Cytology Smears in the Era of Molecular Biomarkers in Non–Small Cell Lung Cancer: Doing More With Less

Lung ADC is characterized by recurrent mutations in EGFR kinase and rearrangements involving anaplastic lymphoma kinase (ALK).³  These genetic alterations are successfully being targeted with specific drugs. Molecular testing for EGFR and ALK is required to select patients with lung ADC for EGFR- and ALK-targeted tyrosine kinase inhibitor therapy, respectively.¹²  KRAS and BRAF function downstream of EGFR in the signaling pathway, and activating mutations have been described to be usually mutually exclusive in EGFR-mutated tumors.¹⁹ 

Activating EGFR mutations are located in the tyrosine kinase domain (exon 18–21). Deletions in exon 19 and the point mutation L858R in exon 21 are the most common activating mutations usually found in lung ADCs, never smokers, females, and Asian patients.²⁰  Secondary EGFR mutations have also been shown to be involved in resistance to EGFR-directed tyrosine kinase inhibitor therapy.²¹  About 50% of cases of acquired resistance are related to a secondary T790M mutation in exon 20 of the EGFR gene.²²  ALK gene rearrangements are often found in never smokers or light ex-smokers and in younger patients.²³ 

KRAS, a Ras GTPase acting as a key molecular switch in the EGFR signaling pathway, is mutated at codons 12, 13, and 61 in 16% to 40% of patients with lung ADC.^19,24  KRAS mutations have been shown to be associated with poor response to EGFR-directed tyrosine kinase inhibitor therapy.¹² 

BRAF, a serine/threonine kinase, is activated by somatic point mutations in exon 15, including V600E, in 1% to 2% of patients with lung cancer.²⁵ 

Further recurrent mutations and amplifications in many potentially targetable oncogenes have been found in lung ADC, including human epidermal growth factor receptor 2 (HER2; also known as ERBB2), mesenchymal-epithelial transition factor (MET), and fibroblast growth factor receptors 1 and 2 (FGFR1 and FGFR2), as well as fusion oncogenes involving the c-ros oncogene 1 (ROS1) receptor tyrosine kinase, neuregulin 1 (NRG1), neurotrophic tyrosine kinase receptor type 1 (NTRK1), and rearranged during transfection (RET) gene.³  These oncogenic changes may predict sensitivity to clinical inhibitors and jointly account for most cases of lung ADC.³  Because of the similarity between the tyrosine kinase domains of ALK and ROS1, many ALK inhibitors also inhibit ROS1. Lung cancers with ROS1 translocations respond to crizotinib and alectinib, which are also used in ALK-positive lung cancers.⁸ 

Activating mutations in EGFR and ALK gene rearrangement are typically not present in squamous NSCLC.³  Lung SCCs are characterized by complex genomic alterations with recurrent mutations in several genes, including TP53, NFE2L2, KEAP1, BAI3, FBXW7, GRM8, MUC16, RUNX1T1, STK11, ERBB4, DDR2, and FGFR1 amplification.²⁶ 

It has been described that DNA hypomethylation leads to elevated mutation rates and is suggested to be an early step in carcinogenesis.³  Epigenetic silencing of genes through DNA promoter methylation of cytosines (C) affects the expression of several genes during initiation and development of lung cancer. Aberrant DNA methylation consists of the addition of a methyl group to the 5′ carbon of C within C-guanine dinucleotide sequences (CpGs), which are mainly concentrated in regions known as CpG islands. In general, methylation studies are based on DNA bisulfite conversion.^27,28  The C in CpG is resistant to bisulfite conversion and remains a C upon bisulfite treatment. The C outside CpG will be changed to thymidine by bisulfite treatment.^27,28 

The terms DNA hypermethylation and hypomethylation are applied to describe the aberrant methylated state of most CpG sites in a specific sequence. DNA methylation level is often described to be inversely correlated to gene expression level.²⁹ 

Silencing of genes by aberrant promoter DNA hypermethylation is likely to influence the transformation of lung epithelial cells.²⁸  In lung cancer, CDKN2A, which encodes INK4A (also known as p16) and ARF (also known as p14), is often silenced through aberrant promoter hypermethylation. The CDKN2A gene is inactivated in ∼67% of ADCs and ∼70% of SCCs.³⁰  Aberrant hypermethylation of p16/CDKN2A is an early event in lung cancer and a potential biomarker for early diagnosis.^27,28 

