Integrated Cellular and Plasma Proteomics of Contrasting B-cell Cancers Reveals Common, Unique and Systemic Signatures*

Approximately 800,000 leukemia and lymphoma cases are diagnosed worldwide each year. Burkitt's lymphoma (BL) and chronic lymphocytic leukemia (CLL) are examples of contrasting B-cell cancers; BL is a highly aggressive lymphoid tumor, frequently affecting children, whereas CLL typically presents as an indolent, slow-progressing leukemia affecting the elderly. The B-cell-specific overexpression of the myc and TCL1 oncogenes in mice induce spontaneous malignancies modeling BL and CLL, respectively. Quantitative mass spectrometry proteomics and isobaric labeling were employed to examine the biology underpinning contrasting Eμ-myc and Eμ-TCL1 B-cell tumors. Additionally, the plasma proteome was evaluated using subproteome enrichment to interrogate biomarker emergence and the systemic effects of tumor burden. Over 10,000 proteins were identified (q<0.01) of which 8270 cellular and 2095 plasma proteins were quantitatively profiled. A common B-cell tumor signature of 695 overexpressed proteins highlighted ribosome biogenesis, cell-cycle promotion and chromosome segregation. Eμ-myc tumors overexpressed several methylating enzymes and underexpressed many cytoskeletal components. Eμ-TCL1 tumors specifically overexpressed ER stress response proteins and signaling components in addition to both subunits of the interleukin-5 (IL5) receptor. IL5 treatment promoted Eμ-TCL1 tumor proliferation, suggesting an amplification of IL5-induced AKT signaling by TCL1. Tumor plasma contained a substantial tumor lysis signature, most prominent in Eμ-myc plasma, whereas Eμ-TCL1 plasma contained signatures of immune-response, inflammation and microenvironment interactions, with putative biomarkers in early-stage cancer. These findings provide a detailed characterization of contrasting B-cell tumor models, identifying common and specific tumor mechanisms. Integrated plasma proteomics allowed the dissection of a systemic response and a tumor lysis signature present in early- and late-stage cancers, respectively. Overall, this study suggests common B-cell cancer signatures exist and illustrates the potential of the further evaluation of B-cell cancer subtypes by integrative proteomics.

Burkitt's lymphoma (BL) 1 and chronic lymphocytic leukemia (CLL) represent the extremes of B cell cancers; BL is highly aggressive and frequently affects children, whereas CLL typically presents as an indolent, slow-progressing leukemia in the elderly (1,2).
BL is a hallmark myc-driven tumor; induced by chromosomal translocation of the transcription factor myc to immunoglobulin (Ig) enhancers (3). Myc influences ϳ15% of the genome, regulating processes such as cell proliferation, metabolism, adhesion, angiogenesis and de-differentiation (4 -6). Such neoplastic-like traits make myc an aggressive oncogene, dysregulated or overexpressed in most cancers (7). Formal proof of its oncogenic properties were demonstrated when human myc was placed into the Ig heavy chain enhancer (E) region of the mouse. These 'E-myc' mice produced high-penetrance lymphomas within around 100 days, recapitulating several molecular and pathological aspects of BL (3, 8 -10).
In contrast, CLL is a relatively indolent B-cell cancer, presenting with a CD5 ϩ CD19 ϩ leukemia (11,12). Up to 90% of CLL cases express TCL1-a protein involved in lymphocyte development (13)(14)(15). TCL1 is suggested to promote cell survival and proliferation by amplification of AKT phosphorylation, induced by growth factor-, cytokine-and B-cell receptorinduced PI3K signaling (16,17). The E-TCL1 mouse, overexpressing TCL1 in B-cells again through the enhancer, was developed as a potential model of CLL (18). E-TCL1 mice recapitulate the expanded bone marrow, splenic and circulatory CD5 ϩ B-cell populations of CLL, detectable from around 3-5 months of age. The disease progresses to lethal splenomegaly and leukemia at ϳ12 months (18,19). Currently it is regarded as one of the most useful preclinical models of CLL (20,21).
Together, these models provide the opportunity to study oncogenesis mediated by two contrasting oncogenes at opposing ends of the proliferative spectrum. E-myc tumors, like BL, are highly aggressive, have rapid onset and form lymphoid tumors, whereas E-TCL1 mice, like CLL, present a relatively indolent and slow-developing leukemia with secondary lymphoid organ involvement (3,18).
In addition to cellular characterizations of tumors, plasma analysis holds the potential to identify additional biological signatures arising during oncogenesis. In particular, extracellular fluids can provide insight into the tumor-host dialogue between the immune system and micro-environment and report on the metabolic and homeostatic aberrations that tumors display. Further, combining cellular and plasma characterization can provide greater insight into the mechanisms by which any biomarkers of disease may be entering the circulation.
Liquid chromatography coupled with mass spectrometry (LC-MS) proteomics currently provides the best means of establishing global differential protein expression profiles of cellular and plasma samples. These approaches are evolving rapidly and yielding ever-increasing proteome coverage (22,23). Nonetheless plasma/serum proteomics still presents significant challenges because of the vast dynamic range of protein concentrations (24). Although advances continue to be made with immunodepletion strategies (25)(26)(27), an alternative method termed, Subproteome Enrichment by Size Exclusion Chromatography (SuPrE-SEC), offers an effective alternative. SuPrE-SEC uses size-dependent protein fractionation to deplete high-abundance proteins and enrich for lowabundance proteins. The resulting reduction of the protein concentration dynamic range facilitates a greater depth of LC-MS proteomics coverage (28,29).
This investigation has applied multiplexed, LC-MS proteomics to the characterization of plasma subproteomes and isolated B-cell material from E-myc, E-TCL1 and wildtype (WT) mice. The resulting proteomes have been interrogated to identify common and tumor-specific signatures and to understand the combined cellular and extracellular characteristics of B-cell tumors.
Dulbecco's modified Eagle's medium (DMEM), glutamine, pyruvate, penicillin and streptomycin were purchased from Life Technologies, Carlsbad, CA. Mouse B-cell isolation kits and magnetic cell sorting columns were purchased from Miltenyi Biotech, Bergisch Gladbach, Germany. Twentythree-gauge needles and 1-ml syringes were purchased from BD Biosciences, Franklin Lakes, NJ, Interleukin 5 (IL5) recombinant protein from Peprotech, Rocky Hill, NJ, Immobilon 0.45 m pore PVDF membrane from Millipore, Billerica, MA and nonfat milk from Marvel, Birmingham, UK. 30ϫ DTT and 3ϫ red loading buffer were purchased for Cell Signaling Technologies, Danvers, MA. Enhanced chemiluminescence reagents, tandem mass tags (TMT) 10-plex isobaric labeling reagents and 2 kDa M w cut-off Slide-A-Lyzer dialysis cassettes and proteomics grade lys-c were purchased from Thermo Scientific, Waltham, MA. Proteomics grade trypsin was purchased from Roche, Basel, Switzerland. Isobaric tags for relative and absolute quantitation (iTRAQ) 8-plex reagents were purchased from ABSciex, Framingham, MA. Antibodies are detailed in Supplementary Methods.
