Crystallographic structure determination of FAS in the absence of the γ-subunit

The FAS sample used for growing crystals was prepared similarly to the one used for cryo-EM sample preparation. I crystallized the FAS using the sitting drop vapor diffusion technique with the following buffer: 0.3 M sodium malonate, 11.5-13% (w/v) PEG3350 and 0.1 M HEPES pH 7. The crystals diffracted X-rays to 2.9 Å and belonged to the primitive monoclinic space group P2¹ with unit cell constants of a = 217.6 Å, b = 347.6 Å, c = 265.3 Å,

β = 107.9° with a single FAS molecule in the asymmetric unit. The structure was solved by molecular replacement using the previously reported 3.1 Å S. cerevisiae structure (Leibundgut et al., 2007). The resulting electron density displayed features corresponding to side chains throughout the whole molecule including the PPT domain, which was absent in the model used for phasing (Figure 20(i)). The PPT domain was thus modelled into the density and was found to have a similar structure to that solved for an isolated PPT domain (Johansson et al., 2009) (Figure 20(ii)). The only sequences that could not be modelled in this structure were the same flexible loop segments which have always been unresolved in previous reported structures. The model after refinement had average B-factors of 92 Ų and R/R^free of 19.7% / 21.8%. The diffraction data was also analyzed using the STARANISO server, where a mild anisotropy in diffraction of ~0.1-0.2 Å was detected (Supplementary Figure S2). Anisotropically truncated data, taking into account diffraction spots up to 2.8 Å in certain directions, was then used for model refinement. The side chain and solvent densities improved resulting in an improved model with average B-factors reduced to ~ 73 Ų and R/R^free of 19.2% / 21.1% (Supplementary Figure S2, Supplementary Table 4).

Figure 20. Crystallographic structure of the FAS in the absence of the γ-subunit. (i) X-ray crystal structure of FAS at 2.9 Å resolution. Several excerpts within the respective enzymatic domains displayed in orange are utilized to show the quality of the final 2mFo-DFc electron density maps contoured at: (a) DH domain 1.5 σ, (b) AT domain 1.5 σ, (c) ER domain 1.5 σ, (d) KR domain 2.0 σ and (e) KS domain 2.0 σ. (ii) PPT domain (red) and an α-helical sequence (grey) along with their corresponding density contoured at 1.0 σ. These regions were not visualized together with the FAS in previous high resolution structures of this complex (Leibundgut et al., 2007).

The crystallographic FAS structure was similar in conformation to the non-rotated cryo-EM model. The Cα atoms of the crystallographic model and the non-rotated cryo-EM model were compared and had an overall RMSD of 0.92 Å (Figure 21). The first 150 residues of the AT domain along with two segments in the MPT domain (1737-1750, 1850-1970) differed the most between the two models. Upon excluding these residues from the calculations, the RMSD decreased to 0.67 Å. To put these variations into context, published structures of identical proteins are known to have a Cα atom RMSD in the range of 0 - 1.2 Å (Kufareva and Abagyan, 2012). These deviations are attributed to inherent protein flexibility and experimental resolution limits. Therefore, RMSD values between 0 - 1.2 Å are considered here to infer that the two structures being compared are nearly identical. Furthermore, the overall RMSD between the published crystallographic FAS structure (Leibundgut et al., 2007) and the one from this study is ~1 Å, whereas, between the published crystallographic FAS structure and the cryo-EM structure from this study is ~ 0.7 (Supplementary Figure S3). The most variations observed in these two comparisons are are also found at the AT and MPT domains. The MPT domain is involved in forming crystal contacts and thus can vary between the EM and crystal structures as well as among the crystals belonging to different space groups. In general, the RMSD values of below 1 Å between the published FAS structure and the ones from this study indicate that the structure of the FAS does not vary significantly between the two structural techniques.

After validating that the structure of the FAS determined by X-ray crystallography and cryo-EM remains almost identical, I then compared the two conformations of the FAS obtained from the endogenous FAS holoenzyme sample and the FAS sample in the absence of the γ-subunit that were described in the previous sections.

Figure 21. Comparison among FAS models derived from X-ray crystallography and cryo-EM. The RMSD between the cryo-EM and X-ray crystal structure from this study as well as between X-ray crystal structure from this study and the published FAS structure (PDB 2UV8) was less than or equal to 1 Å. The AT domains residues 5-150 along with MPT domain residues 1737-1750 and 1850-1970 contributed towards 20-30% of the total deviations among the structures. RMSD values calculated excluding the aforementioned residues are indicated within brackets.

Stabilization of the ACP in the non-rotated conformation using “cycled” FAS

Unlike in the endogenous FAS holoenzyme complex, the non-rotated conformation of FAS without the γ-subunit had the ACP domain stalled at the KS domain (Figure 18). This is attributed to the step where FAS was “cycled” using malonyl-CoA and NADPH to drive all the FAS molecules to the same functional state. The stabilization of the ACP domain in the non-rotated state also improved the resolution of this FAS conformation by 0.3 Å for the cryo-EM map. The 2.9 Å crystal structure obtained from these FAS preparations was also 0.2 Å better than the published FAS structures (Jenni et al., 2007; Leibundgut et al., 2007).

Comparison between rotated FAS conformation with and without the γ-subunit

In the absence of the γ-subunit, FAS was found to adopt both the rotated (~8%) and the non-rotated (~92%) conformation (Figure 18). The ratio of the two states were different from the endogenous FAS holoenzyme complex, which had 60% particles in the rotated and 40%

in the non-rotated state (Figure17). Furthermore, the additional densities next to the ER and KR domain were absent in the FAS when the γ-subunit was absent (Figure 22). The correlation between the presence of the coiled coil density in the structure and the presence of the subunit in the FAS preparation indicates that these densities corresponds to the γ-subunit and not to any segment of the FAS itself. The non-rotated conformations of the FAS, both, with and without the γ-subunit were devoid of any additional densities that might correspond to the γ-subunit. Therefore, it appears as though the γ-subunit is bound to the FAS in a stable manner only in the rotated conformation.

Figure 22. Comparison of rotated FAS conformation of the FAS, FAS-holoenzyme and Gipson et al. (2010) cryo-EM structures. The density maps corresponding to the asymmetric unit of the FAS from the cryo-EM structure of FAS in the absence of the γ-subunit (left, this study), FAS-holoenzyme (middle, this study) and the 5.9 Å FAS structure from Gipson et al. (2010) (right). In the presence of the γ-subunit, the rotated conformation of FAS contains extra densities next to the ER and KR domains (pink). Similar densities were also observed by Gipson et al. (2010).