Abstract
Background: Accumulating data indicate that human epidermal growth factor receptor-2 (HER2)-positive breast cancer is a heterogeneous disease. We undertook a study to correlate lipid profiles with heterogeneous clinicopathological features of HER2-positive breast cancer. Materials and Methods: Histology-directed matrix-assisted laser desorption/ionization (MALDI)-mass spectrometry (MS) analyses were performed on 22 retrospective frozen tissue samples collected from patients with HER2-positive metastatic breast cancer, in order to correlate lipid profiles with clinicopathological characterisitics. Additionally, a pair of tumor and adjacent normal tissue was profiled to identify cancer-associated changes in lipid profiles. Results: Sphingomyelin 34:1, phosphatidylcholine (PC) 32:0, and PC 34:1, and PC 36:2 were overexpressed in HER2-positive breast cancer compared to adjacent normal tissue (HER2 signature). Lipid MALDI-MS profiles were different between Ki-67-high and Ki-67-low tumors. The proliferation signature (Ki-67-high vs. Ki-67-low) and the HER2 signature (cancer vs. normal) did not significantly overlap with each other. Conclusion: For the first time to our knowledge, this study describes lipid profiles correlated with various clinicopathological characteristics of HER2-positive breast cancer. Lipid profiling might be helpful for the molecular characterization of this disease.
Matrix-assisted laser desorption/ionization (MALDI)-mass spectrometry (MS) imaging has been demonstrated to be useful for molecular profiling of common solid tumors (1). In this approach, thin sections of frozen tissues are obtained from surgical resections or biopsies, and mass spectra were obtained from discrete tumor locations on the tissue. We previously reported that protein and lipid MALDI-MS profiles can classify breast cancers according to intrinsic subtypes (2). MALDI imaging MS can classify breast cancer tissue specimens according to human epidermal growth factor receptor-2 (HER2) status (3). Accumulating evidence suggests that an alteration in lipid composition is associated with breast carcinogenesis (4-8). We previously reported that phosphatidylcholine (PC) 34:1 is overexpressed across all intrinsic subtypes of breast cancer (2). The activity of phospholipid-metabolizing enzymes is regulated by the receptor tyrosine kinase cascade (9, 10). While erythroblastic leukemia viral oncogene homolog-2 (ERBB2) is a well-defined oncogene affecting the carcinogenesis, prognosis, and treatment outcome of breast cancer, HER2-positive breast cancer is heterogeneous in response to various therapeutic agents, patterns of recurrence and survival (11). We, therefore, undertook a study to correlate lipid profiles with heterogeneous clinicopathological features of HER2-positive breast cancer. As a result of this research, we have identified HER2-positive breast cancer-associated lipid changes, as well as lipid profiles correlated with the Ki-67 positivity (as a measure of proliferation) and the treatment outcome, using MALDI MS analyses.
Materials and Methods
Tissue preparation and the acquisition of MALDI MS data. Samples were collected from patients with breast cancer with Institutional Review Board-approved informed consent (NCC-NCS09278), at the time of curative surgery (n=12) or the diagnostic percutaneous needle biopsy (n=10), at the National Cancer Center in Korea. Samples were frozen in liquid nitrogen prior to the analysis. A cut-off value of 1% or more of positively-stained nuclei was used to define estrogen receptor (ER) positivity. HER2 immunohistochemistry result was scored as 0-3+ according to the method recommended for the HercepTest (Dako, Glostrup, Denmark) (12). Only tumors with scores of 3+ or ERRBB2 gene amplification by fluorescence in situ hybridization (FISH) were included in this study. Ki-67 immunohistochemistry was performed using MIB-1 antibody (Dako).
For MALDI-MS analyses, thin (10 μm) cryosection slides were obtained from frozen tissues using a cryostat (Leica CM 3050S, Leica Microsystems Inc., Bannockburn, IL, USA). One section was affixed to a standard glass slide, and then stained with hematoxylin and eosin (H&E). The other sections were thaw-mounted onto indium tin oxide (ITO) slides (HST Inc., Newark, NJ, USA), and desiccated in a vacuum for subsequent MALDI MS profiling. Binary matrix solution was prepared by dissolving 7 mg each of 2,5-dihydroxybenzoic acid (DHB) and α-cyano-4-hydroxycinnamic acid (CHCA) in 1 ml of 70% methanol plus 0.1% trifluoroacetic acid and 1% piperidine (13). DHB/CHCA matrix (100 pL) was deposited using a Chip-1000 instrument (Shimadzu, Kyoto, Japan). Mass spectra were acquired in both positive- and negative-ion reflector modes using an UltrafleXtreme MALDI time-of-flight mass spectrometer (Bruker Daltonics, Bremen, Germany). MS data were acquired in the m/z range between 500 and 1,100 Da by averaging signals from 3,000 consecutive laser shots with a frequency of 1,000 Hz. Guided by the H&E-stained cryosection slide, matrix spots representing tumor-rich areas were selected using FlexImaging software (version 2.1, Bruker Daltonics) for each tumor sample (Figure 1). Mass spectra data from selected spots was exported to ClinProTools (version 2.2, Bruker Daltonics) for further data processing. MALDI tandem MS analysis was directly performed on the tissue section after MALDI MS and mapped to public lipid databases (www.lipidmaps.org).
