Although histochemical staining has been believed to inhibit the DNA amplification reaction, no previous study has systematically evaluated the influence of histochemical staining on downstream molecular assays. To evaluate an influence of H&E staining on DNA testing, we isolated DNA from 10 unstained, 10 hematoxylin-stained, 10 eosin-stained, and 10 H&E-stained tissue sections (ie, 4 groups), from each of 5 colon cancers. Among the 4 groups, we did not observe any significant or appreciable difference in DNA fragmentation by agarose gel electrophoresis, in DNA amplification by real-time polymerase chain reaction (PCR), in microsatellite PCR fragment analyses, or in a PCR-pyrosequencing assay. As a proof-of-principle study, we successfully performed microsatellite instability analysis and sequencing of KRAS and BRAF on more than 1,300 colorectal cancers using DNA extracted from H&E-stained tissue sections. Our data provide no evidence for an interfering effect of H&E staining on DNA testing, suggesting that DNA from H&E-stained sections can be effectively used for routine DNA testing.
Polymerase chain reaction
Screening and identification of genetic alterations in formalin-fixed, paraffin-embedded (FFPE) tumor materials have been important in molecular diagnosis and clinical medicine.1–6 Solid tumors are inherently heterogeneous tissues that contain neoplastic cells admixed with non-neoplastic cells, including inflammatory cells, vascular and lymphatic endothelial cells, smooth muscle cells, and stromal fibroblasts. Therefore, an enrichment of neoplastic cells is commonly performed before DNA extraction in molecular testing to avoid false-negative results due to low neoplastic cellularity.7
Macrodissection from tissue sections on glass slides has been a widely used method to exclude pure nonneoplastic areas and enrich neoplastic cellularity for molecular testing. For macrodissection of tumor areas, an accurate morphologic identification of tumor areas is essential.8,9 Thus, macrodissection should be guided by an H&E-stained tissue section with tumor areas marked by a pathologist under a microscope Image 1A. It would be ideal to dissect tumor tissue from an H&E-stained section (without coverslip) because the H&E stain enables us to easily identify tumor areas Image 1B, particularly next to the H&E slide with the marked tumor areas (Image 1A). In contrast, it is difficult to accurately identify tumor areas of an unstained tissue section Image 1C, and tumor areas can be missed.
With regard to the influence of histochemical staining on downstream molecular assays, some previous studies have shown interfering effects of histochemical staining, especially hematoxylin stain,10–12 while another study failed to show a significant difference in polymerase chain reaction (PCR) amplification between DNA specimens from H&E-stained and unstained tissue sections.13 To avoid possible interfering effects of histochemical staining on molecular assays, macrodissection of tumor from unstained tissue sections is commonly performed, despite the difficulty in identifying tumor areas (Image 1C) and the risk of missing tumor areas.1,14–17 To our knowledge, no previous study has systematically evaluated the influence of histochemical staining on downstream molecular assays.
We therefore conducted this study to evaluate various molecular assay results on DNA extracted from H&E-stained, hematoxylin-stained, and eosin-stained FFPE tissue sections, in comparison with DNA extracted from unstained tissue sections. We compared results of agarose gel genomic DNA fragment analysis (without PCR), quantitative real-time PCR assay, PCR-pyrosequencing assay, and PCR–capillary electrophoresis fragment analyses. We did not observe an appreciable difference in any of the results on DNA specimens from stained tissue. Furthermore, as a proof-of-principle study, we successfully used DNA from H&E-stained tissue sections and obtained DNA analysis data from more than 1,300 colorectal cancers. Our data support the feasibility of DNA extraction from H&E-stained tissue sections for routine clinical laboratory workflow.
Materials and Methods
FFPE Tissue Specimens
Specimen collection and analysis in this study were approved by the Harvard School of Public Health, Brigham and Women’s Hospital, and Dana-Farber Cancer Institute institutional review boards (Boston, MA). FFPE colorectal cancer tissue specimens (5 cases) were anonymized after collection from the archival file of the Department of Pathology, Brigham and Women’s Hospital, to assess the effects of histochemical staining on downstream molecular testing Figure 1.
