Toward Improving the Proteomic Analysis of Formalin-fixed, Paraffin-embedded Tissue

Carol B Fowler; Timothy J O'Leary; Jeffrey T Mason


Expert Rev Proteomics. 2013;10(4):389-400. 

In This Article

Workflows for Proteomic Analysis of FFPE Tissue

In recent years, the identification of formaldehyde-induced protein modifications[16,17] and a deeper understanding of the mechanisms of cross-link reversal[43,44] have led to the development of improved protein recovery techniques for FFPE samples. The majority of these extraction techniques are based upon the principles of heat-induced antigen retrieval,[2,45] and yield significant recovery of intact proteins[2,4,30] or tryptic peptides[46] from FFPE tissue. This has led to the wide range of available methods for analyzing archival tissues that are summarized in the workflows shown in Figures 1 & 2. Currently, there are three main workflows used for the proteomic analysis of FFPE tissues. The first two workflows start with the recovery of proteins from FFPE tissue using an extraction buffer and heating protocol. In the first workflow, the extracted proteins are digested using proteolytic enzymes and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). In the second workflow, the extracted whole proteins are analyzed directly by methods such as 2D polyacrylamide gel electrophoresis (2D PAGE), reverse-phase protein microarrays (RPPA) or immunoassays. The third workflow involves the treatment of intact FFPE tissue on slides with proteolytic enzymes followed by direct analysis using MALDI-IMS.

Figure 1.

A typical workflow for tissue section-based isolation of proteins from formalin-fixed, paraffin-embedded tissues. A tissue block with a surface area of 5–10 mm2 is sectioned at 10–50 microns serial sections 1–3) are placed in a capped vial and the tissue is cleared by deparaffinization using two xylene extractions followed by rehydration through a graded series of ethanol solutions terminating in water. The tissue pellet is isolated by centrifugation followed by the addition of AR buffer. Small tissue samples isolated by LCM, as described in Figure 2, enter the workflow at this point. The tissue is then homogenized in the AR buffer by vortexing or sonication on ice. For large samples or rigid fibrous tissues, homogenization can be preceded by grinding the tissue using a mortar and pestle. The homogenized tissue is then subjected to antigen retrieval using one of the methods described in the text and the solubilized proteins are isolated by centrifugation. The resulting supernatant can be used directly for whole-protein based top-down proteomics studies. For bottom-up proteomics, the supernatant is further processed to remove ionic detergent, such as SDS. The proteins are then denatured, reduced and alkylated and treated with trypsin to generate peptides. Following desalting using a reverse-phase media (C18), the peptides are typically analyzed by liquid chromatography-tandem mass spectrometry or matrix-assisted laser desorption-time of flight.AR: Antigen retrieval; LCM: Laser capture microdissection; SDS: Sodium dodecyl sulfate.

Figure 2.

A typical workflow for tissue slide-based isolation of proteins from formalin-fixed, paraffin-embedded tissues. A tissue block is sectioned at 5–10 microns and placed on a glass microscope slide. If the tissue will be used for IMS, a conductive slide coated with indium-tin oxide is used. The tissue is then cleared by deparaffinization using two xylene extractions followed by rehydration through a graded series of ethanol solutions terminating in water. The slide is then placed in AR buffer and subjected to antigen retrieval using one of the methods described in the text. The workflow then splits into two paths. For LCM (blue arrows), the target cells are identified by comparison with a conventionally stained serial section and then removed using a LCM instrument or a hypodermic needle. Typically, 10,000–200,000 cells are collected and placed in a capped vial. The LCM sample then enters the workflow at the point shown in Figure 1. For IMS (green arrows), the slide is coated with trypsin to digest the constituent proteins. Coating is accomplished using aerosol vaporization or acoustic droplet ejection. An internal calibrant (peptide standards) can be coated onto the tissue to correct time-of-flight values for variations in tissue thickness, or alternately, an external calibrant spot may be applied to a clear, tissue-free area of the slide. Finally, the tissue is coated with matrix such as α-cyano-4-hydroxycinnamic acid and analyzed by MALDI-IMS.AR: Antigen retrieval; IMS: Imaging mass spectrometry; LCM: Laser capture microdissection; MALDI-IMS: Matrix-assisted laser desorption/ionization imaging mass spectrometry.

