2005 Summer Scholar Participants-HHMI Undergraduate Program - Carnegie Mellon University

2005 HHMI Summer Scholar Participants

Eva Goellner

Eva Goellner, Carnegie Mellon University
Mentor: Dr. Robert Murphy

Three Color Image Collection for Segmentation and Analysis of Protein Subcellular Location Patterns in Polarized Monolayers of CaCo2 Cells

Automated analysis of the location of proteins is useful in determining statistical differences in protein patterns that may be undetectable to the human eye. The goal of this project is to collect three-dimensional, three-channel fluorescence microscope images of the CaCo2 cell line. CaCo2 are human colon cancer epithelial cells, which will be used as a model system for extending current automated subcellular location pattern analysis methods to polarized epithelia and tissues. Unlike in prior experiments on single cultured cells, the polarized monolayers require a method for determining the boundaries of each cell before pattern analysis can be accomplished. Achieving an acceptable method of data collection to facilitate segmentation has been a main goal of this project. The protocol we have developed involves tagging a protein of interest such as a Golgi or mitochondrial protein using primary antibodies and Alexa-488 conjugated secondary antibodies. In order to perform automated segmentation, it is necessary to also have a nuclear image and a continuous plasma membrane image. DRAQ5 was used to stain the nuclei. A primary antibody against the cadherin family of proteins and a secondary antibody conjugated to Alexa-564 dye provided the continuous plasma membrane staining desired. This staining was homogeneous and easily recognized by automated segmentation algorithms. The resulting three-color images are appropriate for automated segmentation as well as subcellular location pattern analysis of the proteins of interest.

Aaron Goldman

Aaron Goldman, Carnegie Mellon University
Mentor: Dr. Javier López

Detection and Functional Analysis of Recursive Splicing

Splicing is the removal of introns from eukaryotic transcripts to produce functional mRNAs. Ten percent of genes in human and Drosophila , including many genes with important roles in development and disease, contain enormous introns that are longer than 10 kb. Little is known about the mechanisms required for efficient expression of such genes. It is known that many large introns are removed by recursive splicing in Drosophila and probably also in humans. In this process, the upstream exon becomes spliced to a ratcheting point (RP) within the intron. The sequence of the RP consists of overlapping 3' and 5' splice site motifs, so that a donor site is regenerated and used for splicing to the next RP. The process is repeated until the downstream exon is reached. I have demonstrated experimentally the recursive splicing of a terminal exon and a non-exonic element in the frizzled gene. Most RPs are not exonic, they are highly enriched within introns longer than 20 kb, and they have been conserved across large evolutionary distances. These observations suggest that recursive splicing plays an important role specifically in the expression of genes with long introns, possibly by reducing processing mistakes or stimulating transcript elongation by RNA polymerase II. To test this hypothesis I am deleting the single non-exonic RP from intron 2 of the outspread ( osp ) gene in Drosophila melanogaster . This intron spans 24 kb but is only one part of a much larger transcription unit that cannot be manipulated readily. I used high-fidelity PCR to isolate osp intron 2 and its flanking exons from Drosophila genomic DNA and to fuse this fragment to the Act5C promoter and polyadenylation site in a plasmid vector. I will also generate a second version of this construct in which the RP is precisely deleted. I will transfect these constructs into Drosophila SL2 cells and I will compare the efficiency of expression of the wild type and mutant versions. I will use Quantitative RT-PCR assays with different primer combinations to compare accumulation of transcripts, efficiency of splicing, and efficiency of transcript elongation through the intron.

Alexander Greenwood

Alexander Greenwood, Carnegie Mellon University
Mentor: Dr. John Nagle

Neutral Flotation Method for Determination of Lipid and Cholesterol Volumes in Lipid Bilayers

