The Bacterial Flagellar Hook: A Molecular Universal Joint
for the Atlas of Macromolecules in Protein Explorer
Copyright © by Eric Martz, December 2004. Revised July 2006.
Permission is given to use this resource, or portions thereof, in websites or presentations provided the source, proteinexplorer.org, is cited. If the flagellar hook atomic coordinate data files are used, please also cite Samatey et al..
To re-start the animations, reload/refresh this page.

Bacterial flagella (definition) are whip-like organelles that bacteria use to swim about. Flagella are rotated by a molecular motor (see schematic diagram of motor, hook, and proximal flagellum) that penetrates the cell wall and membrane. The relatively stiff flagellar filaments are connected to the motor by a flexible segment called the "flagellar hook". This hook needs to transmit rotational torque (definition) from the motor while bending to allow the angle between the flagellar filament and the motor axis to change. The rotation of the bent hook has been likened to that of a smoke ring. Mechanically, the flagellar hook is analogous to a universal joint (definition and images).

The Protonic Nanomachine Project directed by Prof. Keiichi Namba at the Japan Science and Technology Agency provides gorgeous and highly instructive movies. The third one down on Page One zooms from the whole bacterial cell down to the spinning hook and motor. On Page Five is an animated simulation of the assembly of the entire motor, hook and flagellum from protein monomers. Further explanations and diagrams are available at Namba's Laboratories for Nanobiology at Osaka University, Japan.

The flagellar hook is made of about 120 copies of a protein molecule called "FlgE". In October, 2004, Samatey et al. in the groups of David DeRosier (Rosenstiel Basic Medical Sciences Research Center and Dept. Biology, Brandeis University, Waltham MA USA) and Keiichi Namba (Osaka University, Japan) reported the 3D structure of FlgE31, the mid-portionNote 1 of wild type FlgE from Salmonella typhimurium .

Using their 3D structure for FlgE31, Samatey et al. constructed a model of the entire flagellar hook as shown above. They then simulated its rotation (illustrated below) in order better to understand the ability of the cylindrical structure to bend yet resist twisting, thereby serving as a torque-transmitting universal joint. Note that while the resulting models have a hollow core, Samatey et al. believe that the core is normally filled with the portion of FlgE that had to be removed in FlgE31 to achieve crystallization. (One molecule of FlgE31 has a mass of 31 kilo-Daltons, so the entire hook model has a mass of about 4 mega-Daltons.)

Thanks to Prof. Keiichi Namba of Osaka University, Japan, who kindly provided the atomic coordinates used for the animations below, and gave permission to redistribute them for educational purposes.

Troubleshooting: If the links below don't display something that looks like the small movies below, please try starting Protein Explorer -- during start up, it automatically tests your browser and advises you how to configure it to display the animations below.
To re-start the animations, reload/refresh this page.

Animated multi-GIF's (methods) can be pasted into PowerPoint. Permission is given to copy these multi-GIF's into presentations or websites provided the source, proteinexplorer.org, is cited.
Animated multi-GIF's (methods) can be pasted into PowerPoint. Permission is given to copy these multi-GIF's into presentations or websites provided the source, proteinexplorer.org, is cited.
Protein Contacts within the Flagellar Hook: ( July 2006)
These views of the hook contacts are MolSlides and include analysis of conservation with ConSurf. Surprisingly, at least in this fragment that contains only 74% of full-length FlgE, none of the contacts within the hook are conserved.
  • Adjustment of monomer to fit hook model (a morph in Chime). Samatey et al. adjusted the relative positions of the two domains in 1WLG.pdb in order to achieve optimal fitting of the 121-chain assembly into the cryo-EM density map of the straight hook. In this morph, the domains closer to the center of the hook (the D1 domains) were aligned prior to the morph, making the adjustment appear as movement of the domain that makes up the hook surface (the D2 domain). Use your mouse to rotate the morph to view it from several perspectives in order to see all components of the change. Morph methods.

    No bulk change in conformation was involved in bending the straight hook model.

A morph is planned here that will show the sliding between adjacent chains in the 11-start helix that occurs during hook rotation.
Flagellar Hook in Protein Explorer: (Render and color as you wish)
The command scripts linked above can be used as starting points for coloring and rendering the rotating animations, items 2-4, below.
    Static Models
  1. FlgE31 monomer, 1WLG_1.mmol extracted from the crystal dimer, 1WLG, by Probable Quaternary Structures.
  2. Hook-conforming FlgE31 monomer (straight.pdb) with conformation adjusted for optimal fit into cryo-EM-based model of the flagellar hook: first chain from file hook.pdb.
  3. Complete flagellar hook, straight model:
  4. Complete flagellar hook, bent model, alpha carbon atoms only (rot000.pdb).