Liquid Chromatography-tandem Mass Spectrometry

In recent years, many workflows for LC-MS/MS have employed detergent-assisted extraction of FFPE tissue, followed by a sample preparation step to remove detergents and other buffer components that are not compatible with MS. Shi et al. adapted their heat-induced antigen retrieval procedure to compare extracts from frozen and FFPE tissue sections taken from the same human renal cancer biopsy.[2] Proteins were extracted by heating whole tissue sections in 10 mM Tris-HCl, 2% (w/v) SDS buffer at 100°C for 30 min followed by incubation at 60°C for 2 h. LC-MS/MS identified 2404 and 3236 total proteins in the frozen and FFPE specimens, respectively. Similar studies have been reported by Jiang et al.[36] and Xu et al..[47] Azimzadeh et al. compared several extraction buffers containing CHAPS, Tween-20 or combinations of detergents such as SDS and NP-40.[5] They found that a 20 mM Tris buffer, pH 8.8, containing 2% (w/v) SDS, 1% (w/v) β-octylglucoside, 200 mM dithiothreitol (DTT), 200 mM glycine and protease inhibitors gave the highest protein yields. Protein separation by 1D SDS-PAGE and in-gel tryptic digestion of the excised gel bands was the most reproducible sample preparation method tested for LC-MS/MS protein identification. Addis et al. heated whole tissue sections in a 20 mM Tris buffer pH 8.8, with 2% (w/v) SDS and 200 mM DTT and analyzed the extracts by 1D SDS-PAGE; western blot and LC-MS/MS were performed on tryptic digests of the excised gel bands.[48] More recently, Tanca et al. used this extraction methodology, now known as GeLC-MS/MS, to evaluate matched FFPE and fresh-frozen canine mammary carcinomas and their normal controls by LC-MS/MS. Spectral counting was used to compare expression of specific proteins between samples, and results showing differential expression trends were similar for frozen and FFPE samples. However, higher molecular weight proteins were underrepresented in the FFPE samples.[49]

Wiśniewski et al. developed a method which allows in-solution tryptic digestion of the FFPE tissue extract, followed by filter-aided sample preparation (FASP).[50] After heat-assisted extraction in a 100 mM Tris-HCl, pH 8.0 buffer containing 4% (w/v) SDS and 100 mM DTT, the protein extract was washed multiple times against 8 M urea in 0.1 M Tris-HCl pH 8.5 using an ultracentrifugation spin filter (30 kDa molecular weight cut off) to remove the detergent and DTT. The proteins were alkylated with iodoacetamide and digested with trypsin to release the peptide fragments from the spin filter, followed by peptide identification by LC-MS/MS. This method has been applied to identify phosphorylated and N-glycosylated proteins in FFPE mouse liver tissue sections isolated by LCM.[4] The addition of consecutive incubations with the proteolytic enzymes LsyC and trypsin increased the number of proteins identified by LC-MS/MS by up to 40%.[51,52]

Other strategies have been developed to improve protein extraction from FFPE tissues. The addition of elevated pressure (40,000 psi) to heat and detergent-assisted protocols improved both the protein extraction efficiency (from 26 to 96%) and the reversal of formaldehyde-induced adducts in a five-protein tissue surrogate.[34] When FFPE mouse liver sections were extracted using a combination of heat and elevated pressure (40,000 psi), there was a fourfold increase in protein extraction efficiency, a threefold increase in the extraction of intact proteins and up to a 30-fold increase in the number of non-redundant proteins identified by MS, compared with FFPE tissue extracted with heat and detergent alone. More importantly, the number of non-redundant proteins identified in the FFPE tissue was nearly identical to that of the matched fresh-frozen sample.[30] More recently, Fu et al. reported that the combination of heat and high pressure improved protein yield and the number of proteins identified from FFPE aorta.[53]

