Filtering The Transcriptome Enhancing The Value

The Big Heart Disease Lie

Cure Cardiovascular Disease Forever

Get Instant Access

Gene arrays as applied to cardiovascular disease are inherently noisy experiments as even in a seemingly genetically homogenous population the severity and penetrance of the disease is not necessarily uniform. In order to raise the signal-to-noise ratio so that interpretable data are obtained, some form of filtering is essential. This can occur at either the "back-end" using bioinformatics, or at the "frontend" by careful selection of the sample population ( churchill/research/expression/in dex.html) [45] or by sub selection of only the most relevant portion of the transcriptome (see below). Additionally, the overall experimental design is critical and details are often overlooked or not reported so that it becomes problematic for another experimentalist to exactly repeat the procedure. Seemingly trivial matters can significantly affect a particular experiment. For example, how are the animals sacrificed? If an animal smells the blood of another in the procedure room, a stress response can be initiated. Similarly, the cir-cadian rhythm can have a significant impact on the hormonal status of an animal, which could impact cardiovascular status. These considerations are rarely acted upon but could, if carefully considered, minimize the inherent biological noise of the experiment.

The Back-end - Very few, if any studies have documented the fulfillment of the initial catalogue's promise. For example, hypertrophy is a common adaptive response of the heart to increased workload, injury and stress, and has been characterized in many models as being defined by the activation of a common set of genes, which are normally only expressed during early cardiac development. While a number of different models of hypertrophy [46] or myocardial infarction [47] have been catalogued at the transcriptional level, the difficulty lies in translating what are essentially thousands upon thousands of observations into a coherent, testable model that can be biologically validated either through gain or loss-of function studies. However, one of the more carefully considered cardiovascular studies illustrates the value of multiple comparisons and rigorous mining of the resultant data. Aronow et al. set out to examine the "monolithic" hypothesis, that is, a gene program common to different hypertropies exists, by comparing the transcriptomes of four genetic hypertrophy models that showed varying degrees of the hypertrophic response [48]. DNA microarrays were used to compare approximately 9000 mRNAs in the four transgenic mouse models. Although the total number of regulated genes (defined as a certain - fold up- or down-regulated) var ied between the models with the numbers corresponding to the relative severity of the phenotypic response, no commonality between the four models could be defined. However, by applying a modest amount of analysis using hierarchical-tree and K-means clustering to the data sets, patterns involving 276 genes that were regulated among the four models could be discerned, including a subset associated with the activation of apoptosis. Northern analyses were subsequently used to confirm the microarray patterns and the biological annotation initiated using the existing databases.

The Front-end - In order to enhance the signal-to-noise ratio of an experiment, its design can also apply filters at the "front-end". For example, rather than interrogating the entire RNA complement of the cell, one can attempt to restrict it only to those transcripts that will be processed into the proteome [49]. We recently carried out such a study in order to explore the efficacy of the "front-filter" approach [50]. The ^-agonist, isoproterenol, when administered over a 10-14 day period is a simple and well-characterized protocol that results in a characteristic 20-30% cardiac hypertrophy as measured by the heart to body weight ratios [51]. However, instead of interrogating the entire RNA complements of the treated and untreated animals, only the actively translated RNA was studied. Polysomes derived from the animals' ventricles were loaded on sucrose gradients, size fractionated using velocity density centrifugation and the RNA from these fractions used to select for transcripts that were loaded onto polysomes in response to isoprotere-nol. Four Clontech Atlas 1.2 microarray filters were simultaneously hybridized to radiolabeled cDNA probes derived from either vehicle-treated (control) free or polysome bound RNA, or from isoproterenol-treated free or polysome bound RNA. A numerical value for the shift to polysomes was calculated by taking the ratio of isoproterenol bound/free signal divided by the ratio of vehicle-treated bound/free signal. Signals were normalized to a median filter value to correct for differences in probe specific activities with increases of a particular transcript in the polysome fraction due to chronic isoproterenol infusion resulting in a value of >1.0. Thus, while a particular transcripts steady state level might not be increased during the treatment, if its translational efficiency was affected, it would be detected. This high-throughput screen, designed to identify only the transcripts that are actively translated during cardiac hypertrophy or whose translation is down-regulated during the physiological response, identified a number of genes with established links to hypertrophy, including Sp3, c-jun, annexin II, cathepsin B, and HB-EGF [52-56], confirming the screen's accuracy. However, in order to test the usefulness of the screen, we decided to focus on a candidate transcript that had not been previously linked to hypertrophy and found that protein levels of the tumor suppressor PTEN (phosphatase and tensin homologue on chromosome ten) were increased in the absence of increased messenger RNA levels (Fig. 2.1). While overall, the mRNA levels of PTEN were not increased as result of isoproterenol treatment, the movement of the existing transcripts into the heavy portion of the polysomes was. Quantitative western blot analyses showed that PTEN protein expression is, in fact, induced in isoproterenol-treated mouse hearts relative to vehicle-treated hearts [50], in agreement with the polysome-derived data. Taken to

Fig. 2.1 Polysome-derived RNA levels change although total RNA amounts remain stable. For each array position that had a signal greater than 0.5X the median filter value as well as a signal on all membranes, the ratio (TI/UI)/(TS/US), where: US, untranslated sham (vehicle solvent only); TS, translated sham; UI, untranslated isoproterenol treated; TI, translated isoproterenol treated. A ratio >1 indicates a shift toward polysomes in the isoproterenol treated hearts and <1 indicates a shift away from polysomes. Only candidates that had a ratio of >2 were selected. ARNA.

