[an error occurred while processing this directive]
Ribosome structure:

The path from preparation to structure determination - for ribosomes only

Protein crystallography:
The atomic structure of a protein - actually the structure of any organic or inorganic molecule - can be analyzed with intensive xrays, at a resolution which can hardly be achieved with any other method.

THE prerequisite is however the succesful crystallization of the molecule/protein to be studied. This means that the individual proteins are arranged into a repeating 3 dimensional lattice. This crystalline lattice diffracts light (or xrays) in a characteristic way. The resulting diffraction pattern allows in principle to reconstruct the atomic composition of the crystal, i.e. to determine the structure of the crystallized protein. In reality its not that simple. However, the crystallization is difficult enough and frequently a matter of chance. In many cases, crystallization never succeeds even after years and years of trials, with fatal consequences for the poor phd-student ...

Below, we will just give a glimpse on the problems encountered during structure determination of ribosomal particles. It's not a text-book, not even a tutorial, and not always extremly accurate, but hopefully readable ...

1. Production of ribosomes

How to obtain crystals of a ribosomal particle ?? The best is of course to get the xray data from someone who doesn't know what to do with it. Never works. Another option would be to get the crystals or at least the ribosomes from someone, who doesn't know what to do with it. Also never works. Which just leaves the tedious way. The first step is the production of sufficiently large quantities of ribosomes. Bacteria are fairly easy to grow, so thats the favorite target, and not surprisingly, all ribosomal structures were obtained from bacterial ribosomes ....
A: Grow the bugs
B: Separate the ribosomal particles.
Growth of the bacteria is fast and efficient, just requires the proper conditions like media and temperature.
To separathe the ribosomes from all the other cell constituents, the cells are destroyed (eg with high pressure), and placed into a large centrifuge. Fast rotations in the centrifuge separates the different parts of the cell according to the mass or sedimentation constant. The ribosomal particles of a bacterial ribosome are therefore named 30S, 50S or 70S, since they sediment with velocities of 30, 50 or 70 S (S=Svedberg constant).

(To obtain ribosomal particles, which even crystallize, the procedure is not that simple. Small changes in the protocol can lead to inhomogeneties of the samples. And this can just make the difference of crystallizable or not-crystallizable. Its not a mere coincidence, that only 4 groups so far obtained reasonably well diffracting crystals, after 20 years of trial and error.)

2. Crystallization

A common procedure to obtain protein or ribosomal crystals is - for obvious reasons - called hanging drop method.
A drop containing varying amounts of salts, alcohols and the protein to crystallized is placed onto a small plate (cover slip). The plate and hence the drop is positioned above a reservoir containing essentially the same ingredients, except for the protein of course, but much more concentrated. The whole system is sealed to create a closed system and stored in a room, which is maintained at a constant temperature and humidity, in the ideal case even at a constant air pressure. The temperature varies from experiment to experiment, but the most common temperatures are 4o or 18o celsius.
The difference of concentrations between drop and reservoir leads to vapor diffusion, which slowly changes the concentration in the drop (also in the reservoir, but the volume of the reservoir is usually much larger). This way, the concentration in the drop is hoped to lead at some point to conditions favorable for crystal formation. The time needed for crystallization varies from days over month to eternity ...
Sometimes, introdution of small seed crystals or impurities facilitates crystallization, but the seed crystals also have the tendency to grow in a layered or inhomogenous fashion, which can affect the quality of the diffraction of the crystals - the ultimate goal - quite a bit.

The crystals usually can't stay in the drop for a long time, so they are either immediately frozen and kept at very low temperatures, or they are kept in a solution keeping the crystals intact (as far as possible).

A few examples for crystals of ribosomal particles
50S: Haloarcula marismortui

thin splates
30S: Thermus thermophilus

long needles
50S: Deinococcus radiodurans

ellipsoidal or triangular
Unfortunately, there is no correlation between the beauty of a crystal and its diffraction properties: like in real life, the most beautiful ones are often the worst ones.
So improvement or fine tuning of the crystals is still dominated by trial and error and success can only be determined by measuring diffraction patterns, tedious ....