Experimental Design and Statistical Rationale-Sample pooling, detailed in supplemental Figs. S1, was used to accommodate comparative experiments in single isobaric label sets. Tumorous spleens were collected from 4 E-myc and 4 E-TCL1 mice with terminal tumor presentation. For nontumor controls, 6 spleens, handled as two pools of 3 spleens, were collected from 6 week old and 200 day WT littermates and 6 week-old E-myc and E-TCL1 mice with no signs of tumor or splenomegaly (summarized in supplemental Fig. S1). Nontumor pools were analyzed with six samples per isobaric label, whereas tumor samples were analyzed as biological replicate pools of two tumors.
For plasma, six samples were individually collected for each of the above tumor and nontumor conditions, in parallel with B-cell isolation. In addition, 6 samples were isolated from E-TCL1 mice at a preterminal "30%" stage, when 30% B220 ϩ CD5 ϩ cells were present in the blood. To minimize bias these were selected from a cohort of 14 E-TCL1 mice, from every other mouse reaching the 30% threshold. Tumor and preterminal "30%" tumor plasma were also analyzed as biological replicates, with three samples allocated to each label.
Each pair of tumor pools was compared with two WT sample pools, generating four ratios which could be evaluated for consistency and significance to ensure the reproducibility of quantitative results. Replicate ratios were analyzed for consistency with an FDR-corrected t test and considered significant with a p value of Ͻ0.05. p values were coupled with a measure of magnitude detailed in Quantitative and Statistical Analysis of MS Data, below.
B-cell Isolation-B-cells were isolated from single-cell suspensions using the mouse negative selection B-cell isolation kits and MACS columns according to the manufacturer's instructions, washed once at room temperature (RT) in red blood cell lysis buffer (155 mM NH 4 CL, 10 mM NaHCO 3 , 0.1 mM Na 2 EDTA (pH 7.4) before three washes in PBS. B-cell isolation efficiency was assessed by flow cytometry immunostaining with CD19 and CD3, typically yielding Ͼ90% purity with less than 2% T-cell (CD3 ϩ ) contamination (supplemental Fig. S1). B-cell pellets were snap frozen and stored in liquid nitrogen prior to lysate preparation.
Plasma Isolation-To isolate plasma, maximize sample purity, minimize red blood cell lysis and accommodate for organ displacement from tumors, an adaption of bleeding from the inferior vena cava under terminal anesthesia (30) was utilized. Using a 23-gauge needle, 700 -1000 l of blood was collected for each animal into 50 g/ml sodium heparin in PBS. Blood was immediately placed on ice and centrifuged at 2000 ϫ g for 15 min at 4°C with plasma stored in liquid nitrogen. Samples were rejected if red blood cell lysis was visible by eye.
Plasma Subproteome Enrichment-20 l was taken from each of the six replicate plasma samples described above to give pools of 120 l. For tumor samples, two 120 l pools of 3 ϫ 40 l were formed to provide biological replicates (summarized in Figs. 1 and supplemental Fig. S1).
SuPrE-SEC was adapted from that described previously (28). Each 120 l plasma pool was diluted with 380 l of 6 M guanidine hydrochloride in 10% methanol and separated by 3 KW804 SEC columns in series at 1.2 ml/min and 30°C. The low molecular weight subproteome was isolated during elution from 42-55 min (Supplemental Fig. S1G) and dialyzed (2 kDa M w cut-off) into ultrapure water (18.2 M⍀ cm Ϫ1 ) with five exchanges into 5 liters. Protein was lyophilized and re-solubilized in 0.5 M TEAB with 0.05% SDS with 30 g digested and labeled for MS analysis as described below.
Sample Preparation for MS-Snap frozen cell pellets were lysed on ice by trituration with a 23-gauge needle in 0.5 M TEAB with 0.05% SDS. Disrupted cells were further sonicated and lysates cleared at 16,000 ϫ g for 10 min at 4°C. 100 g of cell lysate or 30 g of plasma subproteome was reduced with 50 mM TCEP and alkylated with 200 mM MMTS, before digestion overnight at RT with a 30:1 ratio of proteomics grade trypsin. Plasma proteins were additionally digested for a further 2 h at 37°C with 100:1 proteomics grade Lys-c. Peptides were incubated with either iTRAQ 8-plex (B-cell peptides) or TMT 10-plex isobaric tags (plasma peptides) according to the manufacturer's instructions. Samples were then lyophilized and labeled peptides serially reconstituted, for each proteome, in 100 l of 2% v/v ACN, 0.1% v/v NH 4 OH.
Peptide Fraction Resolution and Characterization by LC-MS/MS-Lyophilized peptide fractions were individually reconstituted in 2% ACN, 0.1% FA, and ϳ500 ng of peptides were subjected to nano-ultra-performance liquid chromatography by a Dionex Ultimate 3000 (Thermo Scientific). Conditions used varied between the plasma and B-cell proteomes, detailed in Supplementary Methods. In summary, peptides were trapped by C18 and eluted over a reverse phase gradient, of which several lengths were used depending on the peptide fraction abundance and proteome. The total MS time for the iTRAQ-labeled B-cell proteome was ϳ200 h, with the TMT-labeled plasma proteome analyzed over ϳ250 h.
Peptide elution was directly coupled to electrospray ionization at 2.4 kV using a PicoTip nESI emitter (New Objective, Woburn, Massachusetts), and were characterized with an Orbitrap Elite Velos Pro mass spectrometer (Thermo Scientific). MS characterization of eluting peptides was conducted between 350 and 1900 m/z at 120,000 mass resolution. The top 12 ϩ 2 and ϩ3 precursor ions per MS scan (minimum intensity 1000) were characterized by tandem MS with highenergy collisional dissociation (HCD) (30,000 mass resolution (15,000 for the B-cell proteome), 1.2 Da isolation window, 40 keV normalized collision energy) and CID (ion trap MS, 2 Da isolation window, 35 keV). Additionally, the DMSO ion at 401.922718 was used as a MS lockmass (31).
