Next-generation sequencing analysis of endometrial screening liquid-based cytology specimens: a comparative study to tissue specimens

Background Liquid-based cytology (LBC) is now a widely used method for cytologic screening and cancer diagnosis. Since the cells are fixed with alcohol-based fixatives, and the specimens are stored in a liquid condition, LBC specimens are suitable for genetic analyses. Methods Here, we established a small cancer gene panel, including 60 genes and 17 microsatellite markers for next-generation sequencing, and applied to residual LBC specimens obtained by endometrial cancer screening to compare with corresponding formalin-fixed paraffin-embedded (FFPE) tissues. Results A total of 49 FFPE and LBC specimens (n = 24) were analyzed, revealing characteristic mutations for endometrial cancer, including PTEN, CTNNB1, PIK3CA, and PIK3R1 mutations. Eight cases had higher scores for both tumor mutation burden (TMB) and microsatellite instability (MSI), which agree with defective mismatch repair (MMR) protein expression. Paired endometrial LBC, and biopsied and/or resected FFPE tissues from 7 cases, presented almost identical mutations, TMB, and MSI profiles in all cases. Conclusion These findings demonstrate that our ad hoc cancer gene panel enabled the detection of therapeutically actionable gene mutations in endometrial LBC and FFPE specimens. Endometrial cancer LBC specimens offer an alternative and affordable source of molecular testing materials.


Background
Endometrial cancer is the most common gynecologic cancer worldwide, affecting over 500,000 women every year [1]. Since lifestyle changes can affect the risk of developing endometrial cancer, an epidemiological approach is essential to manage the disease [2]. Moreover, genome-wide association studies have identified new susceptibility loci for endometrial cancer in the human genome, providing candidate genes for further studies aiming to unravel the mechanisms driving carcinogenesis, which could offer new opportunities for early screening and identification of therapeutic targets [3]. Unlike squamous cell carcinoma of the uterine cervix, which is characterized by pathogenesis closely related to that of human papilloma virus infection [4], endometrial carcinogenesis is linked to common cancer gene mutations and genome instability [5,6].
Cytology specimens, including conventional smear, cytospin, liquid-based cytology (LBC), and cell block, are valuable sources of routine molecular diagnostics [7]. Extensive studies have been reported that a variety of cytology materials, containing LBC specimens suitable for next-generation sequencing (NSG) testing, especially in non-small cell lung cancer [7][8][9][10][11][12]. Meanwhile, the endometrial cytology is not only a safe and easy clinical procedure but also comparable to suction endometrial biopsy for detecting atypical endometrial hyperplasia and cancers [13][14][15][16]. Since 1987, in Japan, a cytologic examination for endometrial cancer screening has been established by the National Health Insurance law. After that, endometrial cytology, using a specific device, was introduced as a less invasive and less painful tool [17,18] to screen high-risk population and perimenopausal women with abnormal genital bleeding. Conventional endometrial cytology preparation, however, has some disadvantages such as bloody background, cellular overlapping, and thick cell clusters; therefore, the resulting cytological screening would not be widely accepted in endometrial cancer management except in Japan [19]. After the introduction of LBC in endometrial cytology in combination with transvaginal sonography, LBC-based endometrial cytology, a less invasive and expensive procedure, was proposed to screen cancer also in asymptomatic postmenopausal women [20,21].
To date, LBC represents a routine endometrial cytology method for cancer diagnosis and gene analysis [22]. We also previously reported that residual LBC samples could better preserve the genome quality for downstream NSG, as compared to formalin-fixed paraffinembedded (FFPE) tissues, even after several years of storage [23]. In the present study, we established a small custom cancer gene panel with full exon coverage of 60 cancer-related genes and 17 microsatellite foci. We then used this custom panel to comparatively analyze the genetic profiles obtained from residual endometrial LBC specimens and corresponding FFPE tissues with respect to the presence of somatic gene mutations, tumor mutation burden (TMB), and microsatellite instability (MSI). We further conducted immunohistochemistry for detecting the expression of mismatch repair (MMR) proteins in the FFPE sections to compare the results obtained by genomic analysis. The detection of MMR protein expression is known as a reliable proxy of the MSI status [24]. This study demonstrates the feasibility of using a small NGS-based cancer gene panel to screen for endometrial cancer in residual LBC specimens. The reliability of the results was compared with those obtained from biopsied and resected FFPE tissue analysis.