Aberrant patterns of DNA methylation of certain genes are associated with specific histologic phenotypes and are currently used as diagnostic markers. Aberrant short stature homeobox 2 (SHOX2) DNA methylation is being used in lung cancer as a prognostic biomarker as well as a diagnostic marker.^31–33  Patients with lung cancer and aberrant SHOX2 methylation have an adverse overall survival and the hypermethylation levels significantly correlate with a worse outcome.³³  Recently, our research group has described that the transmembrane protease serine 4 (TMPRSS4) promoter region is often hypomethylated in ADC and SCC lung cancer subtypes compared with normal lung tissue, suggesting that the methylation levels of the TMPRSS4 promoter might be useful as a diagnostic tool in lung cancer.³⁴ 

Tumor heterogeneity—genetic, cellular, and epigenetic—characterizes lung cancer.³  For patients with multiple, apparently separate, primary lung ADCs, each tumor may be tested but testing of multiple different areas within a single tumor is not necessary.³⁵ 

Paraffin-embedded tumor biopsy, generally taken from the primary tumor or metastatic lesion, is a diagnostic technique commonly used to determine the molecular profile of a tumor, for instance at a single time point before treatment commences. Most patients with lung cancer are diagnosed with advanced-stage disease or with metastatic lesions, which precludes them from surgical tumor resection. For these patients, the diagnosis is based on the analysis of cytology samples, especially FNA, obtained through minimally invasive procedures.

Cytology samples offer advantages over tissue biopsies for routine molecular testing: (1) cytology direct smears can be either air dried or fixed in alcohol; (2) they allow for rapid on-site adequacy assessment; and (3) DNA is readily extractable, providing high-quality DNA even if archived for a prolonged period.^17,36  Alcohol-fixed and air-dried smears are used for DNA extraction. In both cytologic preparations, the number of intact nuclei that are suitable for FISH analysis is higher and DNA appears to be better preserved.^16,37  Direct smear allows assessment of tumor content, ensuring adequacy for DNA extraction for molecular studies.¹⁷ 

There are a variety of minimally invasive procedures depending on the location of the tumor and/or metastasis. The most-used sampling procedures are computed tomography–guided FNA of the lung (frequently coupled with core needle biopsies), bronchoscopy-guided FNA with or without electromagnetic navigation, endobronchial ultrasound–guided FNAs of lymph nodes, endoscopic ultrasound–guided FNA, and direct aspiration of superficial metastatic lesions.^38,39 

Cytology samples also offer some advantages over their histologic counterpart. Rapid on-site evaluation can redirect the procedure in order to obtain samples adequate for ancillary studies. Rapid on-site evaluation done by a cytopathologist has improved the adequacy of FNA, allowing the assessment of cellular content, tumor cell yields, and adequate material for molecular testing (Figure 2, A through D). Rapid on-site evaluation maximizes cytology samples for molecular testing, including important steps for samples acquisition, processing, and selection.⁴⁰ 

Although there are many studies validating any form of cytology samples as a reliable source for high-quality DNA,^17,39  in reality, we commonly face the challenge of cytology samples with limited amounts of tumor cells. Therefore, validation studies are essential steps for any molecular testing on cytologic samples.^38,41  Cytologic samples pose unique challenges in validation, partly because of the lack of standardization among laboratories for specimen collection, processing, and staining methodologies.^11,38,42 

We have previously demonstrated that on-site evaluation by a cytopathologist is crucial for obtaining adequate cytology samples for molecular studies.³⁷  When preanalytic factors are correctly implemented, cytology samples are an excellent source of DNA for molecular testing. When needed, microdissection or macro-dissection can be easily performed (Figure 1, A). However, some types of specimens, such as brushing and exfoliative samples, have less neoplastic content, and therefore microdissection is not an option. Good-quality DNA can be extracted from stained smears, fresh fluids, cell block, and ThinPrep (Hologic Inc, Marlborough, Massachusetts). Previously, we have demonstrated that the amount of DNA extracted from cytology samples is similar to that extracted from core needle biopsies.³⁷  Core needle biopsy samples yield a higher concentration of DNA than FNA samples, but the DNA quality is similar.³⁷  Although most commercially available DNA extraction kits have been validated for unstained slides from formalin-fixed, paraffin-embedded samples, they can also be used in cytology samples with excellent results.¹⁶ 

