Use of touch imprint cytology as a simple method to enrich tumor cells for molecular analysis




A relative excess of nonneoplastic cells in frozen carcinoma samples is often a cause of false-negative results in molecular assays. Given the greater cohesiveness of epithelial tumor cells compared with nonneoplastic epithelium and mesenchymal stroma, the authors hypothesized that tumor procurement by touch imprinting would provide a simple, cost-effective method of obtaining enriched neoplastic cells compared with frozen whole-tumor samples.


Eleven adenocarcinomas with known KRAS gene mutations were tested. Two sets of 8 touch imprint (TP) slides and 1 frozen whole-tumor sample (FS), both with a corresponding hematoxylin and eosin-stained slide, were obtained from each tumor. DNA from unstained TP and FS samples was tested for KRAS exon 2 mutations by Sanger sequencing. The percentage of carcinoma cells was determined by light microscopy of hematoxylin and eosin-stained slides. The fold increase in the mutant-enriched DNA in TP versus FS samples was determined by calculating the height ratio between the mutant and wild-type peaks on the sequencing electropherogram.


Using light microscopy, TP demonstrated a 1.1-fold to 3.5-fold (mean, 1.8-fold) enrichment in neoplastic cells compared with the FS. The mutant–to–wild-type peak height ratio was 1.4-fold to 7.1-fold (mean, 3.1-fold) higher in TP compared with the corresponding FS samples. The average amount of extracted DNA ranged from 145 ng to 7.9 μg per TP slide.


The procurement of carcinoma samples by TP is rapid, simple, and inexpensive; consistently provides a tumor-enriched sample; is an excellent source of high-quality tumor DNA; and could compensate for the relatively low sensitivity of direct sequencing. Cancer (Cancer Cytopathol) 2013;121:354–360. © 2013 American Cancer Society.


Tumor procurement is an important responsibility of an academic surgical pathology department regardless of whether procured tissues are to be used for research purposes or as clinical samples for molecular diagnostics. One of the common methods for tumor procurement is the banking of frozen tumor samples. Although the great advantage of this method is good preservation of nucleic acids for future analysis, a relative excess of nonneoplastic cells in whole-tumor samples leads to dilution of genetic signals and is often a cause of false-negative results in molecular assays. Thus, obtaining a sample enriched for neoplastic cells is critical for molecular characterization and the accurate detection of genetic abnormalities.

To overcome the inherent heterogeneity of human tumor tissues and to gain the purity needed for molecular analysis, various microdissection techniques have been developed for the separation of the tumor cells from the nonneoplastic tissues (mesenchymal stroma, inflammatory cells, nonneoplastic epithelium, etc).[1] One of the first such techniques involved manual or micromanipulator guidance of a needle to scrape off an area of interest from a thin tissue section.[2] Soon after that, a more sophisticated approach using laser capture microdissection (LCM) was developed.[3] Although it is a powerful tool for obtaining exceptionally pure samples for molecular analysis, LCM requires expensive equipment, special training of the operator, and a considerable amount of time to perform the microdissection.[4] To meet the requirements of a rapidly expanding field of molecular diagnostics and to facilitate clinical and genomics research, an easy, rapid, and cost-effective protocol for obtaining enriched tumor samples for molecular characterization is needed.

Maitra et al took advantage of the inherent property of desmosome-rich epithelial cells to separate from surrounding stromal tissues as tightly adherent clusters and developed a technique to facilitate subsequent LCM that they termed “EASI” (epithelial aggregate separation and isolation).[5] Tumor-enriched specimens were prepared by gentle scraping of the surface of fresh tumor tissue, spreading the material over a glass slide, and rapidly fixing it in an alcohol-based fixative. For smaller volume samples (ie, those measuring < 0.5 cm2), the authors suggested a simple “touch preparation.”

The natural propensity of carcinoma cells to cluster and to adhere to a glass slide has been successfully used in touch imprint cytology. This simple and rapid technique has proven to be useful as an adjunctive intraoperative diagnostic method with a high degree of diagnostic accuracy.[6] For example, touch imprints can be used for the evaluation of sentinel lymph nodes as well as for the assessment of surgical margins in breast cancer specimens.[7-9]

In the current study, we hypothesized that tumor procurement by simple touch imprinting would provide a simple, cost-effective method with which to obtain samples enriched in neoplastic cells compared with frozen whole-tumor samples. We evaluated 11 adenocarcinoma samples, all of which demonstrated neoplastic cell enrichment using touch imprint.



