Detection of PIK3CA Gene Mutation in Head and Neck Squamous Cell Carcinoma Using Droplet Digital PCR and RT-qPCR

Head and neck squamous cell carcinomas (HNSCC) are the seventh cause of human malignancy with low survival rate due to late diagnosis and treatment. Its etiology is diverse; however genetic factors are significant. The most common mutations in HNSCC were found in the genes: PIK3CA (10–12%), BRCA1 (6%), and BRCA2 (7–9%). In some cases, these biomarkers correlate with recurrences or survival showing a potential of prognostic and predictive value. A total of 113 formalin-fixed paraffin embedded (FFPE) tumor samples were collected from patients with HNSCC (oral cavity: 35 (31.0%); oropharynx: 30 (26.0%); larynx: 48 (43.0%)). We examined PIK3CA H1047R mutation by Real Time PCR (RT-qPCR) and droplet digital PCR (ddPCR). BRCA1 and BRCA2 mutations were analyzed by RT-qPCR while p16 protein expression was assessed by immunohistochemistry. Finally, we identified HPV infection by RT-qPCR. The relationships between genomic alterations and clinical parameters were assessed using the Yates’ corrected Chi-squared test or Fisher’s exact test for nominal variables. Kaplan Meier plots were applied for survival analysis. Our results revealed 9 PIK3CA H1047R mutations detected by ddPCR: 8 of them were negative in RT-qPCR. Due to the use of different methods to test the presence of the PIK3CA gene mutation, different treatment decisions might be made. That is why it is so important to use the most sensitive methods available. We confirmed the usefulness of ddPCR in the PIK3CA mutation assessment in FFPE samples.


Introduction
Cancer detection, treatment and prevention was, until recently, something of a "onesize-fits-all". Current practice guidelines recommended chemotherapy based on risk and not on predictive factors. The genetic signature of cancer may help predict the risk of recurrence or identify the best treatment strategy. The number of gene mutations implicated in cancer is growing and the number of drugs being developed to target specific mutations is also rising. Modern technologies offer personalized cancer therapies [1]. However, they are not fully available to patients with head and neck squamous cell carcinomas (HNSCC)-a group of malignant tumors of that region, excluding the brain and eyeball. HNSCC accounts for 5% of all cancer cases and is the seventh cause of human malignancy [2]. Worldwide, the number of patients who are diagnosed with HNSCC is estimated to be 800,000, with 430,000 deaths every year [2,3]. More than 90% of head and neck cancers originate from squamous epithelial cells, and affect mostly the upper respiratory tract, including larynx (24%), oral cavity (40%) and pharynx (35%). They are most often recognized and treated at an advanced stage. Despite advances in diagnostic and therapeutic techniques, the 5-year survival rate has not improved for decades and remains low [4][5][6].
usefulness of selected methods in the assessment of the mutation H1047R of PIK3CA gene. We also assessed the most frequent mutations of BRCA1/2, infection of HPV virus and p16 expression in patients' samples to get a picture of the biological potential of the tumor.

Group Characteristics
Tumor specimens were collected from 113 patients diagnosed with head and neck squamous cell carcinoma surgically treated at the Department of Otolaryngology, Head and Neck Oncology, Medical University of Lodz, from 2010 to 2016. The group consisted of Caucasians from the same residential area. Patients were staged using TNM staging system in accordance with the 7th edition and subsequently re-evaluated in accordance with the 8th edition of the American Joint Committee on Cancer Staging of Head and Neck Cancer [7,34,35]. There was survival data available for all of the group. Patients whose death was not related to the primary disease were excluded from the analysis. The observation time varied from 2 to 167 months with a mean of 54.9 months. Time from primary surgical treatment to the onset of local or nodal recurrences was found in a total of 13 patients (11.60%) over a period from 6 months to 9 years (mean 40.9 months; SD 25.13; median 24 months). Patterns for age, sex, primary tumor location, clinical staging, grading, local and nodal recurrence were correlated with HPV status. Tobacco exposure was expressed as pack-years of smoking. We distinguished between nonsmokers and smokers, among whom we differentiated light and heavy, respectively, below and above the median (7300 cigarettes per year), due to the same value in all groups. Patients due to alcohol consumption were grouped into non-drinkers and those who drank less or more than 6 beverages weekly (median). The demographic and clinical data are described in Table 1. The study was approved by the Local Ethics Committee (RNN/12/06/KE; RNN/69/07/KE). All procedures performed in studies involving human participants were under the ethical standards of the institutional and national research committee and in agreement with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants involved in the study.

