|
|
|
|
The use of Microbatch for Large Scale Crystallization Projects By Douglas Instruments, Hungerford, UK Overview Automation will be required to crystallize a significant proportion of the proteins discovered by the various genome projects. The microbatch crystallization method is very fast, and it can therefore perform screening more thoroughly than the vapor diffusion method. The same equipment can perform microbatch and a variation of vapor diffusion where many samples are equilibrated against a single reservoir solution. Off-the-shelf incubators could be equipped with simple mechanisms for automatic viewing of crystals. Microbatch crystallization would use far fewer personnel than vapor diffusion, and it would cost around 1/3 the cost of alternative systems. Douglas Instruments is seeking a partner to exploit its technology
The need for automated crystallization As the gene sequences of proteins flow from the human genome project (and other genome projects) in increasing numbers, so the need for crystallization of these proteins increases. Whether the functions of these proteins are known or not, determining their structures will make a cost-effective contribution to their characterization. The only practical way of crystallizing a significant proportion of around 100,000 proteins will be to automate the process.
Problems in previous attempts to automate protein crystallization Since the late 1980s there have been at least six major attempts to automate protein crystallization by major US and European companies and research institutes. In a report carried out by Douglas Instruments for the European Commission in 1995, it was found that whilst these robotic systems worked, they were all unsuccessful in that they are not in use today (with the exception of Douglas Instruments’ own system, IMPAX, described below). Around a million dollars was spent on each. The major problems with these systems are or were:
In late 1997 a new automatic crystallization system was introduced by Cyberlab. Although it has some of the above drawbacks, this system seems to perform crystallization by vapor diffusion well. It is however comparatively slow and expensive.
Crystallization methods Three methods can be used to carry out crystallization experiments in large numbers:
Microbatch crystallization [1] involves dispensing droplets of protein with precipitants, buffers and additives in miniature wells under oil. The method was invented by Imperial College, London, in collaboration with Douglas Instruments. It is easily automated, and since 1990, over 80 automatic microbatch systems have been installed worldwide. Currently, half a microliters of protein is used per trial for screening, and 200 trials can be dispensed in an hour. Slow concentration of the sample is achieved by using an oil containing silicone oil [2] which allows gradual evaporation of water. For more details about microbatch see http://www.douglas.co.uk Although we believe that microbatch is very useful for all stages of crystallization, including screening, optimization and production of crystals for data collection, we initially want to emphasize its use for screening (finding initial crystallization conditions). For example, microbatch could be used for screening, followed by vapor diffusion to optimize crystals (note however that microbatch is widely used for optimization and often succeeds where vapor diffusion fails). The equipment that dispenses microbatch experiments can also be used for vapor diffusion experiments, where many samples are equilibrated against a salt or PEG solution in a single reservoir. This can be achieved in standard microbatch plates (HLA plates) - please contact Douglas Instruments for more information. This method has the advantage that if no crystals or precipitates appear the concentration of the reservoir solution can be increased.
Advantages of microbatch for large scale automation More conditions can be screened. Douglas Instruments recently showed [3] that microbatch gave more crystallization leads than vapor diffusion for a given amount of protein and operator time. Microbatch found 43 conditions while vapor diffusion found 41. Vapor diffusion required more work and used more protein. Ease of automation. Microbatch crystallization is physically far simpler than vapor diffusion, which is reflected in the complexity of the automatic systems which carry out the methods. Microbatch systems have far fewer working parts and are very reliable. Cost. Microbatch crystallization systems are less expensive and use less consumables, reagents and protein. Moreover the crystallization plates are smaller and do not require glass coverslips, tape etc. Approximately 22 systems would be needed for this project, whereas 48 Cyberlab systems would be needed since they are slower (see discussion below). (A good solution might be to use both systems.)
Philosophy of automation To automate the whole of crystallization from screening to data collection is a huge task, and is unnecessary. The difficulty arises partly because artificial intelligence would be required to design experiments for optimization and to reliably recognize crystals. Douglas Instruments therefore proposes that rather than trying to automate the whole process, human operators should interact with automatic equipment which they would guide. For example, operators should use user-friendly software with templates for crystallization experiments which can be adapted to each protein. Such a system would still rely on human judgment. One of the most difficult aspects of automatic systems is to move samples from one automatic system to another. We suggest that instead, plates should be transferred to and emptied out of automatic viewing stations by hand. This operation would amount to less than 0.2% of the labor requirement of the project.
Integration with databases Because data can be collected from a standard set-up, it will be much easier to build a database of successful crystallization conditions. The data will also be standardized, rather than reflecting the biases of many individual scientists. It is hoped that analysis of the database will greatly speed up protein crystallization generally by increasing understanding of crystallization
Inspecting plates to see if crystals have grown Douglas Instruments proposes that plates should be scanned by a simple mechanism within
standard readily available incubators. Here the compact size of microbatch plates is a
great advantage. Each incubator will be equipped with around 10 carousels which each have
space for around 40 plates. One or more moveable digital cameras will be able to view the
plates. Each day the plates can be scanned automatically with a minimum of disturbance to
the plates and images of every well can be stored. If microbatch plates were stored for 6
weeks, approximately 28 incubators would be needed for the project (assuming around 5000
structures would be solved, see discussion below.)
An image recognition system could be used to search for crystals. However, it should be used simply to place wells in order of interest so that they can subsequently be viewed by an operator. This approach will allow a higher percentage of proteins to be crystallized. An alternative method will be to train many human operators to recognize crystals in images of wells which they can view in their homes using the internet or intranet.
A sensible target might be to solve the structures of around 5000 representative proteins from e.g. the human genome project. This would required the expression, purification and screening of 20,000 protein preparations, since satisfactory crystals can be obtained from only about one protein prep in four. (On average each protein might have to be modified once by going back to the molecular biology.) The most productive laboratories currently take around 60 man-days per protein to obtain crystals suitable for data collection. This overall figure includes trials with other proteins that cannot be crystallized satisfactorily. Assuming this rate could be speeded up somewhat for larger numbers of proteins, the approximate labor requirements in order to complete the project in 5 years for three different approaches are as follows:
The number of plates per person per day for the manual approach (10) seems low, but this includes the time taken to design experiments. The figure given is approximately 3 times the rate for the most productive laboratories at present. With automated systems, experiments can be designed while experiments that were designed previously are being executed, so no extra time need be allowed for design. However, the above figures for both automatic systems assume that most of the screening takes place automatically at night.
Conclusions Microbatch crystallization is far simpler than alternative methods, but gives greater productivity, especially for screening. We propose that it should be used to crystallize proteins arising out of the various genome projects. It could be used alone or in combination with vapor diffusion with sitting or hanging drop. Sitting drop can be achieved with the same equipment as microbatch by equilibrating many samples against a single reservoir. Using microbatch would reduce both personnel and capital expenditure by a factor of 2 to 3. Douglas Instruments is now seeking a partner to exploit the microbatch method for large-scale crystallization projects.
References [1] N.E. Chayen, P.D. Shaw Stewart, D. L. Maeder and D.M. Blow, J. Appl. Cryst. 23
(1990) 297. Patrick Shaw Stewart and Peter Baldock, August 1998 and February 1999
|
|
