Designs By Kate Dbk Case Study Powerpoint

Gina Milano ENGN 0090 In 2005, Kate Creevey, founder and CEO of Designs by Kate, Inc., created a women’s jewelry company that sold jewelry via direct sales. She modeled her business after the Mary Kay cosmetics direct sales model. Since the direct sales business model had become fairly popular over the past decade, generating $29.7 billion in sales in 2004 alone, Creevey saw this as an opportunity to seize. Further, she believed jewelry fit this model well as it was universal; thus, she would not have to worry about sizing or other similar complications. Creevey strived to price her jewelry below $75 as she wanted to make her jewelry available to everyone, but at the same time she wanted her jewelry to be fashionable and trendy. Therefore, she hired a jewelry maker from New York to design an initial collection of 25 pieces. She also hired Juanita Thomas and John Simpson to help her run the company. Thomas was assigned to oversee the development and management of the sales while Simpson was the Chief Financial Officer of DBK. Her headquarters were located in New York, NY as that was the fashion capital of America. Creevey initialized an extensive word of mouth advertising campaign in order to

Citation: Catenacci DVT, Liao W-L, Thyparambil S, Henderson L, Xu P, Zhao L, et al. (2014) Absolute Quantitation of Met Using Mass Spectrometry for Clinical Application: Assay Precision, Stability, and Correlation with MET Gene Amplification in FFPE Tumor Tissue. PLoS ONE 9(7): e100586. https://doi.org/10.1371/journal.pone.0100586

Editor: Valli De Re, Centro di Riferimento Oncologico, IRCCS National Cancer Institute, Italy

Received: April 14, 2014; Accepted: May 25, 2014; Published: July 1, 2014

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Data Availability: The authors confirm that all data underlying the findings are fully available without restriction. The data are all contained within the paper and/or Supporting Information files.

Funding: This research was supported in part by NIH K12 award (CA139160-01A), CALGB/ALLIANCE YIA Award, Live Like Katie Foundation Award, Oncoplex Dx Research Collaboration, and UCCCC (University of Chicago Comprehensive Cancer Center) Award in Precision Oncology (DVTC); the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research (DB). A scientific research collaborative agreement between Oncoplex and University of Chicago allowed funding for reagents, expenses towards tissue acquisition and processing, and salaries of laboratory members. The role of each individual author at Oncoplex is as previously indicated, and as employees they received salaries from Oncoplex Dx. The specific roles of each author is outlined in the author contributions section.

Competing interests: WLL, ST, KB, JU, MD, DBK, TDV, TH, and JB are/were paid employees and stock owners at Oncoplex Dx. DVTC received collaborative research funding from Oncoplex Dx. This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials.

Background

Hepatocyte growth factor receptor (HGFR), commonly known as Met, is a membrane receptor that possesses tyrosine kinase activity [1], [2]. Binding of HGF ligand to Met activates its kinase activity through autophosphorylation of tyrosine residues 1234 and 1235. This activation of Met engages a number of additional signal proteins (e.g., CREB, ERK1, ERK1/2, ERK2, JNK, STAT3, and various MAPKK) either directly or indirectly, resulting in a variety of Met-driven biological activities that ultimately convey an invasive oncogenic phenotype [3]. Clinically, Met is of wide-spread interest, as overexpression of this protein is associated with aggressive tumor properties and poor patient outcomes [4]–[9]. Met signaling is aberrantly constitutively activated by protein overexpression and/or genetic alteration [10]. Specifically, MET gene amplification and consequent overexpression is an ‘oncogenic driver’ in a subset (∼5%) of gastric and esophageal adenocarcinomas [4], [11]–[14], while MET mutations have been rarely reported in various hereditary and sporadic cancers including gastroesophageal adenocarcinomas (GEC) [15]. A previous study has shown that Met protein is overexpressed in esophageal adenocarcinoma (EA) surgical specimens and EA cell lines, while Met dysregulation can occur early in the progression from Barrett’s dysplasia to adenocarcinoma [16], [17]. The role of Met in GEC and other cancers have made it a prime target for therapeutic strategies [4], [14], [18]. HGF or Met inhibitors currently under development can be broadly subdivided into biological or low molecular weight synthetic compounds, and are currently being tested in clinical trials [14], [19]–[23]. Biological agents are monoclonal antibodies (mAb) that either neutralize the ligand, hepatocyte growth factor receptor (HGF), or bind the receptor itself, effectively blocking the ligand/receptor interaction and activation. These are currently being evaluated in phase I–III trials for various tumor types [18], [20], [24], [25]. In a phase I trial, we described a complete response to onartuzumab, a Met monoclonal antibody, in a patient with stage IV GEC having high MET GCN and Met over-expression [18]. A recent randomized phase II trial in GEC evaluating an anti-HGF antibody, rilotumumab, demonstrated a survival advantage compared to placebo, with predictive benefit particularly in patients’ tumors having high Met expression (Met+) by immunohistochemistry (IHC), in contrast to those lacking expression (Met−) [20]. On the other hand, most synthetic compounds targeted against Met are ATP competitive tyrosine kinase inhibitors (TKI) that inhibit Met autophosphorylation and subsequent downstream signaling activation, with particular sensitivity observed in the setting of Met overexpression as a consequence of MET amplification [13], [14], [22], [26], [27]. Notably, a single arm phase IIa trial of foretenib, a multi-kinase inhibitor including Met, was relatively disappointing, at least as monotherapy in biomarker unselected chemo-refractory GEC patients [23].

