Unsupervised Clustering of Subcellular Protein Expression Patterns in High-Throughput Microscopy Images Reveals Protein Complexes and Functional Relationships between Proteins

Protein subcellular localization has been systematically characterized in budding yeast using fluorescently tagged proteins. Based on the fluorescence microscopy images, subcellular localization of many proteins can be classified automatically using supervised machine learning approaches that have been trained to recognize predefined image classes based on statistical features. Here, we present an unsupervised analysis of protein expression patterns in a set of high-resolution, high-throughput microscope images. Our analysis is based on 7 biologically interpretable features which are evaluated on automatically identified cells, and whose cell-stage dependency is captured by a continuous model for cell growth. We show that it is possible to identify most previously identified localization patterns in a cluster analysis based on these features and that similarities between the inferred expression patterns contain more information about protein function than can be explained by a previous manual categorization of subcellular localization. Furthermore, the inferred cell-stage associated to each fluorescence measurement allows us to visualize large groups of proteins entering the bud at specific stages of bud growth. These correspond to proteins localized to organelles, revealing that the organelles must be entering the bud in a stereotypical order. We also identify and organize a smaller group of proteins that show subtle differences in the way they move around the bud during growth. Our results suggest that biologically interpretable features based on explicit models of cell morphology will yield unprecedented power for pattern discovery in high-resolution, high-throughput microscopy images.     

A paradigm for therapy-induced microenvironmental changes in solid tumors leading to drug resistance

Intrinsic alterations in the tumor microenvironment are known to contribute to various forms of drug resistance. For example, tumor hypoxia, due to abnormal or sluggish blood flow within areas of solid tumors, can result in both microenvironment-mediated radiation and chemotherapeutic drug resistance. In contrast, acquired resistance to chemotherapy is generally considered to be the result of the gradual selection of mutant subpopulations having genetic mutations and biochemical alterations responsible for the resistant phenotype. Here we present a paradigm for therapyinduced microenvironment-mediated acquired drug resistance. It is based on the results showing that tumor cells appear to be heterogeneous in their relative dependence on adjacent tumor-associated vasculature for survival. Some tumor cells are highly vessel dependent, whereas some are significantly less so, and thus can survive in more hypoxic regions of tumors, distal from such tumor vessels. Hence, it is possible that variant tumor cells that are less vessel dependent may therefore be selected for over time by successful antiangiogenic drug therapies. This results in loss of response or attenuated responses to the therapy. Preliminary evidence is summarized in support of this hypothesis, using paired human colon cancer (HCT116) cell lines that contain two copies of either the wild-type or the disrupted p53 tumor suppressor gene. The mutant cells were found to be less responsive to antiangiogenic therapy, compared to the wild-type cells, and could be progressively selected for in mixed cell populations. Because p53 inactivation can lead to resistance to hypoxia-mediated apoptosis, the results suggest that a protracted and successful antiangiogenic therapy may create more hypoxic tumor microenvironments, thereby creating the necessary conditions to accelerate the selection of mutant tumor cells that are more adept in surviving and growing in such environments. As such, consideration might be given to the combined use of bioreductive hypoxic cell cytotoxic drugs and angiogenesis inhibitors to prolong the efficacy of antiangiogenic therapeutics.     

Inhibition of endosomal/lysosomal degradation increases the infectivity of human immunodeficiency virus

Productive entry of human immunodeficiency virus type 1 (HIV-1) into a host cell is believed to proceed via fusion of the viral envelope with the host cell's plasma membrane. Interestingly, the majority of HIV-1 particles that bind to the cell surface are taken up by the host cell via endocytosis; however, this mode of internalization generally does not result in infection. Presumably, virus particles remain trapped in the endocytic pathway and are eventually degraded. Here, we demonstrate that treatment of cells with various pharmacological agents known to elevate the pH of endosomes and lysosomes allows HIV-1 to efficiently enter and infect the host cell. Pretreatment of cells with bafilomycin A1 results in up to a 50-fold increase in the infectivity of HIV-1(SF2). Similarly, pretreatment of target cells with amantadine, concanamycin A, concanamycin B, chloroquine, and ammonium chloride resulted in increases in HIV-1 infectivity ranging between 2- and 15-fold. Analysis of receptor and coreceptor expression, HIV-long terminal repeat (LTR) transactivation, and transduction with amphotropic-pseudotyped murine leukemia virus (MLV)-based vectors suggests that the increase in infectivity is not artifactual. The increased infectivity under these conditions appears to be due to the ability of HIV-1 and MLV particles to enter via the endocytic pathway when spared from degradation in the late endosomes and lysosomes. These results could have significant implications for the administration of current and future lysosmotropic agents to patients with HIV disease.     

