Peter J. Ulintz

Novel gene and gene model detection using a whole genome open reading frame analysis in proteomics

BackgroundDefining the location of genes and the precise nature of gene products remains a fundamental challenge in genome annotation. Interrogating tandem mass spectrometry data using genomic sequence provides an unbiased method to identify novel translation products. A six-frame translation of the entire human genome was used as the query database to search for novel blood proteins in the data from the Human Proteome Organization Plasma Proteome Project. Because this target database is orders of magnitude larger than the databases traditionally employed in tandem mass spectra analysis, careful attention to significance testing is required. Confidence of identification is assessed using our previously described Poisson statistic, which estimates the significance of multi-peptide identifications incorporating the length of the matching sequence, number of spectra searched and size of the target sequence database.ResultsApplying a false discovery rate threshold of 0.05, we identified 282 significant open reading frames, each containing two or more peptide matches. There were 627 novel peptides associated with these open reading frames that mapped to a unique genomic coordinate placed within the start/stop points of previously annotated genes. These peptides matched 1,110 distinct tandem MS spectra. Peptides fell into four categories based upon where their genomic coordinates placed them relative to annotated exons within the parent gene.ConclusionThis work provides evidence for novel alternative splice variants in many previously annotated genes. These findings suggest that annotation of the genome is not yet complete and that proteomics has the potential to further add to our understanding of gene structures.

Analysis of the outer membrane proteome of Caulobacter crescentus by two-dimensional electrophoresis and mass spectrometry.

Caulobacter crescentus, a Gram negative α‐purple bacterium that displays an invariant asymmetric cell division pattern, has become a key model system for the study of bacterial development. Membrane proteins play key roles in cell cycle events, both as components of landmark morphological structures and as critical elements in regulation of the cell cycle. Recent advances for the isolation and solubilization of bacterial membrane proteins prior to isoelectric focusing have significantly improved the separation of outer membrane proteins by two‐dimensional (2‐D) electrophoresis. In this work we describe the analysis of the outer membrane proteome of Caulobacter crescentus. Proteins were identified using 2‐D gel electrophoresis and peptide mass fingerprinting by matrix‐assisted laser desorption/ionization‐time of flight mass spectrometry. We identified 54 unique proteins out of which 41 were outer membrane proteins. Of the outer membrane proteins, 16 were identified as TonB‐dependent receptor proteins. These studies were executed simultaneously with the Caulobacter genome sequencing project and advantages and limitations of proteomic analysis of a nonannotated genome are discussed. Finally, protein levels between cells grown in rich and minimal media are compared which demonstrates that many of the TonB‐dependent receptor proteins are found at higher levels in minimal medium.

Comparison of MS 2-Only, MSA, and MS 2/MS 3 methodologies for phosphopeptide identification

Current mass spectrometers provide a number of alternative methodologies for producing tandem mass spectra specifically for phosphopeptide analysis. In particular, generation of MS(3) spectra in a data-dependent manner upon detection of the neutral loss of a phosphoric acid in MS(2) spectra is a popular technique for circumventing the problem of poor phosphopeptide backbone fragmentation. The newer Multistage Activation method provides another option. Both these strategies require additional cycle time on the instrument and therefore reduce the number of spectra that can be measured in the same amount of time. Additional informatics is often required to make most efficient use of the additional information provided by these spectra as well. This work presents a comparison of several commonly used mass spectrometry methods for the study of phosphopeptide-enriched samples: an MS(2)-only method, a Multistage Activation method, and an MS(2)/MS(3) data-dependent neutral loss method. Several strategies for dealing effectively with the resulting MS(3) data in the latter approach are also presented and compared. The overall goal is to infer whether any one methodology performs significantly better than another for identifying phosphopeptides. On data presented here, the Multistage Activation methodology is demonstrated to perform optimally and does not result in significant loss of unique peptide identifications.

