Metabolic Profiling of Formalin-Fixed Paraffin-Embedded Tissues Discriminates Normal Colon from Colorectal Cancer

Recent development of analytic techniques, such as chromatography and mass spectrometry (MS), has empowered metabolomics as an important field of omics studies. Metabolites are the downstream products of genomic, transcriptomic, and proteomic variability, thus providing useful information on biological conditions (1–4). Accumulating evidence indicates that cancer cells may reprogram cellular metabolism according to external or internal stimuli, such as nutritional starvation and extra- or intracellular stress, to survive in an adverse tissue microenvironment (5–9). Reprogrammed cancer metabolisms have also been suggested to facilitate nutrient production, including nucleotides, amino acids, and lipids, for cancer proliferation and adjustment to disadvantageous environment such as hypoxia and malnutrition. Metabolomics capture snapshots of various combinations of molecular alterations and complex interactions with host cells in the tumor microenvironment (10). Thus, through identifying metabolites and metabolic pathways with differential abundance between cancer and normal tissues, metabolomics could provide an insight into the establishment of novel biomarkers and the development of effective therapeutic strategies.

Tissue-based metabolomics research has mainly been conducted using fresh frozen materials (1, 11–17). However, a major downside of using fresh frozen materials is that many established cohorts lack sufficient fresh frozen samples. This limitation could be overcome by using formalin-fixed paraffin-embedded (FFPE) tissue materials, as these samples are obtained routinely for pathological diagnostic processes. Because FFPE tissues can be easily archived for decades with the corresponding patient data for retrospective assessment, FFPE tissues constitute widely available resources of clinical information, which is essential for retrospective molecular studies. Because of widespread availability and long-term stability of FFPE tissues, accurate profiling of their metabolite content would greatly increase the potential for discovery and validation of clinically useful biomarkers. In this study, we aim to evaluate the precision and performance characteristics of comprehensive metabolomics study using FFPE and matched fresh frozen samples of colorectal cancer by LC/MS-MS (liquid chromatography/mass spectrometry).

Eleven samples from radical colectomies were used in the study. All samples were collected with written informed consent and by approval from Brigham and Women's Hospital Institutional Review Board (Boston, MA). Matched FFPE and fresh frozen samples were collected from each colectomy. Tissue blocks were sectioned at 4 μm and were stained with hematoxylin and eosin (H&E) to identify tumor and normal area in each specimen. Sections of 20 μm were used for the metabolomics analysis. Histopathologic evaluation was performed to assess the tumor cell percentage in each sample.

The metabolite extraction from FFPE and fresh frozen samples was performed, as described previously (18, 19). Briefly, the metabolites from fresh frozen samples were extracted by incubating in 1 mL of 80% methanol at room temperature for 4 hours. After centrifugation at 14,000 × g for 10 minutes, the supernatants were collected and stored at −80°C until metabolic analyses. The metabolites from FFPE samples were extracted by incubating in 1 mL of 80% methanol at 70°C for 45 minutes in 1.5 mL microcentrifuge tubes without any deparaffinization procedure. The samples were placed on ice for 15 minutes and centrifuged at 14,000 × g for 10 minutes at 4°C. The supernatants were transferred into new 1.5 mL microcentrifuge tubes and placed on ice for 10 minutes, followed by centrifugation at 14,000 × g for 5 minutes at 4°C. The supernatants were collected and stored at −80°C until metabolic analyses.

Metabolite profiling was conducted by Metabolon Inc as described previously (20). All methods utilized Waters ACQUITY ultra-performance liquid chromatography (UPLC) and a Thermo Scientific Q-Exactive high resolution/accurate MS interfaced with a heated electrospray ionization source and Orbitrap mass analyzer operated at 35,000 mass resolution.

All statistical analyses were conducted using R software (version 3.5.3), and all P values were two-sided. Our primary aim was to assess the differentially abundant metabolites (DAM) between tumor and normal tissues and associated pathways enrichment. Principal component analysis was performed using the detected metabolites in FFPE and fresh frozen samples, individually. Two-dimensional plots constructed using first and second principal components as well as first and third principal components were used as visual aids for assessing sample quality. Permutation test was used for the identification of DAM whereby the null distribution was constructed using all the possible combination of randomly assigned tumor and normal labels to the 20 (or 22) FFPE (or fresh frozen) samples. Metabolites with P < 0.05 were considered significant and metabolites with P < 0.10 were used in metabolic pathways enrichment analysis. The pathway analysis module in MetaboAnalyst 4.0 was employed and hypergeometric test was used to test metabolites involved in a particular pathway are enriched compared with random hits. False discovery rate (FDR) was applied to reduce false-positive discovery in metabolic pathway enrichment analysis. The module also measures an impact score on each pathway based upon the position of metabolite(s), that is, a centralized position that is connected to many other entities is considered more impactful. As such, we considered pathways satisfying the conditions of number of hits > 1, and either -ln FDR > 2, or impact score > 0.1 as influential pathways.