A number of groups have also reported proteomic studies on FFPE using buffers that do not contain SDS or other ionic detergents, such as Liquid Tissue™,[3,46,54–56] which is commercially available. Commonly, tissue sections on slides were cleared of paraffin and rehydrated through graded alcohols. After an optional staining step with H&E, several thousand cells were isolated by needle dissection or LCM. The cells were then heated (usually at 95°C for up to 90 min) in retrieval buffer, followed by overnight digestion with porcine trypsin and the clarified extract was reduced with DTT. This extraction methodology is usually reserved for relatively small amounts of material (i.e., cells collected by LCM) that may not always represent the full complexity of a tumor. However, extracts obtained in this manner can be analyzed by LC-MS/MS without any subsequent purification steps. Naidoo et al. extracted FFPE pancreatic ductal adenocarcinoma primary tumors and matched lymph node metastases using the Liquid Tissue kit, followed by multi-dimensional LC-MS/MS. Comparison of the proteome from the FFPE samples with data from prior studies showed a considerable number of proteins that had not been previously identified in pancreatic juice, serum and urine.[54] Takadate et al. also utilized this extraction methodology to perform targeted selective reaction monitoring MS (SRM-MS) on FFPE node-positive pancreatic ductal carcinoma. Three novel proteins were overexpressed in poor outcome groups, ECH1, OLFM4 and STML2.[56] Other buffers and methodologies have also been utilized for FFPE tissue. Paulo et al. heated FFPE pancreatic tissue in 6 M guanidine-HCl, 50 mM ammonium bicarbonate, 20 mM DTT, pH 8.5 at moderate heat (70°C) for 1 h, followed by alkylation and tryptic digestion.[57] Alternately, Alkhas et al. incubated LCM endometrial cancer samples in 100 mM ammonium bicarbonate, 20% (v/v) acetonitrile at 95°C for 1 h followed by 65°C for 2 h prior to tryptic digestion.[58] Finally, Wisztorski et al. recently developed a novel extraction methodology employing on-tissue digestion.[59] An intact tissue section on a slide was subjected to heat-induced antigen retrieval, and trypsin was then directly deposited onto the tissue using a chemical inkjet printer. The peptides were then removed by liquid microextraction prior to analysis by LC-MS/MS. Because the trypsin was deposited onto defined areas, they were able to identify 983 non-redundant protein groups in the benign region of the tissue and 792 proteins in the cancerous region.[59] A similar extraction method was reported by Hatakeyama et al..[60]

Quantitative & Targeted Proteomics

Advances in protein extraction methodology, separation technology and MS instrumentation have advanced qualitative proteomics to the point where thousands of proteins can be identified from a single FFPE tissue section. Further advances in technology, such as isotope coded-affinity tags[61] have made quantitation of differentially expressed proteins in FFPE tissue possible. One workflow for quantitation utilizes the commercially available iTRAQ™ kit,[62] which consists of a set of multiplexed isobaric reagents that label amine groups. Jain et al. identified nine proteins that were differentially expressed in FFPE oral lesions in HIV positive (+) and HIV negative (-) patients using this method.[63] Tryptic peptides from two groups of HPV+/HIV+ samples were labeled with iTRAQ tags 116 and 117, and two groups of HPV+/HIV- samples were labeled with iTRAQ tags 114 and 115. MS/MS generated reporter ions could then be used to identify and quantify peptides labeled with each tag. iTRAQ labels have also been used to identify differentially expressed protein in nasopharyngeal carcinoma.[64] More recently, iTRAQ was used to compare proteomes from fresh-frozen and FFPE rat spinal cords with experimental autoimmune encephalomylelitis.[65]