Sometimes, crystallization leads to bizar results. This crystal was grown under zero gravity on a spacelab mission ....
(Foto by Niels Volkmann, ca. 1994)


Measuring diffraction patterns, means to expose the crystals to extremly intensive, energetic x-rays, which leads to destruction of the internal order of the crystals. To reduce the radiation damage, protein crystals are usually kept at very low temperatire (see below). However, to cool the crystal down to 100K or below, the water inside the crystal needs to replaced - at least partially - by cryo protectants, to prevent ice formation, which would destruct the rather fragile protein crystals. The crystals hence need to be transferred to a solution containing one or more cryo protectants: alcohols, high molar salts or sugars ...

3. X-ray structure analysis

3a: Mounting crystals in the x-ray beam.

Protein crystals are - contrary to ordinary crystals composed of very small molecules (diamonds for example) - quite sensitive to mechanical stress. This is mainly a consequence of the loose packing of large and flexible molecules, surrounded by large amounts of solvent, which can easily occupy 50-70% of the volume in a crystal. Therefore, density is much lower and crystal contacts much weaker than in salt crystals.

To account for the fragility, protein crystals are usually mounted in the x-ray beam utilizing specially designed tools:
  1. Capillaries: thin glas tubes, were originally used to mount crystals at room temperature, and sometime even used to grow the crystals. As long as the crystals are stable enough to survive radiation at room temperature, capillaries have the advantage that the solution surrounding the crystal can be exchanged in situ, which allows to screen for heavy atom derivatives on a single crystal. For cryo crystallography, capillaries are less suited and - as far as I know - barely used.
  2. Loops: are easy to produce. It just requires a (thin) hair, which is bent and glued to a metal pin. However, many materials like hair and polymers diffract the x-ray beam in a specific way, producing undesired power diffraction, which can cause some trouble particularly for low resolution data collection which is not so uncommon for ribosomal crystals.
    The crystal is kept inside the loop mainly through surface tension of the surrounding solution. However, surface tension can also act on the crystal itself, so that the loop is not always, but in most cases, the best solution.
  3. Glas spatula, are made of one or two very thin glas plates, which are glued to a glas pin, which is mounted on a brass pin. This design is a sense unique and purely handcrafted. This can be rather time consuming - compared to filling an order form for capillaries or loops - but results in extremly stable, reusable and tailor made sample holders..
    The crystal is usually can be directly fished - like for the loops - into the spatula. Sounds easy, but with little experience, it can well take half an hour, before the crystal is well oriented between the two glass plates.
    There are two major advantages of spatulas compared to loops: very thin crystals tend are better protected against mechanical stress (like surface tension) and its possible to choose a preferred orientation of the crystal in the x-ray beam. In a loop the crystal will almost always be randomly oriented. This is quite advantageous in case of our crystals of the 30S subunit from Thermus thermophilus. These crystals are thin, but very long needles: properly oriented, it's possible to collect data from a number of places on the crystal (provided the beam is sufficiently fine), which allows to collect more data from a single crystal.
    Spatulas are also fairly rigid. If the cold stream cooling the sample is rather strong, but less continous, the sample holder tends to vibrate, which could be more problematic for the comparably fragile loops, but it's an open question if this could affect the diffraction substantially ....
    They only disadvantage of the spatulas is the rather anisotropic absorption of the glass, which could possibly affect very small (anomalous) differences.

    Vertical sandwich spatula, with a H50S crystal mounted between the glass plates.

    Flat spatula with an exposed 30S crystal mounted between the two parallel glass plates.

    Fotos: MPG-AGRS Archiv, Hamburg

A crystal can either be directly froozen at the beamline, just by plunging it into the cold nitrogen stream (100K or -173o C). Alternatively, the crystal can be frozen in liquid nitrogen, or even better liquid propan prior to datacollection. Propan has the advantage to solidify at liquid nitrogen temperature. Handling and (long time) storage of the crystals becomes tvery easy this way, though shipping crystals in propane (explosive) sometimes turns into a nightmare ...

a vessel including a stored crystal and solid propane.

