Mapping Organelle Proteins and Protein Complexes in Drosophila melanogaster

Abstract

Many proteins within eukaryotic cells are organized spatially and functionally into membrane bound organelles and complexes. A protein's location thus provides information about its function. Here, we apply LOPIT, a mass-spectrometry based technique that simultaneously maps proteins to specific subcellular compartments, to Drosophila embryos. We determine the subcellular distribution of hundreds of proteins, and protein complexes. Our results reveal the potential of LOPIT to provide average snapshots of cells.

A Foundation for Reliable Spatial Proteomics Data Analysis

“…change chromatography and analyzed via LC-MS/MS on a QSTAR XL quadrupole-time-of-flight mass spectrometer (Applied Biosystems, Foster City, CA). Tan et al (18) collected Drosophila melanogaster embryos at 0 to 16 h. The material was homogenized and centrifuged to collect the supernatant, thereby removing cell debris and nuclei. Membrane fractionation was performed on an iodixanol gradient, and fractions were quantified using iTRAQ 4-plex isobaric tags (16) as follows: fractions 4/5, 114; fractions 12/13, 115; fraction 19, 116; and fraction 21, 117.…”

“…Missing data and the impact of imputation have not been thoroughly addressed in proteomics, let alone in spatial proteomics. Published LOPIT studies (7,18,20) have excluded proteins that presented missing values across replicated experiments. PCP studies (8,21) have limited the computation of their 2 metric to pairs of fractions (defined as the squared deviation of the normalized profile for all peptides divided by the number of data points), thus increasing the bias by reducing the number of data points.…”

“…organelle markers) from which the applied machine learning system will learn is extremely important, as it can have a significant effect on the success or failure of the learner. To illustrate this paradigm, we investigated the effect that marker choice has upon the application of the phenoDisco novelty detection algorithm in mining a Drosophila melanogaster dataset that had been produced using the LOPIT technology (18). We considered three sources of markers for use as input for the phenoDisco algorithm: (i) a highly and manually curated set from experts in the field (20 endoplasmic reticulum (ER), 6 Golgi, 14 mitochondrial, and 15 plasma membrane markers from our curated marker sets, originally obtained from Ref.…”

“…As an example of an application of protein localization prediction using supervised machine learning, we took the first replicate from Tan et al (18) and applied a weighted SVM classifier for protein classification. The labeled training data (Fig.…”

The Current State of the Golgi Proteomes

“…A total of 329 Drosophila proteins were identified and localized to three subcellular locations; the plasma membrane (94), mitochondria (67) and the ER/Golgi (168) (Tan et al, 2009). The lack of distinction between ER-and Golgi-residing Drosophila proteins by LOPIT underscored the significant challenges faced when dissecting complex and heterogeneous biological samples, as opposed to a simplified system of crude membranes from a relatively homogenous Arabidopsis cell culture.…”