Identification of plasma hsa_circ_0000190 and 0001649 as biomarkers for predicting the recurrence and treatment response of patients with oral squamous cell carcinoma

Abstract

Background:

Oral squamous cell carcinoma (OSCC) is a type of malignancy characterized by high relapse and recurrence rates in the late stage despite optimal surgical intervention and postoperative chemoradiotherapy. Because the management of relapse following definitive treatment is challenging, accurate risk stratification is of clinical significance to improve treatment outcomes. Circular RNAs (circRNAs) are noncoding RNAs featured with cell-type specificity and high stability, owing to their circular structure, making these molecules excellent biomarkers for a variety of diseases.

Methods:

The levels of hsa_circ_0000190 and 0001649 in plasma samples from 30 healthy controls and 66 OSCC patients were determined by droplet digital polymerase chain reaction. The same primer sets were used with PCR to examine the expression of these two circRNAs in cancerous and adjacent normal tissues. A receiver operating characteristics curve was generated to evaluate the diagnostic value. The Kaplan–Meier method with a log-rank test was used for survival analysis.

Results:

We identified two circRNAs as potential biomarkers for OSCC, showing that the plasma level of hsa_circ_0000190 was significantly decreased in the late stage and marginally correlated with the development of second primary OSCC. We also found that the decreased plasma hsa_circ_0001649 was correlated with the recurrence and poor prognosis of patients. Additionally, we found that high plasma hsa_circ_0000190, but not hsa_circ_0001649, possibly predicted a better response of patients to induction chemotherapy.

Conclusion:

Our study demonstrated the potential of biomarkers in plasma to inform not just the tumor but the entire oral cavity, thereby offering a prediction for early recurrence and second primary OSCC. The plasma circRNAs remain valuable for OSCC, albeit the easy accessibility to the oral cavity.

1. INTRODUCTION

Oral squamous cell carcinoma (OSCC) is one of the most common cancers in Central and Southeast Asia.¹ The 5-year relative survival rate for the early disease is 90% but decreases to approximately 40% for those diagnosed at a late stage.^2,3 OSCC patients are treated primarily by curative-intent surgery, with adjuvant chemo-radiotherapy in stages III and IV.⁴ Under certain circumstances, induction chemotherapy is opted for shrinking tumor volume before definitive treatment.⁵

While the treatment of primary tumors is guided by standard principles, the management of early recurrence or second primary diseases often grows in complexity. For instance, the clinical decision to reoperate is complicated by previous radiation and chemotherapy. In addition, the benefit of reirradiation and salvage systemic therapy remains uncertain, and the effectiveness of induction chemotherapy to gain locoregional control may vary between individual patients.^6–8 Consequently, locoregional recurrence and second primary tumor, occurring in up to 40% of cases, contribute to the main cause of cancer-related deaths in OSCC patients.^9,10 Because the suboptimal treatment compromises the quality of life and may allow disease progression, the biomarkers to predict OSCC patients at a high risk of relapse may help formulate the therapeutic strategy and prompt a shift of treatment paradigms to enhance overall survival.

Circular RNAs (circRNAs) are a type of noncoding RNAs formed by back-splicing precursor mRNAs into a covalently closed, single-stranded circRNA molecule without free 5′ and 3′ ends or poly-A tails. CircRNAs may act as the miRNA sponges to titrate miRNA activity or interact with RNA-binding proteins and RNA polymerase II to regulate gene expression.^11–14 Although circRNAs were thought to be by-products of aberrant mRNA splicing, recent studies showed that circRNAs are differentially expressed in a tissue- or disease-specific manner. CircRNAs can be valuable biomarkers. In fact, due to their lack of 5′-3′ polarity and polyadenylate tail to prevent RNase R degradation or RNA exonuclease digestion, circRNAs are stable and can be readily detected in various tissues and body fluids, proving their potential as biomarkers of numerous diseases, including cancers. For OSCC, several potential circRNA biomarkers have been identified, including hsa_circ_009755, 0005379, and 001242, that show high diagnostic accuracy for OSCC detection.^15–17 However, to our knowledge, the circRNAs as predictive or therapeutic biomarkers for OSCC have not yet been identified.

In this study, we examined a list of circRNAs that have been reported as cancer biomarkers in a variety of cancers. We aimed to identify the circRNAs in plasma of OSCC patients that could improve risk stratification for prediction of locoregional recurrence and second primary tumor in OSCC patients.