Improved Classification of Mass Spectrometry Database Search Results Using Newer Machine Learning Approaches

Manual analysis of mass spectrometry data is a current bottleneck in high throughput proteomics. In particular, the need to manually validate the results of mass spectrometry database searching algorithms can be prohibitively time-consuming. Development of software tools that attempt to quantify the confidence in the assignment of a protein or peptide identity to a mass spectrum is an area of active interest. We sought to extend work in this area by investigating the potential of recent machine learning algorithms to improve the accuracy of these approaches and as a flexible framework for accommodating new data features. Specifically we demonstrated the ability of boosting and random forest approaches to improve the discrimination of true hits from false positive identifications in the results of mass spectrometry database search engines compared with thresholding and other machine learning approaches. We accommodated additional attributes obtainable from database search results, including a factor addressing proton mobility. Performance was evaluated using publically available electrospray data and a new collection of MALDI data generated from purified human reference proteins.

Intra-tumor genetic heterogeneity in rectal cancer

Colorectal cancer arises in part from the cumulative effects of multiple gene lesions. Recent studies in selected cancer types have revealed significant intra-tumor genetic heterogeneity and highlighted its potential role in disease progression and resistance to therapy. We hypothesized the existence of significant intra-tumor genetic heterogeneity in rectal cancers involving variations in localized somatic mutations and copy number abnormalities. Two or three spatially disparate regions from each of six rectal tumors were dissected and subjected to the next-generation whole-exome DNA sequencing, Oncoscan SNP arrays, and targeted confirmatory sequencing and analysis. The resulting data were integrated to define subclones using SciClone. Mutant-allele tumor heterogeneity (MATH) scores, mutant allele frequency correlation, and mutation percent concordance were calculated, and copy number analysis including measurement of correlation between samples was performed. Somatic mutations profiles in individual cancers were similar to prior studies, with some variants found in previously reported significantly mutated genes and many patient-specific mutations in each tumor. Significant intra-tumor heterogeneity was identified in the spatially disparate regions of individual cancers. All tumors had some heterogeneity but the degree of heterogeneity was quite variable in the samples studied. We found that 67–97% of exonic somatic mutations were shared among all regions of an individual’s tumor. The SciClone computational method identified 2–8 shared and unshared subclones in the spatially disparate areas in each tumor. MATH scores ranged from 7 to 41. Allele frequency correlation scores ranged from R2=0.69–0.96. Measurements of correlation between samples for copy number changes varied from R2=0.74–0.93. All tumors had some heterogeneity, but the degree was highly variable in the samples studied. The occurrence of significant intra-tumor heterogeneity may allow selected tumors to have a genetic reservoir to draw from in their evolutionary response to therapy and other challenges.

Investigating MS2/MS3 Matching Statistics A Model For Coupling Consecutive Stage Mass Spectrometry Data For Increased Peptide Identification Confidence

Improvements in ion trap instrumentation have made n-dimensional mass spectrometry more practical. The overall goal of the study was to describe a model for making use of MS2 and MS3 information in mass spectrometry experiments. We present a statistical model for adjusting peptide identification probabilities based on the combined information obtained by coupling peptide assignments of consecutive MS2 and MS3 spectra. Using two data sets, a mixture of known proteins and a complex phosphopeptide-enriched sample, we demonstrate an increase in discriminating power of the adjusted probabilities compared with models using MS2 or MS3 data only. This work also addresses the overall value of generating MS3 data as compared with an MS2-only approach with a focus on the analysis of phosphopeptide data.

