Targeted Mass Spectrometry Enables Multiplexed Quantification of Immunomodulatory Proteins in Clinical Biospecimens

Abstract

Immunotherapies are revolutionizing cancer care, producing durable responses and potentially cures in a subset of patients. However, response rates are low for most tumors, grade 3/4 toxicities are not uncommon, and our current understanding of tumor immunobiology is incomplete. While hundreds of immunomodulatory proteins in the tumor microenvironment shape the anti-tumor response, few of them can be reliably quantified. To address this need, we developed a multiplex panel of targeted proteomic assays targeting 52 peptides representing 46 proteins using peptide immunoaffinity enrichment coupled to multiple reaction monitoring-mass spectrometry. We validated the assays in tissue and plasma matrices, where performance figures of merit showed over 3 orders of dynamic range and median inter-day CVs of 5.2% (tissue) and 21% (plasma). A feasibility study in clinical biospecimens showed detection of 48/52 peptides in frozen tissue and 38/52 peptides in plasma. The assays are publicly available as a resource for the research community.

Keywords: cancer; correlative biomarkers; immuno-MRM; immunotherapy; mass spectrometry.

Figure 1

Immuno-MRM enables highly multiplexed protein quantification. (A) The immuno-MRM assay workflow commences with generation of a protein lysate from the biospecimen of interest. Cleavable stable isotope labeled standards unique to each targeted peptide sequence are spiked into the sample at a known concentration. The protein mixture is converted to peptides by enzymatic digestion (Lys-C and trypsin). Custom monoclonal antibodies coupled to magnetic beads are used to enrich the endogenous peptides and labeled standards. The eluate is analyzed by multiple reaction monitoring-mass spectrometry, where analyte peptides and internal standards coelute with equivalent relative areas of monitored transitions. High sensitivity is achieved through analyte enrichment and optimization of mass spectrometer parameters for the targeted peptides. High specificity is maintained through optimal selection of fragment ion transitions. (B) Trypsin-mediated release of peptides was produced by overnight enzymatic digestion of proteins from a pool of cell lysates. The peak area ratios (light:heavy) for the 33 endogenous peptides detected were normalized to the maximum timepoint and plotted over time. Error bars are the standard deviation of three replicates. (C) Performance figures of merit for assays characterized in tissue matrix. (D) Performance figures of merit for assays characterized in plasma matrix. A representative response curve for the peptide VEIIATMK from CXCL10. Each concentration point was measured in triplicate. Distribution of R² values from linear regression of the response curves. Distribution of lower limit of quantification (LOQ) where each point refers to concentration determined by the lowest point on the curve with less than 20% CV. Repeatability is characterized by the distribution of CV values for intra- (within day) and inter- (between day) variability at three concentrations, in addition to the measurement of endogenous peptide. Each point corresponds to the average %CV for a peptide measured at three concentrations, Low (Lo), Medium (Med), and High (Hi) in triplicate over five days (n=15 at each concentration for a peptide). Endogenous measurements refer to the intra (within day) and inter- (between day) variability of endogenous peptides detected above the LOQ in five replicates measured over 5 days (n = 25). Stability shows the distribution of %CV and relative percent difference for 4 conditions compared to immediate analysis: (i) stored at 8 hours at 4°C, (ii) 48 hours at 4°C, (iii) after two freeze-thaw cycles, and (iv) stored at -80°C for 5 weeks. For box plots, the line shows the median value, boxes show the inner quartiles, and the whiskers show 5-95% of data.

Figure 2

The immuno-MRM assays show detection in frozen tissues. (A) Frozen tissues were obtained for 110 tumors from 11 tumor types. The number of each type is indicated in the pie chart. (B) The relative fractions of adipose, lymphocytes, red blood cells, stroma, and tumor cells were plotted for tumors with available images (108 out of 110). Cellular microheterogeneity was determined by using the HALO algorithm. Each point represents an individual tumor. Box plots show median (line), inner quartiles (box), and 5-95% range (vertical lines). (C) Distribution of peptide detection plotted as a histogram, showing the number of peptides detected above LOQ across the 110 frozen tumors. (D) Heatmap showing unsupervised clustering of analytes detected above LOQ in > 50% of tumor specimens. Peak area ratios (light:heavy) were normalized for each peptide analyte, and the z-score was used for clustering. (E) Histogram showing correlation of protein expression measured by immuno-MRM with mRNA transcript level determined by RNAseq (52). The median is indicated by a dotted line (0.559).

Figure 3

Determination of sample requirements for detection in tissue. The number of analytes predicted for decreasing amounts of tissue was determined from the signal-to-noise ratio measured using 500 μg protein digest input. Error bars show the 95% confidence interval.

Figure 4

Utility of the assays for measurement of protein expression in plasma. (A) Plasma samples were obtained for 45 patients with breast, colorectal, or ovarian tumors, as indicated in the pie chart. (B) Distribution of peptide detection plotted as a histogram, showing the number of peptides detected above LOQ across the 45 plasmas using 100 μL aliquots of plasma as input. (C) Heatmap showing unsupervised clustering of analytes detected above LOQ in >50% of plasma samples. Peak area ratios (light:heavy) were normalized for each peptide analyte, and the z-score was used for clustering. (D) Histogram showing the correlation of protein expression levels in patients for whom both tissue and plasma samples were available.