A Double‐Armed, Hydrophilic Transition Metal Complex as a Paramagnetic NMR Probe

Abstract Synthetic metal complexes can be used as paramagnetic probes for the study of proteins and protein complexes. Herein, two transition metal NMR probes (TraNPs) are reported. TraNPs are attached through two arms to a protein to generate a pseudocontact shift (PCS) using cobalt(II), or paramagnetic relaxation enhancement (PRE) with manganese(II). The PCS analysis of TraNPs attached to three different proteins shows that the size of the anisotropic component of the magnetic susceptibility depends on the probe surroundings at the surface of the protein, contrary to what is observed for lanthanoid‐based probes. The observed PCS are relatively small, making cobalt‐based probes suitable for localized studies, such as of an active site. The obtained PREs are stronger than those obtained with nitroxide spin labels and the possibility to generate both PCS and PRE offers advantages. The properties of TraNPs in comparison with other cobalt‐based probes are discussed.


Introduction
Ever since the determination of first metalloprotein structures using paramagnetic NMR restraints, [1] it has been acknowledged that paramagnetism is ap owerful tool for the study of biomolecules.The interactions of unpaired electrons with nuclei generate paramagnetic effects that contain structural information. Thep seudocontact shift (PCS) and paramagnetic relaxation enhancement (PRE) are the paramagnetic effects that most frequently find application. [2] PCS offer long-range distance and conformational information and can be measured easily with high precision. PREs only yield distance information but are strongly distance-dependent, making them exquisitely sensitive to minor states in which the nuclear-electron distance is reduced. [3] In proteins that bind metals naturally,ap aramagnetic center is already present or can be introduced by replacing ad iamagnetic metal ion, like Ca II or Mg II ,with aparamagnetic ion, such as Co II ,M n II ,o ralanthanoid, Ln III . [4] Foro ther proteins,t he introduction of ap aramagnetic center is required, either through genetic means [5] or chemical attachment. To limit the effects on the protein and obtain asingle set of paramagnetic effects,t he ideal chemical probe has no net charge and is hydrophilic,p ositioned rigidly relative to the protein, and of high symmetry.
Over the past years,m any paramagnetic NMR probes were designed and synthesized. [6] Ln III has been the paramagnetic center of choice in many cases because of the range of anisotropic components of the magnetic susceptibilities (described by the Dc tensor) that these ions display,with sizes of 2 10 À32 m 3 for Eu III ,8 10 À32 m 3 for Yb III and up to 50-84 10 À32 m 3 for Tm III ,T b III ,a nd Dy III . [7] Thes imilarity in coordination chemistry makes it possible to use the same probe with different Ln III ions.F ewer probes for transition metals have been reported. Of the transition metal ions,highspin Co II yields among the largest PCSs,w ith tensor sizes in the order of 2-7 10 À32 m 3 , [1c] and displays weak PREs, whereas Mn II causes strong PREs owing to the presence of five unpaired electrons and al ong electronic relaxation time. [8] Theu se of transition metal ions as paramagnetic centers for protein structural studies has already al ong history. [9] However,o nly few site-specific transition metal probes have been designed for protein NMR studies.S -(2-pyridylthio)cysteaminyl-EDTAi sacommercially available (TRC,T oronto,C anada) transition metal probe.H owever,i ty ields multiple PCS for an ucleus owing to the presence of stereoisomers of the complex. [10] Some other probes are attached to the protein through as ingle arm and require additional coordination by ar esidue of the protein near the attachment site. [11] Recently,two-point attachment was introduced for Co II probes to ensure that the metal ion is rigid relative to the protein. [12] Foro ne of these probes,m etal ion exchange between solvent and probe was observed. [12a] This prompted us to develop adouble-armed transition metal ion probe that could tightly bind the metal ion and generate as ingle set of paramagnetic effects.C yclen (1,4,7,10-tetraazacyclododecane) is aw idely used building block for metal ligand design owing to its high metal binding affinity.Alarge number of cyclen derivatives have been developed for metal ions for the application in biomedicine [13] and magnetic resonance imaging (MRI), [14] for which high thermodynamic stability and kinetic inertness are required. Among these reported cyclen-based complexes,t here are,h owever,f ew successful Co II and Mn II complexes,i np articular for Mn II , which can easily be oxidized by air. [15] Herein, we report the design and synthesis of several C 2symmetric cyclen derivatives,w hich can tightly bind the transition metal ions Co II and Mn II and are stable in air and buffers.T hese transition metal NMR probes (TraNPs) were tested using three proteins.T he TraNPs are linked to the proteins trough two disulfide bonds at as pecific location on protein surface,yielding single sets of paramagnetic effects in NMR spectra. Interestingly,the sizes of the Dc tensor of Co-Tr aNP are in the range of 2-5 10 À32 m 3 for different proteins, indicating that the protein environment has an indirect effect on the Co II coordination that influences the Dc tensor. We discuss the Tr aNP properties and compare them with other reported Co II tags.

