Bacterial protease uses distinct thermodynamic signatures for substrate recognition

Abstract

Porphyromonas gingivalis and Porphyromonas endodontalis are important bacteria related to periodontitis, the most common chronic inflammatory disease in humans worldwide. Its comorbidity with systemic diseases, such as type 2 diabetes, oral cancers and cardiovascular diseases, continues to generate considerable interest. Surprisingly, these two microorganisms do not ferment carbohydrates; rather they use proteinaceous substrates as carbon and energy sources. However, the underlying biochemical mechanisms of their energy metabolism remain unknown. Here, we show that dipeptidyl peptidase 11 (DPP11), a central metabolic enzyme in these bacteria, undergoes a conformational change upon peptide binding to distinguish substrates from end products. It binds substrates through an entropy-driven process and end products in an enthalpy-driven fashion. We show that increase in protein conformational entropy is the main-driving force for substrate binding via the unfolding of specific regions of the enzyme (“entropy reservoirs”). The relationship between our structural and thermodynamics data yields a distinct model for protein-protein interactions where protein conformational entropy modulates the binding free-energy. Further, our findings provide a framework for the structure-based design of specific DPP11 inhibitors.

Introduction

Periodontitis is the most common chronic inflammatory disease of humans worldwide, affecting nearly half of adults in the United Kingdom and the United States of America^{1, 2}. The condition is characterized by destruction of the connective tissue and alveolar bone surrounding the teeth and has many negative impacts in life quality³, for instance, loss of permanent tooth. Porphyromonas gingivalis ⁴ is the major causative agent in periodontitis and Porphyromonas endodontalis ⁵ is another abundant bacterium in periodontal sites. Considerable attention has been drawn to these organisms due to recent reports associating periodontitis to systemic diseases⁶ like type II diabetes mellitus⁷, rheumatoid arthritis⁸, oral cancers^{9, 10}, cardiovascular diseases¹¹, Alzheimer et al.¹² and respiratory diseases¹³. In particular, P. gingivalis is a model pathogen for investigating microbial subversion in periodontal host immune response, which causes adverse impacts in systemic health¹⁴.

Both Porphyromonas species are Gram-negative black-pigmented anaerobes that do not ferment carbohydrates; instead, they use proteinaceous substrates as carbon and energy source¹⁵. Proteases with different specificities reduce these extracellular proteins into di- and tri-peptides¹⁶, which are further degraded via specific pathways, producing short-chain fatty acids, ammonia, acetate, propionate and butyrate¹⁷. Together with other P. gingivalis elements such as the recently characterized pili¹⁸, these metabolic end products are also virulence factors causing host tissue damage¹⁹. In P. gingivalis, extracellular proteins are initially degraded to oligopeptides by potent cysteine endopeptidases, i.e., gingipains R (Rgp, Arg-specific) and K (Kgp, Lys-specific)²⁰, mainly localized in the outer membrane. Sequentially, the periplasmic enzymes, four dipeptidyl peptidases (DPPs), (i.e. DPP4, DPP5, DPP7 and DPP11), prolyl tripeptidyl peptidase-A and acylpeptidyl oligopeptidase convert the oligopeptides to di- and tri-peptides¹⁶, which are then incorporated via oligopeptide transporters²¹. These enzymes’ different specificities and their concerted actions secure proper nutrient source and are essential for the bacteria metabolism. However, metabolic regulation for amino acid degradation is not well understood. Furthermore, these dipeptidases are widely distributed in the bacterial kingdom, including the two major phyla Bacteroidetes and Proteobacteria ²², thus it is of ample relevance to elucidate their mechanism of action.

In P. gingivalis, the most utilized peptides contain Asp/Glu²³ and are degraded by dipeptidyl peptidase 11 (DPP11), rendering it a central metabolic role in this microorganism²⁴. The metabolism of glutamate- and aspartate-containing peptides generates cytotoxic products^{23, 25}, such as ammonia and butyrate, which may have a role in this bacterium to adversely impact systemic health. DPP11 is a dimeric 162 kDa (Supplementary Fig. S1) periplasmic serine protease (catalytic triad S652, D226 and H85) recently discovered in P. endodontalis and later identified in P. gingivalis by homology search²⁴ (they share close to 58% identity). Due to its specificity for Asp/Glu in the P1 position (second amino acid from the peptide N-terminus), DPP11 discovery is in line with the observation that aspartate and glutamate are the most intensively consumed amino acids in P. gingivalis ²³. Indeed, P. gingivalis dpp11-knock-out strain shows growth impairment²⁴, suggesting its critical role in the bacterium energy metabolism. Its absence in mammals²⁶ strengthens the enzyme’s potential as an attractive drug target. In this way, we aimed at elucidating the structural basis of peptide recognition by DPP11 in order to establish its mechanism of action.

