Biochimica et Biophysica Acta 1859 (2017) 1859–1871

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

journal homepage: www.elsevier.com/locate/bbamem

Crystallographic and biochemical characterization of the dimeric architecture of site-2 protease

Magdalena Schacherl a,⁎,1, Monika Gompert a, Els Pardon b,c, Tobias Lamkemeyer d,2, Jan Steyaert b,c, Ulrich Baumann a a Institute of Biochemistry, University of Cologne, Otto-Fischer-Str. 14, 50674 Cologne, Germany b Structural Biology Brussels, Vrije Universiteit Brussel (VUB), 1050 Brussels, Belgium c Structural Biology Research Center, VIB, 1050 Brussels, Belgium d Cluster of Excellence in Cellular Stress Responses in Aging-associated Diseases (CECAD), University of Cologne, Cologne, Germany article info abstract

Article history: Regulated intramembrane proteolysis by members of the site-2 protease family (S2P) is an essential signal trans- Received 22 June 2016 duction mechanism conserved from bacteria to humans. There is some evidence that extra-membranous Received in revised form 8 May 2017 domains, like PDZ and CBS domains, regulate the proteolytic activity of S2Ps and that some members act as di- Accepted 10 May 2017 mers. Here we report the crystal structure of the regulatory CBS domain pair of S2P from Archaeoglobus fulgidus, Available online 11 May 2017 AfS2P, in the apo and nucleotide-bound form in complex with a specific nanobody from llama. Cross-linking and SEC-MALS analyses show for the first time the dimeric architecture of AfS2P both in the membrane and in Keywords: fi Site-2 protease detergent micelles. The CBS domain pair dimer (CBS module) displays an unusual head-to-tail con guration CBS domain and nucleotide binding triggers no major conformational changes in the magnesium-free state. In solution, Intramembrane cleavage MgATP drives monomerization of the CBS module. We propose a model of the so far unknown architecture of Nanobody the transmembrane domain dimer and for a regulatory mechanism of AfS2P that involves the interaction of pos- Complex itively charged arginine residues located at the cytoplasmic face of the transmembrane domain with the nega- Dimer tively charged phosphate groups of ATP moieties bound to the CBS domain pairs. Binding of MgATP could promote opening of the CBS module to allow lateral access of the globular cytoplasmic part of the substrate. © 2017 Elsevier B.V. All rights reserved.

1. Introduction cleaving proteases (I-CLiPs) [3]. Thus, it differs radically, for example, from the ATP-dependent transmembrane proteases FtsH or Lon, Proteolysis is involved in the regulation of a variety of fundamental which cleave substrates outside of the lipid bilayer [4–6]. Contrary to cellular processes in living cells, e.g. antigen presentation, protein qual- proteolytic enzymes operating in an aqueous environment, I-CLiPs ity control, apoptosis and stress responses [1]. Thus, proteolysis itself face the mechanistic problem of delivering a water molecule within has to be tightly regulated and balanced with protein synthesis. Besides the hydrophobic membrane to the scissile peptide bond. Therefore, gat- the large variety of soluble proteases, there also exist membrane bound ing mechanisms allowing water and substrate polypeptide to access the proteases. One of them is the site-2 protease (S2P; MEROPS: M50 [2]), a active site have been discussed [7,8], allowing at the same time regula- Zn2+-metalloprotease that cleaves its substrates within the membrane tion of proteolytic activity. plane and thus belongs to the class of the so-called intramembrane Site-2 proteases are conserved in all kingdoms of life. In bacteria they are involved in the regulation of sporulation [9,10],inthetransductionof stress signals [11] or release of peptide pheromones [12].Inmammals Abbreviations: CBS, cystathionine-beta-synthase; CBSD, cystathionine-beta-synthase S2Ps regulate cholesterol synthesis [13] and activate the ER stress re- domain pair; S2P, site-2 protease; Nb, nanobody; CDR, complementarity-determining re- sponse [14]. S2Ps are also found in plants, fungi and protozoa and are gion; SEC-MALS, size exclusion chromatography with multiangle light scattering. ⁎ Corresponding author at: Institute of Biochemistry, University of Cologne, Otto- highly abundant in archaea. For the latter, substrates are mostly un- Fischer-Str. 12-14, 50674 Cologne, Germany. known. S2Ps cleave their α-helical substrates, which are mostly transcrip- E-mail addresses: [email protected], tion factors (e.g. SREBP or ATF-6), in the water-excluding environment of [email protected] (M. Schacherl). the membrane and must therefore unwind their α-helical substrates and 1 Present address: Center of Advanced European Studies and Research (CAESAR), deliver water to the active site. Surprisingly, the residues crucial for catal- Ludwig-Erhard-Allee 2, 53175 Bonn, Germany. 2 Present address: Federal Institute for Drugs and Medical Devices, Kurt-Georg- ysis were found to be the same as in soluble proteases, contained in the Kiesinger-Allee 3, 53175 Bonn, Germany. HEXXH and the NPDG motives (with the three zinc-ligands in bold and

http://dx.doi.org/10.1016/j.bbamem.2017.05.006 0005-2736/© 2017 Elsevier B.V. All rights reserved. 1860 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 the catalytic base in italics) [15]. The crystal structure of the transmem- 2. Material and methods brane domain of an archaeal site-2 protease from Methanocaldococcus janaschii (MjS2P) shows six transmembrane helices (TMH), with TMHs 2.1. Materials α1andα5–α6 are considered to constitute the gate domain [8].Theac- tive site residues are located on TMHs α2andα4, respectively. In the crys- All detergents were purchased from Anatrace and were of tal structure of MjS2P two different states, open and closed, were Anagrade® quality. All other chemicals were purchased from Sigma- observed. In the open state TMH α1andα6 (gate domain) are thought Aldrich, AppliChem or Merck, if not stated otherwise. Enzymes for clon- to move away from each other by 10–12 Å to allow lateral access of the ing were from New England Biolabs and DNA purification kits from substrate helix. From a structural point of view, the open structure Sigma-Aldrich. could represent the catalytically competent state, and in analogy the closed state would be the inactive one. 2.2. Methods S2P take their name from their function: they are setting a cut at site- 2 of a substrate in a cascade that requires prior cleavage at site-1 [16]. 2.2.1. Cloning and protein expression of AfS2P constructs The latter is accomplished by the site-1 protease (S1P), a serine protease Using genomic DNA from Archaeoglobus fulgidus (DSM 4304, DSMZ, located in the periplasm of bacteria or the ER-lumen of eukarya. S1P Braunschweig, Germany) the af0332 gene was PCR amplified and cleaves a substrate just outside the membrane plane and prepares it subcloned into pET-based vectors (used oligonucleotides, restriction for the S2P that can only cleave the shed substrate [17]. enzymes and vectors in Table S1). In this manner, constructs for the There are three subtypes of S2Ps differing in membrane topology full-length protein (AfS2P) and the CBS domain (AfCBSD) only were and domain architecture. Most of the archaeal S2Ps display six trans- generated. membrane helices and contain a CBS (cystathionine-β-synthase) do- Expression was performed with the BL21 (DE3) derivative strains main pair facing the cytoplasm. Most bacterial and mammalian S2Ps C43 (DE3) for full-length protein and BL21 (DE3) pLysS (DE3) for the contain one or several extracytoplasmic PDZ domains and possess four AfCBS domain. A single colony of freshly transformed E. coli cells was in- or more transmembrane helices. Some S2Ps display no additional do- oculated in 50 ml lysogeny broth (LB) with appropriate antibiotics mains beside the transmembrane domain (TMD). Regarding the impor- (Tables S1 and S2) per liter of culture and grown in an orbital shaker tant role S2Ps play, their activity must be strictly controlled. Regulation overnight (~16 h) at 30 °C and 220 rpm. The overnight culture was in- may be exerted via adaptor proteins analogous to the ATP-driven prote- oculated into fresh LB to an optical density at 600 nm (OD600) of 0.1 ases [18] or via the PDZ-domains in the relevant S2Ps, e.g. Rip1 [19].For and further incubated at 37 °C to an OD600 of 0.6. Gene expression those S2Ps that possess CBS domains, a regulatory role of the latter ap- was induced with 1 mM IPTG (isopropyl-β-D-thiogalactopyranoside) pears plausible. CBS domains are known to bind nucleotides like ATP, followed by incubation at 20 °C overnight. Cells were harvested by cen- AMP or S-adenosyl-derivatives and are implicated in protein regulation trifugation (7000 ×g, 4 °C, 20 min), washed with buffer (20 mM Tris [20]. For Bacillus subtilis S2P SpoIVFB, which contains such a CBS do- pH 8.0, 300 mM NaCl), and after a second centrifugation step cell pellets main, a direct activation of proteolytic activity by ATP was demonstrat- were stored at −80 °C until further use. ed [21]. However, it is not clear how exactly ATP influences the conformation of SpoIVFB and its binding to the substrate pro-σK. Bio- 2.2.2. Selection, cloning and expression of llama nanobodies chemical data suggest that ATP-binding influences the oligomerization Llama immunization was performed according to previously de- of SpoIVFB and this allows pro-σK to gain access to its active site [21]. scribed protocols [25,26]. In summary, a llama was injected several The estimated oligomeric state of SpoIVFB after solubilization from times with in total 2 mg of dimeric AfS2P-His in DDM over 6 weeks. membranes is a tetramer. For all other S2P it is unclear, if they act as Phage display library construction and selection have been performed monomers or if they oligomerize analogous to other intramembrane following procedures described by Conrath et al. [27] with minor mod- proteases like for example rhomboids [22]. For CBS-domain containing ifications: total RNA was extracted from the peripheral blood lympho- S2Ps however, dimerization or even tetramerization would make cytes [28],50μg of total RNA was used to prepare cDNA using sense, as CBS domains are often found in these configurations in soluble SuperScript II (Invitrogen) and a dN6 primer according to the manufac- enzymes [23]. CBS domains often appear as tandems of two (CBS1/2), tures instruction [26]. Subsequent to library generation, nanobodies also called Bateman modules or CBS domain pairs, or even four copies were selected on solid phase coated AfS2P-His using phage display. (CBS1–4) on one polypeptide chain [24]. Phages were recovered by incubating the AfS2P-His-coated wells with This lack of insight results to a large part from a shortage on structur- 100 mM triethylamine (pH 10) for 10 min. The wells were washed al information for S2Ps. The only currently available experimental once with Tris-HCl (pH 6.8) and several times with PBS, and freshly three-dimensional structure is the above-mentioned transmembrane grown TG1 cells were added to the wells to recover the non-eluted moiety of the archaebacterial member MjS2P. However, this structure phages. After 2 rounds of selections, 96 clones were screened in ELISA does not comprise the two regulatory CBS domains CBS1 and 2 that and positive clones were sequenced, revealing 8 nanobody families. Fi- are present in this protein. Furthermore, it is unclear if MjS2P nally, all selected nanobody genes were recloned in the pHEN6 vector oligomerizes and how a possible dimer of S2P would look like, as in for expression with a C-terminal His-tag in E. coli [27]. the crystal structure only an inverted dimer was found that is a crystal- Nanobodies binding AfS2P were produced and purified in mg quan- lization artifact. tities from WK6(su−) E. coli strain [29] as described previously [26]. In order to gather more insight into the quaternary structure and regulation of S2Ps, we have investigated S2P from the thermophilic 2.2.3. AfS2P protein purification archaeon Archaeoglobus fulgidus (AfS2P) that displays analogous to Cells were resuspended in lysis buffer (20 mM Tris-HCl pH 7.5, MjS2P a CBS domain pair at its cytoplasmic side. We report here the 300 mM NaCl) with addition of 1 mM PMSF and 10 μg/ml DNaseI crystal structures of the apo and nucleotide-bound form of the AfS2P- (Roche), passed twice through a TS 0.75 (Constant Systems) cell CBS domain pair (AfCBSD), its orientation with respect to the protease disruptor at 2.5 kbar and centrifuged (10,000 ×g, 4 °C, 15 min) to re- domain and the architecture of the protease dimer within the mem- move cell debris. The supernatant was centrifuged using an ultracentri- brane. Monomerization of the regulatory CBS domain pair by MgATP fuge (250,000 ×g, 4 °C, and 60 min) to collect the membranes and positively charged residues at the cytoplasmic face of the trans- containing recombinant AfS2P. Membranes were solubilized in 20 mM membrane domain located near the negatively charged phosphate Tris pH 7.5, 300 mM NaCl, 10 mM imidazole, 5% glycerol (buffer W10) groups of AfCBSD-bound ATP-molecules hint at a possible ATP-driven in a ratio of 1:9 (w/v) first by passing the collected membranes through regulation of the proteolytic activity of AfS2P. a 21-gauge needle with a subsequent addition of β-D-dodecylmatoside M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1861