Results and Discussion
Derivatives of 1,4,7,10-tetraazacyclododecane (cyclen) have been used extensively for metal binding because of their favorable metal coordination properties. [16] Foratwoarmed transition metal probe,C 2 -symmetric derivatives that allow attachment to two cysteiner esidues are required. [7b] A selective partial protection and alkylation strategy was used to synthesize the novel C 2 -symmetric metal binding ligands Tr aNP1-SS and Tr aNP1-RR, as well as the control compounds Tr aNP3-S and Tr aNP5 ( Figure 1a nd the Supporting Information, Schemes S1 and S2). Except for Tr aNP5, which has poor affinity for metals ions,t he metal complexes are stable under atmospheric condition. In addition, N-(carboxymethyl)-S-(pyridin-2-ylthio)cysteine was synthesized (Figure 1, tag 1)a nd was used for comparison with Tr aNPs. [12a] Three proteins were used to characterize the paramagnetic properties of the Tr aNPs,T4lysozyme (T4Lys) with the mutations K147C/T151C, Bacillus circulans xylanase (BCX) with mutations E78Q/T109C/T111C and ubiquitin with the mutations E24C/A28C.The mutation E78Q in BCX abolishes catalytic activity but is otherwise irrelevant for this study. These three proteins contain no cysteineresidues apart from the pair introduced for probe attachment. In T4Lys and ubiquitin, the two cysteine residues are located in an a-helix, whereas in BCX, they are in a b-strand. TheP CS caused by Co II on the amide groups in the proteins were obtained by taking the difference of the 1 H N chemical shifts in the spectra of the Co II and Zn II -tagged proteins (Figure 2and Figures S1 and S2). TheP CS were fitted to Equation (S1) in the Experimental Section (Supporting Information) to obtain the two components of the anisotropic part of the magnetic susceptibility, Dc ax and Dc rh ,t he orientation of the Dc tensor and the position of the paramagnetic center relative to the protein structures.
ForT raNP1 T4Lys (K147C/T151C), BCX (E78Q/T109C/ T111C) and ubiquitin (E24C/A28C) more than 80, 100, and 40 PCS were measured and fitted against published structures.  Theresults are presented in Table 1and Tables S1-S3. Agood correlation was found between the experimental and backcalculated PCS ( Figure S3). Only the PCS of the amide group of T4Lys residue 163 deviates strongly from the calculated value.T his residue is located at the C-terminus and its location in the structure may be ill-defined. It was excluded from the calculations.T he results for different structures of the same protein were essentially the same (Tables S1-S3).