We determined the structures for the inactive constructs PgDPP11_22-720 S655A, PeDPP11_22-717 S652A and its complexes with the dipeptides Arg-Asp and Arg-Glu, as well as the substrate Leu-Asp-Val-Trp, at 2.4, 2.85, 2.2, 2.1 and 2.6 Å resolution, referred to as PgDPP11, PeDPP11, PeDPP11:RD, PeDPP11:RE and PeDPP11:LDVWs, respectively (Table 1). DPP11 crystal structures in complex with peptides disclose a significant domain motion upon ligand binding and allow the elucidation of the enzyme’s specificity and selectivity. The distinct conformational states reported here offer opportunities for the rational development of drugs and molecular tools for DPP11 studies, which are not possible to be fully exploited in the unbound form of the enzyme. Microcalorimetric analyse reveal a dual thermodynamic signature where DPP11 binds substrates through an endothermic/entropy-driven process, and end products in an exothermic/enthalpy-driven fashion. We propose that increase in protein conformational entropy is the main-driving force for substrate recognition and that enzyme plasticity favours substrate promiscuity.

Table 1.

Data collection and refinement statistics.

	PgDPP11	PeDPP11	PeDPP11:RD	PeDPP11:RE
Data collection
X-ray source	BM14/ESRF	ID30A-1/ESRF	ID29/ESRF	ID29/ESRF
Space group	P 2 1 2 1 2 1	P 2 1 2 1 2 1	P 2 1 2 1 2 1	P 2 1 2 1 2 1
Cell dimensions
a, b, c (Å)	103.18, 117.21, 148.35	76.75, 91.83, 229.91	111.81, 114.40, 147.82	111.44, 112.53, 148.26
Resolution (Å)	47.22–2.20 (2.32–2.20)	48.72–2.85 (3.00–2.85)	48.15–2.20 (2.32–2.20)	49.42–2.10 (2.21–2.10)
R pim (%)	4.2 (57.9)	9.4 (46.8)	3.1 (37.9)	4.7 (38.3)
R merge (%)	9.1 (122.4)	22.6 (111.8)	6.1 (74.4)	8.4 (69.6)
CC1/2 (%)	99.8 (96.8)	98.9 (53.5)	99.9 (68.5)	99.6 (69.7)
I / σ(I)	12.3 (0.9)	7.5 (1.7)	14.9 (2.0)	8.8 (2.0)
Completeness (%)	99.7 (99.4)	100 (100)	99.6 (99.1)	99.0 (99.8)
Redundancy	5.6 (5.4)	6.6 (6.6)	4.7 (4.6)	3.9 (4.0)
Refinement
Resolution (Å)	47.22–2.40	46.7–2.85	45.24–2.20	47.47–2.10
No. reflections	70637	38796	9612	107843
R work / R free (%)	20.7/25.9	24.0/27.4	18.1/22.6	18.9/22.9
No. atoms	11516	10832	11235	11093
Protein	11132	10773	10769	10549
Ligand/ion	23	3	40/5	42/2
Water	361	56	421	500
B–factors (A2)
Protein	55.97	46.9	65.3	56.0/
Ligand/ion	72.93	20.78	47.6/62.6	44.6/51.326
Water	51.58	16.6	53.2	47.4
R.m.s. deviations
Bond lengths (Å)	0.002	0.003	0.008	0.008
Bond angles (°)	0.680	0.763	1.141	1.094
Ramachandran analysis
Favoured (%)	96	93.5	95.5	95.8
Allowed (%)	3.5	5.8	3.8	3.7
Outliers (%)	0.5	0.6	0.7	0.5
	PeDPP11:LDVW	PeDPP11:altconf	FpDPP11:RD
Data collection
X–ray source	ID30A–1/ESRF	ID30A–1/ESRF	ID23–1/ESRF
Space group	C2	C2	P21
Cell dimensions
a, b, c (Å)	87.78, 113.33, 111.22	88.02, 103.99, 111.39	126.05, 70.68, 191.59
	β = 106.2°	β = 104.9º	β = 97.3º
Resolution (Å)	47.40–2.60 (2.74–2.60)	46.82–2.50 (2.64–2.50)	47.61-2.10 (2.21–2.10)
R pim (%)	8.2 (75.4)	6.3 (43.8)	3.6 (43.2)
R merge (%)	13.0 (121.1)	9.6 (67.4)	5.5 (66.6)
CC1/2 (%)	99.2 (41.2)	99.5 (73.4)	99.9 (77.5)
I / σ(I)	7.5 (1.0)	9.2 (1.5)	12.1 (1.7)
Completeness (%)	99.2 (95.9)	99.0 (99.6)	98.9 (99.6)
Redundancy	3.4 (3.5)	3.1 (3.0)	3.2 (3.2)
Refinement
Resolution (Å)	43.58–2.60	46.85–2.50	46.83–2.10
No. reflections	32005	33283	193404
R work / R free (%)	21.7/24.7	19.9/24.0	19.6/24.4
No. atoms	5490	5393	22036
Protein	5399	5240	21499
Ligand/ion	23/16	3	38/4
Water	52	150	495
B–factors (A2)
Protein	51.98	48.5	65.5
Ligand/ion	43.29/90.86	58.5	70.3/58.7
Water	43.04	41.7	50.3
R.m.s. deviations
Bond lengths (Å)	0.014	0.003	0.009
Bond angles (°)	1.119	0.705	1.149
Ramachandran analysis
Favoured (%)	94	95.1	94.7
Allowed (%)	6	4.1	4.8
Outliers (%)	0	0.8	0.5