(DDM) to 1% final concentration and incubated on a rotary shaker for Optilab rEX (refractive index measurement, both Wyatt Technology) 60 min at 4 °C. Insoluble membrane proteins and particles were separat- and a Superose 6 10/300 GL column operated with a HPLC system (Wa- ed by a second ultracentrifugation step. ters) was first equilibrated with 20 mM Tris pH 7.5, 150 mM NaCl, 5% In the case of His-tagged proteins, the supernatant of the second ul- glycerol, 0.02% sodium azide, 0.03% DDM for several hours, until the tracentrifugation was diluted threefold with buffer W10 to lower the base line of the Optilab rEX signal was stable. 50 μgBSA(bovine detergent concentration and applied on a Ni-NTA superflow column serum albumin, Sigma-Aldrich) were applied on the column to test (Qiagen) equilibrated with buffer W10 supplemented with 0.03% the stability of the measurement. Subsequently 50 μg AfS2P-His were DDM. The column was washed with buffers containing increasing imid- applied on the column and a three-angle measurement was performed azole concentrations from 10 mM to 70 mM. Bound protein was eluted to determine the molecular weight of the particles (settings of the mea- with buffer containing 250 mM imidazole. All buffers were supplement- surements in Table S3). Data were evaluated using the software ASTRA ed with 0.03% DDM. 5.3.2 by plotting the molar mass vs. volume and the Rayleigh ratio vs. For Strep-tagged proteins, membranes were solubilized in buffer W volume. (20 mM Tris pH 7.5, 300 mM NaCl), the supernatant applied on an equil- For the comparison of ATP- and MgATP-bound forms of AfS2P-His, a ibrated Strep-Tactin superflow column (IBA) and washed with buffer W system was used composed of an 18-angle Heleos-II MALS device with containing detergent until A280 reached the base line. Bound proteins WyattQELS, Optilab T-rEX (both Wyatt Technology) and an AdvanceBio were eluted with buffer W supplemented with 2.5 mM desthiobiotin. Sec 300 Å (4.6 mm id) column operated with a 1260 Infinity II HPLC sys- Eluates were concentrated in Amicon centrifugal filter units tem (both Agilent Technologies). Samples (10 μg AfS2P-His) were (Millipore) with a nominal molecular weight cutoff of 100 kDa to a vol- mixed with 1 mM ATP (final concentration) or 1 mM ATP, 1.5 mM ume of max. 4 ml and applied on a Superdex 200 pg HiLoad 16/600 col- MgCl2 and incubated for 1 h. The column was operated in 20 mM Tris umn (S200, GE Healthcare), or to 0.5 ml and applied on a Superose 6 10/ pH 7.5, 150 mM NaCl, 5% glycerol, 0.02% sodium azide, 0.03% DDM

300 GL column (GE Healthcare), respectively. Size exclusion chroma- with 1 mM ATP or 1 mM ATP, 1.5 mM MgCl2, respectively. Data were tography (SEC) was performed in 20 mM Tris pH 7.5, 150 mM NaCl, evaluated using the software ASTRA 7.1.0.29 as above. 5% glycerol, 0.02% sodium azide and 0.03% DDM. Fractions of the elution peaks were collected and analyzed on 12–15% SDS-PAGE gels [30] with- 2.2.8. Limited proteolysis and EDMAN sequencing sample preparation out boiling of the samples. Protein concentrations were determined Limited proteolysis experiments were performed using the con- using the BCA assay (Thermo) according to the manufacturers protocol. struct AfS2P-His. Protein samples (1.4 mg each) were digested with ei- ther trypsin (type III from bovine pancreas), proteinase K, Glu-C, Lys-C 2.2.4. AfCBSD protein purification (all Sigma-Aldrich), or chymotrypsin (TLCK-treated, Fluka), in a 1:100 Cells were resuspended and processed as described for the full- (w/w) ratio at 30 °C overnight in 20 mM Tris pH 7.5, 300 mM NaCl, 5% length protein. The lysate was centrifuged (250,000 ×g, 4 °C, 60 min) glycerol, 0.03% DDM. After proteolytic digestion samples were incubat- to remove cell debris and membranes. His-tagged AfCBSD was purified ed for 30 min with specific inhibitors (200 μM TLCK for Glu-C and Lys-C; using a HIS-Select HF Nickel Affinity Gel column (Sigma-Aldrich) equil- 5 mM PMSF and 1 mM EDTA for chymotrypsin, trypsin and proteinase ibrated with buffer C10 (20 mM Tris pH 7.5, 150 mM NaCl, 10 mM im- K). Some 30 μg of digested protein were mixed with 2× EDMAN sample idazole) and the column was washed until A280 reached the base line. buffer (1.8 mM Tris, pH 6.8, 1% saccharose, 0.2% SDS, 0.0004% Subsequently the protein was eluted with buffer C10 containing bromophenol blue, 0.2% β-mercaptoethanol) and subjected to a 250 mM imidazole until the absorption reached a steady line. For re- glycine-free SDS-PAGE (running buffer 100 mM Tris, 100 mM Tricine, moval of the N-terminal His-tag (construct His-AfCBSDΔC13) the pro- pH 8.25, 0.1% SDS). The gels were blotted onto PVDF membranes by tein was incubated over night with 2 U Thrombin (Sigma-Aldrich) per electro transfer (transfer buffer 50 mM boric acid/NaOH, pH 9.0, 20% mg protein at 4 °C. methanol, 0.1% SDS) for EDMAN sequencing [31]. The membrane was For Strep-tagged AfCBSD, cells were lysed in buffer C (20 mM Tris stained in 0.25% Coomassie R-250, 45% methanol, 9% acetic acid and pH 7.5, 150 mM NaCl), the supernatant was applied on an equilibrated destained in 50% methanol, 10% acetic acid and the corresponding pro- Strep-Tactin superflow column (IBA) and washed with the same buffer tein bands excised and subjected to EDMAN N-terminal sequencing at until A280 reached the base line. Bound proteins were eluted with buff- the Analytical Research and Services, University of Bern, Department er C supplemented with 2.5 mM desthiobiotin. of Chemistry and Biochemistry, Switzerland. Eluates were concentrated and applied on a S200 16/600 column equilibrated in 20 mM Tris pH 7.5, 150 mM NaCl, 0.02% sodium azide. 2.2.9. Modeling and dimer docking The model of AfS2P-TM (transmembrane domain) was computed 2.2.5. Llama nanobody purification using Modeller 9.12 [32] with the MjS2P transmembrane domain struc- Llama nanobodies were purified according to established protocols ture as template [8] (PDB ID: 3B4R). [26,27]. After Ni-NTA purification of the His-tagged nanobodies, SEC Possible intramembrane dimers were computed using the ClusPro was performed at 4 °C in 20 mM Tris pH 7.5, 150 mM NaCl, 0.02% sodi- 2.0 server (http://cluspro.bu.edu) [33–36]. The model of the closed um azide using a S200 16/600 column. Protein samples were stored for form of MjS2P (PDB ID:3B4R; chain B) was used [8]. Balanced models long term at 4 °C. Protein concentration was determined using the re- (calculated using the balanced scoring function; [36]) were selected spective extinction coefficient at 280 nm. and only dimers were taken into account that showed the same orienta- tion of their monomers within a putative lipid bilayer. Then the mono- 2.2.6. Purification of the AfCBSD-nanobody complex mers of AfS2P-TM were structurally aligned to the MjS2P monomers The dimeric fraction of AfCBSD-Strep was mixed in a 1:1.5 M ratio and the distance between the C-terminal parts of helix α6 was mea- with Nb330, incubated for 10 min at 4 °C, concentrated to a volume of sured. Models with 25–40 Å of distance were selected, as this lies in 5 ml or less using a 10 kDa NMWC Amicon centrifugal filter unit. To re- the distance range between the N-termini of the AfCBSD dimer (36 Å), move excess nanobody a SEC step using an S200 16/600 column in which would be covalently linked to the very C-termini of helices α6 20 mM Tris pH 7.5, 150 mM NaCl, 0.02% sodium azide, was employed. in a full-length AfS2P dimer.