Eleven adenocarcinomas with known KRAS oncogene mutations were studied: 2 primary colorectal carcinomas (CRCs), 3 liver metastases of CRCs, 1 lung metastasis of a CRC, 2 adenocarcinomas of the lung, and 3 ductal adenocarcinomas of the pancreas. Tumor specimens were processed under a waiver of authorization approved by the Memorial Sloan-Kettering Cancer Center Institutional Review Board.

Touch Imprint Protocol

A single 1.0 cm × 1.0 cm × 0.3 cm fresh tumor sample was cut, avoiding surrounding normal tissue and necrotic areas (if possible). Two sets of 10 touch imprint slides were prepared as follows. Each of the 20 slides was prepared by dabbing the cut surface of the tissue against uncharged glass slides to cover at least 60% of the slide surface. The slide was immediately immersed in cold 95% ethanol, and after the last slide was prepared, they were fixed in ethanol for 5 minutes. The first and the last slide in each set (ie, slides 1 and 10) were stained with hematoxylin and eosin (H&E) to ensure the presence of tumor cells, and the remaining 8 slides (ie, slides 2-9) were air-dried and saved to be used for DNA extraction. Touch imprint slides were stored at room temperature for 1 to 8 months before DNA extraction. A matched fresh tumor sample (without major areas of normal tissue) was embedded in OCT, frozen in liquid nitrogen, and stored at −20 C degrees to be used to prepare 1 control H&E slide and for DNA extraction.

DNA Extraction and Direct Sequencing

Frozen tissue was diced with a blade into pieces measuring < 1 mm in diameter. Each touch imprint slide was first moistened with approximately 5 μL of DNA extraction buffer and the cells were scraped with a cell scraper. DNA was extracted using a DNeasy extraction kit (Qiagen, Carlsbad, Calif). The DNA concentration and the 260 nanometers (nm)/280 nm ratio were determined using a spectrophotometer (NanoDrop 1000 Spectrophotemeter; NanoDrop Products, Thermo Fisher Scientific, Wilmington, DE). Activating mutations in KRAS codons 12 and 13 were detected by standard polymerase chain reaction (PCR) sequencing. PCR amplification of a 224-base pair fragment including the entire coding region of exon 2 was performed using HotStar Taq DNA polymerase (Qiagen) and appropriate primers (KRAS2/E2intF: 5′-GTG TGA CAT GTT CTA ATA TAG TCA-3′ and KRAS2/E2extR: 5′-CTG TAT CAA AGA ATG GTC CTG CAC-3′).[10] In case 11, a PCR was also performed using standard primers in conjunction with a 10-mer locked nucleic acid (LNA) oligonucleotide to suppress amplification of the wild-type DNA, as previously described.[11] All PCR amplifications were run in duplicate for each sample. The PCR products were sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems/Life Technologies, Grand Island, NY) on an ABI3730 Genetic Analyzer (Applied Biosystems/Life Technologies).

Evaluation for Tumor Cell Enrichment

Light microscopy

The ratio between the mean percentage of carcinoma cells (vs nonneoplastic cells) on 2 H&E-stained touch imprint slides and the percentage of carcinoma cells on the frozen tissue section provided a fold increase in carcinoma cells in touch imprints versus frozen tissue sections for each case.

Mutation analysis

The relative percentage of the mutant allele in a given sample was determined by calculating the height ratio between the mutant and wild-type peaks on the sequencing electropherogram, in both forward and reverse DNA strands. The fold increase in the mutant-enriched DNA in touch imprints versus frozen samples was first calculated separately for each forward and reverse run; the values were then averaged to determine the fold increase per case.

Statistical Analysis

A Student t test was used to determine statistical significance. P values < .05 were considered to be statistically significant.