Histological, cytological and blood specimens
In this study, 53 tissue and cytology samples from 24 patients with endometrial cancer registered in the Clinical Research of Cancer Gene Panel Analysis of Gynecologic Cancers study from January to August 2019 at Kagoshima University Hospital were used. The specimens included 2 mL of whole blood as a reference, FFPE tissues from 10 biopsied, 29 resected tissues (39 FFPE tissues in total), and 10 residual LBC samples of endometrial cytology. The tissues were fixed with phosphate-buffered neutral 10% formalin within 24 h and routinely processed for paraffin embedding, followed by sectioning for hematoxylin and eosin (H&E) staining. The FFPE blocks were stored in the dark at room temperature (25 ± 2°C). The endometrial cytology samples were obtained using a specific device (Endocyte® sampler, Laboratoire CCD, Paris, France) [17,18], and collected cells were fixed with CytoRich Red solution (Becton Dickinson, Franklin Lakes, NJ, USA). The cytology samples were routinely processed for LBC using BD SurePath liquid-based Pap Test System (Becton Dickinson). The residual LBC specimens were stored at 4°C.
The pathological diagnosis was made by two boardcertified pathologists (IK and AT), according to the World Health Organization classification [25], and the tumor fraction in the sections (% area) was evaluated in 10% increments. Similarly, the cytological diagnosis was made by two board-certified cytopathologists (IK and TH), according to the classification system proposed by the Japanese Society of Clinical Cytology [26]. For cytology specimens, atypical or tumor cell numbers were counted in five highpower fields (HPF) at × 400 magnification. The average counts were calculated and the total tumor cells (T) subjected to NGS were estimated. Similarly, non-tumor cell counts were calculated as the average of five HPF, and total non-tumor cells (N) were estimated, and then the tumor cell ratio (T/N + T) was calculated. The cytological and pathological studies were performed independently from the molecular studies described below.

DNA extraction and quality check
Whole blood DNA was extracted using the QIAamp DNA Blood Mini Kit (Qiagen). DNA from FFPE tissues was obtained from 3 to 6 sections with 10-μm thickness, including more than 30% of the cancer area, or after microdissection using a microdissection laser (Leica Biosystems, Nussloch GmbH, Germany). One mL of residual LBC samples was centrifuged (× 12,000 g), and the pellets were resuspended in 95% ethanol and air-dried for 5 min. Both the FFPE sections and LBC pellets were then incubated with proteinase K (Promega, Madison, WI, USA) for 15 h at 70°C, followed by a 1-h incubation at 98°C in lysis buffer (Promega). After centrifugation (× 12,000 g), the supernatants were applied to the Maxwell RSC DNA FFPE kit and Maxwell 16 system (Promega). After measuring the extracted DNA concentration using the Qubit 3.0 dsDNA BR assay kit (Life Technologies, Grand Island, NY, USA), DNA qualities were monitored using the QIAseq DNA quantimize kit (Qiagen). A quality check score (QC score) less than 0.04 was considered as a high-quality DNA, according to our previous report [23].

Design of the custom gene panel
A total of 60 cancer-related genes and 17 microsatellite foci [27] were selected from QIAseq Targeted DNA Custom Panel (Qiagen [28]), including 2615 primers for the regions of interest, with an average exon coverage of 99.87%. The cancer-related genes were used for the construction of a custom panel (detailed in Table 1). The selection of these 60 genes was validated according to the Catalogue Of Somatic Mutations In Cancer (COS-MIC ver. 90, https://cancer.sanger.ac.uk/cosmic) to include target genes of molecular targeting drugs for solid cancers, and genes frequently expressed in ovarian and endometrial cancers.

