The study of membrane proteins is a difficult topic on its own because of the particular characteristics of these proteins. The study of membrane proteins in a complex mixture where degradation is a natural process is even more challenging.
In this study I successfully developed a robust protocol for the BBM preparation from mouse intestinal mucosa, a tissue whose main function is to degrade nutrients in order to facilitate their ingestion. The inherent protein and RNA degradation which I faced during my early work had to be considered and taken care of to obtain a reproducible BBM preparation. A significant amount of time and effort was spent on developing protocols for the inhibition of protein and RNA degradation, which at the end was partially achieved. It is worth pointing out that none of the published studies previously dealing with BBM analysis using mass spectrometry, immunoassays or microarrays had mentioned any issue regarding tissue degradation.
A triplicate identification of the proteins part of the BBM resulted in the identification of more than 1460 proteins, of which “only” 260 were integral membrane protein, an apparently disappointing result considered that the preparation was thought to be highly enriched in membrane proteins. However, a detailed analysis of the proteins identified in the BBM preparation made apparent that a large extent of the remainder was accounted for membrane-associated, membrane-anchored proteins and cytoskeleton proteins involved in protein trafficking between cytosol and plasma membrane, thus closely related functionally to the BBM function. Even some of the 330 cytosolic proteins, such as the apolipropteins which could have been discounted as contaminant at a first glance, were demonstrated to be specifically co-enriched in the BBM fraction probably due to their interaction with a BBM-specific component.
A rough estimation of protein abundance based on peptide counts confirmed that membrane proteins and membrane-associated proteins represented the dominant species in the BBM preparation. In this analysis we didn’t observe any particular bias against proteins with many transmembrane helices. Nevertheless, the identification of these proteins was based on tryptic peptides exclusively located in the loops, in extracellular areas or in the intracellular domains of the proteins. I believe that the reason for the lack of identification of tryptic transmembrane peptides (assuming that trypsin was equally active in this environment) was due in many cases to the length of the generated peptides, typically of 40-50 amino acids long. Indeed, the elution of highly charges peptides of 5-6 kDa mass were occasionally observed in the elution
chromatograms. Unfortunately, due to software limitation (peptide with charge state above +3 could not be considered in an automatic data analysis), these peptides were not analyzed and further studies will be required to confirm this observation.
This study is the first proteomic approach in which numerous receptors and transporters and many relevant proteins to cholesterol absorption were identified. This is in contrast to several recently published proteomic studies (2, 122) claiming the characterization of specific membrane proteins of BBM. However, their protein lists were mostly restricted to Ras-related proteins and cytosolic or membrane associated-proteins. Aminopeptidase N and Sodium/Glucose co-transporter, two of the most abundant proteins in the BBM, were characterized in a proteomic approach for the first time in this study. Similarly, some proteomic studies have been based on the isolation of the lipid rafts (possible BBM micro domains responsible for cholesterol absorption) to enrich and identify proteins that participate in cholesterol absorption (3). None of these studies reports the identification the Niemann-Pick C1-like protein 1, the target of ezetimibe (a cholesterol absorption inhibitor) and for many researchers the likely transporter of cholesterol. In this study, the Niemann-Pick C1-like protein 1 was one of the most abundant proteins in the purified BBM fraction.
These data re-emphasize the critical importance of a reproducible isolation and fractionation of a membrane preparation in a high-throughput proteomic approach, even if using a mass spectrometer with high resolving power and excellent mass accuracy. In this study, a robust BBM isolation protocol (in which the unspecific proteolytic activity of the preparation was controlled by adding a cocktail of protease inhibitors and peptide substrates) led us to the characterization at the protein level of a large number of cholesterol- or fat absorption-related proteins which, until now, had only been characterized at the gene level or by immunoassays.
Thus, a membrane proteomic approach is feasible and may yield excellent results when the analyzed membrane fraction is optimally purified. Evidently, the requirement for a complex purification and fractionation scheme is accompanied by an increasing analysis complexity.
Preparation of the BBM in sections enabled the identification of some additional BBM-specific proteins and provided some examples of protein localization along the small intestine.
For example, the SR-BI receptor was identified in the BBM fraction prepared from total mucosa with one single peptide (and hence was not included in the final list in appendix A1).