When dry phospholipids lipids are fully hydrated, they tend to self-assemble into multilamellar vesicles, or MLVs, due primarily to the hydrophobic effect. These vesicles are good model systems to obtain physical data on a single bilayer, which is the underlying structure of all cell membranes. The specific volume (and its inverse, the density) of lipid bilayers is a significant quantity of interest to biochemists and biophysicists. This quantity is needed in order to calculate the electron density profile, or structure, of a hydrated lipid. The specific volumes of many biologically relevant lipids lie between those of H2O and D2O, at least at some attainable temperatures. It is possible to determine whether a particular lipid is denser than its solvent by observing whether it sinks or floats. By creating many tubes of varying H2O and D2O concentrations, the specific volume of a lipid can therefore be pinpointed. H2O and D2O mix nearly perfectly, and it has been demonstrated by the Nagle lab that the H2O/D2O concentration inside the MLV is identical with that outside the MLV. The tendency to sink or float is enhanced by spinning the samples in a desktop centrifuge located in a temperature-controlled chamber (±0.5°C). The lipids measured are: POPC, diC22:1PC, Br 4 -DOPC, and mixtures of cholesterol and DPPC, DMPC, DOPC, POPC, and brain sphingomyelin. We pioneered a method of measuring lipids outside of the water/heavy water range by producing mixtures of varying ratios with a measurable lipid and extrapolating to x (mole ratio) = 1. We have determined the partial specific volume of cholesterol in the range of feasible cholesterol concentrations for DOPC, DMPC and POPC at 30° C, DPPC at 50° C, brain sphingomyelin at 45° C, and DOPC and brain sphingomyelin at 24° C.
Andy Hsieh

Andy Hsieh, Carnegie Mellon University
Mentor: Dr. Robert Murphy

Effects of Ras Oncogene on Subcellular Location Patterns in NIH 3T3 Cells

Using fluorescence microscopy to locate proteins that are labeled with fluorescent molecules or antibodies is a great tool for analyzing their distributions in cells. However, most analysis on protein subcellular distributions has been done with visual inspection, which is subject to personal bias and tends to have low reproducibility. An alternative approach would be generating numerical descriptions computationally that capture the essence of subcellular location patterns without being sensitive to the extensive variation that occurs in size, shape, and orientation of the cells in microscope images. In addition, they have to be robust across distinct microscopy methods and cell types. These numerical descriptions can also be expanded to make statistical comparisons of the distribution of a protein between two conditions, such as in the presence and absence of a cancer-promoting gene. Ras , which is mutated in 30% of human cancers, serves as the model oncogene. NIH 3T3 cell lines with unique proteins tagged with green fluorescent protein (GFP) from the Central Dogma-Tagging Project ( http://cdtag.bio.cmu.edu/www/public/index.html ) were transfected with a retroviral vector containing the Ras oncogene. Images were collected for the wildtype and Ras -containing pairs so the effects of Ras oncogene on subcellular location patterns could be examined. Of the 20 cell lines imaged, 14 proteins were found to have statistically significant differences ( P value < 0.05) in subcellular distributions between Ras + and Ras - . These proteins, having distinct properties compared to their wildtype counterparts, are candidates for screening and possible treatment for cancer. Future directions for this project include analyzing the temporal change in subcellular location patterns in Ras -transfected cell lines and evaluating the effectiveness of cancer drugs by determining the restoration of the altered location patterns.

Alexander Hu

Alexander Hu, Carnegie Mellon University
Mentor: Dr. Robert Murphy

The Systemic Analysis of Organelle Patterns in Adherent Cells of Varying Lineages

During development, pluripotent cells differentiate into specialized cell types, each which perform particular functions. Although all cells have basic organelles which perform tasks essential for viability, different specialized cells may have different organelle patterns and distributions in order to satisfy a cell's particular needs. Thus, it can be hypothesized that numerical information describing these organelle distribution patterns in digital images can be used to identify differences between differentiated cell types. In order to image these cells and their organelles, fluorescent probes specific for the mitochondria, lysosomes, Golgi apparatus, and nuclei were imaged using various filter sets in order to obtain multi-channel images. Approximately 13 cell lines were used for imaging. Roughly 50 images of each cell type were used for feature calculation and processing. From these digital images, numerical features regarding texture, pixel intensities, and various distance metrics can be used to classify these images through the use of a support vector machine (SVM). With this numerical information, we can learn how organelle patterns vary between cell lines. Clustering of the numerical features will also give us insight into whether cells from similar lineages have features that are more closely related.