As a proof-of-principle feasibility study, we tested DNA extraction from H&E-stained tissue sections in a large cohort of colorectal cancer cases. We used 2 US nationwide prospective cohort studies, the Nurses’ Health Study (N = 121,701 women followed up since 1976) and the Health Professionals Follow-up Study (N = 51,529 men followed up since 1986).18,19 We collected FFPE tissue blocks from hospitals and pathology laboratories throughout the United States where cohort participants with colorectal cancer underwent cancer resections.18,19 H&E-stained tissue sections were reviewed by a pathologist (S.O.). Genomic DNA was extracted from H&E-stained tissue sections, using another H&E-stained section with marked tumor areas as a guide (as in Images 1A and 1B), for downstream KRAS, BRAF, and MSI testing, and results were available for 1,314, 1,313, and 1,293 cases, respectively.
Examples of H&E-stained and unstained tissue sections. A, An H&E-stained tissue section slide to guide tumor tissue dissection. A pathologist marked a tumor area under a microscope. B, An H&E-stained tissue section without coverslip for subsequent DNA extraction. It is easy to identify the tumor area. C, An unstained tissue section for DNA extraction. It is not as easy to identify the tumor area as in the H&E-stained tissue section (B).
Overall Study Strategy, DNA Extraction, and Agarose Gel Electrophoresis
Deparaffinized and hydrated 10-μm sections from the same tissue block were stained by hematoxylin (Harris hematoxylin, Surgipath, Richmond, IL) alone for 3 minutes (group 2), by eosin (Surgipath) alone for 15 seconds (group 3), or by H&E (group 4; Figure 1). Unstained slides were prepared as control slides (group 1). Figure 1 illustrates our overall strategy of experiments to assess the effects of hematoxylin and/or eosin staining on DNA integrity and downstream molecular assays.
To avoid bias that could be introduced by tumor-directed macrodissection, a whole tissue section (including tumor and adjacent normal tissue) was entirely scraped off from a glass slide by using a sterile needle. Thus, DNA yields were very similar between the groups. The scraped tissue was collected into a microtube, and genomic DNA was extracted using QIAamp DNA Mini Kit (Qiagen, Valencia, CA). As a result, each of 10 tissue sections was “aliquoted” to 1 tube, to yield 10 aliquoted DNA specimens for each group (Figure 1). The 260/280 nm absorbance ratio of extracted DNA was approximately 1.8 (NanoDrop, Thermo Scientific, Waltham, MA). Extracted DNA from unstained tissue and H&E-stained tissue from all 5 cases was analyzed by electrophoresis in 0.8% agarose gel Image 2. Three cases were used for real-time PCR and PCR-pyrosequencing assay, and the other 2 cases were used for microsatellite PCR fragment analysis.
Our overall strategy to assess the influence of tissue staining on downstream DNA testing. DNA was extracted from unstained tissue (group 1), hematoxylin-stained tissue (group 2), eosin-stained tissue (group 3), and H&E-stained tissue (group 4). PCR, polymerase chain reaction; WGA, whole genome amplification.
Real-Time PCR Assay for GAPDH
To assess PCR amplification efficiency, we performed real-time PCR using the primers for GAPDH and SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA) by the ABI 7300 Real-Time PCR system (Applied Biosystems). Primer sequences were 5′-GTCATGGGTGTGAACCATGAGAA-3′ and 5′-TGGT-CATGAGTCCTTCCACGAT-3′. The PCR reaction was repeated 4 times on each of the 10 DNA aliquots for groups 1 through 4, and cycle threshold (Ct) values were compared.
Microsatellite PCR Fragment Analysis
To assess the effects of hematoxylin and/or eosin staining on PCR fragment analysis, we performed PCR for dinucleotide markers, D2S123 and D5S346,20 after PCR-based whole genome amplification on genomic DNA as previously described.21 Forward primers were labeled with fluorescence, and PCR product sizes were 180 base pairs (bp; D2S123) and 129 bp (D5S346). PCR products were electrophoresed and analyzed by the ABI 3730 DNA Analyzer (Applied Biosystems). PCR fragment analysis was repeated twice on each of the 10 DNA aliquots.
Agarose gel (0.8%) electrophoresis to assess integrity of DNA from unstained and H&E-stained tissue sections. There was no appreciable difference in DNA integrity between DNA specimens from unstained tissue and H&E-stained tissue sections. The representative case (10 sections) was shown. Lane 1, size marker; lanes 2–11, group 1 (unstained tissue); lanes 12–21, group 4 (H&E-stained tissue). bp, base pairs.