2.2.7. Size exclusion chromatography with multi angle light scattering 2.2.10. Crosslinking and MS-analysis (SEC-MALS) Isolated membranes from E. coli C43 (DE3) cells overexpressing Multiangle static light scattering measurements were performed AfS2P-His were homogenized in PBS (resulting protein concentration using a system composed of a miniDAWN Tristar MALS device and 1 mg/ml) and membrane proteins were cross-linked with 1 mM and 1862 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871

10 mM dimethyl suberimidate (DMS) for 5, 15, 120 min and 16 h at 4 °C, Before setting up the experiment, the capillary (0.3 mm inner diameter, 20 °C and 37 °C, respectively. Samples were either nucleotide-free, or 60 mm length) was filled with AfCBSD-Nb330 complex concentrated to contained 1 mM ATP or 1 mM ATP, 1.5 mM MgCl2. DMS was quenched 10 mg/ml. The capillaries were separated from the liquid precipitant so- with 100 mM Tris pH 8.0 for 5 min, and membranes were solubilized lution by a 1 cm layer of 1% low melting agarose (LMA; Fig. S10A), with addition of DDM to a final concentration of 1%. Insoluble material through which the precipitants slowly diffused into the capillary, drag- was cleared at 200.000 ×g and supernatants were analyzed in anti- ging some of the LMA molecules with them. In this manner, the crystals penta His immunoblots. could grow within the LMA matrix and did not touch the glass surface of For MS-analysis isolated membranes (protein concentration the capillaries during growth. By blowing out the LMA matrix all crys- 9 mg/ml) from E. coli C43 (DE3) cells overexpressing AfS2P-Strep tals could be harvested from the capillary (Fig. S10B, C). Crystals were were homogenized in PBS and membrane proteins were cross-linked flash frozen in 0.1 M Bis-Tris pH 6.0, 0.2 M (NH4)2SO4, 16% PEG 3350, with 8 mM dimethyl suberimidate (DMS) for 15 min at 20 °C. DMS 25% sucrose. was quenched with 100 mM Tris pH 8.0, and membranes were solubi- For the nucleotide-bound form of AfCBSD-Nb330, ATP was added lized with addition of DDM to a final concentration of 1%. Cross-linked immediately after elution from the affinity column to a concentration AfS2P-Strep protein was purified as described above for native AfS2P- of 1 mM and was maintained at this concentration during size exclusion Strep. 25 μg of dimeric SEC-purified protein was precipitated by addi- chromatography. Crystals were obtained at a protein concentration of tion of 4 volumes (vol) of ice-cold acetone, incubated for 20 min at 10 mg/ml in 0.1 M imidazole pH 7.25, 0.2 M MgCl2, 12% ethanol with −80 °C and pelleted for 10 min at 16.000 ×g and 4 °C. Pellets were asizeof5×10×40μm3. Crystals were flash frozen in 0.1 M imidazole washed with 1 volume of ice-cold acetone, dried under the fume hood pH 7.25, 0.2 M MgCl2, 12% ethanol, 30% sucrose. for 10 min and solubilized in 20 μl50mMNH4HCO3, 0.2% ProteasMax Diffraction data of the apo form were collected at the Swiss Light surfactant (Promega) for trypsin-digest or 20 μl 100 mM Tris pH 7.8, Source, PSI, Villigen, Switzerland on beamline X06DA using a Pilatus

10 mM CaCl2, 0.2% ProteasMax for chymotrypsin-digest, respectively. 2M detector, and of the nucleotide-bound form on beamline X06SA After 1 h at 25 °C and 700 rpm, the same buffers were filled up to 100 using a Pilatus 6M detector. All data were processed using the XDS soft- μl and supplemented with ProteasMax to 2% final concentration. Modi- ware package [39]. Data collection and refinement statistics are summa- fied trypsin or chymotrypsin (both sequencing grade, Promega) were rized in Table 1. added in a 1:50 ratio (w/w) and incubated o/n at 37 °C. The samples were supplemented with formic acid to a final concentration of 0.5% 2.2.13. Structure determination and refinement

(final pH 4–5) and centrifuged for 10 min at 16.000 ×g and 4 °C. C18 The apo structure was determined by molecular replacement (MR) StageTips (3M) were prepared and loaded with sample as described with the program Phaser [40] implemented in the Phenix package in [37]. Peptides from 25 μg initial protein were finally contained in 10 [41] in a two-step procedure. First a full rotational and translational μl 0.5% acetic acid. Peptides (2 μl) were separated using a 90-min nLC- search was performed with the structure of a DHFR-directed nanobody MS/MS gradient on a 15-cm C18 column (Dr. Maisch) and sprayed (PDB ID: 4FHB) [42] as search model, which was manually truncated in into a Q Exactive Plus mass spectrometer (Thermo) at the CECAD Prote- all three CDR regions. This yielded a partial solution explaining the elec- omics Facility, Cologne, Germany. The settings for peptide fragmenta- tron density of the half of the asymmetric unit with a translation func- tion were as follows. An MS1 scan (resolution, 70,000; scan range, tion Z-score (TFZ) score of 11 and log-likelihood gain (LLG) of 100. For 300–2000 m/z) was followed by 20 MS2 scans acquired in the Orbitrap the AfCBSD moieties, none of the tested models from protein structures (resolution, 75,000; isolation window, 3 m/z) using higher energy colli- with N20% sequence identity yielded a molecular replacement solution. sional dissociation fractionation. Dynamic exclusion was enabled (10 s). Thus, an improved model was prepared with phenix.sculptor [43] using Cross-linked peptides were identified using the StavroX software [38]. the structure of the CBS-domain protein MJ0100 (PDB ID: 3KPB) [44] covering 50% of the AfCBSD sequence (sequence identity of 45%). With 2.2.11. Fluorescence polarization assays this model and the partial solution from the first MR run, a good solution The affinity of AfCBSD variants towards ATP in absence and presence for the composite structure was obtained (TFZ score of 13.8, LLG of 868). of Mg2+ was determined employing fluorescence polarization of γ-(6- The rest of the AfCBSD molecule and the three CDRs of the Nb were built Aminohexyl)-ATP-ATTO-495 (γ-6AH-ATP-ATTO-495; Jena Bioscience) manually using Coot [45] with alternating iterative cycles of refinement in a Synergy H4 plate reader (Biotek). Serial dilutions (1:1) of AfCBSD- with phenix.refine [41]. For cross validation, a randomly selected subset Strep wild type and R343E variant in TBS or TBS supplemented with (5%) of reflections was set aside in order to monitor the progress of re-

2.5 mM MgCl2 starting from 2 mM protein were prepared and mixed finement using the free R factor [46]. Finally, the model could be refined with 400 nM of γ-6AH-ATP-ATTO-495. After equilibration at room tem- to an Rwork/Rfree of 20.6%/26.5%. The structure of the nucleotide-bound perature for 10 min fluorescence polarization was measured with exci- form was solved by MR with Phaser with a TFZ score of 9.9 and a LLG tation at 485 ± 20 nm and emission at 528 ± 20 nm. The G-value of 0.87 of 736 using the apo structure as model. The model was refined as de- was used. The mPOL values were determined automatically using the scribed for the apo form to an Rwork/Rfree of 24.8%/29.9%. The final struc- GEN5 software (v2.01, BioTek) and the equilibrium dissociation con- tures were validated using MolProbity [47],WHATCHECK[48] and the stants KD were fitted using a four-parameter fit with the ORIGIN soft- corresponding tools in Coot. ware (OriginLab). 2.3. Accession numbers 2.2.12. Crystallization of the AfCBSD-Nb330 complexes and data collection The complex of AfCBSD-Nb330 apo was initially crystallized by The atomic coordinates and structure factors have been deposited in sitting drop vapor diffusion at 293 K in a condition composed the Protein Data Bank [49] under the accession numbers PDB ID: 5G5R of 0.1 M Bis-Tris pH 6.5, 0.2 M (NH4)2SO4, 20% PEG 3350. Crystals (apo AfCBSD-Nb330) and PDB ID: 5G5X (nucleotide-bound AfCBSD- (20 × 10 × 10 μm3) were flash frozen in liquid nitrogen in 0.1 M Bis- Nb330).

Tris pH 6.5, 0.2 M (NH4)2SO4, 20% PEG 3350, 20% glycerol and diffraction data collected at beamline X06DA (Swiss Light Source, PSI, Villigen, 3. Results and discussion Switzerland). Due to the low resolution diffraction properties various crystal optimization techniques were applied. Only the counter diffu- 3.1. The isolated soluble CBS domain of AfS2P (AfCBSD) sion technique using Granada Domino boxes (Triana) in combination with 0.14 M Bis-Tris pH 6.0, 0.28 M (NH4)2SO4 and 19% PEG 3350 The following crystal structures were determined in this study: yielded crystals of good quality (diffraction up to 2.3 Å resolution). AfCBS domain pair (AfCBSD) in complex with Nb330 without bound M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1863