NGS analysis
Forty ng of DNA from whole blood cells and LBC specimens and 100-200 ng of DNA from FFPE tissue sections were used to construct the NGS library based on QIAseq Targeted DNA Custom Panel (Qiagen). The quality of the libraries was monitored using an Agilent highsensitivity DNA kit (Agilent Technologies, Santa Clara, CA, USA) to confirm the successful generation of 300bp PCR products. The libraries were then applied to MiSeq sequencer (Illumina, San Diego, CA, USA) after dilution with a hybridization buffer to a final concentration of 20 pM. The obtained sequencing data were analyzed by the Qiagen Web Portal service (https://www. qiagen.com/us/shop/genes-and-pathways/data-analysiscenter-overview-page/ ). Qiagen Web Portal service, equipped with a smCounter analyzing pipeline [29], indicates that a mean unique molecular index (UMI) sequence depth of 2,000 is necessary for detection of 0.5% variant allele frequency (VAF), and the cut-off value of VAF varies, inversely proportional to the mean sequence depth. Our NGS, with a condition of tumor fraction ≥20% and cluster passing filter rate ≥ 90%, yielding a mean UMI depth of 500, at which the SNP detection were analyzed as controls, and the genomic sequence data were compared to those deposited in the COSMIC database. In this study, DNA sequence data obtained from whole blood were used only for reference, whereas germline analysis was not performed.

Calculation of TMB and MSI scores
The numbers of missense mutations, including nonsynonymous mutations and internal deletions, with more than 10% VAF, were counted as somatic mutations. TMB was calculated as the number of single nucleotide variants/Mbp of DNA sequence [30,31]. MSI scores were determined by MSIsensor (ver. 1.0) [32,33].

Immunohistochemistry for MMR proteins
To evaluate the expression of MMR proteins, FFPE tissue sections were applied to immunohistochemistry (IHC) using antibodies against MLH1 (clone ES05 M3640), MSH2 (clone FE11 M3639), MSH6 (clone EP49 M3646) and PMS2 (clone EP51 M3647) purchased from DAKO (Tokyo, Japan) [34,35]. Staining was performed on representative 5-μm-thick FFPE sections with Envision FLEX High pH K8000 system (DAKO) according to the manufacturer's protocol. Positive nuclear staining of lymphocytes in the tissue sections was monitored as a positive control. MMR deficiency was defined as complete loss of nuclear staining for both MLH1 and PMS2, both MSH2 and MSH6, MSH6 only, or PMS2 only. When partial loss of each protein was observed, the case was defined as showing heterogeneous expression.

Statistical analyses
All values are expressed as the mean ± standard deviation. Significant differences were analyzed using Welch's t-test and Mann-Whitney U test. Values of p < 0.05 were considered statistically significant. The cut-off values for the evaluation of TMB-high and MSI-high conditions were determined by receiver operator characteristic (ROC) curves.

Ethical approval for the genome studies
The studies using clinical samples were approved by the Ethics Committees for Clinical and Epidemiologic Research at Kagoshima University, and written-informed consent was obtained from each participant. The studies did not include participants younger than the age of 20.

DNA quality and quantity obtained from FFPE and LBC specimens
The results of the pathological diagnosis, DNA quality, and input DNA for NGS of the 24 endometrial cancer cases are summarized in Table 2. The storage period of FFPE and LBC specimens ranged from 2 weeks to 3 years. The tumor fraction of the FFPE sections ranged from 30 to 90%. In endometrial LBC specimens, estimated tumor cell count (T) subjected to NGS varied from 1500 to 140,000, and the tumor cell ratio varied from 27 to 95% (T/N + T). All of the genomic DNA extracted from the FFPE and LBC samples demonstrated high-quality and sufficient quantity for library construction and successful sequencing.

