by Jason Hale

Mona Norcum, Ph.D., is a Professor of Biochemistry at the University of Mississippi Medical Center, and the director of the UMMC Electron Microscope Facility. She has been a dedicated MCSR researcher for several years, and represents the UMMC on the Mississippi Supercomputer User Advisory Group (MSUAG). In a poster session at last month's Mississippi Academy of Sciences meeting in Biloxi, Dr. Norcum highlighted some of her current research efforts in macromolecular structural analysis. She agreed to let us spotlight her MCSR-related research efforts this month, and patiently entertained a number of unenlightened questions. This article attempts to capture the essentials of her research in layman's terms, so that even nonbiologists may appreciate it. I would refer the more scientifically-specialized reader to the meaty contents of Dr. Norcum's poster.

Our Questions:

What is the purpose of the research?
What is the big picture?
What is the research method?
What steps are required to transform 2-D datasets into 3-D images?
What would a picture of the process look like?
Can you explain what is meant by "cross-correlating the images of a dataset with the projections of a reference"?
What software tools are used?
What is the history of UMMC's electron microscope?
What is the current and future research focus?
What papers, presentations, or publications have been produced from MCSR-related research?

Answers

What is the purpose of the research?
To better understand the structure of proteins by creating high-resolution, three-dimensional images that can be manipulated with computer visualization tools. Her poster states: "For a variety of large biological macromolecules, determination of three-dimensional structures by the "standard" method of X-ray crystallography is not feasible. In these cases, computational microscopy is a useful alternative. "
Back to Questions

What is the big picture? (What applications are there for an improved understanding of protein structures?)
Understanding a protein's structure allows researchers to better determine "...how the protein (or macromolecule) actually works in its biological role." This is essential for designing new drugs that will be eeffective at interacting with the protein in some desired way.
Back to Questions

What is the research method?
An electron microscope is used to create several thousand 2-D images of a protein. The negatives are digitized into a gallery of digital 2-D images of a given structure, which is uploaded as a dataset to Dr. Norcum's SGI Octane Workstation, or to her account on the Origin 2000 supercomputer. Then, methods of single particle reconstruction and computer image analysis are applied to the dataset, ultimately creating and refining 3-D images, until the best possible resolution can be obtained.
Back to Questions

What steps are required to transform 2-D datasets into 3-D images?
There are several phases.

Selection
First, thousands of 2-D particle images are abstracted from the micrograph. Each image in this gallery represents a 3-D volume as projected onto a flat service from a different angle.

Alignment
Next the computer program attempts to align all the images. "That is, it attempts to place them in the center and rotate them to match each other as closely as possible."

Classification
"The program then classifies the aligned images by grouping those that are most similar or separating the images into classes depending on the choice of program used."

Averaging
Next the program creates a composite of each class by averaging the images in that class.

Orientation
"The angles that relate the individual classes are then determined by a computer search among all possible orientations for the 'best fit' as determined by the closest match in a cross-correlation calculation."

Three-dimensional Reconstruction
Using the angles determined in the orientation step, a three-dimensional reference structure is calculated using the image averages.

Refinement
"The reference structure is then projected into several hundred images, each from a different projection angle. Cross-correlation calculations are again used to match each of the several thousand original electron microscopic images with a particular projection. This again gives the proper orientation angles for each image, which is used in calculating a refined structure. This procedure can require millions of calculations and must be repeated until the highest resolution or 'best' structure is computed."
Back to Questions

What would a picture of the process look like?

Here's one from the poster session:

blah

And here's another:

Back to Questions

Can you explain what is meant by "cross-correlating the images of a dataset with the projections of a reference"?
"What it means is that the computer makes projections of the reference volume at about 194 angles. Think of it of as taking a three-dimensional object and shining a light onto from 194 different positions and seeing what the shadow looks like. However, in this case it is a projection of the three-dimensional density into a two-dimensional representation. Then, the cross-correlation calculation means that the computer takes each electron microscopic image and finds the projection that matches it best by maximizing the peak of the cross-correlation function.

