Mechanisms of Cdc42 Polarization in Yeast

Polarization is important for the function and morphology of many different cell types. The keys regulators of polarity in eukaryotes are the Rho-family GTPases. In the budding yeast Saccharomyces cerevisiae, which must polarize in order to bud and to mate, the master regulator is the highly conserved Rho GTPase, Cdc42. During polarity establishment, active Cdc42 accumulates at a site on the plasma membrane characterizing the “front” of the cell where the bud will emerge. The orientation of polarization is guided by upstream cues that dictate the site of Cdc42 clustering. However, in the absence of upstream cues, yeast can still polarize in a random direction during symmetry breaking. Symmetry breaking suggests cells possess an autocatalytic polarization mechanism that can amplify stochastic fluctuations of polarity proteins through a positive feedback mechanism.

Two different positive feedback mechanisms have been proposed to polarize Cdc42 in budding yeast. One model posits that Cdc42 activation must be localized to a site at the plasma membrane. Another model posits that Cdc42 delivery must be localized to a particular site at the plasma membrane. Although both mechanisms could work in parallel to polarize Cdc42, it is unclear which mechanism is critical to polarity establishment. We directly tested the predictions of the two positive feedback models using genetics and live microscopy. We found that localized Cdc42 activation is necessary for polarity establishment.

While this explains how active Cdc42 localizes to a particular site at the plasma membrane, it does not address how Cdc42 concentrates at that site. Several different mechanisms have been proposed to concentrate Cdc42. The GDI can extract Cdc42 from membranes and selective mobilize GDP-Cdc42 in the cytoplasm. It was proposed that selectively mobilizing GDP-Cdc42 in combination with local activation could locally concentrate total Cdc42 at the polarity site. Although the GDI is important for rapid Cdc42 accumulation at the polarity site, it is not essential to Cdc42 concentration. It was proposed that delivery of Cdc42 by actin-mediated vesicle can act as a backup pathway to concentrate Cdc42. However, we found no evidence for an actin-dependent concentrating pathway. Live microscopy experiments reveal that prenylated proteins are not restricted to membranes, and can enter the cytoplasm. We found that the GDI-independent concentrating pathway still requires Cdc42 to exchange between the plasma membrane and the cytoplasm, which is supported by computational modeling. In the absence of the GDI, we found that Cdc42 GAP became essential for polarization. We propose that the GAP limits GTP-Cdc42 leak into the cytoplasm, which would be prohibitive to Cdc42 polarization.

Defining the Role of the Histone Methyltransferase, PR-Set7, in Maintaining the Genome Integrity of Drosophila Melanogaster

The complete and faithful duplication of the genome is essential to ensure normal cell division and organismal development. Eukaryotic DNA replication is initiated at multiple sites termed origins of replication that are activated at different time through S phase. The replication timing program is regulated by the S-phase checkpoint, which signals and repairs replicative stress. Eukaryotic DNA is packaged with histones into chromatin, thus DNA-templated processes including replication are modulated by the local chromatin environment such as post-translational modifications (PTMs) of histones.

One such epigenetic mark, methylation of lysine 20 on histone H4 (H4K20), has been linked to chromatin compaction, transcription, DNA repair and DNA replication. H4K20 can be mono-, di- and tri-methylated. Monomethylation of H4K20 (H4K20me1) is mediated by the cell cycle-regulated histone methyltransferase PR-Set7 and subsequent di-/tri- methylation is catalyzed by Suv4-20. Prior studies have shown that PR-Set7 depletion in mammalian cells results in defective S phase progression and the accumulation of DNA damage, which may be partially attributed to defects in origin selection and activation. Meanwhile, overexpression of mammalian PR-Set7 recruits components of pre-Replication Complex (pre-RC) onto chromatin and licenses replication origins for re-replication. However, these studies were limited to only a handful of mammalian origins, and it remains unclear how PR-Set7 impacts the replication program on a genomic scale. Finally, the methylation substrates of PR-Set7 include both histone (H4K20) and non-histone targets, therefore it is necessary to directly test the role of H4K20 methylation in PR-Set7 regulated phenotypes.

I employed genetic, cytological, and genomic approaches to better understand the role of H4K20 methylation in regulating DNA replication and genome stability in Drosophila melanogaster cells. Depletion of Drosophila PR-Set7 by RNAi in cultured Kc167 cells led to an ATR-dependent cell cycle arrest with near 4N DNA content and the accumulation of DNA damage, indicating a defect in completing S phase. The cells were arrested at the second S phase following PR-Set7 downregulation, suggesting that it was an epigenetic effect that coupled to the dilution of histone modification over multiple cell cycles. To directly test the role of H4K20 methylation in regulating genome integrity, I collaborated with the Duronio Lab and observed spontaneous DNA damage on the imaginal wing discs of third instar mutant larvae that had an alanine substitution on H4K20 (H4K20A) thus unable to be methylated, confirming that H4K20 is a bona fide target of PR-Set7 in maintaining genome integrity.