Tissue biopsy is often not an option for patients with advanced stage cancer or metastatic NSCLC.⁴³  Then, FNA cytology offers a minimally invasive, accurate, simple, and cost-effective method for diagnosing NSCLC.³⁹  Moreover, rapid on-site evaluation helps to overcome the limitations of sampling in FNA cytology. The relatively lower amount of DNA yielded by a cytologic smear is compensated for by more sensitive genomic platforms, including NGS, that require less sample material and time.^15,36 

We have previously demonstrated that the sensitivity and specificity of EGFR and KRAS mutation analyses by Sanger sequencing, real-time polymerase chain reaction (PCR), and Cobas (Roche Molecular Systems Inc, Pleasanton, California) EGFR mutation test from cytologic smears and core needle biopsies are similar.^16,37 

It has been shown that PCR analysis using cytologic slides has better results when compared with biopsy specimens, regardless of the amount of tumor cells analyzed.^44,45  This could be explained by differences in fixation method used in the 2 types of specimens, favoring alcohol fixatives over formalin for mutational studies.⁴⁴  Therefore, as cytology samples are often collected in non–formalin-based fixatives, they offer the possibility of testing of higher-quality nucleic acids.⁴² 

In general, the amount of extracted DNA depends on the technique used.¹⁶  For DNA extraction, cells from cytology smears can be removed from glass slides by scraping. It has been shown that scraping yields significantly more DNA than cell lifting.⁴⁶  We have previously shown that cytology-based EGFR and KRAS mutational tests are as effective as tumor tissue–based tests. In addition, various types of cytology specimens, including FNA, bronchoscopy-guided FNA, direct aspirations, and pleural effusion, are shown to be suitable for molecular testing.^37,45  Direct smear preparations provide a high overall tumor fraction and higher average DNA yields than cell block preparation.³⁷  Routinely in our laboratory, after classic xylene coverslip removal, Papanicolaou-stained smears are first rinsed in alcohol and then scraped into Eppendorf tubes. Slides are not destained prior to DNA extraction.¹⁶ 

If possible, it is recommended that cytology smears be digitally scanned before performing a molecular analysis to record the most representative and diagnostic slide.^14,36,40 

Next-generation sequencing technology is a high-throughput method that allows massive parallel sequencing of multiple targeted genomic regions in multiple samples.⁴⁷  It requires a very low input of DNA, in contrast to traditional sequencing methods.⁴⁷  A variety of genomic alterations can be detected with high accuracy and sensitivity. Moreover, it allows quantitative evaluation of the mutated allele.^15,47 

Next-generation sequencing results from stained cytologic samples are similar to the ones from their snap-frozen, dry pellet counterparts, indicating that preanalytical factors, such as fixation and staining of cytologic specimens, do not induce significant detectable modifications in nucleic acid.⁴⁸ 

In a proof-of-concept study, an NGS gene panel that targets 568 clinically relevant mutations in 6 genes (EGFR, KRAS, NRAS, BRAF, cKIT, and PDGFRα) was designed and tested using samples with a very low DNA content.⁴¹  The NGS assay showed excellent sensitivity and specificity, being able to detect relevant mutations in cell-free DNA purified from samples with a very low DNA, including serum and plasma.¹⁴ 

In the last 2 years, our institution has adopted the Oncomine Focus Assay (Thermo Fisher Scientific, San Francisco, California), a targeted, multi-biomarker NGS assay that analyzes clinically relevant gene alterations including single-nucleotide variants, short insertions and deletions, somatic copy number variations, and gene fusions from DNA and RNA in a single workflow (Figure 1, F and G). This commercially available assay with low DNA requirements (∼10 ng, approximately 1500 intact cells) is being increasingly adopted in many laboratories for testing samples with limited material. This assay is able to detect amplifications and fusions in addition to mutations, including ALK and ROS1 gene fusions.