"To go back a step or two, an EM image is a 2d projection of a 3d protein. It is a sum of all the image density through the structure because the EM beam goes through the particle. So, when we reconstruct a 3d volume and make projections, we are trying to mimic what the microscope does. This way, we can pick the projected image of our 3d volume that most closely matches a 'real' EM projection and determine the angles that are needed to relate it to the 3d volume. So the process is to map thousands of 'real' images with a set of a few hundred computer made projections of the 3d volume. Then the 3d structure can be recalculated to higher resolution because of the increase in number of input images.

"For example, a normal calculation would be to project a 3d reference volume at 194 or 798 angles (10 degree or 5 degree increments) and map a data set of 10,000 EM images to these. So in this example there would be 194 X 10000 = 1,940,000 or 798 X 10000 = 7,980,000 cross correlation calculations just in this step of picking the matching projection, which is actually picking the right angles for the next step of 3d reconstruction of the structure."
Back to Questions

What software tools are used?
SPIDER, which was developed using the IRIX, SGI's implementation of Unix, is the image analysis program that does all of the computations. For some of the simpler structures, Dr. Norcum can use SPIDER on her SGI Octane workstation, and she also does her visualization on this platform. However, for most structures, the processes detailed above require so much computation that the MCSR's SGI Origin 2000 supercomputer is the only way to go.
Back to Questions

What is the history of UMMC's electron microscope?
"The University Medical Center purchased the LEO 912AB transmission electron microscope in 1998. Being fully equipped for cryoelectron microscopy and fitted with an integrated energy filter, it is the only microscope of its kind in the state of Mississippi. The microscope, high resolution scanner, digital printer and computers necessary for this type of work cost approximately $535,000. Since the opening of the EM facility, the resultant research funding includes $314,000 for the multisynthetase project in this lab and $52,000 from collaborators at Penn State University School of Medicine and Medical College of Ohio. "
Back to Questions

What is the current and future research focus?
Current projects include:
- "Study of the protein distribution and active site arrangement within the mammalian multienzyme complex of aminoacyl-tRNA synthetases" (funded at UMMC by NSF)
- "Determination of structures and subunit arrangements within isoforms of mammalian meprin proteases (funded at Penn State University by NIH)."

These are described in some detail in Dr. Norcum's MAS Poster Presentation.
Back to Questions

What papers, presentations, or publications have been produced from MCSR-related research?
From Dr. Norcum's web page:

Recent Publications Norcum, M.T. and Boisset, N. 2002. Three-dimensional architecture of the eukaryotic multisynthetase complex determined from negatively stained and cryo electron micrographs. FEBS Lett. 512: 298-302.
Traxler, K.W., Norcum, M.T., Hainfeld, J.F., Carlson, G.M. 2001. Direct visualization of the calmodulin subunit of phosphorylase kinase via electron microscopy following subunit exchange. J. Struct. Biol. 135: 231-238.
Norcum, M.T. and Warrington, J.A. 2000. The cytokine portion of p43 occupies a central position within the eukaryotic multisynthetase complex. J Biol Chem Accelerated Publication 275:17921-17924.
Norcum, M.T. and Dignam, J.D. 1999. Immunoelectron microscopic localization of glutamyl-/prolyl-tRNA synthetase within the eukaryotic multisynthetase complex. J. Biol. Chem. 274: 12205-12208.
Norcum, M.T. 1999. Ultrastructure of the eukaryotic multisynthetase complex derived from two-dimensional averaging and classification of negatively stained electron micrographs. FEBS Lett. 447: 217-222.
Norcum, M.T. and Warrington, J.A. 1998. Structural analysis of the multienzyme amino-acyl tRNA synthetase complex: a three domain model based on reversable crosslinking. Prot. Sci. 7: 79-87.
Back to Questions