NMR analysis suggests the terminal domains of Robo1 remain extended but are rigidified in the presence of heparan sulfate

Protein expression and purification

The hRobo1-Ig1, hRobo1-Ig1-2, and hRobo1-Ig1-2-LBP4 constructs were synthesized using codons optimized for mammalian cell expression by GeneArt (Regensburg, Germany). The one-domain construct comprised residues 61 to 169 (Uniprot Q9Y6N7), while both two-domain constructs comprised residues 61 to 266. The lanthanide-binding loop was derived from the original Imperiali construct by replacement of amino acids at both ends with cysteines, deletion of an isoleucine group and replacement of a tryptophan with an alanine15. This resulted in the amino acid sequence CADTNNDGAYEGDELC, which was inserted between residues R221 and K224 while residues G222 and G223 were deleted (numbering from Uniprot entry Q9Y6N7). The constructs were inserted into pGEn2 expression vectors25. The vectors contained codons for an N-terminal hexahistidine-tag and GFP protein separated by a short linker and TEV cleavage site.

Both protein expression and purification proceeded as previously described13.26. Cleavage with TEV left a short scar on the N-terminus of the product (GSGG). A preliminary expression was performed in HEK293F cells, while a second isotope-labeled batch was expressed in HEK293S (GnT1MGAT1 knockout) cells (ATCC), which add primarily Man5GlcNAc2 glycans to N-glycosylation sites. Growth and expression of this batch proceeded in 500 mL of a custom version of FreeStyle 293 expression media (Gibco, ThermoFisher Scientific) which lacked both glucose and a select set of amino acids, namely lysine, phenylalanine, tyrosine, and valine. To this medium we added 2.5 g of 13C1-glucose and 75 mg of 13C-methyl-valine (Cambridge Isotope Labs, Tewksbury, MA), along with supplied supplements of lysine, phenylalanine and tyrosine. After two rounds of metal affinity chromatography and a round of size exclusion chromatography, the final yield was ~ 10 mg. Protein concentration was determined by UV/Vis absorbance measurement at 280 nm. An extinction coefficient of 20.315 M−1 cm−1 was predicted using the ProtParam tool on the Expasy webserver and used in the calculations27.

Native mass spectrometry

Purified hRobo1-Ig1-2-LBP4 from HEK293F cells was buffer exchanged into 10 mM ammonium acetate, pH 6.8 using Amicon microconcentrators (10 kDa molecular weight cutoff) for native MS and IM-MS measurements. Samples were immediately frozen and stored at − 20 °C prior to analysis.

Assessment of lanthanide binding was performed using a 12 T Bruker Solarix FT-ICR-MS instrument. The instrument was calibrated for high-m/z using 1 mg/mL sodium perfluoroheptanoate (Sigma-Aldrich) in 50:50 (v/v) % acetonitrile/water. Samples were prepared with 10 µM protein and 0, 10, 20, and 30 µM of LuCl3. Samples were infused by syringe pump at a rate of 2.0 µL/min and ionized via electrospray (ESI) at a voltage drop of 4500 V. Ions were desolvated with a skimmer 1 voltage of 100 V and accumulated in the collision cell for 1.0 s prior to injection into the ParaCell, where ions were trapped after a 1.4 ms time of flight delay. Ions were excited for broadband detection by a chirp waveform at 40% excitation power. Spectra were collected from m/z 500 to 5000 with 1 M points and a transient duration of 1.4 s. Each spectrum was the sum of 100 scans. Spectra were analyzed using Bruker DataAnalysis software.

Ion mobility mass spectrometry data were collected using a Waters Synapt G2-S instrument. The protein concentration was 10 µM. hRobo1-2-LBP4 + Arixtra spectra were collected on a solution prepared with an additional 10 µM Arixtra (fondaparinux sodium salt, Sigma-Aldrich). Solutions were directly infused at flow rates of 0.5–1.0 µL/min. ESI was achieved using fused silica emitter tips (ID 15 um, new objective) and a Waters Zspray source operating at capillary voltages of 1.8–2.5 kV. To ensure gentle ionization conditions the source block temperature was set to 30 °C, the sampling cone was kept at 30 V, and the extraction cone was set to 1.0 V. The traveling wave in the IMS cell had a velocity of 300 m/s and an amplitude of 21 V. IM-MS data was exported to a csv file using TWIMExtract28. Collision cross-sections were derived from arrival times, based on calibration with protein standards as described previously9.

NMR spectroscopy

NMR spectroscopy was performed on 13C1-Glucose, 13C-dimethyl-valine, labeled hRobo1-Ig1-2-LBP4. The protein was initially exchanged into a buffer composed of 25 mM Tris, 100 mM NaCl, pH 7.4, 0.02% NaN3, 10 µM DSS, 90/10% H2O/D2O at a final protein concentration of 300 µM. For PCS data a near molar equivalent (~ 0.9) of DyCl3 was added. Addition of diamagnetic LuCl3 lead to prohibitive levels of protein precipitation and a sample containing no lanthanide was used as a reference instead. For Arixtra binding experiments, Arixtra was titrated to a final concentration of 600 µM.

NMR spectra were acquired on a Bruker AVANCE NEO 900 MHz spectrometer using a triple resonance TXO cryo probe that is optimized for 13C and 15N observation. 13c, 1H-HETCOR spectra were recorded with 2048 × 192 points using the standard Bruker hxinepph sequence. Sweep widths of 61 ppm and 3.0 ppm were used for the 13C and 1H dimensions, respectively. INEPT delays were set to 1/(4 J) (2.0 ms), while the refocusing delays were set to 1/(12 J) (0.667 ms), where J was assumed to be 125 Hz. 96 scans were averaged. All spectra were processed using nmrPipe29. Peak picking was performed with NMRFAM-Sparky30.

