Supplementary Information
Supplementary Figures
Supplementary Figure S1. Peak-picking of the HSQC spectrum of human ubiquitin in Bruker TopSpin 3.2 for setup of the F1F2-selective experiments. The numbers are temporary indices before assignment. After peak-picking, 72 F1F2-selective (HNCA)NH experiments were set up and recorded, each of which selectively probes one of the amide resonances shown here.
Supplementary Figure S2. Establishing the amino-acid connectivity along the primary sequence of ubiquitin by the F1F2-selective (HNCA)NH experiment. As shown in Figure 3 of the main text, each F1F2-selective (HNCA)NH experiments yields a connectivity between three amino acids (i−1), (i) and (i+1) (hereafter referred to as: triplet). These triplets are shown here in orange and aligned in such a way that a sequential connection along the primary sequence is established due to the overlap between two triplets (see main text for details). The numbers of the triplets refer to the peaks picked as shown in Supporting Figure S1. Additional information about the amino acid type from the amino-acid type-selective experiments (“MUSIC”) is shown in blue. Combining the information from amino-acid type-selective experiments and the connectivity between the triplets, one can fill in the primary sequence of ubiquitin into the missing gaps (yellow). The connectivity breaks at proline residues (Pro19, Pro38 and Pro39) or residues that are not observed in the HSQC spectrum due to conformational exchange broadening (Glu24 and Gly53). Thus, the assignment is obtained separately in 7 fragments. Note that only a few of the MUSIC experiments were used to complete the assignment. The remaining experiments can be used for validation. X: connectivity not unambiguous or end of cluster (N- or C- terminus, proline residue or conformational exchange peak).
Supplementary Figure S3. F1F2-selective NOESY spectra of human FABP4 taken at temperatures of 298 K (upper panel) and 287 K (lower panel).
Supplementary Figure S4. F1F2-selective R1ρ relaxation dispersion experiments on human FABP4. A) An HSQC spectrum of FABP4 indicating examples of resonances for which relaxation dispersion experiments were conducted. B) R1ρ relaxation dispersion profiles obtained for Gly35 and Ile63 at a temperature of 298 K using the pulse sequence of Figure 5.
Supplementary Figure S5. Algorithm for automated backbone assignment of proteins from 2D (HNCA)NH spectra exclusively. Additional information from amino-acid type-selective experiments can be incorporated at the step “read amino acid information”.
Supplementary Tables
Supplementary Table S1. Time domain and spectral width of the F1F2-selective experiments discussed here.
F1F2-selective experiment / Complex points(direct / indirect dim.) / Sweep width [ppm]
(direct / indirect dim.)
(HNCA)NH / 1024 / 24 / 14.0 / 34.0
NOESY-[1H, 15N]-HSQC / 1024 / 24 / 12.0 / 40.0
NOESY-[1H, 13C]-HSQC / 1024 / 48 / 12.0 / 40.0
[15N]-R1ρ relaxation dispersion / 1024 / 8 / 12.0 / 1.2
Supplementary Table S2. Relative sensitivity of the experiments reported in this study as compared to a sensitivity improved HSQC experiment taken under equivalent conditions (number of scans, the time domain, sweep width, receiver gain, relaxation delay etc.). The downfield amide resonance of Ile13 of ubiquitin was used for intensity quantification. In the NOESY experiment, the peak corresponding to auto-relaxation was quantified (mixing time: 20 ms). In the case of the R1ρ relaxation dispersion experiment, the spin-lock was omitted.
Experiment / Intensity ratio [Iexperiment / IHSQC](HNCA)NH / 3 %
NOESY-[1H, 15N]-HSQC / 43 %
[15N]-R1ρ relaxation dispersion / 72 %
Supplementary Methods
De novo assignment from F1F2-selective (HNCA)NH
Without any prior assignment information on ubiquitin, we quickly connected the amino acid triplets obtained from the F1F2-selective (HNCA)NH spectra to form seven larger fragments in ubiquitin, thereby establishing sequential connections between all the amide resonances observed in the HSQC spectrum (Supplementary Figures S1 and S2). To assign these fragments to the primary sequence of ubiquitin, we obtained 2D amino acid-type selective spectra (Schubert et al. 2000; Schubert et al. 2005; Schubert et al. 2001a; Schubert et al. 2001b; Schubert et al. 2001c; Schubert et al. 1999). By simple overlay with the HSQC spectrum, these experiments showed quickly which of the amide cross-peaks in the HSQC spectrum of ubiquitin were glycine, alanine, arginine, aromatic (Phe/Tyr/His), glutamine, serine, aspartate, asparagine and proline-preceding (Pro−1) residues (Supplementary Figure S2). The combined information of the F1F2-selective (HNCA)NH experiments and the amino acid type selective experiments allowed unambiguous complete assignment of all observed peaks in the HSQC spectrum of ubiquitin (Supplementary Figures S1 and S2).
