MSACL 2018 US Abstract

Topic: Tissue Imaging & Analysis

Identifying Predictive Markers of Breast Cancer Receptor Status and Molecular Subtypes Using DESI-MS Imaging

Kyana Garza (Presenter)
University of Texas at Austin

Authorship: Kyana Garza (1), Raquel Mary Rodrigues Peres (2), Andréia M. Porcari (2), Jialing Zhang (1), Jonathan Young (1), Robert J. Tibshirani (4), Geisilene Russano de Paiva (2), Wendong Yu (3), Chandandeep Nagi (3), Marcos N. Eberlin (2), Luís O. Zanatta Sarian (2), and Livia S. Eberlin (1)
(1) Department of Chemistry, University of Texas at Austin, Austin, Texas, USA, (2) State University of Campinas, Sao Paulo, Brazil, (3) Department of Pathology, Baylor College of Medicine, Houston, Texas, USA, (4) Department of Statistics, Stanford University

Short Abstract

Breast cancer receptor status is indicative of patient prognosis as well as determines the appropriate treatment option for the patient. Here, desorption electrospray ionization mass spectrometry, together with the Lasso, were used to characterize breast cancer receptor status. Statistical classifiers were generated for the prediction of breast cancer receptor status, and an overall accuracy of 88.7%, 89.0%, and 34.9% was achieved for estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) status, respectively. Metabolic signatures predictive of breast cancer receptor status were also identified using the Lasso method.

Long Abstract

Introduction

Breast cancer is the most common cancer among women worldwide, with approximately 12% of women in the United States at risk of developing the disease during their lifetime (1). Breast cancer is a highly heterogeneous disease with distinct molecular subtypes, each of which is defined by the overexpression, or status, of certain cell receptor proteins. ER, PR, HER2 are predictive biomarkers of breast cancer prognosis and heavily influence decisions regarding treatment options (2). Therefore, the determination of breast cancer receptor status is essential in the clinical management of breast cancer as it allows personalized treatment/therapy options for breast cancer patients (2-5). Immunohistochemistry (IHC) and fluorescence in situ hybridization (FISH) are commonly used approaches to determine ER, PR, and HER2 status of breast cancer tissue in the clinic. However, these methods can be time and labor intensive as they require lengthy procedures (6-7), and FISH requires experienced personnel to conduct and interpret the results (7). Ambient ionization mass spectrometry imaging (MSI) has been widely used to analyze and visualize a variety of molecular species, including fatty acids, metabolites and complex lipids, in biological tissue with minimal sample preparation (8-10). More specifically, desorption electrospray ionization mass spectrometry imaging (DESI-MSI) has been used to obtain molecular information that is diagnostic of various types of human cancer in near real time (11). Statistical tools have been used to evaluate the molecular information obtained from DESI-MSI, allowing for the selection of discriminatory molecules as well as the prediction of disease state (12).

Here, we employed DESI-MSI to correlate lipid and metabolic profiles with ER, PR, and HER2 status for enhanced breast cancer diagnosis. DESI-MSI allows for the spatial visualization of molecular species that are characteristic of each receptor status in near real time. The Least absolute shrinkage and selector operator (Lasso) statistical method was used to build classification models based on lipid and metabolite species that discriminate between the positive and negative status of each receptor. High mass accuracy/high mass resolution and tandem mass spectrometry measurements were used to tentatively identify species selected by Lasso as important biomolecules in ER, PR and HER2 breast cancer subtypes. In summary, our results indicate that DESI-MSI and Lasso enables the detection and identification of differing species within the ER, PR, and HER2 class that may be used for molecular subtyping of breast cancers.

Methods

Breast cancer tissue samples were obtained from the Cooperative Human Tissue Network (CHTN) and CAISM hospital (Campinas, Brazil). Samples were sectioned at 16 µm thickness and analyzed in both positive and negative ion mode using a LTQ-Orbitrap Elite mass spectrometer (Thermo Fisher Scientific, CA) coupled to a DESI-MSI platform (Prosolia Inc., IN). DESI-MSI was performed using a ACN:DMF (1:1) in negative ion mode and pure ACN in positive ion mode. The analyzed tissue sections were H&E stained for pathological evaluation, where histological pure regions were identified. Tissue samples were further characterized for ER, PR and HER2 status using molecular biology assays including IHC, FISH and gene sequencing. The least absolute shrinkage and selection operator (Lasso) statistical method was used to identify molecular markers and build classification models.

