Phosphotyrosine signaling analysis in human tumors is confounded by systemic ischemia-driven artifacts and intra-specimen heterogeneity

Abstract

Tumor protein phosphorylation analysis may provide insight into intracellular signaling networks underlying tumor behavior, revealing diagnostic, prognostic or therapeutic information. Human tumors collected by The Cancer Genome Atlas program potentially offer the opportunity to characterize activated networks driving tumor progression, in parallel with the genetic and transcriptional landscape already documented for these tumors. However, a critical question is whether cellular signaling networks can be reliably analyzed in surgical specimens, where freezing delays and spatial sampling disparities may potentially obscure physiologic signaling. To quantify the extent of these effects, we analyzed the stability of phosphotyrosine (pTyr) sites in ovarian and colon tumors collected under conditions of controlled ischemia and in the context of defined intratumoral sampling. Cold-ischemia produced a rapid, unpredictable, and widespread impact on tumor pTyr networks within 5 minutes of resection, altering up to 50% of pTyr sites by more than 2-fold. Effects on adhesion and migration, inflammatory response, proliferation, and stress response pathways were recapitulated in both ovarian and colon tumors. In addition, sampling of spatially distinct colon tumor biopsies revealed pTyr differences as dramatic as those associated with ischemic times, despite uniform protein expression profiles. Moreover, intratumoral spatial heterogeneity and pTyr dynamic response to ischemia varied dramatically between tumors collected from different patients. Overall, these findings reveal unforeseen phosphorylation complexity, thereby increasing the difficulty of extracting physiologically relevant pTyr signaling networks from archived tissue specimens. In light of this data, prospective tumor pTyr analysis will require appropriate sampling and collection protocols to preserve in vivo signaling features.

Figure 1

Post-excision ischemia induces rapid, widespread, patient-dependent alterations to pTyr networks. (a) Heatmaps of all quantified peptides from individual patient-derived ovarian tumor 4-timepoint ischemia study. Rows indicate mass spectrometry-derived quantitative levels of pTyr sites in log₂ scale relative to t=0 min, rank ordered by value at t=60 min. Onset of color in the heatmaps corresponds to changes ≥3 s.d. from value at t=0 (red, increasing; blue, decreasing). See Supplementary Table 1 for details. (b)Hierarchical-clustered heatmap of quantitative pTyr profiles of individual ovarian tumors at resection. Quantitative levels of pTyr sites grouped in rows, in log₂ mean normalized scale. (c, d) Select examples of distinct inter-patient ischemia regulated temporal dynamics. Colored lines represent relative inter-patient phosphorylation levels, mean ± s.d. of technical replicates is shown. Temporal data values shown are derived from independent patient datasets and normalized with t=0 value measured in b (See Supplementary Fig. 3 for additional examples). (e) Affinity propagation-derived clusters for pTyr sites detected in patient 27 (See Supplementary Fig. 4 for affinity propagation clusters of other patients). Subplots signify groups of pTyr sites with similar quantitative trends across time (x-axis). Solid lines denote trends of individual pTyr sites. Mean log₂-temporal values relative to t=0 min are shown. Exemplar pTyr site of each cluster shown in blue. (f) Ischemia-regulated phosphorylation sites common to all patients with functional categorization. Ischemia index (y-axis) for each site represents absolute sum of the log₂ changes relative to t=0 at5, 30, 60 minutes. Mean value ± SEM from all patients is plotted. Asterisks denote sites known to be implicated in protein activity or function.

Figure 2

Spatial phosphorylation heterogeneity is apparent in human colon tumors. (a) Separate unsupervised hierarchical-clustered heatmaps represent the spatial and temporal sample set of each patient-derived colon tumor. Patient identifier shown to top left of each heatmap. Rows indicate mass spectrometry-derived quantitative levels of pTyr sites in log₂ mean normalized scale. Onset of saturated color in the heatmaps corresponds to changes ≥2-fold from the mean (red, increasing; blue, decreasing). (b) Pearson correlation analysis was used to quantify the direction and magnitude of correlation among the spatially distinct colorectal tumor samples. Data points presented in each plot are the log₂mean-centered value for a given pTyr site between paired spatially-distinct samples at the indicated timepoints in patient 323. r=Pearson correlation coefficient; 95% CI= 95% confidence interval of r; P=two-tailed p-value. See Supplementary Fig. 7 for correlation analysis of other patients. (c) Phosphorylation levels of clinically useful pTyr sites are plotted for each patient spatial and temporal sample set. Samples are shaded according to patient and columns grouped to indicate identical timepoint samples that are spatially distinct. Mean value ± SEM is plotted. Statistically significant differences are indicated. Note: ** indicates P=0.001 to 0.01, and * indicates P=0.01 to 0.05.

Figure 3

Spatial and temporal protein expression differences are minimal in colorectal tumors. (a)Separate unsupervised hierarchical-clustered heatmaps represent the spatial and temporal sample set of each patient-derived colon tumor. Patient identifier shown to top left of each heatmap. Rows indicate mass spectrometry-derived quantitative levels of protein expression in log₂ mean normalized scale. Onset of saturated color in the heatmaps corresponds to changes ≥2-fold from the mean (red, increasing; blue, decreasing). (b) Quantitative levels of protein (from a) and matching pTyr sites (from 2a) on the same protein. Rows represent values in log₂ scale relative to t=0_a min. Onset of color in the heatmaps represents changes greater than ∼1.4-fold from the mean (red, increasing; blue, decreasing) to emphasize uniformity in protein expression values compared to pTyr values. Plotted pTyr values are also quantitatively normalized to the protein expression levels. See Supplementary Table 6 for details. (c) Selected examples of corresponding protein and pTyr levels from b. Columns are shaded to group identical timepoint samples that are spatially distinct. Mean value ± SEM is plotted.

Figure 4

Ischemia alterations occur across tumor types impacting a core set of functional classes yet has the potential to affect biologically diverse pTyr pathways. (a) Heatmaps of individual patient-derived colorectal tumor 4-timepoint ischemia study. Rows indicate mass spectrometry-derived quantitative levels of pTyr sites in log₂ scale relative to t=0 min, rank ordered by value at t=60_av min. Average value of paired timepoints is plotted. Onset of color in the heatmaps corresponds to changes ≥3 s.d. from value at t=0_av (red, increasing; blue, decreasing). See Supplementary Table 4 for details. (b) Venn diagram indicating ischemia-phosphorylation signature observed in multiple tumors. 56 ischemia-regulated sites were detected in ≥4 ovarian tumor samples and 29 ischemia-regulated sites were detected in ≥4 colorectal tumor samples. (c) Kinase enrichment analysis (from union of data in b) was performed to identify kinases with high-confidence enrichment and plotted onto a map of the human kinome using Kinome Cluster to highlight the diverse set of kinases affected by delayed freezing. Green circle=kinases significantly altered in ≥4 colon tumors and/or ≥4 ovarian tumors.