DNA Isolation
Formalin-fixed paraffin embedded (FFPE) tumor samples were obtained from HNSCC tumors collected between 2010-2016. Written consent was obtained from all of the patients. All tumor samples were assessed by sectioning, hematoxylin and eosin staining under a microscope to ensure that the tumor content in each sample was higher than 50%. DNA was isolated from samples using Maxwell RSC DNA FFPE Kit (Promega Corporations, Madison, WI, USA, cat No AS1450) and automatic isolation on AS4500 device (Promega Corporations, Madison, WI, USA). The purity of extracted genomic DNA was checked by measuring the absorbance at 260 nm(A 260 ), 280 nm (A 280 ) and 230 nm (A 230 ) with a Nanodrop 2000 spectrophotometer. Extracted genomic DNA with a A 260 /A 280 ratio between 1.8-2.0 and A 260 /A 230 ratio over 2.0 were considered satisfactory to conduct analysis.

p16 Immunohistochemical Expression
The CINtec Histology Kit (Ventana Medical Systems, Oro Valley, AZ, USA, Roche) was used for p16 immunohistochemistry, following manufacturer's protocol. The evaluation was performed according to the previously published criteria [36]. The sample was considered positive when greater than 70% of carcinoma tissue was detected with strong and diffuse nuclear and cytoplasmic immunostaining for p16. Lack of reactivity or faintly diffuse was counted as p16-negative. Independent assessment by two investigators was used for staining intensity and the percentage of positively stained cells for each analyzed slide.   The RT-qPCR for the amplification of exon 20 of PIK3CA gene and detection the mutation H1047R was performed using 25 ng DNA in total volume of 25 µL as previously described [37]. Briefly, the reaction includes 12.5 µL TaqMan Universal Master Mix (Thermo Fisher Scientific, Waltham, MA, USA, cat no A30872), 2.5 µL 10× primers/probes mix (Thermo Fisher Scientific, Waltham, MA, USA, Cat no 4351379), 5 µL nuclease free water and 5 µL DNA 5 ng/µL. The PCR protocol was an initial enzyme activation step of 2 min at 5 • C and 10 min at 95 • C for denaturation, followed by 50 cycles of 95 • C for 15 s and 60 • C for 60 s. All reactions were performed in duplicates using the Real Time PCR System CFX 96 (BioRad Laboratories Inc., Hercules, CA, USA). Mutations of genes BRCA1/2 were detected using Oncogenetic BRCA Panel (Sacace Biotechnologies, Milan, Italy, cat no R-27/P-48FRT) according to manufacturer's protocol. The panel includes mutations: gene BRCA1 (185delAG, 4153delA, 5382insC, 3819delGTAAA, 3875delGTCT, T181G (Cys61Gly), 2080delA) and gene BRCA2 (6174delT). Briefly, it is a quantitative test that allows the detection by RT-qPCR based on the amplification of the genome-specific region using specific primers. In each reaction mutant (FAM probe) and wild type (HEX probe) of the allele is assessed. The PCR reactions were run in the following temperature program: 80 • C for 2 min, 94 • C for 3 min followed by 40 cycles of 95 • C 10 s and 60 • C for 40 s.

Droplet Digital PCR (ddPCR)
Analysis by ddPCR was performed on the QX100 Droplet Digital System (BioRad Laboratories Inc., Hercules, CA, USA), including primers and probes (FAM-mutant type BioRad cat no 10031246 and HEX-wild type BioRad cat no 10031249) and Supermix for probes (No dUTP) according to manufacturer's protocol [38]. Briefly, reactions were performed in 25 µL volume using ddPCR 2× Master mix (BioRad Laboratories Inc., Hercules, CA, USA), 1.25 µL 20× primer and probe mix, 8.75 µL nuclease free water and 2.5 µL template. Then 20 µL of each sample was transferred into the middle well of the cartridge (BioRad cat no 1864008) and 70 µL of droplet generation oil (BioRad cat no 1863005) was added to the lower well. Once the process of generating droplets was completed, 40 µL of mixture was transferred into the wells of a 96-well plate, sealed (BioRad cat no 1814040) and loaded in a thermal cycler (T100-BioRad Laboratories Inc., Hercules, CA, USA). PCR was performed using following conditions: 95 • C for 10 min, followed by 40 cycles of 94 • C for 30 s and 57 • C for 60 s. After thermal cycling the plate was read in the QX200 Droplet Reader and based on positive droplets and the Poisson distribution, the absolute copy number of the mutant allele of PIK3CA H1047R and wild type was calculated (QuantaSoft analysis system BioRad Laboratories Inc., Hercules, CA, USA).