    Rotations

  5. Complete flagellar hook, simplified, rotating. 121 copies of FlgE31 represented as 3 "atoms" each, times 60 frames representing one full rotation in 6 degree increments. Coordinates shrunk ten-fold, so interatomic distances are incorrect.
  6. Complete flagellar hook, simplified, rotating (showing 6-start helices). 121 copies of FlgE31 represented as 5 "atoms" each, times 60 frames representing one full rotation in 6 degree increments. Coordinates shrunk 5.4-fold, so interatomic distances are incorrect.
  7. Portion of flagellar hook rotating, all alpha carbon atoms. Two rings, 11 copies of FlgE31 in each ring, alpha carbon atoms only, times 45 frames representing a half-turn in 4 degree increments.
  8. Two copies of FlgE31 rotating, all alpha carbon atoms. Two copies of FlgE31 alpha carbon atoms only, rotating about the (invisible) flagellar hook axis, times 45 frames representing a half-turn in 4 degree increments.

    Morphs between Conformations

  9. Single chain: Crystal (1WLG) to straight hook assembly
      Samatey et al. adjusted the relative positions of the two domains in 1WLG.pdb in order to achieve optimal fitting of the 121-chain assembly into the cryo-EM density map of the straight hook. In this morph, the domains closer to the center of the hook (D1 domains) were aligned prior to the morph, making the adjustment appear as movement of the surface domain (D2 domain). Use your mouse to rotate the morph to view it from several perspectives in order to see all components of the change.

        Methods: The starting and ending models for this morph were, from static models above, 1wlg_1.mmol to hook.pdb, using the first chain out of 121 chains in the latter file. Residues comprising domain D1 (71-142 plus 287-363) were aligned using DeepView with RMS 0.82 Å. Only the alpha carbon atoms from the aligned PDB files were used in the morph. The DOS program morph2.exe (available from PDBTools) was used to make a linear interpolation of 12 intermediate frames between the starting and ending models.

  10. Single chain: Straight hook assembly to curved (bent) hook assembly (a "toggle" between two models, not a morph).
      As can be seen in this 2-model comparison, no significant bulk movements were made in the monomer chain in order to bend the straight flagellar hook model. Note that the slight differences seen between the two models here may be an artefact of using MaxSprout (see methods below).

        Methods: The starting and ending models for this morph were, from static models above, hook.pdb to rot000.pdb, using the first chain out of 121 chains in each file. Since DeepView will not align models that contain only alpha carbons, the chain from rot000.pdb was processed by MaxSprout, which trimmed off the carboxy-terminal residues 362-3. Residues comprising domain D1 (71-142 plus 287-361) were then aligned using DeepView with RMS 0.28 Å. Only the alpha carbon atoms from the aligned PDB files are displayed in this two-model "toggle" file. N.B.: The output of MaxSprout modifies alpha carbon positions slightly (Liisa Holm, personal communication), so the small differences seen may be an artefact of utilizing MaxSprout.

  11. Two chains: Sliding between two chains during rotation of the hook.
      Data for this comparison were not available to me in a form I know how to align. If someone wants to align two alpha-carbon-only models for me, please contact me.


Notes.

1. FlgE is a protein of 402 amino acids. Solving its structure by X-ray crystallography required removing 26% of its length from the ends. This was required in order to coax it to crystallize, instead of forming filaments. The crystalline FlgE31 included amino acids 71-369, of which 71-363 were resolved.

2. One can trace various helices around the surface of the flagellar hook. Different helices connect monomers in paths at different angles from the common helical axis. There are 11 columns of molecules, each assigned one of 11 distinct colors, in the above animations. Each color forms a helix around the surface of the hook, although less than half a turn is made within the length of the hook. Since 11 of these helices start at the base of the cylinder (or at any plane formed by transecting the cylinder perpendicular to the helical axis), these are called 11-start helices. The FlgE domains that form the surface of the hook (the D2 domains) form close contacts in helices that have 6-starts. These 6-start helices make almost two complete turns within the length of the flagellar hook. Illustration of 11-start and 6-start helices. (Thanks to Dennis R. Thomas for help here.)


Methods for Animations of Rotations.

Prof. Keiichi Namba of Osaka University, Japan, kindly provided me with his models of the flagellar hook and gave permission to share them publically on this website. He included 360 curved models of the flagellar hook, each representing a one degree rotation. These models contained only alpha carbon atoms, and together represented about one gigabyte of data.

Each model is a curved cylinder whose circular perimeter is made up of 11 copies of the FlgE31 protein. Eleven layers of these 11-chain circles were stacked to make the cylinder, which thus contained 121 copies of FlgE31.

Reduction of Hook Rotation Dataset Size. In order to convey the gross structural features of hook rotation, yet with a small enough amount of data feasible to play as an animation in MDL Chime, I simplified Namba's models by representing each protein chain with only three atoms. These were chosen to represent the positions of the two protein domains and the elbow connecting them. Further, instead of using 360 models, spaced one degree apart, for a full rotation, I used only 60, spaced six degrees of rotation apart. These simplifications reduced the size of the dataset 586-fold, from 12,763,080 atoms (293 alpha carbons/chain x 121 chains x 360 rotation positions) to 21,780 atoms (3 atoms/chain x 121 chains x 60 rotation positions).

Preparing Data for Animation in Protein Explorer.

Protein Explorer (proteinexplorer.org) includes instructions on viewing and saving animations and making animated multi-GIF files.


Feedback to Eric Martz.