One possible source of DNA damage due to loss of PR-Set7 is reduced origin activity. I used BrdU-seq to profile the genome-wide origin activation pattern. However, I found that deregulation of H4K20 methylation states by manipulating the H4K20 methyltransferases PR-Set7 and Suv4-20 had no impact on origin activation throughout the genome. I then mapped the genomic distribution of DNA damage upon PR-Set7 depletion. Surprisingly, ChIP-seq of the DNA damage marker γ-H2A.v located the DNA damage to late replicating euchromatic regions of the Drosophila genome, and the strength of γ-H2A.v signal was uniformly distributed and spanned the entire late replication domain, implying stochastic replication fork collapse within late replicating regions. Together these data suggest that PR-Set7-mediated monomethylation of H4K20 is critical for maintaining the genomic integrity of late replicating domains, presumably via stabilization of late replicating forks.

In addition to investigating the function of H4K20me, I also used immunofluorescence to characterize the cell cycle regulated chromatin loading of Mcm2-7 complex, the DNA helicase that licenses replication origins, using H4K20me1 level as a proxy for cell cycle stages. In parallel with chromatin spindown data by Powell et al. (Powell et al. 2015), we showed a continuous loading of Mcm2-7 during G1 and a progressive removal from chromatin through S phase.

The Role of Dopaminergic Systems in the Neurobehavioral Teratology of Organophosphates in Zebrafish

Background: Organophosphate (OP) pesticides are well-known developmental neurotoxicants that have been linked to abnormal cognitive and behavioral endpoints through both epidemiological studies and animal models of behavioral teratology, and are implicated in the dysfunction of multiple neurotransmitters, including dopamine. Chemical similarities between OP pesticides and organophosphate flame retardants (OPFRs), a class of compounds growing in use and environmental relevance, have produced concern regarding whether developmental exposures to OPFRs and OP pesticides may share behavioral outcomes, impacts on dopaminergic systems, or both. Methods: Using the zebrafish animal model, we exposed developing fish to two OPFRs, TDCIPP and TPHP, as well as the OP pesticide chlorpyrifos, during the first 5 days following fertilization. From there, the exposed fish were assayed for behavioral abnormalities and effects on monoamine neurochemistry as both larvae and adults. An experiment conducted in parallel examined how antagonism of the dopamine system during an identical window of development could alter later life behavior in the same assays. Finally, we investigated the interaction between developmental exposure to an OPFR and acute dopamine antagonism in larval behavior. Results: Developmental exposure to all three OP compounds altered zebrafish behavior, with effects persisting into adulthood. Additionally, exposure to an OPFR decreased the behavioral response to acute D2 receptor antagonism in larvae. However, the pattern of behavioral effects diverged substantially from those seen following developmental dopamine antagonism, and the investigations into dopamine neurochemistry were too variable to be conclusive. Thus, although the results support the hypothesis that OPFRs, as with OP pesticides such as chlorpyrifos, may present a risk to normal behavioral development, we were unable to directly link these effects to any dopaminergic dysfunction.

Distributed Feature Selection in Large n and Large p Regression Problems

Fitting statistical models is computationally challenging when the sample size or the dimension of the dataset is huge. An attractive approach for down-scaling the problem size is to first partition the dataset into subsets and then fit using distributed algorithms. The dataset can be partitioned either horizontally (in the sample space) or vertically (in the feature space), and the challenge arise in defining an algorithm with low communication, theoretical guarantees and excellent practical performance in general settings. For sample space partitioning, I propose a MEdian Selection Subset AGgregation Estimator ({\em message}) algorithm for solving these issues. The algorithm applies feature selection in parallel for each subset using regularized regression or Bayesian variable selection method, calculates the `median' feature inclusion index, estimates coefficients for the selected features in parallel for each subset, and then averages these estimates. The algorithm is simple, involves very minimal communication, scales efficiently in sample size, and has theoretical guarantees. I provide extensive experiments to show excellent performance in feature selection, estimation, prediction, and computation time relative to usual competitors.

While sample space partitioning is useful in handling datasets with large sample size, feature space partitioning is more effective when the data dimension is high. Existing methods for partitioning features, however, are either vulnerable to high correlations or inefficient in reducing the model dimension. In the thesis, I propose a new embarrassingly parallel framework named {\em DECO} for distributed variable selection and parameter estimation. In {\em DECO}, variables are first partitioned and allocated to m distributed workers. The decorrelated subset data within each worker are then fitted via any algorithm designed for high-dimensional problems. We show that by incorporating the decorrelation step, DECO can achieve consistent variable selection and parameter estimation on each subset with (almost) no assumptions. In addition, the convergence rate is nearly minimax optimal for both sparse and weakly sparse models and does NOT depend on the partition number m. Extensive numerical experiments are provided to illustrate the performance of the new framework.

For datasets with both large sample sizes and high dimensionality, I propose a new "divided-and-conquer" framework {\em DEME} (DECO-message) by leveraging both the {\em DECO} and the {\em message} algorithm. The new framework first partitions the dataset in the sample space into row cubes using {\em message} and then partition the feature space of the cubes using {\em DECO}. This procedure is equivalent to partitioning the original data matrix into multiple small blocks, each with a feasible size that can be stored and fitted in a computer in parallel. The results are then synthezied via the {\em DECO} and {\em message} algorithm in a reverse order to produce the final output. The whole framework is extremely scalable.