Mutations detected in FFPE specimens
The detected mutations in endometrial cancers and genomic information of the variants are summarized in supplemental Tables S1 and S2. Among the 24 cases, 18 were finally diagnosed as endometrioid carcinoma (EC) by biopsy or surgical resection, including 12 cases of G1, 5 cases of G2, and 1 case of G3 EC. The other 6 cases consisted of 3 dedifferentiated carcinomas (DC), and 1 case each of mixed EC/serous carcinoma (SC), SC, and clear cell carcinoma (CCC). These endometrial cancers showed common mutation profiles, including PTEN, CTNNB1, PIK3CA, and PIK3R1 mutations.
The cases of mixed EC/SC, CCC, and SC had additional TP53 mutations. Two different FFPE sections relative to case no. 6 were analyzed, revealing G1 and G2 EC. While NGS analysis unveiled common mutations in PTEN, CTNNB1, and ARID1A, different PIK3CA and PIK3R1 mutations were also detected, suggesting the existence of at least two cancer clones. The three DC cases harbored PTEN, CTNNB1, PIK3CA, or PIK3R1 mutations along with multiple mutations in receptortype tyrosine kinase genes, such as FGFR, ERBB, RET, and FLT. The mutation profiles of the EC and DC sections were not completely comparable but exhibited similarity with respect to PTEN mutation.

Mutations detected in LBC specimens
Endometrial LBC specimens contained abundant atypical cells, resulting in a higher frequency of mutation detection in the endometrial LBC specimens (9 out of 10 cases; supplemental Table S1). Mutations in PTEN, CTNNB1, and PIK3CA were identified in 1 case of atypical cell cytology (case no. 14), in which the diagnosis of G1 EC was confirmed by endometrial curettage biopsy.

Relations between MMR protein expression, TMB, and MSI
The overall relationships between MMR protein deficiency (MMR-D), TMB, and MSI status are shown (Fig. 1). In cases of MMR-D (9 cases, 21 samples), the TMB score was significantly higher than in cases of MMR protein proficiency (MMR-P) (15 cases, 27 samples) (p < 0.001; Fig. 2a, left). The MSI score of MMR-D cases was also significantly higher than that in MMR-P cases (p < 0.001; Fig. 2b, left). Estimated cut-off values for the evaluation of TMB-high (TMB-H) and MSI-high (MSI-H) were > 31.1 and > 4.2, respectively, as determined by the ROC curve ( Fig. 2a and b). The IHC results, TMB, and MSI status are summarized in supplemental Table S3. The majority of MMR-D cases (8 of 9 cases), but not MMR-P cases, showed both TMB-H and MSI-H. The exception was the case  Fig. 3. Very similar findings were obtained from cases no. 20 and 22, and both the TMB and MSI scores were over the cut-off values in these cases.

Correlation of the genetic diagnosis from LBC and FFPE specimens
In 10 cases (case nos. 9,12,14,16,18,19,[21][22][23][24], the FFPE tissues from biopsy and/or resection along with endometrial LBC, were subjected to gene panel analysis. For cases no. 14 and 22, biopsied and endometrial LBC samples were analyzed together, revealing the same mutations. In 8 cases (case nos. 9, 12, 16, 18, 21, 23, and 24), 3 paired endometrial LBC, biopsied, and resection FFPE specimens were available for genetic studies. Seven of these 8 cases (except for case no. 18) exhibited almost identical mutation profiles and a similar TMB/MSI status. The VAF detected from 7 endometrial LBC specimens (case nos. 9, 12, 14, 16, 21, 23, and 24) showed a remarkably similar VAF to that of the same gene mutations from the corresponding biopsied FFPE (Fig. 4a). In case no. 22, VAF of APC markedly changed from 85.6% in LBC to 58.9% in biopsied FFPE, possibly because chemotherapy resulted in tumor clonal selection after LBC sampling and before the biopsy, and, therefore, the data were excluded from the Fig. 4a. All together with the results obtained from these 7 cases, a fine concordance was observed between endometrial LBC, and biopsied and/or resected FFPE tissue specimens in numbers of mutations, TMB and MSI scores, and MMR protein expression (Fig. 4b).