MS Data Processing-Target-decoy searching of raw spectra data was performed with Proteome Discoverer software version 1.4.1.14 (Thermo Scientific). Spectra were subject to a two stage search, both using SequestHT (version 1.1.1.11), with Percolator used to estimate FDR with a threshold of q Յ 0.01. The first allowed only a single missed cleavage, minimum peptide length of 7, precursor mass tolerance of 5 ppm, no variable modifications and searched against the mouse UniProt Swissprot database, supplemented with the human sequences for (downloaded 01/15, 16,676 protein sequences). The second search used only spectra with q Ͼ 0.01 from the first search, allowed two missed cleavages, minimum peptide length of 6, searched against the human/mouse Uni-Prot trembl database (with human myc and TCL1, downloaded 01/15, 52,469 protein sequences), precursor mass tolerance of 10 ppm and a maximum of two variable (1 equal) modifications of; TMT or iTRAQ (Tyr), oxidation (Met), deamidation (Asn, Gln) or phospho (Ser, Thr, Tyr). In both searches, fragment ion mass tolerances of 0.02 Da and 0.5 Da were used for HCD and CID spectra, respectively. Fixed modifications of Methythio (Cys), TMT or iTRAQ (Lys and N terminus) were used. PhosphoRS was used to predict the probability of specific phosphorylated residues. Reporter ion intensities were extracted from nonredundant PSMs with a tolerance of 20 ppm. To reduce ratio compression, peptide spectrum match data for proteins (qϽ0.01) were exported from Proteome Discoverer and submitted to Statistical Processing for Isobaric Quantitation Evaluation (SPIQuE) at spiquetool-.com. This method weighted the contributions of each PSM quantitation to a protein's quantitation on the basis of PSM features (manuscript in preparation). For example, high-intensity peptides with low isolation interference were given a greater weighting factor. Although the effects upon the ratios were minimal, the overall trend demonstrated efficient ratio decompression (Supplementary Methods). An example of the effect on the ratio decompression to the human TCL1 protein, discretely expressed in E-TCL1 B-cells, is outlined in Supplementary Methods.
The raw data and processed outputs have been deposited to the ProteomeXchange Consortium (32) via the PRIDE partner repository with the data set identifier PXD004608.
Quantitative and Statistical Analysis of MS Data-Log 2 (ratios) were generated describing each sample pool relative to the two WT controls. To define those proteins with the great-est fold change for each biological state, an FDR-corrected one sample t test was performed and average values determined. To derive a single, robust measure, representative of both the magnitude and consistency of differential expression, a ratio was defined between that of the mean and the standard deviation of the 4 log 2 ratios, termed the regulation score (Rs ϭ mean/(standard deviation ϩ1)). For proteins with no, or inconsistent differential expression, the Rs tends toward 0. The Rs also correlated more strongly with p values for each set of ratios (Supplementary Methods). In all instances, differentially expressed proteins (DEPs) were defined as those with an FDR-corrected p value of Ͻ0.05 and an Rs threshold of Ͼ0.5 or Ͻ-0.5. Although arbitrary, evaluation of this threshold demonstrated a minimum average fold change of 1.4, for which a fold change variation of just 6% was observed. This therefore fulfilled the power analysis principles described by Levin, 2011 (33), but on a case-by-case basis, by incorporating variation into each Rs via the S.D. as a denominator. The outliers for this threshold were additionally highlighted to demonstrate that consistent differential expression was still observed (Supplementary Methods). Other p values were determined by a false discovery rate-corrected 1-or 2-sample, two tailed, t test with no assumption of equal variance.
Bioinformatics Analyses-Proteins reaching the thresholds outlined above were submitted to either Ingenuity Pathway Analysis (IPA) or Database for Annotation, Visualization and Integrated Discovery (DAVID). For DAVID analyses, the default settings were used for pathway and gene ontology (GO) term enrichment, with Benjamini-corrected p values of Ͻ0.05 considered significant. For B-cell/tumor proteome enrichments, all quantified proteins were used as a background. For IPA analyses, default settings were used. Upstream regulator analysis was determined using all DEPs (RsϾ0.5/Ͻ-0.5, p Ͻ 0.05). Annotations of biomarkers and drug targets were conducted by IPA. Plasma protein deconvolution was conducted by removing any proteins annotated by IPA as predominantly cellular in addition to any proteins with more than 30 PSMs from the B-cell proteome.
SDS-PAGE and Western blotting-Cells were lysed in radioimmunoprecipitation assay (RIPA) cell lysis buffer (0.15 M NaCl, 1% v/v octylphenoxypolyethoxyethanol (IgePalCA630), 0.5% w/v sodium deoxycholate (DOC), 0.1% w/v sodium dodecyl sulfate (SDS), 0.05 M TRIS (pH 8) and 1% v/v protease inhibitor) and as described previously (34). Lysate supernatants were isolated by centrifugation (16,000 ϫ g for 10 min at 4°C). Protein concentrations were determined by Bradford assay. Lysates were reduced in loading buffer with 2-mercaptoethanol at 95°C for 5 min. Lysates were resolved by SDS-PAGE and transferred to PVDF. Membranes were blocked in 5% (w/v) nonfat milk for 1 h before probing with primary antibodies, as detailed in Supplementary Methods. Expression was detected by incubation with a horseradish peroxidase-conjugated secondary antibody using enhanced chemiluminescence reagents and a ChemiDoc-It imaging system (UVP). GAPDH or tubulin were used as loading controls. Relative quantification of protein band intensity was determined using Image J, normalized against loading controls and ratios to WT B-cell lysates were derived. For E-myc tumor validations, lysate protein concentrations were determined by bicinchoninic acid assay (BCA), reduced with DDT and diluted in 3x loading buffer. Western blots were visualized with fluorescent antibodies and an Odyssey Imaging System (Li-cor) and quantified using Image Studio 2.0 (Li-cor).
Cell Culture-Cells derived from E-myc or E-TCL1 tumors were cultured in DMEM supplemented with 2 mM glutamine, 1 mM pyruvate, 45 units/ml penicillin, 45 l/ml streptomycin, 200 M asparagine, 50 M 2-mercaptoethanol, and 10% FCS. E-myc tumors were cultured a density of 5 ϫ 10 7 cells/ml at 37°C in 10% CO 2 . E-TCL1 cells were cultured at a density of 5 ϫ 10 6 cells/ml at 37°C in 5% CO 2 . Cells were cultured for 24 h prior to treatment. Serum starvation was conducted for 4 h in the above media with 0.5% BSA replacing FCS.
Flow Cytometry-Cells were stained with either the manufacturer's recommended concentration, or 10 g/ml, of antibody (Supplementary Methods) for 30 min in the dark, washed and analyzed by flow cytometry with a FACScan or FACScalibur (BD) (35,36). Relative expression was determined using the geometric means. Cell cycle status was assessed by hypotonic PI (50 g/ml PI, 0.1% w/v sodium citrate, 0.1% w/v triton x-100) incubated at 4°C for 15 min. PI fluorescence was measured by FL2 on a linear scale at the lowest flow rate. Cell division was tracked by CFSE dilution of cells stained with a concentration of 5 M for 15 min at room temperature.

Quantitative Proteomics of E-myc and E-TCL1 Tumors
and Plasma-To characterize the global proteome expression of E-myc and E-TCL1 tumors, an 8-plex iTRAQ experiment was designed incorporating terminal tumor, premalignant and age-matched WT B-cells ( Fig. 1). Sample pooling in combination with biological replicates enabled the averaging of biological variability of several samples within a single isobaric tag experiment. Splenic tumors samples, because of a greater anticipated variability, were pooled in two pairs of two tumors from each model; derived from a total of 8 terminal mice (supplemental Fig. S1). Nontumor controls consisted of samples pooled from six animals. For plasma proteomics, controls were again characterized as pools of six samples, with two pools of three plasma samples for each tumor condition. Additionally, plasma from preterminal E-TCL1 tumors was analyzed, derived from mice reaching a CD5 ϩ B220 ϩ leukemia threshold of 30%, termed "30%" samples. To accommodate these additional samples, TMT 10-plex isobaric labels were used.