The FISH assay is used for detection of ALK and ROS1 rearrangements, using dual-labeled, break-apart probes.^11,49  Direct smears are currently being used for FISH assays. Nuclei on direct smears are not truncated, which allows for the detection of the true number of FISH signals (Figure 2, A and B).⁵⁰ 

Immunohistochemistry detects the expression of ALK and ROS1 protein and is a screening tool for testing NSCLC samples for ALK and ROS1 rearrangement.^11,18,49  ROS1 IHC should be followed by confirmatory FISH testing.¹⁸  ALK or ROS1 FISH testing is positive if rearrangement is seen in at least 15% of cancer cells.¹⁸  In specimens positive for ALK (Figure 2, C and D), (1) if ALK rearrangement has occurred, nuclei will contain “broken-apart” red and green signals, which appear separated by at least 2 signal diameters; and (2) if ALK deletion has occurred, nuclei will contain single red signals.²³  In specimens positive for ROS1, (1) in the classical pattern, 1 fusion signal (native ROS1) and broken-apart green and red signals are seen; (2) in the atypical pattern, 1 fusion signal (native ROS1) and 1 green signal without the corresponding red signal are seen.⁴⁹ 

Cell block preparation, Diff-Quik– and Papanicolaou-stained smears, and ThinPrep slides are used for ALK rearrangement analysis by FISH.³⁸  Cell blocks have been traditionally used for ALK rearrangement analysis, because the same protocols as for histology can be used. Ethanol-based fixation of cytologic smears is suitable for FISH assay, although laboratory protocols may well differ from those for formalin-fixed tissue.³⁵ 

In our laboratory, we favor direct smears over cell blocks for FISH analysis (Figure 2, A through D). On direct smears, the entire cell nucleus is analyzed, eliminating signal loss from truncation artifacts.⁴⁵  We select an area of the smear where cells are not overlapped in order to analyze entire and individual nuclei. When performing a FISH assay, we mark the area previously selected and add the hybridization probe only in this area. It is neither necessary nor cost-effective to use the whole slide. In our experience, overlapped cells and areas of nuclear debris should be avoided. Monolayered areas with at least 100 tumor cells with entire nuclei should be used to assess the FISH signals.

The pathologic diagnosis of lung cancer has become a multistep process.⁴³  Although assays have not been validated on cytology specimens, cytopathologists are increasingly being asked to assess the expression of PD-L1 in NSCLC to select patients for treatment with PD-1/PD-L1 inhibitors.

PD-L1 protein expression, as detected by IHC testing, has been used as a predictive biomarker assay for anti–PD-1/PD-L1 therapies.⁵¹  An assay for assessing PD-L1 expression is approved by the US Food and Drug Administration and the European Medicines Agency for both first-line and second-line therapy with pembrolizumab.⁷ 

The tissue-handling procedure for PD-L1 IHC does not differ from that for other diagnostic or predictive IHC markers.⁹  There are several recommendations regarding preanalytic conditions for preparation and storage of tissues for future PD-L1 staining. Critical steps are (1) minimum fixation time (6–48 hours for biopsies and 24–48 hours for resections), (2) storage time for cut sections (fewer than 2 months), and (3) storage time for blocks (less than 3 years).⁵²  PD-L1 IHC on decalcified tissue should be avoided or interpreted with great caution until further validation studies on PD-L1 IHC have become available.

PD-L1 protein expression is seen on dendritic cells, macrophages, mast cells, T- and B-lymphocytes, endothelial cells, and tumor cells. The different commercial assays available define PD-L1 staining as complete circumferential or partial linear cell membrane staining of tumor cells at any intensity.^9,43  Only cytoplasmic staining in tumor cells is not considered positive for scoring purposes. PD-L1 expression on tumor cells and immune cells is scored as a percentage of total cells.

The use of cytology samples for assessing PD-L1 protein expression is not advocated because none of the assays are validated for this purpose. PD-L1 staining and quantitation on cytology samples could be feasible, provided that appropriate protocols and validation studies are in place. In the case of alcohol-fixed samples, the PD-L1 IHC protocol may require some adjustments.