The left image shows the local setup at beamline 19ID at Argonne National Lab in Chicago, one of the best beamlines worldwide.
Below: The control panel of the beamline. Hitting the right buttons, allows to start data collection. Software, detectors, goniometers and all other technical details usually differ from beamline to beamline, since most beamlines have at least some inhouse developments to improve the experimental setup (which usually does an excellent job).
Hitting the wrong buttons, or the right buttons at the wrong time, usually disables the experiment.
Sometimes it just dumps the whole synchrotron, disabling hundreds of experiments. An easy way to become very popular ... .
View over the beamlines around the synchrotron at ESRF in Grenoble.

3b: Radiation damage, cryo cooling

In the early days, ribosome (and also protein) crystals were measured at room temperature, the way it was usually done for small molecules. However, the crystals literally evaporated in the x-ray beam within seconds, even at so called first generation synchrotrons. Data collection was simply impossible.

It was a truly ingenious idea to transfer the cryo method from electron microscopy to protein crystallography. To keep the crystals at low temperature, a heating element (essentially just a 10 Ohm resistor) was placed into liquid nitrogen.

X-ray diffraction pattern of a ribosome crystal.
A 10 Watt electrical current through the resistor was sufficient to produce a strong, reliable source of gaseous nitrogen with a temperature of roughly -180o C. Keeping the crystal at such temperatures reduces radiation damage to a minimum.

This way it became possible to collect data for up to 6 days from a single crystal at DORIS (DESY, Hamburg/Ger) - without apparent radiation damage.

Slowly third generation synchrotrons became available, and older beamlines were constantly upgraded. This was a highly desired development, since the speed of data collection and more important the maximum resolution collectable improved substantially. But there is a price to pay: protein and particularly ribosome crystals started to feel the enhanced intensity of the x-ray beam: better resolution at limited lifetime.
The intense beam sometimes burns crystals visibly, as shown in the small image on the left. The darker parts were already exposed to x-rays, the lighter parts are just waiting for it ...
It's easy to demonstrate radiation damage: the upper panel shows a diffraction between 3.4-5Å, at the beginning of datacollection. The lower panel shows the same diffraction pattern after some repeated exposures. Despite cryo cooling, the diffraction deteriorates quickly.
So nowadays, we need again a couple of crystals to collect a complete dataset (not to mention redundant data), which is however still much faster than it used to be in ancient times - at considerably better resolution.

We also tried to collect data at much lower temperature (4K or -269oC), but these were rather expensive and complicated trials, which just proofed that helium doesn't improve decay significantly.