Though labeling approaches have gained favor in proteomics, with over 150 published studies using the iTRAQ reagents alone, FFPE tissues present special challenges to quantitative proteomics. Formaldehyde modifies lysine and arginine residues, which are primary labeling sites for isotope affinity tags. As discussed above, variations in fixation time and tissue quality can complicate tissue extraction and reversal of formaldehyde-induced adducts and cross-links. Several groups reported proteomic studies for FFPE tissue using quantification methods that do not rely upon stable isotope levels.[3,4,11,66] For example, spectral counting MS determines relative protein concentrations by counting the number of spectra identified for a peptide of interest and calculating a normalized spectral abundance factor. Spectral counting has been used to compare relative protein abundance in normal and cancerous head and neck squamous epithelium[3] and more recently to identify differentially expressed proteins in FFPE metastatic melanoma tissue.[66] Another strategy, label-free quantification, which is usually carried out on orbitrap or Fourier transform ion cyclotron mass analyzers, detects and extracts peptide signals at the MS1 level, matches the peptides across multiple LC-MS/MS scans to normalize peptide intensities, and then selects quantifiable discriminatory features, whose intensities are different among groups of samples.[67] Recently, Azimzadeh et al. described this computational method to detect deregulated proteins in irradiated mouse heart. The data for FFPE tissue were in agreement with quantitative data obtained from fresh-frozen tissue.[11]

Another promising method for quantitative proteomics by MS/MS is SRM or multiple reaction monitoring (MRM) MS,[10,55,68–70] which is most commonly carried out on triple quadrupole mass spectrometers. In SRM, a peptide precursor for a protein of interest is first isolated by MS and then fragmented to yield product ions whose signal abundances indicate the relative abundance of the peptide in the sample. In MRM, multiple transitions (precursor/product ion pairs) are monitored. DeSouza et al. extracted laser dissected FFPE tissue sections with the Liquid Tissue kit.[10] The tryptic peptides were labeled with a pair of light and heavy mTRAQ™ reagents, a non-isobaric variation of the iTRAQ labels. LC-MRM MS was then used to quantitate differential expression levels of polymeric Ig receptor and pyruvate kinase isoform M2 in FFPE endometrial carcinoma tissue and matched frozen tissue from the same tumor. MRM quantification was also used to correlate napsin-A and hAG-2 proteins with patient outcome in stage IA and IIIA lung adenocarcinoma.[55,69] More recently, Guzel et al. developed an MRM assay for calcyclin as a biomarker for pre-eclampsia.[68] Two synthetic calcyclin peptides with extra glycines were spiked into the sample extract as internal references. Finally, Gamez-Pozo et al. employed SRM monitoring to quantitate phosphopeptides in FFPE renal cell carcinoma and non-small cell lung carcinoma biopsies.[70] Comparable phosphorylation rates were found in matched FFPE and fresh-frozen samples.


The extraction of non-degraded, full-length immunoreactive proteins from FFPE tissues was first reported by Ikeda et al.[9] and Becker et al..[71] Becker et al. successfully measured HER2 in extracts of FFPE breast cancer biopsy specimens using western blots and RPPA. This result was verified by detection of the receptor by IHC on matched FFPE tissue slides and by detection of the HER2 transcript using FISH. 2G-PAGE analysis of protein extracts from skeletal muscle and liver were reported in two articles by the Addis et al..[72,73] The FFPE gel revealed 250 protein spots compared with 400 for the matched fresh tissue gel using silver staining. More importantly, the protein spots on the FFPE gel matched those on the fresh tissue gel by western blot and LC-MS/MS following in situ enzymatic hydrolysis.[72] Using 2D difference gel electrophoresis (2D DIGE), it was shown that the protein spots on the FFPE gel exhibited an acid shift that directly correlated with their pI values and a reduction in spot intensity that directly correlated with protein molecular weight and density of lysine residues.[73] These patterns were highly reproducible, which led the authors to suggest that 2D-PAGE of FFPE tissue extracts were best suited to differential proteomic investigations of archival tissues. Recently, the authors' group found that extraction of FFPE mouse liver using heat supplemented with elevated pressure (40,000 psi) yielded a threefold increase in the extraction of non-degraded, full-length proteins compared with matched tissue extracted with heat alone.[30] The ability to extract intact proteins from FFPE tissues increases the diagnostic and prognostic efficacy of proteomic-based biomarker discovery by allowing biomarker validation using orthogonal methods such as western blotting, IHC, immunoassays and possibly structural and interaction proteomic studies.