Table 1 Statistics of diffraction data collection and structure refinement

apo Nucleotide-bound

Crystal parameters

Space group P6122 P6122 Unit-cell parameters (Å, °) a = b = 68.44, c = 196.8, α = β = 90, γ = 120 a = b = 70.91, c = 201.44, α = β = 90, γ = 120 Za (molecules per ASU) 2 2 Matthews coefficient (Å3 Da−1) 2.31 2.53 Solvent content (%) 46.7 51.4 Data-processing statistics Temperature of measurement (K) 100 100 Wavelength (Å) 0.98 1.00 Resolutionb (Å) 45.18–2.40 (2.46–2.40) 61.41–2.80 (2.97–2.80) Total reflections 104,269 (7012) 76,368 (11,417) Unique reflections 11,381 (807) 7966 (1231) Completeness (%) 99.7 (97.9) 99.7 (99.3) c Rmerge (%) 7.3 (88.1) 11.1 (88.9) d CC1/2 (%) 100.0 (79.5) 99.9 (76.9) 〈I/σ(I)〉 24.24 (2.31) 14.21 (2.16) Refinement statistics Resolution range (Å) 43.98–2.40 52.43–2.80 e Rwork/Rfree (%) 20.7/24.9 24.6/26.9 No. of non-H protein atoms 1848 1802 No. of water molecules 31 – 2− No. of ligand molecules 2 SO4 1 ATP, 1 AMP Root-mean-square deviations Bond lengths (Å) 0.002 0.003 Bond angles (°) 0.443 0.675 Average B factor (Å2) All protein atoms 59.06 87.18 Waters 52.87 – Ligands 77.46 100.23 Ramachandran plotf (%) Most favored 97.48 95.71 Additionally allowed 2.52 3.86 Outliers 0.00 0.43 PDB accession number 5g5r 5g5x

a There is one molecule nanobody and one molecule of AfCBSD in the asymmetric unit. b Highest resolution shell in parentheses. c ¼ ∑ ∑ j ð Þ−h ð Þij=∑ ∑ ð Þ fl 〈 〉 Rmerge hkl j Ii hkl I hkl hkl iIi hkl ,where Ii(hkl)istheith measurement of the intensity of the unique re ection (hkl) and I(hkl) is the mean over all symmetry- related measurements. d CC(1/2) [78]. e Random 5% of working set of reflections [46]. f MolProbity [47]. nucleotide (termed apo), and the nucleotide-bound form (with bound nanobodies raised against the full-length AfS2P protein were employed ATP and AMP) of the same nanobody complex (termed nucleotide- for co-crystallization with the AfCBSD. We were successful with Nb330 bound). The structures were refined at diffraction limits of 2.4 Å and that exhibits the immunoglobulin-fold with the prototypical 9 β- 2.8 Å, respectively, with satisfactory statistics (Table 1). strands A–B–C–C′–C″–D–E–F–G [50] forming a four-stranded and a five stranded antiparallel β-sheet (plus one very short additional β- 3.1.1. Overall structure of AfCBSD in complex with Nb330 strand Η). Furthermore, it displays one disulfide-bond between the con- Despite all efforts, all crystals of the AfCBSD alone, with and without served cysteines (Cys22, Cys95) located on strands Β and Φ,stacked ATP, diffracted X-rays only to a resolution of 10 Å and worse. Thus, against a conserved tryptophan (Trp36); and three hypervariable

Fig. 1. Overall structure and dimer of the AfCBSD-Nb330 complex. (A) Cartoon representation; AfCBSD is colored from N- to C-terminus (blue to red). AfCBSD consists of two domains CBS1 and CBS2. The secondary structure elements are depicted. The CDRs (orange) and the His-tag of Nb330 (red) are highlighted. Sulfate ions and the disulfide bridge are shown as sticks. (B) Two views of the AfCBSD-Nb330 dimer that is formed in a head-to-tail manner, in which CBS1 interacts with CBS2* and CBS2 with CBS1*. Residues contributing to the dimer interface are located on helices α7, α8andα11 of both monomers. 1864 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 regions (or complexity determining regions; CDR) located in the loops Alignment of Nb330 with 25 nanobodies with known 3D structure B–Χ,C′–C″ and F–G, respectively (Figs. 1A and S1). It additionally ex- found in the Protein Data Bank (PDB, www.rcsb.org) and with the hibits one short 310-helix between strands E and F, and an additional highest sequence similarity to Nb330 revealed that the CDR3 of Nb330 β-strand H at the very C-terminus. is one of the shortest ones with only 9 residues (Fig. S1). The nanobody The CBS domain pair or Bateman domain of AfS2P (here named binds in a parallel manner to AfCBS (Fig. S2), considering its longest axis AfCBSD for simplification) consists of two consecutive pyramid- and the pseudo 2-fold axis of the AfCBSD. Binding of the epitope is made shaped CBS domains with the typical β1-α1-β2-β3-α2 topology [24] possible by a bending of CDR3 by 90° in respect to the nanobody's lon- (Fig. 1A) and comprises AfS2P residues 236–362. In domain 1 (CBS1), gest axis (also called the stretched twisted conformation of CDR3) helix 2 is very short and exhibits a 310-conformation (η8), followed by (Fig. 2). About 40% of the top 25 similar nanobodies show a comparable the linker that contains a second 310-helix (η9), that is sometimes mode of binding to their epitope (parallel) whereas another 40% bind found in CBS domain pairs and often called the α0′ helix. In domain 2 perpendicular to their target protein, in which the CDR3-loop is in an (CBS2), both helices (α10 and α11) are α-helices. Electron density elongated (or protruding) conformation (often stabilized by the exten- was missing for the last 6 amino acids (R357-S362) following helix sion of the flanking β-strands). For the remaining 20% of nanobodies, α11 as well as the C-terminal StrepII-tag probably due to high flexibility only unligated structures are available, so it cannot be exactly assessed, of this part of the structure. Together, both CBS domains form a which conformation they would adopt when bound to their epitope. butterfly-shaped overall structure with pseudo-C2 symmetry with the 2-fold axis running parallel to the central β-sheets. 3.1.3. Mediation of crystal contacts by Nb330 Difficulties in crystallizing CBS domain containing proteins, indepen- dent of the presence or absence of their respective ligands, have been 3.1.2. The epitope of Nb330 and mode of binding on AfCBSD reported [23]. As mentioned above, the crystals of the AfCBSD alone Nb330 binds to an epitope located mainly on CBS1 (Figs. 2 and S2), diffracted X-rays weakly also in our study, which points to their poor which is composed of the N-terminal linker preceding helix α7, strand inner order and/or high flexibility. In crystallography, nanobodies are β5, helix α7 and the loop Lβ6β7, and by one residue of strand β10 on used to stabilize certain conformations and to promote crystal contacts CBS2. The interface is characterized by several hydrogen bonds, one [51–53]. Crystal contacts within the AfCBSD-Nb330 crystals are mostly salt bridge, and diverse van-der-Waals interactions and covers a surface mediated by the nanobody (Fig. S3). The AfCBSD interacts mainly with area of 720.4 Å2. From the three CDRs, CDR2 and CDR3 contribute most itself along the c-axis (like a cord, Fig. S3A–B), whereas the nanobody to the antigen-binding surface (Figs. 2 and S2). Asn30′ on CDR1 (the molecules mediate the cross connections along the a- and b-axes apostrophe indicates residues of the Nb) interacts with Lys240 (α7). (Fig. S3A–C), through which the third dimension of the crystal is On CDR2, Thr52′ and Ser56′ interact with Glu253 (α7), and Asn58′ generated. with Gly271 (Lβ6β7) as well as on CDR3 Ala101′ and Thr102′ with Thr244, and Ser105′ with Glu241. Three additional interactions are 3.1.4. AfCBSD forms a dimer in the crystal lattice formed by some so-called framework region residues (FR) of Nb330, In both structures, the AfCBSD forms a dimer (Fig. 1B) in the crystal namely the salt bridge between Glu46′ and Arg331 (β10), as well as lattice in accordance with its migration behavior on size-exclusion chro- the interaction of Leu47′ with Glu272 and Asp61′ with Arg273 (both matography during purification as described below (Fig. 4). CBS domain on Lβ6β7). pairs display three different kinds of dimer configurations (baptized CBS

Fig. 2. The epitope of Nb330 on AfCBSD. The three CDRs of Nb330 (CDR1 in orange, CDR2 in yellow, CDR3 in green) and some residues of the framework region 2 and 3 (FR2–3) bind to an epitope on AfCBSD, mainly composed of residues of CBS1. Those are located on the loop preceding β5, β5, Lβ5α7andLβ6β7. On CBS2 Arg331 (Lβ9β10) is also involved. Coloring scheme of AfCBSD as in Fig. 1. M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1865 modules in [54]): parallel (head-to-head), anti-parallel (head-to-tail), in CBSX2 from A. thaliana, PDB ID: 4GQV [56]), especially in their apo and V-shaped CBS modules [23]. The dimer of AfCBSD displays the states. The configuration of the CBS module observed in the AfCBSD re- rare anti-parallel configuration in which CBS1 interacts with the CBS2* ported here is unique among anti-parallel CBS modules with an α = domain of the other pair and, analogously, CBS2 with CBS1* (Fig. 1B). 105° between the two CBS domain pairs (or 75° deviation from 180°; The dimer interface comprises an area of 996 Å2 per monomer as deter- Fig. S4) including a slight tilt (which results from the shift of the mole- mined by the PISA server [55] and involves mainly van-der-Waals inter- cule by 9 Å along the rotation axis, determined using Chimera [57]). In actions between helices α7/α8andα11* and the symmetry-equivalent this way, AfCBSD is able to accommodate a large nucleotide like ATP in between α7*/α8* and α11, respectively. As an example, van-der-Waals its S1 ligand-binding site (binding site nomenclature as in [58]) that interactions are formed between Ile258, Leu280 and Ile347* (and the would otherwise clash with the other ATP molecule in the neighboring corresponding residues on the other protomer). Only two hydrogen- protomer if AfCBSD would adopt a more flat configuration. bonds are formed, namely between Lys262(O) and Arg343*(NH1) and, analogously, Lys262*(O) and Arg343(NH1) (Fig. 3A). 3.1.5. Nucleotide binding of AfCBSD in the crystal Parallel CBS modules form flat discs (180° angle “α” between the In the nucleotide-bound form, AfCBSD binds one ATP molecule in the two pseudo-2-fold axes), whereas anti-parallel modules often show S1 site and one AMP molecule in the S2 site (Fig. 3B). Both molecules some degree of distortion of the α-angle to either side (e.g. α =125° originate from the buffers used during purification that were