Results and Discussion

As previously reported²⁷, the overall fold of DPP11 comprises a bilobal architecture (Fig. 1a,b). The upper helical domain dictates the specificity of the enzyme and caps the catalytic domain, which has a typical chymotrypsin double β-barrel fold²⁸. PeDPP11 and PgDPP11 superposition yielded a root mean square deviation (r.m.s.d.) of 1.4 Å for 629 out of 685 superimposed Cα-atoms. A notable difference between the unbound PeDPP11 and its complexes with peptides is the conformational change bringing the helical and catalytic domains closer (Fig. 1b). This movement yields an approximate rotation of 22° of one domain relative to the other with a negligible translational component^{29, 30}. Notably, the helical domain undergoes larger structural changes reflected in higher r.m.s.d. and B-factor values, when compared to the catalytic domain, which behaves as a rigid body (Supplementary Table S1a,b).

Figure 1.

Structure of Porphyromonas endodontalis DPP11. (a) Domain architecture of PeDPP11. SP is signal peptide. The locations of catalytic triad amino acids are indicated by “red stars”. (b) Ribbon representation of PeDPP11 structure. Domains are coloured as in item (a) and helix α14 is shown in dark blue. Upper panel shows two perpendicular views of unbound PeDPP11. Lower panel shows two perpendicular views of PeDPP11 as in complex with peptides (binding pocket shown as yellow surface). (c) Active site of PeDPP11:RD (peptide RD shown in green). Catalytic triad is underlined. Note that S652 is mutated to alanine. (d) Active site of PeDPP11:LDVW (peptide LDVW shown in magenta), peptide omit map contoured at 3σ, shown in blue.

The active site of DPP11 lays in a wide cleft running through the middle of the protein between the catalytic and helical domains, which contributes to the formation of the substrate binding subsites (Supplementary Table S2a–c). The bound-peptide is anchored at its N-terminus primarily by N332 (N-anchor) located in the helical domain. It moves approximately 4.0 Å (Cα) towards the catalytic domain upon peptide binding (Supplementary Fig. S2a). The distance between the N-anchor and the catalytic S652 permits accommodation of only two amino acid residues, revealing how the enzyme acquires its dipeptidyl peptidase specificity (Figs. 1c,d). Evolutionary conserved R670 is responsible for the Asp/Glu specificity at subsite S1: its guanidinium group directly interacts with the substrate carboxyl group of Asp/Glu (Fig. 1c, Supplementary Fig. S2b). R670 and R336 confer a dominant positive charge to subsite S1 further explaining its P1 acidic specificity (Supplementary Fig. S2c). Indeed, the substitution R670D completely abolished PeDPP11 activity²⁴. In PeDPP11:LDVW, the third and fourth amino acids of the substrate (Val and Trp at positions P1′ and P2′, respectively) exhibit few interactions with the enzyme. For instance, Val (P1′) displays a weakly defined electron density with only 40% of its solvent accessible area buried by DPP11, while Trp (P2′) completely lacks electron density (Fig. 1d). This active site design renders the enzyme’s specificity more relaxed, with selectivity imposed mainly at P1 and P2 residues of the substrate. This promiscuous feature of DPP11 helps to provide nutrients for P. gingivalis and P. endodontalis given the scarce resources in the subgingival plaque³¹. However, the strategy to increase enzyme promiscuity comes with a price: the affinities for substrates and end products are strikingly similar (Fig. 2).

Figure 2.

Microcalorimetric analysis. Isothermal titration calorimetry experiments performed by titrating LD (left panel) and LDVW (right panel) into PeDPP11. Upper panel shows time-dependent deflection of heat for each injection (top). Integrated calorimetric data for the respective interactions (bottom). The continuous curve represents the best fit using a one-site binding model. Lower panel shows the graphical representation of thermodynamics parameters.