2. METHODS

2.1. Patients and samples

This study was approved by the IRB (NO. 2017-12-015BCF), and written informed consent was obtained from all participants. This study included a total of 30 healthy controls and 66 OSCC patients who were diagnosed as squamous cell carcinoma of buccal mucosa, tongue, jawbone, lip, mouth floor, and palate. The clinicopathological features are shown in Table 1. Among these patients, six (9%) were stage I, eight (12%) were stage II, seven (11%) were stage III, 30 (45%) were stage IVa, and 15 (23%) were stage IVb. The mean (SD) patient age was 58 (9.9) years (range 32-78). All patients underwent primary surgical resection with tumor-free margins. The patients who were diagnosed with distant metastasis, and therefore, not indicated for surgical treatment were excluded.

Table 1.

Association between the expression of hsa_circ_0000190 and hsa_circ_0001649 with clinicopathological characteristics

Characteristics	Patients	has_circ_0000190		has_circ_0001649
		Mean ± SD	p	Mean ± SD	p
Gender  Male  Female	61 5	307 ± 249 362 ± 254	0.63	1569 ± 1686 1298 ± 614	0.72
Age
<60	44	318 ± 274	0.74	1669 ± 1807	0.40
>60	22	296 ± 191		1308 ± 1198
Tumor site
Buccal mucosa	27	367 ± 327	0.17	1456 ± 1620	0.69
Gingiva	15	242 ± 172		1863 ± 1273
Tongue  Other sites	16 8	235 ± 144 402 ± 164		1674 ± 1222 1043 ± 641
Tumor size
T1	9	331 ± 161	0.94	1425 ± 917	0.77
T2	10	348 ± 197		2013 ± 1564
T3	8	304 ± 212		1229 ± 1180
T4	39	298 ± 286		1524 ± 1570
Lymph node involvement
N0	45	329 ± 227	0.70	1666 ± 621	0.55
N1	5	256 ± 218		621 ± 340
N2	11	311 ± 163		1366 ± 1405
N3	5	201 ± 135		1823 ± 967
Differentiation
Well	7	301 ± 162	0.63	2722 ± 2772	0.13
Moderate	49	298 ± 274		1429 ± 1434
Poor	10	381 ± 142		1313 ± 1363
Pathological features
Lymphovascular invasion (–)	45	321 ± 275	0.61	1671 ± 1781	0.38
Lymphovascular invasion (+)	21	288 ± 181		1286 ± 1241
Perineural invasion (–)	45	325 ± 279	0.49	1652 ± 1804	0.45
Perineural invasion (+)	21	279 ± 167		1326 ± 1179

Blood samples were collected before surgery or, in some cases, before and after induction chemotherapy. All patients underwent surgical treatment regardless of whether induction chemotherapy was given. Blood samples collected in Vacutainer EDTA tubes (Becton Dickinson, Franklin Lakes, NJ, USA) were centrifuged for 10 minutes at 1000g within 30 minutes upon harvesting plasma under room temperature. Plasma was stored at −80°C in 1 mL aliquots until further processing. The tumor tissues and adjacent normal tissues were immediately frozen with liquid nitrogen and then stored at −80°C until use. All procedures of specimen acquisition have followed the tenets of the Declaration of Helsinki and are reviewed by Institutional Review Committee at Taipei Veterans General Hospital.

2.2. Cell culture

The parental and cisplatin-resistant OECM1 cancer cell lines were grown in Roswell Park Memorial Institute Medium (RPMI; #11-100; Biological Industries, Beit Haemek, Israel) supplemented with 10% Fetal Bovine Serum (FBS; #10438-028; Thermo Fisher Scientific, Waltham, MA, USA), 1% penicillin-streptomycin (#15140-122; Thermo Fisher Scientific), and 1% glutamine. The characteristics of these cells were described in the previous study¹⁸ Subculturing was performed using trypsin-EDTA. The medium was refreshed every 2 days. All cell lines were tested negative for mycoplasma contamination.