2− Fig. 3. Nucleotide binding of AfCBSD. (A) Nucleotide binding site of one protomer in the apo form (wheat) is occupied by one sulfate ion (SO4 ) originating from the crystallization solution. Arg343* from the second protomer of the dimer (slate blue) participates in the binding of the sulfate ion and binds directly to Lys262 (O). (B) In the nucleotide-bound form ATP binds in the S1 site and AMP in the S2 site of AfCBSD. Arg343* stabilizes the α-phosphate of AMP. (C) Detailed view of nucleotide binding of AfCBSD using LigPlot+ [77].Inbothnucleotidestheadenine rings are stabilized by interactions to the backbone of AfCBSD (Val243, Gly265, Val301, Arg323), the hydroxyl groups of the pentoses are bound to the side chains of Asp282/Ser297 and Asp340, respectively. The phosphates of ATP are coordinated by the side chains of Thr279, His281 and Arg338 as well as by the backbone of Gly322. 1866 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 supplemented with 1 mM ATP from microbial source, which can con- 3.1.6. Nucleotide binding of AfCBSD in solution tain AMP as contaminant. Alternatively, the AMP could result from Purification of AfCBSD-Strep constructs in the presence of spontaneous ATP or ADP hydrolysis. The electron density map is in full magnesium-free ATP (addition of 1 mM ATP after elution from affinity agreement with the interpretation of ATP in site S1 and AMP in site chromatography) triggers a change in the distribution of their oligomer- S2. Placing, for example, ADP in both sites resulted in strong features ic states observed in SEC (Fig. 4). For AfCBSD-Strep supplemented with in the difference density map. Both nucleotides are bound in hydropho- ATP (Fig. 4A, solid line) the elution peak is sharpened in shape as no bic pockets (Fig. 3C). fronting (tetrameric form, peak at 77.8 ml) is observed in contrast to The adenine rings are stabilized by interactions to the backbone of AfCBSD-Strep without ATP (Fig. 4A, dashed line). Still a distribution of AfCBSD. For the AMP-bound S2 site, a typical “reverse” adenine- dimer and monomer is observed with ATP. The resolution of the main binding motif [59] is found with Val243 interacting simultaneously via elution peak is improved in such as the monomeric form (peak at its backbone amino hydrogen atom and carbonyl oxygen atom with 87.8 ml) becomes more visible compared to the very broad peak of the N1 nitrogen atom and the NH2 group of AMP, respectively. Addition- AfCBSD-Strep. This indicates that without bound nucleotide the ally, the carbonyl oxygen atoms of Gly265 and Val243 accept a hydro- monomer-dimer-distribution is very dynamic and by addition of ATP gen bond from the NH2 group. His263 sits at the back and makes the equilibrium is slightly shifted to the AfCBSD dimer. To ensure com- stacking interactions with the adenine ring of AMP. Additionally, parability, the SEC experiments were performed using the same param- Ile335 makes C-H··π-interactions with the latter via its δ-andγ- eters (FPLC, column, injection volume, buffer, flow rate and methyl-groups, coming from the other side of the adenine ring. temperature). The adenine ring of ATP bound to site S1 is coordinated via C-H··π- interactions by Ile277 and Ile321. The ATP-binding site shows an alter- 3.1.7. The C-terminal helix α11 plays a major role in AfCBSD dimerization ation from the “reverse” adenine-binding motif. The NH2 group is From the crystal structure we concluded that the C-terminal helix bound by the carbonyl oxygen of Arg323 and the N1 nitrogen hydrogen might exert an influence on the oligomeric state of our AfCBSD con- bonds with the backbone amide hydrogen atom of Val301. In other struct. We analyzed the crystal structure of AfCBSD comprising amino nucleotide-bound CBS domains, like CBSX2 ([56], PDB ID:4GQY) and acids M1-M354 (G355-S362 not resolved in the crystal structure) MJ1225 ([60], PDB ID:3LFZ), a similar interaction pattern is found. In with the PISA server [55] and compared the dimer interface with a trun- CBSX2 Arg204 and Leu182 correspond to Arg323 and Val301 in AfCBSD, cated model comprising amino acids M1-E349 (corresponding to the respectively. There, the distance of the NH2 group of the bound AMP to construct AfCBSDΔC13). The C-terminal helix α11 seems to strongly Leu182(O) is 3.0 Å and to Arg204(O) is 3.3 Å. In MJ1225 Gly117 and contribute to the dimer interface as by removing a large part of it Ile95 are the corresponding residues. In both structures (MJ1225/3LFZ (A350-M354), the solvation free energy gain upon interface formation and CBSX2/4GQY), no second nucleotide is bound to the S2 site. Binding decreases from −19.8 kcal/mol (AfCBSD) to −9.7 kcal/mol of AMP to site S2 of AfCBSD could lead to the small distortion of the (AfCBSDΔC13) and the interface area decreases from 996 Å2 to 687 Å2. binding site towards Arg323, especially as Arg323 is involved in the In order to test this experimentally, we examined a thrombin- binding of the α-phosphoryl group of the neighboring AMP via the cleavable, N-terminally tagged construct with the 13 C-terminal guanidinium group of its side chain. This would explain the shorter dis- amino acids missing, yielding after thrombin digest construct tance of 2.3 Å of the Arg323 carbonyl oxygen to ATP(NH2)inAfCBSD. AfCBSDΔC13. ATP-saturated AfCBSDΔC13 is exclusively a monomer At both sites, the hydroxyl groups of the ribose moieties are hydro- (Fig. 4A, dotted line), which is in agreement with the PISA analysis gen bonded to the side chains of Asp282/Ser297 (ATP) and Asp340 and the inferred contribution of the C-terminus. (AMP), respectively. Arg343* stabilizes the α-phosphoryl group of The C-terminal helix α11 harbors Arg343, a residue that directly AMP, and thus bridges the second protomer of the dimer to the nucleo- bridges both monomers by reaching over the nucleotide-binding site tide binding site. The α-phosphoryl group of ATP is hydrogen bonded to and interacting with the carbonyl oxygen of Lys262 (Fig. 3A). In the the side chains of Thr279 and His281. The backbone of nucleotide-bound structure Arg343 bridges both monomers by addi-

Gly322(N) stabilizes the β-phosphoryl group and Arg338(NH2)theγ- tional binding to the α-phosphate of the neighboring AMP (Fig. 3B, C). phosphoryl group of ATP. All the above-mentioned residues are part of We have prepared mutants replacing Arg343 by alanine and glutamic the linker regions preceding strands β5andβ8, the AMP-binding acid and have investigated their oligomeric state (Fig. 4B). While the ex- motif G-h-h′-T/S-x-x′-D/N and the adenosine-binding motif h-y-y′-h′- change to alanine has no effect on dimerization (Fig. 4B, dotted line), ex- P(“h” is a hydrophobic and “x” or “y” are any residue) (definitions as change to glutamic acid yields a monomer independent of nucleotide in [23],Fig.S7). binding (Fig. 4B, solid and dashed lines). The change from a positive to CBS domains regulate enzyme function by binding to nucleotides a negative charge at this position seems to have a repulsive effect that like ATP, AMP, SAM and their derivatives [61]. For example, the isolated leads to the dissociation of the dimer. This may be mediated by a local CBS domain of Bacillus subtilis site-2 protease SpoIVFB binds specifically electrostatic repulsion within the nucleotide-binding site (Fig. 3)orby to ATP, and Zn and ATP stimulate the proteolytic cleavage of the sub- the dislocation of the whole helix α11. The SEC experiments of the strate Pro-σK S20G by liposome-reconstituted full-length SpoIVFB R343 variants (Fig. 4B) were performed on a different column than for [21]. While conformational changes upon binding of the ligand have the above-mentioned constructs (Fig. 4A). Thus the elution volumes been described for many of CBSDs [23], there are little structural differ- for the peaks are shifted by 2–3ml. ences observable in our crystal structures of the nucleotide-bound and unbound forms of AfCBSD. The overall RMS deviation of 117 superposed 3.1.8. Effect of magnesium ions on the oligomeric state and ATP-affinity of Cα-atoms is 0.7 Å. Thus, any regulation governed by ligand binding of AfCBSD the AfCBSD module must be mediated in another fashion. AfCBSD supplemented with 1 mM ATP shows a distribution of di- A concern during structure determination was, that nanobody- meric and monomeric state, similar to the nucleotide-free form binding could impose an artificial conformation onto the nucleotide- (Fig. 4A). However, in cells many proteins bind ATP in its magnesium- bound AfCBSD, thus preventing conformational changes natively bound form (MgATP). In the bacterial cytoplasm free Mg2+ is estimated occurring. As mixing of AfCBSD with nucleotides and purification of to be 0.5–2mM[62,63] and the ATP-concentration 1–10 mM [64,65]. the dimeric species is performed before complex formation with Therefore the unbound form of nucleotides (NTP4−) will not be satu- Nb330, this seems rather unlikely. In the case that major conforma- rated with Mg2+ under physiological conditions, also considering the tional changes of AfCBSD upon nucleotide binding would occur in affinity of ATP for Mg2+ being 103 μM [66]. AfCBSD could therefore solution, Nb330 would probably not recognize its epitope and no bind free ATP under physiological conditions as found in the here de- complex would be formed. scribed crystal structure. There we found no counter-ion in the complex M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1867

Fig. 4. Oligomerization behavior and nucleotide binding of AfCBSD constructs. (A) Size exclusion chromatograms of nucleotide-free (dashed line) and nucleotide-bound AfCBSD-Strep (solid line). The nucleotide-free sample displays a broad distribution of mainly monomer and dimer, hinting at a dynamic equilibrium between the two. Only a slight sharpening of curve shape towards the dimeric species is observed upon binding of 1 mM ATP. Removal of the C-terminal 13 residues in AfCBSDΔC13 (dotted line) disrupts dimer formation. (B) Exchange of R343 (located on α11) to alanine (dotted line) does not disrupt the dimer. In contrast, an exchange to glutamic acid yields monomeric (solid line) protein independent of nucleotide binding (dashed line). (C) MgATP drastically shifts the equilibrium to the monomeric AfCBSD species (dashed and black solid lines) when compared to nucleotide-free AfCBSD (dotted line). In contrast, magnesium ions alone (grey solid line) do not have this effect. The theoretical molecular weight of the AfCBSD-Strep monomer is