IHC analysis of formalin-fixed paraffin-embedded (FFPE) tissue sections is routinely used for measuring Met expression in patient samples. While IHC has proven useful over the past several decades, mass spectrometry (MS) methods that measure the absolute levels of proteins with high specificity for clinical application are rapidly emerging [28]. Selected reaction monitoring (SRM)-MS assays are presently widely utilized to objectively quantitate metabolites in biological samples. In contrast to IHC which has limitations in specificity, reproducibility and sensitivity, the measurements provided by SRM-MS are highly specific since several performance characteristics of the analyte are measured [29]. These characteristics include its mass, the masses of several of its fragment (or transition) ions, chromatographic retention time, as well as how well these characteristics match to a heavy isotope labeled internal standard of the analyte. Addition of the heavy internal standard also allows for absolute abundance of the target analyte to be measured.

Liquid-Tissue-SRM has been developed which is a MS-based technology platform that measures the absolute abundance of targeted proteins in patient-derived FFPE tissue [30], [31]. In this method, laser microdissection is used to isolate tumor cells from FFPE sections and a cell lysate is prepared and digested into tryptic peptides (Figure 1). A known amount of a heavy peptide representing the targeted analyte is added to the lysate. Analysis using a SRM-MS method will then quantify the amount of analyte in the sample.

Figure 1. Liquid Tissue-Selected Reaction Monitoring (SRM) workflow for quantification of proteins from formalin-fixed paraffin-embedded (FFPE) tissue.

Tissue sections are cut onto Director microdissection slides. After they are deparaffinized, areas of interest are identified by pathologists. Laser microdissection is used to procure areas of interest and the targeted cellular material is collected in a tube. The cellular material is processed using the Liquid Tissue protocol, which includes trypsinization, to produce a soluble peptide lysate. A known amount of heavy internal standard peptide is added to the lysate and the sample is analyzed using SRM to measure the absolute abundance of the endogenous peptide of interest.

https://doi.org/10.1371/journal.pone.0100586.g001

In this study we present a Liquid Tissue-SRM assay for measuring the absolute abundance (amol/µg) of Met expression in FFPE tumor tissues ungoing laser microdissection. We tested the reproducibility across sample replicates as well as between different instrument platforms and operators. We also tested Liquid-Tissue-SRM Met levels in serial tissue sections re-analyzed one year after the initial testing; lack of temporal reproducibility is a recognized limitation of IHC. We then applied the developed assay to measure Met levels in 130 GEC FFPE tumors (in both small endoscopic and core biopsies, as well as curative surgical resection specimens), and results were correlated with Met expression as either an H-score or as a binary (positive/negative) score as determined by IHC, as well as MET gene copy number (GCN) and MET/CEP7 ratio as determined by fluorescence in situ hybridization (FISH). Finally, we sought to define a SRM expression level cut-off that would optimally distinguish between Met expression consistent with MET gene amplification as compared to expression without MET gene amplification.