Using light microscopy, the average percentage of tumor cells in touch imprints and whole frozen tissue samples was 73% (range, 50% to 95%) and 50% (range, 20% to 80%), respectively. The 2 touch imprints stained with H&E (slides 1 and 10) demonstrated similar tumor cellularity in all cases, suggesting that the intervening slides used for molecular analysis also contained consistent amounts of neoplastic cells. The average amount of extracted DNA ranged from 145 ng to 7.9 μg per touch imprint slide and the lowest DNA yield was obtained from pancreatic adenocarcinoma with a high degree of desmoplasia (case 6). The 260 nm/280 nm ratio of touch imprint DNA and frozen tissue DNA were ≥ 1.86 and ≥ 1.71, respectively. Table 1 summarizes the enrichment in neoplastic cells for touch imprinting versus whole frozen sections. On light microscopy, touch imprints demonstrated enrichment in the neoplastic cell percentage compared with the frozen tissue in all cases. A 1.1-fold to 3.3-fold (mean, 1.8-fold) enrichment was obtained, based on the morphologic estimation of the relative amount of carcinoma.

Table 1. Average Fold Increase in Neoplastic Cell Enrichment in Touch Imprints Versus Frozen Tissue
Case No.DiagnosisKRAS MutationTP/FS Mutant-to-Wild-Type Peak Height Ratio (Sequencing Electropherogram)TP/FS Tumor Percentage Ratio (H&E)
  1. Abbreviations: CRC, colorectal adenocarcinoma; FS, frozen tissue sample; H&E, hematoxylin and eosin; LC, adenocarcinoma of lung; met, metastasis; NA, not available; PC, ductal adenocarcinoma of pancreas; TP, touch imprint.

  2. a

    Average fold increase for the first 10 cases.

1CRCG12V (35G>T)4.41.9
2CRC, liver metG12V (35G>T)1.51.5
3CRC, liver metG12S (34G>A)1.81.7
4CRCG13D (38G>A)6.43.2
5LCG12V (35G>T)2.41.4
6PCG12R (34G>C)3.03.3
7CRC, lung metG13D (38G>A)5.21.2
8CRC, liver metG12C (34G>T)1.61.1
9PCG12V (35G>T)2.11.4
10PCG12D (35G>A)2.31.6
11LCG12C (34G>T)NA1.4

Using standard PCR sequencing, KRAS mutations were detected in 10 cases in both touch imprints and frozen tissue samples. In these 10 cases, the mutant–to–wild-type peak height ratio was 1.5-fold to 6.4-fold (mean, 3.1-fold) higher in touch imprint samples compared with the corresponding frozen samples (Fig. 1). In case 11, the frozen sample was composed of approximately 60% tumor cells, and the remainder were nonneoplastic inflammatory cells. In case 11, the KRAS mutation was detected by standard PCR sequencing in the touch imprints only but initially not in the matched frozen tissue; the mutation in the frozen tissue sample was subsequently confirmed by LNA-PCR sequencing (Fig. 1C). Although not statistically significant (likely due to the small number of cases), the fold increase in mutant-enriched DNA was relatively lower in touch imprints of 3 largely necrotic liver metastases from CRCs compared with carcinomas with predominantly viable tumor (1.6-fold vs 3.7-fold; P = .07 according to the Student t test for unpaired data).

Figure 1.

Sequencing electropherograms of (upper row) touch imprints (TP) and (lower row) frozen sections (FS) of cases (A) 1, (B) 10, and (C) 11 are shown. Sequencing electropherograms indicated that the mutant–to–wild-type peak ratio was higher in TP samples from cases 1 and 10. In case 11, the mutation was detected only in TP samples by standard sequencing, and was detected only by locked nucleic acid-polymerase chain reaction (LNA-PCR) assay in the FS samples. FWD indicates forward; REV, reverse.


In the current study, we demonstrated that tumor procurement by touch imprinting has multiple advantages over frozen whole-tumor tissue sampling. Neoplastic cell enrichment was found to be greater in touch imprints than in matched frozen tissue in all cases studied, and this was confirmed by 2 different methods: light microscopy and sequencing electropherogram analysis. The quality of DNA based on the 260 nm/280 nm ratio was similar in both sample types and the amount of extracted DNA from touch imprints was sufficient for a PCR-based assay in all cases, including those tumors with relatively abundant fibrosis/desmoplastic reaction such as ductal adenocarcinomas of the pancreas.