Discussion
We showed that LBC specimens offered sufficient DNA quality and quantity for NGS analyses. Endometrial LBC samples contained a much higher tumor cell ratio, allowing for NGS analysis to be efficiently performed. The endometrial LBC specimen containing at least 1450 tumor cells, and 10 ng of input DNA were successfully analyzed with the present protocol. A previous report suggested that 5000 viable cells are necessary for successful NGS using a 50-cancer gene hotspot panel from conventional cytology samples [36]. Therefore, appropriate cytology specimens containing more than several thousand cells with a higher tumor ratio could be an alternative source for genetic analysis to FFPE and conventional cytology specimens. However, the use of LBC specimens for NGS analysis faces a few disadvantages [37,38]. Unlike NGS in scraped cells from smears, in which the same cells observed by cytopathologists are subjected to molecular testing, NGS analysis from LBC specimens is started with pooled residual samples. Therefore, although we checked for the presence of tumor cells in the residual LBC specimens and then applied to DNA extraction, false-negative materials may be applied to NGS. A rapid on-site evaluation to ensure the sufficient preservation of tumor cells are unavailable in LBC specimens, either. To the best of our knowledge, another disadvantage is that an efficient method to enrich tumor cell fraction in LBC specimens has not yet been developed. In this regard, the LBC specimens, such as inflammatory cell-rich ascites, may not be always suitable for NGS. Conventional smears and cell blocks are superior to LBC for increasing tumor cell fraction by trimming techniques [38,39]. Conversely, one of the advantages of LBC specimens is the well-preserved DNA quality and yield even after long-term storage [23]. The absence of diagnostic material loss due to cell scraping and less complicated procedures represent additional advantage of LBC specimens.
Since the loss of MMR protein expression can occur by methylation and mutations in the promoter region, in addition to loss-of-function mutations in the coding region, a validation of the exon sequence would not be sufficient to evaluate the MMR expression and function [47][48][49]. Actually, in the present study, only 3 of 9 MMR-D cases presented pathogenic mutations in MLH1, MSH6, or PMS2. Although IHC is considered a practical tool for speculation of TMB and MSI conditions as a first screening step [50,51], the results of MMR protein IHC were not always comparable to those of MMR gene mutation identification. Thus, evaluation of TMB and MSI using NGS-based genome analysis along with IHC would be a practical strategy for clinical testing [52][53][54]. NGS analysis with endometrial LBC specimens, which were not usually used for immunocytochemistry of MMR proteins, consistently detected TMB-H and MSI-H. Therefore, our custom-made panel would possibly be beneficial to determine TMB-H and MSI-H status. However, the sample size was relatively small, and the NGS panel is also insufficient to comprehensively detect TMB and MSI at the whole-exome sequencing level, in which TMB calculation algorithm often weights down mutation data from hotspot to prevent TMB over-estimation [32,33,55,56].

Conclusion
The NGS-based panel with coverage of 60 cancer genes and 17 microsatellite foci demonstrated highly recurrent somatic mutations, TMB, and MSI in FFPE and endometrial LBC specimens. In addition, the endometrial cytology can efficiently collect cells from the entire part of the endometrial cavity to predict histological subtypes [16] and also is available to detect asymptomatic endometrial cancer by the less invasive cytologic screening [14]. Therefore, the small size cancer gene panel and endometrial LBC specimens would be an alternative tool for genetic testing as a diagnostic or therapeutic strategy. As in the case of NGS analysis from FFPE tissues, it is vital that cytopathologists correctly estimate the amounts of tumor cells, to select suitable LBC specimens for molecular diagnosis appropriately [7].