Proteomic analysis of FFPE tissue: barriers to clinical impact

KEYWORDS:

• Proteomics
• formalin-fixed, paraffin-embedded (FFPE) tissue
• preanalytical factors
• protein extraction
• bioinformatics

1. Proteomic analysis of archival FFPE tissues

Archival formalin-fixed, paraffin-embedded (FFPE) tissues and their associated diagnostic records represent an invaluable source of retrospective proteomic information on diseases for which clinical outcome and response to treatment are known [1]. Over the past few years, advances in methodology have made it possible to recover proteins or peptides from FFPE tissue, yielding a reasonable representation of the proteins recovered from matched fresh or frozen specimens. Early qualitative approaches have given way to quantitative proteomic analysis of FFPE tissues [2]. An encouraging recent trend is the validation of quantitative FFPE proteomics findings using orthogonal methods such as immunohistochemistry and immunoblotting [3,4]. A study by Ostasiewicz et al. [5] combined quantitative proteomic analysis of FFPE colorectal carcinoma specimens with messenger RNA gene expression analysis using PCR. These developments suggest that FFPE proteomics has matured sufficiently that translation of FFPE proteomic biomarker findings into clinical assays is on the horizon. This editorial focuses on the barriers that must be overcome for FFPE tissues to impact clinical molecular diagnosis. An overview of FFPE tissue proteomics is provided in the excellent review of Gustafsson et al. [2].

2. The role of preanalytical factors in FFPE proteomic analysis

The ability to study FFPE archival tissue by proteomic methods is now clearly established. What is less clear is how accurately proteomic analyses reflect the in vivo tissue proteome rather than artifacts introduced in subsequent sample preparation steps. Preanalytical factors, including warm and cold ischemic time, tissue size, fixation conditions, tissue processing, and storage time and conditions, can have a dramatic impact on proteomic analysis of FFPE tissue [6]. A further aspect rarely considered is biological (patient) variability in tissue proteomes. The importance of these preanalytical factors to the meaningful translation of proteomic methods and findings to clinical practice has been eloquently discussed in a review by Becker [7]. The few systematic studies addressing these factors have yielded contradictory results. For example, top-down proteomic studies report up to a 50% loss in recoverable protein from FFPE tissue blocks stored for 20 years [8], while bottom-up studies report little effect of storage time on protein identification by mass spectrometry (MS) [9]. One might question the value of considering preanalytical factors for FFPE proteomics because information on sample preparation is rarely available. One reason to consider preanalytical factors is to identify proteomic analyses for which FFPE archival tissue would yield misleading results. Phosphoproteins levels, for example, drop rapidly between excision and fixation with alterations of ±20% for time-to-fixation periods of 4 to 40 min [10]. Investigation of phosphorylation status of proteins is one case where the use of archival tissue would yield unreliable information. A second reason is to determine the variability introduced in proteomic analyses by preanalytical factors in order to establish statistical parameters for interpreting proteomic data from FFPE tissue. This approach is illustrated in a study of FFPE extraction buffers by Broeckx et al. [11]. Using murine FFPE tissue, the authors measured an overall coefficient of variation of 48% for the reproducibility of the MS-based proteomic analysis. The MS analysis step contributed 12% to the overall variation, and the preanalytical factors contributed the remaining 36%. At this level of variation, a power analysis concludes that about 100 samples would be required to detect a change of 50% in biomarker levels with 90% certainty. The authors found a total sample variation (sample preparation and biological variation) of at least 68% for human colon FFPE tissue specimens. Studies such as this establish that rigorous systematic investigation of FFPE total sample variation will establish a foundation for statistically reliable analysis of proteomic results, despite a lack of knowledge about sample preparation or patient origin.