Nicolette Louissaint

Nicolette Louissaint, Carnegie Mellon University
Mentor: Dr. Frederick Lanni

Large-scale Remodeling of Collagen Matrix by Swiss 3T3 Fibroblasts

In connective tissue, fibroblasts are in contact with extracellular matrix (ECM) which they make, maintain and repair. Fibroblasts are able to move through the ECM by extension and contraction, which enables them to carry out the biochemical and mechanical remodeling that converts relatively unstructured collagen into dense connective tissue having a high degree of fiber alignment. The main ECM component is type I collagen. Collagen is composed of long, high-molecular weight polypeptides that form triple helix structures which further assemble into fibrils and gels. In order to determine how fibroblasts utilize their actin-based cytoskeletal machinery to carry out motile processes (such as protrusion, migration and contraction) in an ECM, we form in vitro specimens composed of cells and collagen gel to model the ECM and observe the mechanical behavior of the cells. I have used time-lapse differential interference contrast (DIC) microscopy to observe the elastic deformations caused by single cells, and large-scale irreversible contractions caused by the collective action of groups of cells. The deformation quantification and analysis (DQA) website ( http://dqa.web.cmu.edu ) is capable of quantifying these deformations by image processing. My project also uses polarization microscopy to measure large-scale deformation and compaction of collagen caused by the cells. The deformation of the collagen gel depends on the density of cells, approximately 340 cells/ m L being optimal for Swiss 3T3 fibroblasts. The DQA site is used to track the collagen and compute the two-dimensional density change, which is the sum of the eigenvalues of the strain tensor. As the cells contract collagen, patterns of cell-fibril alignment become evident in the polarization image. Individual cells cause anisotropy by deforming the gel--affecting nearby cells, providing a mechanism for cooperative cell interaction. Our hypothesis is that the contractility of single cells causes ECM anisotropy, which is sensed by distant cells, and affects their behavior in a positive feedback loop. Contracted collagen gels produce aligned cells, an aligned actin cytoskeleton, and oriented collagen along the regions with the most strain. This research was supported by the HHMI Undergraduate Research Initiative.

James Malatack

James Malatack, Carnegie Mellon University
Mentor: Dr. William Brown

Optimizing Cell Surface Expression of Single Chain Variable Fragments (ScFv) on Yeast

Single Chain Variable Fragments (ScFvs) are engineered antibodies targeted against a specific antigen. Using a modified Enzyme-Linked Immunosorbent Assays (ELISA) a specific, engineered mouse antibody (ScFv12) was selected that targets a specific antigen. ScFv12 was then spliced into a plasmid (pCANTAB 5 E) and grown up in XL1Blue E. coli . ScFv12 was then spliced into a plasmid (pNL6) which, upon being transformed into Saccharomyces cerevisiae, was expressed as a cell surface fusion protein. ScFv12's presence on the cell wall was confirmed by a c-myc tag, which had been linked to the ScFv12, being located to the cell wall. However, ScFv12 antigen did not bind to the ScFv12, which was determined by using a Fluorescence Activated Cell Sorter (FACS). In order to solve this problem, S cFv12 will be spliced, using restriction enzymes and ligation, into a plasmid (pNL9) which, upon transformation into S. cerevisiae, secrets ScFv12 out of the cell. Agarose gels will be used to confirm DNA fragment sizes during the digestions and to determine if ligation has occurred. Purification of the ScFv from these transformed cells will enable analysis of antigen binding to the ScFv linked complex to determine whether the ScFv is functional when linked to the cell surface. Upon isolation of the ScFv12 from the media of the newly constructed pNL9/ScFv12 secretion yeast strain it was determined via a Western blot and ELISA that the ScFv12 was at a very high molecular weight and not functional in binding to the epitope, respectively. Sequencing will be done to help unravel why the pNL9/ScFv12 yeast are secreting such a large construct with the ScFv12.

Andreas Pfenning

Andreas Pfenning, Carnegie Mellon University
Mentor: Dr. Alison Barth

A Comparative Genomic Approach to Identifying the Plasticity Transcriptome

We used a computational genomic approach to identify genes that may exhibit activity-dependent or learning-related expression in mouse. The transcription factors CREB, c-Fos, and Zif268/Krox24/egr-1 are known to upregulate genes in response to neural activity. The presence of a binding site for one of these transcription factors in a gene's promoter therefore makes it a candidate for activity-dependent induction. We searched the sequences of more than 10,000 mouse and 10,000 human promoters from the Database of Transcription Start Sites (dbtss.hgc.jp/) for CREB, AP-1(c-Fos related binding site), and Zif268 binding sites. To perform this search, position specific scoring matrices (PSSMs) representing the specific binding sites were generated using experimentally determined binding sites from the Transfac Public Database ( www.gene-regulation.com ). Cross-validation of human and mouse promoters was used to identify candidate genes that were likely to have an activity-regulated component, under the assumption that a conserved transcription factor binding site is more likely to have regulatory function.