We performed a series of isothermal titration calorimetry experiments to further characterize peptide binding to PeDPP11. Binding of LD and RD dipeptides/end products to PeDPP11 was largely exothermic (ΔH _bind of −22.0 and −15.5 kJ.mol⁻¹, respectively) at 25 °C indicating an enthalpy-driven process [Fig. 2 (left panel), Supplementary Fig. S3a]. A favourable change in entropy due to water displacement caused by peptide binding and concomitant domain motion was observed (Fig. 1b). Analysis of PeDPP11:RD using Naccess³² revealed a large loss of solvent-accessible area upon peptide binding, approximately 1430 Ų.

In contrast, binding of the LDVW and LDL substrates was largely endothermic (ΔH _bind of +23.8 and +17.0 kJ.mol⁻¹, respectively) at 25 °C indicating an entropy-driven process to overcome the unfavourable enthalpic contribution [Fig. 2 (right panel), Supplementary Fig. S3b]. The binding of peptides to PgDPP11 induces the dimerization of its monomeric population (Supplementary Fig. S4), masking the real thermodynamic contributions involved in the binding process. In this way, we focused our thermodynamic analysis solely on PeDPP11.

Next, we asked what governs the opposite thermodynamic signatures observed for DPP11 binding of end products and substrates. To address this question, we dissected the contributions of the three possible components influencing the binding energetics: solvent, ligand and the protein itself. The most apparent answer would point to hydrophobic effects, which is the release of well-ordered water molecules from interfaces to the bulk solvent, resulting in system’s entropy increase upon ligand binding³³. However, our crystal structures of DPP11 in complex with LDVW and dipeptide RD are both in closed conformation, excluding the possibility that solvent released from the protein’s cleft would explain the larger increase in entropy upon substrate binding. Then, we analysed the ligand’s contribution to the process. The presence of only one additional amino acid in the peptide LDL (ΔH _bind of +17.0 kJ.mol⁻¹) compared to LD (ΔH _bind of −22.0 kJ.mol⁻¹) results in the outstanding difference of +39.0 kJ.mol⁻¹ in binding enthalpy. Due to the peptides similarity, energetic effects originating from the ligands alone do not suffice to explain the distinct thermodynamic binding forces reported. In light of the analyse above, we concluded that the major contribution for the opposite thermodynamic signatures must arise from the protein itself, via changes in conformational entropy, as demonstrated below.

In the free energy equation: ΔG _tot = ΔH _tot − TΔS _tot, the total binding entropy (ΔS _tot) is deconvoluted into the sum of changes in ΔS _conf (conformational entropy), ΔS _sol (solvation entropy) and ΔS _RT (rotational and translational entropy)³⁴. Based on experimentally-measured heat capacity changes (ΔC_p) for PeDPP11:LDVW (−1.6 kJ.mol⁻¹.K⁻¹) and PeDPP11:LD (−3.3 kJ.mol⁻¹.K⁻¹) interactions, we calculated a ΔS _sol of +417.8 and +844.9 J.mol⁻¹.K⁻¹ and a ΔS _conf of −198.2 and −775.6 J.mol⁻¹.K⁻¹, respectively (Fig. 3). The data indicate that in both binding events the solvent provides a favourable contribution to the observed entropy and shows a 3.5-fold more prohibitive change in overall ΔS _conf for PeDPP11 interaction with LD compared to LDVW. We propose that the +677.3 J.mol⁻¹.K⁻¹ difference in ΔS _conf is associated with the unfolding of DPP11 specific regions upon binding to LDVW.

Figure 3.

Thermodynamic analysis. (a) PeDPP11 binding to LD. (b) PeDPP11 binding to LDVW. Upper panels: Temperature dependence of ∆G, ∆H and −T∆S. Middle panel: Table with thermodynamic data derived from the ITC measurements at different temperatures. Lower panel: Entropy parameters estimations. Conformational entropy was calculated using the following equation: ∆S _conf = ∆S _tot − ∆S _sol − ∆S _rt ³⁴. Where ∆S _sol = ∆Cp ln (298 K/385 K)⁶⁵ and ∆S _rt is estimated using the “cratic entropy” value of −33.3 J.mol.⁻¹K^{−1  66}.

The helical domain displays a high diversity of structural states across all solved structures in this work. When compared to unbound PeDPP11, the r.m.s.d. of the helical domain is 5-fold higher than that of the catalytic domain for PeDPP11:LDVW and 2-fold higher for PeDPP11:RD and PeDPP11:RE (Supplementary Table S1a). Particularly, the unfolding of helix α14 (residues 320–346) and loop F441-K451 upon LDVW binding corroborates our hypothesis that protein conformational change is the determining factor for the opposed thermodynamics signatures observed upon peptide binding (Fig. 4a,b and Supplementary Fig. S5). We postulate that substrate binding leads to higher protein ΔS _conf, which overcompensates for the unfavourable enthalpic contribution.