Global Topology Analysis of Pancreatic Zymogen Granule Membrane Proteins

The zymogen granule is the specialized organelle in pancreatic acinar cells for digestive enzyme storage and regulated secretion and is a classic model for studying secretory granule function. Our long term goal is to develop a comprehensive architectural model for zymogen granule membrane (ZGM) proteins that would direct new hypotheses for subsequent functional studies. Our initial proteomics analysis focused on identification of proteins from purified ZGM (Chen, X., Walker, A. K., Strahler, J. R., Simon, E. S., Tomanicek-Volk, S. L., Nelson, B. B., Hurley, M. C., Ernst, S. A., Williams, J. A., and Andrews, P. C. (2006) Organellar proteomics: analysis of pancreatic zymogen granule membranes. Mol. Cell. Proteomics 5, 306–312). In the current study, a new global topology analysis of ZGM proteins is described that applies isotope enrichment methods to a protease protection protocol. Our results showed that tryptic peptides of ZGM proteins were separated into two distinct clusters according to their isobaric tag for relative and absolute quantification (iTRAQ) ratios for proteinase K-treated versus control zymogen granules. The low iTRAQ ratio cluster included cytoplasm-orientated membrane and membrane-associated proteins including myosin V, vesicle-associated membrane proteins, syntaxins, and all the Rab proteins. The second cluster having unchanged ratios included predominantly luminal proteins. Because quantification is at the peptide level, this technique is also capable of mapping both cytoplasm- and lumen-orientated domains from the same transmembrane protein. To more accurately assign the topology, we developed a statistical mixture model to provide probabilities for identified peptides to be cytoplasmic or luminal based on their iTRAQ ratios. By implementing this approach to global topology analysis of ZGM proteins, we report here an experimentally constrained, comprehensive topology model of identified zymogen granule membrane proteins. This model contributes to a firm foundation for developing a higher order architecture model of the ZGM and for future functional studies of individual ZGM proteins.

Nuclear lamina genetic variants, including a truncated LAP2, in twins and siblings with nonalcoholic fatty liver disease

Nonalcoholic fatty liver disease (NAFLD) is becoming the major chronic liver disease in many countries. Its pathogenesis is multifactorial, but twin and familial studies indicate significant heritability, which is not fully explained by currently known genetic susceptibility loci. Notably, mutations in genes encoding nuclear lamina proteins, including lamins, cause lipodystrophy syndromes that include NAFLD. We hypothesized that variants in lamina‐associated proteins predispose to NAFLD and used a candidate gene‐sequencing approach to test for variants in 10 nuclear lamina‐related genes in a cohort of 37 twin and sibling pairs: 21 individuals with and 53 without NAFLD. Twelve heterozygous sequence variants were identified in four lamina‐related genes (ZMPSTE24, TMPO, SREBF1, SREBF2). The majority of NAFLD patients (>90%) had at least one variant compared to <40% of controls (P < 0.0001). When only insertions/deletions and changes in conserved residues were considered, the difference between the groups was similarly striking (>80% versus <25%; P < 0.0001). Presence of a lamina variant segregated with NAFLD independently of the PNPLA3 I148M polymorphism. Several variants were found in TMPO, which encodes the lamina‐associated polypeptide‐2 (LAP2) that has not been associated with liver disease. One of these, a frameshift insertion that generates truncated LAP2, abrogated lamin–LAP2 binding, caused LAP2 mislocalization, altered endogenous lamin distribution, increased lipid droplet accumulation after oleic acid treatment in transfected cells, and led to cytoplasmic association with the ubiquitin‐binding protein p62/SQSTM1. Conclusion: Several variants in nuclear lamina‐related genes were identified in a cohort of twins and siblings with NAFLD; one such variant, which results in a truncated LAP2 protein and a dramatic phenotype in cell culture, represents an association of TMPO/LAP2 variants with NAFLD and underscores the potential importance of the nuclear lamina in NAFLD. (Hepatology 2018;67:1710‐1725).

Lymph Node Metastases in Colon Cancer are Polyclonal

Purpose: Recent studies have highlighted the existence of subclones in tumors. Lymph nodes are generally the first location of metastasis for most solid epithelial tumors, including colorectal cancer. We sought to understand the genetic origin of lymph node metastasis in colorectal cancer by evaluating the relationship between colorectal cancer subclones present in primary tumors and lymph nodes. Experimental Design: A total of 33 samples from seven colorectal cancers, including two or three spatially disparate regions from each primary tumor and one to four matched lymph nodes for each tumor, underwent next-generation whole-exome DNA sequencing, Affymetrix OncoScan SNP arrays, and targeted deep confirmatory sequencing. We performed mapping between SNPs and copy number events from the primary tumor and matched lymph node samples, allowing us to profile heterogeneity and the mutational origin of lymph node metastases. The computational method PyClone was used to define subclones within each tumor. The method Clonality Inference in Tumors Using Phylogeny (CITUP) was subsequently used to infer phylogenetic relationships among subclones. Results: We found that there was substantial heterogeneity in mutations and copy number changes among all samples from any given patient. For each patient, the primary tumor regions and matched lymph node metastases were each polyclonal, and the clonal populations differed from one lymph node to another. In some patients, the cancer cell populations in a given lymph node originated from multiple distinct regions of a tumor. Conclusions: Our data support a model of lymph node metastatic spread in colorectal cancer whereby metastases originate from multiple waves of seeding from the primary tumor over time. Clin Cancer Res; 24(9); 2214–24. ©2017 AACR. See related commentary by Gerlinger, p. 2032