Statistical Methods
Relationship between genomic alterations and clinical parameters were assessed by using the Yates' corrected Chi-squared test or Fisher's exact test for nominal variables. (STATISTICA 13, Stat-Soft Inc.). Differences were considered significant at confidence intervals greater than 95% (p < 0.05). Survival distribution were estimated by the Kaplan−Meier method. Overall survival (OS) was determined from the time of initial diagnosis to the time of death whereas recurrence-free survival (RFS) as time from the initial diagnosis to recurrence. The statistical significance of differences between survival rates was ascertained using the log-rank test. Univariate Cox proportional hazard models were performed to assess the prognostic value of clinical and genomic markers on OS. The results are presented as hazard ratio and 95% confidence interval.

Results
The 113 cases represent primary tumors located in the oral cavity (n = 35; 31.0%), the oropharynx (n = 30; 26.0%) and in the larynx (n = 48; 43.0%). With regard to the patients' data, male to female ratio was 76% versus 24%, respectively, with the average age of the patients 63.4 (ranged 42−91). This represents the normal distribution in that disease. In terms of habits, 74% were smokers or former smokers and 49% had a history of alcohol consumption. This cohort included comparable numbers of tumor stage T1−T2 (52%) and T3−T4 (48%) tumors. The prevalence of patients without metastases to the neck lymph nodes (70%) was observed.
All patients were pathologically diagnosed, and the collected tissues were tested for HPV (26/113 HPV-positive); p16 protein (11/113 abnormal expression); PIK3CA mutation (9/113-8%) and BRCA1/2 mutation (0/113). Data concerning HPV infection and p16 expression in correlation with PIK3CA mutation is shown in Figure 1. Most of the observed PIK3CA mutations (8/9, 88.9%) occurred in patients without confirmed HPV infection (8/87, 9.2% in the HPV negative group) and with normal p16 expression (N = 85) in patients with laryngeal cancer. In the HPV positive group there was one patient with PIK3CA mutations (1/26, 3.85%), diagnosed with tongue cancer and abnormal expression of p16 protein ( Figure 1). Analyzing the variables presented in Table 2, we obtained significant differences only for the correlation with the clinical stage of the disease, the stage of histopathological grade and metastases. We observed a two-fold higher risk of neck metastases in the group of women compared to men. At the same time, people with neck nodal involvement had over thirteen times higher risk of death. Patients with advanced grade (G3) of the disease had more than four times higher risk of neck metastasis. However, HPV-negative patients had more than three times lower chance of neck metastases. This is interesting because usually a better prognosis is observed in the group of people who develop the disease based on HPV infection. Moreover, patients with normal expression of the p16 protein in tumor tissue had a more than five times higher chance for a better disease course (stage T1−T2). Analyzing the variables presented in Table 2, we obtained significant differences only for the correlation with the clinical stage of the disease, the stage of histopathological grade and metastases. We observed a two-fold higher risk of neck metastases in the group of women compared to men. At the same time, people with neck nodal involvement had over thirteen times higher risk of death. Patients with advanced grade (G3) of the disease had more than four times higher risk of neck metastasis. However, HPV-negative patients had more than three times lower chance of neck metastases. This is interesting because usually a better prognosis is observed in the group of people who develop the disease based on HPV infection. Moreover, patients with normal expression of the p16 protein in tumor tissue had a more than five times higher chance for a better disease course (stage T1−T2).
Univariate Cox regression analysis presented in Table 3 did not show any significant correlations between the analyzed variables. Kaplan−Meier survival curves for recurrence free survival (RFS) in the group of patients with HNSCC showed no correlation between the studied variables (PIK3CA mutation, HPV infection and p16 expression). However, in the group of patients with the normal expression of the p16, we observed a better outcome (log-rank test p = 0.086) (Figure 2).  free survival (RFS) in the group of patients with HNSCC showed no correlation between the studied variables (PIK3CA mutation, HPV infection and p16 expression). However, in the group of patients with the normal expression of the p16, we observed a better outcome (log-rank test p = 0.086) (Figure 2).  Comparison between RT-qPCR and ddPCR shows that for only one case did both methods give equivalent results. Eight ddPCR-positive cases were negative at RT-qPCR. We did not observe any correlation for the presence of the PIK3CA H1047R gene mutation and tested variables.