Theiso-surfaces of the Dc tensors,plotted on T4Lys,BCX, and ubiquitin, are shown in Figure 3a nd Figure S4. The Dc tensors are mostly axial, with only am inor rhombic component. Theiso-surfaces of the two enantiomers are very similar. TheC o II ions are located between the cysteiner esidues,a s expected for at wo-armed probe,a nd in line with results obtained for the lanthanoid probes,C LaNP5 and CLaNP7. [7b,17] Thedistances between the Co II and the cysteine C a atoms are between 7.6 and 9.8 (Table 1). Thed istance between Co II ions in Tr aNP1-RR and TraNP1-SS bound to T4Lys is only 1.4 and the Dc tensor orientations are similar ( Figure 3). However, the Dc ax of Tr aNP1-SS is 12 %s maller than that of Tr aNP1-RR (Table 1). Also on BCX, Tr aNP1-RR and TraNP1-SS position the Co II ions in similar locations and the Dc tensor frame orient in the same way.Inthis case, Dc ax of Tr aNP1-SS is even 32 %s maller than for TraNP1-RR. Interestingly,the Dc ax of Tr aNP1-RR (SS) attached to BCX is 27 %( 44 %) smaller than for this probe attached to T4Lys (Table 1). Foru biquitin, the Dc tensor of Co-TraNP1-SS labeled on ubiquitin was even smaller than for the other two proteins (Table 1). These differences will be discussed later.
To investigate whether Tr aNP1 can also be used to generate PREs with at ransition metal, Tr aNP1-SS was loaded with Mn II and linked to T4Lys K147C/T151C.P REs were obtained by comparing the intensities of amide resonances in HSQC spectra of Mn II and Zn II -tagged T4Lys samples ( Figure S5). Figure 4A shows the intensity ratios. From these,t he PREs and Mn II -1 H N distances were derived [Supporting Information, Experimental Section, Eqs.(2) and 2lzm [18] 2lzm [18] 2bvv [19] 2bvv [19] 2mjb [20] [a] in 10 À32 m 3 ;[ b] Adjusted Q-value, see Eq. (S2).  ( 3)].T hese distances were compared with those obtained from the PCS-based Co II position, assuming that Mn II and Co II in TraNP1-SS sit in the same position relative to T4Lys.A good correlation, with am argin of AE 3 ,f or distances between 19 and 31 was found ( Figure 4B). Forp eaks that broadened beyond detection, the observed distance was set to 19 and for the amide groups with unaffected peak intensities the distance was set to 31 ,explaining the points on these two horizontal distance lines in Figure 4B.T hree clear outliers were observed (residues 10, 31, and 32). It is not obvious why these amide groups give deviating results.T he assignments appear correct, and the structure is well-defined for these residues.T he position of the Mn II ion was also determined by fitting it to the experimental distances (see Supporting Information for details). These calculations place the Mn II ion 2.5 away from the Co II position based on the PCS data ( Figure S6 A). Thee xperimental and back-calculated distances correlate within the AE 3 range for most residues,e xcept for residues 10, 31, and 32, which deviate 4-5 (Figure S9 B). Exclusion of these three residues yielded ab etter fit (Figure S6 C). These and other calculations showed that the exact calculated position of the Mn II ion is strongly dependent on the input data set ( Figure S6 A). It is estimated that the PRE-based position of the metal has aprecision of 2-3 and is less precise than the location based on the PCS data. However, the metal positions obtained through both approaches are consistent, being in between the two cysteine residues.
To determine which of the pending arms coordinate the metal in Tr aNP1, Tr aNP3-S and Tr aNP5 were synthesized, lacking one or both hydroxy-propionic acid groups,r espectively.T raNP5 failed to bind metals,w hereas Tr aNP3-S was capable of coordinating Co II .W hent his complex was linked to T4Lys,two sets of PCS were observed for each amide group and the PCS were smaller than those observed for Tr aNP1 (Figure S7 A). As compared to TraNP1, the C 2 -symmetry is broken in TraNP3, resulting in two isomers upon attachment to the protein, which could be the cause of the double resonances.Itsuggests that the amide groups in the other two pending arms are incapable of coordinating the Co II ion, so these arms have additional rotational freedom compared to the coordination arms of the lanthanoids binding counterparts,C LaNP5 and CLaNP7. [7b,17] Lanthanoids require eight or nine ligands,s oa ll pending arms are involved in metal coordination. We also tested lanthanoid binding to Tr aNP1. Theaffinity for these metals is poor.