Protein Microarrays

Though LC-MS/MS-based workflows can identify thousands of proteins from a single archival sample, tandem MS is not a high-throughput methodology and thus may be more suited to biomarker discovery than to the clinical laboratory. MS is also an expensive, instrument-intensive methodology that requires specialized training. Several recent studies have shown that RPPAs have the potential to fill this gap.[74–81] The RPPA format allows lysates from multiple patient samples to be arrayed on nitrocellulose-coated slides, and the sandwich-antibody format is familiar to clinical laboratories which routinely perform IHC. Protein arrays on untreated glass slides or carbon-coated plates for an electrochemiluminescence-based detection system[37] have also been reported. Several protein–antibody interactions can be detected in a single experimental run, thus allowing for the analysis of complex protein networks. Since the samples are spotted in a dilution series and in triplicate, it is possible to quantify protein expression when normalized against purified proteins spotted on the same slide. Ideally, a reference standard should be included to serve as a positive control for antibody staining and to control for inter-slide variability.[77] Protein expression profiles can be determined more accurately than by IHC, and results from RPPAs can complement gene microarrays and anatomical pathology in the clinic.[82]

Guo et al. noted that even when extracted proteins were partially degraded, the overall signal from FFPE tissues was comparable with frozen samples.[74] As mentioned above, Becker et al. were able to show that HER2 levels could be reproducibly quantitated in FFPE breast cancer biopsies,[71] and more recently, Wulfkuhle et al. reported levels of total and phosphorylated HER2 in agreement with FISH and IHC data.[80] Berg et al. have also used protein microarrays to demonstrate correlations of HER2 levels with other biomarkers, such as HER3, epidermal receptor 1 and urokinase plasminogen receptor,[78] as well as progesterone receptor and estrogen receptor.[79] RPPA was also used to examine the expression levels of 17 cancer-related molecules in esophageal adenocarcinoma. Correlation of protein expression levels with clinical outcome data identified a new molecular subtype of esophageal adenocarcinoma that expressed low levels of HSP27 proteins and high levels of HER proteins.[76] Assadi et al. were also able to show the RPPA signal intensities for HER2 correlated well with ordinal scores obtained by IHC for the same tissue specimens. An ordinal multinomial logistic regression model showed good correlation between RPPA signal intensity and IHC scores of 2+ and 3+, while correlation with IHC scores of 0 and 1+ were less reliable.[81] These results indicate that RPPAs could be integrated into the clinical setting for both biomarker validation and analysis of patient samples to help predict clinical outcome and response to treatment. However, successful implementation of RPPAs in the clinic requires efficient extraction of high-quality proteins from FFPE tissue, and highly specific antibodies that represent binding epitopes under denaturing conditions (i.e., linear epitopes).[44,74] For accurate quantitation of proteins, it is especially important to include protein standards and to demonstrate stoichiometric binding of antibodies.[77,79]