For the large feasibility cohort, we performed microsatellite instability (MSI) analysis using a 10-marker panel (D2S123, D5S346, D17S250, BAT25, BAT26, BAT40, D18S55, D18S56, D18S67, and D18S487).20 MSI-high was defined as the presence of instability in 30% or more of the markers and MSI-low/microsatellite stability as instability in fewer than 30% of the markers.20
PCR-Pyrosequencing Assay for KRAS and BRAF
To assess the effect of hematoxylin and/or eosin staining on pyrosequencing, we performed a PCR-pyrosequencing assay for KRAS (codons 12 and 13), which was previously developed and validated.21 The size of PCR product was 82 bp, and 10 μL of each was sequenced by using the Pyrosequencing PSQ96 HS System (Qiagen). The PCR-pyrosequencing reaction was repeated 3 times on each of the 10 aliquots. In addition, in the feasibility cohort, the PCR-pyrosequencing assay for BRAF (codon 600) was performed as previously described.22
For all statistical analyses, we used the SAS program (version 9.1, SAS Institute, Cary, NC). All P values were 2-sided. The Ct in real-time PCR, which reflected amplification efficiency of each PCR reaction given similar amounts of scraped tissue and extracted DNA, was compared by using the analysis of variance (ANOVA) test, adjusting for case, spot of each reaction, and plate. We performed the ANOVA test for comparing peak heights of the fluorescence signal on capillary electrophoretograms in microsatellite analysis. The distribution of the peak height values (median, 3,028.5; range, 0–10,618) was normalized by logarithmic transformation after adding 0.5 to each value (to log-transform the value of “0”). A deviation from the null hypothesis in any of the comparisons might imply an interfering effect of histochemical staining on the DNA testing process.
Integrity of DNA From Unstained and Stained Tissue
Our overall strategy to assess the influence of H&E staining on subsequent DNA testing for FFPE tissue is shown in Figure 1. First, we assessed the influence of H&E staining on DNA integrity. Extracted DNA specimens of group 1 (unstained) and group 4 (H&E-stained) were loaded onto 0.8% agarose gel (Image 2). There was no appreciable difference in DNA integrity between the DNA specimens from H&E-stained and unstained tissue sections from all 5 cases.
Comparison of Ct Values in Real-Time PCR
We compared Ct values in real-time PCR for GAPDH in the 4 groups using 3 cases (× 10 sections × 4 repeated runs) by the ANOVA test, which adjusted for case, spot of each reaction, and plate Figure 2A. There was no significant difference in Ct values between groups (P = .19).
Tissue staining and subsequent polymerase chain reaction (PCR) analysis. A, Mean cycle threshold (Ct) values in quantitative real-time PCR for GAPDH on DNA specimens (3 cases × 10 sections × 4 repeated runs) in each group. The error bars indicate the SD. There was no significant difference in the mean Ct values between groups (P = .19). B, Mean peak heights in electrophoretograms in PCR-fragment analysis of the microsatellite markers (D2S123 and D5S346) on DNA specimens (2 cases × 10 sections × 2 repeats) from each group. The error bars indicate the SD. There was no significant difference in the mean peak heights in the 4 groups (overall P = .35).
Microsatellite PCR Fragment Analysis
We performed PCR fragment analysis for the 2 microsatellite markers, D2S123 and D5S346, and assessed the potential influence of tissue staining by using 2 cases (× 10 sections × 2 repeats). We measured the peak height in electrophoretograms, and compared the mean peak height value for the 4 groups. There was no significant difference in the mean peak height in the 4 groups (overall P = .35) Figure 2B.
PCR-Pyrosequencing Assay for KRAS
We performed the PCR-pyrosequencing assay for KRAS codons 12 and 13 on 3 cases (× 10 sections × 3 repeats) and compared results for the 4 groups. In 2 cases, there was a c.35G>A (p.G12D) mutation, and 1 case showed only a wild-type sequence Figure 3. All pyrograms for KRAS codons 12 and 13 showed clear peaks and little background, and there was no appreciable difference in the 4 groups.