Discussion
Current evidence suggests that PIK3CA mutations have predictive and prognostic value, although their clinical significance remains unclear [14]. The mutational spectrum that has been reported in the literature by tissue tumor sequencing, with TP53 and genomic alterations in the PI3K pathway being among the most frequent events in HNSCC Comparison between RT-qPCR and ddPCR shows that for only one case did both methods give equivalent results. Eight ddPCR-positive cases were negative at RT-qPCR. We did not observe any correlation for the presence of the PIK3CA H1047R gene mutation and tested variables.

Discussion
Current evidence suggests that PIK3CA mutations have predictive and prognostic value, although their clinical significance remains unclear [14]. The mutational spectrum that has been reported in the literature by tissue tumor sequencing, with TP53 and genomic alterations in the PI3K pathway being among the most frequent events in HNSCC [39]. In our study, we compared two sensitive molecular methods and we found that ddPCR detected mutations with higher sensitivity in FFPE samples from HNSCC patients (RT-qPCR detected mutation in one case and ddPCR in nine cases). A total of nine mutations were detected in the patients' group, which is in line with the frequency of these changes reported for head and neck cancers (8%). The majority of them (8/9, 88.9%) were observed in HPV negative cases of HNSCC. PIK3CA mutations in the helical domain (E542K, E545K) were more often observed in HPV positive samples of HNSCC while in HNSCC HPV negative cases, PIK3CA mutations were mostly found across the entire gene. [22]. We confirmed the H1047R mutation (located in kinase domain) in exon 20 of PIK3CA gene mainly in HPV negative HNSCC (only 1/9 in HPV positive HNSCC). Nichols et al. found three H1047R mutations in 87 analyzed patients, which is 3.45% (all three mutations were detected in HPV negative cases) [22]. In line with our results other studies have reported a low concordance rate for these two methods [40][41][42]. Lui et al reported 12.6% mutations of PIK3CA in HNSCC (46/151), which was substantially fewer than the number reported by TCGA (in total 21%, 58/279 cases, including: 32/279, 11.7% of E545K/E542K mutations and 26/279, 9.3% of H4047R mutation) [9]. Our result of 8% for the H1047R mutation in the PIK3CA gene remains consistent with the data from the TCGA database [9,12]. Beavers et al. reported that a point mutation (PIK3CA E545K), detected at a fractional abundance of 28.9% in the primary tumor tissue by ddPCR, could not be identified by Sanger sequencing. Due to the use of different methods to test for the presence of the PIK3CA gene mutation and the failure to detect an existing mutation, the wrong patient treatment decision might be made. These authors suggested that when FFPE DNA was used for PCR and Sanger sequencing, a phenomenon of "allelic drop out" could occur, due to sample degradation. ddPCR was less prone to this effect because of the smaller amplicon size used in this technique. The ddPCR was also a reliable, highly sensitive alternative method for the detection of the BRAF V600E mutation in papillary thyroid carcinoma (it confirmed the presence of the mutation in 92 samples whereas Sanger sequencing in 67 samples) [43]. The authors concluded that Sanger sequencing is a gold standard and a widely used method in laboratories for detecting mutations but because of its relatively low sensitivity (in 25 samples Sanger sequencing indicated wild-type BRAF instead of mutated) the detection requires a large amount of tumor DNA in the sample [43]. Wang et al. compared the results from Sanger direct sequencing (as the standard), ddPCR, and quantitative real-time PCR (qRT-PCR) [43]. Their data indicated that ddPCR was much more sensitive and much easier to interpret than qRT-PCR [44]. The results published by Zhang et al. demonstrated that the ddPCR approach reliably detected plasmid samples with 5% (398 copies), 1% (57 copies), 0.5% (24 copies) and even 0.1% (average 6 copies) mutation rate whereas qPCR at the levels of 5 and 1% [45]. The other results revealed that both methods had high analytical sensitivity and they were capable of detecting the JAK2 V617F mutation with a limit of detection of 0.12% for RT-qPCR and 0.01% for ddPCR [46]. The latter work showed the significant differences in the possibilities of detecting mutations of both techniques. This is in line with the results we obtained. The differences between methods are presented in Table 4. They show that for HNSCC and breast cancer, the PIK3CA gene mutation detection tests were conducted in various biological materials (FFPE, serum, plasma, frozen tissue, cell lines) and with the use of various research techniques (RT-qPCR, Sanger sequencing, NGS, ddPCR) which affected the received sensitivity of detected changes [19,20,23,[47][48][49][50][51][52][53][54][55][56]. This translates into the quality and quantity of the obtained results (including false negative and false positive). At the moment the Sanger sequencing has been the gold standard for verifying such mistakes. Unfortunately, this is no longer possible due to the aforementioned limitations of this method. Patients waiting for treatment need a selected type of therapy, based on the molecular profile of the tumor. Techniques such as ddPCR and NGS make it possible more and more often, although of these two methods the latter is certainly cheaper [57][58][59]. Table 4. PIK3CA mutations and amplifications in HNSCC, breast cancer and PROS (Abbreviations: HNSCC-head and neck squamous cell carcinoma, OSCC-oral squamous cell carcinoma, PROS-PIK3CA-related overgrowth spectrum, HPV-human papilloma virus, PIK3CA-phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha gene, FFPE-formalin-fixed paraffin-embedded, FB-fibroblast, qPCR-real time quantitative polymerase chain reaction, UDT-Seq-ultra-deep targeted sequencing, NGS-next generation sequencing, ddPCR-droplet digital PCR, BEAMing-beads, emulsification, amplification and magnetics).