We first examined and compared the metabolite profiles of the FFPE and fresh frozen samples using untargeted UPLC-MS; an overview of the experiment design is depicted in Fig. 1A. A total of 539 metabolites were detected from the samples. Among them, 197 metabolites were identified in both FFPE and fresh frozen samples, 339 in only fresh frozen samples, and 3 in only FFPE samples (Fig. 1B). We next categorized the identified metabolites according to the Human Metabolome Database. As shown in Fig. 1C, the ratio of detected metabolites in FFPE samples to that in fresh frozen samples varied according to the class of the metabolites. In general, similar metabolite profile was observed in both FFPE and fresh frozen samples, except for the classes of peptide and energy (Fig. 1D). There was no peptide detected in FFPE samples, whereas metabolites associated with energy were highly preserved in FFPE samples (78%; Fig. 1C). To investigate whether the preservation of metabolites in FFPE samples varied according to molecular weight of metabolites, we examined the correlation between metabolites molecular weight and amount detected in FFPE and fresh frozen samples (Fig. 2A and B). There was a stronger negative correlation between metabolites molecular weight and amount detected in FFPE samples (r = −0.16) compared with that in fresh frozen samples (r = −0.014). The detection rate of metabolites with larger than molecular weight of 500 in FFPE samples was 9.5%, whereas that in fresh frozen samples was 21.5% (P < 0.001, Fig. 2C). Approximately 37% of the metabolites detected in fresh frozen samples were detected in FFPE samples. We visualized in a heatmap that the abundance profile of the 197 common metabolites, which could be detected in both FFPE and fresh frozen samples, tended to be depleted in FFPE samples, as compared with that derived from fresh frozen samples (Fig. 1E). Similar analyses for the component subclasses, specifically 97 lipids, 53 amino acids, 18 nucleotides, 8 carbohydrates, 7 xenobiotics, 7 cofactors and vitamins, and 7 energy metabolites shared by FFPE and fresh frozen samples, are provided in Supplementary Figs. S1A–S1C and S2A–S2D. These results suggest that the reproducibility of metabolite detection in FFPE samples compared with fresh frozen samples varies according to the class of the metabolites.

We next investigated the difference of metabolite profiles by comparing the metabolite abundance between tumor and normal tissues in FFPE samples. As a quality assessment, we performed principal component analysis (PCA) using the 200 metabolites detected in FFPE samples (Fig. 3A; Supplementary Fig. S3A) and 536 metabolites detected in fresh frozen samples (Supplementary Fig. S3C and S3D). While the metabolic profiling derived from the tumor fresh frozen tissues were separated from the ones from normal fresh frozen tissues in the PC1-PC2 and PC1-PC3 space, the ones from two FFPE samples (P7_T_FFPE and P11_N_FFPE) exhibited different metabolite profiles compared with the ones from the others. Better separation between tumor and normal tissues within FFPE samples was achieved after excluding these two outlier samples (Fig. 3B; Supplementary Fig. S3B). As such, P7_T_FFPE and P11_N_FFPE were not included in downstream differentially abundant metabolites (DAM) and metabolic pathways analysis. A total of 193 metabolites were differentially abundant (P < 0.05) between tumor and normal tissues in either FFPE and fresh frozen samples (Fig. 3C). Among them, 15 metabolites were identified in both FFPE and fresh frozen samples, while 159 were exclusive to fresh frozen samples and 19 were only detected in FFPE samples. These 34 DAM belong to different classes of metabolites in which higher proportions of amino acid and energy-related metabolites were observed in FFPE samples (Fig. 3D). Among tumor-accumulated DAM, 6 metabolites, including branched chain amino acids (BCAA) isoleucine and leucine, were identified from both FFPE and fresh frozen samples (Fig. 3E). Among tumor-depleted DAM, 9 metabolites were identified in both FFPE and fresh frozen samples (Fig. 3F). Therefore, metabolite profile aberrations caused by tumorigenesis as captured in DAM were partially preserved in FFPE samples as compared with fresh frozen samples.