3c: phase problem

Our measurements record the diffracted intensity, but thats not the whole story. In essence, the diffraction produces a fourier transformed image of the electron density inside our crystal. We just need to apply the inverse fourier transformation to restore the electron density, interpret the density and - the structure is solved. Unfortunately, we only measure the intensites (or amplitudes) but not the corresponding phases. Half of the information - and actually the more important half - is lost. Thats frequently called the phase problem in crystallography. One therefore needs to find a way to restore the lost phase information. Fortunately, at least 3-4 fundamental methods to solve this problem have been developed over the last 50 years or so ...
  1. Molecular replacement is probably the easiest and fastest way to solve a structure. However, it requires that someone else solved a structure of a similar protein earlier. This homologeous model is used as a search model. The correct placement of the search model allows the calculation of initial phases, which are usually sufficient to solve the new structure. Provded the search model is similar enough and the known fragment (it doesn't need to be the whole protein) large enough ....
  2. Direct methods (or ab initio phasing) try to derive the phase information from phase relation ships and plausible constraints (we usually know at least the sequence of the protein). It work beautifully to deduce small structures like the heavy atom substructure, and even medium sized proteins have been solved this way. But ribosome structures are still much to large and the resolution much too low to explore the structure ab initio.
  3. Isomorphous replacement (and its close relative, anomalous dispersion) is usually the method of choice. This method requires the incorpation of a heavy atom into the crystal structure, which leads to small but significant changes in the diffracted intensities. The differences between native and derivative data allows to determine the positions of the heavy atoms, which allows to calculate heavy atom phases. But at least with more than one derivative or additional anomalous differences, the heavy atom phases are sufficient to determine the full phase information (extremly smart procedure).
  4. Anomalous difference based structure determination is in a sense very similar to isomorphous replacement. The intensity differences in this case result however from the diffraction of atoms measured close (in terms of energy) to an absorption edge, which breaks the mirror symmetry of the so called Friedel mates. Change of energies can also change the diffraction properties of the atoms, which allows to get extremly accurate estimates of the protein phase information. This method has a number of huge advantages: all necessary information can - in principle - be obtained from a single crystal. Non-isomorphism between native and derivative crystals is not an issue in this case. Frequently, naturally occuring atoms (eg. Se, S) can even be used, so that not even a non-native heavy atom needs to be incorporated.



Lacking a homologeous model of the 30S ribosomal subunit, we used a combination of isormorphous replacement and anomalous dispersion to reconstruct the phase information. To be sure that the signal (=isomorphous differences) were large enough to be detetcted we used extremly large and electron desnse heavy atom cluster like W18 und Ta6Br12. That turned out to be a rather fortunate choice, since the incorporation of W18-clusters improved the reolsution of the 30S crystals dramatically. Instead of 7-9Å, wie could collect data up to 3Å.



heavy atom cluster:
W18 and Ta6Br122-
Unfortunately, the native and W18 crystals now became completely non-isomorphous, the resulting differences between the two crystal forms were not usable. But the W18-cluster are apparently not strongly bound. By washing the crystals in a solution deploit of W18 compounds, some of the heavy atom sites became much less occupied without altering the diffraction or the crystal lattice. Now we got significant and usable differences between W18 and W18-washed crystals. The positions of the best occupied W18-sites could easily be derived from a so called difference pattersons, which allowed to calculate initial phases, which again could be used to detect all the missing, less well bound W18 clusters hidden in the crystal. Though it was generally believed that the large cluster will not provide good phase information at high resolution, it tunred out that the isomorphous and anomalous differences provided sufficiently accurate phases all the way down to 3Å. But it was a very tedious process to get that far .

Though the whole process might appear to be rather technical, it doesn't really require a deep inside into physics of scattering processes (though it won't hurt). Fortunately, for most of the task - from data processing to structure determination and even modelling - well written programs are freely available, which do most of the job (reading the manuals is uncommon, however still highly desirable) ....

Tja, ribosomes are still not really routine work - which is one of the really nice aspects of this work. The resolution is low, the diffraction lousy, the decay severe and so on ... One possible way out: wait until someone else did all the work, jump on the waggon as late as possible, and become the king of ribosome structure !
Science and politics are just the same game ...


Getting back to the fabulous W18 cluster. The compound binds exclusively to ribosomal proteins, preferably arginine or lysine rich regions. Many of the ribosomal proteins (eg S2, S11, S18) contain a compact core and quite flexible arms of low complexity. W18 binding utilizes these arms for binding and incuces a kind of preferred conformation. Might well be, that co-factors like initiation or elongation factors utilize the extended protein arms in a similar fashion....


Links: S18 (gelb=unseres, blau=Cambridge) mit einem gebundenen W18. Die 'Arme' binden anders und deutlich weniger flexibel (d.h. man kann längere Stücke erkennen.)