Fig. 2.1 Polysome-derived RNA levels change although total RNA amounts remain stable. For each array position that had a signal greater than 0.5X the median filter value as well as a signal on all membranes, the ratio (TI/UI)/(TS/US), where: US, untranslated sham (vehicle solvent only); TS, translated sham; UI, untranslated isoproterenol treated; TI, translated isoproterenol treated. A ratio >1 indicates a shift toward polysomes in the isoproterenol treated hearts and <1 indicates a shift away from polysomes. Only candidates that had a ratio of >2 were selected. ARNA.

Total RNA changes were calculated by summing signals from all four array membranes. With the exception of the serine/threonine kinase, pim-1 (gene #23), the candidate genes exhibited minimal RNA fluctuation. APolysomes. Significant changes occurred in the degree of polysome loading (movement to the heavy fraction). PTEN was selected on the basis of the subsequent, biological annotation. Note that the constitutive markers, GAPDH and alpha actin, remained relatively unchanged.

gether with the shift of PTEN mRNA into the heavy polysome fractions during hypertrophy and the minimal change of total PTEN mRNA, this finding is consistent with regulation of PTEN expression by increased translational initiation.

PTEN was originally identified as a human tumor-suppressor gene and is also called MMAC1/TEP1 (MMAC1, mutated in multiple advanced cancers-1; TEP1, TGF-/5 regulated, epithelial cell enriched phosphatase). The gene is either deleted or inactivated in a high percentage of breast, endometrial, brain and prostate cancers [57-59]. A potent tumor-suppressor function has been confirmed by performing an in vivo loss-of-function via gene ablation studies in mice. Mice with only one functional copy of the gene are more likely to develop tumors of multiple origins, while loss of both alleles leads to embryonic lethality [60-62].

PTEN is a dual-specificity phosphatase with homology to the focal adhesion-associated protein tensin [63]. In vitro, PTEN can dephosphorylate acidic polypeptides, focal adhesion kinase (FAK), and the adaptor protein, Shc. However, the major in vivo substrate for PTEN appears to be phosphatidylinositol 3, 4, 5-tripho-sphate (PIP3), as embryonic fibroblasts taken from PTEN null mouse strains have abnormally high levels of PIP3 and are resistant to apoptosis [61]. The PTEN-/- fibroblasts have very high levels of activated Akt, a serine/threonine kinase that is regulated by PIP3 and phosphatidylinositol 3, 4-biphosphate (PIP2). Intriguingly, Akt is an important regulator of both cell survival and growth [64], and PTEN has been defined genetically and biochemically to act as a negative regulator of Akt in opposition to the evolutionarily conserved IGF-1/PI3K/Akt signaling pathway [6567]. Thus, the biological annotation for this candidate is extraordinarily rich, although the data are from systems other than the heart. Nevertheless, the overall functional annotation of PTEN was of sufficient value as to warrant further exploration.

High throughput functional screens are critical for assigning biological value to candidates identified by genome or proteome wide screens. A general limitation of the cardiovascular field is the paucity of accurate screens for determining a candidate protein's functional role, unless one assumes in vitro binding assays with putative partners (such as defined transcriptional factor activation domains, etc.) are truly accurate representations of a candidate's biological role. Therefore, to determine the (potential) role PTEN might play during cardiac hypertrophy, we restricted our approach to either cell culture or transgenic animals using complementary gain- and loss-of-function approaches whenever possible.

To explore the possibility of a biological function for PTEN in the hypertrophic response, adenovirus was used to overexpress the protein in primary cultures of neonatal rat cardiomyocytes. Overexpression resulted in fewer viable cells as a result of apoptotic pathway activation. More interestingly, expression of a catalyti-cally inactive form of PTEN [63] we termed H123YPTEN, led to cardiomyocyte hypertrophy with a well-ordered sarcomeric structure being conserved in the cardiomyocytes as shown by a-actinin staining [50]. Molecular markers, cell volume and shape, as well as protein synthetic rates were all consistent with the inactive form mediating a robust hypertrophy in the cultured cardiomyocytes [50].

There are several possible mechanisms by which H123YPTEN might act as a dominant suppressor of endogenous PTEN, including the sequestration of PTEN binding partners needed for full activity. For example, PTEN can bind to focal adhesion kinase directly, leading to its dephosphorylation and inactivation. Consistent with the ability of H123YPTEN to sequester FAK from endogenous PTEN, FAK tyrosine phosphorylation in AdH123Y infected cells was increased. We are currently analyzing whether the H123Y mutation stabilizes the interaction with FAK in cardiomyocytes.

Although these data are intriguing they do not provide enough information about the way PTEN acts within the cellular networks to make any firm conclu sions about its role in a physiologically relevant hypertrophy. Rather, the data illustrate that cell culture experiments represent yet another biological filter against which to test putative candidates isolated from the screens. The data do, however, justify a more extensive exploration of PTEN's activity within the whole organ and whole animal contexts using drug-inducible, cardiac-specific transgenesis.

Was this article helpful?

0 0
Your Heart and Nutrition

Your Heart and Nutrition

Prevention is better than a cure. Learn how to cherish your heart by taking the necessary means to keep it pumping healthily and steadily through your life.

Get My Free Ebook

Post a comment