Screening for the effect of a potent new anti-HIV compound on HIV infected cells using oligonucleotide arrays to measure gene expression. Sanjive Qazi, Parker-Hughes Cancer Center, Roseville MN.

Introduction  Advances in microarray technology will enable the measurement of thousands of genes simultaneously.  We can now ask questions that relate to how networks of genes act together such as: What are the functions of this gene (functional annotation)? Which genes regulate this gene? Which genes are responsible for this disease? Which drugs will treat this disease?  In trying to answer these questions two types of modeling frameworks are being developed.  Genetic information flow. Genetic code to structure to function.  Complex dynamic systems. Function as manifested in the dynamic behavior of biochemical networks.  I will present some data from studies that are beginning to reveal in astounding molecular detail the response of a host cell to viral invasion, and how this will help us to develop therapies to either stop the infection or help the host cell defeat the invader.

Step 1: Binding A virus consists of an outer envelope of protein, fat and sugar wrapped around a set of genes (in the case of HIV, genetic information is carried as RNA instead of DNA) and special enzymes. HIV has proteins on its envelope that are strongly attracted to the CD4+ surface receptor on the outside of the T4-cell. When HIV binds to a CD4+ surface receptor, it activates other proteins on the cell's surface, allowing the HIV envelope to fuse to the outside of the cell. The HIV Life Cycle

Step 2: Reverse Transcription HIV's genes are carried in two strands of RNA, while the genetic material of human cells is found in DNA. In order for the virus to infect the cell, a process called "reverse transcription" makes a DNA copy of the virus's RNA. After the binding process, the viral capsid (the inside of the virus which contains the RNA and important enzymes) is released into the host cell. A viral enzyme called reverse transcriptase makes a DNA copy of the RNA. This new DNA is called "proviral DNA." Reverse transcription can be blocked by: Nucleoside Reverse Transcriptase Inhibitors (NRTIs), Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTIs), and Nucleotide Reverse Transcriptase Inhibitors.

Step 3: Integration The HIV DNA is then carried to the cell's nucleus (center), where the cell's DNA is kept. Then, another viral enzyme called integrase hides the proviral DNA into the cell's DNA. Then, when the cell tries to make new proteins, it can accidentally make new HIVs. Integration can be blocked by integrase inhibitors, a new class of drugs that are in the earliest stage of research.

Step 5: Translation The mRNA carries instructions for making new viral proteins from the nucleus to a kind of workshop in the cell. Each section of the mRNA corresponds to a protein building block for making a part of HIV. As each mRNA strand is processed, a corresponding string of proteins is made. This process continues until the mRNA strand has been transformed or "translated" into new viral proteins needed to make a new virus.

Step 6: Viral Assembly Finally, a new virus is assembled. Long strings of proteins are cut up by a viral enzyme called protease into smaller proteins. These proteins serve a variety of functions; some become structural elements of new HIV, while others become enzymes, such as reverse transcriptase. Once the new viral particles are assembled, they bud off the host cell, and create a new virus. This virus is then able to infect new cells. Each infected cell can produce a lot of new viruses. Viral assembly can be blocked by Protease Inhibitors (PIs).

DNA Transcription Transcription factors mRNA levels (Modified from Brazma (2000)

The Affymetrix Oligonucleotide Chip. Segment of a piece of DNA Oligonucleotides (25mer) 12,625 genes per chip. 20 oligonucleotides per gene. 64 pixels per probe cell (3  m/pixel) The chip contains numerous spiked controls, housekeeping genes, background cells, noise determination and normalization methods.

Data from a probe set on the chip.

Comparing data from two experiments.

NO DRUG 1nM Drug 1  M Drug Statistical filters used: The genes present (Presence Call in Affymetrix) in drug treated, ANOVA p<0.02 between groups. Red indicates increased expression, and green is decreased expression (Log(fold change)). Genesight 3 (Biodiscovery Software, Clustering to extract genes which tightly co-express.

Statistical filters used: The genes present (Presence Call in Affymetrix) in absence of drug, ANOVA p<0.02 between groups. NO DRUG 1nM Drug 1  M Drug

No drug1 nM Drug 1  M Drug Genes that increased levels of expression are pooled with those that showed decreases, and plotted in three dimensions representing levels of expression in the three viral strains. Red: INCREASES Purple: DECREASES

(Weng et al., 1999, Science, pp 92) Control of transcription.

Modeling complex network dynamics. Differential equations: (Modified from Brazma (2000)

RN + G R + N RG + N R + G RGN  GTP  S kcat The agonist (N) or the inactive g- protein (G) can independently bind to the receptor with kinetic constants shown. Binding of the agonist increases the affinity of the receptor for the binding of the g-protein, and vice versa. On formation of the g- protein / receptor / agonist complex (RGN) the g-protein dissociates with rate kcat to form an active  -subunit which is bound to a non- hydrolysable  -GTP  S (  GTP  S). Receptor-G protein Interactions.

kcat = 1 sec -1 [receptor] = 1 nM[receptor] = 100 nM The amplification and dynamics of  -GTP  S appearance depends on receptor concentration and G-protein dissociation rate (kcat)

[receptor] = 1 nM[receptor] = 100 nM kcat = 5 sec -1

Boolean Networks: (Modified from Brazma (2000)

Hybrid models: (Modified from Brazma (2000)

Summary  Gene expression clustering to extract regulatory motifs, inference of functional annotation and using the pattern as a molecular signature. Hierarchical clustering. K-means. Self-Organized Maps. Autoclass. Correlation metric construction. Mutual Information.  Functional annotation will require good experimental and statistical design.  Need to consider integrated behavior of regulatory network. Differential equations – need to solve for hundreds of parameters even in relatively simple networks. Boolean – Gene expression levels tend to be continuous. Hybrid models – Still will require a large amount of computation.

mRNA levels – Microarray Protein levels – SDS gel Metabolite concentrations - Spectrophotometry Build models Simulation software Network dynamics Optimization software Reverse Engineering

Acknowledgements. Dr. Fatih Uckun – President/Director of the Parker-Hughes Institute. Virology team. Sharon Pendergrass. Danielle Maher. Tricia Gill. Nicole Burkhardt. Jizhong Jin.