The effect of sample age on the metabolic information extracted from formalin-fixed and paraffin embedded tissue samples using desorption electrospray ionization mass spectrometry imaging.
Author(s): Isberg OG, Xiang Y, Bodvarsdottir SK, Jonasson JG, Thorsteinsdottir M, Takats Z
Publication: J Mass Spectrom Adv Clin Lab, 2021, Vol. 22, Page 50-55
PubMed ID: 34939055 PubMed Review Paper? No
Purpose of Paper
This paper investigated the effects of formalin-fixed, paraffin-embedded (FFPE) block storage on the analysis of the metabolome using desorption electrospray ionization mass spectrometry imaging (DESI-MSI) by evaluating logistic regression models that classify specimens based on the decade of specimen collection (block storage duration). Features not affected by storage were validated for tissue analysis using frozen specimens.
Conclusion of Paper
The developed linear regression model had an average classification accuracy of 96%. When storage interval groupings were shortened from 10to 3 or 5- years, the classification accuracy increased to 98% (both), and when storage interval groupings were lengthened to 25 or 50 years, the classification accuracy increased to 99%. Out of MS features identified (number nor specified), 152 features were found to be significant when data was separated by the duration of FFPE block storage. A linear regression model based only on these 152 features had an area under the curve of 0.97. Investigation of the intensities of high (500—900 m/z) versus low (100-500 m/z) mass range features found that while the high mass range features were relatively stable during FFPE block storage, the intensities of the low mass range features declined with storage. Of the 87 high mass range features, 60 were stable in FFPE specimens during block storage, of which 7 were found to overlap with the frozen TMA within 10 ppm.
Studies
-
Study Purpose
This study investigated the effects of FFPE block storage on the analysis of the metabolome using DESI-MSI by evaluating logistic regression models that classify specimens based on the decade of specimen collection (FFPE block storage duration). Features not affected by storage were validated for tissue analysis using frozen specimens. Nine tissue microarrays containing 583 FFPE breast cancer cores collected between 1935 and 2013 from 216 patients and seven frozen tissue microarrays containing breast tumor and normal adjacent specimens embedded in hydrogel stored at -80°C were obtained from the department of pathology. Cryosections (12 μm) were stored at -80°C until analysis. FFPE sections (thickness not specified) were deparaffinized by incubating for 1 h at 60°C and rinsing twice in xylene. The metabolites were quantified using a Xevo G2-XS QTOF mass spectrometer in negative ion mode (mass range 50-1000 m/z) coupled to a two-dimensional DESI stage set to a spatial resolution of 85 µm. Consecutive sections from FFPE blocks were H&E stained. For the analysis, cores from all slides were grouped together based on collection decade (1940-1949, 1950-1959, 1960-1969, 1970-1979, 1980-1989, 1990-1999, 2000-2009, and 2010-2020). Weighted logistic regression models were then constructed in Python and refined based on univariate and multivariate statistical methods. Significantly affected features were identified (P<0.05) and the relative importance of the top 30 features were assessed based on ranking the logistic regression coefficients. Overlapping features between the univariate and multivariate analyses were considered to be features significantly affected by FFPE block storage duration. Features found to be stable were validated using the frozen specimens.
Summary of Findings:
The developed linear regression model had an average classification accuracy of 96%. When storage interval groupings were shortened from 10 to 3 or 5- years, the classification accuracy increased to 98% (both), and when storage interval groupings were lengthened to 25 or 50 years, the classification accuracy increased to 99%. Out of MS features identified (number nor specified), 152 features were found to be significant when data was separated by the duration of FFPE block storage. A linear regression model based only on these 152 features had an area under the curve of 0.97. Investigation of the intensities of high (500—900 m/z) versus low (100-500 m/z) mass range features found that while the high mass range features were relatively stable during FFPE block storage, the intensities of the low mass range features declined with storage. Of the 87 high mass range features, 60 were stable in FFPE specimens during block storage, of which 7 were found to overlap with the frozen TMA within 10 ppm.
Biospecimens
Preservative Types
- Formalin
Diagnoses:
- Neoplastic - Not specified
- Neoplastic - Carcinoma
Platform:
Analyte Technology Platform Small molecule ESI MS Lipid ESI MS Lipid Tissue microarray Small molecule Tissue microarray Pre-analytical Factors:
Classification Pre-analytical Factor Value(s) Biospecimen Preservation Type of fixation/preservation Formalin (buffered)
Frozen
Storage Storage duration Collected in 1940-1950
Collected 1950-1960
Collected 1960-1970
Collected 1970-1980
Collected 1980-1990
Collected 1990-2000
Collected 2000-2010
Collected 2010-2020