15.5 kDa. (D) Fluorescence polarization assay of AfCBSD wildtype and R343E variant in absence and presence of 2 mM MgCl2.Affinities of AfCBSD-Strep to ATP and MgATP are virtually the same and monomerization does not prevent nucleotide-binding. Fluorescent nucleotide γ-6AH-ATP-ATTO-495 was used at 400 nM concentration. Relative polarization in mPOL is plotted against protein concentration in μM. (Inset) Curve fitting and calculation of equilibrium dissociation constants with Origin. structure, in similarity with other CBS domains, e.g. those of yeast AMPK R343E variant to a fluorescent ATP derivative (γ-6AH-ATP-ATTO-495) [67] and human Clc-5 [68]. using fluorescence polarization (Fig. 4D; Table S4). We used the mono- To investigate a nonetheless possible influence of Mg2+ onto ATP- meric R343E variant to study whether monomerization has an effect on binding of AfCBSD, we have determined the oligomeric state of AfCBSD ATP-binding. We found only a slightly higher affinity to ATP compared and its affinity to ATP in the presence of Mg2+ ions in solution. to MgATP (340 ± 10 μM vs. 374 ± 12 μM, Table S4) for the wild-type The effect of MgATP on AfCBSD in the SEC (Fig. 4C, dashed line) is in- protein. Interestingly, the monomeric R343E variant shows only a triguing. MgATP-binding seems to drive monomerization, similar to the minor reduction in affinity compared to the wild-type protein (447 ± effect of the R343E mutation (Fig. 4B). In contrast, free Mg2+ alone does 20 μMforATPand419±53μM for MgATP, respectively; Table S4). To- not have this effect (Fig. 4C, grey solid line). 20 mM Mg2+ in the buffer is gether, these data clearly show that monomerization does not prevent a rather non-physiological condition. The overall Mg2+-concentration and that Mg2+ is not required for efficient ATP-binding by AfCBSD. in the cytoplasm of different mammalian and bacterial cell types was The electrostatic surface around the nucleotide-binding site could determined to be 20–30 mM [69], but this is the whole cellular Mg2+. also play an important role in the requirement of Mg2+ for ATP- Most of the Mg2+ is bound to phospholipids, protein, DNA and nucleo- binding. CNNM2 has an electronegative surface around the tides, among others. As mentioned above, the free magnesium is esti- nucleotide-binding site, whereas Clc-5 and IMPDH2 display an electro- mated to 0.5–2 mM. That is why we also tested a buffer containing positive surface and thus do not require the positive charge of Mg2+ for 2+ 1.5 mM MgCl2 and 1 mM ATP, leaving 1 mM Mg as counter-ion for efficient ATP-binding. MgATP binding was found also in CBS domains of ATP and 0.5 mM free Mg2+. There we observe the same effect, as with human AMPK [72],buttheiraffinity to MgATP did not differ from that to

20 mM MgCl2 +1mMATP(Fig. 4C, black solid line). ATP (Table S4), and the surface in the nucleotide-binding site was also Many proteins bind MgATP instead of free ATP. This is different for electropositive. AfCBSD also shows an electropositive surface in the many CBS domains. There, only few were found to bind MgATP, for in- nucleotide-binding site, as described below. stance the CBS domains of the magnesium transporters CNNM [70,71], where Mg2+ seems to be required for efficient ATP-binding, which is 3.2. Probing the quaternary structure of AfS2P probably linked to their function as magnesium transporters. On the other hand, the CBS domains of Clc-5 and IMPDH2 show a higher affin- 3.2.1. Oligomerization within the membrane ity to ATP than to MgATP (KD of 37.5 μM and 56.4 μM for ATP and 177 To investigate the oligomeric state of AfS2P in its native environ- μM and 242 μM for MgATP, respectively (Table S4)). We have therefore ment, we cross-linked full-length AfS2P-Strep directly in E. coli mem- determined the affinity of AfCBSD wild-type protein and the monomeric branes using the membrane permeable amine-reactive cross-linker 1868 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871

DMS. SDS-PAGE and immunoblot analysis of cross-linked samples M. jannaschii (MjS2P-TM) [8]. The loop Lα1α2inAfS2P (UniProt: showed equally populated species at approx. 35 kDa and 70 kDa, re- O29915) is considerably longer with 14 residues than in MjS2P spectively - corresponding to a monomer and a dimer of AfS2P (UniProt: Q57837), with only one residue (Fig. S7). Thus an exact (Fig. S5C). It is important to recognize that AfS2P shows a typical “gel modeling of this loop is not easily possible. shifting” behavior for membrane proteins on SDS-gels [73] as its nomi- nal molecular weight is 41 kDa per monomer, i.e. it migrates slightly 3.2.4. Characterization of the transmembrane dimer interface faster than expected. Additionally, SEC-MALS experiments of DDM- To characterize the possible dimer interface within the transmem- solubilized AfS2P-His resulted in a size of approximately 82 kDa for brane domain (TMD) we isolated the DMS-cross-linked dimeric fraction the protein fraction of the detergent-protein micelle, which would cor- and subjected its proteolytic fragments to mass spectrometric analysis. respond to a dimer of AfS2P (Fig. S5D–F, Table S3). It is important to note We were able to find one crosslink within the membrane, namely be- that the cross-linking and SEC-MALS experiments were performed at tween Lys6 of one protomer and Lys181* of the other protomer (resi- temperatures ranging from 4 °C to 37 °C. Due to technical reasons, dues marked by an asterisk) (Figs. 5C and S5A–B). Together with in work at 83 °C, the optimal growth temperature of Archaeoglobus silico protein-protein docking using ClusPro [33,34,36] we could pre- fulgidus, is not possible. pare a rough model of the dimer within the membrane (Fig. 5B, TMD). Dimerization is not unusual for intramembrane proteases. Dimers We additionally took into consideration the dimeric structure of the cy- and even tetramers were found for members of the rhomboid family toplasmic AfCBSD described in this work. of intramembrane proteases [22]. Recent studies show that From the various dimers calculated by ClusPro, we discarded those cooperativity/homotropic allostery plays a role in the regulation of in which the two protomers did not show the same relative orientation rhomboid specificity [74], although the protease is also active against within a possible lipid bilayer (pseudo or inverted dimers). This resulted certain substrates in its monomeric form. As the natural substrate of in dimers showing a back-to-back orientation (interaction between AfS2P is unknown to date, suitable sensors will be designed to be able TMHs α2andα4 of both monomers) or a face-to-face orientation (in- to study AfS2P's proteolytic activity in the future. teraction of α1-β1-β2 with α5*-α6*). In the crystal structure of the AfCBSD the N-termini of each protomer that directly connect to the C- 3.2.2. Neither ATP nor MgATP influence the oligomeric state of full-length termini of TMHs α6 in a full-length protein are 36 Å away from each AfS2P other. In back-to-back dimers calculated by ClusPro the distance of the Due to the dynamic monomer-dimer equilibrium of the AfCBSD C-termini of TMHs α6 ranges from 40 to 70 Å, while in face-to-face di- (Fig. 4A) and the influence of MgATP on the AfCBSD (Fig. 4C), we inves- mers this distance is between 25 and 40 Å. Additionally, Lys6 and tigated if MgATP had also effects on the oligomeric state of the full- Lys181* have to come into close contact, as the length of the spacer length protein and if this would shed light on the regulatory mechanism arm in the cross-linker DMS is 11 Å. Both conditions are satisfied only of proteolysis by AfS2P. Therefore, we performed cross-linking and SEC- in a TMD dimer with face-to-face orientation (Fig. 5B, TMD). As the MALS experiments (Fig. S6) with full-length AfS2P-His protein in the two-fold symmetry of the AfCBSD module determined in this work presence of ATP and MgATP, respectively. could impose a two-fold symmetry on the TMD of AfS2P, two-fold sym- Cross-linking and SEC-MALS experiments both show no differences metry was another criterion for the selection of the TMD dimer model. in the oligomeric state between the nucleotide-free, ATP- or MgATP- bound forms. SEC-MALS of detergent-solubilized AfS2P-His in the pres- 3.2.5. Solvent exposition of solubilized AfS2P ence of ATP and MgATP yields the same molecular weight, peak distri- Limited proteolysis can be used to probe local structure and to locate bution and peak shape (Fig. S6A–C; Table S3). Cross-linked exposed and flexible regions in proteins. To investigate the domain membrane-bound AfS2P-His exposed to MgATP exhibits the same dis- structure (architecture) of AfS2P and to map potential loops extending tribution of oligomeric states as ATP-exposed and nucleotide-free out of the detergent-protein micelle, limited proteolysis was performed AfS2P at temperatures from 20 to 37 °C (Fig. S6D–F). using five different proteases. Proteolysis sensitive sites determined by EDMAN sequencing of proteolytic fragments (Table S5) were mapped 3.2.3. Structural model of AfS2P-TM onto the proposed AfS2P model (Fig. 6). Seven protease sensitive sites A homology model of the transmembrane domain of AfS2P was pre- were identified in AfS2P - four at the cytoplasmic face of the transmem- pared (Fig. 5A, TMD) using the structure of the homologous S2P from brane domain (TMD) and three within the AfCBSD (Fig. 6A). Exposed