Results

SRM assay development and preclinical validation

For development of the Met Liquid Tissue-SRM assay, multiple peptides obtained from a tryptic digest of recombinant Met were measured using MS. The resulting three candidate peptides, NLNSVSVPR, GDLTIANLGTSEGR, and TEFTTALQR, were then extensively screened in multiple formalin-fixed cell lines and FFPE clinical samples. The peptide 418TEFTTALQR426 provided the most reproducible peak heights, retention times, chromatographic ion intensities, clean elution profile, and distinctive/reproducible transition ion ratios; therefore, this peptide was selected for clinical assay development. SRM transitions used for the quantification of Met was selected based on the representative MS fragmentation spectrum for peptide 418TEFTTALQR[13C6,15N4]426 (Figure 2A). Standard curves were generated to determine the assay’s limits of detection and quantitation (LOD and LOQ) by spiking various amounts of the light peptide into eight aliquots of lysate prepared from formalin-fixed (FF) SK-BR-3 cells containing consistent amount of the heavy isotopic version of the peptide. The final concentrations of light peptide in the lysates ranged from 150 to 25000 amol. The amount of light peptide added to each of the eight calibration samples along with the amount detected in each is listed in Table 1. Coefficients of variation (CV’s) for the various concentration points ranged from 2.8% to 20% for samples analyzed in triplicate. The amount of light peptide recovered (amol) was plotted against the amount of light peptide spiked (amol) to create a standard curve. The LOD and LOQ was 150 and 200 amol, respectively, with a linear regression value of R2 = 0.9998 (y = 0.9627x+80.025) for the assay standard curve when all 8 spiking concentration points were included (Figure 2B) and R2 = 0.9989 (y = 1.1391x–7.7041) when excluding the highest point of 25000 amol (Figure 2Binset). The standard curve showed linearity and low variations over the 2 orders of magnitude concentration range tested. The chromatographic profile of the light and heavy versions of the peptide TEFTTALQR are shown in Figure 2C. Both peptides eluted at 9.67 minutes within the SRM-MS analysis and the corresponding SRM transition ion ratios are shown in Figure 2D. The precursor ions for the light and heavy peptides are m/z 533.78 and 538.78, respectively and three fragment ions per precursor were chosen for the identification and quantification of Met. The fragment ions for the light and heavy peptides and their corresponding optimized collision energy are m/z 588.35 (19V)/689.39 (20V)/836.42 (19V) and 598.35 (19V)/699.39 (20V)/846.47 (19V), respectively.

Figure 2. Development of SRM assay for Met showing the fragmentation spectrum for heavy TEFTTALQR peptide

(A), the standard curve generated in human SK-BR-3 cell lysate (B); inset: the standard curve generated without the highest spiking point (25000 amol). The total ion chromatograms for the light and heavy isotopically labeled peptides are shown (C) along with the transition ions used to identify and quantitate each peptide (D).

https://doi.org/10.1371/journal.pone.0100586.g002

Table 1. Characteristics of Met light peptide spiked SK-BR-3 lysates used to prepare calibration curve to determine limit of detection (LOD) and quantitation (LOQ) for development of Met Liquid Tissue-SRM assay.

https://doi.org/10.1371/journal.pone.0100586.t001

To test assay precision, Met was measured in 9 human NSCLC and 11 human GEC FFPE tissues. The precision study was conducted on two different LC-MS systems (System R and System S) operated by two different scientists. Five of the nine NSCLC samples and four out of eleven GEC tissues showed Met levels above the LOD. The levels of Met in these tissues ranged from 275.5–2401.2 amol/µg. The CVs for these measurements ranged from 1.5–17.7%. A plot (Figure 3) comparing the results from systems R and S had a linear regression value of 0.9968 (y = 0.9883x–17.728) demonstrating that the SRM-MS method generated extremely low level of variance between the two systems.

Figure 3. Precision assessment for measuring the absolute abundance of Met in 9 NSCLC and 11 GEC FFPE tissues.

Each sample was analyzed on two different liquid chromatography-mass spectrometry systems operated by two different scientists. Red square: Met positive GEC tumors (4/11). Blue square: Met positive NSCLC tumors (5/9).

https://doi.org/10.1371/journal.pone.0100586.g003

Comparison of Liquid Tissue-SRM and ECL Met Quantitation

The levels of Met in five different cell lines (B5/589, H596, SKBr3, MKN45, and T24) were measured using both Liquid Tissue-SRM and electrochemiluminescence (ECL) immunoassay. The results obtained using the two methods are plotted against each other in Figure 4. The correlation coefficient (R2) between the two sets of measurements was 0.998 when all five cell lines were compared, and 0.716 when excluding MKN45, a MET gene amplified line with extremely high expression. (Figure 4 inset). Overall there was good correlation of the measurements provided by Liquid Tissue-SRM and ECL.

Figure 4. Comparison of Met levels measured in five different cell lines using Liquid Tissue-SRM and an electrochemiluminescent (ECL) immunoassay.

Inset: comparison of SRM and ECL Met levels measured within the four cell lines containing the lowest concentration of Met.

https://doi.org/10.1371/journal.pone.0100586.g004

Resistance of Liquid Tissue-SRM Met Assay to Preanalytical Handling and Time from Tissue Sectioning

IHC is known to be very sensitive to preanalytical variables, including fixation time, warm and cold ischemia time, and the age of tissue sections. To test the robustness the SRM assay, Met protein expression was measured on tissues dissected from sections cut from ten NSCLC tumor blocks, where the first microdissection/analysis occurred immediately after sectioning and the subsequent microdissection/analysis from the adjacent serial sections occurred 13 months later. A plot comparing the Met levels detected in these paired tissues is shown in Figure 5. Met concentration values measured in these tissues at the different time points are also shown (Figure 5 inset table). The correlation coefficient between these two groups of samples was 0.893, showing that the Liquid Tissue-SRM process provides reproducible results for FFPE samples that have been previously sectioned up to 13 months prior to analysis.