Tumor specimens may vary greatly in the abundance of viable neoplastic cells relative to stromal elements, desmoplastic reaction, mucin content, inflammatory infiltrates, nonneoplastic epithelium, and necrosis. These factors may largely influence the usability of case material; according to the Cooperative Human Tissue Network, insufficient tumor cellularity was found to be one of the most common reasons why approximately 15% of tissues sampled for specific research could not be used as originally intended.[12-14] In the current study, we examined 3 liver metastases of CRCs involved by extensive necrosis due to neoadjuvant chemotherapy and demonstrated that touch imprint procurement of largely necrotic tumors resulted in significant neoplastic cell enrichment in all cases compared with frozen whole-tumor samples.

In many molecular diagnostic laboratories, manual macrodissection of formalin-fixed paraffin-embedded (FFPE) tumor tissue sections is a standard method to enrich tumor cells for molecular analysis. Three of the cases in the current study had been subjected to routine clinical testing for KRAS mutations using microdissected paraffin sections for DNA extraction. Based on the sequencing electropherogram analysis, the tumor cell enrichment in macrodissected FFPE sections was higher than in matched frozen tissue but was notably lower than in touch imprints in all 3 cases (data not shown). Although superior to frozen whole-tumor tissue, manual macrodissection may not be helpful in cases in which tumors have a diffusely infiltrative growth pattern without discrete tumor nests or in the presence of abundant inflammatory cells. Microscopic entrapped nonneoplastic elements, such as pancreatic parenchyma within ductal adenocarcinomas, cannot be separated from the neoplastic cells without much more labor-intensive microdissection such as LCM.

In our experience, exuberant inflammation within a tumor is not an uncommon reason for the failure of PCR sequencing assays. The inflammatory cells may contribute a significant amount of wild-type DNA and therefore “dilute” the tumor DNA. Case 11 in the current study is an example of how this issue could be resolved by touch imprint tumor procurement. Although the tumor cell content in the frozen tissue was approximately 60%, extensive coexisting inflammation was the likely reason why the KRAS mutation was not detected by standard PCR sequencing in this material. Conversely, KRAS mutation was detected in the matching touch imprints that consisted predominantly (approximately 90%) of carcinoma cell clusters, with only scattered inflammatory cells. These results suggest that touch imprints may be particularly useful in the procurement of carcinomas with a heavy inflammatory infiltrate. The same principle could be applied to procurement from lymph node metastases.

In the current study, the presence of a mutation in the frozen tissue sample from case 11 was subsequently confirmed by LNA-PCR sequencing. Although the technical sensitivity of a standard PCR sequencing for KRAS mutations is 20% to 25% and requires at least 50% of tumor DNA in the test sample to ensure the detection of a heterozygous point mutation, the LNA-PCR sequencing assay has markedly improved the technical sensitivity to 0.5%, thereby requiring only 1% of tumor DNA in a tested sample. LNA-PCR sequencing could certainly be the assay of choice in cases with low tumor cell content. However, to our knowledge, to date this method has been evaluated only for the detection of point mutations and it cannot be applied to the type of large-scale mutational profiling performed on some of the most recent mass spectrometry-based genotyping platforms, which offer testing at intermediate technical sensitivity (ie, 5%-10%).[11, 15] Therefore, to ensure wide usability of the test material in various molecular assays with a broad range of technical sensitivities, neoplastic cell-enriched samples are required.

Tumor volume is often a limiting factor in tumor procurement. Neoadjuvant chemotherapy or radiation along with earlier detection of malignant processes often results in a relatively small tumor size at the time of surgical resection, requiring routine histologic processing in toto to ensure an accurate morphologic diagnosis. Touch imprint cytology uses minimal tumor tissue and does not compromise routine histology diagnosis. Thus, touch imprinting could be the method of choice for the procurement of small, limited-volume tumors.

Several studies have demonstrated that a sufficient amount of good-quality DNA for molecular analysis could be obtained from cytology smears.[16, 17] However, in these reports, the smears were prepared by scraping the entire tumor surface, thereby requiring a microdissection step for tumor cell enrichment and potentially including nonneoplastic stromal elements among the scraped material. In the current study, instead of mechanical scraping of the entire tumor surface, we took advantage of the inherent property of carcinoma cells to cluster and to adhere to a glass slide to avoid microdissection as an additional time-consuming step.