Figure 4.

PeDPP11 conformational changes. (a) Close-up view of the main PeDPP11 regions that unfold upon binding to LDVW, as observed in the crystal structures. (b) Loop F441-K451 region superposition of unbound PeDPP11 (blue), PeDPP11:LDVW (magenta, dashed line) and PeDPP11:RD (green). Unbound PeDPP11 is represented as ribbons and peptide binding pocket as yellow surface. (c) Cartoon representation depicting a DPP11 helix unfolding. Upon substrate binding, energy is absorbed from the solution to break polar contacts, which causes helix destabilization. In the disordered stage, the helix accesses different structural states, increasing system entropy. (d) Close-up view of the helix α14 missing region in PeDPP11_altconf. Intra-main chain polar contacts are indicated with orange dashed lines.

Consistent with the measured endothermic binding, we propose that energy is absorbed from the solution to break key interactions, such as those stabilizing loop F441-K451 and intra-main chain polar interactions that stabilize helix α14, but possibly in additional regions of the helical domain. These events permit the motion of the helical domain between different structural states leading to increased protein ΔS _conf (Fig. 4c). Usually, protein unfolding yields a positive ΔC_p ³⁵, which explains the difference of +1.7 kJ.mol⁻¹.K⁻¹ in ΔC_p between PeDPP11:LDVW and PeDPP11:RD interactions. Upon DPP11-substrate binding, the increase in protein conformational entropy counterbalances the overall entropic costs in protein-peptide interactions (including loss of protein and peptide degrees of freedom). Interestingly, we obtained an unbound PeDPP11 crystal form, called here PeDPP11_altconf, which lacks electron density for helix α14, indicating its susceptibility to unfold (Fig. 4d). Similar to PeDPP11 complexes, this structure is also closed (rotation angle of 27° of helical domain relative to the catalytic domain), illustrating the enzyme flexibility.

Protein-peptide interactions often occur in a way that minimizes the conformational changes of the protein partner, while maximizing their enthalpic potential via its packing and formation of hydrogen bonds³⁶ (Fig. 5a). This strategy helps to decrease the entropic costs associated with the peptide loss of conformational entropy upon binding. The process can also be entropy-driven with the solvent providing the main driving-force, in this case, conformational flexibility may accompany peptide binding³⁷ (Fig. 5b). Here, increased ΔS _conf in DPP11 establishes endothermic substrate binding via partial enzyme de-structuring associated with an increase in helical domain entropy, which acts as an “entropy reservoir” (Fig. 5c). DPP11 active site design displays stereochemical specificity only for P1 and P2 positions of the ligand. This arrangement favours substrate entropy-driven binding by limiting the enthalpic contributions of protein-peptide interaction (i.e. limiting the number of polar contacts) for only the two first amino acids of the incoming peptide. Additionally, our data also illustrate how conformational plasticity enables enzyme promiscuity; for instance, by closing differently around different ligands³⁸.

Figure 5.

DPP11 conformational entropy in peptide binding. This cartoon illustrates two previously described models and DPP11 binding model reported in this work. (a) In the enthalpy-driven binding mode depicted, there are no major conformational changes and the active site is prearranged. The process is mainly governed by protein-peptide interactions, resulting in favourable enthalpy. (b) In this entropy-driven binding mode, the displacement of solvent molecules “entropy reservoir” provides the main driving-force for peptide-binding, and increases in system entropy outweighs the unfavourable enthalpy. In this case, peptide binding may be accompanied by protein conformational changes. (c) In DPP11 entropy-driven binding mode, protein conformational entropy is the main driving-force for substrate binding. De-structuring of parts of the helical domain “entropy reservoir” contributes to the increase in entropy necessary to compensate for the unfavourable enthalpy.

Due to experimental challenges, the role of conformational entropy in molecular recognition by proteins has begun to be elucidated only recently, mainly by nuclear magnetic resonance (NMR) relaxation methods³⁹. Using NMR techniques and molecular dynamics simulations, Veglia and colleagues observed in cAMP-dependent protein kinase A (PKA-C) a similar binding mode to that of DPP11. They showed that the substrate PLN_1-20 (phospholamban) binds to (PKA-C) in an entropically driven way, resulting in protein increased conformational dynamics. Conversely, binding of the inhibitor PKI_5-24 (protein kinase inhibitor) to PKA-C is enthalpically driven and stabilizes the protein, quenching the enzyme dynamics, which is important to prime the active for catalysis⁴⁰.