In a separate experiment, the levels of EGFR in A431 xenografts fixed for various lengths of time (4 hour-1 week) ranged from 2–3.01 fmol/ug; similarly, EGFR expressions ranged from 2.3–3.0 fmol/ug when compared between immediate and 1 hour-delay fixation. The CVs for these measurements ranged from 3.7–24%. This demonstrated the excellent assay robustness regardless of fixation time (4 hour–1 week), as well as prefixation cold ischemia (1 hour immediately before fixation).

Met Expression in the Gastroesophageal Cancer (GEC) Patient Cohort

Met levels were quantitated in a cohort of 130 GEC tumors, comprised of primary tumor resections, endoscopic primary tumor biopsies, and core needle biopsies of metastatic sites (liver, peritoneum) (Table S1). The Met levels in the tumors were above the LOD in 45/130 (34.6%) samples with the levels in these samples ranging from 215.88 to 4669.5 amol/µg (Figure 6). The presence of Met could be unequivocally detected above the LOD in four additional samples; however, as their quantities were below the LOQ they were labeled as “positive” ( = ∼150 amol/µg). While the majority of the samples had Met levels on the order of a few hundred amol/µg total protein, the concentration of Met in 7 tissues ranged from over 2000 amol/µg to almost 5000 amol/µg total protein. The CV for triplicate measurement of all samples was less than 20%. In 73% of the samples, the CVs were less than 10%.

Figure 6. Absolute levels of Met in those GEC tumors above the LOD (45/130) as measured using Liquid Tissue-SRM.

Blue indicates MET gene amplified samples (ratio of MET/CEP7 ≥2) as determined by FISH, and green indicates those not gene amplified (ratio<2). Samples in red were not FISH tested. Sample 36 (526.93 amol/ug) is taken from reference 18 (sample obtained at disease recurrence after initial onartuzumab treatment).

https://doi.org/10.1371/journal.pone.0100586.g006

Comparison of Met SRM and MET gene copy number (GCN) in GEC cell lines & tissues

Met SRM results were compared with those of MET GCN in order to identify an optimal SRM cut-off level that was consistent with MET amplification (i.e. FISH ratio 2 and GCN≥4, see methods). We first assessed 22 GEC cell lines and 1 lymphoblast control line for MET GCN by FISH, and Met expression by SRM (Table S2, Figure S1). Met levels were detected in 16/22 (72.7%) GEC lines, and ranged from 150 to 6290 amol/µg. FISH confirmed three previously reported MET amplified GEC lines (SNU5, MKN-45, and OE33) [4]. Using a cut-off value of 1500 amol/µg, derived from these data, SRM demonstrated a 100% (3/3) sensitivity (95% CI 0.29–1) and 100% (20/20) specificity (95% CI 0.83–1) in discerning MET amplification status. This preliminary cut-off was therefore used to evaluate 30 GEC tissue samples prospectively.

We performed MET FISH in a subset (n = 30) of the total 130 GEC cohort (Table S3). Twenty-three samples (23/30) had mean MET GCN per nucleus >2. Among these elevated GCN samples, 7 tumors (23%) had MET gene amplification (ratio MET:CEP7 ≥2 and GCN≥4); the high rate of MET gene amplification was due to intentional enrichment in this 30 sample cohort with patients having previously established MET amplification by FISH. The Met SRM result for the same 30 cases is plotted against MET:CEP7 ratio (red) or MET GCN (blue) (Figure 7). Met expression correlated well with both parameters in all but two samples (Sample 23 and 24). The correlation coefficient (R2) between the two sets of measurements in all 30 samples were 0.651 for SRM and MET:CEP7 ratio, and 0.774 for SRM and mean MET GCN per nucleus (figure not shown). While Sample 24 had a MET:CEP7 ratio of 1.18, it demonstrated high polysomy with a mean 7.35 copies of MET per nucleus, corresponding with higher expression compared to low polysomy/disomy samples, but lower than MET amplified samples. Sample 23 represented a primary gastric cancer; only ∼20–30% of the tumor cells revealed clustered gene amplification (MET:CEP7 ratio 14.93) (sample 23a), while the remainder of tumor cells were low polysomic (sample 23b) (Table S3, Figure S2). Although the gene amplified subclonal area demonstrated a mean 41.8 copies of MET per nucleus, the remaining non-amplified tumor cells were also microdissected during processing for MS analysis based tumor cells observed by H&E staining. This tumor molecular heterogeneity, therefore, likely led to a lower ‘diluted’ SRM Met read. This explanation was verified by examining the metastatic lymph node of this patient (Sample 23c), where the hypothesis of possible clonal selection of the MET amplified cells from the primary tumor to the lymph node was evaluated. The metastatic lymph node (Sample 23c) did indeed reveal a mean of 15.33 MET copies per nucleus (ratio MET:CEP7 6.31) with 100% of the scored tumor cells having elevated MET GCN. The SRM detected a high amount of Met (3836 amol/µg) (Table S3, Figure S2).