In addition to assessing individual DAM, we further examined the metabolite pathways enriched in the DAM. Our interpretation focused on the influential pathways which were selected on the basis of three conditions, that is, -ln FDR > 2 and/or impact score > 0.1, and hit > 1. There was one pathway, glyoxylate and dicarboxylate metabolism pathway, enriched in tumor-depleted DAM between FFPE and fresh frozen samples (Fig. 4A and B) and 3 pathways, which were enriched in tumor-accumulated DAM detected simultaneously in both FFPE and fresh frozen samples (Fig. 4C and D). These included arginine and proline, glycerophospholipid, and glycine, serine, and threonine metabolism pathways. Hence, differences of profiling metabolic pathways between tumor and normal tissue in FFPE samples were, in part, similar to that in fresh frozen samples.

Experimental evidence indicates that genetic and epigenetic alterations occurring during carcinogenesis alter metabolic programming in favor of cancer cell survival and growth in tumor microenvironment (4, 7, 8, 16, 17, 21–24). The detection of changes in metabolic profiles has the potential to advance early detection of cancer, as well as the prediction of prognosis or drug sensitivity (14, 25–28). In view of the availability of FFPE archival samples compared with fresh frozen samples, we aimed to assess the feasibility of utilizing the FFPE samples, in addition to fresh frozen samples, for metabolomics analysis. Our data demonstrated tumorigenesis-related metabolic profiling in FFPE samples and is, thus, suggestive of the potential of utilizing FFPE samples for metabolomics studies, particularly for small molecular metabolites, although fresh frozen sample-derived metabolite profile was only partially recapitulated from FFPE samples. Our data gives insight into feasibility of metabolomics analyses for molecular pathological epidemiology studies (29–31).

Accumulating evidence supports differential metabolic programming between tumor and normal tissues and the biological importance of metabolites (25, 27, 32). One remarkable example is the activation of glycolysis in cancer tissue, known as Warburg effect (33, 34). Interestingly, activated glycolysis in cancer cells appear to be beneficial to cancer cell survival and growth by several mechanisms, such as increased nucleotide production utilizing activating pentose phosphate pathway (21, 22). Consequently, depleted citric acid is replaced by utilizing alpha-ketoglutarate, which is metabolized to citric acid by IDH (isocitrate dehydrogenase) and ACO (aconitase; ref. 35). This study detected the depletion of alpha-ketoglutarate, which is one of the metabolites in tricarboxylic acid cycle, in tumor tissues. Alpha-ketoglutarate is also decomposed into 2-hydroxyglutarate by IDH, which often harbors point mutation in several cancers (4, 23). The depletion of alpha-ketoglutarate in this study may indicate the activation of citric acid synthesizing and IDH pathways. In addition to energy metabolites, the depletion and accumulation of amino acid in tumor tissues were robustly detected in both FFPE and fresh frozen samples, and this study suggests that BCAA including isoleucine and leucine are accumulated in tumor tissues. Because these essential ketogenic BCAAs are degraded directly into acetyl-CoA and succinyl-CoA in human mitochondria, their accumulation reiterates mitochondrial dysfunction in tumor cells to completely oxidize ketogenic substrates into carbon dioxide that is accompanied by inherent transfer of protons with less deuterium from nutrients to metabolic matrix water (36, 37). The fundamental mitochondrial function of curtailing deuterium oncoisotope accumulation in intermediary metabolites and nucleotides to prevent cell transformation has been argued as a surrogate marker for predicting response in personalized treatment and as a prognostic marker for patient survival in several clinical cancer studies (38, 39). Our study thus ingeminates again the clinical importance of insufficient mitochondrial deuterium depletion with resulting oncogenic transformation that can be unearthed from paraffin embedded tissue samples via metabolic profiling. Moreover, BCAT1 (BCAA transaminase 1) plays a role in maintaining the stemness of cancer stem cells in producing and accumulating BCAA, whereby BCAA enhances cancer progression (4, 7, 23–25). Therefore, this study demonstrated the profiling of amino acid and energy was partially recapitulated in FFPE samples. Further investigations using larger cohorts are required to clarify the exact mechanisms that underlie the relationship between metabolic alterations and carcinogenesis in colorectal cancer.

Conventionally, a fresh frozen tissue material, frozen immediately after surgical resection, is believed to have much better preservation of metabolites than a FFPE sample, which undergoes the fixation. In this study, the detectable metabolites in FFPE samples were approximately 40% of that in fresh frozen samples. Of note, large molecular metabolites, such as peptides, were undetected in FFPE samples. Our results are consistent with a recent prostate cancer study, which reported that FFPE samples have approximately 45% of metabolites compared with fresh frozen samples, but small molecular amino acids and a part of lipids are relatively well preserved and detected in FFPE materials compared with that in fresh frozen tissue materials (19). Therefore, these findings support the potential utility of FFPE samples for comprehensive analyses targeting small molecular metabolites.