The structural and thermodynamics data presented here provide a distinct model for protein-protein interaction, particularly in cases where increase in protein conformational entropy significantly contributes to the free energy of binding. Together with PKA-C, DPP11 binding mode may represent a general mechanism for biomolecular recognition, allowing the identification of proteins that share similar features and that have evolved to promiscuously bind numerous ligands. These findings further provide an innovative framework for structure-based drug design to develop compounds that target the “entropy reservoirs”. For instance, molecules able to prevent the unfolding of helix 14 and loop F441-K451 could display efficient inhibitor properties. Alternatively, it is also conceivable the identification of effectors that increase catalytic power by promoting protein dynamics.

Methods

Protein expression and purification

E. coli codon-optimized genes encoding for C-terminal 6xHis-tagged PeDPP11_22-717 S652A and PgDPP11_22-720 S655A in the pET-22b(+) vector (cloning sites NdeI and XhoI) were purchased from GenScript (Piscataway, USA). The construct Flavobacterium psychrophilum DPP11 (called here FpDPP11_17-713) encoding the N-terminal fusion sequence (MGGSHHHHHHGMASMTGGQQMGRDLYDDDDKDPTL) was cloned into the expression vector pTricHis.

All plasmids were transformed in BL21(DE3)pLysS. The cells were grown in LB-medium containing 100 μg ml⁻¹ ampicillin and 34 μg ml⁻¹ chloramphenicol. After 3 h at 37 °C, temperature was reduced to 30 °C and protein expression was induced by the addition of 0.5 mM isopropyl-1-thio-D-galactopyranoside (IPTG). Cells were then allowed to grow for 4 h and were harvested by centrifugation at 4,000 g for 10 minutes. For protein purification, cells were resuspended in 50 mM Hepes-NaOH pH 8.0, 150 mM NaCl. Cell debris was removed by centrifugation at 25,000 g for 45 minutes at 4 °C, and the supernatant was subjected to affinity chromatography on 5 ml HisTrap^TM (GE Healthcare) equilibrated with lysis buffer. Bound protein was eluted in lysis buffer containing 500 mM imidazole. Further purification was performed by size exclusion chromatography (SEC) on a HiLoad 26/60 Superdex 200 (GE Healthcare) column previously equilibrated with 10 mM Hepes-NaOH pH 7.4, 100 mM NaCl. Purified protein was concentrated using 20 ml concentrators with an appropriate molecular weight cut-off (Vivaspin® 50,000 MWCO, Sartorius).

Crystallization

To enhance the crystallizability of PgDPP11 and PeDPP11, truncated forms of the enzymes were designed lacking the first 21 amino acid residues (called here PgDPP11_22-720 and PeDPP11_22-717) which were predicted to be signal peptides⁴¹. The following crystallization trials used the nanodrop-dispensing robot (Phoenix RE; Rigaku Europe) employing the sitting drop vapour diffusion technique by mixing equal volumes of protein (200 nl) and reservoir solutions (200 nl) at 20 °C in a 96-well Intelli-Plate (ArtRobbins Instruments®). All crystals were cryoprotected in a solution consisting of reservoir solution supplemented with 20% glycerol before flash-cooling in liquid nitrogen. X-ray diffraction data were collected at 100 K.

• PgDPP11 _22-720 S655A (PgDPP11) was crystallized at 10 mg ml⁻¹ using the Morpheus⁴² screen condition D11: 0.12 M alcohols, buffer system 3 pH 8.5, 40% v/v glycerol, 20% w/v PEG 4000.

• PeDPP11 _22-717 S652A (PeDPP11:RD) at 10 mg ml⁻¹ was incubated with 1.2 mM dipeptide Arg-Asp on ice for 15 minutes. Crystals were obtained in the Morpheus screen condition E10: 0.12 M ethylene glycols, 0.1 M buffer system 3 pH 8.5, 40% v/v ethylene glycol, 20% w/v PEG 8000.

• PeDPP11 _22-717 S652A (PeDPP11:RE) at 10 mg ml⁻¹ was incubated with 1.2 mM dipeptide Arg-Glu on ice for 15 minutes. Crystals were obtained in the Morpheus screen condition F12: 0.12 M monosacharides, 0.1 M buffer system 3 pH 8.5, 25% v/v MPD, 25% PEG 1000, 25% w/v PEG 3350.

• FpDPP11 _17-713 was crystallized at 10 mg ml⁻¹ in the PACT Premier screen (Molecular Dimensions®) condition G2: 0.2 M NaBr, 0.1 M bistris propane, pH 7.5, 20% PEG 3350.