Figure 7. Correlation of Met levels using Liquid Tissue-SRM and MET gene amplification by FISH in 30 GEC tumors.

This cohort includes diploid/low polyploid, high polyploid, and amplified samples. The left y-axis (blue diamond) represents the MET copy number per nucleus and the right y-axis (red square) indicates MET:CEP7 ratio.

https://doi.org/10.1371/journal.pone.0100586.g007

After discounting this molecularly heterogeneous primary tumor sample 23a, and replacing with sample 23c, SRM and MET:CEP7 ratio showed high correlation R2 = 0.898; SRM and MET GCN per nucleus had R2 = 0.829 (Figure 7). The SRM Met level of 1500 amol/µg, as determined from the cell line data, was 100% (7/7) sensitive (95% CI 0.59–1) and 100% (23/23) specific (95% CI 0.85–1) for discerning MET amplification status by FISH in the FFPE tissues.

Overall, the pooled data from the cell lines and GEC tissues (using 23c) demonstrated high concordance between MET GCN (or amplification ratio) and SRM Met measurement. SRM identified GEC lines and samples having MET amplification with 100% (10/10) sensitivity (95% CI 0.69–1) and 100% (43/43) specificity (95% CI 0.92–1).

Comparison of Met IHC to SRM in GEC cell lines and FFPE tissues

We assessed Met IHC in various cell line FFPE pellets (Figure S3), which under-appreciated the expression intensity level differences between MET amplified and non-amplified lines. We then assessed the subset (n = 44) of the total 130 GEC cohort having both Met IHC and SRM. A comparison between the two in terms of any expression revealed a correlation coefficient R2 = 0.54 (Figure S4). Using an H-score of ≥10 and SRM level of ≥LOD (150 amol/ug) to define ‘Met+’, there was high discordance between the two assays (Table S4, Figure 8A). Using SRM as the reference, IHC identified 20/23 (sensitivity 87%; 95% CI 0.66–0.97) positive SRM samples. However, IHC was negative in only 4/21 SRM negative samples (specificity 19%, 95% CI 0.05–0.42). This was only marginally improved when using an alternate IHC scoring system where Met is considered positive when ≥25% of tumor cells stain with ≥1+ intensity (Table 2). Using this system, of the 23 positive SRM samples, IHC identified 19 positive (sensitivity 82.6%, 95% CI 0.61–0.95); of 21 negative SRM samples, IHC identified 7 negative (specificity 33.3%, 95% CI 0.15–0.57). Increasing the IHC cut-off to ≥50% tumor cells staining ≥1+ improved specifity to 42.9% (95% CI 0.22–0.66) at the expense of sensitivity 78.3% (95% CI 0.56–0.93), as would be expected. In summary, IHC and SRM for Met expression were relatively discordant.

Figure 8. IHC to SRM or FISH

assay correlations. Correlation of IHC Met positive H-score to (A) Met SRM (amol/µg) or (B) MET/CEP7 ratio by FISH. Inset tables assess sensitivity/specificity of IHC H-score, assuming SRM (A) and FISH (B) as the comparative standards.

https://doi.org/10.1371/journal.pone.0100586.g008

Comparison of Met IHC to MET FISH in GEC FFPE tissues

We assessed a subset of 31 GEC samples with both Met IHC and MET FISH scores to assess the sensistivy/specificity of H-score to identify MET amplified tumors (Table S5). Using an H-score of ≥150 and FISH ratio of ≥2 with GCN≥4 as ‘amplified’, there was discordance between the two assays (Figure 8B). The correlation coefficient comparing IHC to FISH was R2 = 0.48 (FISH ratio) and R2 = 0.51 (FISH GCN) (Figure S5). Using FISH as the gold standard, IHC identified 6 of 8 FISH amplified tumors resulting in a sensitivity of 75% (95% CI 0.35–0.97). IHC identified tumors with H-score ≥150 in 5 FISH negative samples resulting in specificity of 78.3% (95% CI 0.56–0.83). In summary, using an IHC H-Score of ≥150 to discern MET amplification status resulted in a sensitivity of 75% and specificity of 78%. Altering the H-score cut-off could increase the sensitivity at the expense of specificity, or vice versa. H-score was out-performed by SRM in the ability to accurately discern MET amplification status in this study.