MALDI-IMS as a Tool for Morphology-based Proteomic Tissue Analysis

While LC-MS/MS can identify thousands of proteins in a single tissue sample, and quantitative MS and RPPAs can be used to identify and quantitate differentially expressed proteins, these analyses are performed on protein extracts. Thus, differential protein expression cannot be localized to specific tissue regions and data cannot be directly related to IHC. In MALDI-IMS, the morphological distribution of individual peptides, proteins and small molecules can be visualized. This is particularly useful in differentiating areas of the tissue that are not histologically distinct, but do exhibit different protein expression profiles. With recent progress in instrumentation and informatics software, MALDI-IMS has emerged as a promising tool for tissue-based proteomics and biomarker discovery.[83] In MALDI-IMS, a tissue section, normally fresh-frozen, is mounted onto a microscope slide with an electrically conductive indium-tin oxide coating. After the tissue section is washed with a solvent, such as ethanol or acetone to remove salts and any other non-desired molecules, such as lipids, matrix solution is applied.[83] The matrix may be applied using an automated spotter or sprayer, and varies by application. α-cyano-4-hydroxycinnamic acid (CHCA) is normally employed for peptides and small molecular-weight compounds, while 2,5-dihydroxybenzoic acid (DHB) is useful for larger molecular weight molecules, such as small proteins. Imaging software is then used to register the dimensions of the tissue section and MALDI-time of flight (MALDI-TOF) spectra are acquired for each x, y coordinate; laser raster spots are generally 50–200 µm in diameter. The MALDI data are then compiled into a 2D ion density map that shows the localization of individual m/z peaks.

MALDI-IMS of FFPE tissue requires several additional preparative steps. In early studies, FFPE tissue sections were cleared of paraffin with xylene, rehydrated through a graded series of alcohols and then digested with trypsin.[84,85] Ronci et al. showed that adding a heat-induced antigen retrieval step after deparaffinization increased the number of peptide identifications. When FFPE breast cancer specimens were left untreated, were heated in 0.1 M EDTA or were treated with trypsin alone, eight, 70 and 22 proteins were identified, respectively.[86] Other groups have also added an antigen retrieval step to the tissue preparation workflow. After deparaffinization and rehydration in alcohol and ammonium bicarbonate, Gustafsson et al. boiled FFPE tissue sections in 10 mM citric acid buffer, pH 6, for 10 min and then incubated the slide at 98°C for 30 min prior to tryptic digestion.[87] Groseclose et al. heated FFPE tissue microarray (TMA) sections in 10 mM Tris buffer, pH 9 prior to tryptic digestion and MALDI-IMS.[88] More recently, Casadonte and Caprioli published a complete protocol for FFPE tissue preparation, antigen retrieval in Tris buffer, and MALDI-IMS of a 300 tissue core TMA.[89] The addition of an antigen retrieval step to the MALDI imaging workflow has greatly improved peptide ion detection and made biomarker discovery and validation using FFPE tissue by this technology possible. However, with improved detection come two challenges: determining which m/z peaks are of interest and determining protein identification. A typical MALDI-IMS data set comprises thousands of MS spectra; for a review on currently available computational methods, including unsupervised data mining methods such as principal component analysis and spatial segmentation, see.[90] Selected m/z ions are then subjected to MS/MS (typically MALDI-TOF/TOF), and the MS/MS spectra are searched using a database such as MASCOT to identify proteins of interest. Instrument sensitivity and the variability of TOF m/z measurements[91,92] are two major challenges to the unambiguous identification of MS/MS spectra in situ. The variability of the TOF analyzer is most typically addressed by calibrating the instrument with an external calibration standard prior to each MALDI-IMS experiment. However, differences in tissue section thickness may affect mass accuracy, especially when using a calibration spot external to the tissue section. Gustafsson et al. addressed this challenge by depositing a peptide calibration mixture directly onto the tissue section prior to MALDI imaging. All spectra were calibrated to internal peak features to reduce m/z errors between MALDI-IMS values and values for tryptic peptides previously determined by LC-MS/MS.[92] With improvements to tissue preparation, and computation, MALDI-IMS is rapidly becoming accessible for FFPE tissue. By combining the ability to visualize the morphological distribution of peptides, lipids and other molecules in a tissue section with MS-based molecular identification, MALDI-IMS is emerging a complement to diagnostic pathology and LC-MS/MS. In MALDI-IMS, the spatial distribution of peptide fragments are often used as surrogate biomarkers in place of the parent protein. This approach must always be validated by ensuring that the distribution of the peptide fragment matches that of the parent protein and that the peptide fragment distribution is confirmed by an orthogonal method, such as IHC.