3. Workflow for FFPE proteomic analysis

Given the large total sample variation of FFPE tissues, it is essential that all steps under control of the researcher be standardized. Of these steps, the most critical is protein extraction from the FFPE tissue sample. There is currently no consensus on the optimal protocol for extraction or accepted standards for quantitative evaluation of the extracts. There are few studies comparing the efficacy of the numerous protein extraction protocols that have appeared in the literature. However, these comparative studies have found that the quantity and identity of the recovered proteins depend upon the extraction buffer and method used [11,12]. The principal conclusion drawn from these comparative studies is that no one extraction buffer is ideal for all proteins and all tissue types. It has been suggested by Shi et al. [13] that ensuring complete solubilization of tissue samples, so that no reside remains, is the best way to avoid extraction bias and achieve the goal of standardizing the recovery of proteins from FFPE tissues. This may be achieved by sequential extraction of the tissue pellet using buffers differing in pH, ionic strength, and detergent. Clearly more work is required to develop evidence-based guidelines to ensure quantitative and qualitative reproducibility in the recovery of proteins from FFPE tissues.

4. Bioinformatics analysis of FFPE proteomic data

It is generally assumed that the application of antigen retrieval methods to FFPE tissues results in the reversal of most formaldehyde-induced cross-links and adducts that would otherwise prevent the identification of proteins through their proteolytic peptides. Top-down methods such as gel electrophoresis that are based on the separation of intact proteins provide a different interpretation [14]. Studies of FFPE tissue protein extracts by 2-D difference gel electrophoresis exhibit high molecular weight complexes arising from residual cross-linked proteins. Relative to fresh tissue, FFPE gel profiles also exhibit an acid shift correlated to protein isoelectric point values and a reduction in spot intensity correlated to protein molecular weight and lysine content [15]. These results indicate the presence of residual formaldehyde adducts on lysine and possibly other amino acid residues. The existence of residual formaldehyde-induced protein adducts suggests that protein coverage can be improved by incorporating protein modifications induced by formaldehyde into the bioinformatics analysis of FFPE mass spectral data. This can be accomplished using two bioinformatics approaches. The more general approach uses an unrestricted modification search to extract the modifications in the sample directly from the MS/MS data without the need for prior configuration [16]. Using this method, unexpected modifications are detected in an unbiased manner. The second approach incorporates known formaldehyde-induced protein adducts into the posttranslational modification search module when analyzing the mass spectral data [11,17]. When applied to FFPE mass spectral data, these bioinformatics enhancements improve peptide identification by up to 8%. Although the unrestricted modification search is the more robust approach, introducing known formaldehyde-induced protein modifications into the search algorithm helps reduce both the search space and false positive identifications. A greater understanding of formaldehyde-induced protein adducts in FFPE tissues would improve these bioinformatics methods and enable the identification of more proteins and reduce false positive identifications. Unfortunately, most studies of protein-formaldehyde chemistry have been restricted to aqueous solutions. It is likely that formaldehyde-induced protein adducts undergo further reactions during histological processing. Using an unrestricted modification search method, Zhang et al. [16] identified the presence of methylated lysine in FFPE proteomes: a reaction not seen in aqueous solutions of protein and formaldehyde. Ethoxymethylation of the primary amine is observed in 2′-deoxyadenosine 5′-monophosphate when the nucleotide is transferred to ethanol after reaction with formaldehyde in aqueous solution [18]. Clearly, additional studies are necessary to fully characterize the spectrum of formaldehyde-protein adducts in FFPE tissues.

5. Conclusions

The above discussion describes the workflow that starts at the time of excision of tissue from the patient and ends with the proteomic analysis of the archival FFPE specimens derived from the tissue. Many of the preanalytical variables within this workflow are beyond the control of the researcher. While this may limit proteomic studies of certain posttranslational modifications, meaningful proteomic results can be obtained by observing the statistical constraints imposed by variability in sample preparation and patient biological variation. The remaining aspects of the workflow under the control of the researcher should be standardized to the extent possible. One impediment to achieving standardization is the small community of researchers with a long-standing interest in the elements of the FFPE workflow rather than just the proteomic outcome [19]. One way to overcome this obstacle is to establish a consortium of interested researchers organized under an umbrella organization such as the Biosciences Research Network of the National Cancer Institute (NCI). Many of the studies of formaldehyde-protein chemistry and antigen retrieval were sponsored by the Innovative Molecular Analysis Technologies program of the NCI. A coordinated research effort by the members of this consortium would accelerate the research necessary to develop evidenced-based best practices for the proteomic analysis of FFPE tissue and translation of these findings into meaningful clinical assays.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

The author was supported by the U.S. Department of Veterans Affairs [VA01-0109201].