Comparison of Met IHC, SRM Met, and MET FISH in GEC tissues

We assessed the correlation of IHC, SRM and FISH in 24 GEC FFPE tissues that had been tested using all three methods (Table S6). A stronger correlation of SRM with FISH MET/CEP7 ratio (R2 = 0.89) was observed compared to SRM with IHC H-Score (R2 = 0.68) (Figure 9A). Similarly there was a stonger correlation of SRM with FISH MET/CEP7 ratio (R2 = 0.89) compared to IHC H-Score with FISH (R2 = 0.53) (Figure 9 B).

Figure 9. Three-way SRM-IHC-FISH assay correlations.

(A) Correlation of SRM Met level (amol/µg, x-axis) to Met H-Score by IHC (blue, left y-axis) and to FISH MET/CEP7 ratio (red, right y-axis) in GEC FFPE tissues having all three tests performed. Correlation coefficients for the comparisons are in their respective colors. (B) Correlation of FISH MET/CEP7 ratio (x-axis) to Met H-Score by IHC (blue, left y-axis) and to SRM Met level (red, right y-axis) in GEC FFPE tissues having all three tests performed. Correlation coefficients for the comparisons are in their respective colors.

https://doi.org/10.1371/journal.pone.0100586.g009

Discussion

Detection of relevant protein biomarkers in tumor samples has been routinely accomplished through IHC analysis of FFPE tissues. While IHC has proven useful, it is limited by preanalytical variability, antibody specificity, imperfect sensitivity, low throughput, and inter-laboratory differences in methodology and analysis/scoring standardization.

IHC is used to semi-quantify proteins in clinical samples and can be a critically important feature that often dictates diagnostic or therapeutic decisions. The colorimetric signal produced by IHC, however, is nevertheless subjectively “quantitated” as interpreted by the Pathologist. To generate an IHC “score”, the staining intensity and percentage of cells must be taken into account as well as the extent of perceived background staining. Despite recent application of image software and signal quantitation hardware platforms for IHC analysis, the lack of standardization and subjective quantitation are challenges of IHC being uniformly, systematically, and successfully applied for characterizing target proteins in cancer patient tissues.

To overcome these limitations we have developed a targeted Liquid Tissue-SRM assay to measure Met directly from FFPE patient tissue. The assay requires only a small amount of sample, and is routinely accomplished using laser microdissection of a single 10 µm tissue section. Indeed, many of the GEC samples analyzed in this study were acquired as endoscopic or core biopsies of metastatic lesions. The assay is epitope-independent and therefore does not suffer from antibody-related issues such as degree of specificity, manufacturing variability, and tissue processing and fixation. Since the absolute quantity of Met in the sample is determined by software analysis of the MS signal, the result is objective and absolutely quantitative.

Archival FFPE sections are extremely valuable resources for translational clinical studies and for the validation of clinical utility for companion diagnostics. However, IHC analysis of previously cut tissue sections is limited, deeming its usefulness, reliability, and clinical significance on archival FFPE slides compromised over time. There was observed decrease in the reactivity for p53 in FFPE tissue sections that had been cut from a paraffin block and stored at room temperature for only two months [32]. Slides cut two months previously and stored at room temperature were stained simultaneously with those that had been freshly cut and results showed that for all cases, the immunoreactivity was weaker on slides that had been stored. In three instances where p53 staining was observed on freshly cut slides, no p53-immunreactivity was observed on the corresponding stored slides. This deterioration of immunoreactivity has been demonstrated to occur as early as 2 weeks after storage [33]. A loss of immunoreactivity has been reported for estrogen receptor (ER), progesterone receptor (PR), cyclin D1, E-cadherin, and Her2 on sections stored for 6 months at 4°C [34]. Most procedures for conducting IHC analysis recommend that sections be cut from the tissue block immediately before testing to preserve antigenicity. For example, Ventana recommends that sections stored longer than 6 weeks should not be used for IHC testing using its breast panel kit, which contains antibodies that target ER, Her2, PR, and Ki-67 (www.ventanamed.com).