Mechanisms of Protein Recovery from FFPE & Their Implications

The authors' group has investigated the mechanism of heat-induced antigen retrieval and the recovery of proteins from FFPE tissues. Studies of single proteins in 10% formalin revealed that, following removal of excess formaldehyde, proteins with normal structure, function and immunoreactivity could be recovered by heating at 60°C for 30 min.[93,94] Additional studies with single proteins and tissue surrogates revealed that ethanol-dehydration of formaldehyde-treated proteins during tissue processing further confounds the recovery of proteins and antigens from FFPE tissue for IHC and proteomic analysis. Ethanol was found to promote conformational changes in formaldehyde-fixed proteins resulting in their forming amyloid-like protein aggregates stabilized by formaldehyde cross-links, hydrogen bonds and van der Waals interactions. Substantial energy was required to rehydrate this aggregated protein network in order to reverse formaldehyde adducts and cross-links.[7,30,34] This conclusion was further supported by studies indicating that heat-induced antigen retrieval causes protein unfolding, which supports a linear epitope model of recovered immunoreactivity.[44] These findings help explain the role of heat, detergents, protein denaturants, reducing agents, sonication[95] and pressure[33] in promoting protein recovery from FFPE tissues as they all promote the physical disruption of protein aggregates, protein denaturation and penetration of water into the protein core. These findings also suggest that increasing the surface area to volume ratio of FFPE tissue specimens is likely to result in better protein recovery. Consequently, specimens consisting of small numbers of cells collected by LCM yield more recovered protein per tissue weight than do whole tissue sections.[8] Although LCM specimens are suitable for evaluation of known biomarkers, the discovery of new protein biomarkers is likely to require larger FFPE specimens. Ultimately, MALDI-IMS may represent the best compromise between proteomic assessment of large FFPE tissue sections and good protein recovery.

Studies of the reaction of formaldehyde with whole proteins indicate that accessibility plays a significant role in the formation of cross-links.[18] A probabilistic model of formaldehyde-induced adducts and cross-links in insulin by Metz et al. suggests that normal conformational fluctuations of the protein in solution result in a heterogeneous population of insulin molecules; each with a unique pattern of formaldehyde modifications.[17] These studies have important implications for the proteomic analysis of proteolytic digests of FFPE protein extracts. Among the set of redundant proteins that constitute a protein's total population, it is highly likely that every tryptic cleavage site and every tryptic fragment contained within the protein will be present in a form devoid of formaldehyde adducts or cross-links in a subset of this population. Consequently, complete reversal of all formaldehyde adducts and cross-links may be unnecessary in order to correctly identify the presence of the protein in the FFPE tissue extract. This is advantageous for qualitative proteomic analysis of FFPE tissues, but it also requires that stringent conditions be used for protein identification. For example, the overlap of proteins identified as being present in both FFPE and matched fresh tissue decreases significantly when two constitutive peptides are required for protein identification rather than one.[28] This fact also complicates quantitative analysis of FFPE tissues as many formaldehyde-modified peptides will not be identified during database searches. Three approaches can be taken to minimize this complication when performing quantitative analysis. The first approach is to ensure complete extraction of the FFPE tissue to avoid extraction bias and maximize the total protein for analysis.[8] This approach may require multiple extractions of the sample using the same or different buffers. The second approach is to investigate quantitative differences in protein expression by comparing sets of FFPE tissues rather than comparing FFPE tissues with matched fresh or frozen tissues. The third approach is to verify results derived from the analysis of FFPE tissues by LC-MS/MS and MALDI-IMS with orthogonal methods, such as IHC, western blots or protein arrays.