We have demonstrated that ethanol-fixed and air-dried touch preparation slides can be stored at room temperature and remain a source of high-quality DNA for at least 8 months. It is important to note that the ethanol fixation step may not be required if the material is to be used for DNA extraction only, and touch imprints may be stored for a prolonged period of time because air-dried cytology smears have been shown to remain a good source of quality DNA for several years.[17]

Touch imprints do not require costly ultra-cold storage systems or special preparation reagents such as liquid nitrogen. No sectioning is required to prepare touch imprint H&E controls, nor is preparation of aliquots needed. Although 8 touch imprint slides were used for DNA extraction, we suggest that the number of slides could easily be reduced by approximately one-half if procured carcinomas appear “cellular” macroscopically and lack macroscopic evidence of extensive fibrosis, necrosis, or excessive mucin production.

A major disadvantage of tumor procurement by touch imprinting would be failure to obtain RNA of sufficient quality or quantity. We tested several RNA extraction methods including chloroform/ispropanol extraction and commercial kits (RNeasy [Qiagen, Carlsbad, Calif] and PicoPure RNA Isolation Kit [Applied Biosystems&Life Technologies, Grand Island, NY]) without satisfactory results. This was surprising, because procured touch imprints were immediately fixed in cold ethanol to ensure optimal tissue (and RNA) preservation, and general protocols and precautions for RNA extraction and handling were followed. The quality of frozen-tissue RNA was only slightly better based on housekeeping gene expression (phosphoglycerate kinase) and RNA quality analysis using the Agilent 2100 Bioanalyzer (Agilent Technologies Inc, Santa Clara, Calif) (data not shown). One possible reason for failure could be the faster degradation of RNA in a thin layer of tissue at room temperature. Inadequate RNA quality and quantity would be a limiting factor in studies requiring RNA, such as reverse transcriptase-polymerase chain reaction to detect gene rearrangement fusion transcripts. Alternative preservation media (eg, PAXgene preservative; PreAnalytiX GmbH, Hombrechtikon, Switzerland) also can be investigated as a method with which to better preserve RNA from touch preparation samples.[18] However, this problem may be overcome to some degree by using alternative methods to identify chromosomal translocations, such as fluorescence in situ hybridization, which has been successfully applied to touch imprints.[19] Despite these considerations, for clinical molecular diagnostic purposes, DNA is often sufficient given that the majority of genetic defects found in carcinomas (eg, point mutations, insertions, deletions) are easily targeted by one of the PCR-based methods that use genomic DNA (such as PCR sequencing, restriction fragment length polymorphism, or fragment analysis).

Other potential limitations of the current study are the relatively small size and the inclusion of different tumor types with various degrees of compounding factors (such as fibrosis, necrosis, or inflammation). Furthermore, we have tested only a limited number of adenocarcinomas and cannot be certain that similar results would be obtained with other carcinoma types.

Our ever-growing understanding of the key molecular mechanisms that are affected during cancer development and progression, and the identification of therapeutic targets and prognostic markers, have changed our approach to the diagnosis of and therapy for malignant tumors. Translation from basic science to clinical applicability occurs regularly, and the number of molecular diagnostic tests is increasing accordingly. Despite rapid technological advancement, the accuracy of the molecular results depends largely on the selection of appropriate study material. A neoplastic cell-enriched sample is necessary to ensure an accurate molecular diagnosis. The results of the current proof-of-principle study demonstrated that procurement of carcinomas by touch imprinting is rapid, simple, and inexpensive; consistently provides a tumor-enriched sample; and is an excellent source of high-quality tumor DNA. This method is superior to frozen whole-tumor tissue (and FFPE macrodissected tumor), especially in the presence of extensive inflammation or necrosis, and it is a potential method of choice for procurement from small, limited-volume tumors. Neoplastic cell enrichment by touch imprinting may compensate for the relatively low sensitivity of direct PCR sequencing and possibly other molecular methods.


No specific funding was disclosed.


The authors made no disclosures.