2.3. RNA isolation and ddRT-PCR

Total RNA was isolated from the tumor, adjacent normal tissues, and OECM1 cancer cells using the RNeasy Mini Kit (QIAGEN, Hilden, Germany). Oligonucleotides were designed using the computer software package Primer Express 2.0 (Applied Biosystems, Foster City, CA, USA). All oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA, USA). For examination of the circRNA expression, the divergent primers were designed to specifically amplify the circRNAs across the back spliced junctions but not their linear mRNA counterparts. The primer sequences for PCR are as follows: hsa_circ_0000190, 5′GGCAGCTGAAGTCACACATGAA3′ and 5′ CCAGTGCAATGACATGAGCAGT3′; hsa_circ_0001649, 5′ AATGCTGAAAACTGCTGAGAGAA3′ and 5′ TTGAGAAAACGAGTGCTTTGG3′; and GAPDH, 5′ GGCAGCTGAAGTCACACATGAA3′ and 5′ CCAGTGCAATGACATGAGCAGT3′. The same primer sets were used in the PCR and droplet digital polymerase chain reaction (ddPCR) experiments.

Blood samples collected in Vacutainer EDTA tubes (Becton Dickinson) were processed as described.¹⁹ Briefly, total RNA was isolated from 1 mL of plasma by using the QIAamp Circulating Nucleic Acid Kit (QIAGEN), eluted with spin columns in 15 μL of nuclease-free water, followed by synthesis of complementary DNA (cDNA) in a 20-µL reaction using the cDNA Synthesis Kit (Thermo Fisher Scientific). The droplets for the ddPCR reactions were generated according to the manufacturer’s protocol (Bio-Rad, Hercules, CA, USA), and the samples were transferred into a 96-well PCR plate. The ddPCR was conducted on the QX200 Droplet System with EvaGreen Supermix (Bio-Rad) using 1 μL of synthesized cDNA in a 20 μL PCR reaction buffer. The same procedures were performed for all test specimens and negative controls. At the end of the PCR reaction, PCR-positive and PCR-negative droplets were counted by the QX200 Droplet Reader (Bio-Rad), and the data were analyzed by QuantaSoft software (Bio-Rad).

2.4. Statistical analysis

Quantifiable data are expressed as mean ± SD of the mean (SD). Differences between the groups were analyzed using one-way analysis of variance followed or Student’s t-test. The Kaplan–Meier method with a log-rank test was used for survival analysis. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) 18.0 Software (SPSS, Chicago, IL, USA) and PRISM (GraphPad Software Inc., San Diego, CA, USA). All the p values were two-sided, and a value of <0.05 was considered statistically significant.

3. RESULTS

3.1. The plasma levels of hsa_circ_0000190 in OSCC patients were downregulated and correlated to the progression to late stage

To identify the circRNA biomarkers in plasma for OSCC patients, we began by examining a list of circRNAs that have been reported to be differentially expressed in cancer patients and healthy individuals. Among a total of 15 circRNAs tested, we found that the hsa_circ_0000190 and 0001649 were detected in the tumor and adjacent normal oral tissues by PCR, although either or both these two circRNAs were frequently decreased in the tumor in several patients (Fig. 1A). The amplicons of the expected size (145 bp for hsa_circ_0000190 and 115 bp for hsa_circ_0001649) were featured with resistance to the RNase R (Fig. 1B) and the conjunction sites (Fig. 1C), indicating their conformation of a covalently closed loop. However, since we were unable to detect these candidate circRNAs in the plasma by PCR, RT-ddPCR was further used to measure the plasma levels of hsa_circ_0000190 and 0001649 in a cohort of 30 healthy controls and 66 OSCC patients (Supplementary Fig. 1). We found that the plasma levels of hsa_circ_0000190 and 0001649 in OSCC patients were downregulated compared to that in healthy individuals, with a highly significant decrease (p < 0.0001) of hsa_circ_0000190 in the late stage (Fig. 2A). Receiver operating characteristic curve (ROC) analysis revealed that hsa_circ_0000190 effectively differentiated OSCC patients from healthy individuals with the area under ROC (AUC) of 0.84 (95% confidence interval [CI], 0.75-0.93); however, the hsa_circ_0001649 was unable to differentiate OSCC patients from healthy individuals (AUC, 0.55; 95% CI, 0.40-0.70) (Fig. 2B). The best cut-off value for hsa_circ_0000190 based on Youden’s index (sensitivity + specificity − 1) indicated the sensitivity and specificity were 83.71% and 71.73%. Meanwhile, we found no association between their levels with the clinicopathological features of OSCC patients, such as gender, age, tumor sites, tumor size, lymph node involvement, differentiation, lymphovascular invasion, and perineural invasion (Table 1). Therefore, the plasma hsa_circ_0000190 may reflect the degree of malignancy rather than a particular clinicopathological feature. The decrease in plasma hsa_circ_0000190 could serve as a superior OSCC diagnostic biomarker for OSCC patients, particularly in the late stage.