The tumors selected for proteomics analysis, detailed in supplemental Fig. S1A-S1F, were representative of the range of characteristics observed for the model. The median terminal presentations, for example (supplemental Fig. S1A), of 92 and 318 days for E-myc and E-TCL1 tumors, respectively, agreed with earlier reports (3,8,18,19).
Splenic B-cells from tumors and controls were lysed, trypsin digested, and assigned to the eight isobaric labels of iTRAQ 8-plex ( Fig. 1) to provide relative protein expression quantitation. The labeled B-cell peptides were pooled and characterized by two-dimensional (2D) LC-MS identifying 9260 proteins (qϽ0.01). 8270 proteins were relatively quantitated across all eight sample pools. The depth of proteome coverage and quantitation is summarized in supplemental Fig.  S1H.
Plasma samples were subjected to an adapted form of SuPrE-SEC (28), reducing the dynamic range and facilitating deeper proteome coverage by the exclusion of the majority of high-abundant proteins (supplemental Table S1, supplemental Fig. S1G). 2D LC-MS characterization of the plasma subproteome provided relative quantitation for 2095 of the 2688 identified proteins (qϽ0.01). Peptide spectrum matches (PSMs) to the higher-M w protein albumin (n ϭ 3835, M w ϭ 68.7kDa), for example, were of far lower abundance than transthyretin (n ϭ 7769, M w ϭ 15.8kDa) or apolipoprotein A-II (n ϭ 7129, M w ϭ 11.3kDa). Given approximate plasma concentrations of 40 mg/ml, 260 g/ml and 300 g/ml for these proteins (37), respectively, this represented around a 275-fold enrichment of low M w proteins.
Reproducibility and Validation of Quantitative Proteomics Results-The relative quantitative proteomes demonstrated tumor-dependent hierarchical clustering identifying common trends of differential tumor expression compared with both WT controls ( Fig. 2A). Evaluation of the reproducibility between tumor sample pools by linear regression highlighted the strong correlation between the 5 pairs of independent biological replicates (Fig. 2B). Additionally, an overlap of 1288 proteins were quantitated in both the B-cell and plasma proteomes.
Individual protein expression was considered for a number of candidates to confirm the successful characterization and analysis of the quantitative proteome. The results presented an anticipated model-specific overexpression of the myc and TCL1 human transgenes (Fig. 2C, supplemental Fig. S2A, S2B), in addition to characteristic CD5 overexpression and B220 underexpression for E-TCL1 tumors. Several DEPs were validated by antibody-based methods to further verify the relative quantitative accuracy of the proteomics; 10 for E-myc tumors and 7 for E-TCL1 tumors (supplemental Fig.  S2G, S2H). The log 2 ratios to WT of these validations were plotted alongside the quantitative proteomics results (Fig. 2D). Additionally, previously reported DEPs in E-myc or E-TCL1 tumors were compared with the proteomics results (supplemental Fig. S2D, S2F). A correlation was also observed with a study describing relative mRNA expression in E-myc tumors (38) (supplemental Fig. S2J). Overall, the quantitative pro-teomics correctly identified over or underexpression for the vast majority of the 38 DEPs observed by antibody-based quantitations. This comparison also highlighted reporter ion ratio compression, a commonly reported feature of isobaric tag quantitation (39,40). Anticipated plasma proteins, such as IgM overabundance in E-TCL1 tumors, was also observed (supplemental Fig. S2I). These results therefore supported the overall reliability of the DEP observations and the characterization and analysis by which they were generated.
Proteomic Characterization of B-cell Tumor Phenotypes-Proteins were considered DEPs when both tumor pools demonstrated a clear, consistent over-or underexpression to both 6-week and 200-day WT samples. Rather than an average of the four derived ratios for each tumor, potentially misrepre-  FIG. 1. E-myc and E-TCL1 model proteomic characterization workflow. B-cells and plasma were isolated from splenocytes and blood, respectively, derived from terminal and preterminal E-myc and E-TCL1 mice and WT controls (supplemental Fig. S1C). Pooling was used to accommodate all samples within single isobaric-labelled experiments for B-cell (iTRAQ 8-plex) and plasma (TMT 10-plex) proteomics. Pools of 2, 3, or 6 samples were used for tumors, tumor plasma and pretumor controls, respectively. For all tumor samples, biological replicate pools were analysed. Plasma pools, totalling 120 l per pool, were subjected to size exclusion chromatography (SEC) to isolate the low molecular weight subproteome (supplemental Fig. S1G). B-cells were isolated by negative selection and were lysed, quantified and 100 g of protein pooled. Plasma and B-cell proteins were subjected to reducing conditions, cysteine alkylation, trypsin proteolysis and isobaric labelling and pooling of the generated peptides. The labeled peptides for each proteome were resolved by two-dimensional liquid chromatography and quantitatively characterized by mass spectrometry (MS). The MS data were then subject to target-decoy analysis and reporter ion quantitation to generate two quantitative proteomes.

A.
Characteristic model expression sentative of variable findings, a ratio of average to standard deviation (mean/(S.D.ϩ1)) was calculated, termed the regulation score (Rs). This provided a single value expressive of both magnitude and consistency for either, or both, tumors relative to WT B-cells. In combination with FDR-corrected p values, the Rs allowed careful selection of the most confidently and consistently DEPs (detailed in Supplementary Methods).
The range of significantly DEPs for each tumor were represented by volcano plot, comparing Rs to -log 10 (p values) (Fig.   3A, supplemental Fig. S3A). This analysis illustrated the broad extent of DEP signatures observed in each tumor, with ϳ3000 and 1500 DEPs in E-myc and E-TCL1 tumors, respectively. The 6-week E-myc and E-TCL1 B-cells exhibited signatures broadly indistinguishable from terminal E-myc tumors and WT B-cells, respectively (supplemental Fig. S3B). The 6-week E-myc B-cell signature did however, present a lesser extent of differential expression. When compared with one another, a common signature became apparent between E-   myc and E-TCL1 tumors (Fig. 3B) demonstrating that many of the E-myc tumor-overexpressed proteins were also overexpressed in E-TCL1 tumors, but to a lesser magnitude. Approximately 700 and 200 proteins were considered overand underexpressed in the tumors of both models, respectively (Fig. 3C). Examples of the 10 most consistently DEPs from this common signature were presented (Fig. 3D); and annotated as predicted cell surface markers with potential as immunotherapy targets; drug targets, annotated by IPA; and novel DEPs with no previous published links with cancer (detailed further in supplemental Fig. S3C-S3G). Among the DEPs common to both tumors were, for instance signatures of cell cycle upregulation, such as the three kinesin proteins, KIF11, KIF20A, and KIF23 and an overall trend of cell surface protein underexpression such as CD23, CR2, CD200, CD40, IgG receptor FCGRT, CD38, CD22, and IL21R. The most consistently overexpressed surface proteins, HMMR/ CD168, has previously been associated with B-cell cancers (41,42).