Data processing and statistical analysis. ClinProTools was used for baseline subtraction, spectral recalibration, and spectral area calculation. A resolution of 300 was applied to the peak detection method. The Top Hat baseline with 10% minimal baseline width was used for baseline subtraction. Data reduction was performed at a factor of four. Spectra were recalibrated with a maximal peak shift of 2,000 ppm between reference and peak masses. The value of the % Match to Calibrant Peaks parameter was set to 20%. Spectra that were not re-calibratable were excluded. All data with signal-to-noise ratios of more than 5 were acquired, and the peak area was used for the peak calculation with zero level integration. To identify MALDI- MS peaks differentially expressed between cancer and normal tissue, the average peak area (±standard deviation) was compared using Student t-test. For the correlation study, an average peak list was set up for each tissue sample by choosing peaks on the calculated total average spectrum for each tissue sample to create one average spectrum per patient. After excluding peak m/z 616.3 (non-lipid) in the positive mode, we normalized positive-mode datasets to the average peak area. After excluding peak m/z 540.2 (non-lipid), we also normalized negative-mode datasets to the average peak area. Average-normalized datasets (i.e. positive- and negative-mode lipid datasets) were then combined into a single dataset and subjected to statistical analysis using BRB-ArrayTools (version 4.1; NCI, Bethesda, USA) (14). Ki-67 positivity was determined based on the percentage of cells stained positive for Ki-67 and dichotomized as ≥15% or <15%. A principal component analysis (PCA) plot was generated using multi-dimensional scaling analysis of BRB-ArrayTools, which graphically represents correlation coefficients among samples without forcing the samples into specific clusters. The three primary principal components were used as the axes for the 3-dimensional scaling representation. Class prediction analyses were also performed using all classifier functions built into BRB-ArrayTools [compound covariate predictor (CCP), diagonal linear discriminant analysis (LDA), 1- and 3-nearest neighbors (NN), nearest centroid (NC), and support vector machines (SVM)]. To evaluate whether classes had different lipid profiles or not, class prediction analyses were performed using all samples as a training set. The 0.632+ bootstrap cross-validated misclassification rate was computed for all classifier functions in the training set. Class labels were then randomly shuffled and the cross-validated misclassification rate was computed for each random dataset. The permutation p-value, which is defined as the proportion of random datasets that give as small misclassification rate as is obtained with real class labels, was then calculated. MALDI-MS profiles of the classes were considered different if this permutation p-value was <0.05. In order to compare signatures, the 4-peak index was calculated by summing up the peak area of four informative peaks in a given dataset.
To correlate lipid MALDI-MS profiles with treatment outcome, the survival analysis tool (of the BRB-ArrayTools) was used, which identified peaks whose expression was correlated with the time to progression (TTP) fits to proportional hazards models relating TTP to each peak. Time to progression is defined as the time from treatment (trastuzumab in combination with either docetaxel or paclitaxel) to disease progression according to response evaluation criteria in solid tumor (RECIST, version 1.1) (15). The survival analysis tool computed the p-value for each peak for testing the hypothesis that the TTP was independent of the expression level for that gene. Class labels were then randomly permutated, and for each random permutation, the p-value was re-computed for each peak. The tool then computed the proportion of the random permutations that gave as many peaks significant at feature selection p<0.05 as were found in comparing the true class labels. Lipid profiles were considered to be significantly correlated with TTP if the permutation p-value was less than 0.05.
Results
Cancer vs. normal tissues. This histology-directed lipid MALDI-MS analysis is based primarily on 22 patients with retrospective frozen tissue samples collected from HER2-positive metastatic breast cancer. Additionally, paired tumor and adjacent normal samples were obtained from a 54-year-old female patient (ER−, HER2 3+) who underwent modified radical mastectomy, in order to identify peaks differentially expressed between normal and tumor [invasive and ductal carcinoma in situ (DCIS) components, separately]. At feature selection p<0.0001, 12 peaks were differentially expressed between the invasive component of this HER2-positive breast carcinoma and adjacent normal tissue (designated as the HER2 signature) (Table I and Figure 2). All of these peaks were overexpressed in cancer. MS/MS analyses identified four peaks among them. Peaks at m/z 741.6, m/z 772.7, m/z 798.6, and m/z 824.7 in the positive ion mode were identified as sphingomyelin (SM) 34:1, phosphatidylcholine (PC) 32:0, and PC 34:1, and PC 36:2, respectively. When the DCIS component of the same patient was compared with adjacent normal tissue for lipid profiles, 12 peaks were differentially expressed between DCIS and adjacent normal tissue. The same four lipids were identified (SM 34:1, PC 32:0, and PC 34:1, and PC 36:2) from these 12 peaks. No peaks were differentially expressed between invasive components and DCIS components, at feature selection p<0.001.