Discordant result provided by older slides is a major concern, particularly due to the increasing importance of retrospective diagnostic tissue analysis. It is estimated that more than one billion FFPE tissue blocks are archived in tissue banks worldwide. Since the first description of formalin fixation and storage was described more than a century ago, this approach has become the technique of choice for preservation of tissue. Most importantly, in many cases previously cut sections are the only form in which remaining archived FFPE tissue is stored. Archiving the tissue should enable retrospective biomarker studies of various cancers; however, the temporal stability of IHC epitopes is unsatisfactory, with unreliable results. RNA expression analysis via next-generation sequencing (NGS) is also subject to wide variability, often lacking concordance with protein expression even in the best of circumstances, let alone when evaluating aged FFPE tissue sections [35], [36].

Simultaneous IHC analyses of multiple samples concurrently using tissue microarrays (TMAs) can be used to increase assay throughput, yet will often analyze tissues that have been stored for several weeks/months after sectioning from their paraffin block. Adjusting the IHC protocols for fresh or stored slides (e.g. different antigen retrieval methods) may offer a potential solution; however, this strategy eliminates the study’s consistency and introduces the presence of potentially undetectable artifacts that may arise by altering the protocol for a selected group of sections. Using different antibodies has also been suggested as a possible remedy to this problem. Different antibodies are known to provide highly variant results, even when used on identical slides [37]. Previous studies have also failed to demonstrate how storage conditions can be improved to alleviate the discrepancy between results provided by fresh and stored slides [33]. Finally, construction of TMAs in routine clinical settings is not performed for logistical reasons.

In our study we showed that Liquid Tissue-SRM results obtained at least one year apart of cut tissue sections are highly reproducible, providing a more robust platform for conducting both proteomic biomarker studies on archival samples. The robustness of Liquid Tissue-SRM results compared to IHC results is most likely due to the fact that the SRM analyte is a linear tryptic peptide instead of a three-dimensional epitope. As such, the SRM assay is less susceptible to preanalytical handling issues by heating the tissue to 95 degrees, followed by trypsinization. In addition to the age of cut sections, we have also assessed time of fixation (4 hour to 1 week) and cold ischemia (1 hour) prior to fixation. In each case, SRM results were similarly not impacted by these factors.

We then demonstrated a strong correlation with ECL expression and SRM for Met in FFPE cell lines. Next, analysis of 130 GEC samples using Liquid Tissue-SRM showed that Met at any level was detected in approximately 35% of the tumors. This rate of Met expression is similar to other studies, where Met ‘overexpression’ by IHC ranged from 23.7% to 70% [4], [38]–[40]. In addition to random patient selection, this wide variability of Met expression rates in the literature may be due to several factors, including a lack of standardization in the age of the tissue section, the staining procedure, and, very notably, the interpretation/scoring of the IHC analysis. Additionally, different antibodies (polyclonal and monoclonal) used for detecting Met also very likely contribute to the variability in Met expression rates observed across these different studies. On the other hand, studies in our MS lab have shown high intra-laboratory reproducibility across different operators and MS instrumentation (Figure 3), and this level of reproducibility should easily translate to other laboratories since the Liquid Tissue-SRM process does not rely on affinity reagents or subjective interpretation of the MS results. Finally, as demonstrated by Sample 23 in our series, intratumoral molecular heterogeneity (through space within the primary tumor, or from primary tumor to lymph node and/or distant site) may also contribute to variable rates of Met overexpression (as well as variable gene copy number). Therefore the site of the tumor sample should also be taken into account when determining the rate of Met overexpression within a given patient cohort, since metastastatic lesions may possess higher Met expression levels [41].