3D-13C-edited-HSQC-TOCSY data were collected identically on 300 µM samples of both hRobo1-Ig1 and hRobo1-Ig1-2 constructs using the mlevhsqcetgp3d pulse sequence and recorded with 2048, 64, and 128 points in t3t2and t1 dimensions, respectively. The indirect dimensions were sampled with a 5% NUS scheme generated within TopSpin. A TOCSY mixing time of 60 ms was employed. The resulting spectra were processed in nmrPipe and reconstructed using the SMILE algorithm29.31.

Resonance assignment

Resonance assignments for valine methyl groups were made by initially focusing on a truncated hRobo1-Ig1 construct. The remaining valine methyl groups were then assigned to the Ig2 domain using a hRobo1-Ig1-2 construct without an inserted lanthanide binding loop. Assignments were further constrained by mutagenesis of six of the valines (V71, V133, V165, V188, V241 and V246) isoleucine. In both cases remaining assignments were determined using a custom script based on the Asssign_SLP program11. Experimental Chuhhuhand H chemical shifts measured from 3D-HSQC-TOCSY spectra were compared with predicted values ​​determined using ShiftX2 (Table S1)32, while mutagenesis information was included as constraints on the scoring function. Crosspeak assignments were readily transferred to the hRobo1-Ig1-2-LBP4 construct. The assigned valine methyl resonances were used to assist assigning alanine methyl signals by comparing their PCS values.

Model building and molecular dynamics simulations

A model of hRobo1-Ig1-2-LBP4 was built starting from an available X-ray crystal structure (PDB 2V9R) using UCSF chimera5.33. Initial coordinates for the LBP were taken from the X-ray crystal structure of interleukin-1β with a lanthanide-binding insertion sequence (PDB 3LTQ, residues 53A through 53P) and modified to match our modified construct12. Phi and psi angles near the fusion were adjusted to smoothly extend the beta sheet. Hydrogen atoms were added with the reduce tool34. Aspartic acid residue 97 was changed to the native asparagine and a Man5GlcNAc2 glycan was added using tleap35.

The system for a conventional molecular dynamics (cMD) simulation was prepared and run using Amber2018 with the ff99SB forcefield for amino acids and Glycam-06j forcefield for carbohydrates. The hRobo1-Ig1-2-LBP4 model was solvated with a truncated octahedron of TIP5P water and neutralized by the addition of sodium cations36. The resulting system was energy minimized with the Sander module using 25,000 steps of steepest descent minimization followed by 46,996 steps of conjugate gradient minimization. The system was slowly heated to 300 K over 1 ns and then the density was equilibrated for 1 ns in the NPT ensemble. The system was further equilibrated by 50 ns of NVT simulation. The production cMD run consisted of 1 µs of NVT simulation.

A gaussian accelerated MD (GaMD) simulation was performed starting from the same equilibrated system with randomized initial velocities19. Acceleration was performed with a dual-boost scheme applied to both the total potential and the dihedral potential. Energy statistics were calculated from an initial 14 ns NVT simulation and both ({sigma }_{0D}) and ({sigma }_{0P}) were set to 6.0. Likely conformers were selected by clustering the frames via their C RMSD in MATLAB using functions from MDToolBox37. Clusters containing more than 300 frames were chosen for energy reweighting, which was also performed in MATLAB using the approach described by Miao et al.19.

PCS analysis

PCS data were fit to PDB structures using custom MATLAB code. The magnetic susceptibility tensor was determined from the following linear system of equations:

where M is an × 5 matrix with row corresponding to a single atomic nucleus, Δχ is a 5 × 1 matrix containing five independent elements of the anisotropic part of the magnetic susceptibility tensor, and P is an × 1 matrix containing the PCS values ​​measured in ppms. The elements are constructed from the cartesian coordinates of the ion-nucleus vector as follows, where r is the vector length in meters:

$$M=frac{1}{{12pi r}^{5}}left[begin{array}{ccccc}frac{1}{2}left({2z}^{2}-{x}^{2}-{y}^{2}right)& frac{1}{2}left({x}^{2}-{y}^{2}right)& 2xy& 2xz& 2yz\ vdots & vdots & vdots & vdots & vdots end{array}right]$$


With at least 5 PCS values, the above system can be solved for Δχ. This approach is quite similar to that used by REDCAT (albeit for RDC calculations) and Paramagpy38.39. Agreement between a given structure and the PCS measurements was assessed by calculating a Q-factor, which is given by Eq. (3).

$$Q=frac{sum {left({PCS}_{obs}-{PCS}_{calc}right)}^{2}}{sum {left({PCS}_{obs }right)}^{2}}$$


Ig1 orientational grid search

A grid search over possible Ig1 orientations was conducted using a MATLAB script. A starting structure was produced using the representative frame of the most probable cluster from the GaMD simulation. This frame was then aligned with an X-ray crystal structure (PDB 2V9R) via the Ig2 domain and saved relative to the X-ray coordinate system. Different orientations of the Ig1 domain were generated using Euler angle rotations (ZXZ convention) about the alpha carbon within residue Ile 104. For each Euler angle 50 values ​​were used, creating a 3-dimensional grid of 125,000 possible orientations. Each orientation was scored by calculating a Q-factor (Eq. 3) for the Ig1 PCS data. the (Delta chi ) tensor used to back-calculate PCS values ​​was determined from an Ig2 dataset made by combining data from both Arixtra-bound and unbound samples. The principal values ​​of the tensor in m3 are: ({Delta chi }_{xx}=-4.3times {10}^{-32}), ({Delta chi }_{yy}=-34times {10}^{-32})and ({Delta chi }_{zz}=38times {10}^{-32}).

Leave a Comment

Your email address will not be published. Required fields are marked *