The following crystallization trials used the sitting-drop vapour diffusion method and were conducted at the High Throughput Crystallization Laboratory of the EMBL Grenoble Outstation (https://embl.fr/htxlab)⁴³. Drops of 100 nl sample and 100 nl crystallization solution were set up in CrystalDirect plates (MiTeGen, Ithaca, USA) with a Cartesian PixSys robot (Cartesian Technologies, Irvine, USA). The experiments were incubated at 20 °C in a RockImager system (Formulatrix Inc., Bedford, USA). Automated high-throughput crystal cryo-cooling and harvesting were performed with CrystalDirect^TM Technology as described by Zander et al., 2016⁴⁴. Crystals were stored in liquid nitrogen for data collection. Data collection was done in a fully automated fashion at MASSIF-1, ESRF⁴⁵. X-ray diffraction data were collected at 100 K.

• PeDPP11 _22-717 S652A (unbound) was crystallized at 10 mg ml⁻¹ initially in the condition D11 of ProComplex screen (Qiagen®). The condition was further optimized to 0.1 M Tris-HCl pH 7.5, 15% PEG 6000.

• PeDPP11 _22-717 S652A in the alternate conformation (PeDPP11 _altconf ) was crystallized initially in the condition B2 of ProComplex screen (Qiagen®): 0.1 M calcium acetate, 10% w/v PEG 4000, 0.1 M sodium acetate pH 4.5. The condition was further optimized to: 0.1 M calcium acetate, 15% w/v PEG 4000, 0.1 M sodium acetate pH 5.0.

• PeDPP11 _22-717 S652A (PeDPP11:LDVW) at 22 mg ml⁻¹ was incubated for 30 minutes on ice with 1.0 mM Leu-Asp-Val-Trp. Crystals were initially obtained in condition D11 of ProComplex screen (Qiagen®). The condition was further optimized to 0.1 M Tris-HCl pH 7.5, 15% PEG 6000.

Structure determination

Our initial attempts to solve the structure by molecular replacement using the coordinates of dipeptidyl aminopeptidase BII (DAP BII)⁴⁶ from Pseudoxanthomonas mexicana WO24, the closest homologue (37% identity) with known 3D structure, were ineffective. We suspected that different conformations adopted by the protein in the crystal could be rendering molecular replacement trials unsuccessful. So, a DPP11 homologue (37% identical to PgDPP11) from FpDPP11_17-713 was employed to grow monoclinic crystals (space group P2₁). FpDPP11_17-713 structure was then determined by molecular replacement using the coordinates of DAP BII (PDB code: 3WOJ, 27% identity). Subsequently, FpDPP11_17-713 structure was successfully used as template to solve the structures of PgDPP11_22-720 S655A and PeDPP11_22-717 S652A (Supplementary Fig. S6). Two out of four subunits in the crystal asymmetric unit of FpDPP11_17-713 were in complex with Arg-Asp, a dipeptide co-purified from E. coli. This finding led us to grow co-crystals of PeDPP11_22-717 S652A in complex with dipeptides Arg-Asp and Arg-Glu.

Data were processed with XDS⁴⁷, Scala⁴⁸ and Pointless⁴⁹. All structures were solved by molecular replacement with PHASER⁵⁰, refined with PHENIX⁵¹ and manually adjusted in COOT⁵². R_free-values⁵³ were computed from 5% randomly chosen reflections not used for the refinement. The structure stereochemistry was checked using Molprobity⁵⁴. Details of data collection and refinement statistics are provided in Table 1. Peptide omit maps are depicted in Supplementary Fig. S7. All figures were prepared using the program PyMOL (http://www.pymol.org). Poisson-Boltzmann calculations were performed using the software APBS^{55, 56}.

Isothermal Titration Calorimetry

All experiments were carried out in 10 mM Hepes-NaOH pH 7.4, 100 mM NaCl. Both the enzymes and the peptides were dissolved in the same buffer. The bindings were analysed with a MicroCal^TM iTC₂₀₀ microcalorimeter (GE Healthcare, Life Sciences) equilibrated at the respective temperature. Typically, a total of one aliquot of 0.4 μl and 19 aliquots of 2.0 μl of the peptide solution were injected at a rate of 0.5 μl/s into 200 μl of the protein solution under constant stirring at 750 rpm at the specified concentrations. The following titrations were performed: 600 μM LDVW to 60 μM PeDPP11, 810 μM LD to 75 μM PeDPP11 (at 10 °C, 1.28 mM LD to 120 μM PeDPP11 was employed), 1 mM of LDL to 75 μM PeDPP11, 1.0 mM of RD to 80 μM M PeDPP11. As a control to exclude buffer-dependent effects, we additionally performed the binding of 600 μM LDVW to 60 μM PeDPP11 and 600 μM LD to 75 μM PeDPP11 in 50 mM sodium phosphate pH 7.4 with 100 mM NaCl (Supplementary Fig. S8). Every injection was carried out over a period of 4 s with a spacing of 110 s between the injections. The corresponding heats of binding were determined by integrating the observed peaks after correcting for the heat dilution of the peptide determined in a reference measurement (peptide injected into buffer). These corrected values were plotted against the ratio peptide vs. protein concentration in the cell to generate the binding isotherm. Nonlinear least-squares fitting using Origin version 7.0 (Microcal) was used to obtain the association constants (K _a), heats of binding (ΔH) and stoichiometries. K _d and Gibbs free energy (ΔG) were calculated according to: K _d = 1/K _a and ΔG = −RT ln K _a = RT ln K _d. The reported values are averages of at least two independent measurements. The stoichiometry obtained in all experiments is within the range 0.7–1.1, which is in agreement with the crystal structures (stoichiometry 1).