Results

DESI-MSI was used to investigate the differences in estrogen receptor positive (ER+) and estrogen receptor negative (ER-), progesterone receptor positive (PR+) and progesterone receptor negative (PR-), and human epidermal growth factor receptor 2 positive (HER2+) and human epidermal growth factor receptor negative (HER2-) breast cancer samples in negative and positive ion mode. A variety of molecular species, such as metabolites, fatty acids, sphingolipids (such as ceramides) and glycerophospholipids (GP) were observed in the mass spectra obtained in the negative ion mode. Using the Lasso method, we performed feature selection to identify predictive markers that are significant in differentiating the positive and negative status of each receptor. Several lipid classes significantly contributed to the models generated, including glycerophosphoinositols (PI), glycerophosphoserines (PS), glycerophosphoethanolamines (PE). DESI-MS imaging enabled the clear visualization of these species, allowing for the discrimination of ER+ from ER- regions as well as PR+ from PR- regions within the same tissue sample. For example, PE 34:2 (m/z 700.529) and PI 36:1 (m/z 863.566) were observed at higher relative abundances in ER+ tissue samples whereas PS 32:1 (m/z 807.503) was observed to be more abundant in ER- samples. Several predictive markers were selected by Lasso to differentiate between PR+ and PR- tissue including PI 34:1 (m/z 835.34), PS 38:4 (m/z 810.529), and PE 36:2 (m/z 742.53). Our data shows that PI were more abundant in both ER+ and PR+ tissue, while PG were more abundant in ER- and PR- tissue. More species were detected in positive ion mode, such as triaclyglycerides (TG) glycerophosphocholines (PC). TG were observed to be more abundant in normal breast tissue, which typically contains more fatty tissue compared to breast tumors, while PC were observed to be more abundant in breast cancer tissue. Currently, a classifier is being generated for the prediction of breast cancer receptor status using our positive ion mode data.

Using Lasso, a classification model was built using cross validation to perform per pixel and per patient analysis of ER and PR status in samples. For ER status prediction, 30,478 pixels from 77 independent samples were evaluated and yielded an overall agreement of 88.7% with the ER status in breast cancer samples. For PR status prediction, 30,420 pixels from 77 independent patients were evaluated and yielded an overall agreement of 89.0% with the PR status in breast cancer samples. Moderate results were achieved for HER2 status prediction, with an overall accuracy of 34.9% for 25,427 pixels.

Our current efforts include optimizing IHC for breast cancer frozen sections as well increasing our number of samples. Although the receptor status is provided in pathology reports, IHC is performed on a different section of the sample than what is received, meaning that the receptor status for the sample received could be different than what is provided in the pathological report. Optimizing IHC for breast cancer frozen sections will identify regions of ER+, ER-, PR+, PR-, HER2+, and HER2- cells. Identifying regions of positive and negative receptor status cells will further improve our statistical classifiers for the determination breast cancer receptor status. Based on the results using previously determined receptor status information, we expect Lasso to achieve higher performance once an improved classification model is generated. We are also currently investigating why a lower performance was observed for HER2 prediction as well as generating statistical classifiers for the prediction of breast cancer molecular subtypes.

Conclusions & Discussion

DESI-MSI enabled the spatial visualization of ER+, ER-, PR+ and PR- regions within a tissue sample, allowing for the characterization of breast cancer receptor status. Using the Lasso statistical method, overall accuracies of 88.7% and 89.0% were achieved for ER and PR breast cancer prediction, respectively. DESI-MSI enabled the detection and identification of potential predictive markers within the ER and PR class that may be used for molecular subtyping of breast cancers. Our current results indicate that DESI-MSI has the potential to be applied towards the determination of breast cancer receptor status directly from tissue samples, which is critical for breast cancer diagnosis and treatment selection for breast cancer patients.