Time course of NADH oxidase, inducible nitric oxide synthase and peroxynitrite in diabetic retinopathy in the BBZ/WOR rat.

This study investigated the time course of NADH oxidase, a source of superoxide in the vascular endothelium, inducible nitric oxide synthase (iNOS), and peroxynitrite (ONOO(-)) in the BBZ/Wor rat, a spontaneous model of noninsulin dependent diabetes (NIDDM). Colloidal gold-labeled immunocytochemical studies of iNOS and nitrotyrosine, a marker for OONO(-), were done on sections of retinas from male BBZ/Wor rats in which NADH oxidase was localized by cerium derived cytochemistry at three time points: pre-diabetes (prior to the onset of hyperglycemia); new onset diabetes (2-6 days after onset of hyperglycemia); and chronic diabetes (4-18 months after onset of hyperglycemia). Control retinas were from age matched non-diabetic BB(DR)/Wor rats. The percentage of blood vessels positive for NADH oxidase increased significantly (P = 0.05) in new onset (64.2 +/- 6.5%) and chronic diabetes (83.2 +/- 11.4%), as compared to pre-diabetes (25.8 +/- 5.6%) and nondiabetic controls (33.6 +/- 15.9%). The percentage of blood vessels positive for iNOS immunoreactivity was significantly higher in new onset diabetic retinas (69.6 +/- 5.88%, P = 0.0001; 8.9 +/- 3.29 colloidal gold particles (cgp) /50 microm(2)) than in chronic diabetic retinas (49.9 +/- 9.75%; 7.9 +/- 5.12 cgp) and both were significantly higher (P = 0.0001) than in prediabetic (3.7 +/- 0.81%; 0.4 +/- 0.56 cgp) and nondiabetic control retinas (8.7 +/- 4.66%; 1.2 +/- 1.40 cgp). In new onset diabetes, levels of nitrotyrosine immunoreactivity (60.8 +/- 16.91 cgp) were significantly higher (P = 0.0001) than those in chronic diabetes (29.5 +/- 4.31 cgp); both were significantly higher (P = 0.0001) than those in prediabetic (8.2 +/- 1.70 cgp) and nondiabetic retinas (9.0 +/- 1.87 cgp). There was no cumulative increase in nitrotyrosine in the chronic diabetic retinas as a function of time. In rats with diabetes there was disruption of the inner blood-retinal barrier. These results suggest that iNOS and ONOO(-) may contribute to retinal damage in diabetes from the onset of hyperglycemia in NIDDM.     

Role of conserved transmembrane cationic amino acids in the prostaglandin transporter PGT

The prostaglandin transporter "PGT" interacts electrostatically with its anionic substrate, based on inhibition by the disulfonic stilbenes [Chan, B. S. (1998) J. Biol. Chem. 273, 6689-6697], inhibition by the thiol-reactive anion sodium (2-sulfonatoethyl)methanethiosulfonate (MTSES) [Chan, B. S. (1999) J. Biol. Chem. 274, 25564-25570], and the requirement for a negatively charged 1-position carboxyl on the substrate [Itoh, S. (1996) Mol. Pharm. 50, 736-742]. Here we found that modification of positively charged residues on wild-type PGT by arginine- and lysine-specific reagents significantly inhibited transport. We previously found that the binding site of PGT is formed, at least in part, by its membrane-spanning segments [Chan, B. S. (1999) J. Biol. Chem. 274, 25564-25570]. Three charged residues within predicted transmembrane spans (E78, R560, and K613) are conserved in PGT and in related transporters. Substitution of the anionic residue E78 (E78D and E78C) produced an essentially functional transporter, whereas substitution of the cationic residues with neutral residues (R560N and K613Q) resulted in poorly functional transporters. Immunoblotting revealed similar expression levels of wild-type and mutant transporters, and immunostaining indicated correct targeting. Conservative charge substitutions (R560K, K613R, and K613H) resulted in generally functional transporters. In contrast, R560N was nonfunctional, whereas the substrate affinity of K613G decreased greater than 50-fold. Conservative substitutions retaining the charge at position 613 (K613R and K613H) restored the substrate affinity, suggesting a direct role of K613 in substrate binding. Double-neutral mutants E78G/R560C and E78G/K613C were inactive, indicating that these residues are not simply charge-paired. Our results suggest that an arginine at position 560 is critical for maximal substrate translocation, and that a positively charged side chain at position 613 contributes to electrostatic binding of the anionic substrate.     

