Nature sub-journal: Spatial proteomics explores glioblastoma prognostic markers

Figure 1 Article

• Experimental methods: MALDI-MSI, label-free spatial proteomics

• Experimental samples: 46 glioblastoma samples

• Spatial area segmentation: red area, yellow area, blue area

• Areas of interest for pathology after HE staining: tumor area, very dense and dense infiltrated area, necrotic area, and microangioproliferation (MVP) area

Figure 2 Spatial area segmentation

Conclusion of the article

1

MALDI-MSI allows for the patient based on molecular characteristics

Group

Given the heterogeneity of GBM, the investigators performed spatially resolved proteomics studies
of 46 tumors from prospective patient cohorts under the guidance of mass spectrometry imaging (MSI).
To understand the stratification of patients in the cohort, the researchers considered global segmentation and considered the segmentation plot of major molecular differences, and identified 3 major regions, red region A, yellow region B, and blue region C
.
Each colored region has a common molecular signature, indicating that the spectra of these regions are similar
.

Figure 3 MALDI-MS imaging combined with microproteomics for the general workflow of GBM intertumoral and intratumor heterogeneity characterization

2

Identify specific signal pathway characteristics for each group

In order to understand the molecular differences between the three regions, the researchers selected 2-5 specific microextraction points on each tissue based on the molecular regions determined by the spatial segmentation of MALDI-MSI data to analyze the tumor heterogeneity and microenvironment with specific protein characteristics in each group, and obtained a total of 135 microextraction points
.
Of all tumor samples, 28 extractions were performed in red region A, 20 in yellow region B, and 87 in
blue region C.
These 3 regions show histologic heterogeneity:

• Red region A: 28 extractions: 39.
  3% tumor area, 14.
  3% very dense invasion area, 28.
  6% dense invasion area, 10.
  7% tumor area including a small number of necrotic cells, 3.
  6% tumor area with MVP, 3.
  6% tumor area with inflammation;

• Yellow region B: 20 extractions: 45% tumor area, 25% tumor area including a small number of necrotic cell area, 5% tumor area with MVP, 20% necrotic area;

• Blue region C: 87 extractions: 73.
  6% tumor area, 5.
  7% very dense invasion area, 2.
  3% dense invasion area, 3.
  4% tumor area including a small number of necrotic cells, 12.
  6% tumor area with MVP, 1.
  1% tumor area with inflammation, 1.
  1% necrotic area
  .
  

Figure 4 Spatial proteomic analysis

Proteomic analysis of these extracted samples using label-free revealed a total of 4936 proteins
.
The researchers first determined the correlation between all extraction points in 46 GBM samples through Pearson correlation analysis, which confirmed the existence of different proteomic GBM isoforms and cross-validated the results
of MALDI-MSI.
It was further found that a total of 1183 proteins were expressed
significantly differently between the three groups.
Three specific clusters of overexpressed proteins in each region are identified by heat map, i.
e.
, cluster 1 corresponds to the overexpressed protein in group B, cluster 2 corresponds to the overexpressed protein in group A, and cluster 3 corresponds to the overexpressed protein
in group C.
Pathway analysis found that group A overexpressed proteins were associated with neurogenesis and axon guidance, group B overexpressed proteins were associated with microglial activation and more generally immune system activation, and group C overexpressed proteins were mainly involved in tumor growth, developmental regulation, viral infection and transcriptome modification
at the RNA level.

3

Correlation between TCGA and proteomics data

The researchers compared nearly 5,000 proteins identified to the TCGA database, where 682 genes showed elevated expression in gliomas, and 282 proteins
from these 682 genes were found in the sample in this article.
In the proteome data, the researchers found that 12 proteins were associated with poor prognosis, of which 7 proteins were overexpressed in region A, 2 proteins were overexpressed in region B, and 3 proteins were overexpressed
in region C.
The researchers also found 9 proteins associated with a good prognosis, of which 7 proteins were overexpressed in region A and 2 proteins were overexpressed
in region C.

4

Proteome and survival data correlation analysis

In order to find potential prognostic proteins, the researchers performed ANOVA test on the entire proteome dataset based on overall survival (OS), divided the cohort into three groups according to the patient's OS, and used the Cox model to find that 5 proteins maintained a high significant correlation
with patient survival.
Based on the expression of these five proteins, it was found that the OS of the two groups (Cluster 1 and Cluster 2) was significantly
different.
Among the five proteins that were highly correlated with survival, IP_652563, ALCAM, and ANXA11 were adverse prognostic markers for tumors, while PPP1R12A and RPS14 were good prognostic markers
.

Fig.
5 Five proteins that were highly correlated with patient survival

The investigators further evaluated
5 proteins in an independent retrospective cohort of 50 patients by immunohistochemistry.
Patients are grouped according to their survival time and divided into three groups with different survival periods
.
Thirteen patients had low survival (less than 1 year), 25 had moderate survival (1-2 years), and 12 had high survival (more than 2 years).

RPS14 and PPP1R12A were expressed at higher levels in tumors in patients with longer survival compared to patients with low and moderate survival, and ANXA11 was expressed at higher levels in tumors in patients with low and medium survival compared with patients with longer survival, confirming the effectiveness of
prognostic markers.

GBM Past Interpretation

Article Interpretation | Regional maps of the glioblastoma proteome reveal a triaxial model of intratumor heterogeneity

Westlake Omi spatial proteomics