From these candidates, specific gene targets were chosen for experimental verification of activity-dependent gene regulation. We hypothesized that genes would show increased protein levels after increased activity induced by epileptic seizure. As an initial approach, candidates from both sets were evaluated using Western blotting of mouse cortical tissue homogenates 6 hours and 24 hours after seizure compared to control. Unexpectedly, none of the candidates showed a clear increase at these time points. It may be that increased protein levels occur less than 6 hours after seizure and are transient; thus our assay conditions may not be optimized to detect these changes. Further validation of our results could be performed using microarray data sets generated at various timepoints after seizure.

Jared Wenger

Jared Wenger, Carnegie Mellon University
Mentor: Dr. Elizabeth Jones

Mutational Analysis of the Saccharomyces cerevisiae Gene PBN1 through Alanine Scanning Mutagenesis

The study of budding yeast ( Saccharomyces cerevisiae ) genetics helps us understand important cellular pathways and protein functions, the knowledge of which can then be applied to human cells and the treatment of disease. S. cerevisiae PBN1 is an essential gene which encodes for an ER integral membrane protein (Pbn1p) required for maturation of Protease B (PrB), a vacuolar hydrolase. Previous work in our lab has shown that PBN1 must have other critical roles in the cell, as Protease B itself is not essential for viability. In order to investigate these other possible roles (which are suggested to be in protein folding and ER quality control), I hope to generate conditional mutations in PBN1 using an alanine scanning method. This method targets residues expected to be on the surface of the protein in order to disrupt critical areas for protein-protein interactions and ultimately the function of Pbn1p. This method uses an algorithm that scans the amino acid sequence of the protein in windows of 5 amino acids, searching for 2 or more charged residues. Wherever this condition is met, those residues are changed to alanine (to generate numerous, individual mutations). Alanine is a small, uncharged amino acid, so replacing charged amino acids with alanine will change the properties of that part of Pbn1p and may disrupt its function. Conditional phenotypes that we will be screening for at various temperatures are growth, sensitivity to DTT (dithiothreitol, a reducing agent), sensitivity to tunicamycin, PrB activity, and upregulation of Unfolded Protein Response. The alanine scanning method has been proven to work effectively to generate conditional mutations. I have currently created 16 individual mutations (using site-directed mutagenesis) that are dispersed throughout the protein. The next step of the project is to integrate these pbn1 mutant alleles into yeast strains and screen for conditional phenotypes. Once conditional mutants are isolated, the critical roles of PBN1 can be elucidated and more can be learned about protein folding, chaperone proteins, and ER quality control in this organism.

Elizabeth Wiltrout

Elizabeth Wiltrout, Carnegie Mellon University
Mentor: Dr. Danith Ly

Synthesis of Monomer for Formation of Peptide Nucleic Acid Microtubular Channels

Cyclic peptide nucleic acid (PNA) can be formed from linear alternating D- and L- peptides linked cyclically. Self-assembling tubular structures formed by hydrogen bonding of the cyclic PNA backbone have been reported. These tubular structures were described to have controlled diameters by alternating the number of peptides in the linear chain. Lengths of the tubular structures could not be controlled due to spontaneous hydrogen bonding between the backbones. We propose a synthesis for a monomer and cyclic PNA that can hydrogen bond the side chains by Watson-Crick base pairing to form nanotubes whose lengths may be controlled by varying the ratio of cyclic PNAs in solution. Synthesis schemes for a monomer capable of Watson-Crick base pairing from both top and bottom of the ring is proposed using standard organic syntheses. This monomer will be used to synthesize cyclic PNA with eight of the monomers by solid phase synthesis. Formation of nanotubes made from the cyclic PNA can be monitored by atomic force microscopy. Nanotubes with a controlled length have practical uses when applied to biological systems. A nanotube can act as a transmembrane ion channel and may be used for drug delivery to specific cells.