References & Acknowledgements:

References

1. DeSantis, C.; Ma, J.; Bryan, L.; Jemal, A., Breast cancer statistics, 2013. CA Cancer J Clin 2014, 64 (1), 52-62.

2. Kos, Z.; Dabbs, D. J., Biomarker assessment and molecular testing for prognostication in breast cancer. Histopathology 2016, 68 (1), 70-85.

3. Nguyen, P. L.; Taghian, A. G.; Katz, M. S.; Niemierko, A.; Abi Raad, R. F.; Boon, W. L.; Bellon, J. R.; Wong, J. S.; Smith, B. L.; Harris, J. R., Breast cancer subtype approximated by estrogen receptor, progesterone receptor, and HER-2 is associated with local and distant recurrence after breast-conserving therapy. J Clin Oncol 2008, 26 (14), 2373-8.

4. Yager, J. D.; Davidson, N. E., Estrogen carcinogenesis in breast cancer. N Engl J Med 2006, 354 (3), 270-82.

5. Dai, X.; Li, T.; Bai, Z.; Yang, Y.; Liu, X.; Zhan, J.; Shi, B., Breast cancer intrinsic subtype classification, clinical use and future trends. Am J Cancer Res 2015, 5 (10), 2929-43.

6. Taylor, C. R.; Levenson, R. M., Quantification of immunohistochemistry--issues concerning methods, utility and semiquantitative assessment II. Histopathology 2006, 49 (4), 411-24.

7. Yaziji, H.; Goldstein, L. C.; Barry, T. S.; Werling, R.; Hwang, H.; Ellis, G. K.; Gralow, J. R.; Livingston, R. B.; Gown, A. M., HER-2 testing in breast cancer using parallel tissue-based methods. Jama-J Am Med Assoc 2004, 291 (16), 1972-1977.

8. Dill, A. L.; Eberlin, L. S.; Costa, A. B.; Zheng, C.; Ifa, D. R.; Cheng, L.; Masterson, T. A.; Koch, M. O.; Vitek, O.; Cooks, R. G., Multivariate statistical identification of human bladder carcinomas using ambient ionization imaging mass spectrometry. Chemistry 2011, 17 (10), 2897-902.

9. Eberlin, L. S.; Dill, A. L.; Costa, A. B.; Ifa, D. R.; Cheng, L.; Masterson, T.; Koch, M.; Ratliff, T. L.; Cooks, R. G., Cholesterol sulfate imaging in human prostate cancer tissue by desorption electrospray ionization mass spectrometry. Anal Chem 2010, 82 (9), 3430-4.

10. Dill, A. L.; Ifa, D. R.; Manicke, N. E.; Ouyang, Z.; Cooks, R. G., Mass spectrometric imaging of lipids using desorption electrospray ionization. J Chromatogr B Analyt Technol Biomed Life Sci 2009, 877 (26), 2883-9.

11. Ifa, D. R.; Eberlin, L. S., Ambient Ionization Mass Spectrometry for Cancer Diagnosis and Surgical Margin Evaluation. Clin Chem 2016, 62 (1), 111-23.

12. Sans, M.; Gharpure, K.; Tibshirani, R.; Zhang, J. L.; Liang, L.; Liu, J. S.; Young, J. H.; Dood, R. L.; Sood, A. K.; Eberlin, L. S., Metabolic Markers and Statistical Prediction of Serous Ovarian Cancer Aggressiveness by Ambient Ionization Mass Spectrometry Imaging. Cancer Research 2017, 77 (11), 2903-2913.

Acknowledgements

We would like to thank the National Institute of Health (Grant 4R00CA190783-02) for funding this project, and the Cooperative Human Tissue Network for providing the tissue samples (funded by the NCI). We would also like to acknowledge John Lin for his assistance in data and statistical analysis.


Financial Disclosure

DescriptionY/NSource
Grantsno
Salaryno
Board Memberno
Stockno
Expensesno

IP Royalty: no

Planning to mention or discuss specific products or technology of the company(ies) listed above:

no