The Proteome in Pictures

In the largest project of its kind, Swedish scientists are studying normal and cancerous tissues to discover not only the location and abundance of all human proteins, but also how these are affected by disease states.

Mar 14, 2005
Helen Dell

Courtesy of Peter Nilsson

Instead of testing each tissue sample on its own slide, the protein profiling is done using tissue microarrays generated by assembling large number of patient biopsies onto a single glass slide.

In the largest project of its kind, Swedish scientists are studying normal and cancerous tissues to discover not only the location and abundance of all human proteins, but also how these are affected by disease states. "We want to make a comprehensive protein atlas for every protein," says project leader Mathias Uhlen of the Royal Institute of Technology in Stockholm. "My vision is that we will have a complete atlas in 10 years."

The team is using fluorescent antibodies to localize each protein with high-resolution imaging. The pictures they take are sharp enough to see where the protein resides within each cell, and also to estimate its abundance. "You can use the computer as a virtual microscope to zoom in and out," says Uhlen.

Using data from the Human Genome Project, the researchers cloned predicted genes from chromosomes X, Y, 14, and 22 (more will follow). The project will help to confirm gene predictions and to define gene functions, they suggest.

From each cloned gene, the team produces polyhistidine-tagged recombinant protein and uses it to generate polyclonal antibodies. "In formalin-fixed tissue samples on a slide, the proteins are denatured, unfolded," notes Uhlen. "If you're using monoclonal antibodies, you're lucky if the one [antibody-binding] site is exposed."

The group is producing 100–150 new recombinant proteins each week, and five new antibodies a day. They are generating about 150 gigabytes of data every day, and have written 60,000 lines of code in the past year to create a database of the images and associated experimental data. The first release of the public database is planned for the Human Proteome Organization (HUPO) conference in August, and Uhlen envisions updated releases every six months.

The database is not the only resource from the HUPO-funded project. "A lot of groups worldwide will want to work with these antibodies," says Uhlen. The team uses less than one percent of each antibody for its experiments. "There is plenty left for other people to use for other purposes," he says.

He and his collaborators have begun to set up the Swedish Human Proteome Resource to make the antibodies available to the wider research community, although they are still discussing the finer points of how the antibodies will be disseminated. "This is potentially a tremendous resource," says Ruedi Aebersold, professor of systems biology at the Swiss Federal Institute of Technology, Zurich, who was not involved in the study. "The images are of course useful to annotate the protein and learn where it is, but for the community there will also be huge value from the availability of the reagents."

Aebersold has slight concerns, however, about antibody specificity. "One of the real challenges with the antibody approach is that although it is easy enough to make antibodies, it is very difficult to figure out if the antibody is specific ... or if there are any cross-reactivities," he says.

Uhlen says he is confident that the specialized purification and validation procedure his team uses will weed out nonspecific antibodies and limit false-positives. The team purifies each antibody using its particular protein antigen; they test each antibody preparation against an array of hundreds of different proteins to check for cross-reactivity. Additionally, the researchers plan to produce two different antibodies to two different parts of the protein; the binding pattern of the two antibodies will validate each other, Uhlen says.

Each antibody is used to produce more than 700 pictures from different tissue samples, amounting to 10,000 images each week. The fully automated procedure probes tissue microarray samples from 48 normal tissues, including heart and brain, and from the 20 cancers that occur most frequently in the Western world, including breast, prostate, and liver. Uhlen and his colleagues hope that the comparison between normal and cancerous tissues will provide insights into the proteins that underlie the disease process.

Helmut Meyer, director of the Medical Proteome Center at the Ruhr University of Bochum, Germany, and one of the leaders of the HUPO-funded Human Brain Proteome Project, cautions that finding protein expression differences between normal and disease states does not necessarily mean those proteins are the cause of the disease. Anything the proteome projects highlight will need validation by inspection at a much closer level with more traditional techniques, such as knocking out the gene in an animal model.

Says Uhlen, "We are already seeing a significant number of potential diagnostic markers, but it's early days to say whether any of these are druggable targets."