Fig. 5. Structural model of AfS2P. (A) Homology model of AfS2P TMD monomer combined with the apo structure of AfCBSD. The dashed line represents residues connecting the TMD and AfCBSD. Secondary structure elements are highlighted. (B) Model of the AfS2P dimer calculated by ClusPro. (C) Close-up view of Lys6 and Lys181 that were found to be in close proximity by cross-linking analysis. (D) Close-up view of the possible interaction site between the TMD and the ATP bound by AfCBSD comprising Arg76/77 and Arg81/83 of the TMD. M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1869 sites were located in the loop Lβ4α3(MM|D97EL, the sequenced N- transmembrane dimer. This in turn would allow accommodating the terminal amino acid is numbered) connecting the edge stand β4 and globular cytoplasmic domain of a substrate and releasing it into the cy- helix α3, and within the first amino acids of helix α3(EE|L107VV). An- toplasm. But how then would the transmembrane helix of such a sub- other two protease accessible sites could be determined in Lα4-Cα5 strate gain access to the active site? (EK|R174 |S175YA)aswellasinthefirst amino acids of helix α5 The idea for our current model came to us when we examined the (AE|A179T|K181IA). The CBS1 domain showed one cleavage site within electrostatic surface of the AfCBSD dimer that is in contact with the cy- the loop Lβ6β7(VE|G271ER). The same was found for the CBS2 domain toplasmic face of the TMD. The surface of the latter is positively charged (VV|E328 |H329GR|V332VG; Lβ9β10), which additionally exhibits a with exposed lysine and arginine residues, namely Arg76, 77, 81, 83, cleavage site 17 amino acids from the C-terminus (IK|E346IL) on helix 104, 174 and Lys101, 102, 173, which is in accordance with the α11. Interestingly, no cleavage could be observed for the loops Lα1α2 positive-inside rule [76]. The counterpart surface of the AfCBSD dimer as well as Lα5α6 at the periplasmic face of the TMD. This fact indicates displays a deep, mainly positively charged groove traversing the that in DDM-solubilized AfS2P the micelle covers these loops to an ex- whole surface (Fig. S9). Binding of the two ATP molecules within this tent that no proteolysis is possible. Especially for Lα1α2, which is 14 groove masks the positive charge and could mediate direct interactions residues long, it could mean that it inserts into the membrane plane to the arginine residues within the cytoplasmic face of the TMD and interacts with residues from the inner of the TMD. As we could ob- (Fig. 5D). Thus we oriented the AfCBSD dimer to the putative TMD serve a cleavage of the AfCBSD at only one face of the domain (Fig. 6A), dimer in such a way that the γ-phosphate of ATP could interact with tight dimerization and association with the cytoplasmic face of the TMD Arg76/77 and/or Arg81/83 and helices α6 would link to the very N- seem to render its α-helical side protease-inaccessible. The found limit- termini of the AfCBSD modules by threading through between helices ed proteolysis cleavage sites additionally support our proposed dimer α10 and α11 (Fig. 5B, dashed lines). In this orientation binding of ATP model (Fig. 6B). would stabilize the TMD conformation, which we believe would be the open, proteolytically active conformation. This would go in line 3.2.6. First insights into a possible nucleotide-driven regulation with findings for B. subtilis SpoIVFB for which ATP-binding was shown The relative orientation of AfS2P-TM and AfCBSD to each other has to activate the protease when reconstituted in proteoliposomes [21]. been unclear as the crystal structure of the full-length protein has not In our model, ATP binding to the AfCBSD would bring the above- been determined yet. Nevertheless, based on our findings we feel confi- mentioned TMD arginines, located at the end of helix 2 and in the dent to propose a possible relative orientation of the two domains to succeeding loop Lα2β3(Fig. 5D), in larger distance to the “immobile” each other. We also discuss a possible regulation of AfS2P-activity by helices α6 and would hence open up the active site grooves for sub- its CBS module upon nucleotide binding. strate binding. Thus, helices α6 would act as hinges around which the The CBS modules of membrane proteins like MgtE [75],CNNM2[70] TMD dimer would open and close (Fig. 5A); and in analogy the TMD ar- or CLC-5 [68] undergo large conformational changes upon ligand bind- ginines would act as the “handles”. The lack of the stabilizing interac- ing (Mg2+ in MgtE/CNNM2, ATP in CLC-5) which trigger other confor- tions of ATP in the apo form would lead to a sampling of TMD mational changes in the corresponding transmembrane domains. conformations between open and closed, but none of them would be Structural alignment of both AfCBSD forms (apo and nucleotide- discriminated. In this way substrate binding would be impeded. bound) shows that their conformation is virtually the same (Fig. S8), It is not entirely clear how exactly MgATP would accomplish as the RMSD of the dimers is about 0.7 Å calculated for 117 aligned Cα monomerization and thus opening of the CBS module from a structural atoms with a maximal local deviation of 1.8 Å (Cα of Met319). A drastic point of view. One possibility is that it is mediated through electrostatic change in oligomeric state of AfCBSD was observed upon MgATP- repulsion on Arg343 (Fig. 3B). This could provoke the abolishment/ binding (Fig. 4C), and not upon ATP-binding. If AfS2P binds MgATP in clearing of the steric hindrance by the CBSD in the cytoplasmic site of the cell, which is fairly possible, this would then induce the opening of AfS2P, allowing to accommodate a globular domain (e.g. transcription the CBS module without changing the oligomeric state of the factor) on the cytoplasmic side. Additionally, it could also lead to the

Fig. 6. Mapping of protease-accessible sites on the AfS2P model. (A) Monomeric AfS2P model showing the cleavage sites from limited proteolysis experiments and EDMAN sequencing (red). Cleavage occurs between the two residues of each cleavage site shown as sticks and is marked by a vertical line in the respective protein sequence. (B) Mapped cleavage sites (red) on the AfS2P dimer model. The dashed lines represent residues connecting the TMD and AfCBSD. 1870 M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 exposition of a new binding surface recognizing the substrate and oth- [4] T. Tomoyasu, et al., Topology and subcellular localization of FtsH protein in Escherichia coli, J. Bacteriol. 175 (1993) 1352–1357. erwise buried within the dimer interface, adding specificity and affinity [5] C. Bieniossek, et al., The molecular architecture of the metalloprotease FtsH, Proc. to substrate binding by AfS2P besides the sequence specificity of the ac- Natl. Acad. Sci. U. S. A. 103 (2006) 3066–3071. tive site. [6] I. Botos, et al., The catalytic domain of Escherichia coli Lon protease has a unique fold and a Ser-Lys dyad in the active site, J. Biol. Chem. 279 (2004) 8140–8148. [7] R.P. Baker, et al., Enzymatic analysis of a rhomboid intramembrane protease impli- 4. Conclusions cates transmembrane helix 5 as the lateral substrate gate, Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 8257–8262. [8] L. Feng, et al., Structure of a site-2 protease family intramembrane metalloprotease, Site-2 proteases play important roles in many biological processes Science 318 (2007) 1608–1612. [9] S. Cutting, et al., Sporulation operon spoIVF and the characterization of mutations by cleaving membrane-bound transcription factors. An in-depth under- that uncouple mother-cell from forespore gene expression in Bacillus subtilis,J. standing of the mechanisms of substrate recognition, proteolytic cleav- Mol. Biol. 221 (1991) 1237–1256. age and regulation is of great importance. This paper reports the first [10] D.H. Green, S.M. Cutting, Membrane topology of the Bacillus subtilis pro- – fi sigma(K) processing complex, J. Bacteriol. 182 (2000) 278 285. crystal structure of the regulatory CBS domain tandem and the rst [11] Y. Akiyama, et al., RseP (YaeL), an Escherichia coli RIP protease, cleaves transmem- characterization of the oligomeric state and dimer interface of a site-2 brane sequences, EMBO J. 23 (2004) 4434–4442. fi protease family member. Further study is needed to identify the sub- [12] F.Y. An, et al., Identi cation and characterization of a determinant (eep) on the En- terococcus faecalis chromosome that is involved in production of the peptide sex strates and study the exact mechanisms of substrate recognition and pheromone cAD1, J. Bacteriol. 181 (1999) 5915–5921. regulation of archaeal CBS-domain containing members of this interest- [13] R.B. Rawson, et al., Complementation cloning of S2P, a gene encoding a putative metalloprotease required for intramembrane cleavage of SREBPs, Mol. Cell 1 ing class of intramembrane cleaving proteases. (1997) 47–57. [14] J. Ye, et al., ER stress induces cleavage of membrane-bound ATF6 by the same prote- ases that process SREBPs, Mol. Cell 6 (2000) 1355–1364. Ethics statement [15] N.G. Zelenski, et al., Membrane topology of S2P, a protein required for intramembranous cleavage of sterol regulatory element-binding proteins, J. Biol. All llama vaccination experiments were executed according to EU Chem. 274 (1999) 21973–21980. [16] J. Sakai, et al., Sterol-regulated release of SREBP-2 from cell membranes requires two animal welfare legislation (2010/63/EU) and after approval by the sequential cleavages, one within a transmembrane segment, Cell 85 (1996) local ethics committee. 1037–1046. [17] E.A. Duncan, et al., Cleavage site for sterol-regulated protease localized to a Leu-Ser bond in the lumenal loop of sterol regulatory element-binding protein-2, J. Biol. Transparency document Chem. 272 (1997) 12778–12785. [18] D.A. Dougan, et al., ClpS, a substrate modulator of the ClpAP machine, Mol. Cell 9 (2002) 673–683. The Transparency document associated with this article can be [19] J.S. Schneider, et al., Site-2 protease substrate specificity and coupling in trans by a found, in online version. PDZ-substrate adapter protein, Proc. Natl. Acad. Sci. U. S. A. 110 (2013) 19543–19548. [20] A.A. Baykov, et al., The CBS domain: a protein module with an emerging prominent Acknowledgements role in regulation, ACS Chem. Biol. 6 (2011) 1156–1163. [21] R. Zhou, et al., Intramembrane proteolytic cleavage of a membrane-tethered tran- scription factor by a metalloprotease depends on ATP, Proc. Natl. Acad. Sci. U. S. A. We are grateful to Alex Cameron and Liz Carpenter (formerly) from 106 (2009) 16174–16179. the Membrane Protein Lab, Harwell Science and Innovation Campus, [22] P. Sampathkumar, et al., Oligomeric state study of prokaryotic rhomboid proteases, – Didcot, UK for fruitful discussions. We thank Dr. Petra Parizek, Depart- Biochim. Biophys. Acta 1818 (2012) 3090 3097. [23] J. Ereno-Orbea, et al., CBS domains: ligand binding sites and conformational variabil- ment of Biochemistry, University of Zürich, Switzerland, Dr. Yasar L. ity, Arch. Biochem. Biophys. 540 (2013) 70–81. Ahmed, BZH, University of Heidelberg, Germany and Dr. Felix Deluweit, [24] A. Bateman, The structure of a domain common to archaebacteria and the homocystinuria disease protein, Trends Biochem. Sci. 22 (1997) 12–13. Wyatt Technology Europe, Dernbach, Germany for their support during [25] M. Arbabi Ghahroudi, et al., Selection and identification of single domain antibody SEC-MALS measurements. We thank Elena Brunstein for excellent tech- fragments from camel heavy-chain antibodies, FEBS Lett. 414 (1997) 521–526. nical assistance, and Astrid Wilbrand-Hennes (CECAD proteomics facil- [26] E. Pardon, et al., A general protocol for the generation of Nanobodies for structural biology, Nat. Protoc. 9 (2014) 674–693. ity) for support in mass spectrometry. We thank Nele Buys for the [27] K.E. Conrath, et al., b-Lactamase inhibitors derived from single-domain antibody technical assistance during nanobody discovery. M. Schacherl thanks fragments elicited in the camelidae, Antimicrob. Agents Chemother. 45 (2001) – Dr. Elmar Behrmann, Research Center Caesar Bonn, Germany for sup- 2807 2812. [28] P. Chomczynski, N. Sacchi, Single-step method of RNA isolation by acid port in the last stages of the project. The help of the staff at the guanidinium thiocyanate-phenol-chloroform extraction, Anal. Biochem. 162 beamlines X06SA and X06DA of the Swiss Light Source, PSI, Villigen, (1987) 156–159. [29] R. Zell, H. Fritz, DNA mismatch-repair in Escherichia coli counteracting the hydrolytic Switzerland is highly appreciated. The project was supported by grant deamination of 5-methyl-cytosine residues, EMBO J. 6 (1987) 1809–1815. INST 216/682-1 FUGG from the German Research Foundation. The re- [30] U.K. Laemmli, Cleavage of structural proteins during the assembly of the head of search leading to these results has received funding from the bacteriophage T4, Nature 227 (1970) 680–685. [31] P. Edman, G. Begg, A protein sequenator, Eur. J. Biochem. 1 (1967) 80–91. European Community's Seventh Framework Program (FP7/2007- [32] N. Eswar, et al., Protein structure modeling with MODELLER, Methods Mol. Biol. 426 2013) under grant agreement no. 283570 (BioStruct-X). We thank (2008) 145–159. [33] S.R. Comeau, et al., ClusPro: a fully automated algorithm for protein-protein docking, INSTRUCT, part of the European Strategy Forum on Research Infrastruc- Nucleic Acids Res. 32 (2004) W96–W99. tures (ESFRI) and the Hercules Foundation Flanders for their support to [34] S.R. Comeau, et al., ClusPro: an automated docking and discrimination method for the nanobody discovery. the prediction of protein complexes, Bioinformatics 20 (2004) 45–50. [35] D. Kozakov, et al., PIPER: an FFT-based protein docking program with pairwise po- tentials, Proteins 65 (2006) 392–406. Appendix A. Supplementary data [36] D. Kozakov, et al., How good is automated protein docking? Proteins 81 (2013) 2159–2166. [37] J. Rappsilber, et al., Protocol for micro-purification, enrichment, pre-fractionation Supplementary data to this article can be found online at http://dx. and storage of peptides for proteomics using StageTips, Nat. Protoc. 2 (2007) doi.org/10.1016/j.bbamem.2017.05.006. 1896–1906. [38] M. Gotze, et al., StavroX–a software for analyzing crosslinked products in protein in- teraction studies, J. Am. Soc. Mass Spectrom. 23 (2012) 76–87. References [39] W. Kabsch, XDS, Acta Cryst D66 (2010) 125–132. [40] Airlie J. McCoy, et al., Phaser crystallographic software, J. Appl. Crystallogr. 40 (2007) [1] A. Ciechanover, A.L. Schwartz, The ubiquitin-mediated proteolytic pathway: mecha- 658–674. nisms of recognition of the proteolytic substrate and involvement in the degrada- [41] Paul D. Adams, et al., PHENIX: a comprehensive Python-based system for macromo- tion of native cellular proteins, FASEB J. 8 (1994) 182–191. lecular structure solution, Acta Cryst D66 (2010) 213–221. [2] N.D. Rawlings, et al., MEROPS: the database of proteolytic enzymes, their substrates [42] D. Oyen, et al., Mechanistic analysis of allosteric and non-allosteric effects arising and inhibitors, Nucleic Acids Res. 42 (2014) D503–D509. from nanobody binding to two epitopes of the dihydrofolate reductase of [3] M.S. Wolfe, et al., Are presenilins intramembrane-cleaving proteases? Implications Escherichia coli, Biochim. Biophys. Acta 1834 (2013) 2147–2157. for the molecular mechanism of Alzheimer's disease, Biochemistry 38 (1999) [43] G. Bunkoczi, R.J. Read, Improvement of molecular-replacement models with Sculp- 11223–11230. tor, Acta Crystallogr. D Biol. Crystallogr. 67 (2011) 303–312. M. Schacherl et al. / Biochimica et Biophysica Acta 1859 (2017) 1859–1871 1871