Bioinformatics Reveals Signatures Common to Contrasting B-cell Tumors-Although individual DEPs can suggest functional insight, approaches simultaneously considering all DEPs, potentially offer a broader understanding of biological mechanisms. Accordingly, topological, proteome-wide expression patterns were investigated by bioinformatics.
To visualize these processes and highlight the proteins contributing to the enriched GO terms, networks were displayed descriptive of individual proteins and their interrelationships, defined by STRING (43). Fig. 4B demonstrates a network generated from proteins annotated with "chromosome segregation" highlighting the scale and complexity of the dysregulation of this process seen in both tumors. "Ribosome biogenesis" was also illustrated in this way (Fig.  4C), highlighting those overexpressed proteins responsible for the term's enrichment. A cluster of interacting ribosome proteins was identified, alongside several assembly regulators. Several canonical pathways, including ribosome-, translation-, and cell cycle-related pathways demonstrated significant enrichment in both tumors (supplemental Fig.   S4A). A network detailing all underexpressed proteins revealed a highly interrelated series of interactions highlighting major histocompatibility complex proteins, interferon response proteins and B-cell-related signaling molecules (Fig. 4D).
To simultaneously analyze both proteome-wide over-and underexpression, upstream regulators were evaluated using IPA. Upstream regulator activity was inferred by the comparison of anticipated downstream expression profiles of proteins with those expression profiles derived by quantitative proteomics for both tumors (supplemental Table S3, supplemental Fig. S4B-S4E). Protein functionality was then inferred from a combination of the resulting upstream regulator activation z-scores-a value proportional to overall predicted activation (44)-and the proteomics-determined differential expression (Fig. 4E). Several regulators not quantitated by proteomics, including miRNA were also inferred to be activated or inactivated (supplemental Fig. S4C, S4E).
It was additionally possible to compare differential protein phosphorylation to infer activation or inhibition (supplemental Fig. S4F-S4H). HDAC2 overexpression in E-myc tumors, for instance, was accompanied by a signature of downstream inactivation (supplemental Fig. S4B) and a decreased phosphorylation of Ser 394 (supplemental Fig. S4H), a modification previously associated with the activation of HDAC2 (45). Similarly, in E-TCL1 tumors, RBL1 overexpression was accompanied by apparent inactivation and decreased Thr 385 phosphorylation. Differential phosphorylation of BCR signaling pathway components demonstrated an increased phosphorylation of downstream proteins such as NF-B, OCT2 and ETS1 (supplemental Fig. S4J).
Characteristics Specific to E-myc B-cell Tumors-Although the common tumor signature in both models was strong, several tumor-specific differences were also apparent, particularly in the more aggressive E-myc tumors. To evaluate tumor-specific expression, proteins were filtered to include those which exhibited significant over-or underexpression in one tumor model (RsϾ0.5/Ͻ-0.5, p Ͻ 0.05) but not the other (RsϾ0.25/Ͼ-0.25). This identified 572 and 537 proteins with specific over-or underexpression in E-myc tumors, respectively. The 10 most over-and underexpressed proteins

Activated regulators No/inconsistent activity Inhibited regulators
Upstream regulator enrichment A. E.

B.
C. in addition to cell surface proteins were illustrated to highlight examples of this specific differential expression pattern in E-myc tumors (Fig. 5A).

D.
The over-and underexpressed proteins specific to E-myc tumors were evaluated, as for Fig. 4A, for GO term enrichment (Fig. 5B, supplemental Fig. S3). The majority of GO terms Eμ-myc-specific underexpressed proteins Additionally, the top 10 proteins with specific cell surface expression are shown. B, GO term enrichment analysis for differential expression specific to E-myc tumors (as described for Fig.  4). C-D, StringDB networks derived from E-myc-specific differentially expressed proteins annotated with C, "methylation" (GO:0032259) (n ϭ 50), and D, "actin cytoskeleton organization" (GO:0030036) (n ϭ 49). E, Western blot evaluation of the three "actin cytoskeleton organization"annotated proteins, coronin 1a (COR1A), lymphocyte cytosolic protein 1 (LCP1) and actin regulatory protein CAPG, expression in E-myc tumors, relative to nontumor B-cells. Two addition cytoskeletal proteins, myosin-9 (MYH9) and moesin (MOE) are also evaluated alongside myc. observed for both tumors were also observed for the E-mycspecific proteins, further highlighting trends of increased proliferation and growth of cells. Exceptions included trends of epigenetic-related processes and underexpressed cytoskeletal processes.
Characteristics Specific to E-TCL1 B-cell Tumors-Protein expression specific to E-TCL1 tumors was also evaluated, with the top 10 DEPs and membrane proteins presented (Fig. 6A). As validation this illustrated the discrete expression of human TCL1 (supplemental Fig. S2A) in E-TCL1-derived B-cell samples. GO term enrichment identified upregulated processes relating to intracellular compartments, such as ER stress, vesicular transport and glycosylation. Extracellular interactions, including signaling, locomotion and adhesion appeared as the strongest underexpressed process in E-TCL1 tumors (Fig. 6B). A network formed from E-TCL1-specific overexpressed proteins (Fig. 6C) presented clusters of associated proteins contributing to enriched GO terms, such as ER and golgi proteins. Several signaling proteins were also specifically overexpressed, such as phosphatidylinositol kinases and phosphatases. Associated with these signaling proteins were the alpha and beta subunits of the interleukin 5 receptor (IL5R␣ and IL5R␤), also apparent among the top 10 membrane proteins, highlighted in Fig. 6A.

Interleukin 5 Receptor is Overexpressed by E-TCL1 B-cell Tumors-
The overexpression of the IL5R was first evaluated by plotting the individual iTRAQ ratios for each unique peptide matching IL5R␣ and IL5R␤ (Fig. 6D, supplemental Fig. S5A-S5E). For both IL5R subunits a clear, specific overexpression was apparent. Furthermore, the signature for IL5 activity was apparent through various bioinformatics analyses, including IPA regulator analyses (supplemental Fig. S4C, S5F). A signature additionally became apparent in E-TCL1 terminal and preterminal plasma indicative of an IL5-induced proliferation signature, as well as plasma signatures of quantitated proteins and inferred regulators potentially upstream of IL5 induction (supplemental Fig. S5G-S5I).
Given the strength of this evidence, the surface expression of IL5R␣, the subunit specific to IL5 recognition, was evaluated by flow cytometry for terminal and pretumor E-TCL1 B-cells, relative to WT B-cells (Fig. 6E, supplemental S5J-S5K). Although IL5R␣ expression was observed across a wide distribution for WT B-cells, for pretumor E-TCL1 B-cells, the emerging CD5 ϩ population was observed with high IL5R␣ expression. The E-TCL1 tumor cells exhibited the same CD5 ϩ IL5R␣ ϩ expression as the emerging pretumor E-TCL1 B-cell population.