Gene copy number (GCN) elevation, and more specifically MET gene amplification have been shown to be both prognostic and predictive in preclinical and early phase clinical trials [4], [5], [11]–[14], [18]. FISH assessment of GCN is low-throughput, laborious, and subjective, similar to IHC. It is possible that certain Met inhibitors may be more efficacious in MET amplified tumors (eg. TKIs) while other inhibitors may be more efficacious in MET non-amplified tumors (eg. HGF ligand antibodies). Therefore, a multiplex analysis using SRM, including Met and other relevant proteins, that is able to provide a linear expression level of Met, as well as validated cut-off levels consistent with gene amplification, would be an ideal ‘one-time’ clinical test for economical use of scarce tissue. Moreover, the level of gene amplification and consequent level of overexpression may also be important with respect to prognostic and predictive benefit, as recently was shown for HER2/Her2 [42]. We demonstrated a strong linear correlation with Met SRM levels and both MET GCN and MET/CEP7 ratio in cell lines and FFPE tissues, while IHC poorly correlated with SRM and FISH analyses. We demonstrated that ≥1500 amol/µg of Met had excellent (100%) sensitivity and specificity in all samples evaluated to detect MET FISH amplification, and further validation of this cut-off is ongoing prospectively in an independent large patient tumor cohort. One case (23a) was initially determined to be negative for Met expression by SRM and IHC, as well as non-amplified by FISH, NGS, and array comparative genomic hybridization (aCGH) (data not shown). However, a lymph node metastasis of this patient was shown to be highly amplified by both SRM and FISH, suggesting that the primary tumor harbored a subclonal population of MET amplified cells which seeded the lymph node metastasis. This intra-tumoral molecular heterogeneity underscores the need to molecularly profile lesions (metastases) that will ultimately become the most detrimental to the patient. This occurred in only 1 of 7 (15%) identified MET amplified FFPE samples; the true rate at which this observed tumor evolution occurs with respect to MET amplification is an area of active research in our laboratory. Efforts to biopsy metastatic lesions for molecular profiling may help limit biomarker misclassification due to intra-patient tumor molecular heterogeneity through space [41]. ‘High’ expression in non-amplified samples, albeit lower than MET amplified tissues, was observed with high polysomy samples (eg sample 24), representing a subset of samples having expression levels less than 1500 amol/µg but more ∼1000 amol/µg. Thus, ‘gene dosage’ clearly corresponded with SRM-MS Met expression levels in the cell lines and FFPE samples evaluated. Prospective evaluation of the prognostic and predictive properties of each of the observed expression groups (none: <150 amol/µg, low: 150–1000 amol/µg, high: 1000–1500 amol/µg, very high: >1500 amol/µg) will be valuable.

One limitation of this study is the use of a polyclonal anti-MET antibody for IHC testing. An ongoing study evaluating currently available monoclonal anti-MET antibody companion diagnostics will evaluate the performance of those antibodies compared to SRM and FISH assays. Another limitation is the current lack of correlation of the three assays (FISH, IHC, and SRM) with clinical outcome data (prognostic or predictive); this analysis is also ongoing. Despite a promising randomized phase II trial [43], a recent press release regarding anti-Met therapy in the phase III trial for ‘MET+’ lung cancer showed futility, and may hinge on the manner in which ‘MET+’ expression is defined; the ability of SRM to provide absolute expression that is linear and quantitative may overcome challenges posed by the subjectivity of IHC scoring [44]. Ultimately, after direct comparison, the companion diagnostic that is able to reliably quantify Met/MET with corresponding clinical relevance (prognostic and/or predictive) in a timely, tissue-economic, and cost-efficient manner will prevail.

As biomarker discoveries and the establishment of their clinical relevance continue, it will be critical that the companion diagnostic assays be accurate but also expandable. While assays used to determine biomarker expression will be performed using many different technologies (e.g., MS, IHC, FISH, DNA/RNA-seq etc.), this expansion must not only preferably include the ability to multi-plex additional important analytes within the assay [45], [46], but should also be able to translate easily to other tumor types. Since the Liquid Tissue-SRM method developed for Met requires no antibodies or other types of affinity reagents, and is performed using an unbiased sample preparation method, this ‘next-generation’ proteomic companion diagnostic is easily applied to other tumor types besides GEC. Moreover, additional SRM processes able to simultaneously quantitate up to 80 protein biomarkers from limited tissue samples have been developed, along with Met, in a single assay from a single 10 µm tissue section cut onto a microdissection slide [30], [31], [41], [45]. The ability to provide the protein expression data of many actionable biomarkers will enable better economic use of these scarce clinical tissue samples as well as more accurate multivariate prognostic/predictive capabilities [47], [48]. Upon receiving a tissue sample, generation of a quantititative report is on the order of 5 days, which competes with IHC turnaround time - particularly when taking into account the numerous advantages of SRM compared to IHC discussed above. We are currently evaluating the companion diagnostic SRM ‘GEC-plex’ [45], including Met, in a large clinically linked series of GEC samples as a prognostic and predictive tool, and to further evaluate the cut-off level of ≥1500 amol/µg for MET amplification as a surrogate for ‘Met-addicted’ tumors. Additionally, the ‘GEC-plex’ is incorporated as an integral diagnostic test, along with targeted NGS, to prospectively assign GEC patients into predefined molecular subsets in the planned PANGEA clinical trial [41].

Materials and Methods

Ethics Statement Human Samples

All research involving human tissue samples was approved by the University of Chicago Review Board (IRB16146B and IRB16294A) and conducted in accordance with these protocols and the Declaration of Helsinki. Written informed consent from the donor or next of kin was obtained in accordance with the tissue procurement protocols.

Tissue and laser microdissection

One thought on “Designs By Kate Dbk Case Study Powerpoint

Leave a Reply

Your email address will not be published. Required fields are marked *