Proteomic changes in cerebrospinal fluid from primary central nervous system lymphoma patients are associated with protein ectodomain shedding

Figures

• 

  Figure 1. Clinical data of participating individuals The mean values and corresponding standard deviations are shown. (A) Concentrations of albumin, IgG, IgA and IgM in CSF and serum. (B) CSF/serum concentration quotients of albumin, IgG, IgA and IgM. Significance levels: p < 0.05 (*), p < 0.01 (**).

• 

  Figure 2. Hierarchical cluster analysis of the CSF proteome in PCNSL and control patients For the comparison, only proteins (306 proteins) not correlating with the albumin concentration were considered. The comparable proteomic signatures of PCNSL patients treated (10 patients) or untreated (7 patients) with steroids led to the decision to include all PCNSL patients in one group for further data interrogation.

• 

  Figure 3. Characterization of the PCNSL CSF proteome Altogether, 601 and 438 proteins were identified and quantified in the CSF, respectively. A/B. Distribution of quantified proteins in PCNSL patients (A) and tumor-free controls (B) in percentages. Proteins with concentrations >10-5 μg/μL are shown separately, whereas proteins < 10-5 μg/μL are summed. (C) Abundance range of quantified proteins. Abundances are shown in log10 scale (mean concentration of each group; blue: control group, red: PCNSL patients).

• 

  Figure 4. Statistical analysis of quantified proteins (A) Volcano plot of proteins significantly correlating with CSF albumin, measured by LC-MS (Pearson correlation, p-values corrected with Benjamini-Hochberg correction, Supplementary Table 2). The red line indicates a p-value of p = 0.001. Proteins marked as “true” (p < 0.001) correlated significantly with CSF albumin (positive fold-change, more abundant in PCNSL patients; negative fold-change, more abundant in the control group). (B) Distributions of confidence intervals of proteins not correlating with albumin (Supplementary Table 4). The green line indicates the upper and lower limits calculated from the technical variance (three times the standard deviation). Unchanged proteins (104) are indicated in red. Proteins were ranked according to their abundances (from high to low). (C) Volcano plot of proteins not correlating with albumin (306 proteins) (Supplementary Table 5). The red line indicates a p-value of p = 0.05. Proteins marked as “true” (p < 0.05, fold-change > 2) differed significantly in abundance between the control and PCNSL patients, whereas proteins marked with “false” did not differ significantly (positive fold-change,…

• 

  Figure 5. Comparison of differentially abundant CSF protein data and tumor tissue data The tumor tissue proteome (930 proteins, 1% false discovery rate, ≥ 2 peptides per protein) consisted of 6 independent PCNSL patients (unpublished mass spectrometry data), and the transcriptome study consisted of 21 independent PCNSL patients (3773 genes, occurring in at least 75% of the patients, as published by Montesinos-Rongen in 2008 [24]).

• 

  Figure 6. Network enrichment analysis of differentially abundant proteins In total, 66 proteins were used to generate the network of GO biological processes. Thirty-three proteins (50%) were successfully mapped to 21 unique biological processes. The proteins (gene names) associated with the GO terms are highlighted in red. The direction of alteration is indicated by an arrow, and the occurrence at the proteome or mRNA level is indicated by an asterisk. The color code indicates the group affiliation. The term of each group with the highest significance is shown in bold letters.

• 

  Figure 7. Sequence mapping of identified peptides onto the specific amino acid sequences of transmembrane proteins Proteins that were more abundant in the CSF of PCNSL patients are marked with an asterisk. The number represents the length of the protein (amino acids). Color code: blue, extracellular; yellow, transmembrane region; green, cytoplasmic; orange, luminal; magenta, signal peptide; black, identified peptides.