gs-DCN1L. Since the 9–10 loop is highly mobile in this region it is difficult to assess which hydrophobic residues are absolutely exposed. However, the proximity of the hydrophobic pocket to the cullin-binding site is intriguing (Figure 2C), and may serve as a binding site for molecules that would modulate DCN1L function or availability.

Overall, notwithstanding low sequence identity and loop-size differences, the Gs-DCN1L fold is similar to Sc-DCN1 at the proposed cullin and the NEDD8-E2 conjugate binding regions.⁸ This result and the similarity of Gs-DCN1L with mammalian DCN1Ls suggest that mammalian DCN1Ls share high secondary and tertiary structural homology at their respective substrate binding sites with both Gs-DCN1L and Sc-DCN1. Given the relatively high sequence identity of Gs-DCN1L for mammalian DCN1Ls, especially types 1 and 2, our structure will provide additional insight to guide experiments designed to probe DCN1L function in mammals. The most glaring difference between Gs-DCN1L and Sc-DCN1 is the exposed hydrophobic pocket near the 9–10 loop in Gs-DCN1L. The sequence similarity among Gs-DCN1L and mammalian DCN1Ls in this region suggest that mammalian DCN1Ls will also have an exposed hydrophobic cavity in this region. This striking feature of Gs-DCN1L suggests that some DCN1Ls may bind additional molecules, which could modulate their function.