Fig. 1.

The expression and characterization of hsa_circ_0000190 and 0001649 in OSCC patients. A, PCR analysis of the expression of the hsa_circ_0000190 and 0001649 in paired tumor and adjacent normal tissues. B, qRT-PCR analysis of hsa_circ_0000190 and 0001649 in the RNase R-treated RNA samples, with GAPDH serving as the control. C, Sanger sequencing of the hsa_circ_0000190 and 0001649 PCR amplicon, showing that the hsa_circ_0000190 and 0001649 comprises the exon 4 and 3 of CNIH4 and the exon 29 and 26 of SHPRH, respectively. *p < 0.001 (Student’s t-test).

Fig. 2.

The hsa_circ_0000190 is downregulated in the plasma of OSCC patients. A, The levels of hsa_circ_0000190 and 0001649 were assessed in the plasma of 30 healthy individuals, 21 stages I to III OSCC patients, and 45 stage IV OSCC patients by RT-ddPCR. B, ROC curve analysis of the diagnostic power of the hsa_circ_0000190 and 0001649. ***p < 0.001; ****p < 0.0001. ddPCR = droplet digital polymerase chain reaction; ROC = receiver operating characteristic.

3.2. Decreased plasma level of hsa_circ_0001649 correlates with recurrence and poor prognosis of OSCC patients

To examine whether hsa_circ_0000190 and 0001649 could be prognostic biomarkers for OSCC patients, we correlated the levels of these circRNAs with the risk of recurrence and/or development of second primary OSCC after treatment. We found that the plasma hsa_circ_0000190 in patients with second primary OSCC was decreased compared to non-recurrent/second primary OSCC (Fig. 3A, left). On the other hand, the level of plasma hsa_circ_0001649 was significantly lower in patients with recurrent OSCC than those without recurrence (Fig. 3A, right). Overall, OSCC patients who were at risk of disease progression tended to have low levels of plasma hsa_circ_0000190 and 0001649 (Fig. 3B). Meanwhile, we set 198 copies/mL and 1029 copies/mL (1 SD below the mean of healthy individuals) as the cut-off value for the hsa_circ_0000190 and 0001649, respectively, to select patients for further analysis. We found that the decreased plasma hsa_circ_0001649, but not hsa_circ_0000190, correlated with a short recurrence interval (Fig. 3C). We also found that patients with low plasma hsa_circ_0001649, but not hsa_circ_0000190, were more likely to have a poor disease-free survival (Fig. 4, left and middle). Notably, the tendency for patients with low levels of both hsa_circ_0000190 and 0001649 to have a worse overall survival did not reach statistical significance, likely due to the small number of cases (Fig. 4, right). Collectively, the decrease in plasma hsa_circ_0001649 to a low level is correlated with an increased risk of early recurrence and poor overall survival.

Fig. 3.

The downregulation of hsa_circ_0000190 and 0001649 correlates with an increased rate of second primary tumor and early recurrence, respectively. A, The levels of hsa_circ_0000190 and 0001649 were assessed by ddPCR in the plasma of 42 patients without recurrence or second primary tumor, nine patients with recurrence, 11 patients with second primary tumor, and four patients with both recurrence and second primary tumor. B, Scatter plot of plasma hsa_circ_0000190 vs 0001649 in 30 healthy individuals, 38 patients without recurrence or second primary tumor, and 28 patients with recurrence and/or second primary tumor. C, Correlation between the levels of hsa_circ_0000190 (left) or hsa_circ_0001649 (right) with recurrence days. *p < 0.05. ddPCR = droplet digital polymerase chain reaction.

Fig. 4.

Kaplan-Meier survival analysis for overall survival of OSCC patients with high (above cut-off value) and low (below cut-off value) levels of plasma hsa_circ_0000190 and/or 0001649. The cut-off value is defined as the circRNA level one standard deviation lower than the mean of healthy individuals.