[44] M. Lucas, et al., Binding of S-methyl-5′-thioadenosine and S-adenosyl-L-methionine [62] T. Alatossava, et al., Manipulation of intracellular magnesium content in polymyxin to protein MJ0100 triggers an open-to-closed conformational change in its CBS B nonapeptide-sensitized Escherichia coli by ionophore A23187, J. Bacteriol. 162 motif pair, J. Mol. Biol. 396 (2010) 800–820. (1985) 413–419. [45] P. Emsley, et al., Features and development of Coot, Acta Crystallogr. D66 (2010) [63] A. Romani, A. Scarpa, Regulation of cell magnesium, Arch. Biochem. Biophys. 298 486–501. (1992) 1–12. [46] Free R value: cross-validation in crystallography, in: A.T. Brünger, W.C.J. Carter, et al., [64] I. Beis, E.A. Newsholme, The contents of adenine nucleotides, phosphagens and (Eds.), Methods in Enzymology, Macromolecular Crystallography Part B, Academic some glycolytic intermediates in resting muscles from vertebrates and inverte- Press 1997, pp. 366–396. brates, Biochem. J. 152 (1975) 23–32. [47] V.B. Chen, et al., MolProbity: all-atom structure validation for macromolecular crys- [65] H. Yaginuma, et al., Diversity in ATP concentrations in a single bacterial cell popula- tallography, Acta Crystallogr. D Biol. Crystallogr. 66 (2010) 12–21. tion revealed by quantitative single-cell imaging, Sci. Rep. 4 (2014) 6522. [48] G. Vriend, WHAT IF: a molecular modeling and drug design program, J. Mol. Graph. [66] J.E. Wilson, A. Chin, Chelation of divalent cations by ATP, studied by titration calo- 8 (1990) 52–56 (29). rimetry, Anal. Biochem. 193 (1991) 16–19. [49] H. Berman, et al., Announcing the worldwide Protein Data Bank, Nat. Struct. Biol. 10 [67] R. Townley, L. Shapiro, Crystal structures of the adenylate sensor from fission yeast (12) (2003) 980. AMP-activated protein kinase, Science 315 (2007) 1726–1729. [50] S. Muyldermans, Nanobodies: natural single-domain antibodies, Annu. Rev. [68] S. Meyer, et al., Nucleotide recognition by the cytoplasmic domain of the human Biochem. 82 (2013) 775–797. chloride transporter ClC-5, Nat. Struct. Mol. Biol. 14 (2007) 60–67. [51] K.V. Korotkov, et al., Crystal structure of the N-terminal domain of the secretin GspD [69] H. Ebel, T. Günther, Magnesium metabolism: a review, J. Clin. Chem. Clin. Biochem. from ETEC determined with the assistance of a nanobody, Structure 17 (2009) 18 (1980) 257–270. 255–265. [70] M.A. Corral-Rodriguez, et al., Nucleotide binding triggers a conformational change of [52] A.Y. Lam, et al., Nanobody-aided structure determination of the EpsI:EpsJ the CBS module of the magnesium transporter CNNM2 from a twisted towards a flat pseudopilin heterodimer from Vibrio vulnificus, J. Struct. Biol. 166 (2009) 8–15. structure, Biochem. J. 464 (2014) 23–34. [53] S.G. Rasmussen, et al., Structure of a nanobody-stabilized active state of the [71] Y. Hirata, et al., Mg2+-dependent interactions of ATP with the cystathionine-β- beta(2) adrenoceptor, Nature 469 (2011) 175–180. synthase (CBS) domains of a magnesium transporter, J. Biol. Chem. 289 (2014) [54] N.A. Mahmood, et al., Engineering of ion sensing by the cystathionine beta-synthase 14731–14739. module of the ABC transporter OpuA, J. Biol. Chem. 284 (2009) 14368–14376. [72] B. Xiao, et al., Structural basis for AMP binding to mammalian AMP-activated protein [55] E. Krissinel, K. Henrick, Inference of macromolecular assemblies from crystalline kinase, Nature 449 (2007) 496–500. state, J. Mol. Biol. 372 (2007) 774–797. [73] A. Rath, et al., Detergent binding explains anomalous SDS-PAGE migration of mem- [56] B.C. Jeong, et al., Change in single cystathionine beta-synthase domain-containing brane proteins, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 1760–1765. protein from a bent to flat conformation upon adenosine monophosphate binding, [74] E. Arutyunova, et al., Allosteric regulation of rhomboid intramembrane proteolysis, J. Struct. Biol. 183 (2013) 40–46. EMBO J. 33 (2014) 1869–1881. [57] E.F. Pettersen, et al., UCSF Chimera–a visualization system for exploratory research [75] M. Hattori, et al., Crystal structure of the MgtE Mg2+ transporter, Nature 448 (2007) and analysis, J. Comput. Chem. 25 (2004) 1605–1612. 1072–1075. [58] B.E. Kemp, et al., AMPK structure and regulation from three angles, Structure 15 [76] G. von Heijne, Membrane protein structure prediction. Hydrophobicity analysis and (2007) 1161–1163. the positive-inside rule, J. Mol. Biol. 225 (1992) 487–494. [59] K.A. Denessiouk, et al., Adenine recognition: a motif present in ATP-, CoA-, NAD-, [77] R.A. Laskowski, M.B. Swindells, LigPlot+: multiple ligand-protein interaction dia- NADP-, and FAD-dependent proteins, Proteins 44 (2001) 282–291. grams for drug discovery, J. Chem. Inf. Model. 51 (2011) 2778–2786. [60] I. Gomez-Garcia, et al., The crystal structure of protein MJ1225 from [78] K. Diederichs, P.A. Karplus, Better models by discarding data? Acta Crystallogr. D Methanocaldococcus jannaschii shows strong conservation of key structural features Biol. Crystallogr. 69 (2013) 1215–1222. seen in the eukaryal gamma-AMPK, J. Mol. Biol. 399 (2010) 53–70. [61] J.W. Scott, et al., CBS domains form energy-sensing modules whose binding of adenosine ligands is disrupted by disease mutations, J. Clin. Invest. 113 (2004) 274–284.