To compare Tr aNP1 with another two armed Co II -tag,we synthesized the published Co II probe named tagging agent 1( tag 1), N-(carboxymethyl)-S-(pyridin-2-ylthio)cysteine, Figure 1. In this case,the two cysteine residues on the protein each react with one probe molecule and the Co II ion is sandwiched in between the two attached groups.S warbrick et al. [12a] attached this probe to ubiquitin E24C/A28C and reported ar emarkably large Dc tensor (À7.4 10 À32 m 3 ). We repeated the experiment with the same ubiquitin variant and also attached tag 1t oT 4Lys K147C/T151C.L abeling was confirmed with mass spectrometry (Figures S8 and S9). Forall the samples linked to either tag 1, Co II -tag 1, or Zn II -tag 1, the same mass was observed, of the free protein plus 409 Da. This mass difference equals the mass of two attached tag 1 molecules (354 Da) plus an additional 55 Da. As this extra mass was present independent of the presence of either Co II or Zn II in the sample,wea ssume that tag 1loses these metal ions and picks up additional mass in the process of the mass spectroscopy measurement, for example,F e III or Mn II ions. 1 H-15 NH SQC spectra of the Co II and Zn II -tagged ubiquitin and T4Lys were recorded and the PCS measured. Thes ize and orientation of the Dc tensor of tag 1d erived from the experimental PCS analysis of the ubiquitin NMR spectra were the same as those obtained using the published PCS values (Figures S10 Aa nd S11 and Table S4).
In our spectra, some of the residues showed more than asingle paramagnetic peak, such as residues K6, T7, and H68 ( Figure S11). Moreover,C o II loading appeared to be incomplete.I nt he NMR spectra of ubiquitin tagged with Co IIloaded tag 1, diamagnetic peaks were present (Figure S11 B), even if 1.2 equiv of Co II was added, rather than the reported 0.6 equiv.T he tagged, metal-free ubiquitin also behaved curiously,s howing many double peaks in the HSQC spectrum, compared to untagged ubiquitin. Upon addition of Zn II , single peaks remain in the HSQC spectrum (Figure S11 C). Thus,the metal-free tag 1caused the presence of two forms of the protein. Also for T4Lys K147C/T151C,t ag 1s howed partial loading with Co II or Zn II ,e ven with 10 equiv of the metal ion added. Again, more than one peak with PCS were observed for some of the residues ( Figure S12). From the tag 1-T4Lys HSQC spectrum, around 50 PCS were measured and used for Co II positioning and Dc tensor calculation. The Co II is located between the two cysteine residues and the distances to the two cysteine C a atoms are around 7 (Figure S10 B), which is similar to the results for the ubiquitin variant. As observed with TraNP1, the size of the tensor was affected by the protein because the Dc ax component of Co IItag 1was,though still quite large,somewhat smaller than for the ubiquitin variant, and the Dc rh component was more than three times smaller (Table S4).
Thes ynthesis and characterization of an ew two-armed transition metal binding NMR probe,T raNP1, are reported. Loaded with Co II and attached to the proteins T4Lys,B CX and ubiquitin through two disulfide bridges,itcauses PCS of the resonances of amide nuclei. TheP CS can be fitted well and yield the position of the metal relative to the protein as well as the orientation and size of the Dc tensor. Whereas metal position and tensor orientation are similar, interesting differences are observed in the sizes of the Dc ax and Dc rh between the two enantiomers,a sw ell as between T4Lys, BCX, and ubiquitin. We are puzzled by the large variation in these tensor sizes for cobalt ions that are expected to have the same coordination environments.Weattribute these effects to slight differences in coordination of the cobalt ion. The binding of the probe to the protein and interactions with protein side-chains may introduce slight strain on the Co II ligands,l eading to changes in the electron distribution and thus in the paramagnetic effect. It is likely that the cobalt ion is coordinated by the four ring nitrogen atoms and two carboxy oxygen atoms in ap seudo-octahedral conformation.