MSI Testing and Pyrosequencing for KRAS and BRAF in the Large Feasibility Cohort
Finally, we performed a proof-of-principle feasibility study on FFPE tissue blocks that were fixed and processed in numerous pathology laboratories throughout the United States, using the 2 US nationwide prospective cohort studies.18 We dissected tumor tissue from H&E-stained sections (without coverslip) using a guide H&E slide with marked tumor areas (as in Images 1A and 1B), and extracted DNA for subsequent MSI analysis and the PCR-pyrosequencing assay for KRAS and BRAF. KRAS and BRAF mutations were detected in 470 (35.8%) of 1,314 cases and 186 (14.2%) of 1,313 cases, respectively Table 1. MSI-high was present in 197 (15.2%) of 1,293 cases. These frequencies are compatible with the published data on these molecular features in colorectal cancers,23–27 particularly in population-based studies.28–33 There was no evidence for an interfering effect of H&E staining on these molecular assays. These data suggest that DNA from H&E-stained sections can be used for routine tumor DNA testing on FFPE tissue.
We conducted this study to evaluate the potential influence of H&E staining on subsequent DNA testing. Solid tumor tissue is inherently heterogeneous tissue with an admixture of neoplastic cells and nonneoplastic cells, including inflammatory, stromal, endothelial, and smooth muscle cells, even within tumor tissue areas. Macrodissection from H&E-stained tissue sections enables accurate identification and dissection of tumor areas to obtain DNA specimens with a high content of neoplastic cell DNA. Especially for small tumor nests or metastatic tumor foci that must be carefully and precisely dissected, it is very useful to stain and clearly visualize the tissue section. H&E staining is a common and simple method to visualize the tumor area of interest. Although hematoxylin staining has been believed to inhibit DNA amplification reaction, only a small number of studies have been performed to assess the potential influence of tissue staining on amplification of DNA by PCR.10,11,13 Notably, no previous study has systematically evaluated the influence of H&E staining on downstream molecular assays. Our current study provided no evidence for interfering effects of H&E staining on subsequent DNA assays. Our data support routine clinical use of H&E staining of tissue sections for subsequent tumor tissue dissection, DNA extraction, and molecular analyses.
Tissue staining and subsequent polymerase chain reaction (PCR)-pyrosequencing assay for KRAS. A, Case 1 with mutant KRAS codon 12 (c.35G>A) admixed with wild-type sequence. The assay was repeated 3 times on 120 DNA aliquot specimens (3 cases × 4 groups × 10 sections), and representative results are shown. There was no appreciable difference between DNA specimens from the 4 groups. B, Case 3 with wild-type KRAS. The assay was repeated 3 times on 120 DNA aliquot specimens (3 cases × 4 groups × 10 sections), and representative results are shown. There was no appreciable difference between DNA specimens from the 4 groups.
Previous studies evaluated potential influence of tissue staining on DNA assays.10,11,13 Burton et al10 and Chen et al34 showed that hematoxylin staining inhibited DNA amplification by PCR, and Diss et al11 reported that DNA from H&E-stained tissue was amplified less efficiently than DNA from unstained tissue. On the other hand, Murase et al13 reported no significant difference in PCR amplification between DNA specimens from unstained and H&E-stained tissue. However, none of the previous studies10,11,13 quantitatively evaluated the efficiencies of PCR amplification on DNA specimens from unstained and stained tissues. In the present study, there was no statistically significant difference in GAPDH amplification by quantitative real-time PCR between DNA specimens from unstained tissue, hematoxylin-stained tissue, eosin-stained tissue, and H&E-stained tissue.
Pyrosequencing is a nonelectrophoretic sequencing by nucleotide extension. Pyrosequencing is useful for sequencing of relatively short nucleotides (up to 40 to 50 bp). Therefore, pyrosequencing has been applied for single nucleotide polymorphism genotyping,35 bacterial strain typing,36 mutation detection in tumors,21,37,38 quantitative promoter CpG island methylation analysis,39,40 and measurement of LINE-1 methylation.41 We evaluated the influence of H&E staining for the PCR-pyrosequencing assay for KRAS and showed no appreciable difference in results between DNA specimens from unstained and H&E-stained tissues.
Capillary electrophoresis is commonly used for microsatellite analysis in FFPE clinical tissue specimens (eg, MSI and loss of heterozygosity analyses).42–45 We showed no significant difference in peak heights of the electrophoretograms of 2 microsatellite markers (D2S123 and D5S346) between DNA specimens from unstained and stained tissues.