We note that there is a convenient degeneracy in the F1F2-selective (HNCA)NH dataset, because a given residue u “sees” two residues v and w, but u is also seen from w and from v. Thus, if the experimental data connecting u → v somewhat ambiguous (e.g. due to peak overlap or low signal-to-noise ratio), one can check if u is seen from v to resolve the ambiguity. Moreover, since not all of the 16 amino-acid-type selective experiments had to be used to obtain complete assignment of ubiquitin, these data allow an independent consistency check of the obtained assignment. Together with the arguments presented in the main text, this establishes that the F1F2-selective experiments can be used to backbone assignment of proteins de novo.
Amino acid-type selective experiments
All experiments were acquired using the default Bruker pulse programs available in TopSpin 3.2 acquisition software: music_de_3d (Asp/Glu), music_fhyw_3d (Phe/His/Tyr/Trp), music_gly_3d (Gly), music_ser_3d (Ser), music_ile_3d (Ile), music_qn_3d (Asn/Gln), music_kr_3d (Lys/Arg), music_lavia_3d (Leu/Ala; Val/Ile/Ala), music_tavi3d (Val/Ile; Thr/Ala) music_pro_1_3d (Pro−1). In total, 16 datasets were recorded with different labeling schemes. All of these spectra were recorded as 2D experiments by setting the number of data points to be acquired in the F1 dimension to 1. Note that only a few of the spectra were actually used to establish the assignment (Supplementary Figure S2; blue rows labeled “MUSIC”). Accordingly, the additional spectra are beneficial for validation of the assignment.
New pulse programs
The pulse program and parameters files (in Bruker format) are available on the homepage of our laboratory: http://www.moleng.kyoto-u.ac.jp/~moleng_01/
F1F2-selective (HNCA)NH Pulse Program Parameters
The parameters used were as follows. Most of these parameters were adopted from the original publication of the HN(CA)NNH experiment (Weisemann et al. 1993) or the Bruker file hncannhgp3d.2.
φ1 / yφ2 / x
φ3 / y, −y
φ4 / 2(x) 2(−x)
φ5 / 8(x), 8(−x)
φ6 / 4(x), 4(−x)
φ7 / 4(−y), 4(y)
φrec / x, −x, −x, x, −x, x, x, −x
G1 / 25 G/cm
G2 / 25 G/cm
G3 / 11.5 G/cm
G4 / 40 G/cm
G5 / 4.05 G/cm
F1F2-selective NOESY-[1H, X]-HSQC Pulse Program Parameters
The parameters used were as follows. Most of these parameters were adopted from the sensitivity-improved NOESY-HSQC experiment or the Bruker file noesyhsqcf3gpsi3d.
φ1 / −yφ2 / 2(−x) 2(x)
φ3 / 4(x) 4(−x)
φ4 / x
φ5 / x (−x)
φ6 / 8(x) 8(−x)
φ7 / 8(x) 8(−x)
φ8 / 8(y) 8(−y)
φrec / x, −x, −x, x, −x, x, x, −x
Δ / 1 / 4JHX
G1 / 25 G/cm
G2 / 15 G/cm
G3 / 25 G/cm
G4 / 40 G/cm
G5 / 2.5 G/cm
G6 / −1 G/cm
G7 / 4.05 G/cm
F1F2-selective 15N relaxation dispersion Pulse Program Parameters
The parameters used were as follows.
φ1 / 2(y) 2(−y)φ2 / x, −x
φ3 / y
φ4 / −y
φ5 / x
φ6 / −y
φrec / x, −x, −x, x
G1 / 3 G/cm
G2 / 15 G/cm
G3 / −30 G/cm
G4 / 40 G/cm
G5 / 2.5 G/cm
G6 / −1 G/cm
G7 / 4.05 G/cm
1