Based on the crystal structure of as imilar compound, [21] the structure of Tr aNP1 was modelled using Spartan '14 & Spartan'14 Parallel Suite (www.wavefun.com). Thet wo enantiomers indeed show slight differences ( Figure S13). Thel inkage of Tr aNP1 to the proteins was also modelled, using XPLOR-NIH. [22] In this model, the coordination obtained in the Spartan model was constrained and the arms for attachment were free to rotate.The protein backbone was fixed, and side chains were allowed to rotate.T he position of the Co II ion was restrained to the experimental position. An acceptable model was obtained in which the plane of the cyclen ring is roughly perpendicular to the surface of the protein and the arms for attachment point away from the ring, relative to the hydroxy-propionic acid groups ( Figure 5). In the BCX-TraNP1-SS model, Ty r113 is located within hydrogen-bonding distance of one of the hydroxy groups.W e speculate that such an interaction with as ide chain could affect the ligand coordination of the cobalt ion and influence the size of the Dc tensor. It is concluded that the paramagnetic properties of Co II are very sensitive to the environment of the ligands,s trongly affecting the size of the anisotropic component of the magnetic susceptibility,i nl ine with the large differences reported for the size of paramagnetic effects in other Co II -compounds. [4a, 11a, 23] These observations are strikingly different from lanthanoid probes.F or rigidly attached probes,s uch as CLaNP-5 and CLaNP-7, usually similar sizes for Dc ax and Dc rh are found, independent of site of attachment. [7b, 17] As ac onsequence,the Dc ax and Dc rh need to be determined for Tr aNP1 in each system, whereas for CLaNP,t he sizes can, to first approximation, be estimated on the basis of literature values. Figure 6presents acomparison of the results for Co II -TraNP1-SS and Yb III -CLaNP5 attached to T4Lys at residues 147 and 151. [24] Them etals are 2.6 apart, though both are located between the Cys residues.Also,the direction of the z-axis of Figure 5. Model of Co-TraNP1-SS attached to BCX E78Q/T109C/ T111C. The protein is represented in cartoon mode and the two cysteine residues and the tag were modelled in the structure (PDB entry 2bvv). [19] The oxygen of Tyr113 (in cyan sticks) can readily be brought into hydrogen bond distance of one of the hydroxy groups of TraNP1-SS. The cysteine residues and the probe are shown in sticks, with carbon atoms in yellow and the nitrogen, oxygen, and sulfur atoms in blue, red, and dark yellow,respectively.The Co II ion is shown as asphere. the tensor and the degree of rhombicity differ considerably. Figure 6C shows the models of the probes attached to T4Lys. In CLaNP,a ll the four pending arms are involved in the coordination of the metal and thus are oriented in the same direction, placing them in such aw ay that the cyclen ring is roughly parallel to the protein surface and the metal in between the cyclen ring and the protein. In Tr aNP,the cyclen ring is perpendicular to the surface and the metal is thus on one side of the ring, relative to the protein.
Thec omparison of TraNP1 and tag 1c onfirmed that the coordination of the Co II atom has alarge effect on the size of the Dc tensor. Fort ag 1, four carboxyl groups and two secondary amines are likely involved in the coordination, whereas TraNP1 has two carboxyl groups and four tertiary amines contributing to the coordination. However,both tags are likely to provide adistorted octahedral environment, so it is unclear whether the different type of ligands are the cause of the large Dc tensor of tag 1, yielding larger PCS than obtained with Tr aNP1 and approaching those obtained with Yb III probes.The metal is roughly located at the same position in both probes,inbetween the cysteines and 7-8 away from the C a atoms.H owever,t he tensor axes are oriented very differently ( Figure S10).