4. Electron density, modeling

Ok, after lenghty calculations, endless nights at the synchrotron, hektoliter of coffee, something like an interpretable electron density pops up, which somehow needs to be correlated with the chemical composition of the protein or ribosome, which is at medium resolution (3.5-4Å) a nightmare ... and success depends a lot on trial and error ....
Right: Just an example to show, that the proper choice of parameters during phase determination can make quite a difference (image from Dr. Marco Glühmann).
It's obvious, that the density shown in the lower left part of the image, provides the best electron density. It's more detailed and also less fragmented than the other two electron densities ....

Naturally, people don't like to do this tedious process manually, so there are a couple of excellent programs available, which allow automatic tracing of the electron density. Unfortunately, not for multi-meric complexes and only at sufficiently high resolution (2.5Å or better) ...

Ribosomal nucleo-protein-assemblies are giant complexes made of a several entities. To interpret the electron density of such a monster, one need to use at least some independent information:

  • Crosslinks can tell which proteins/nucleotides are close contact with other proteins/nucleotides. Usually these data are not really accurate on a topological scale, but very reliable.
  • Biochemical data provide information about the involvement of certain substructures of the ribosome in biological functions. The interplay of L10-L11-L7/L12 with elongation factors is just one example.
  • The structure of a number of proteins have been determined earlier. Though there are usually some differences between the structure in isolation and the structure inside the ribosome, similarity is usually high enough to explore the homology.
  • Certain proteins and rRNA crosstalk to components in the other ribosomal subunit (30S to 50S and the other way round). Tell you accurately, which part should be close to the subunit interface.
  • Predictions of the secondary or tertiary structure can be quite handy.
  • Biochemical data of antibiotic binding sites are also sometimes very informative. But can also be quite misleading, since some compounds posess several binding sites (eg tetracycline) even on both ribosomal subunits, which can lead to really strange interpretations ...

... and so on ...

Naturally, the quality of the atomic interpretation strongly depends on the resolution of the data. At 3Å or better its quite easy to distinguish RNA and protein and even to roughly assign sequences to a strech of density. Beginning with the larger double helical parts of the rRNA structure, it becomes fairly easy to correlate the partial structure with the 2D structure of the RNA. To model the single stranded parts connecting the individual helices becomes sometimes a bit tricky, and its easy to get out of registry.

Of course, usually more than one person is trying to model the density (trying compete with the competitors :-)), which allows to model RNA and proteins pretty much in parallel. Even in the very early stages working at fairly low resolution, mis-interpretations (heh, you can't place a RNA there !! It's part of my protein !!) hardly ever occured. It's a kind of weird 3-D puzzle ... finding the right, missing piece after weeks of seeking and searching is just fantastic ( - kind of minimal compensation for the weeks before). The rest is finetuning, refinement, finetuning and refinement ... and we still find bits and pieces which ask for improvements ...

I

Left: An example for a RNA-helix in an earlier stage of the structure determination. The calculated electron density is shown together with the model (the interpretation of the electron density). Clearly, many details are not yet visible.
Right: An example for a (flexible) protein loop, at 3Å. Many, but not all, side chains are visible and interpretable ...

5. Finally, a structure

The ultimate outcome is the structure, sometimes still containing small errors (depending on resolution and experience), but for sure: it's a nice one !

Just need to publish it faster than your competitors (if there are any).

Once the work is done, an amazing number of helpers and contributors appear out of the mysts ... the beauty of the structure doesn't last long, though there are some rumors that it's not always like this ...

And then ?

Once the structure is solved, there is still a whole world to be explored, unfortunately never enough time to travel through it, sigh. The function and mechanism needs to be deduces, the specific interactions to be analyzed, various ligand and co-factors to be attached and so on. But just on time, another project, another subunit, another organism shows up ... its a fascinating, endless journey. Oops, did I say endless ? Of course, there is a natural limit, retirement. The Max-Plnack-Society is still propagating the Harnack principle: if the head of a department retires, the department is shut down as well, with very rare exceptions and mostly independent of scientific prospects. Of course, a noble prize is a reasonable excuse, but that's a difficult target, as the MPG encountered during the last couple of years :-) ...
<-- back to top

[an error occurred while processing this directive]