Study
Year In our group of patients, we did not detect any of the eight BRCA1/2 mutations tested. Comparable results were obtained by Moslehi et al. in a study of Ashkenazy Jews. In the group of patients with HNSCC, they found an exceptionally low (0%) frequency of BRCA1/2 germline mutations despite the fact that in this population there is a founder effect for these changes [60]. In the Polish population we also have such an effect described. Perhaps mutations inversely correlate with cancers of the neck and head. In terms of the characteristics of the study group, with regard to HPV and p16 protein expression, our results were consistent with those described earlier. Only the correlation of the absence of HPV infection with a better prognosis is an interesting observation. Taking into account the survival curves, we did not find any correlation with the analyzed variables either, except for the trend related to better prognosis in patients with normal expression of the p16 protein. In Garcia-Esudero et al. studies the presence of mutation or amplification of PIK3CA gene was also not associated with an increased risk of recurrence [61].
It is necessary to mention some limitations of our study. First of them is that we did not confirm the presence of mutations in samples by sequencing. The reason for this was the fact that in recent studies, ddPCR was found to be more precise than sequencing for detecting rare alleles as molecular markers. Sequencing strategies applied to FFPE samples for mutation detection pose serious problems derived from polymerase error rates. This enzyme introduces mistakes during amplification, generating point mutations. An estimated error rate of 1% gives rise to hundreds of millions of sequencing mistakes in a single experiment [62]. As a result, at present in comparison of mutation detection in the clinical setting there is no gold standard available for assessment of tumor heterogeneity. Furthermore, we assessed only one type of mutations of PIK3CA and we did not check expression on mRNA levels which would be extremely valuable, and we intend to do so because Garcia-Escudero et al. revealed that PIK3CA overexpression, but not mutations, is a poor prognostic marker in HNSCC. There is a need for a larger study to validate our results [61].

Conclusions
We confirmed the utility of ddPCR for detection of PIK3CA mutation in FFPE samples. Differences in somatic events between individuals as well as within the same single biopsy site, and the method of detection contribute to treatment failure and drug resistance because of the vast genetic variety between samples. The ddPCR technique appears to be a more sensitive than RT-qPCR and should be routinely implemented to determine the presence of genetic markers because it prevents false negative results.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Conflicts of Interest:
The authors declare no conflict of interest.