Systematic evaluation of muscle coenzyme Q10 content in children with mitochondrial respiratory chain enzyme deficiencies

Coenzyme Q10 content, pathology evaluation, and electron transport chain (ETC) enzyme analysis were determined in muscle biopsy specimens of 82 children with suspected mitochondrial myopathy. Data were stratified into three groups: "probable" ETC defects, "possible" ETC defects, and disease controls. Muscle total, oxidized, and reduced coenzyme Q10 concentrations were significantly decreased in the probable defect group. Stepwise logistic regression indicated that only total coenzyme Q10 was significantly associated with probable ETC defect. Receiver operator characteristic (ROC) analysis suggested that total muscle coenzyme Q10 was the best predictor of an ETC complex abnormality. Determination of muscle coenzyme Q10 deficiency in children with suspected mitochondrial disease may facilitate diagnosis and encourage earlier supplementation of this agent.     

Reciprocal regulation of syndecan-2 and Notch signaling in vascular smooth muscle cells

Vascular cell interactions mediated through cell surface receptors play a critical role in the assembly and maintenance of blood vessels. These signaling interactions transmit important information that alters cell function through changes in protein dynamics and gene expression. Here, we identify syndecan-2 (SDC2) as a gene whose expression is induced in smooth muscle cells upon physical contact with endothelial cells. Syndecan-2 is a heparan sulfate proteoglycan that is known to be important for developmental processes, including angiogenesis. Our results show that endothelial cells induce mRNA expression of syndecan-2 in smooth muscle cells by activating Notch receptor signaling. Both NOTCH2 and NOTCH3 contribute to the increased expression of syndecan-2 and are themselves sufficient to promote its expression independent of endothelial cells. Syndecan family members serve as coreceptors for signaling molecules, and interestingly, our data show that syndecan-2 regulates Notch signaling and physically interacts with NOTCH3. Notch activity is attenuated in smooth muscle cells made deficient in syndecan-2, and this specifically prevents expression of the differentiation marker smooth muscle α-actin. These results show a novel mechanism in which Notch receptors control their own activity by inducing the expression of syndecan-2, which then acts to propagate Notch signaling by direct receptor interaction.     

Selection of synthetic proteins to modulate the human frataxin function

Frataxin is a kinetic activator of the mitochondrial supercomplex for iron-sulfur cluster assembly. Low frataxin expression or a decrease in its functionality results in Friedreich's Ataxia (FRDA). With the aim of creating new molecular tools to study this metabolic pathway, and ultimately, to explore new therapeutic strategies, we have investigated the possibility of obtaining small proteins exhibiting a high affinity for frataxin. In this study, we applied the ribosome display approach, using human frataxin as the target. We focused on Affi_224, one of the proteins that we were able to select after five rounds of selection. We have studied the interaction between both proteins and discussed some applications of this specific molecular tutor, concerning the modulation of the supercomplex activity. Affi_224 and frataxin showed a K_D value in the nanomolar range, as judged by surface plasmon resonance analysis. Most likely, it binds to the frataxin acidic ridge, as suggested by the analysis of chemical shift perturbations (nuclear magnetic resonance) and computational simulations. Affi_224 was able to increase Cys NFS1 desulfurase activation exerted by the FRDA frataxin variant G130V. Importantly, Affi_224 interacts with frataxin in a human cellular model. Our results suggest quaternary addition may be a new tool to modulate frataxin function in vivo. Nevertheless, more functional experiments under physiological conditions should be carried out to evaluate Affi_224 effectiveness in FRDA cell models.     

Distinct self-interaction domains promote Multi Sex Combs accumulation in and formation of the Drosophila histone locus body

Nuclear bodies (NBs) are structures that concentrate proteins, RNAs, and ribonucleoproteins that perform functions essential to gene expression. How NBs assemble is not well understood. We studied the Drosophila histone locus body (HLB), a NB that concentrates factors required for histone mRNA biosynthesis at the replication-dependent histone gene locus. We coupled biochemical analysis with confocal imaging of both fixed and live tissues to demonstrate that the Drosophila Multi Sex Combs (Mxc) protein contains multiple domains necessary for HLB assembly. An important feature of this assembly process is the self-interaction of Mxc via two conserved N-terminal domains: a LisH domain and a novel self-interaction facilitator (SIF) domain immediately downstream of the LisH domain. Molecular modeling suggests that the LisH and SIF domains directly interact, and mutation of either the LisH or the SIF domain severely impairs Mxc function in vivo, resulting in reduced histone mRNA accumulation. A region of Mxc between amino acids 721 and 1481 is also necessary for HLB assembly independent of the LisH and SIF domains. Finally, the C-terminal 195 amino acids of Mxc are required for recruiting FLASH, an essential histone mRNA-processing factor, to the HLB. We conclude that multiple domains of the Mxc protein promote HLB assembly in order to concentrate factors required for histone mRNA biosynthesis.