Ad isadvantage of tag 1i st hat some amides show more than one resonance in the spectrum of the paramagnetic sample.F urthermore,t he metal affinity appears to be relatively low,leading to the presence of peaks of metal-free tagged protein in the spectra of the paramagnetic sample, increasing spectral crowding that could be problematic for larger proteins.N ext to the doubly anchored tag 1, several tags with as ingle attachment point were reported, like (2pyridylthio)-cysteamine-EDTA, 2-vinyl-8-hydroxyquinoline (2V-8HQ), [11a] and 3-mercapto-2,6-pyridinedicarboxylic acid (3MDPA). [11b] As ac ommercially available probe,( 2-pyridylthio)-cysteamine-EDTAi sw idely used in protein paramagnetic NMR after it was first reported with Fe III as the paramagnetic center. [25] Further research found that the Co IIloaded (2-pyridylthio)-cysteamine-EDTAgenerated two sets of PCS,i no ne case [10] but not in other, [26] owing to the presence of stereoisomers of the complex. The2 V-8HQ is ar igid and small probe for Co II and requires additional ligands from ap rotein side-chain, making the metal location less predictable than for atwo-armed probe.Slow exchange of Co II ions between the solvent and the 2V-8HQ tag on ubiquitin was reported. Thea ffinity and exchange rate depend on the coordinating side chain. For3 MDPA, which can bind Ln III and Co II ,t he tensor orientations for all the metal ions are similar but the metal affinity is very weak.
To reduce the probe attachment flexibility,g enetic incorporation of natural or unnatural amino acids in the protein sequence has been proposed. Theu nnatural amino acid bipyridylalanine (BpyAla) [27] and 2-amino-3-(8-hydroxyquinolin-3-yl)propanoic acid dihydrochloride (HQA) [28] , which both have as ide chain strongly chelating transition metal ions,were successfully introduced into West Nile virus NS2B-NS3 protease (WNVpro) and membrane proteins (1TM-CXCR1 and p7), respectively.S imilar to 2V-8HQ, both require additional ligands provided by protein side chains.HQA was used for Mn II to measure PRE. Recently,it was reported that also adihistidine (diHis) motif can be used to bind Co II . [12b] Themotif was introduced to ubiquitin on an a-helix, as well as a b-strand of GB1. Also in this study,t he dHis motif generated different Dc tensor values and orientations for the different protein variants.

Conclusions
Tr aNP1 adds an ew probe to the range of paramagnetic probes available for NMR on biomolecules. [2e, 6a] Co II has as maller anisotropic magnetic susceptibility than most lanthanoids,p lacing it close to Pr III . [4b] Its application can lie in studying small and local structural changes in proteins and protein complexes,s uch as can occur in enzyme active sites during catalysis or in protein pockets upon ligand binding. Thee ffects of stronger lanthanoids are less suitable for studying such small structural changes.The reason is that such probes cannot detect changes close to the probe owing to PRE effects and events further away require larger structural changes to cause significant PCS changes.A1 change at al arge distance from ap aramagnetic metal leads to as mall relative change in angle and distance and thus asmall change in PCS,both in absolute and relative terms,even for strongly paramagnetic ions.This point is illustrated with an example in Table S5. Thus,Y b III and the stronger Ho III ,D y III ,T m III ,a nd Tb III are suited for studying domain motions and determination of structures of complexes, [2e] whereas probes with smaller Dc tensors are suitable for detection of nearby,small structural changes.T he relatively low paramagnetic anisotropy of Co-probes makes the measurement of paramagnetically induced RDC inconvenient at routine fields,s uch as 14 T (600 MHz), with maximal predicted values for 1 H- 15 No f1 .7 and 4Hzf or Tr aNP1 and tag 1, respectively.A tt he highest fields achievable,o btaining these RDC becomes realistic (7 and 16 Hz, respectively,a t2 8T ,1 200 MHz), offering possibilities for the study of protein mobility.F inally,the fact that Tr aNP1 can also bind Mn II to measure PRE is convenient because PRE-derived distances complement the restraints obtained from PCS.