3.3. Ascending trend of plasma hsa_circ_0000190 possibly predicts a better response of OSCC to induction chemotherapy

We further explore the potential of hsa_circ_0000190 and 00016149 to contribute clinical-relevant information. Specifically, the levels of these circRNAs in plasma samples from OSCC patients collected before and after induction chemotherapy were correlated with the change in gross tumor volumes measured by computed tomography or magnetic resonance imaging. Of the 7 OSCC patients, the pretreatment plasma hsa_circ_0000190 in one patient (#6) was above that of all other patients, and the level of which continued to rise after treatment. On the contrary, the pre-treatment plasma hsa_circ_0000190 in another patient (#7) was low and further decreased to barely detectable levels after treatment (Fig. 5A, right). Remarkably, following the completion of induction chemotherapy, the size tumor of patient #1 regressed by 45%, whereas the tumor size of patient #7 progressed by 25% (Fig. 5A, B). Such correlation, however, was not observed for the plasma hsa_circ_0001649. Additionally, we found that the expression of hsa_circ_0000190, but not hsa_circ_0001649, was decreased in the OECM1-derived cisplatin-resistant cells compared to their parental OECM1 cells (Fig. 5C). Taken together, high plasma hsa_circ_0000190 with a tendency to ascend may favor a better response of OSCC to induction chemotherapy.

Fig. 5.

A high pre-treatment hsa_circ_0000190 predicts superior response to induction chemotherapy. A, The levels of hsa_circ_0000190 (left) and 0001649 (middle) were assessed by ddPCR in seven paired plasma samples from OSCC patients before and after induction chemotherapy. The percentage change of tumor gross volume is calculated by dividing the estimated tumor size before induction chemotherapy with that after induction chemotherapy. The maximal increase or decrease in size in this cohort is indicated in red and green, respectively. B, The radiographic imaging (CT or MRI) of OSCC patients to estimate tumor size before and after induction chemotherapy. C, PCR analysis of the expression of the hsa_circ_0000190 and 0001649 in parental and cisplatin-resistant OECM1 cancer cell lines. CT = computed tomography; ddPCR = droplet digital polymerase chain reaction.

4. DISCUSSION

In this study, we demonstrated that hsa_circ_0000190 and 0001649 were significantly decreased in the plasma samples of OSCC patients in the late stage, with a higher diagnostic value of hsa_circ_0000190 than 00016149 for this cancer type. We also found that the patients with low plasma hsa_circ_0000190 were more likely to develop second primary OSCC, and those with low plasma hsa_circ_0001649 were at risk of early tumor recurrence and poor prognosis. Additionally, OSCC patients with high hsa_circ_0000190 appeared to have a favorable response to induction chemotherapy, whereas those with a declining level of hsa_circ_0000190 were less responsive to such treatment. Therefore, hsa_circ_0000190 and 0001649 are of clinical relevance as the blood-based biomarkers in the monitoring and treatment of OSCC.

The ability of circRNAs as biomarkers for OSCC diagnosis and prognosis has been previously reported. Like other cancer types, most of these studies examined the differential expression of circRNAs between tumor and adjacent normal tissues, predominantly reporting downregulation of the identified circRNA biomarkers in OSCC. The propensity of cancer-related circRNAs to be decreased in OSCC is, however, distinct from many other cancers. Indeed, the upregulation of multiple circRNAs has been commonly reported in lung cancer (LC), colorectal cancer (CRC), breast cancer, and hepatocellular cancer (HCC), except gastric cancer (GC). Not only is the expression profile of circRNAs in different cancers distinct, but the role of a given circRNA across cancer types can also be different. In terms of hsa_circ_0000190, although its downregulation was found in GC,²⁰ osteosarcoma,²¹ and multiple myeloma²² as our observation with OSCC, the opposite finding was reported in LC.²³ Likewise, hsa_circ_0001649 was found to be downregulated in HCC,²⁴ CRC,^25,26 non-small-cell LC,²⁷ and cholangiocarcinoma²⁸ but upregulated in LC.²³ Accordingly, the circRNA biomarker is tissue-specific and closely related to tumor characteristics.