Emerging data from cell blocks and matched histologic specimens suggest that cytologic material is as good as histologic material for PD-L1 IHC.⁵³  Actually, several preliminary studies have shown that the development of cytologic criteria and interpretation of the PD-L1 IHC on cell blocks is feasible, but may require adjustments of standard PD-L1 interpretation parameters to account for differences in processing method.^6,53  In cytology samples, quantification of PD-L1–positive immune cells will likely be challenging.⁹  In our institution, cytology samples are an important method for diagnosing lung cancer. We are currently optimizing PD-L1 staining on direct smear (Figure 3). The PD-L1 IHC is being performed and evaluated in the same way as the corresponding clinically validated commercial assay. We are evaluating 100 or more viable tumor cells, and positive PD-L1 staining is defined as complete circumferential or partial linear cell membrane staining of tumor cells at any intensity (Figure 3).

Investigating promoter methylation in patients with lung cancer is, in general, challenging. In the majority of cases, lung cancer is linked to smoking, and smoking itself leads to gene methylation changes.⁵⁴  Therefore, finding good and reliable methylation-specific biomarkers may rely on the choice of an appropriate control group or on having a combination of 2 or more genes that could reduce the probability of having a false-positive or false-negative result.

The detection of the methylation level of certain genes is currently being used as a diagnostic tool and a prognostic biomarker in lung cancer, but only a few have been validated in cytology samples. In general, methylation studies are based on DNA bisulfite conversion, which leads to DNA fragmentation. After bisulfite conversion, the fraction of circulating tumor cell DNA obtained from a cytology sample is often very low and can be difficult to detect with current approaches. Protocols for bisulfite conversion and DNA methylation–specific PCR have been optimized for cytology smear.⁵⁵ 

Quantitative methylation-specific real-time PCR is currently being used to detect aberrant patterns of DNA methylation of several genes, including p16-INK4a and SHOX2.^33,56,57  This technique needs a relatively high amount of bisulfite-converted DNA (∼50 ng), which is often difficult to obtain from small specimens. Additionally, bisulfite conversion will reduce even more the final fraction of good DNA needed for methylation studies. Therefore, the development of new assays using more sensitive and quantitative techniques with a very low input of DNA are needed.

Recently, droplet digital PCR (ddPCR) has been shown to reliably detect genomic alteration at low allele frequencies and with high sensitivity (Figure 4).⁵⁸  Droplet digital PCR is achieved by partitioning a standard PCR into many replicate reactions prior to thermal cycling.⁵⁹  This results in the physical separation of DNA molecules, thus allowing binary classification of compartments that either contain or do not contain the amplified target sequence.⁵⁹  Copy numbers can then be calculated by applying the Poisson distribution.⁶⁰  Therefore, ddPCR allows absolute measure of target DNA.⁶⁰ 

Using ddPCR for the detection of DNA methylation in the TMPRSS4 promoter gene in bisulfite-converted samples, we were able to detect TMPRSS4 promoter hypomethylation with very low input of DNA, as low as 0.5 ng (3.3 copies/μL; Figure 4). It has also been shown that ddPCR can detect amounts as low as 0.5% methylated DNA in the background of a nonmethylated control DNA. Therefore, ddPCR is a very promising tool, suitable for measuring methylation in samples with low DNA concentrations, including cytology samples.

The promoter DNA methylation ddPCR assay for cytology sample analysis could improve lung cancer diagnosis and staging.

Adequate management of cytology samples, including direct smears, is beneficial for patients with NSCLC, especially when tissue biopsy is not an option. Cytology samples are suitable for determining the genetic profile of a tumor, eliminating the need for invasive tissue biopsies. This is particularly important in lung cancer, where a very high percentage of patients may have no biopsies taken or have biopsies with insufficient tumor tissue for molecular studies. Rapid on-site evaluation from FNA procedures has improved the adequacy and management of cytology samples.

More sensitive approaches, such as NGS and ddPCR, combined with minimally invasive procedures, such as FNA, may be used to better understand the disease, treat patients, and monitor treatment response in NSCLC patients.

We thank Nerea Gomez, BSc; Mariam Maset, BSc; María Mercedes Aguirre, BSc; and Maria Eugenia Echarri, BSc, for the expert technical help. We also thank Maria Villalba, MSc; Miriam Redrado, MSc, PhD; and Francisco Expósito, MSc, for the methylation studies. We thank CIMA Labs Diagnostics for assistance in the molecular work. This work was supported by PIUNA (University of Navarra; ref. 12028701) and ERA-NET on Translational Cancer Research (TRANSCAN; ref. AC14/00034).