Interleukin 5 Drives Proliferation in E-TCL1 B-cell Tumors via AKT-
The observation of substantial overexpression of both IL5R␣ and IL5R␤, strongly suggested a functional role for the IL5R and IL5 in driving E-TCL1 tumors. To investigate and functionally validate the IL5R overexpression, terminal tumors from E-TCL1 mice were treated with 0, 10 and 100 ng/ml of IL5 for 48 h in vitro. First, cell density indicated a dose-dependent expansion induced by IL5 (Fig. 7A). CFSE labeling confirmed that cells were proliferating in a dose-dependent manner with as many as 70% of cells being postmitotic after 48 h of 100 ng/ml IL5 treatment (Fig. 7B). Furthermore, evaluation of cell cycle phases by hypotonic PI staining demonstrated that almost three times as many cells (14.5%) were in S/G 2 /M phase after 100 ng/ml IL5 treatment, compared with no treatment (4.9%) (Fig. 7C).
Given the proposed role of TCL1 in amplifying AKT signaling (16,17), AKT activation by IL5 was investigated (Fig. 7D). Serum starvation and the subsequent addition of IL5 induced AKT phosphorylation in a dose-dependent manner at doses of greater than 5 ng/ml. Downstream signaling (summarized in Fig. 7E) was also apparent, with dose-dependent phosphorylation of the S6 ribosomal protein and S6 kinase, indicative of mTOR activation.
Plasma proteomics reveals signatures of tumor lysis and immune response-To investigate the overall plasma proteome signature, those proteins quantified commonly or discretely within the B-cell and plasma proteomes (supplemental Table S2) were dissected based on canonical protein localizations (Fig. 8A). The vast majority (82%) of the 1288 proteins observed as common to both proteomes, were annotated with a canonical cellular localization, compared with just 40% of proteins identified discretely in the plasma proteome. Furthermore, evaluation of the relative abundances in terminal tumor plasma revealed a striking signature of overabundant cell-derived proteins (Fig. 8B). The tumor origin of this plasma

-4 4
Fold change from WT FIG. 6. E-TCL1-specific tumor characteristics. A, Heat maps for proteins and surface proteins specific to E-TCL1 tumors (Rs E-TCL1 Ͼ0.5/Ͻ-0.5, pϽ0.05, Rs E-myc Ͻ0.25/Ͼ-0.25). B, GO term enrichment analysis for differential expression specific to E-TCL1 tumors. C, A StringDB network derived from E-TCL1-specific differentially expressed proteins demonstrating interaction scores of Ͼ0.7 highlighting processes in protein clusters. (ER; endoplasmic reticulum). D, Individual iTRAQ quantifications (as log 2 (ratios to WT)) for the peptides uniquely matching to IL5RA and it receptor partner CSF2RB (IL5RB). E, Flow cytometry evaluation of IL5 receptor alpha subunit (IL5R␣) expression on an E-TCL1 tumor (representative of four evaluated tumors (supplemental Fig. S5K)), compared to splenic B cells pooled and isolated from three 2-month-old WT and E-TCL1 mice. signature was additionally indicated by GO term enrichment (Fig. 8C), with highly similar trends to that of the B-cell tumors (Fig. 4A), and an inter-proteome correlation between the approximate protein abundances (supplemental Fig. S6A). Additionally proteins were illustrated with biomarker applications (supplemental Fig. S6H), specific tumor plasma signatures (supplemental Fig. S6J-S6P), differential abundance relative to approximate relative plasma concentrations (supplemental Fig. S6Q), over-and underabundance in both the tumor and plasma proteomes (supplemental Fig. S6R) and with potential as biomarkers derived from functional tumor regulators (supplemental Fig. S6S-S6T).
To explore differential protein abundance resulting from the extrinsic response to the tumor, the plasma proteome was subject to deconvolution, removing proteins with a clear cellular origin. The remaining proteins, termed the "lysis-free" plasma proteome (supplemental Table S2, supplemental Fig. S6F), provided a signature descriptive of an immune response. Terms including "defense response" (n ϭ 16, p ϭ 4.0 ϫ 10 Ϫ5 ) and "extrinsic apoptotic signaling   pathway via death domain receptors" (n ϭ 6, p ϭ 1.4 ϫ 10 Ϫ4 ) were identified as enriched among these overabundant "lysis-free" plasma proteins (Fig. 8C). To detail this lysis-free signature further, a network was generated from all overabundant lysis-free plasma proteins (Fig. 8D). Protein clusters included chemotaxis regulators Ccl2, Ccl9 and

FIG. 8. Model-dominant plasma signatures of E-myc and E-TCL1 tumors.
A, Interproteome comparison detailing the canonical cellular localizations, annotated by Ingenuity Pathway Analysis (IPA), of proteins commonly or discretely quantified in plasma or B-cells. B, The relative plasma protein abundances relative to WT plasma for terminal E-myc and E-TCL1 tumors, highlighting those proteins annotated as intracellular (cytosolic/nuclear proteins) and extracellular. C, GO term enrichment for plasma proteins overabundant in both tumors, considering all plasma proteins and additional analyzing only those extracellular proteins not traceable to the B-cell tumor proteome, termed the "lysis-free" plasma. D. StringDB networks highlighting interactions and relationships for proteins overabundant in the "lysis-free" plasma for both tumors. E, GO term enrichment for plasma proteins overabundant specific to each tumor (Rs Ͼ 0.5 and p Ͻ 0.05, Rs other tumor Ͻ0.5). F, Proteins demonstrating progressive increases in abundance correlating with tumorigenesis in E-TCL1 mice.
Comparison between the terminal plasma signatures for each tumor model (Fig. 8B) demonstrated the dominance of this tumor lysis signature in E-myc tumors, whereas E-TCL1 terminal plasma contained a dominant signature of extracellular proteins. In both cases these signatures were observed for the opposing tumor, but to a lesser extent. GO term enrichment for overabundant plasma proteins specific to E-myc and E-TCL1 terminal tumors illustrated model-specific signatures of tumor lysis and extracellular proteins, respectively (Fig. 8E). E-myc tumor plasma specific GO terms had a strong resemblance to that of the B-cell tumors, highlighting cellular processes similar to both the tumors and the common plasma signature (Fig. 4A and 8C). Overabundant plasma proteins specific to terminal E-TCL1 tumors enriched for several GO terms related to immune processes, such as "immune response" (n ϭ 17, p ϭ 1.5 ϫ 10 Ϫ4 ) and "lymphocyte activation" (n ϭ 12, p ϭ 4.8 ϫ 10 Ϫ4 ).