Correlation with clinicopathological variables. Twenty-two HER2-positive breast carcinomas were subject to the correlation study (Table II). When lipid MALDI MS profiles were compared between HER2 2+ and HER2 3+ tumors, there was no significant difference in lipid profiles. Lipid profiles were not different either according to the ER status or the menopausal status. Notably, however, lipid profiles were significantly different according to the Ki-67 positivity (Ki-67 ≥15% vs. <15%). Permutation p-values for 0.632+ bootstrap cross-validation misclassification rates were ≤0.05 for all classifiers except support vector machines (0.05, 0.04, 0.01, 0.03, 0.05, and 0.08 for CCP, LDA, 1-NN, 3-NN, NC, and SVM, respectively). Table III lists lipid peaks with positive correlation with the percentage of Ki-67-positive cells (designated as the proliferation signature). We addressed whether the proliferation signature (Table III) overlaps with the HER2 signature [SM 34:1 (m/z 741.6), PC 32:0 (m/z 772.7), and PC 34:1 (m/z 798.6), and PC 36:2 (m/z 824.7)]. When the 4-peak index was generated for each sample by summing up the peak area of these four identified lipids in the HER2 signature, the 4-peak index did not significantly correlate with the percentage of Ki-67 positive cells of each sample (R=−0.18; p=0.46). Thus, the proliferation signature does not significantly overlap with the HER2 signature.
We then planned to correlate lipid profiles with the clinical response to trastuzumab. Given the small sample size of patients undergoing uniform treatment (taxane/trastuzuamb), supervised prediction analyses could not be performed in this study. Instead, a PCA plot showed that patients who responded to the taxane/trastuzuamb treatment according to the RECIST criteria (n=13) appeared to be clustered separately from those who did not respond (n=3) (Figure 3). There were 23 peaks correlated with the TTP after taxane/trastuzumab therapy at the feature selection p<0.05 (designated as the trastuzumab response signature), and the probability of obtaining at least 23 peaks as being significant by chance (at the 0.05 level), if there were no relationship between the expression profile and survival was 0.049, suggesting that lipid profiles may be significantly different according to the treatment outcome, although this finding needs to be validated in a larger dataset. Neither the HER2 signature nor the proliferation signature segregated patients according to the TTP after therapy (log-rank p=0.90 and p=0.48, respectively).
Discussion
The current study, as far as we are aware, is the first to evaluate lipid profiles in HER2+ breast cancer using MALDI-MS analyses. Tissue MALDI-MS approach is a fast and relatively simple procedure that requires only a small amount of tissue material, yet generates a global view of lipid expression in clinical tissue samples as was also demonstrated in the previous articles (1, 2). SM 34:1, PC 32:0, and PC 34:1, and PC 36:2 contents were found to be increased in HER2-positive breast cancer, in both invasive and DCIS components. These lipids are constituents of cell membrane, and therefore their requirement is increased in rapidly growing cells (16). In addition, this study suggests that lipid profiles are different between patients with Ki-67-high HER2+ tumor and patients with Ki-67-low HER2+ tumor.
It has been suggested that regulation of phospholipid may be linked to the development and progression of cancer (17-19). Three out of four identified lipids in the HER2 signature were phosphatidylcholines. Choline kinase, which generates phosphatidylcholines, plays a role in carcinogenesis (17, 18). RNA interference-mediated choline kinase knock-down induced a significant decrease in phosphatidylcholine level in breast cancer cells with reduced proliferation (19). Given that the activity of phospholipid-metabolizing enzymes is regulated by receptor tyrosine kinase cascade (9, 10), it was hypothesized in our study that a HER2-driven, cancer-associated change in lipid profiles might overlap with the proliferation signature in HER2-positive breast cancer which should also be linked to growth factor signal transduction pathways. According to our findings, however, the proliferation signature and the HER2 signature did not overlap. It may be postulated that once mammary epithelial cells become malignant, lipid profiles may evolve as the accumulation of genetic abnormalities introduce a variable proliferative potential into each HER2-positive breast carcinoma. The HER2 signature did not segregate patients according to the TTP after trastuzumab/taxane therapy either. The limited sample size precluded the identification of lipids correlated with trastuzumab response, but this proof-of-concept study suggests that lipid profiles may be different according to the clinical response to trastuzumab. Thus, this study provides lipid portraits of HER2-positive breast cancer and suggests that lipid profiling might be helpful for the molecular characterization of this disease.
Acknowledgements
This work was supported by grants from the Converging Research Center Program (2012K001506) through the Ministry of Education, Science and Technology of Korea and by the Proteogenomic Research Program through the National Research Foundation of Korea funded by the Korean Ministry of Education, Science and Technology. Dr. Y. H. Kim was supported by a grant of Korea Basic Science Institute (G31123).
Footnotes
-
↵* These Authors contributed equally to this paper.
- Received March 31, 2013.
- Revision received April 24, 2013.
- Accepted April 25, 2013.
- Copyright© 2013 International Institute of Anticancer Research (Dr. John G. Delinassios), All rights reserved