Abstract PR07: Complex sub-clonal populations in colorectal cancer lymph node metastasis

A number of studies have highlighted the role of sub-clones in tumor progression across multiple cancer types including colorectal cancer (CRC). However, the genetic origin of lymph node (LN) metastasis is not well understood. Given that the prognosis for CRC patients worsens once metastasis occurs, we aimed to better understand the relationship between detected sub-clones in the primary tumor and the LN metastasis. Two or three spatially disparate regions from each of 5 colon cancers were dissected and subjected to next-generation whole exome DNA sequencing (average read depth of 50x) followed by targeted confirmatory deep sequencing (average read depth of 420x). The number of mutations per tumor ranged from 95 to 334 in the 4 non-hypermutated tumors and was 1208 in the single hypermutated tumor. DNA was then isolated from the metastatic areas of two to four matched metastatic LN9s from each tumor. Deep-targeted sequencing was also performed on the LN samples targeting the same somatic variant locations identified in the primary tumors. Affymetrix OncoScan SNP arrays were performed on all primary tumor and LN samples to assess copy number (CN) variation and to permit more accurate inference of the cellular frequency of each somatic variant using the computational method PyClone. The PyClone software implements a probabilistic model to estimate the cellular frequency of each mutation in a population of cells using observed alternate allele frequencies as well as copy number and loss of heterozygosity (LOH) information, resulting in a clustering of variants based on the cellular frequency patterns of variants in each sample. Sub-clones are inferred corresponding to the distinct groupings of variants across all tumor and LN samples from a single patient. Additionally, the Clomial software was utilized to complement and validate PyClone results. Clomial implements an expectation-maximization algorithm to infer, given a count of distinct clones expected in a sample, profiles of the variant genotypes of each sample and sample clonal compositions. Generating Clomial profiles at several possible expected clone counts provides an orthogonal measure of clonal genotypes, as well as generating tumor purity estimates and facilitating phylogenetic profiling of variants and clones. CN and LOH differences were identified between different spatial locations in some individual tumors as well as between different LN9s from the same tumor. Some LN9s contained the same CN changes as a particular spatial region of the matched primary tumor and others contained CN variants that were distinct from the tested regions of the primary tumor. As defined by PyClone, each region in the primary tumor contained multiple (2–13, average 6.1) distinct sub-clones and each LN contained 3 - 13 sub-clones (average 5.9). We demonstrate evidence that individual LN metastasis are comprised of multiple sub-clones from the primary tumor. When there are multiple independent LN metastases in an individual patient, they do not all have the same composition of sub-clones. Some metastatic sub-clones were shared among all of the LN9s from the same tumor but there were also sub-clones that were private to a portion of the tumor and to an individual LN. This may be evidence of individual sub-clones metastasizing at different times to different LN9s. Further study of the origins of metastasis may help elucidate how to better target it therapeutically. This abstract is also being presented as Poster A10. Citation Format: Karin M. Hardiman, Peter J. Ulintz, Joel K. Greenson, Rong Wu, Eric R. Fearon. Complex sub-clonal populations in colorectal cancer lymph node metastasis. [abstract]. In: Proceedings of the AACR Special Conference on Colorectal Cancer: From Initiation to Outcomes; 2016 Sep 17-20; Tampa, FL. Philadelphia (PA): AACR; Cancer Res 2017;77(3 Suppl):Abstract nr PR07.