Finally, consideration was given to proteins emerging in the preterminal "30%" E-TCL1 plasma (Fig. 8F). Plasma proteins demonstrating a correlation between overabundance and tumorigenesis were plotted, highlighting the extracellular immune-response proteins haptoglobin (Hp) and ITIH1, both also observed in the lysis free signature. Several additional proteins demonstrating overabundance in preterminal "30%" E-TCL1 plasma are detailed in supplemental Fig. S6I. DISCUSSION B-cell tumors have been intensively investigated by genomics and transcriptomics in recent years, advancing clinical and biological understanding (46 -49). Such information, is however sometimes limited, especially in a functional context, because of the weak correlations observed between mRNA and protein expression (50,51). A comparison between our proteomics data and a former E-myc tumor mRNA expression data set (supplemental Fig. S2J) highlighted an example of this, demonstrating an anticipated but limited correlation (52). Furthermore, genomics and transcriptomics are limited in capturing organism-wide tumor biology from acellular samples such as plasma.
This study aimed to combine and implement recent advances in quantitative MS proteomics to comprehensively characterize the tumors and plasma of contrasting mouse B-cell cancer models; E-myc and E-TCL1. By comparing tumors driven by contrasting oncogenes, with differing phenotypes and rates of progression, any apparent common signatures could suggest the most conserved and essential B-cell cancer mechanisms. Model-specific signatures could provide insight into the molecular basis of the discrete cancer phenotypes and oncogenes, potentially of relevance to BL and CLL. Simultaneous plasma characterization offered potential insight into tumor-host dialogue, systemic cancer im-pacts and biology of biomarker emergence. Furthermore, greater understanding of these models offers an opportunity to appreciate their strengths and weaknesses in preclinical applications.
Overall, isobaric-labeled LC-MS proteomics of B-cell and plasma samples (Fig. 1) provided biologically reproducible ( Fig. 2A, 2B), high-depth (supplemental Fig. S1H) and representative (supplemental Fig. S1) characterizations of the Emyc and E-TCL1 tumor models. The accuracy of the quantitative results was validated by anticipated transgene expression (Fig. 2C), Western blotting and flow cytometry (Figs. 2D, supplemental Fig. S2G, S2H) alongside comparisons to previously published protein expressions (supplemental Fig. S2D, S2F) and mRNA expression data (supplemental Fig. S2J). Interestingly, a brief analysis revealed that those proteins overexpressed without corresponding mRNA overexpression were enriched with proteins annotated with the term "translation." A Common B-cell Tumor Signature in Divergent Models-The expression correlation between the contrasting E-myc and E-TCL1 tumors (Fig. 3B) was indicative of a common B-cell tumor signature that was proportional to aggression. Indeed, this signature strongly highlighted canonical tumor characteristics such as upregulated cell proliferation and growth (Fig. 4). Furthermore, frequent observations of these overexpressed proteins in several human cancers (supplemental Fig. S3C) highlighted potential relevance to non-B-cell cancers. This was particularly noteworthy, given the many clinically relevant cell surface proteins and drug targets characterized (Fig. 3D, supplemental S3F, S3G). HMMR/CD168, for instance, has previously been considered as an immunotherapy target in B-cell cancers (41,53). Similarly, amino acid and zinc transporter overexpression could offer targets of metabolic inhibition or immunotherapy. The observation of proteins, and even putative proteins, in this common B-cell tumor signature with no prior links to cancer offers several new hypotheses (supplemental Fig. S3E). DDX49 and MAK16, for instance, are poorly characterized proteins overexpressed in all four tumor pools, and could be inferred to have a role in ribosome biogenesis, given their homology, orthologous functions and the prevalence of ribosome biogenesis upregulation in both tumors.
Downregulation of immune functions were indicative of the loss of B-cell characteristics, not essential and potentially even inhibitory toward tumor development. Constitutive Ig synthesis, for instance, would provide a negative selective pressure on resource-limited tumor cells. MHC component underexpression suggested a clear mechanism of immune evasion, described previously (54,55).
Model Specific Signatures of E-myc and E-TCL1 Tumors-E-myc-and E-TCL1-specific DEPs (Figs. 5-6) highlighted distinct expression patterns responsible for the contrast in tumor phenotypes; such as the vast protein dysregulation and proliferation produced by the aggressive and pleiotropic nature of myc. The contrast between these models was also apparent prior to tumorigenesis (supplemental Fig. S3B), with little to no tumor signature in 6-week E-TCL1 B-cells, whereas E-myc 6-week old B-cells were almost indistinguishable from terminal tumors. Although none of these 6-week E-myc mice presented with splenomegaly, metastases or ill health, it remains possible that tumor development was present at an early stage in one or more of the mice. E-myc-specific DNA and histone methylation dysregulation (Figs. 5B, 5C) suggests dynamic epigenomic instability as a basis of myc-induced tumor characteristics, previously observed in B-cell lymphomas (56,57). This potentially drives accelerated trait acquisition via nonmutational evolution (58).
Broad underexpression of actin cytoskeletal organizing components was potentially indicative of defects to normal B-cell migration and adhesion (Fig. 5D), identifying a possible mechanism promoting lymph node metastases; frequently observed for E-myc tumors. These proteins may also have been underexpressed as a result of redundant immune-related functions such as those described in Fig. 4D. Of the three inter-connected actin cytoskeletal organizing proteins validated with underexpression ( Fig. 5E), all three demonstrated both immune and cytoskeletal function. Cor1A promotes F-actin disassembly and cell motility (59) and mutations have been shown to impair lymphocyte function (60,61). LCP1 regulates actin bundling and B-cell development (62,63), whereas LCP1 knock out impaired B-cell migration (64). CAPG caps actin filaments with CAPG Ϫ/Ϫ mice exhibiting immune defects (65). Given observations of the pro-metastatic nature of these proteins in nonlymphocyte tumors (66,67), this indicated metastasis may be a more passive event for E-myc tumors. Overexpression of adhesion components such as integrin beta 1 may have facilitated this. Noncanonical cytoskeletal component AHNAK, a titin-related and cancer-associated cytoskeletal protein, shown to be essential for migration and invasion (68); was overexpressed in E-myc tumors, potentially highlighting an alternative metastatic pathway. Additionally, other cytoskeletal components such as those directing chromosome segregation were overexpressed. Validation of two further underexpressed proteins extended the observation of cytoskeletal dysregulation, with moesin and myosin-9 annotated with roles in leukocyte migration. Together these observations suggest a trend for further investigation.
The E-TCL1-specific features, such as ER stress response and aberrant glycosylation upregulation (Fig. 6B-6C) have previously been observed (69). Additional features, such as integrin-␣2 underexpression, transmembrane protein processing, post-translational modification, transport and membrane lipid composition, may suggest several potential mechanisms by which circulatory, leukemic cell formation may be promoted in E-TCL1 tumors.
Generally unaltered BCR pathway component expression, relative to the underexpression observed for E-myc tumors, indicates a role for antigen-receptor signaling in E-TCL1 tumors, observed previously (70) (supplemental Fig. S4J). This highlights potential examples of negative regulation or redundant functions for tumor growth and survival e.g. BCL6, FOXO1, and CD45. Upregulated phosphorylations, such as PI3Ks and FOXO1, specifically in E-TCL1 tumors may also illustrate tumor mechanisms (supplemental Fig. S4J, supplemental Table S4).