Since the characteristics vary between different cancer types, the benefit of biomarkers deserves unique considerations based on each cancer type. For OSCC, the standard of care is primary surgical tumor resection, and some patients may benefit from induction chemotherapy.²⁹ Meanwhile, up to 45% of patients will develop locoregional recurrence and/or second primary tumors, contributing to the leading cause of death.^30,31 As such, the biomarkers that predict recurrence, second primary tumor, or treatment response are of value for OSCC. To date, the potential biomarkers are pretreatment serum interleukin-6 (IL-6),^32,33 CD44,³⁴ and miR-29a for recurrence;³⁵ IL-24 for second primary OSCC;³⁶ and GDF15 for cisplatin-based induction chemotherapy.³⁷ To our knowledge, the current study was the first to report circRNA biomarkers for these purposes. Notably, hsa_circ_0000190 and 0001649 relate to a different pattern of disease relapse, suggesting their roles in different pathways. The hsa_circ_0000190 has been implicated in the repression of miR-767-5p,²² an onco-miRNA through MAPK4 pathway, as well as miR-142-5p,³⁸ a tumor-suppressive miRNA by targeting cyclin-dependent kinases 4/6. The hsa_circ_0001649 has been linked to the suppression of multiple miRNAs, including miR-338-5p, miR-647, and miR-942, that involves in cell proliferation and STAT pathways.³⁹ Apparently, the circRNA-miRNA-mRNA network is complex, making elucidating the causal links between hsa_circ_0000190/0001649 and recurrence/second primary OSCC very difficult. Nonetheless, from the clinical perspective, identifying the biomarkers for patients with increased risk of relapse remains valuable.

Unlike other visceral cancers, OSCC located in the oral cavity is readily accessible for visual inspection, casting doubt on the usefulness of diagnostic biomarkers in body fluid. Indeed, although tissue biopsy is invasive and usually associated with considerable discomfort, the histological examination remains the gold standard for definitive OSCC diagnosis. However, concerning OSCC as a solid tumor exhibiting heterogeneity, a single biopsy may not fully reflect the status of the whole tumor or reliably predict the development of second primary tumor at other sites. In contrast to tissue biopsy, a liquid biopsy may inform about not just the entire tumor but also the relatively normal tissue adjacent or distant to the tumor. Supporting this notion, a meta-analysis on 16 studies demonstrated that biomarkers in serum, plasma, or saliva exhibited a higher diagnostic power for OSCC than that in tissue specimens. Thus, in accordance with the concept of field cancerization, these circulating biomarkers are believed to hold great promise to identify patients at high risk of early relapse.

Although we demonstrated that the plasma hsa_circ_0000190 and 0001649 combinations are of value to predict the risk of the recurrence and development of second primary OSCC, this study did not investigate the expression of their parental genes (CNIH4 and SHPRN, respectively) in the OSCC tissues of our patient cohort. The CNIH4 gene is involved in G protein-coupled receptors trafficking from the endoplasmic reticulum, and the SHPRH gene encodes motifs associated with the DNA repair process. Thus far, the definitive roles of CNIH4 and SHPRN in oral carcinogenesis have not been established. Interestingly, the Cancer Genome Atlas (TCGA) database showed a significant upregulation of CNIH4 in OSCC patients, whereas the level of SHPRH was not significantly different between healthy individuals and OSCC patients. Because the hsa_circ_0000190 and CNIH4 are driven by the same promoter, it is worth further investigation how cancer cells differentially modulate the expression of parental genes and their derived circRNAs to enhance the malignancy. Other limitations include the lack of oral potentially malignant disorders (OPMDs) patients and the small sample size (especially for patients receiving induction chemotherapy) in this study. However, as we showed that the hsa_circ_0000190 was significantly decreased in the late stage, implying its correlation with the degree of malignancy, it is plausible that the expression of these circRNAs would not be significantly altered in OPMDs. In addition, since induction chemotherapy is not routinely performed for OSCC patients, we would not expect to recruit many patients with this treatment.

In conclusion, we demonstrated that hsa_circ_0000190 and 0001649 were downregulated in OSCC patients as evidenced by the comparison between paired OSCC and adjacent normal tissue, as well as between the plasma samples of OSCC patients and healthy controls. These circRNA biomarkers hold high predictive value for recurrent and second primary OSCC. Importantly, our study supports the notion that circRNAs in plasma can be reliable biomarkers and are of great potential to stratify the risk of OSCC patients and improve the treatment strategy.

ACKNOWLEDGMENTS

This research was supported by the Ministry of Science and Technology, Taiwan, grant number MOST-110-2314-B-075-022-MY3, MOST 110-2320-B-075 -006 -MY3, MOST-109-2314-B-075-006, MOST-108-2314-B-303-024, and Taipei Veterans General Hospital, V111E-001-4 V111C-157.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data related to this article can be found at http://links.lww.com/JCMA/A132.

Supplementary Material

ca9-85-431-s001.pdf ^{(564.1KB, pdf)}