Murphy et al46 reported an artifact (a peak at 71 bp) in capillary electrophoresis due to autofluorescence from eosin, which was probably contaminated by nonstringent DNA purification. A notable difference from our study was that Murphy et al46 experienced the artifact after biopsy tissue (not a tissue section on a slide) was stained with eosin “to enable identification of small tissue fragments during sectioning.” We have not experienced this problem, despite the fact that we have performed microsatellite PCR fragment analysis on more than 1,300 cases using DNA from H&E-stained tissue. Nevertheless, we agree with Murphy et al46 that artifacts caused by eosin should be avoided by using stringent DNA purification steps, and that artifacts from histochemical staining should be considered when peaks of the same product size (eg, 71 bp) are present in multiple specimens or primary peaks contain additional underlying peaks of other colors.
As a potential limitation of our study, we used a relatively small number of tissue blocks (N = 5) to systematically compare the effects of histochemical staining in downstream molecular assays. Nevertheless, we used as many as 10 sections and 10 DNA aliquots for each condition to achieve a precise measurement estimate. Moreover, as the proof-of-principle feasibility study, we used DNA extracted from H&E-stained tissue of more than 1,300 colorectal cancers in 2 US nationwide prospective cohort studies (the Nurses’ Health Study and the Health Professionals Follow-up Study) for MSI analysis and pyrosequencing of KRAS and BRAF (Table 1). The FFPE tissue specimens were retrieved from numerous pathology laboratories throughout the United States. Despite the diversity of laboratories that processed and stored tissues, we were successful in performing those molecular assays in more than 95% of cases tested, and the frequencies of MSI-high and mutations of KRAS and BRAF were compatible with published data on colorectal cancer,23–27 particularly in population-based studies.28–33 In addition, using DNA from H&E-stained tissue sections, it has been shown that BRAF mutation and MSI-high in colon cancer are associated with the CpG island methylator phenotype (CIMP20,28,47–52) and patient outcome,20,53 and KRAS mutation is associated with CIMP-low in our cohort studies.22 Our data indicate that H&E-staining is useful for DNA analysis on FFPE tissues from numerous pathology laboratories in multicenter clinical trials or population-based cohort studies,42,54 which can contribute to improvement of patient care and public heath.
Our present study represents the first systematic investigation on the effects of H&E staining in downstream molecular assays. Our results provide no evidence to support the common conception that H&E staining interferes with molecular assays on DNA from FFPE tissue. Our data support routine clinical use of H&E staining on tumor sections for subsequent tissue dissection and DNA extraction. H&E staining enables easy and accurate identification of tumor areas to obtain DNA with a high content of neoplastic cellular DNA. H&E staining can be easily incorporated into routine pathology workflow, so that laboratory personnel can easily identify and dissect tumor areas from H&E-stained tissue sections, guided by an H&E-stained slide with marked tumor areas. Thus, our data may have considerable influence on clinical molecular diagnostics and routine tumor molecular testing toward personalized medicine.
We deeply thank the Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS) cohort participants who have agreed to provide us with information through questionnaires and biological specimens; hospitals and pathology departments throughout the United States for generously providing us with tissue specimens; the staff of the NHS and the HPFS for their valuable contributions; and all of the US state cancer registries for their help.
↵* Drs Morikawa, Shima, and Kuchiba contributed equally to this work.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or NIH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Supported by US National Institute of Health (NIH) grants P01 CA87969 (S.E. Hankinson), P01 CA55075 (W.C. Willett), P50 CA127003 (Dr Fuchs), R01 CA118553 (Dr Fuchs), R01 CA124908 (Dr Fuchs), and R01 CA151993 (Dr Ogino).
Detection of c-KIT and PDGFRA gene mutations in gastrointestinal stromal tumors: comparison of DHPLC and DNA sequencing methods using a single population-based cohort. Am J Clin Pathol. 2010;133:149–155.
Identification of patients with (atypical) MUTYH-associated polyposis by KRAS2 c.34G > T prescreening followed by MUTYH hotspot analysis in formalin-fixed paraffin-embedded tissue. Clin Cancer Res. 2008;14:139–142.
Quality assessment of DNA derived from up to 30 years old formalin fixed paraffin embedded (FFPE) tissue for PCR-based methylation analysis using SMART-MSP and MS-HRM. BMC Cancer. 2009;9:453.doi:10.1186/1471-2407-9-453.
Effects of KRAS, BRAF, NRAS, and PIK3CA mutations on the efficacy of cetuximab plus chemotherapy in chemotherapy-refractory metastatic colorectal cancer: a retrospective consortium analysis. Lancet Oncol. 2010;11:753–762.