Malaria's Pragmatic Approach to Gene Expression

UNEXPECTED CONCORDANCE

Elizabeth Winzeler at the Scripps Research Institute in San Diego led a group that created an expression profile from nine different stages throughout the parasite life cycle,² while Joseph DeRisi's group at the University of California, San Francisco (see profile, p. 40) focused on the asexual blood stage, taking hourly time points over two days.³ The two teams used different strains of the parasite, different microarray platforms, and ultimately, separate methods of analyzing the patterns they found. Regardless, the expression profiles "were amazingly similar," says Karine Le Roch at Scripps, the first author of the paper from Winzeler's lab.

Even more surprising, the work exposed an on-demand expression pattern in which proteins are made specifically when the parasite needs them. "Plasmodium shows a much, much greater cell-cycle dependence than any other organism known, period," says David Roos at the University of Pennsylvania.

Winzeler's group faced significant hurdles to array and analyze the organism, including the A-T richness of the genome and the difficulty in obtaining large amounts of RNA. When she first proposed the project to the National Institutes of Health, reviewers told her it would never work, she says. But according to Le Roch, once they had spent the bulk of their time obtaining "enough sample to put on the chip," the microarrays showed immediate results. "We were really surprised it worked as well as it did," Winzeler says.

Winzeler had previously done test expression profiles on P. falciparum's chromosome 2. This time, they used 25-mer oligonucleotides in a custom-made, high-density array to assess more than 95% of predicted P. falciparum genes. With two synchronization methods to evaluate true cell-cycle control, 88% of predicted genes were expressed at some stage, and all but 885 were expressed at the blood stage. Cluster analysis revealed that genes working on similar functions had similar expression profiles. The researchers were thus able to organize them into 15 hierarchical clusters, including DNA replication, division, and invasion. Their results hinted at the practicality of the parasite's expression pattern: Proteins appear to be expressed when needed.

PROTEINS ON DEMAND

While Winzeler looked at absolute transcription levels, DeRisi analyzed ratios, comparing expression levels to those found at time zero. Using this analysis, the DeRisi lab concluded that most genes are expressed in a cyclical, "just in time" manner.

DeRisi's group looked only at the asexual blood stage, which is responsible for malaria's pathogenicity. In addition to its therapeutic implications, it is the most tractable stage to study, says Manuel Llinas, coauthor on the DeRisi Hot Paper and now at Princeton University.

from PlasmoDB.org

Expression profiles from each of the Hot Papers shows remarkable concordance despite different strategies. These graphs show expression levels for merozoite surface protein 1 during Plasmodium falciparum's intraerythrocytic stage. Expression percentiles from the Winzeler lab (orange, red, and blue lines)correlate highly with those from the DeRisi lab (green and purple lines).

Working for almost two years to develop the 70-nt array and culturing methodology, they decided that "just a few time points would be woefully inadequate" to capture the parasite's complexity, says DeRisi. So, the team monitored the parasites hourly over the two-day blood stage. Llinas remembers being "blown away" when he lined up the data and could see periodicity right away.

DeRisi says he opted against cluster analysis when he noticed a continuous cascade of gene expression. Instead, the Fourier transform method, used in studying yeast cycles, helped them to "analyze the phasic nature of expression," They found that "transcripts are made essentially right before the protein is needed," he says. "This brings forward a very intriguing hypothesis, that Plasmodium may have a fairly rigid transcriptome, and that it may lack the ability to respond to wide ranges of environmental perturbations," he says.

DeRisi says this may be because intraerythrocytic malaria stays "in a very homogenous environment, where conditions are incredibly constant," unlike free-living eukaryotes such as yeast. With its dedicated life purpose – blood-cell invasion, lysis, and reinvasion – malaria may never need to respond dynamically to environmental surprises.

According to Roos, each team's platform had a significant impact on their results. He says DeRisi's glass-spotted array method is cheap and flexible, whereas Winzeler's photolithographic arrays provide greater resolution and can probe a greater range of genes. The strength of DeRisi's data lies in the number of time points, 50, to Winzeler's nine, says Roos.

The distinct capabilities of each team's approach provided complementary information, which Roos' team integrated and made publicly available on the malaria genome database, PlasmoDB. "The fact that their concordance is extremely high gives you great confidence in both of them," he says. Roos estimates that the expression profile component of PlasmoDB gets more than 5,000 hits daily.

DIGGING DEEPER

The authors' interpretations warrant closer scrutiny, however, says Rosenthal, particularly because they use models based on expression peaks. "The tendency of the community now is to see that as black and white," he says, when in reality, lower points of expression may be very meaningful.

Rosenthal says the papers' "real value," is that they provide a powerful tool to study genes of interest, such as potential drug and vaccine targets, and to predict functions of uncharacterized genes based on coexpression with known genes. "The next issue," says Nirbhay Kumar at the Johns Hopkins School of Public Health, "is to ... investigate their functional significance."

In a recent study, Stanley Fields at the University of Washington adapted the yeast two-hybrid assay to describe a network of nearly 3,000 P. falciparum protein-protein interactions.⁴ Most involve one or two previously uncharacterized proteins, and combining the findings with gene-expression clusters from the Winzeler Hot Paper, the team began to parse these proteins' functions, particularly the potential roles of some in erythrocytic invasion.

Earlier this year, a team including Winzeler characterized a protein that allows malaria to switch between alternate invasion pathways.⁵ "If you go back and look at expression data," says Winzeler, "its expression pattern is correlated with those of other proteins with a role in invasion." Because the parasites are exposed in the bloodstream during this stage, such proteins often make good vaccine target candidates.

Malaria researchers are also asking: What controls gene expression in the first place? The "big question," says Llinas, is "how is transcription actually being governed to drive the transcriptional cascade." According to Winzeler, the expression profiles allow researchers to look for motifs associated with coexpressed genes that are putative transcription-binding sites; few regulatory regions have been found thus far.⁶

In the future, Winzeler hopes to integrate current data in order to "create more robust models." Meanwhile, DeRisi says his team is "systematically probing the parasite's transcriptome with various stimuli" to gauge its range of response, though according to Rosenthal, "word of mouth" says that "you don't see much." Both Llinas and Le Roch cite knockouts of interesting genes as a time-consuming but important means of validating current data. After all, says Llinas, "genome-wide approaches are only as good as the specific molecular-biology follow-up."