In contrast, it was robustly characterized in the BBM fraction prepared from the duodenum segment of the small intestine with three different peptides. The benefit of the additional fractionation was obvious in respect to the confident identification of this receptor. However, it required three times more measurement time and a rather large increase of starting material.
The cost/benefit of the selected isolation scheme needs to be carefully considered in respect to how many biological replicates or differential experiments are planned.
The specific identification of the ApoE protein in the BBM fraction captured our interest and led us to a prototype comparison study between BBM preparations from ApoE knockout mice and wild type mice of the same genetic background. Following the same preparation protocol and a well established analytical strategy, we were able to see clear differences in the expression level of several proteins between the two different species. Since no quantification technique was applied and there was no biological replicate, we focused our interest in the
“black and white” differences. Among them, the Ileal Bile Acid Transporter (IBAT) protein and the ApoAI protein, which were robustly identified in the wild type animals, couldn’t be detected in the ApoE knockout mice. This finding was not described in any of the earlier published studies that have attempted to elucidate the reasons why the ApoE knockout mice appear hypercholesterolemic, have high plasma cholesterol and high cholesterol absorption.
In particular, the striking downregulation of the IBAT transporter in the ApoE knockout animal strongly suggests a disruption of the bile acids metabolism. This hypothesis is supported by a separate study that showed that the bile acid pool size in the liver and the intestine of the ApoE knockout mice was significantly increased compare to the wild type mice. The increase of the bile acids pool available in the small intestine of the ApoE knockout mice might explain the higher cholesterol absorption of these animals since cholesterol is better solubilized and, consequently, transported more easily through the BBM. This part of my study provides a completely new perspective to the high cholesterol absorption of these animals and opens a new field of research in this topic.
Similarly to this prototype study, I am convinced that the analytical strategy described in this study is sufficiently mature to perform comprehensive comparative analysis of mice that have for example been treated with specific compound or that have been subjected to different diets.
The simplistic analysis method used in the ApoE knockout example was sufficient to pinpoint to obvious changes. However, in a more complex dataset, a proper quantification strategy should provide a better leverage to discover differences with statistical significance. The 15N full metabolic labeling of mice might represent an ideal strategy for this type of samples. The mixing of an internal standard in the form of a “heavy” BBM preparation in all the samples of interest enables very accurate protein quantification (in principle, all BBM proteins could be normalized to their specific internal standard) over the whole experiment. In addition, the co-purification of the 15N internal standard with the regular BBM preparation provides a mean to compensate for variability in degradation across the experiment.
A significant amount of time and effort was spent in this study for evaluating the stability of the BBM preparation protocol, the reproducibility of the analytical steps leading to the protein identification and, further, the consistence of these analyses with an initial comparative investigation of how to extract the same information directly from the LC-MS precursor ion signals. Overall, the BBM isolation protocol that was developed in this study, combined with a reproducible analytical strategy, enabled us to identify almost 90% of all the 1639 proteins identified in the BBM fraction in at least two of three triplicates. This level of reproducibility was beyond our expectations and encouraged us to individually monitor the variability of each of the technical steps involved in the procedure. The data confirmed the critical stability and reproducibility of the Orbitrap mass spectrometer for the high throughput analysis of such samples. The study pinpointed to an unexpected source of variation in the form of the LC buffers, which need to be tightly controlled to obtain reproducible chromatographic separation conditions. Ideally, samples to be compared must be analyzed using identical buffer and column batches on the LC system.
My study in this respect initiated a discussion for better understanding the analytical requirements and the type of software tools that might allow in the future label-free quantification approaches based on LC-MS data alone. Quantification at the precursor mass signal intensity might provide a more general method of comparison for LC-MS-based proteomics studies than the comparison of protein lists, which only consider approximately 10 to 20% of the available data. Protein identification of signals of interest can then be obtained using MS/MS inclusion lists, where an ion signal of specific m/z and RT is targeted for tandem mass spectrometric analysis. Most importantly, the ability to differentially compare samples across an experiment first requires the knowledge of what is common to the technical and biological replicates of samples so that the significance of a given change can be statistically appreciated. Within this study, we determined what was achievable with regards to reproducibility using standard protein purification technique and advanced analytical strategies but we also described the current limitations that will need to be addressed in the future to fully enable this strategy.