The dose-dependent proliferation of E-TCL1 tumor cells upon IL5 treatment (Fig. 7A-7C) provided additional, functional validation of IL5R overexpression. IL5R signaling has been suggested to induce activation of Lyn, JAK2, Syk, BTK, NF-B, and PI3K (76 -78). Given the suggested role of AKT amplification by TCL1, PI3K signaling was a likely candidate for the effectuation of the IL5 response. Indeed, the PI3K catalytic subunit type 2 beta (PIK3C2B), alongside other PI3K isoforms, appeared as the most overexpressed of the IL5R downstream signaling molecules. Dose-dependent AKT phosphorylation upon IL5 treatment confirmed a role for the PI3K pathway, highlighting TCL1 as a potential factor in the amplification of IL5R signaling (Fig. 7D-7E).
The observation of IL5RA expression by the expanded CD5 ϩ peritoneal and splenic B-cell populations in young E-TCL1 mice (supplemental Fig. S5J) strongly suggests a precursor cell population with characteristics of B-1 B-cells. Given the induction of B-cell proliferation observed with IL5 (79,80) and that IL5 Ϫ/Ϫ mice are deficient for CD5 ϩ B-1 B-cells (81), IL5R signaling emerges as a probable component driving E-TCL1 tumorigenesis from CD5 ϩ IL5RA ϩ B-1 Bcells.
The role for IL5 in B-cells, and B-cell tumors, appears to not be conserved between species (80, 82) highlighting a limitation in the recapitulation of human CLL with previous reports suggesting IL5 induces spontaneous apoptosis in these cells (83). However, the similarities of IL5R signaling in E-TCL1 tumors with IL4 signaling in CLL, suggest continued potential for E-TCL1 mice in modeling many aspects of the human disease.
Plasma Proteomics and Systemic Tumor Signatures-The Ͼ250-fold depletion of high-abundance proteins by SuPrE-SEC and the identification of a wide range of tumor and systemic proteins (supplemental Fig. S6, Table S1) suggested proteomic characterization of the low M w subproteome pro-vided an effective approach to plasma analysis. The capture of the degradome and peptidome, portions of plasma containing fragmented proteins and potential biomarkers, likely contributed to the success of this method (84 -86).
The overabundance of tumor-derived proteins in terminal plasma strongly suggested tumor lysis as a dominant mechanism (Figs. 8A-8E, supplemental S6), a process described in aggressive therapies and lymphoma (87)(88)(89). Lysis product overabundance correlated with tumor aggression and higher rates of cell death, most clearly in E-myc tumors. A correlation between total protein abundance in the B-cell proteome and overabundance in the plasma was observed, regardless of tumor overexpression (supplemental Fig. S6G-S6I). Several of these proteins were annotated biomarkers (supplemental Fig. S6H) highlighting tumor-lysis products in latestage, aggressive tumors, as the dominant biomarker signature. It is likely that exosomes and apoptotic blebs also contributed to this signature.
The systemic immune response signature observed for terminal E-TCL1 plasma suggested that slower tumor development elicited a greater inflammatory response, perhaps because of gradual accumulation, chronic inflammation and a loss of immune regulation (Figs. 8E). An anti-tumor immune signature emerged for both models, when deconvoluted to give the "lysis-free" plasma proteome, which suggested several potential systemic biomarkers (Fig. 8C, 8D).
The contrasting signatures of tumor lysis and immune response in E-myc and E-TCL1 terminal plasma respectively, was reflective of the vastly differing rates of tumor development. However, in both models the signatures were observed, to a lesser extent, in the opposing tumor.
Several systemic signatures were also observed upon integration of the B-cell tumor and plasma proteomes. Overabundance of the carriers of the extracellular matrix component hyaluronan ITIH1-4 ( Fig. 8E), substantial tumor overexpression of CD168 (HMMR, hyaluronan-mediated motility receptor) (Fig. 3D) and links between hyaluronan, inflammation and tumor growth (90) suggest hyaluronan as a component of the B-cell tumor microenvironment. Given their high concentrations in plasma (supplemental Table S1, supplemental Fig.  S6Q) (37), ITIH proteins hold potential as biomarkers of B-cell tumors and aberrant hyaluronan metabolism and transport. Fibronectin (FN1), recognized in the promotion of cancers (91), was observed as marginally overabundant and inferred to be functional (supplemental Fig. S6S). Given the overexpression of FN1-binding integrin ␤1 on tumors, this may represent another microenvironment signature of pro-oncogenic interactions with B-cell tumors.
Functional biomarker analysis for overabundant plasma proteins with inferred functionality in tumors highlighted S100A6, a previously proposed biomarker (92,93), and EIF4G1 (supplemental Fig. S6T) suggesting candidates for noninvasive testing directly related to tumor function.
Preterminal E-TCL1 plasma additionally offered several interesting findings, describing a more clinically-relevant time point in tumorigenesis when a leukemia is present, but is otherwise asymptomatic. The observation of overabundant preterminal "30%" E-TCL1 plasma proteins common to the terminal signature (Fig. 8F, supplemental S6I, S6O, S6P) suggested the characterization of progressive, early biomarkers of B-cell tumors. Components of the tumor lysis-free signature describing a systemic immune response were more prevalent, such as Hp and ITIH1. Proteins correlating with tumor progression, traceable to tumor lysis were, however, also present to some extent, suggesting that both lysis and immune signatures offer a source of biomarkers at earlier stages of tumor development. The more detailed evaluation of the multiple stages of tumor progression, of these and other tumor models, could therefore offer considerable insight into the dynamics of biomarker emergence, better informing biomarker discovery and application.

CONCLUDING REMARKS
In conclusion, this study has detailed the protein expression changes present across highly contrasting B-cell tumors, highlighting dominant features of proliferation and growth. E-myc tumors demonstrated a specific trend of epigenomic instability and cytoskeletal component underexpression, whereas E-TCL1 tumors specifically exhibited ER stress, dysregulated signaling, and IL5-driven proliferation. Many specific targets of kinase inhibitors, nucleoside analogues and immunotherapy emerged for each tumor type. Plasma proteomics indicated lysis products as a major late-stage tumor biomarker signature, dominantly in the more aggressive Emyc tumors, whereas E-TCL1 tumor plasma had a dominant signature of an anti-tumor immune response. Integration of plasma and B-cell proteomics data, alongside tumor lysis, highlighted IL5, HMMR/CD168, ITIH proteins, FN1, S100A6, and EIF4G1 as candidate biomarkers for further investigation. These findings reinforce that aggressive, targeted, chemotherapy and immunostimulatory antibodies promise effective means of treating E-myc and E-TCL1 tumors, respectively with the results offering exciting possibilities when combined and integrated with further additional 'omics and functional data. Finally, these findings suggest that global, integrative evaluation of tumors can provide systemic insight into tumor biology not captured by the evaluation of cells or plasma in isolation.