We acknowledge that there were several limitations in our study. First, as our sample size was relatively small, the statistical power is low. Hence, our findings need to be validated in independent datasets with larger number of samples. Second, because tissue samples include both cancer cells and stromal cells, it would be impossible to test whether the detected metabolic changes were exclusively caused by the cancer cells. Third, the clinicopathologic data were limited. Because colorectal cancer represents a heterogeneous group of neoplasms, the metabolic profiles of colorectal cancer may differ by tumor location, disease stage, microsatellite instability status, and somatic mutation status. However, as long as both FFPE and fresh frozen tissue materials were collected from the same patient and from adjacent tumor location, the comparison study between FFPE and fresh frozen materials should serve its purpose. Fourth, we have excluded two outliers for DAM analyses because these outliers demonstrated metabolic profiles that were different from the other specimens based on the principal component analysis. These outliers might represent different tumor subtype, such as microsatellite instability-high; therefore, further research is warranted to investigate the association of tumor subtype with metabolic profiling. Fifth, more comprehensive analyses including treatment and clinicopathologic features using a larger sample size were not feasible in this study.

In conclusion, our study suggests the potential utility of FFPE samples in studying small-molecule metabolites in colorectal cancer and at the same time cautions its deficiency in preserving larger metabolites such as peptides. These findings need to be validated in additional large populations. Through leveraging the rich sources of FFPE archival samples, our study may help to shape future translational studies through developing strategies for colorectal cancer prevention and treatment through targeting metabolites.

R. Nishihara is a principal scientist at Pfizer. J.A. Meyerhardt is a medical advisory board member at COTA, Taiho, and grant review committee through NCCN and is an advisory board member at Ignyta. M. Giannakis reports receiving commercial research grants from Bristol-Myers Squibb and Merck. C.S. Fuchs is a director at CytomX Therapeutics and Agios, is a consultant at Unum Therapeutics, Amylin Pharma, Bain Capital, Daiichi-Sankyo, Eli Lilly, Entrinsic Health, Genentech, Merck, and Taiho, has ownership interest (including patents) in Cytomx Therapeutics, Entrinsic Health, and Evolveimmune Therapeutics, and has provided expert testimony for Eli Lilly and Amylin Pharma. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conception and design: K. Arima, M.C. Lau, K. Kosumi, A.T. Chan, C.S. Fuchs, S. Ogino

Development of methodology: K. Kosumi, S. Ogino

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Haruki, K. Kosumi, K. Mima, J.P. Väyrynen, T.S. Twombly, K. Fujiyoshi, W.G. Richards, A.T. Chan, M. Giannakis, C.S. Fuchs, S. Ogino

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Arima, M.C. Lau, M. Zhao, K. Kosumi, M. Gu, Y. Baba, C. Guo, H. Baba, A.T. Chan, R. Nishihara, J.A. Meyerhardt, S. Ogino

Writing, review, and/or revision of the manuscript: K. Arima, M.C. Lau, M. Zhao, K. Haruki, K. Kosumi, K. Mima, M. Gu, J.P. Väyrynen, T.S. Twombly, K. Fujiyoshi, J. Kishikawa, H. Baba, A.T. Chan, J.A. Nowak, M. Giannakis, S. Ogino

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Kosumi, T.S. Twombly, K. Fujiyoshi, J.A. Meyerhardt, S. Ogino

Study supervision: J.A. Nowak, C.S. Fuchs, S. Ogino

This work was supported by NIH grants (P50 CA127003, to C.S. Fuchs; R01 CA118553, to C.S. Fuchs; R01 CA169141, to C.S. Fuchs; R01 CA137178, to A.T. Chan; K24 DK098311, to A.T. Chan; R35 CA197735, to S. Ogino; R01 CA151993, to S. Ogino; R21 CA230873, to S. Ogino, K07 CA190673, to R. Nishihara); by Nodal Award (2016-02) from the Dana-Farber Harvard Cancer Center (to S. Ogino); by Stand Up to Cancer Colorectal Cancer Dream Team Translational Research Grant (grant number SU2C-AACR-DT22-17; to C.S. Fuchs and M. Giannakis); and by grants from the Project P Fund, The Friends of the Dana-Farber Cancer Institute, Bennett Family Fund, and the Entertainment Industry Foundation through National Colorectal Cancer Research Alliance. Stand Up To Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C. K. Arima was supported by grants from Overseas Research Fellowship from Japan Society for the Promotion of Science (201860083). K. Haruki was supported by a fellowship grant from the Uehara Memorial Foundation and the Mitsukoshi Health and Welfare Foundation. K. Fujiyoshi was supported by a fellowship grant from the Uehara Memorial Foundation. A.T. Chan is a Stuart and Suzanne Steele MGH Research Scholar. We thank the Department of Surgery, Brigham and Women's Hospital for providing us with tissue specimens.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.