Small-angle X-ray scattering (SAXS) data collection and analysis

SAXS experiments were performed at 0.9918 Å wavelength ESRF at BioSAXS beamline BM29 (Grenoble, France) equipped with PILATUS 1 M⁵⁷. The detector distance was set at 2.864 m. The range of scattering vector 0.03 nm⁻¹ < q < 4.5 nm⁻¹ was covered. For PeDPP11_22-717 S652A, the data were collected using the following protein concentrations: 1.1 mg ml⁻¹, 1.9 mg ml⁻¹, 9.62 mg ml⁻¹ and 16.37 mg ml⁻¹.

The samples were in a buffer containing 10 mM Hepes-NaOH pH 7.4, 100 mM NaCl, and the measurements were performed at 20 °C. The automated sample changer⁵⁸ was employed to load the samples and constantly remove the irradiated sample. Twenty successive exposures of 1 second were collected and compared to detect and discard possible radiation damage effects. For PgDPP11_22-720 S655A, on-line HPLC-mode was used with a Superdex^TM 200 10/300 GL column (GE Healthcare), and SAXS data was recorded directly on the sample eluted. A 0.5 ml min⁻¹ flow was used. The protein concentration applied onto the column was 50 mg ml⁻¹.

The data were processed and analyzed using the online analysis pipeline⁵⁹. Subsequent manual processing was done with the ATSAS 2.6 program package⁶⁰. The forward scattering I(0) and the radius of gyration Rg were extracted from the Guinier approximation calculated with the AutoRG function within PRIMUS^{61, 62}. The maximum particle dimension Dmax and P(r) function were evaluated using the program GNOM⁶³. For PeDPP11_22-717 S652A, the analysis of SAXS data by Guinier approximation showed no concentration dependence effect, indicating the samples were homogeneous and free of aggregation. For this construct, SAXS analyses were performed by merging data from all concentrations measured. For both PeDPP11_22-717 S652A and PgDPP11_22-720 S655A, the theoretical scattering from the crystallographic structures was calculated using the program CRYSOL⁶⁴ and compared with the respective scattering profiles.

Size exclusion chromatography followed by Multiangle Laser Light Scattering

To assess the oligomeric state and molecular weight, the samples were applied onto a Superdex^TM 200 10/300 GL column (GE Healthcare) at the respective concentrations, using a flow of 0.5 ml/min. The column was connected to a miniDAWN Tristar light scattering instrument (Wyatt Technologies, Santa Barbara, CA) and pre-equilibrated with 10 mM Hepes-NaOH pH 7.4, 100 mM NaCl. Data analysis was performed using the manufacturer’s software ASTRA.

Hydrolyzing activity toward MCA-dipeptides

Purification of recombinant active forms of PgDPP11 and PeDPP11 was performed according to Ohara-Nemoto et al.²⁴. PgDPP11 and PeDPP11 (2-20 ng) were used for measurement of dipeptidyl peptidase activity in 200 μl of reaction solution composed of 50 mM sodium phosphate pH 7.0 and 5 mM EDTA. The reaction was started with an addition of 20 μM Leu-Asp-, Arg-Asp or Leu-Glu-MCA and continued at 37 °C for 30 min (Supplementary Fig. S9). Fluorescence intensity was measured with excitation at 380 nm and emission at 460 nm with a Fluorescence Photometer F-4000 (Hitachi).

Author Contributions

G.A.B., Y.O.N., S.F., I.C., G.H. and T.K.N. conceived and performed experiments; G.A.B. performed structure determination and ITC measurements; G.A.B., Y.O.N., S.F., A.R., I.C., J.A.M., T.K.N., K.D.C. analysed the data; G.A.B. wrote the manuscript with comments from all authors.

Competing Interests

The authors declare that they have no competing interests.